Merge branch 'linus' into x86/xen
[linux-2.6] / 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 <net/net_namespace.h>
44 #include <net/snmp.h>
45 #include <net/ipv6.h>
46 #include <net/ip6_fib.h>
47 #include <net/ip6_route.h>
48 #include <net/ndisc.h>
49 #include <net/addrconf.h>
50 #include <net/tcp.h>
51 #include <linux/rtnetlink.h>
52 #include <net/dst.h>
53 #include <net/xfrm.h>
54 #include <net/netevent.h>
55 #include <net/netlink.h>
56
57 #include <asm/uaccess.h>
58
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62
63 /* Set to 3 to get tracing. */
64 #define RT6_DEBUG 2
65
66 #if RT6_DEBUG >= 3
67 #define RDBG(x) printk x
68 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
69 #else
70 #define RDBG(x)
71 #define RT6_TRACE(x...) do { ; } while (0)
72 #endif
73
74 #define CLONE_OFFLINK_ROUTE 0
75
76 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
78 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
79 static void             ip6_dst_destroy(struct dst_entry *);
80 static void             ip6_dst_ifdown(struct dst_entry *,
81                                        struct net_device *dev, int how);
82 static int               ip6_dst_gc(struct dst_ops *ops);
83
84 static int              ip6_pkt_discard(struct sk_buff *skb);
85 static int              ip6_pkt_discard_out(struct sk_buff *skb);
86 static void             ip6_link_failure(struct sk_buff *skb);
87 static void             ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
88
89 #ifdef CONFIG_IPV6_ROUTE_INFO
90 static struct rt6_info *rt6_add_route_info(struct net *net,
91                                            struct in6_addr *prefix, int prefixlen,
92                                            struct in6_addr *gwaddr, int ifindex,
93                                            unsigned pref);
94 static struct rt6_info *rt6_get_route_info(struct net *net,
95                                            struct in6_addr *prefix, int prefixlen,
96                                            struct in6_addr *gwaddr, int ifindex);
97 #endif
98
99 static struct dst_ops ip6_dst_ops_template = {
100         .family                 =       AF_INET6,
101         .protocol               =       __constant_htons(ETH_P_IPV6),
102         .gc                     =       ip6_dst_gc,
103         .gc_thresh              =       1024,
104         .check                  =       ip6_dst_check,
105         .destroy                =       ip6_dst_destroy,
106         .ifdown                 =       ip6_dst_ifdown,
107         .negative_advice        =       ip6_negative_advice,
108         .link_failure           =       ip6_link_failure,
109         .update_pmtu            =       ip6_rt_update_pmtu,
110         .local_out              =       __ip6_local_out,
111         .entry_size             =       sizeof(struct rt6_info),
112         .entries                =       ATOMIC_INIT(0),
113 };
114
115 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
116 {
117 }
118
119 static struct dst_ops ip6_dst_blackhole_ops = {
120         .family                 =       AF_INET6,
121         .protocol               =       __constant_htons(ETH_P_IPV6),
122         .destroy                =       ip6_dst_destroy,
123         .check                  =       ip6_dst_check,
124         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
125         .entry_size             =       sizeof(struct rt6_info),
126         .entries                =       ATOMIC_INIT(0),
127 };
128
129 static struct rt6_info ip6_null_entry_template = {
130         .u = {
131                 .dst = {
132                         .__refcnt       = ATOMIC_INIT(1),
133                         .__use          = 1,
134                         .obsolete       = -1,
135                         .error          = -ENETUNREACH,
136                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
137                         .input          = ip6_pkt_discard,
138                         .output         = ip6_pkt_discard_out,
139                 }
140         },
141         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
142         .rt6i_metric    = ~(u32) 0,
143         .rt6i_ref       = ATOMIC_INIT(1),
144 };
145
146 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
147
148 static int ip6_pkt_prohibit(struct sk_buff *skb);
149 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
150
151 static struct rt6_info ip6_prohibit_entry_template = {
152         .u = {
153                 .dst = {
154                         .__refcnt       = ATOMIC_INIT(1),
155                         .__use          = 1,
156                         .obsolete       = -1,
157                         .error          = -EACCES,
158                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
159                         .input          = ip6_pkt_prohibit,
160                         .output         = ip6_pkt_prohibit_out,
161                 }
162         },
163         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
164         .rt6i_metric    = ~(u32) 0,
165         .rt6i_ref       = ATOMIC_INIT(1),
166 };
167
168 static struct rt6_info ip6_blk_hole_entry_template = {
169         .u = {
170                 .dst = {
171                         .__refcnt       = ATOMIC_INIT(1),
172                         .__use          = 1,
173                         .obsolete       = -1,
174                         .error          = -EINVAL,
175                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
176                         .input          = dst_discard,
177                         .output         = dst_discard,
178                 }
179         },
180         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
181         .rt6i_metric    = ~(u32) 0,
182         .rt6i_ref       = ATOMIC_INIT(1),
183 };
184
185 #endif
186
187 /* allocate dst with ip6_dst_ops */
188 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
189 {
190         return (struct rt6_info *)dst_alloc(ops);
191 }
192
193 static void ip6_dst_destroy(struct dst_entry *dst)
194 {
195         struct rt6_info *rt = (struct rt6_info *)dst;
196         struct inet6_dev *idev = rt->rt6i_idev;
197
198         if (idev != NULL) {
199                 rt->rt6i_idev = NULL;
200                 in6_dev_put(idev);
201         }
202 }
203
204 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
205                            int how)
206 {
207         struct rt6_info *rt = (struct rt6_info *)dst;
208         struct inet6_dev *idev = rt->rt6i_idev;
209         struct net_device *loopback_dev =
210                 dev_net(dev)->loopback_dev;
211
212         if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
213                 struct inet6_dev *loopback_idev =
214                         in6_dev_get(loopback_dev);
215                 if (loopback_idev != NULL) {
216                         rt->rt6i_idev = loopback_idev;
217                         in6_dev_put(idev);
218                 }
219         }
220 }
221
222 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
223 {
224         return (rt->rt6i_flags & RTF_EXPIRES &&
225                 time_after(jiffies, rt->rt6i_expires));
226 }
227
228 static inline int rt6_need_strict(struct in6_addr *daddr)
229 {
230         return (ipv6_addr_type(daddr) &
231                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK));
232 }
233
234 /*
235  *      Route lookup. Any table->tb6_lock is implied.
236  */
237
238 static inline struct rt6_info *rt6_device_match(struct net *net,
239                                                     struct rt6_info *rt,
240                                                     struct in6_addr *saddr,
241                                                     int oif,
242                                                     int flags)
243 {
244         struct rt6_info *local = NULL;
245         struct rt6_info *sprt;
246
247         if (!oif && ipv6_addr_any(saddr))
248                 goto out;
249
250         for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
251                 struct net_device *dev = sprt->rt6i_dev;
252
253                 if (oif) {
254                         if (dev->ifindex == oif)
255                                 return sprt;
256                         if (dev->flags & IFF_LOOPBACK) {
257                                 if (sprt->rt6i_idev == NULL ||
258                                     sprt->rt6i_idev->dev->ifindex != oif) {
259                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
260                                                 continue;
261                                         if (local && (!oif ||
262                                                       local->rt6i_idev->dev->ifindex == oif))
263                                                 continue;
264                                 }
265                                 local = sprt;
266                         }
267                 } else {
268                         if (ipv6_chk_addr(net, saddr, dev,
269                                           flags & RT6_LOOKUP_F_IFACE))
270                                 return sprt;
271                 }
272         }
273
274         if (oif) {
275                 if (local)
276                         return local;
277
278                 if (flags & RT6_LOOKUP_F_IFACE)
279                         return net->ipv6.ip6_null_entry;
280         }
281 out:
282         return rt;
283 }
284
285 #ifdef CONFIG_IPV6_ROUTER_PREF
286 static void rt6_probe(struct rt6_info *rt)
287 {
288         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
289         /*
290          * Okay, this does not seem to be appropriate
291          * for now, however, we need to check if it
292          * is really so; aka Router Reachability Probing.
293          *
294          * Router Reachability Probe MUST be rate-limited
295          * to no more than one per minute.
