[NET] NETNS: Omit net_device->nd_net without CONFIG_NET_NS.
[linux-3.10.git] / net / ipv6 / route.c
1 /*
2  *      Linux INET6 implementation
3  *      FIB front-end.
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
9  *
10  *      This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  */
15
16 /*      Changes:
17  *
18  *      YOSHIFUJI Hideaki @USAGI
19  *              reworked default router selection.
20  *              - respect outgoing interface
21  *              - select from (probably) reachable routers (i.e.
22  *              routers in REACHABLE, STALE, DELAY or PROBE states).
23  *              - always select the same router if it is (probably)
24  *              reachable.  otherwise, round-robin the list.
25  *      Ville Nuorvala
26  *              Fixed routing subtrees.
27  */
28
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/times.h>
33 #include <linux/socket.h>
34 #include <linux/sockios.h>
35 #include <linux/net.h>
36 #include <linux/route.h>
37 #include <linux/netdevice.h>
38 #include <linux/in6.h>
39 #include <linux/init.h>
40 #include <linux/if_arp.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <linux/nsproxy.h>
44 #include <net/net_namespace.h>
45 #include <net/snmp.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
51 #include <net/tcp.h>
52 #include <linux/rtnetlink.h>
53 #include <net/dst.h>
54 #include <net/xfrm.h>
55 #include <net/netevent.h>
56 #include <net/netlink.h>
57
58 #include <asm/uaccess.h>
59
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63
64 /* Set to 3 to get tracing. */
65 #define RT6_DEBUG 2
66
67 #if RT6_DEBUG >= 3
68 #define RDBG(x) printk x
69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
70 #else
71 #define RDBG(x)
72 #define RT6_TRACE(x...) do { ; } while (0)
73 #endif
74
75 #define CLONE_OFFLINK_ROUTE 0
76
77 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
78 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void             ip6_dst_destroy(struct dst_entry *);
81 static void             ip6_dst_ifdown(struct dst_entry *,
82                                        struct net_device *dev, int how);
83 static int               ip6_dst_gc(struct dst_ops *ops);
84
85 static int              ip6_pkt_discard(struct sk_buff *skb);
86 static int              ip6_pkt_discard_out(struct sk_buff *skb);
87 static void             ip6_link_failure(struct sk_buff *skb);
88 static void             ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
89
90 #ifdef CONFIG_IPV6_ROUTE_INFO
91 static struct rt6_info *rt6_add_route_info(struct net *net,
92                                            struct in6_addr *prefix, int prefixlen,
93                                            struct in6_addr *gwaddr, int ifindex,
94                                            unsigned pref);
95 static struct rt6_info *rt6_get_route_info(struct net *net,
96                                            struct in6_addr *prefix, int prefixlen,
97                                            struct in6_addr *gwaddr, int ifindex);
98 #endif
99
100 static struct dst_ops ip6_dst_ops_template = {
101         .family                 =       AF_INET6,
102         .protocol               =       __constant_htons(ETH_P_IPV6),
103         .gc                     =       ip6_dst_gc,
104         .gc_thresh              =       1024,
105         .check                  =       ip6_dst_check,
106         .destroy                =       ip6_dst_destroy,
107         .ifdown                 =       ip6_dst_ifdown,
108         .negative_advice        =       ip6_negative_advice,
109         .link_failure           =       ip6_link_failure,
110         .update_pmtu            =       ip6_rt_update_pmtu,
111         .local_out              =       ip6_local_out,
112         .entry_size             =       sizeof(struct rt6_info),
113         .entries                =       ATOMIC_INIT(0),
114 };
115
116 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
117 {
118 }
119
120 static struct dst_ops ip6_dst_blackhole_ops = {
121         .family                 =       AF_INET6,
122         .protocol               =       __constant_htons(ETH_P_IPV6),
123         .destroy                =       ip6_dst_destroy,
124         .check                  =       ip6_dst_check,
125         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
126         .entry_size             =       sizeof(struct rt6_info),
127         .entries                =       ATOMIC_INIT(0),
128 };
129
130 static struct rt6_info ip6_null_entry_template = {
131         .u = {
132                 .dst = {
133                         .__refcnt       = ATOMIC_INIT(1),
134                         .__use          = 1,
135                         .obsolete       = -1,
136                         .error          = -ENETUNREACH,
137                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
138                         .input          = ip6_pkt_discard,
139                         .output         = ip6_pkt_discard_out,
140                 }
141         },
142         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
143         .rt6i_metric    = ~(u32) 0,
144         .rt6i_ref       = ATOMIC_INIT(1),
145 };
146
147 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
148
149 static int ip6_pkt_prohibit(struct sk_buff *skb);
150 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
151
152 struct rt6_info ip6_prohibit_entry_template = {
153         .u = {
154                 .dst = {
155                         .__refcnt       = ATOMIC_INIT(1),
156                         .__use          = 1,
157                         .obsolete       = -1,
158                         .error          = -EACCES,
159                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
160                         .input          = ip6_pkt_prohibit,
161                         .output         = ip6_pkt_prohibit_out,
162                 }
163         },
164         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
165         .rt6i_metric    = ~(u32) 0,
166         .rt6i_ref       = ATOMIC_INIT(1),
167 };
168
169 static struct rt6_info ip6_blk_hole_entry_template = {
170         .u = {
171                 .dst = {
172                         .__refcnt       = ATOMIC_INIT(1),
173                         .__use          = 1,
174                         .obsolete       = -1,
175                         .error          = -EINVAL,
176                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
177                         .input          = dst_discard,
178                         .output         = dst_discard,
179                 }
180         },
181         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
182         .rt6i_metric    = ~(u32) 0,
183         .rt6i_ref       = ATOMIC_INIT(1),
184 };
185
186 #endif
187
188 /* allocate dst with ip6_dst_ops */
189 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
190 {
191         return (struct rt6_info *)dst_alloc(ops);
192 }
193
194 static void ip6_dst_destroy(struct dst_entry *dst)
195 {
196         struct rt6_info *rt = (struct rt6_info *)dst;
197         struct inet6_dev *idev = rt->rt6i_idev;
198
199         if (idev != NULL) {
200                 rt->rt6i_idev = NULL;
201                 in6_dev_put(idev);
202         }
203 }
204
205 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
206                            int how)
207 {
208         struct rt6_info *rt = (struct rt6_info *)dst;
209         struct inet6_dev *idev = rt->rt6i_idev;
210         struct net_device *loopback_dev =
211                 dev_net(dev)->loopback_dev;
212
213         if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
214                 struct inet6_dev *loopback_idev =
215                         in6_dev_get(loopback_dev);
216                 if (loopback_idev != NULL) {
217                         rt->rt6i_idev = loopback_idev;
218                         in6_dev_put(idev);
219                 }
220         }
221 }
222
223 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
224 {
225         return (rt->rt6i_flags & RTF_EXPIRES &&
226                 time_after(jiffies, rt->rt6i_expires));
227 }
228
229 static inline int rt6_need_strict(struct in6_addr *daddr)
230 {
231         return (ipv6_addr_type(daddr) &
232                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
233 }
234
235 /*
236  *      Route lookup. Any table->tb6_lock is implied.
237  */
238
239 static inline struct rt6_info *rt6_device_match(struct net *net,
240                                                     struct rt6_info *rt,
241                                                     int oif,
242                                                     int strict)
243 {
244         struct rt6_info *local = NULL;
245         struct rt6_info *sprt;
246
247         if (oif) {
248                 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
249                         struct net_device *dev = sprt->rt6i_dev;
250                         if (dev->ifindex == oif)
251                                 return sprt;
252                         if (dev->flags & IFF_LOOPBACK) {
253                                 if (sprt->rt6i_idev == NULL ||
254                                     sprt->rt6i_idev->dev->ifindex != oif) {
255                                         if (strict && oif)
256                                                 continue;
257                                         if (local && (!oif ||
258                                                       local->rt6i_idev->dev->ifindex == oif))
259                                                 continue;
260                                 }
261                                 local = sprt;
262                         }
263                 }
264
265                 if (local)
266                         return local;
267
268                 if (strict)
269                         return net->ipv6.ip6_null_entry;
270         }
271         return rt;
272 }
273
274 #ifdef CONFIG_IPV6_ROUTER_PREF
275 static void rt6_probe(struct rt6_info *rt)
276 {
277         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
278         /*
279          * Okay, this does not seem to be appropriate
280          * for now, however, we need to check if it
281          * is really so; aka Router Reachability Probing.
282          *
283          * Router Reachability Probe MUST be rate-limited
284          * to no more than one per minute.
285          */
286         if (!neigh || (neigh->nud_state & NUD_VALID))
287                 return;
288         read_lock_bh(&neigh->lock);
289         if (!(neigh->nud_state & NUD_VALID) &&
290             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
291                 struct in6_addr mcaddr;
292                 struct in6_addr *target;
293
294                 neigh->updated = jiffies;
295                 read_unlock_bh(&neigh->lock);
296
297                 target = (struct in6_addr *)&neigh->primary_key;
298                 addrconf_addr_solict_mult(target, &mcaddr);
299                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
300         } else
301                 read_unlock_bh(&neigh->lock);
302 }
303 #else
304 static inline void rt6_probe(struct rt6_info *rt)
305 {
306         return;
307 }
308 #endif
309
310 /*
311  * Default Router Selection (RFC 2461 6.3.6)
312  */
313 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
314 {
315         struct net_device *dev = rt->rt6i_dev;
316         if (!oif || dev->ifindex == oif)
317                 return 2;
318         if ((dev->flags & IFF_LOOPBACK) &&
319             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
320                 return 1;
321         return 0;
322 }
323
324 static inline int rt6_check_neigh(struct rt6_info *rt)
325 {
326         struct neighbour *neigh = rt->rt6i_nexthop;
327         int m;
328         if (rt->rt6i_flags & RTF_NONEXTHOP ||
329             !(rt->rt6i_flags & RTF_GATEWAY))
330                 m = 1;
331         else if (neigh) {
332                 read_lock_bh(&neigh->lock);
333                 if (neigh->nud_state & NUD_VALID)
334                         m = 2;
335 #ifdef CONFIG_IPV6_ROUTER_PREF
336                 else if (neigh->nud_state & NUD_FAILED)
337                         m = 0;
338 #endif
339                 else
340                         m = 1;
341                 read_unlock_bh(&neigh->lock);
342         } else
343                 m = 0;
344         return m;
345 }
346
347 static int rt6_score_route(struct rt6_info *rt, int oif,
348                            int strict)
349 {
350         int m, n;
351
352         m = rt6_check_dev(rt, oif);
353         if (!m && (strict & RT6_LOOKUP_F_IFACE))
354                 return -1;
355 #ifdef CONFIG_IPV6_ROUTER_PREF
356         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
357 #endif
358         n = rt6_check_neigh(rt);
359         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
360                 return -1;
361         return m;
362 }
363
364 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
365                                    int *mpri, struct rt6_info *match)
366 {
367         int m;
368
369         if (rt6_check_expired(rt))
370                 goto out;
371
372         m = rt6_score_route(rt, oif, strict);
373         if (m < 0)
374                 goto out;
375
376         if (m > *mpri) {
377                 if (strict & RT6_LOOKUP_F_REACHABLE)
378                         rt6_probe(match);
379                 *mpri = m;
380                 match = rt;
381         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
382                 rt6_probe(rt);
383         }
384
385 out:
386         return match;
387 }
388
389 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
390                                      struct rt6_info *rr_head,
391                                      u32 metric, int oif, int strict)
392 {
393         struct rt6_info *rt, *match;
394         int mpri = -1;
395
396         match = NULL;
397         for (rt = rr_head; rt && rt->rt6i_metric == metric;
398              rt = rt->u.dst.rt6_next)
399                 match = find_match(rt, oif, strict, &mpri, match);
400         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
401              rt = rt->u.dst.rt6_next)
402                 match = find_match(rt, oif, strict, &mpri, match);
403
404         return match;
405 }
406
407 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
408 {
409         struct rt6_info *match, *rt0;
410         struct net *net;
411
412         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
413                   __func__, fn->leaf, oif);
414
415         rt0 = fn->rr_ptr;
416         if (!rt0)
417                 fn->rr_ptr = rt0 = fn->leaf;
418
419         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
420
421         if (!match &&
422             (strict & RT6_LOOKUP_F_REACHABLE)) {
423                 struct rt6_info *next = rt0->u.dst.rt6_next;
424
425                 /* no entries matched; do round-robin */
426                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
427                         next = fn->leaf;
428
429                 if (next != rt0)
430                         fn->rr_ptr = next;
431         }
432
433         RT6_TRACE("%s() => %p\n",
434                   __func__, match);
435
436         net = dev_net(rt0->rt6i_dev);
437         return (match ? match : net->ipv6.ip6_null_entry);
438 }
439
440 #ifdef CONFIG_IPV6_ROUTE_INFO
441 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
442                   struct in6_addr *gwaddr)
443 {
444         struct net *net = dev_net(dev);
445         struct route_info *rinfo = (struct route_info *) opt;
446         struct in6_addr prefix_buf, *prefix;
447         unsigned int pref;
448         u32 lifetime;
449         struct rt6_info *rt;
450
451         if (len < sizeof(struct route_info)) {
452                 return -EINVAL;
453         }
454
455         /* Sanity check for prefix_len and length */
456         if (rinfo->length > 3) {
457                 return -EINVAL;
458         } else if (rinfo->prefix_len > 128) {
459                 return -EINVAL;
460         } else if (rinfo->prefix_len > 64) {
461                 if (rinfo->length < 2) {
462                         return -EINVAL;
463                 }
464         } else if (rinfo->prefix_len > 0) {
465                 if (rinfo->length < 1) {
466                         return -EINVAL;
467                 }
468         }
469
470         pref = rinfo->route_pref;
471         if (pref == ICMPV6_ROUTER_PREF_INVALID)
472                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