296          */
297         if (!neigh || (neigh->nud_state & NUD_VALID))
298                 return;
299         read_lock_bh(&neigh->lock);
300         if (!(neigh->nud_state & NUD_VALID) &&
301             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
302                 struct in6_addr mcaddr;
303                 struct in6_addr *target;
304
305                 neigh->updated = jiffies;
306                 read_unlock_bh(&neigh->lock);
307
308                 target = (struct in6_addr *)&neigh->primary_key;
309                 addrconf_addr_solict_mult(target, &mcaddr);
310                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
311         } else
312                 read_unlock_bh(&neigh->lock);
313 }
314 #else
315 static inline void rt6_probe(struct rt6_info *rt)
316 {
317         return;
318 }
319 #endif
320
321 /*
322  * Default Router Selection (RFC 2461 6.3.6)
323  */
324 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
325 {
326         struct net_device *dev = rt->rt6i_dev;
327         if (!oif || dev->ifindex == oif)
328                 return 2;
329         if ((dev->flags & IFF_LOOPBACK) &&
330             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
331                 return 1;
332         return 0;
333 }
334
335 static inline int rt6_check_neigh(struct rt6_info *rt)
336 {
337         struct neighbour *neigh = rt->rt6i_nexthop;
338         int m;
339         if (rt->rt6i_flags & RTF_NONEXTHOP ||
340             !(rt->rt6i_flags & RTF_GATEWAY))
341                 m = 1;
342         else if (neigh) {
343                 read_lock_bh(&neigh->lock);
344                 if (neigh->nud_state & NUD_VALID)
345                         m = 2;
346 #ifdef CONFIG_IPV6_ROUTER_PREF
347                 else if (neigh->nud_state & NUD_FAILED)
348                         m = 0;
349 #endif
350                 else
351                         m = 1;
352                 read_unlock_bh(&neigh->lock);
353         } else
354                 m = 0;
355         return m;
356 }
357
358 static int rt6_score_route(struct rt6_info *rt, int oif,
359                            int strict)
360 {
361         int m, n;
362
363         m = rt6_check_dev(rt, oif);
364         if (!m && (strict & RT6_LOOKUP_F_IFACE))
365                 return -1;
366 #ifdef CONFIG_IPV6_ROUTER_PREF
367         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
368 #endif
369         n = rt6_check_neigh(rt);
370         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
371                 return -1;
372         return m;
373 }
374
375 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
376                                    int *mpri, struct rt6_info *match)
377 {
378         int m;
379
380         if (rt6_check_expired(rt))
381                 goto out;
382
383         m = rt6_score_route(rt, oif, strict);
384         if (m < 0)
385                 goto out;
386
387         if (m > *mpri) {
388                 if (strict & RT6_LOOKUP_F_REACHABLE)
389                         rt6_probe(match);
390                 *mpri = m;
391                 match = rt;
392         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
393                 rt6_probe(rt);
394         }
395
396 out:
397         return match;
398 }
399
400 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
401                                      struct rt6_info *rr_head,
402                                      u32 metric, int oif, int strict)
403 {
404         struct rt6_info *rt, *match;
405         int mpri = -1;
406
407         match = NULL;
408         for (rt = rr_head; rt && rt->rt6i_metric == metric;
409              rt = rt->u.dst.rt6_next)
410                 match = find_match(rt, oif, strict, &mpri, match);
411         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
412              rt = rt->u.dst.rt6_next)
413                 match = find_match(rt, oif, strict, &mpri, match);
414
415         return match;
416 }
417
418 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
419 {
420         struct rt6_info *match, *rt0;
421         struct net *net;
422
423         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
424                   __func__, fn->leaf, oif);
425
426         rt0 = fn->rr_ptr;
427         if (!rt0)
428                 fn->rr_ptr = rt0 = fn->leaf;
429
430         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
431
432         if (!match &&
433             (strict & RT6_LOOKUP_F_REACHABLE)) {
434                 struct rt6_info *next = rt0->u.dst.rt6_next;
435
436                 /* no entries matched; do round-robin */
437                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
438                         next = fn->leaf;
439
440                 if (next != rt0)
441                         fn->rr_ptr = next;
442         }
443
444         RT6_TRACE("%s() => %p\n",
445                   __func__, match);
446
447         net = dev_net(rt0->rt6i_dev);
448         return (match ? match : net->ipv6.ip6_null_entry);
449 }
450
451 #ifdef CONFIG_IPV6_ROUTE_INFO
452 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
453                   struct in6_addr *gwaddr)
454 {
455         struct net *net = dev_net(dev);
456         struct route_info *rinfo = (struct route_info *) opt;
457         struct in6_addr prefix_buf, *prefix;
458         unsigned int pref;
459         unsigned long lifetime;
460         struct rt6_info *rt;
461
462         if (len < sizeof(struct route_info)) {
463                 return -EINVAL;
464         }
465
466         /* Sanity check for prefix_len and length */
467         if (rinfo->length > 3) {
468                 return -EINVAL;
469         } else if (rinfo->prefix_len > 128) {
470                 return -EINVAL;
471         } else if (rinfo->prefix_len > 64) {
472                 if (rinfo->length < 2) {
473                         return -EINVAL;
474                 }
475         } else if (rinfo->prefix_len > 0) {
476                 if (rinfo->length < 1) {
477                         return -EINVAL;
478                 }
479         }
480
481         pref = rinfo->route_pref;
482         if (pref == ICMPV6_ROUTER_PREF_INVALID)
483                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
484
485         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
486
487         if (rinfo->length == 3)
488                 prefix = (struct in6_addr *)rinfo->prefix;
489         else {
490                 /* this function is safe */
491                 ipv6_addr_prefix(&prefix_buf,
492                                  (struct in6_addr *)rinfo->prefix,
493                                  rinfo->prefix_len);
494                 prefix = &prefix_buf;
495         }
496
497         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
498                                 dev->ifindex);
499
500         if (rt && !lifetime) {
501                 ip6_del_rt(rt);
502                 rt = NULL;
503         }
504
505         if (!rt && lifetime)
506                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
507                                         pref);
508         else if (rt)
509                 rt->rt6i_flags = RTF_ROUTEINFO |
510                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
511
512         if (rt) {
513                 if (!addrconf_finite_timeout(lifetime)) {
514                         rt->rt6i_flags &= ~RTF_EXPIRES;
515                 } else {
516                         rt->rt6i_expires = jiffies + HZ * lifetime;
517                         rt->rt6i_flags |= RTF_EXPIRES;
518                 }
519                 dst_release(&rt->u.dst);
520         }
521         return 0;
522 }
523 #endif
524
525 #define BACKTRACK(__net, saddr)                 \
526 do { \
527         if (rt == __net->ipv6.ip6_null_entry) { \
528                 struct fib6_node *pn; \
529                 while (1) { \
530                         if (fn->fn_flags & RTN_TL_ROOT) \
531                                 goto out; \
532                         pn = fn->parent; \
533                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
534                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
535                         else \
536                                 fn = pn; \
537                         if (fn->fn_flags & RTN_RTINFO) \
538                                 goto restart; \
539                 } \
540         } \
541 } while(0)
542
543 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
544                                              struct fib6_table *table,
545                                              struct flowi *fl, int flags)
546 {
547         struct fib6_node *fn;
548         struct rt6_info *rt;
549
550         read_lock_bh(&table->tb6_lock);
551         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
552 restart:
553         rt = fn->leaf;
554         rt = rt6_device_match(net, rt, &fl->fl6_src, fl->oif, flags);
555         BACKTRACK(net, &fl->fl6_src);
556 out:
557         dst_use(&rt->u.dst, jiffies);
558         read_unlock_bh(&table->tb6_lock);
559         return rt;
560
561 }
562
563 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
564                             const struct in6_addr *saddr, int oif, int strict)
565 {
566         struct flowi fl = {
567                 .oif = oif,
568                 .nl_u = {
569                         .ip6_u = {
570                                 .daddr = *daddr,
571                         },
572                 },
573         };
574         struct dst_entry *dst;
575         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
576
577         if (saddr) {
578                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
579                 flags |= RT6_LOOKUP_F_HAS_SADDR;
580         }
581
582         dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
583         if (dst->error == 0)
584                 return (struct rt6_info *) dst;
585
586         dst_release(dst);
587
588         return NULL;
589 }
590
591 EXPORT_SYMBOL(rt6_lookup);
592
593 /* ip6_ins_rt is called with FREE table->tb6_lock.
594    It takes new route entry, the addition fails by any reason the
595    route is freed. In any case, if caller does not hold it, it may
596    be destroyed.
597  */
598
599 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
600 {
601         int err;
602         struct fib6_table *table;
603
604         table = rt->rt6i_table;
605         write_lock_bh(&table->tb6_lock);
606         err = fib6_add(&table->tb6_root, rt, info);
607         write_unlock_bh(&table->tb6_lock);
608
609         return err;
610 }
611
612 int ip6_ins_rt(struct rt6_info *rt)
613 {
614         struct nl_info info = {
615                 .nl_net = dev_net(rt->rt6i_dev),
616         };
617         return __ip6_ins_rt(rt, &info);
618 }
619
620 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
621                                       struct in6_addr *saddr)
622 {
623         struct rt6_info *rt;
624
625         /*
626          *      Clone the route.
627          */
628
629         rt = ip6_rt_copy(ort);
630
631         if (rt) {
632                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
633                         if (rt->rt6i_dst.plen != 128 &&
634                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
635                                 rt->rt6i_flags |= RTF_ANYCAST;
636                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
637                 }
638
639                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
640                 rt->rt6i_dst.plen = 128;
641                 rt->rt6i_flags |= RTF_CACHE;
642                 rt->u.dst.flags |= DST_HOST;
643
644 #ifdef CONFIG_IPV6_SUBTREES
645                 if (rt->rt6i_src.plen && saddr) {
646                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
647                         rt->rt6i_src.plen = 128;
648                 }
649 #endif
650
651                 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
652
653         }
654
655         return rt;
656 }
657
658 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
659 {
660         struct rt6_info *rt = ip6_rt_copy(ort);
661         if (rt) {
662                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
663                 rt->rt6i_dst.plen = 128;
664                 rt->rt6i_flags |= RTF_CACHE;
665                 rt->u.dst.flags |= DST_HOST;
666                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
667         }
668         return rt;
669 }
670
671 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
672                                       struct flowi *fl, int flags)
673 {
674         struct fib6_node *fn;
675         struct rt6_info *rt, *nrt;
676         int strict = 0;
677         int attempts = 3;
678         int err;
679         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
680
681         strict |= flags & RT6_LOOKUP_F_IFACE;
682
683 relookup:
684         read_lock_bh(&table->tb6_lock);
685
686 restart_2:
687         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
688
689 restart:
690         rt = rt6_select(fn, oif, strict | reachable);
691
692         BACKTRACK(net, &fl->fl6_src);
693         if (rt == net->ipv6.ip6_null_entry ||
694             rt->rt6i_flags & RTF_CACHE)
695                 goto out;
696
697         dst_hold(&rt->u.dst);
698         read_unlock_bh(&table->tb6_lock);
699
700         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
701                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
702         else {
703 #if CLONE_OFFLINK_ROUTE
704                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
705 #else
706                 goto out2;
707 #endif
708         }
709
710         dst_release(&rt->u.dst);
711         rt = nrt ? : net->ipv6.ip6_null_entry;
712
713         dst_hold(&rt->u.dst);
714         if (nrt) {
715                 err = ip6_ins_rt(nrt);
716                 if (!err)
717                         goto out2;
718         }
719
720         if (--attempts <= 0)
721                 goto out2;
722
723         /*
724          * Race condition! In the gap, when table->tb6_lock was
725          * released someone could insert this route.  Relookup.
726          */
727         dst_release(&rt->u.dst);
728         goto relookup;
729
730 out:
731         if (reachable) {
732                 reachable = 0;
733                 goto restart_2;
734         }
735         dst_hold(&rt->u.dst);
736         read_unlock_bh(&table->tb6_lock);
737 out2:
738         rt->u.dst.lastuse = jiffies;
739         rt->u.dst.__use++;
740
741         return rt;
742 }
743
744 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
745                                             struct flowi *fl, int flags)
746 {
747         return ip6_pol_route(net, table, fl->iif, fl, flags);
748 }
749
750 void ip6_route_input(struct sk_buff *skb)
751 {
752         struct ipv6hdr *iph = ipv6_hdr(skb);
753         struct net *net = dev_net(skb->dev);
754         int flags = RT6_LOOKUP_F_HAS_SADDR;
755         struct flowi fl = {
756                 .iif = skb->dev->ifindex,
757                 .nl_u = {
758                         .ip6_u = {
759                                 .daddr = iph->daddr,
760                                 .saddr = iph->saddr,
761                                 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
762                         },
763                 },
764                 .mark = skb->mark,
765                 .proto = iph->nexthdr,
766         };
767
768         if (rt6_need_strict(&iph->daddr))
769                 flags |= RT6_LOOKUP_F_IFACE;
770
771         skb->dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input);
772 }
773
774 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
775                                              struct flowi *fl, int flags)
776 {
777         return ip6_pol_route(net, table, fl->oif, fl, flags);
778 }
779
780 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
781                                     struct flowi *fl)
782 {
783         int flags = 0;
784
785         if (rt6_need_strict(&fl->fl6_dst))
786                 flags |= RT6_LOOKUP_F_IFACE;
787
788         if (!ipv6_addr_any(&fl->fl6_src))
789                 flags |= RT6_LOOKUP_F_HAS_SADDR;
790         else if (sk) {
791                 unsigned int prefs = inet6_sk(sk)->srcprefs;
792                 if (prefs & IPV6_PREFER_SRC_TMP)
793                         flags |= RT6_LOOKUP_F_SRCPREF_TMP;
794                 if (prefs & IPV6_PREFER_SRC_PUBLIC)
795                         flags |= RT6_LOOKUP_F_SRCPREF_PUBLIC;
796                 if (prefs & IPV6_PREFER_SRC_COA)
797                         flags |= RT6_LOOKUP_F_SRCPREF_COA;
798         }
799
800         return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
801 }
802
803 EXPORT_SYMBOL(ip6_route_output);
804
805 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
806 {
807         struct rt6_info *ort = (struct rt6_info *) *dstp;
808         struct rt6_info *rt = (struct rt6_info *)
809                 dst_alloc(&ip6_dst_blackhole_ops);
810         struct dst_entry *new = NULL;
811
812         if (rt) {
813                 new = &rt->u.dst;
814
815                 atomic_set(&new->__refcnt, 1);
816                 new->__use = 1;
817                 new->input = dst_discard;
818                 new->output = dst_discard;
819
820                 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
821                 new->dev = ort->u.dst.dev;
822                 if (new->dev)
823                         dev_hold(new->dev);
824                 rt->rt6i_idev = ort->rt6i_idev;
825                 if (rt->rt6i_idev)
826                         in6_dev_hold(rt->rt6i_idev);
827                 rt->rt6i_expires = 0;
828
829                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
830                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
831                 rt->rt6i_metric = 0;
832
833                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
834 #ifdef CONFIG_IPV6_SUBTREES
835                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
836 #endif
837
838                 dst_free(new);
839         }
840
841         dst_release(*dstp);
842         *dstp = new;
843         return (new ? 0 : -ENOMEM);
844 }
845 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
846
847 /*
848  *      Destination cache support functions
849  */
850
851 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
852 {
853         struct rt6_info *rt;
854
855         rt = (struct rt6_info *) dst;
856
857         if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
858                 return dst;
859
860         return NULL;
861 }
862
863 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
864 {
865         struct rt6_info *rt = (struct rt6_info *) dst;
866
867         if (rt) {
868                 if (rt->rt6i_flags & RTF_CACHE)
869                         ip6_del_rt(rt);
870                 else
871                         dst_release(dst);
872         }
873         return NULL;
874 }
875
876 static void ip6_link_failure(struct sk_buff *skb)
877 {
878         struct rt6_info *rt;
879
880         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
881
882         rt = (struct rt6_info *) skb->dst;
883         if (rt) {
884                 if (rt->rt6i_flags&RTF_CACHE) {
885                         dst_set_expires(&rt->u.dst, 0);
886                         rt->rt6i_flags |= RTF_EXPIRES;
887                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
888                         rt->rt6i_node->fn_sernum = -1;
889         }
890 }
891
892 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
893 {
894         struct rt6_info *rt6 = (struct rt6_info*)dst;
895
896         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
897                 rt6->rt6i_flags |= RTF_MODIFIED;
898                 if (mtu < IPV6_MIN_MTU) {
899                         mtu = IPV6_MIN_MTU;
900                         dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
901                 }
902                 dst->metrics[RTAX_MTU-1] = mtu;
903                 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
904         }
905 }
906
907 static int ipv6_get_mtu(struct net_device *dev);
908
909 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
910 {
911         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
912
913         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
914                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
915
916         /*
917          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
918          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
919          * IPV6_MAXPLEN is also valid and means: "any MSS,
920          * rely only on pmtu discovery"
921          */
922         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
923                 mtu = IPV6_MAXPLEN;
924         return mtu;
925 }
926
927 static struct dst_entry *icmp6_dst_gc_list;
928 static DEFINE_SPINLOCK(icmp6_dst_lock);
929
930 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
931                                   struct neighbour *neigh,
932                                   const struct in6_addr *addr)
933 {
934         struct rt6_info *rt;
935         struct inet6_dev *idev = in6_dev_get(dev);
936         struct net *net = dev_net(dev);
937
938         if (unlikely(idev == NULL))
939                 return NULL;
940
941         rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
942         if (unlikely(rt == NULL)) {
943                 in6_dev_put(idev);
944                 goto out;
945         }
946
947         dev_hold(dev);
948         if (neigh)
949                 neigh_hold(neigh);
950         else
951                 neigh = ndisc_get_neigh(dev, addr);
952
953         rt->rt6i_dev      = dev;
954         rt->rt6i_idev     = idev;
955         rt->rt6i_nexthop  = neigh;
956         atomic_set(&rt->u.dst.__refcnt, 1);
957         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
958         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
959         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
960         rt->u.dst.output  = ip6_output;
961
962 #if 0   /* there's no chance to use these for ndisc */
963         rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
964                                 ? DST_HOST
965                                 : 0;
966         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
967         rt->rt6i_dst.plen = 128;
968 #endif
969
970         spin_lock_bh(&icmp6_dst_lock);
971         rt->u.dst.next = icmp6_dst_gc_list;
972         icmp6_dst_gc_list = &rt->u.dst;
973         spin_unlock_bh(&icmp6_dst_lock);
974
975         fib6_force_start_gc(net);
976
977 out:
978         return &rt->u.dst;
979 }
980
981 int icmp6_dst_gc(void)
982 {
983         struct dst_entry *dst, *next, **pprev;
984         int more = 0;
985
986         next = NULL;
987
988         spin_lock_bh(&icmp6_dst_lock);
989         pprev = &icmp6_dst_gc_list;
990
991         while ((dst = *pprev) != NULL) {
992                 if (!atomic_read(&dst->__refcnt)) {
993                         *pprev = dst->next;
994                         dst_free(dst);
995                 } else {
996                         pprev = &dst->next;
997                         ++more;
998                 }
999         }
1000
1001         spin_unlock_bh(&icmp6_dst_lock);
1002
1003         return more;
1004 }
1005
1006 static int ip6_dst_gc(struct dst_ops *ops)
1007 {
1008         unsigned long now = jiffies;
1009         struct net *net = ops->dst_net;
1010         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1011         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1012         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1013         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1014         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1015
1016         if (time_after(rt_last_gc + rt_min_interval, now) &&
1017             atomic_read(&ops->entries) <= rt_max_size)
1018                 goto out;
1019
1020         net->ipv6.ip6_rt_gc_expire++;
1021         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1022         net->ipv6.ip6_rt_last_gc = now;
1023         if (atomic_read(&ops->entries) < ops->gc_thresh)
1024                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1025 out:
1026         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1027         return (atomic_read(&ops->entries) > rt_max_size);
1028 }
1029
1030 /* Clean host part of a prefix. Not necessary in radix tree,
1031    but results in cleaner routing tables.