473
474         lifetime = ntohl(rinfo->lifetime);
475         if (lifetime == 0xffffffff) {
476                 /* infinity */
477         } else if (lifetime > 0x7fffffff/HZ) {
478                 /* Avoid arithmetic overflow */
479                 lifetime = 0x7fffffff/HZ - 1;
480         }
481
482         if (rinfo->length == 3)
483                 prefix = (struct in6_addr *)rinfo->prefix;
484         else {
485                 /* this function is safe */
486                 ipv6_addr_prefix(&prefix_buf,
487                                  (struct in6_addr *)rinfo->prefix,
488                                  rinfo->prefix_len);
489                 prefix = &prefix_buf;
490         }
491
492         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
493                                 dev->ifindex);
494
495         if (rt && !lifetime) {
496                 ip6_del_rt(rt);
497                 rt = NULL;
498         }
499
500         if (!rt && lifetime)
501                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
502                                         pref);
503         else if (rt)
504                 rt->rt6i_flags = RTF_ROUTEINFO |
505                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
506
507         if (rt) {
508                 if (lifetime == 0xffffffff) {
509                         rt->rt6i_flags &= ~RTF_EXPIRES;
510                 } else {
511                         rt->rt6i_expires = jiffies + HZ * lifetime;
512                         rt->rt6i_flags |= RTF_EXPIRES;
513                 }
514                 dst_release(&rt->u.dst);
515         }
516         return 0;
517 }
518 #endif
519
520 #define BACKTRACK(__net, saddr)                 \
521 do { \
522         if (rt == __net->ipv6.ip6_null_entry) { \
523                 struct fib6_node *pn; \
524                 while (1) { \
525                         if (fn->fn_flags & RTN_TL_ROOT) \
526                                 goto out; \
527                         pn = fn->parent; \
528                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
529                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
530                         else \
531                                 fn = pn; \
532                         if (fn->fn_flags & RTN_RTINFO) \
533                                 goto restart; \
534                 } \
535         } \
536 } while(0)
537
538 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
539                                              struct fib6_table *table,
540                                              struct flowi *fl, int flags)
541 {
542         struct fib6_node *fn;
543         struct rt6_info *rt;
544
545         read_lock_bh(&table->tb6_lock);
546         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
547 restart:
548         rt = fn->leaf;
549         rt = rt6_device_match(net, rt, fl->oif, flags);
550         BACKTRACK(net, &fl->fl6_src);
551 out:
552         dst_use(&rt->u.dst, jiffies);
553         read_unlock_bh(&table->tb6_lock);
554         return rt;
555
556 }
557
558 struct rt6_info *rt6_lookup(struct net *net, struct in6_addr *daddr,
559                             struct in6_addr *saddr, int oif, int strict)
560 {
561         struct flowi fl = {
562                 .oif = oif,
563                 .nl_u = {
564                         .ip6_u = {
565                                 .daddr = *daddr,
566                         },
567                 },
568         };
569         struct dst_entry *dst;
570         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
571
572         if (saddr) {
573                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
574                 flags |= RT6_LOOKUP_F_HAS_SADDR;
575         }
576
577         dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
578         if (dst->error == 0)
579                 return (struct rt6_info *) dst;
580
581         dst_release(dst);
582
583         return NULL;
584 }
585
586 EXPORT_SYMBOL(rt6_lookup);
587
588 /* ip6_ins_rt is called with FREE table->tb6_lock.
589    It takes new route entry, the addition fails by any reason the
590    route is freed. In any case, if caller does not hold it, it may
591    be destroyed.
592  */
593
594 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
595 {
596         int err;
597         struct fib6_table *table;
598
599         table = rt->rt6i_table;
600         write_lock_bh(&table->tb6_lock);
601         err = fib6_add(&table->tb6_root, rt, info);
602         write_unlock_bh(&table->tb6_lock);
603
604         return err;
605 }
606
607 int ip6_ins_rt(struct rt6_info *rt)
608 {
609         struct nl_info info = {
610                 .nl_net = dev_net(rt->rt6i_dev),
611         };
612         return __ip6_ins_rt(rt, &info);
613 }
614
615 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
616                                       struct in6_addr *saddr)
617 {
618         struct rt6_info *rt;
619
620         /*
621          *      Clone the route.
622          */
623
624         rt = ip6_rt_copy(ort);
625
626         if (rt) {
627                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
628                         if (rt->rt6i_dst.plen != 128 &&
629                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
630                                 rt->rt6i_flags |= RTF_ANYCAST;
631                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
632                 }
633
634                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
635                 rt->rt6i_dst.plen = 128;
636                 rt->rt6i_flags |= RTF_CACHE;
637                 rt->u.dst.flags |= DST_HOST;
638
639 #ifdef CONFIG_IPV6_SUBTREES
640                 if (rt->rt6i_src.plen && saddr) {
641                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
642                         rt->rt6i_src.plen = 128;
643                 }
644 #endif
645
646                 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
647
648         }
649
650         return rt;
651 }
652
653 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
654 {
655         struct rt6_info *rt = ip6_rt_copy(ort);
656         if (rt) {
657                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
658                 rt->rt6i_dst.plen = 128;
659                 rt->rt6i_flags |= RTF_CACHE;
660                 rt->u.dst.flags |= DST_HOST;
661                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
662         }
663         return rt;
664 }
665
666 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
667                                       struct flowi *fl, int flags)
668 {
669         struct fib6_node *fn;
670         struct rt6_info *rt, *nrt;
671         int strict = 0;
672         int attempts = 3;
673         int err;
674         int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
675
676         strict |= flags & RT6_LOOKUP_F_IFACE;
677
678 relookup:
679         read_lock_bh(&table->tb6_lock);
680
681 restart_2:
682         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
683
684 restart:
685         rt = rt6_select(fn, oif, strict | reachable);
686
687         BACKTRACK(net, &fl->fl6_src);
688         if (rt == net->ipv6.ip6_null_entry ||
689             rt->rt6i_flags & RTF_CACHE)
690                 goto out;
691
692         dst_hold(&rt->u.dst);
693         read_unlock_bh(&table->tb6_lock);
694
695         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
696                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
697         else {
698 #if CLONE_OFFLINK_ROUTE
699                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
700 #else
701                 goto out2;
702 #endif
703         }
704
705         dst_release(&rt->u.dst);
706         rt = nrt ? : net->ipv6.ip6_null_entry;
707
708         dst_hold(&rt->u.dst);
709         if (nrt) {
710                 err = ip6_ins_rt(nrt);
711                 if (!err)
712                         goto out2;
713         }
714
715         if (--attempts <= 0)
716                 goto out2;
717
718         /*
719          * Race condition! In the gap, when table->tb6_lock was
720          * released someone could insert this route.  Relookup.