1032
1033    Remove it only when all the things will work!
1034  */
1035
1036 static int ipv6_get_mtu(struct net_device *dev)
1037 {
1038         int mtu = IPV6_MIN_MTU;
1039         struct inet6_dev *idev;
1040
1041         idev = in6_dev_get(dev);
1042         if (idev) {
1043                 mtu = idev->cnf.mtu6;
1044                 in6_dev_put(idev);
1045         }
1046         return mtu;
1047 }
1048
1049 int ip6_dst_hoplimit(struct dst_entry *dst)
1050 {
1051         int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1052         if (hoplimit < 0) {
1053                 struct net_device *dev = dst->dev;
1054                 struct inet6_dev *idev = in6_dev_get(dev);
1055                 if (idev) {
1056                         hoplimit = idev->cnf.hop_limit;
1057                         in6_dev_put(idev);
1058                 } else
1059                         hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1060         }
1061         return hoplimit;
1062 }
1063
1064 /*
1065  *
1066  */
1067
1068 int ip6_route_add(struct fib6_config *cfg)
1069 {
1070         int err;
1071         struct net *net = cfg->fc_nlinfo.nl_net;
1072         struct rt6_info *rt = NULL;
1073         struct net_device *dev = NULL;
1074         struct inet6_dev *idev = NULL;
1075         struct fib6_table *table;
1076         int addr_type;
1077
1078         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1079                 return -EINVAL;
1080 #ifndef CONFIG_IPV6_SUBTREES
1081         if (cfg->fc_src_len)
1082                 return -EINVAL;
1083 #endif
1084         if (cfg->fc_ifindex) {
1085                 err = -ENODEV;
1086                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1087                 if (!dev)
1088                         goto out;
1089                 idev = in6_dev_get(dev);
1090                 if (!idev)
1091                         goto out;
1092         }
1093
1094         if (cfg->fc_metric == 0)
1095                 cfg->fc_metric = IP6_RT_PRIO_USER;
1096
1097         table = fib6_new_table(net, cfg->fc_table);
1098         if (table == NULL) {
1099                 err = -ENOBUFS;
1100                 goto out;
1101         }
1102
1103         rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1104
1105         if (rt == NULL) {
1106                 err = -ENOMEM;
1107                 goto out;
1108         }
1109
1110         rt->u.dst.obsolete = -1;
1111         rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1112                                 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1113                                 0;
1114
1115         if (cfg->fc_protocol == RTPROT_UNSPEC)
1116                 cfg->fc_protocol = RTPROT_BOOT;
1117         rt->rt6i_protocol = cfg->fc_protocol;
1118
1119         addr_type = ipv6_addr_type(&cfg->fc_dst);
1120
1121         if (addr_type & IPV6_ADDR_MULTICAST)
1122                 rt->u.dst.input = ip6_mc_input;
1123         else
1124                 rt->u.dst.input = ip6_forward;
1125
1126         rt->u.dst.output = ip6_output;
1127
1128         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1129         rt->rt6i_dst.plen = cfg->fc_dst_len;
1130         if (rt->rt6i_dst.plen == 128)
1131                rt->u.dst.flags = DST_HOST;
1132
1133 #ifdef CONFIG_IPV6_SUBTREES
1134         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1135         rt->rt6i_src.plen = cfg->fc_src_len;
1136 #endif
1137
1138         rt->rt6i_metric = cfg->fc_metric;
1139
1140         /* We cannot add true routes via loopback here,
1141            they would result in kernel looping; promote them to reject routes
1142          */
1143         if ((cfg->fc_flags & RTF_REJECT) ||
1144             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1145                 /* hold loopback dev/idev if we haven't done so. */
1146                 if (dev != net->loopback_dev) {
1147                         if (dev) {
1148                                 dev_put(dev);
1149                                 in6_dev_put(idev);
1150                         }
1151                         dev = net->loopback_dev;
1152                         dev_hold(dev);
1153                         idev = in6_dev_get(dev);
1154                         if (!idev) {
1155                                 err = -ENODEV;
1156                                 goto out;
1157                         }
1158                 }
1159                 rt->u.dst.output = ip6_pkt_discard_out;
1160                 rt->u.dst.input = ip6_pkt_discard;
1161                 rt->u.dst.error = -ENETUNREACH;
1162                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1163                 goto install_route;
1164         }
1165
1166         if (cfg->fc_flags & RTF_GATEWAY) {
1167                 struct in6_addr *gw_addr;
1168                 int gwa_type;
1169
1170                 gw_addr = &cfg->fc_gateway;
1171                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1172                 gwa_type = ipv6_addr_type(gw_addr);
1173
1174                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1175                         struct rt6_info *grt;
1176
1177                         /* IPv6 strictly inhibits using not link-local
1178                            addresses as nexthop address.
1179                            Otherwise, router will not able to send redirects.
1180                            It is very good, but in some (rare!) circumstances
1181                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1182                            some exceptions. --ANK
1183                          */
1184                         err = -EINVAL;
1185                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1186                                 goto out;
1187
1188                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1189
1190                         err = -EHOSTUNREACH;
1191                         if (grt == NULL)
1192                                 goto out;
1193                         if (dev) {
1194                                 if (dev != grt->rt6i_dev) {
1195                                         dst_release(&grt->u.dst);
1196                                         goto out;
1197                                 }
1198                         } else {
1199                                 dev = grt->rt6i_dev;
1200                                 idev = grt->rt6i_idev;
1201                                 dev_hold(dev);
1202                                 in6_dev_hold(grt->rt6i_idev);
1203                         }
1204                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1205                                 err = 0;
1206                         dst_release(&grt->u.dst);
1207
1208                         if (err)
1209                                 goto out;
1210                 }
1211                 err = -EINVAL;
1212                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1213                         goto out;
1214         }
1215
1216         err = -ENODEV;
1217         if (dev == NULL)
1218                 goto out;
1219
1220         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1221                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1222                 if (IS_ERR(rt->rt6i_nexthop)) {
1223                         err = PTR_ERR(rt->rt6i_nexthop);
1224                         rt->rt6i_nexthop = NULL;
1225                         goto out;
1226                 }
1227         }
1228
1229         rt->rt6i_flags = cfg->fc_flags;
1230
1231 install_route:
1232         if (cfg->fc_mx) {
1233                 struct nlattr *nla;
1234                 int remaining;
1235
1236                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1237                         int type = nla_type(nla);
1238
1239                         if (type) {
1240                                 if (type > RTAX_MAX) {
1241                                         err = -EINVAL;
1242                                         goto out;
1243                                 }
1244
1245                                 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1246                         }
1247                 }
1248         }
1249
1250         if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0)
1251                 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1252         if (!dst_mtu(&rt->u.dst))
1253                 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1254         if (!dst_metric(&rt->u.dst, RTAX_ADVMSS))
1255                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1256         rt->u.dst.dev = dev;
1257         rt->rt6i_idev = idev;
1258         rt->rt6i_table = table;
1259
1260         cfg->fc_nlinfo.nl_net = dev_net(dev);
1261
1262         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1263
1264 out:
1265         if (dev)
1266                 dev_put(dev);
1267         if (idev)
1268                 in6_dev_put(idev);
1269         if (rt)
1270                 dst_free(&rt->u.dst);
1271         return err;
1272 }
1273
1274 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1275 {
1276         int err;
1277         struct fib6_table *table;
1278         struct net *net = dev_net(rt->rt6i_dev);
1279
1280         if (rt == net->ipv6.ip6_null_entry)
1281                 return -ENOENT;
1282
1283         table = rt->rt6i_table;
1284         write_lock_bh(&table->tb6_lock);
1285
1286         err = fib6_del(rt, info);
1287         dst_release(&rt->u.dst);
1288
1289         write_unlock_bh(&table->tb6_lock);
1290
1291         return err;
1292 }
1293
1294 int ip6_del_rt(struct rt6_info *rt)
1295 {
1296         struct nl_info info = {
1297                 .nl_net = dev_net(rt->rt6i_dev),
1298         };
1299         return __ip6_del_rt(rt, &info);
1300 }
1301
1302 static int ip6_route_del(struct fib6_config *cfg)
1303 {
1304         struct fib6_table *table;
1305         struct fib6_node *fn;
1306         struct rt6_info *rt;
1307         int err = -ESRCH;
1308
1309         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1310         if (table == NULL)
1311                 return err;
1312
1313         read_lock_bh(&table->tb6_lock);
1314
1315         fn = fib6_locate(&table->tb6_root,
1316                          &cfg->fc_dst, cfg->fc_dst_len,
1317                          &cfg->fc_src, cfg->fc_src_len);
1318
1319         if (fn) {
1320                 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1321                         if (cfg->fc_ifindex &&
1322                             (rt->rt6i_dev == NULL ||
1323                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1324                                 continue;
1325                         if (cfg->fc_flags & RTF_GATEWAY &&
1326                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1327                                 continue;
1328                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1329                                 continue;
1330                         dst_hold(&rt->u.dst);
1331                         read_unlock_bh(&table->tb6_lock);
1332
1333                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1334                 }
1335         }
1336         read_unlock_bh(&table->tb6_lock);
1337
1338         return err;
1339 }
1340
1341 /*
1342  *      Handle redirects
1343  */
1344 struct ip6rd_flowi {
1345         struct flowi fl;
1346         struct in6_addr gateway;
1347 };
1348
1349 static struct rt6_info *__ip6_route_redirect(struct net *net,
1350                                              struct fib6_table *table,
1351                                              struct flowi *fl,
1352                                              int flags)
1353 {
1354         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1355         struct rt6_info *rt;
1356         struct fib6_node *fn;
1357
1358         /*
1359          * Get the "current" route for this destination and
1360          * check if the redirect has come from approriate router.
1361          *
1362          * RFC 2461 specifies that redirects should only be
1363          * accepted if they come from the nexthop to the target.
1364          * Due to the way the routes are chosen, this notion
1365          * is a bit fuzzy and one might need to check all possible
1366          * routes.
1367          */
1368
1369         read_lock_bh(&table->tb6_lock);
1370         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1371 restart:
1372         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1373                 /*
1374                  * Current route is on-link; redirect is always invalid.
1375                  *
1376                  * Seems, previous statement is not true. It could
1377                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1378                  * But then router serving it might decide, that we should
1379                  * know truth 8)8) --ANK (980726).
1380                  */
1381                 if (rt6_check_expired(rt))
1382                         continue;
1383                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1384                         continue;
1385                 if (fl->oif != rt->rt6i_dev->ifindex)
1386                         continue;
1387                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1388                         continue;
1389                 break;
1390         }
1391
1392         if (!rt)
1393                 rt = net->ipv6.ip6_null_entry;
1394         BACKTRACK(net, &fl->fl6_src);
1395 out:
1396         dst_hold(&rt->u.dst);
1397
1398         read_unlock_bh(&table->tb6_lock);
1399
1400         return rt;
1401 };
1402
1403 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1404                                            struct in6_addr *src,
1405                                            struct in6_addr *gateway,
1406                                            struct net_device *dev)
1407 {
1408         int flags = RT6_LOOKUP_F_HAS_SADDR;
1409         struct net *net = dev_net(dev);
1410         struct ip6rd_flowi rdfl = {
1411                 .fl = {
1412                         .oif = dev->ifindex,
1413                         .nl_u = {
1414                                 .ip6_u = {
1415                                         .daddr = *dest,
1416                                         .saddr = *src,
1417                                 },
1418                         },
1419                 },
1420                 .gateway = *gateway,
1421         };
1422
1423         if (rt6_need_strict(dest))
1424                 flags |= RT6_LOOKUP_F_IFACE;
1425
1426         return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1427                                                    flags, __ip6_route_redirect);
1428 }
1429
1430 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1431                   struct in6_addr *saddr,
1432                   struct neighbour *neigh, u8 *lladdr, int on_link)
1433 {
1434         struct rt6_info *rt, *nrt = NULL;
1435         struct netevent_redirect netevent;
1436         struct net *net = dev_net(neigh->dev);
1437
1438         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1439
1440         if (rt == net->ipv6.ip6_null_entry) {
1441                 if (net_ratelimit())
1442                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1443                                "for redirect target\n");
1444                 goto out;
1445         }
1446
1447         /*
1448          *      We have finally decided to accept it.
1449          */
1450
1451         neigh_update(neigh, lladdr, NUD_STALE,
1452                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1453                      NEIGH_UPDATE_F_OVERRIDE|
1454                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1455                                      NEIGH_UPDATE_F_ISROUTER))
1456                      );
1457
1458         /*
1459          * Redirect received -> path was valid.
1460          * Look, redirects are sent only in response to data packets,
1461          * so that this nexthop apparently is reachable. --ANK
1462          */
1463         dst_confirm(&rt->u.dst);
1464
1465         /* Duplicate redirect: silently ignore. */
1466         if (neigh == rt->u.dst.neighbour)
1467                 goto out;
1468
1469         nrt = ip6_rt_copy(rt);
1470         if (nrt == NULL)
1471                 goto out;
1472
1473         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1474         if (on_link)
1475                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1476
1477         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1478         nrt->rt6i_dst.plen = 128;
1479         nrt->u.dst.flags |= DST_HOST;
1480
1481         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1482         nrt->rt6i_nexthop = neigh_clone(neigh);
1483         /* Reset pmtu, it may be better */
1484         nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1485         nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1486                                                         dst_mtu(&nrt->u.dst));
1487
1488         if (ip6_ins_rt(nrt))
1489                 goto out;
1490
1491         netevent.old = &rt->u.dst;
1492         netevent.new = &nrt->u.dst;
1493         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1494
1495         if (rt->rt6i_flags&RTF_CACHE) {
1496                 ip6_del_rt(rt);
1497                 return;
1498         }
1499
1500 out:
1501         dst_release(&rt->u.dst);
1502         return;
1503 }
1504
1505 /*
1506  *      Handle ICMP "packet too big" messages
1507  *      i.e. Path MTU discovery
1508  */
1509
1510 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1511                         struct net_device *dev, u32 pmtu)
1512 {
1513         struct rt6_info *rt, *nrt;
1514         struct net *net = dev_net(dev);
1515         int allfrag = 0;
1516
1517         rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1518         if (rt == NULL)
1519                 return;
1520
1521         if (pmtu >= dst_mtu(&rt->u.dst))
1522                 goto out;
1523
1524         if (pmtu < IPV6_MIN_MTU) {
1525                 /*
1526                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1527                  * MTU (1280) and a fragment header should always be included
1528                  * after a node receiving Too Big message reporting PMTU is
1529                  * less than the IPv6 Minimum Link MTU.
1530                  */
1531                 pmtu = IPV6_MIN_MTU;
1532                 allfrag = 1;
1533         }
1534
1535         /* New mtu received -> path was valid.
1536            They are sent only in response to data packets,
1537            so that this nexthop apparently is reachable. --ANK
1538          */
1539         dst_confirm(&rt->u.dst);
1540
1541         /* Host route. If it is static, it would be better
1542            not to override it, but add new one, so that
1543            when cache entry will expire old pmtu
1544            would return automatically.
1545          */
1546         if (rt->rt6i_flags & RTF_CACHE) {
1547                 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1548                 if (allfrag)
1549                         rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1550                 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1551                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1552                 goto out;
1553         }
1554
1555         /* Network route.
1556            Two cases are possible:
1557            1. It is connected route. Action: COW
1558            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1559          */
1560         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1561                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1562         else
1563                 nrt = rt6_alloc_clone(rt, daddr);
1564
1565         if (nrt) {
1566                 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1567                 if (allfrag)
1568                         nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1569
1570                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1571                  * happened within 5 mins, the recommended timer is 10 mins.
1572                  * Here this route expiration time is set to ip6_rt_mtu_expires
1573                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1574                  * and detecting PMTU increase will be automatically happened.
1575                  */
1576                 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1577                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1578
1579                 ip6_ins_rt(nrt);
1580         }
1581 out:
1582         dst_release(&rt->u.dst);
1583 }
1584
1585 /*
1586  *      Misc support functions
1587  */
1588
1589 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1590 {
1591         struct net *net = dev_net(ort->rt6i_dev);
1592         struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1593
1594         if (rt) {
1595                 rt->u.dst.input = ort->u.dst.input;
1596                 rt->u.dst.output = ort->u.dst.output;
1597
1598                 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1599                 rt->u.dst.error = ort->u.dst.error;
1600                 rt->u.dst.dev = ort->u.dst.dev;
1601                 if (rt->u.dst.dev)
1602                         dev_hold(rt->u.dst.dev);
1603                 rt->rt6i_idev = ort->rt6i_idev;
1604                 if (rt->rt6i_idev)
1605                         in6_dev_hold(rt->rt6i_idev);
1606                 rt->u.dst.lastuse = jiffies;
1607                 rt->rt6i_expires = 0;
1608
1609                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1610                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1611                 rt->rt6i_metric = 0;
1612
1613                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1614 #ifdef CONFIG_IPV6_SUBTREES
1615                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1616 #endif
1617                 rt->rt6i_table = ort->rt6i_table;
1618         }
1619         return rt;
1620 }
1621
1622 #ifdef CONFIG_IPV6_ROUTE_INFO
1623 static struct rt6_info *rt6_get_route_info(struct net *net,
1624                                            struct in6_addr *prefix, int prefixlen,
1625                                            struct in6_addr *gwaddr, int ifindex)
1626 {
1627         struct fib6_node *fn;
1628         struct rt6_info *rt = NULL;
1629         struct fib6_table *table;
1630
1631         table = fib6_get_table(net, RT6_TABLE_INFO);
1632         if (table == NULL)
1633                 return NULL;
1634
1635         write_lock_bh(&table->tb6_lock);
1636         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1637         if (!fn)
1638                 goto out;
1639
1640         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1641                 if (rt->rt6i_dev->ifindex != ifindex)
1642                         continue;
1643                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1644                         continue;
1645                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1646                         continue;
1647                 dst_hold(&rt->u.dst);
1648                 break;
1649         }
1650 out:
1651         write_unlock_bh(&table->tb6_lock);
1652         return rt;
1653 }
1654
1655 static struct rt6_info *rt6_add_route_info(struct net *net,
1656                                            struct in6_addr *prefix, int prefixlen,
1657                                            struct in6_addr *gwaddr, int ifindex,
1658                                            unsigned pref)
1659 {
1660         struct fib6_config cfg = {
1661                 .fc_table       = RT6_TABLE_INFO,
1662                 .fc_metric      = IP6_RT_PRIO_USER,
1663                 .fc_ifindex     = ifindex,
1664                 .fc_dst_len     = prefixlen,
1665                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1666                                   RTF_UP | RTF_PREF(pref),
1667                 .fc_nlinfo.pid = 0,
1668                 .fc_nlinfo.nlh = NULL,
1669                 .fc_nlinfo.nl_net = net,
1670         };
1671
1672         ipv6_addr_copy(&cfg.fc_dst, prefix);
1673         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1674
1675         /* We should treat it as a default route if prefix length is 0. */
1676         if (!prefixlen)
1677                 cfg.fc_flags |= RTF_DEFAULT;
1678
1679         ip6_route_add(&cfg);
1680
1681         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1682 }
1683 #endif
1684
1685 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1686 {
1687         struct rt6_info *rt;
1688         struct fib6_table *table;
1689
1690         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1691         if (table == NULL)
1692                 return NULL;
1693
1694         write_lock_bh(&table->tb6_lock);
1695         for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1696                 if (dev == rt->rt6i_dev &&
1697                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1698                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1699                         break;
1700         }
1701         if (rt)
1702                 dst_hold(&rt->u.dst);
1703         write_unlock_bh(&table->tb6_lock);
1704         return rt;
1705 }
1706
1707 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1708                                      struct net_device *dev,
1709                                      unsigned int pref)
1710 {
1711         struct fib6_config cfg = {
1712                 .fc_table       = RT6_TABLE_DFLT,
1713                 .fc_metric      = IP6_RT_PRIO_USER,
1714                 .fc_ifindex     = dev->ifindex,
1715                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1716                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1717                 .fc_nlinfo.pid = 0,
1718                 .fc_nlinfo.nlh = NULL,
1719                 .fc_nlinfo.nl_net = dev_net(dev),
1720         };
1721
1722         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1723
1724         ip6_route_add(&cfg);
1725
1726         return rt6_get_dflt_router(gwaddr, dev);
1727 }
1728
1729 void rt6_purge_dflt_routers(struct net *net)
1730 {
1731         struct rt6_info *rt;
1732         struct fib6_table *table;
1733
1734         /* NOTE: Keep consistent with rt6_get_dflt_router */
1735         table = fib6_get_table(net, RT6_TABLE_DFLT);
1736         if (table == NULL)
1737                 return;
1738
1739 restart:
1740         read_lock_bh(&table->tb6_lock);
1741         for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1742                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1743                         dst_hold(&rt->u.dst);
1744                         read_unlock_bh(&table->tb6_lock);
1745                         ip6_del_rt(rt);
1746                         goto restart;
1747                 }
1748         }
1749         read_unlock_bh(&table->tb6_lock);
1750 }
1751
1752 static void rtmsg_to_fib6_config(struct net *net,
1753                                  struct in6_rtmsg *rtmsg,
1754                                  struct fib6_config *cfg)
1755 {
1756         memset(cfg, 0, sizeof(*cfg));
1757
1758         cfg->fc_table = RT6_TABLE_MAIN;
1759         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1760         cfg->fc_metric = rtmsg->rtmsg_metric;
1761         cfg->fc_expires = rtmsg->rtmsg_info;
1762         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1763         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1764         cfg->fc_flags = rtmsg->rtmsg_flags;
1765
1766         cfg->fc_nlinfo.nl_net = net;
1767
1768         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1769         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1770         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1771 }
1772
1773 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1774 {
1775         struct fib6_config cfg;
1776         struct in6_rtmsg rtmsg;
1777         int err;
1778
1779         switch(cmd) {
1780         case SIOCADDRT:         /* Add a route */
1781         case SIOCDELRT:         /* Delete a route */
1782                 if (!capable(CAP_NET_ADMIN))
1783                         return -EPERM;
1784                 err = copy_from_user(&rtmsg, arg,
1785                                      sizeof(struct in6_rtmsg));
1786                 if (err)
1787                         return -EFAULT;
1788
1789                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1790
1791                 rtnl_lock();
1792                 switch (cmd) {
1793                 case SIOCADDRT:
1794                         err = ip6_route_add(&cfg);
1795                         break;
1796                 case SIOCDELRT:
1797                         err = ip6_route_del(&cfg);
1798                         break;
1799                 default:
1800                         err = -EINVAL;
1801                 }
1802                 rtnl_unlock();
1803
1804                 return err;
1805         }
1806
1807         return -EINVAL;
1808 }
1809
1810 /*
1811  *      Drop the packet on the floor
1812  */
1813
1814 static int ip6_pkt_drop(struct sk_buff *skb, int code, int ipstats_mib_noroutes)
1815 {
1816         int type;
1817         switch (ipstats_mib_noroutes) {
1818         case IPSTATS_MIB_INNOROUTES:
1819                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1820                 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1821                         IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1822                         break;
1823                 }
1824                 /* FALLTHROUGH */
1825         case IPSTATS_MIB_OUTNOROUTES:
1826                 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1827                 break;
1828         }
1829         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1830         kfree_skb(skb);
1831         return 0;
1832 }
1833
1834 static int ip6_pkt_discard(struct sk_buff *skb)
1835 {
1836         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1837 }
1838
1839 static int ip6_pkt_discard_out(struct sk_buff *skb)
1840 {
1841         skb->dev = skb->dst->dev;
1842         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1843 }
1844
1845 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1846
1847 static int ip6_pkt_prohibit(struct sk_buff *skb)
1848 {
1849         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1850 }
1851
1852 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1853 {
1854         skb->dev = skb->dst->dev;
1855         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1856 }
1857
1858 #endif
1859
1860 /*
1861  *      Allocate a dst for local (unicast / anycast) address.