721          */
722         dst_release(&rt->u.dst);
723         goto relookup;
724
725 out:
726         if (reachable) {
727                 reachable = 0;
728                 goto restart_2;
729         }
730         dst_hold(&rt->u.dst);
731         read_unlock_bh(&table->tb6_lock);
732 out2:
733         rt->u.dst.lastuse = jiffies;
734         rt->u.dst.__use++;
735
736         return rt;
737 }
738
739 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
740                                             struct flowi *fl, int flags)
741 {
742         return ip6_pol_route(net, table, fl->iif, fl, flags);
743 }
744
745 void ip6_route_input(struct sk_buff *skb)
746 {
747         struct ipv6hdr *iph = ipv6_hdr(skb);
748         struct net *net = dev_net(skb->dev);
749         int flags = RT6_LOOKUP_F_HAS_SADDR;
750         struct flowi fl = {
751                 .iif = skb->dev->ifindex,
752                 .nl_u = {
753                         .ip6_u = {
754                                 .daddr = iph->daddr,
755                                 .saddr = iph->saddr,
756                                 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
757                         },
758                 },
759                 .mark = skb->mark,
760                 .proto = iph->nexthdr,
761         };
762
763         if (rt6_need_strict(&iph->daddr))
764                 flags |= RT6_LOOKUP_F_IFACE;
765
766         skb->dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input);
767 }
768
769 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
770                                              struct flowi *fl, int flags)
771 {
772         return ip6_pol_route(net, table, fl->oif, fl, flags);
773 }
774
775 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
776                                     struct flowi *fl)
777 {
778         int flags = 0;
779
780         if (rt6_need_strict(&fl->fl6_dst))
781                 flags |= RT6_LOOKUP_F_IFACE;
782
783         if (!ipv6_addr_any(&fl->fl6_src))
784                 flags |= RT6_LOOKUP_F_HAS_SADDR;
785         else if (sk) {
786                 unsigned int prefs = inet6_sk(sk)->srcprefs;
787                 if (prefs & IPV6_PREFER_SRC_TMP)
788                         flags |= RT6_LOOKUP_F_SRCPREF_TMP;
789                 if (prefs & IPV6_PREFER_SRC_PUBLIC)
790                         flags |= RT6_LOOKUP_F_SRCPREF_PUBLIC;
791                 if (prefs & IPV6_PREFER_SRC_COA)
792                         flags |= RT6_LOOKUP_F_SRCPREF_COA;
793         }
794
795         return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
796 }
797
798 EXPORT_SYMBOL(ip6_route_output);
799
800 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
801 {
802         struct rt6_info *ort = (struct rt6_info *) *dstp;
803         struct rt6_info *rt = (struct rt6_info *)
804                 dst_alloc(&ip6_dst_blackhole_ops);
805         struct dst_entry *new = NULL;
806
807         if (rt) {
808                 new = &rt->u.dst;
809
810                 atomic_set(&new->__refcnt, 1);
811                 new->__use = 1;
812                 new->input = dst_discard;
813                 new->output = dst_discard;
814
815                 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
816                 new->dev = ort->u.dst.dev;
817                 if (new->dev)
818                         dev_hold(new->dev);
819                 rt->rt6i_idev = ort->rt6i_idev;
820                 if (rt->rt6i_idev)
821                         in6_dev_hold(rt->rt6i_idev);
822                 rt->rt6i_expires = 0;
823
824                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
825                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
826                 rt->rt6i_metric = 0;
827
828                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
829 #ifdef CONFIG_IPV6_SUBTREES
830                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
831 #endif
832
833                 dst_free(new);
834         }
835
836         dst_release(*dstp);
837         *dstp = new;
838         return (new ? 0 : -ENOMEM);
839 }
840 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
841
842 /*
843  *      Destination cache support functions
844  */
845
846 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
847 {
848         struct rt6_info *rt;
849
850         rt = (struct rt6_info *) dst;
851
852         if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
853                 return dst;
854
855         return NULL;
856 }
857
858 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
859 {
860         struct rt6_info *rt = (struct rt6_info *) dst;
861
862         if (rt) {
863                 if (rt->rt6i_flags & RTF_CACHE)
864                         ip6_del_rt(rt);
865                 else
866                         dst_release(dst);
867         }
868         return NULL;
869 }
870
871 static void ip6_link_failure(struct sk_buff *skb)
872 {
873         struct rt6_info *rt;
874
875         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
876
877         rt = (struct rt6_info *) skb->dst;
878         if (rt) {
879                 if (rt->rt6i_flags&RTF_CACHE) {
880                         dst_set_expires(&rt->u.dst, 0);
881                         rt->rt6i_flags |= RTF_EXPIRES;
882                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
883                         rt->rt6i_node->fn_sernum = -1;
884         }
885 }
886
887 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
888 {
889         struct rt6_info *rt6 = (struct rt6_info*)dst;
890
891         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
892                 rt6->rt6i_flags |= RTF_MODIFIED;
893                 if (mtu < IPV6_MIN_MTU) {
894                         mtu = IPV6_MIN_MTU;
895                         dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
896                 }
897                 dst->metrics[RTAX_MTU-1] = mtu;
898                 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
899         }
900 }
901
902 static int ipv6_get_mtu(struct net_device *dev);
903
904 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
905 {
906         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
907
908         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
909                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
910
911         /*
912          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
913          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
914          * IPV6_MAXPLEN is also valid and means: "any MSS,
915          * rely only on pmtu discovery"
916          */
917         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
918                 mtu = IPV6_MAXPLEN;
919         return mtu;
920 }
921
922 static struct dst_entry *icmp6_dst_gc_list;
923 static DEFINE_SPINLOCK(icmp6_dst_lock);
924
925 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
926                                   struct neighbour *neigh,
927                                   struct in6_addr *addr)
928 {
929         struct rt6_info *rt;
930         struct inet6_dev *idev = in6_dev_get(dev);
931         struct net *net = dev_net(dev);
932
933         if (unlikely(idev == NULL))
934                 return NULL;
935
936         rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
937         if (unlikely(rt == NULL)) {
938                 in6_dev_put(idev);
939                 goto out;
940         }
941
942         dev_hold(dev);
943         if (neigh)
944                 neigh_hold(neigh);
945         else
946                 neigh = ndisc_get_neigh(dev, addr);
947
948         rt->rt6i_dev      = dev;
949         rt->rt6i_idev     = idev;
950         rt->rt6i_nexthop  = neigh;
951         atomic_set(&rt->u.dst.__refcnt, 1);
952         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