1862  */
1863
1864 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1865                                     const struct in6_addr *addr,
1866                                     int anycast)
1867 {
1868         struct net *net = dev_net(idev->dev);
1869         struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1870
1871         if (rt == NULL)
1872                 return ERR_PTR(-ENOMEM);
1873
1874         dev_hold(net->loopback_dev);
1875         in6_dev_hold(idev);
1876
1877         rt->u.dst.flags = DST_HOST;
1878         rt->u.dst.input = ip6_input;
1879         rt->u.dst.output = ip6_output;
1880         rt->rt6i_dev = net->loopback_dev;
1881         rt->rt6i_idev = idev;
1882         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1883         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1884         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1885         rt->u.dst.obsolete = -1;
1886
1887         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1888         if (anycast)
1889                 rt->rt6i_flags |= RTF_ANYCAST;
1890         else
1891                 rt->rt6i_flags |= RTF_LOCAL;
1892         rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1893         if (rt->rt6i_nexthop == NULL) {
1894                 dst_free(&rt->u.dst);
1895                 return ERR_PTR(-ENOMEM);
1896         }
1897
1898         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1899         rt->rt6i_dst.plen = 128;
1900         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1901
1902         atomic_set(&rt->u.dst.__refcnt, 1);
1903
1904         return rt;
1905 }
1906
1907 struct arg_dev_net {
1908         struct net_device *dev;
1909         struct net *net;
1910 };
1911
1912 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1913 {
1914         struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1915         struct net *net = ((struct arg_dev_net *)arg)->net;
1916
1917         if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1918             rt != net->ipv6.ip6_null_entry) {
1919                 RT6_TRACE("deleted by ifdown %p\n", rt);
1920                 return -1;
1921         }
1922         return 0;
1923 }
1924
1925 void rt6_ifdown(struct net *net, struct net_device *dev)
1926 {
1927         struct arg_dev_net adn = {
1928                 .dev = dev,
1929                 .net = net,
1930         };
1931
1932         fib6_clean_all(net, fib6_ifdown, 0, &adn);
1933 }
1934
1935 struct rt6_mtu_change_arg
1936 {
1937         struct net_device *dev;
1938         unsigned mtu;
1939 };
1940
1941 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1942 {
1943         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1944         struct inet6_dev *idev;
1945         struct net *net = dev_net(arg->dev);
1946
1947         /* In IPv6 pmtu discovery is not optional,
1948            so that RTAX_MTU lock cannot disable it.
1949            We still use this lock to block changes
1950            caused by addrconf/ndisc.
1951         */
1952
1953         idev = __in6_dev_get(arg->dev);
1954         if (idev == NULL)
1955                 return 0;
1956
1957         /* For administrative MTU increase, there is no way to discover
1958            IPv6 PMTU increase, so PMTU increase should be updated here.
1959            Since RFC 1981 doesn't include administrative MTU increase
1960            update PMTU increase is a MUST. (i.e. jumbo frame)
1961          */
1962         /*
1963            If new MTU is less than route PMTU, this new MTU will be the
1964            lowest MTU in the path, update the route PMTU to reflect PMTU
1965            decreases; if new MTU is greater than route PMTU, and the
1966            old MTU is the lowest MTU in the path, update the route PMTU
1967            to reflect the increase. In this case if the other nodes' MTU
1968            also have the lowest MTU, TOO BIG MESSAGE will be lead to
1969            PMTU discouvery.
1970          */
1971         if (rt->rt6i_dev == arg->dev &&
1972             !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1973             (dst_mtu(&rt->u.dst) >= arg->mtu ||
1974              (dst_mtu(&rt->u.dst) < arg->mtu &&
1975               dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1976                 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1977                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
1978         }
1979         return 0;
1980 }
1981
1982 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1983 {
1984         struct rt6_mtu_change_arg arg = {
1985                 .dev = dev,
1986                 .mtu = mtu,
1987         };
1988
1989         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
1990 }
1991
1992 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1993         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
1994         [RTA_OIF]               = { .type = NLA_U32 },
1995         [RTA_IIF]               = { .type = NLA_U32 },
1996         [RTA_PRIORITY]          = { .type = NLA_U32 },
1997         [RTA_METRICS]           = { .type = NLA_NESTED },
1998 };
1999
2000 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2001                               struct fib6_config *cfg)
2002 {
2003         struct rtmsg *rtm;
2004         struct nlattr *tb[RTA_MAX+1];
2005         int err;
2006
2007         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2008         if (err < 0)
2009                 goto errout;
2010
2011         err = -EINVAL;
2012         rtm = nlmsg_data(nlh);
2013         memset(cfg, 0, sizeof(*cfg));
2014
2015         cfg->fc_table = rtm->rtm_table;
2016         cfg->fc_dst_len = rtm->rtm_dst_len;
2017         cfg->fc_src_len = rtm->rtm_src_len;
2018         cfg->fc_flags = RTF_UP;
2019         cfg->fc_protocol = rtm->rtm_protocol;
2020
2021         if (rtm->rtm_type == RTN_UNREACHABLE)
2022                 cfg->fc_flags |= RTF_REJECT;
2023
2024         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2025         cfg->fc_nlinfo.nlh = nlh;
2026         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2027
2028         if (tb[RTA_GATEWAY]) {
2029                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2030                 cfg->fc_flags |= RTF_GATEWAY;
2031         }
2032
2033         if (tb[RTA_DST]) {
2034                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2035
2036                 if (nla_len(tb[RTA_DST]) < plen)
2037                         goto errout;
2038
2039                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2040         }
2041
2042         if (tb[RTA_SRC]) {
2043                 int plen = (rtm->rtm_src_len + 7) >> 3;
2044
2045                 if (nla_len(tb[RTA_SRC]) < plen)
2046                         goto errout;
2047
2048                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2049         }
2050
2051         if (tb[RTA_OIF])
2052                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2053
2054         if (tb[RTA_PRIORITY])
2055                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2056
2057         if (tb[RTA_METRICS]) {
2058                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2059                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2060         }
2061
2062         if (tb[RTA_TABLE])
2063                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2064
2065         err = 0;
2066 errout:
2067         return err;
2068 }
2069
2070 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2071 {
2072         struct fib6_config cfg;
2073         int err;
2074
2075         err = rtm_to_fib6_config(skb, nlh, &cfg);
2076         if (err < 0)
2077                 return err;
2078
2079         return ip6_route_del(&cfg);
2080 }
2081
2082 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2083 {
2084         struct fib6_config cfg;
2085         int err;
2086
2087         err = rtm_to_fib6_config(skb, nlh, &cfg);
2088         if (err < 0)
2089                 return err;
2090
2091         return ip6_route_add(&cfg);
2092 }
2093
2094 static inline size_t rt6_nlmsg_size(void)
2095 {
2096         return NLMSG_ALIGN(sizeof(struct rtmsg))
2097                + nla_total_size(16) /* RTA_SRC */
2098                + nla_total_size(16) /* RTA_DST */
2099                + nla_total_size(16) /* RTA_GATEWAY */
2100                + nla_total_size(16) /* RTA_PREFSRC */
2101                + nla_total_size(4) /* RTA_TABLE */
2102                + nla_total_size(4) /* RTA_IIF */
2103                + nla_total_size(4) /* RTA_OIF */
2104                + nla_total_size(4) /* RTA_PRIORITY */
2105                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2106                + nla_total_size(sizeof(struct rta_cacheinfo));
2107 }
2108
2109 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2110                          struct in6_addr *dst, struct in6_addr *src,
2111                          int iif, int type, u32 pid, u32 seq,
2112                          int prefix, int nowait, unsigned int flags)
2113 {
2114         struct rtmsg *rtm;
2115         struct nlmsghdr *nlh;
2116         long expires;
2117         u32 table;
2118
2119         if (prefix) {   /* user wants prefix routes only */
2120                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2121                         /* success since this is not a prefix route */
2122                         return 1;
2123                 }
2124         }
2125
2126         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2127         if (nlh == NULL)
2128                 return -EMSGSIZE;
2129
2130         rtm = nlmsg_data(nlh);
2131         rtm->rtm_family = AF_INET6;
2132         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2133         rtm->rtm_src_len = rt->rt6i_src.plen;
2134         rtm->rtm_tos = 0;
2135         if (rt->rt6i_table)
2136                 table = rt->rt6i_table->tb6_id;
2137         else
2138                 table = RT6_TABLE_UNSPEC;
2139         rtm->rtm_table = table;
2140         NLA_PUT_U32(skb, RTA_TABLE, table);
2141         if (rt->rt6i_flags&RTF_REJECT)
2142                 rtm->rtm_type = RTN_UNREACHABLE;
2143         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2144                 rtm->rtm_type = RTN_LOCAL;
2145         else
2146                 rtm->rtm_type = RTN_UNICAST;
2147         rtm->rtm_flags = 0;
2148         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2149         rtm->rtm_protocol = rt->rt6i_protocol;
2150         if (rt->rt6i_flags&RTF_DYNAMIC)
2151                 rtm->rtm_protocol = RTPROT_REDIRECT;
2152         else if (rt->rt6i_flags & RTF_ADDRCONF)
2153                 rtm->rtm_protocol = RTPROT_KERNEL;
2154         else if (rt->rt6i_flags&RTF_DEFAULT)
2155                 rtm->rtm_protocol = RTPROT_RA;
2156
2157         if (rt->rt6i_flags&RTF_CACHE)
2158                 rtm->rtm_flags |= RTM_F_CLONED;
2159
2160         if (dst) {
2161                 NLA_PUT(skb, RTA_DST, 16, dst);
2162                 rtm->rtm_dst_len = 128;
2163         } else if (rtm->rtm_dst_len)
2164                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2165 #ifdef CONFIG_IPV6_SUBTREES
2166         if (src) {
2167                 NLA_PUT(skb, RTA_SRC, 16, src);
2168                 rtm->rtm_src_len = 128;
2169         } else if (rtm->rtm_src_len)