953         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
954         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
955         rt->u.dst.output  = ip6_output;
956
957 #if 0   /* there's no chance to use these for ndisc */
958         rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
959                                 ? DST_HOST
960                                 : 0;
961         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
962         rt->rt6i_dst.plen = 128;
963 #endif
964
965         spin_lock_bh(&icmp6_dst_lock);
966         rt->u.dst.next = icmp6_dst_gc_list;
967         icmp6_dst_gc_list = &rt->u.dst;
968         spin_unlock_bh(&icmp6_dst_lock);
969
970         fib6_force_start_gc(net);
971
972 out:
973         return &rt->u.dst;
974 }
975
976 int icmp6_dst_gc(int *more)
977 {
978         struct dst_entry *dst, *next, **pprev;
979         int freed;
980
981         next = NULL;
982         freed = 0;
983
984         spin_lock_bh(&icmp6_dst_lock);
985         pprev = &icmp6_dst_gc_list;
986
987         while ((dst = *pprev) != NULL) {
988                 if (!atomic_read(&dst->__refcnt)) {
989                         *pprev = dst->next;
990                         dst_free(dst);
991                         freed++;
992                 } else {
993                         pprev = &dst->next;
994                         (*more)++;
995                 }
996         }
997
998         spin_unlock_bh(&icmp6_dst_lock);
999
1000         return freed;
1001 }
1002
1003 static int ip6_dst_gc(struct dst_ops *ops)
1004 {
1005         unsigned long now = jiffies;
1006         struct net *net = ops->dst_net;
1007         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1008         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1009         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1010         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1011         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1012
1013         if (time_after(rt_last_gc + rt_min_interval, now) &&
1014             atomic_read(&ops->entries) <= rt_max_size)
1015                 goto out;
1016
1017         net->ipv6.ip6_rt_gc_expire++;
1018         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1019         net->ipv6.ip6_rt_last_gc = now;
1020         if (atomic_read(&ops->entries) < ops->gc_thresh)
1021                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1022 out:
1023         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1024         return (atomic_read(&ops->entries) > rt_max_size);
1025 }
1026
1027 /* Clean host part of a prefix. Not necessary in radix tree,
1028    but results in cleaner routing tables.
1029
1030    Remove it only when all the things will work!
1031  */
1032
1033 static int ipv6_get_mtu(struct net_device *dev)
1034 {
1035         int mtu = IPV6_MIN_MTU;
1036         struct inet6_dev *idev;
1037
1038         idev = in6_dev_get(dev);
1039         if (idev) {
1040                 mtu = idev->cnf.mtu6;
1041                 in6_dev_put(idev);
1042         }
1043         return mtu;
1044 }
1045
1046 int ip6_dst_hoplimit(struct dst_entry *dst)
1047 {
1048         int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1049         if (hoplimit < 0) {
1050                 struct net_device *dev = dst->dev;
1051                 struct inet6_dev *idev = in6_dev_get(dev);
1052                 if (idev) {
1053                         hoplimit = idev->cnf.hop_limit;
1054                         in6_dev_put(idev);
1055                 } else
1056                         hoplimit = ipv6_devconf.hop_limit;
1057         }
1058         return hoplimit;
1059 }
1060
1061 /*
1062  *
1063  */
1064
1065 int ip6_route_add(struct fib6_config *cfg)
1066 {
1067         int err;
1068         struct net *net = cfg->fc_nlinfo.nl_net;
1069         struct rt6_info *rt = NULL;
1070         struct net_device *dev = NULL;
1071         struct inet6_dev *idev = NULL;
1072         struct fib6_table *table;
1073         int addr_type;
1074
1075         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1076                 return -EINVAL;
1077 #ifndef CONFIG_IPV6_SUBTREES
1078         if (cfg->fc_src_len)
1079                 return -EINVAL;
1080 #endif
1081         if (cfg->fc_ifindex) {
1082                 err = -ENODEV;
1083                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1084                 if (!dev)
1085                         goto out;
1086                 idev = in6_dev_get(dev);
1087                 if (!idev)
1088                         goto out;
1089         }
1090
1091         if (cfg->fc_metric == 0)
1092                 cfg->fc_metric = IP6_RT_PRIO_USER;
1093
1094         table = fib6_new_table(net, cfg->fc_table);
1095         if (table == NULL) {
1096                 err = -ENOBUFS;
1097                 goto out;
1098         }
1099
1100         rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1101
1102         if (rt == NULL) {
1103                 err = -ENOMEM;
1104                 goto out;
1105         }
1106
1107         rt->u.dst.obsolete = -1;
1108         rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1109
1110         if (cfg->fc_protocol == RTPROT_UNSPEC)
1111                 cfg->fc_protocol = RTPROT_BOOT;
1112         rt->rt6i_protocol = cfg->fc_protocol;
1113
1114         addr_type = ipv6_addr_type(&cfg->fc_dst);
1115
1116         if (addr_type & IPV6_ADDR_MULTICAST)
1117                 rt->u.dst.input = ip6_mc_input;
1118         else
1119                 rt->u.dst.input = ip6_forward;
1120
1121         rt->u.dst.output = ip6_output;
1122
1123         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1124         rt->rt6i_dst.plen = cfg->fc_dst_len;
1125         if (rt->rt6i_dst.plen == 128)
1126                rt->u.dst.flags = DST_HOST;
1127
1128 #ifdef CONFIG_IPV6_SUBTREES
1129         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1130         rt->rt6i_src.plen = cfg->fc_src_len;
1131 #endif
1132
1133         rt->rt6i_metric = cfg->fc_metric;
1134
1135         /* We cannot add true routes via loopback here,
1136            they would result in kernel looping; promote them to reject routes
1137          */
1138         if ((cfg->fc_flags & RTF_REJECT) ||
1139             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1140                 /* hold loopback dev/idev if we haven't done so. */
1141                 if (dev != net->loopback_dev) {
1142                         if (dev) {
1143                                 dev_put(dev);
1144                                 in6_dev_put(idev);
1145                         }
1146                         dev = net->loopback_dev;
1147                         dev_hold(dev);
1148                         idev = in6_dev_get(dev);
1149                         if (!idev) {
1150                                 err = -ENODEV;
1151                                 goto out;
1152                         }
1153                 }
1154                 rt->u.dst.output = ip6_pkt_discard_out;
1155                 rt->u.dst.input = ip6_pkt_discard;
1156                 rt->u.dst.error = -ENETUNREACH;
1157                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1158                 goto install_route;
1159         }
1160
1161         if (cfg->fc_flags & RTF_GATEWAY) {
1162                 struct in6_addr *gw_addr;
1163                 int gwa_type;
1164
1165                 gw_addr = &cfg->fc_gateway;
1166                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1167                 gwa_type = ipv6_addr_type(gw_addr);
1168
1169                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1170                         struct rt6_info *grt;
1171
1172                         /* IPv6 strictly inhibits using not link-local
1173                            addresses as nexthop address.