2170                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2171 #endif
2172         if (iif) {
2173 #ifdef CONFIG_IPV6_MROUTE
2174                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2175                         int err = ip6mr_get_route(skb, rtm, nowait);
2176                         if (err <= 0) {
2177                                 if (!nowait) {
2178                                         if (err == 0)
2179                                                 return 0;
2180                                         goto nla_put_failure;
2181                                 } else {
2182                                         if (err == -EMSGSIZE)
2183                                                 goto nla_put_failure;
2184                                 }
2185                         }
2186                 } else
2187 #endif
2188                         NLA_PUT_U32(skb, RTA_IIF, iif);
2189         } else if (dst) {
2190                 struct inet6_dev *idev = ip6_dst_idev(&rt->u.dst);
2191                 struct in6_addr saddr_buf;
2192                 if (ipv6_dev_get_saddr(idev ? idev->dev : NULL,
2193                                        dst, 0, &saddr_buf) == 0)
2194                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2195         }
2196
2197         if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2198                 goto nla_put_failure;
2199
2200         if (rt->u.dst.neighbour)
2201                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2202
2203         if (rt->u.dst.dev)
2204                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2205
2206         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2207
2208         if (!(rt->rt6i_flags & RTF_EXPIRES))
2209                 expires = 0;
2210         else if (rt->rt6i_expires - jiffies < INT_MAX)
2211                 expires = rt->rt6i_expires - jiffies;
2212         else
2213                 expires = INT_MAX;
2214
2215         if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2216                                expires, rt->u.dst.error) < 0)
2217                 goto nla_put_failure;
2218
2219         return nlmsg_end(skb, nlh);
2220
2221 nla_put_failure:
2222         nlmsg_cancel(skb, nlh);
2223         return -EMSGSIZE;
2224 }
2225
2226 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2227 {
2228         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2229         int prefix;
2230
2231         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2232                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2233                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2234         } else
2235                 prefix = 0;
2236
2237         return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2238                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2239                      prefix, 0, NLM_F_MULTI);
2240 }
2241
2242 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2243 {
2244         struct net *net = sock_net(in_skb->sk);
2245         struct nlattr *tb[RTA_MAX+1];
2246         struct rt6_info *rt;
2247         struct sk_buff *skb;
2248         struct rtmsg *rtm;
2249         struct flowi fl;
2250         int err, iif = 0;
2251
2252         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2253         if (err < 0)
2254                 goto errout;
2255
2256         err = -EINVAL;
2257         memset(&fl, 0, sizeof(fl));
2258
2259         if (tb[RTA_SRC]) {
2260                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2261                         goto errout;
2262
2263                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2264         }
2265
2266         if (tb[RTA_DST]) {
2267                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2268                         goto errout;
2269
2270                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2271         }
2272
2273         if (tb[RTA_IIF])
2274                 iif = nla_get_u32(tb[RTA_IIF]);
2275
2276         if (tb[RTA_OIF])
2277                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2278
2279         if (iif) {
2280                 struct net_device *dev;
2281                 dev = __dev_get_by_index(net, iif);
2282                 if (!dev) {
2283                         err = -ENODEV;
2284                         goto errout;
2285                 }
2286         }
2287
2288         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2289         if (skb == NULL) {
2290                 err = -ENOBUFS;
2291                 goto errout;
2292         }
2293
2294         /* Reserve room for dummy headers, this skb can pass
2295            through good chunk of routing engine.
2296          */
2297         skb_reset_mac_header(skb);
2298         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2299
2300         rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2301         skb->dst = &rt->u.dst;
2302
2303         err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2304                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2305                             nlh->nlmsg_seq, 0, 0, 0);
2306         if (err < 0) {
2307                 kfree_skb(skb);
2308                 goto errout;
2309         }
2310
2311         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2312 errout:
2313         return err;
2314 }
2315
2316 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2317 {
2318         struct sk_buff *skb;
2319         struct net *net = info->nl_net;
2320         u32 seq;
2321         int err;
2322
2323         err = -ENOBUFS;
2324         seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2325
2326         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2327         if (skb == NULL)
2328                 goto errout;
2329
2330         err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2331                                 event, info->pid, seq, 0, 0, 0);
2332         if (err < 0) {
2333                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2334                 WARN_ON(err == -EMSGSIZE);
2335                 kfree_skb(skb);
2336                 goto errout;
2337         }
2338         err = rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2339                           info->nlh, gfp_any());
2340 errout:
2341         if (err < 0)
2342                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2343 }
2344
2345 static int ip6_route_dev_notify(struct notifier_block *this,
2346                                 unsigned long event, void *data)
2347 {
2348         struct net_device *dev = (struct net_device *)data;
2349         struct net *net = dev_net(dev);
2350
2351         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2352                 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2353                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2354 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2355                 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2356                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2357                 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2358                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2359 #endif
2360         }
2361
2362         return NOTIFY_OK;
2363 }
2364
2365 /*
2366  *      /proc
2367  */
2368
2369 #ifdef CONFIG_PROC_FS
2370
2371 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2372
2373 struct rt6_proc_arg
2374 {
2375         char *buffer;
2376         int offset;
2377         int length;
2378         int skip;
2379         int len;
2380 };
2381
2382 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2383 {
2384         struct seq_file *m = p_arg;
2385
2386         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2387                    rt->rt6i_dst.plen);
2388
2389 #ifdef CONFIG_IPV6_SUBTREES
2390         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2391                    rt->rt6i_src.plen);
2392 #else
2393         seq_puts(m, "00000000000000000000000000000000 00 ");
2394 #endif
2395
2396         if (rt->rt6i_nexthop) {
2397                 seq_printf(m, NIP6_SEQFMT,
2398                            NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2399         } else {
2400                 seq_puts(m, "00000000000000000000000000000000");
2401         }
2402         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2403                    rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2404                    rt->u.dst.__use, rt->rt6i_flags,
2405                    rt->rt6i_dev ? rt->rt6i_dev->name : "");
2406         return 0;
2407 }
2408
2409 static int ipv6_route_show(struct seq_file *m, void *v)
2410 {
2411         struct net *net = (struct net *)m->private;
2412         fib6_clean_all(net, rt6_info_route, 0, m);
2413         return 0;
2414 }
2415
2416 static int ipv6_route_open(struct inode *inode, struct file *file)
2417 {
2418         return single_open_net(inode, file, ipv6_route_show);
2419 }
2420
2421 static const struct file_operations ipv6_route_proc_fops = {
2422         .owner          = THIS_MODULE,
2423         .open           = ipv6_route_open,
2424         .read           = seq_read,
2425         .llseek         = seq_lseek,
2426         .release        = single_release_net,
2427 };
2428
2429 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2430 {
2431         struct net *net = (struct net *)seq->private;
2432         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2433                    net->ipv6.rt6_stats->fib_nodes,
2434                    net->ipv6.rt6_stats->fib_route_nodes,
2435                    net->ipv6.rt6_stats->fib_rt_alloc,
2436                    net->ipv6.rt6_stats->fib_rt_entries,
2437                    net->ipv6.rt6_stats->fib_rt_cache,
2438                    atomic_read(&net->ipv6.ip6_dst_ops->entries),
2439                    net->ipv6.rt6_stats->fib_discarded_routes);
2440
2441         return 0;
2442 }
2443
2444 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2445 {
2446         return single_open_net(inode, file, rt6_stats_seq_show);
2447 }
2448
2449 static const struct file_operations rt6_stats_seq_fops = {
2450         .owner   = THIS_MODULE,
2451         .open    = rt6_stats_seq_open,
2452         .read    = seq_read,
2453         .llseek  = seq_lseek,
2454         .release = single_release_net,
2455 };
2456 #endif  /* CONFIG_PROC_FS */
2457
2458 #ifdef CONFIG_SYSCTL
2459
2460 static
2461 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2462                               void __user *buffer, size_t *lenp, loff_t *ppos)
2463 {
2464         struct net *net = current->nsproxy->net_ns;
2465         int delay = net->ipv6.sysctl.flush_delay;
2466         if (write) {
2467                 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2468                 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2469                 return 0;
2470         } else
2471                 return -EINVAL;
2472 }
2473
2474 ctl_table ipv6_route_table_template[] = {
2475         {
2476                 .procname       =       "flush",
2477                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2478                 .maxlen         =       sizeof(int),
2479                 .mode           =       0200,
2480                 .proc_handler   =       &ipv6_sysctl_rtcache_flush