1174                            Otherwise, router will not able to send redirects.
1175                            It is very good, but in some (rare!) circumstances
1176                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1177                            some exceptions. --ANK
1178                          */
1179                         err = -EINVAL;
1180                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1181                                 goto out;
1182
1183                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1184
1185                         err = -EHOSTUNREACH;
1186                         if (grt == NULL)
1187                                 goto out;
1188                         if (dev) {
1189                                 if (dev != grt->rt6i_dev) {
1190                                         dst_release(&grt->u.dst);
1191                                         goto out;
1192                                 }
1193                         } else {
1194                                 dev = grt->rt6i_dev;
1195                                 idev = grt->rt6i_idev;
1196                                 dev_hold(dev);
1197                                 in6_dev_hold(grt->rt6i_idev);
1198                         }
1199                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1200                                 err = 0;
1201                         dst_release(&grt->u.dst);
1202
1203                         if (err)
1204                                 goto out;
1205                 }
1206                 err = -EINVAL;
1207                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1208                         goto out;
1209         }
1210
1211         err = -ENODEV;
1212         if (dev == NULL)
1213                 goto out;
1214
1215         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1216                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1217                 if (IS_ERR(rt->rt6i_nexthop)) {
1218                         err = PTR_ERR(rt->rt6i_nexthop);
1219                         rt->rt6i_nexthop = NULL;
1220                         goto out;
1221                 }
1222         }
1223
1224         rt->rt6i_flags = cfg->fc_flags;
1225
1226 install_route:
1227         if (cfg->fc_mx) {
1228                 struct nlattr *nla;
1229                 int remaining;
1230
1231                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1232                         int type = nla_type(nla);
1233
1234                         if (type) {
1235                                 if (type > RTAX_MAX) {
1236                                         err = -EINVAL;
1237                                         goto out;
1238                                 }
1239
1240                                 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1241                         }
1242                 }
1243         }
1244
1245         if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1246                 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1247         if (!rt->u.dst.metrics[RTAX_MTU-1])
1248                 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1249         if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1250                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1251         rt->u.dst.dev = dev;
1252         rt->rt6i_idev = idev;
1253         rt->rt6i_table = table;
1254
1255         cfg->fc_nlinfo.nl_net = dev_net(dev);
1256
1257         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1258
1259 out:
1260         if (dev)
1261                 dev_put(dev);
1262         if (idev)
1263                 in6_dev_put(idev);
1264         if (rt)
1265                 dst_free(&rt->u.dst);
1266         return err;
1267 }
1268
1269 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1270 {
1271         int err;
1272         struct fib6_table *table;
1273         struct net *net = dev_net(rt->rt6i_dev);
1274
1275         if (rt == net->ipv6.ip6_null_entry)
1276                 return -ENOENT;
1277
1278         table = rt->rt6i_table;
1279         write_lock_bh(&table->tb6_lock);
1280
1281         err = fib6_del(rt, info);
1282         dst_release(&rt->u.dst);
1283
1284         write_unlock_bh(&table->tb6_lock);
1285
1286         return err;
1287 }
1288
1289 int ip6_del_rt(struct rt6_info *rt)
1290 {
1291         struct nl_info info = {
1292                 .nl_net = dev_net(rt->rt6i_dev),
1293         };
1294         return __ip6_del_rt(rt, &info);
1295 }
1296
1297 static int ip6_route_del(struct fib6_config *cfg)
1298 {
1299         struct fib6_table *table;
1300         struct fib6_node *fn;
1301         struct rt6_info *rt;
1302         int err = -ESRCH;
1303
1304         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1305         if (table == NULL)
1306                 return err;
1307
1308         read_lock_bh(&table->tb6_lock);
1309
1310         fn = fib6_locate(&table->tb6_root,
1311                          &cfg->fc_dst, cfg->fc_dst_len,
1312                          &cfg->fc_src, cfg->fc_src_len);
1313
1314         if (fn) {
1315                 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1316                         if (cfg->fc_ifindex &&
1317                             (rt->rt6i_dev == NULL ||
1318                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1319                                 continue;
1320                         if (cfg->fc_flags & RTF_GATEWAY &&
1321                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1322                                 continue;
1323                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1324                                 continue;
1325                         dst_hold(&rt->u.dst);
1326                         read_unlock_bh(&table->tb6_lock);
1327
1328                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1329                 }
1330         }
1331         read_unlock_bh(&table->tb6_lock);
1332
1333         return err;
1334 }
1335
1336 /*
1337  *      Handle redirects
1338  */
1339 struct ip6rd_flowi {
1340         struct flowi fl;
1341         struct in6_addr gateway;
1342 };
1343
1344 static struct rt6_info *__ip6_route_redirect(struct net *net,
1345                                              struct fib6_table *table,
1346                                              struct flowi *fl,
1347                                              int flags)
1348 {
1349         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1350         struct rt6_info *rt;
1351         struct fib6_node *fn;
1352
1353         /*
1354          * Get the "current" route for this destination and
1355          * check if the redirect has come from approriate router.
1356          *
1357          * RFC 2461 specifies that redirects should only be
1358          * accepted if they come from the nexthop to the target.
1359          * Due to the way the routes are chosen, this notion
1360          * is a bit fuzzy and one might need to check all possible
1361          * routes.
1362          */
1363
1364         read_lock_bh(&table->tb6_lock);
1365         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1366 restart:
1367         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1368                 /*
1369                  * Current route is on-link; redirect is always invalid.
1370                  *
1371                  * Seems, previous statement is not true. It could
1372                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1373                  * But then router serving it might decide, that we should
1374                  * know truth 8)8) --ANK (980726).
1375                  */
1376                 if (rt6_check_expired(rt))
1377                         continue;
1378                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1379                         continue;
1380                 if (fl->oif != rt->rt6i_dev->ifindex)
1381                         continue;
1382                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1383                         continue;
1384                 break;
1385         }
1386
1387         if (!rt)
1388                 rt = net->ipv6.ip6_null_entry;
1389         BACKTRACK(net, &fl->fl6_src);
1390 out:
1391         dst_hold(&rt->u.dst);
1392
1393         read_unlock_bh(&table->tb6_lock);
1394
1395         return rt;
1396 };
1397
1398 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1399                                            struct in6_addr *src,
1400                                            struct in6_addr *gateway,
1401                                            struct net_device *dev)
1402 {
1403         int flags = RT6_LOOKUP_F_HAS_SADDR;
1404         struct net *net = dev_net(dev);
1405         struct ip6rd_flowi rdfl = {
1406                 .fl = {
1407                         .oif = dev->ifindex,
1408                         .nl_u = {
1409                                 .ip6_u = {
1410                                         .daddr = *dest,
1411                                         .saddr = *src,
1412                                 },
1413                         },
1414                 },
1415                 .gateway = *gateway,
1416         };
1417
1418         if (rt6_need_strict(dest))
1419                 flags |= RT6_LOOKUP_F_IFACE;
1420
1421         return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1422                                                    flags, __ip6_route_redirect);
1423 }
1424
1425 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1426                   struct in6_addr *saddr,
1427                   struct neighbour *neigh, u8 *lladdr, int on_link)
1428 {
1429         struct rt6_info *rt, *nrt = NULL;
1430         struct netevent_redirect netevent;
1431         struct net *net = dev_net(neigh->dev);
1432
1433         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1434
1435         if (rt == net->ipv6.ip6_null_entry) {
1436                 if (net_ratelimit())
1437                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1438                                "for redirect target\n");
1439                 goto out;
1440         }
1441
1442         /*
1443          *      We have finally decided to accept it.
1444          */
1445
1446         neigh_update(neigh, lladdr, NUD_STALE,
1447                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1448                      NEIGH_UPDATE_F_OVERRIDE|
1449                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1450                                      NEIGH_UPDATE_F_ISROUTER))
1451                      );
1452
1453         /*
1454          * Redirect received -> path was valid.
1455          * Look, redirects are sent only in response to data packets,
1456          * so that this nexthop apparently is reachable. --ANK
1457          */
1458         dst_confirm(&rt->u.dst);
1459
1460         /* Duplicate redirect: silently ignore. */
1461         if (neigh == rt->u.dst.neighbour)
1462                 goto out;
1463
1464         nrt = ip6_rt_copy(rt);
1465         if (nrt == NULL)
1466                 goto out;
1467
1468         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1469         if (on_link)
1470                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1471
1472         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1473         nrt->rt6i_dst.plen = 128;
1474         nrt->u.dst.flags |= DST_HOST;
1475
1476         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1477         nrt->rt6i_nexthop = neigh_clone(neigh);
1478         /* Reset pmtu, it may be better */
1479         nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1480         nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1481                                                         dst_mtu(&nrt->u.dst));
1482
1483         if (ip6_ins_rt(nrt))
1484                 goto out;
1485
1486         netevent.old = &rt->u.dst;
1487         netevent.new = &nrt->u.dst;
1488         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1489
1490         if (rt->rt6i_flags&RTF_CACHE) {
1491                 ip6_del_rt(rt);
1492                 return;
1493         }
1494
1495 out:
1496         dst_release(&rt->u.dst);
1497         return;
1498 }
1499
1500 /*
1501  *      Handle ICMP "packet too big" messages
1502  *      i.e. Path MTU discovery
1503  */
1504
1505 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1506                         struct net_device *dev, u32 pmtu)
1507 {
1508         struct rt6_info *rt, *nrt;
1509         struct net *net = dev_net(dev);
1510         int allfrag = 0;
1511
1512         rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1513         if (rt == NULL)
1514                 return;
1515
1516         if (pmtu >= dst_mtu(&rt->u.dst))
1517                 goto out;
1518
1519         if (pmtu < IPV6_MIN_MTU) {
1520                 /*
1521                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1522                  * MTU (1280) and a fragment header should always be included
1523                  * after a node receiving Too Big message reporting PMTU is
1524                  * less than the IPv6 Minimum Link MTU.
1525                  */
1526                 pmtu = IPV6_MIN_MTU;
1527                 allfrag = 1;
1528         }
1529
1530         /* New mtu received -> path was valid.
1531            They are sent only in response to data packets,
1532            so that this nexthop apparently is reachable. --ANK
1533          */
1534         dst_confirm(&rt->u.dst);
1535
1536         /* Host route. If it is static, it would be better
1537            not to override it, but add new one, so that
1538            when cache entry will expire old pmtu
1539            would return automatically.
1540          */
1541         if (rt->rt6i_flags & RTF_CACHE) {
1542                 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1543                 if (allfrag)
1544                         rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1545                 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1546                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1547                 goto out;
1548         }
1549
1550         /* Network route.
1551            Two cases are possible:
1552            1. It is connected route. Action: COW
1553            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1554          */
1555         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1556                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1557         else
1558                 nrt = rt6_alloc_clone(rt, daddr);
1559
1560         if (nrt) {
1561                 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1562                 if (allfrag)
1563                         nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1564
1565                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1566                  * happened within 5 mins, the recommended timer is 10 mins.
1567                  * Here this route expiration time is set to ip6_rt_mtu_expires
1568                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1569                  * and detecting PMTU increase will be automatically happened.