2481         },
2482         {
2483                 .ctl_name       =       NET_IPV6_ROUTE_GC_THRESH,
2484                 .procname       =       "gc_thresh",
2485                 .data           =       &ip6_dst_ops_template.gc_thresh,
2486                 .maxlen         =       sizeof(int),
2487                 .mode           =       0644,
2488                 .proc_handler   =       &proc_dointvec,
2489         },
2490         {
2491                 .ctl_name       =       NET_IPV6_ROUTE_MAX_SIZE,
2492                 .procname       =       "max_size",
2493                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2494                 .maxlen         =       sizeof(int),
2495                 .mode           =       0644,
2496                 .proc_handler   =       &proc_dointvec,
2497         },
2498         {
2499                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2500                 .procname       =       "gc_min_interval",
2501                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2502                 .maxlen         =       sizeof(int),
2503                 .mode           =       0644,
2504                 .proc_handler   =       &proc_dointvec_jiffies,
2505                 .strategy       =       &sysctl_jiffies,
2506         },
2507         {
2508                 .ctl_name       =       NET_IPV6_ROUTE_GC_TIMEOUT,
2509                 .procname       =       "gc_timeout",
2510                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2511                 .maxlen         =       sizeof(int),
2512                 .mode           =       0644,
2513                 .proc_handler   =       &proc_dointvec_jiffies,
2514                 .strategy       =       &sysctl_jiffies,
2515         },
2516         {
2517                 .ctl_name       =       NET_IPV6_ROUTE_GC_INTERVAL,
2518                 .procname       =       "gc_interval",
2519                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2520                 .maxlen         =       sizeof(int),
2521                 .mode           =       0644,
2522                 .proc_handler   =       &proc_dointvec_jiffies,
2523                 .strategy       =       &sysctl_jiffies,
2524         },
2525         {
2526                 .ctl_name       =       NET_IPV6_ROUTE_GC_ELASTICITY,
2527                 .procname       =       "gc_elasticity",
2528                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2529                 .maxlen         =       sizeof(int),
2530                 .mode           =       0644,
2531                 .proc_handler   =       &proc_dointvec_jiffies,
2532                 .strategy       =       &sysctl_jiffies,
2533         },
2534         {
2535                 .ctl_name       =       NET_IPV6_ROUTE_MTU_EXPIRES,
2536                 .procname       =       "mtu_expires",
2537                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2538                 .maxlen         =       sizeof(int),
2539                 .mode           =       0644,
2540                 .proc_handler   =       &proc_dointvec_jiffies,
2541                 .strategy       =       &sysctl_jiffies,
2542         },
2543         {
2544                 .ctl_name       =       NET_IPV6_ROUTE_MIN_ADVMSS,
2545                 .procname       =       "min_adv_mss",
2546                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2547                 .maxlen         =       sizeof(int),
2548                 .mode           =       0644,
2549                 .proc_handler   =       &proc_dointvec_jiffies,
2550                 .strategy       =       &sysctl_jiffies,
2551         },
2552         {
2553                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2554                 .procname       =       "gc_min_interval_ms",
2555                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2556                 .maxlen         =       sizeof(int),
2557                 .mode           =       0644,
2558                 .proc_handler   =       &proc_dointvec_ms_jiffies,
2559                 .strategy       =       &sysctl_ms_jiffies,
2560         },
2561         { .ctl_name = 0 }
2562 };
2563
2564 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2565 {
2566         struct ctl_table *table;
2567
2568         table = kmemdup(ipv6_route_table_template,
2569                         sizeof(ipv6_route_table_template),
2570                         GFP_KERNEL);
2571
2572         if (table) {
2573                 table[0].data = &net->ipv6.sysctl.flush_delay;
2574                 table[1].data = &net->ipv6.ip6_dst_ops->gc_thresh;
2575                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2576                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2577                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2578                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2579                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2580                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2581                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2582         }
2583
2584         return table;
2585 }
2586 #endif
2587
2588 static int ip6_route_net_init(struct net *net)
2589 {
2590         int ret = -ENOMEM;
2591
2592         net->ipv6.ip6_dst_ops = kmemdup(&ip6_dst_ops_template,
2593                                         sizeof(*net->ipv6.ip6_dst_ops),
2594                                         GFP_KERNEL);
2595         if (!net->ipv6.ip6_dst_ops)
2596                 goto out;
2597         net->ipv6.ip6_dst_ops->dst_net = hold_net(net);
2598
2599         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2600                                            sizeof(*net->ipv6.ip6_null_entry),
2601                                            GFP_KERNEL);
2602         if (!net->ipv6.ip6_null_entry)
2603                 goto out_ip6_dst_ops;
2604         net->ipv6.ip6_null_entry->u.dst.path =
2605                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2606         net->ipv6.ip6_null_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2607
2608 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2609         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2610                                                sizeof(*net->ipv6.ip6_prohibit_entry),
2611                                                GFP_KERNEL);
2612         if (!net->ipv6.ip6_prohibit_entry) {
2613                 kfree(net->ipv6.ip6_null_entry);
2614                 goto out;
2615         }
2616         net->ipv6.ip6_prohibit_entry->u.dst.path =
2617                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2618         net->ipv6.ip6_prohibit_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2619
2620         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2621                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
2622                                                GFP_KERNEL);
2623         if (!net->ipv6.ip6_blk_hole_entry) {
2624                 kfree(net->ipv6.ip6_null_entry);
2625                 kfree(net->ipv6.ip6_prohibit_entry);
2626                 goto out;
2627         }
2628         net->ipv6.ip6_blk_hole_entry->u.dst.path =
2629                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2630         net->ipv6.ip6_blk_hole_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2631 #endif
2632
2633 #ifdef CONFIG_PROC_FS
2634         proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2635         proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2636 #endif
2637         net->ipv6.ip6_rt_gc_expire = 30*HZ;
2638
2639         ret = 0;
2640 out:
2641         return ret;
2642
2643 out_ip6_dst_ops:
2644         release_net(net->ipv6.ip6_dst_ops->dst_net);
2645         kfree(net->ipv6.ip6_dst_ops);
2646         goto out;
2647 }
2648
2649 static void ip6_route_net_exit(struct net *net)
2650 {
2651 #ifdef CONFIG_PROC_FS
2652         proc_net_remove(net, "ipv6_route");
2653         proc_net_remove(net, "rt6_stats");
2654 #endif
2655         kfree(net->ipv6.ip6_null_entry);
2656 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2657         kfree(net->ipv6.ip6_prohibit_entry);
2658         kfree(net->ipv6.ip6_blk_hole_entry);
2659 #endif
2660         release_net(net->ipv6.ip6_dst_ops->dst_net);
2661         kfree(net->ipv6.ip6_dst_ops);
2662 }
2663
2664 static struct pernet_operations ip6_route_net_ops = {
2665         .init = ip6_route_net_init,
2666         .exit = ip6_route_net_exit,
2667 };
2668
2669 static struct notifier_block ip6_route_dev_notifier = {
2670         .notifier_call = ip6_route_dev_notify,
2671         .priority = 0,
2672 };
2673
2674 int __init ip6_route_init(void)
2675 {
2676         int ret;
2677
2678         ret = -ENOMEM;
2679         ip6_dst_ops_template.kmem_cachep =
2680                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2681                                   SLAB_HWCACHE_ALIGN, NULL);
2682         if (!ip6_dst_ops_template.kmem_cachep)
2683                 goto out;;
2684
2685         ret = register_pernet_subsys(&ip6_route_net_ops);
2686         if (ret)
2687                 goto out_kmem_cache;
2688
2689         /* Registering of the loopback is done before this portion of code,
2690          * the loopback reference in rt6_info will not be taken, do it
2691          * manually for init_net */
2692         init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2693         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2694   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2695         init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2696         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2697         init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2698         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2699   #endif
2700         ret = fib6_init();
2701         if (ret)
2702                 goto out_register_subsys;
2703
2704         ret = xfrm6_init();
2705         if (ret)
2706                 goto out_fib6_init;
2707
2708         ret = fib6_rules_init();
2709         if (ret)
2710                 goto xfrm6_init;
2711
2712         ret = -ENOBUFS;
2713         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2714             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2715             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2716                 goto fib6_rules_init;
2717
2718         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2719         if (ret)
2720                 goto fib6_rules_init;
2721
2722 out:
2723         return ret;
2724
2725 fib6_rules_init:
2726         fib6_rules_cleanup();
2727 xfrm6_init:
2728         xfrm6_fini();
2729 out_fib6_init:
2730         fib6_gc_cleanup();
2731 out_register_subsys:
2732         unregister_pernet_subsys(&ip6_route_net_ops);
2733 out_kmem_cache:
2734         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2735         goto out;
2736 }
2737
2738 void ip6_route_cleanup(void)
2739 {
2740         unregister_netdevice_notifier(&ip6_route_dev_notifier);
2741         fib6_rules_cleanup();
2742         xfrm6_fini();
2743         fib6_gc_cleanup();
2744         unregister_pernet_subsys(&ip6_route_net_ops);
2745         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2746 }