1570                  */
1571                 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1572                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1573
1574                 ip6_ins_rt(nrt);
1575         }
1576 out:
1577         dst_release(&rt->u.dst);
1578 }
1579
1580 /*
1581  *      Misc support functions
1582  */
1583
1584 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1585 {
1586         struct net *net = dev_net(ort->rt6i_dev);
1587         struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1588
1589         if (rt) {
1590                 rt->u.dst.input = ort->u.dst.input;
1591                 rt->u.dst.output = ort->u.dst.output;
1592
1593                 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1594                 rt->u.dst.error = ort->u.dst.error;
1595                 rt->u.dst.dev = ort->u.dst.dev;
1596                 if (rt->u.dst.dev)
1597                         dev_hold(rt->u.dst.dev);
1598                 rt->rt6i_idev = ort->rt6i_idev;
1599                 if (rt->rt6i_idev)
1600                         in6_dev_hold(rt->rt6i_idev);
1601                 rt->u.dst.lastuse = jiffies;
1602                 rt->rt6i_expires = 0;
1603
1604                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1605                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1606                 rt->rt6i_metric = 0;
1607
1608                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1609 #ifdef CONFIG_IPV6_SUBTREES
1610                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1611 #endif
1612                 rt->rt6i_table = ort->rt6i_table;
1613         }
1614         return rt;
1615 }
1616
1617 #ifdef CONFIG_IPV6_ROUTE_INFO
1618 static struct rt6_info *rt6_get_route_info(struct net *net,
1619                                            struct in6_addr *prefix, int prefixlen,
1620                                            struct in6_addr *gwaddr, int ifindex)
1621 {
1622         struct fib6_node *fn;
1623         struct rt6_info *rt = NULL;
1624         struct fib6_table *table;
1625
1626         table = fib6_get_table(net, RT6_TABLE_INFO);
1627         if (table == NULL)
1628                 return NULL;
1629
1630         write_lock_bh(&table->tb6_lock);
1631         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1632         if (!fn)
1633                 goto out;
1634
1635         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1636                 if (rt->rt6i_dev->ifindex != ifindex)
1637                         continue;
1638                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1639                         continue;
1640                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1641                         continue;
1642                 dst_hold(&rt->u.dst);
1643                 break;
1644         }
1645 out:
1646         write_unlock_bh(&table->tb6_lock);
1647         return rt;
1648 }
1649
1650 static struct rt6_info *rt6_add_route_info(struct net *net,
1651                                            struct in6_addr *prefix, int prefixlen,
1652                                            struct in6_addr *gwaddr, int ifindex,
1653                                            unsigned pref)
1654 {
1655         struct fib6_config cfg = {
1656                 .fc_table       = RT6_TABLE_INFO,
1657                 .fc_metric      = IP6_RT_PRIO_USER,
1658                 .fc_ifindex     = ifindex,
1659                 .fc_dst_len     = prefixlen,
1660                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1661                                   RTF_UP | RTF_PREF(pref),
1662                 .fc_nlinfo.pid = 0,
1663                 .fc_nlinfo.nlh = NULL,
1664                 .fc_nlinfo.nl_net = net,
1665         };
1666
1667         ipv6_addr_copy(&cfg.fc_dst, prefix);
1668         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1669
1670         /* We should treat it as a default route if prefix length is 0. */
1671         if (!prefixlen)
1672                 cfg.fc_flags |= RTF_DEFAULT;
1673
1674         ip6_route_add(&cfg);
1675
1676         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1677 }
1678 #endif
1679
1680 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1681 {
1682         struct rt6_info *rt;
1683         struct fib6_table *table;
1684
1685         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1686         if (table == NULL)
1687                 return NULL;
1688
1689         write_lock_bh(&table->tb6_lock);
1690         for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1691                 if (dev == rt->rt6i_dev &&
1692                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1693                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1694                         break;
1695         }
1696         if (rt)
1697                 dst_hold(&rt->u.dst);
1698         write_unlock_bh(&table->tb6_lock);
1699         return rt;
1700 }
1701
1702 EXPORT_SYMBOL(rt6_get_dflt_router);
1703
1704 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1705                                      struct net_device *dev,
1706                                      unsigned int pref)
1707 {
1708         struct fib6_config cfg = {
1709                 .fc_table       = RT6_TABLE_DFLT,
1710                 .fc_metric      = IP6_RT_PRIO_USER,
1711                 .fc_ifindex     = dev->ifindex,
1712                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1713                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1714                 .fc_nlinfo.pid = 0,
1715                 .fc_nlinfo.nlh = NULL,
1716                 .fc_nlinfo.nl_net = dev_net(dev),
1717         };
1718
1719         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1720
1721         ip6_route_add(&cfg);
1722
1723         return rt6_get_dflt_router(gwaddr, dev);
1724 }
1725
1726 void rt6_purge_dflt_routers(struct net *net)
1727 {
1728         struct rt6_info *rt;
1729         struct fib6_table *table;
1730
1731         /* NOTE: Keep consistent with rt6_get_dflt_router */
1732         table = fib6_get_table(net, RT6_TABLE_DFLT);
1733         if (table == NULL)
1734                 return;
1735
1736 restart:
1737         read_lock_bh(&table->tb6_lock);
1738         for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1739                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1740                         dst_hold(&rt->u.dst);
1741                         read_unlock_bh(&table->tb6_lock);
1742                         ip6_del_rt(rt);
1743                         goto restart;
1744                 }
1745         }
1746         read_unlock_bh(&table->tb6_lock);
1747 }
1748
1749 static void rtmsg_to_fib6_config(struct net *net,
1750                                  struct in6_rtmsg *rtmsg,
1751                                  struct fib6_config *cfg)
1752 {
1753         memset(cfg, 0, sizeof(*cfg));
1754
1755         cfg->fc_table = RT6_TABLE_MAIN;
1756         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1757         cfg->fc_metric = rtmsg->rtmsg_metric;
1758         cfg->fc_expires = rtmsg->rtmsg_info;
1759         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1760         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1761         cfg->fc_flags = rtmsg->rtmsg_flags;
1762
1763         cfg->fc_nlinfo.nl_net = net;
1764
1765         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1766         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1767         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1768 }
1769
1770 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1771 {
1772         struct fib6_config cfg;
1773         struct in6_rtmsg rtmsg;
1774         int err;
1775
1776         switch(cmd) {
1777         case SIOCADDRT:         /* Add a route */
1778         case SIOCDELRT:         /* Delete a route */
1779                 if (!capable(CAP_NET_ADMIN))
1780                         return -EPERM;
1781                 err = copy_from_user(&rtmsg, arg,
1782                                      sizeof(struct in6_rtmsg));
1783                 if (err)
1784                         return -EFAULT;
1785
1786                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1787
1788                 rtnl_lock();
1789                 switch (cmd) {
1790                 case SIOCADDRT:
1791                         err = ip6_route_add(&cfg);
1792                         break;
1793                 case SIOCDELRT:
1794                         err = ip6_route_del(&cfg);
1795                         break;
1796                 default:
1797                         err = -EINVAL;
1798                 }
1799                 rtnl_unlock();
1800
1801                 return err;
1802         }
1803
1804         return -EINVAL;
1805 }
1806
1807 /*
1808  *      Drop the packet on the floor
1809  */
1810
1811 static int ip6_pkt_drop(struct sk_buff *skb, int code, int ipstats_mib_noroutes)
1812 {
1813         int type;
1814         switch (ipstats_mib_noroutes) {
1815         case IPSTATS_MIB_INNOROUTES:
1816                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1817                 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1818                         IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1819                         break;
1820                 }
1821                 /* FALLTHROUGH */
1822         case IPSTATS_MIB_OUTNOROUTES:
1823                 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1824                 break;
1825         }
1826         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1827         kfree_skb(skb);
1828         return 0;
1829 }
1830
1831 static int ip6_pkt_discard(struct sk_buff *skb)
1832 {
1833         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1834 }
1835
1836 static int ip6_pkt_discard_out(struct sk_buff *skb)
1837 {
1838         skb->dev = skb->dst->dev;
1839         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1840 }
1841
1842 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1843
1844 static int ip6_pkt_prohibit(struct sk_buff *skb)
1845 {
1846         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1847 }
1848
1849 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1850 {
1851         skb->dev = skb->dst->dev;
1852         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1853 }
1854
1855 #endif
1856
1857 /*
1858  *      Allocate a dst for local (unicast / anycast) address.
1859  */
1860
1861 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1862                                     const struct in6_addr *addr,
1863                                     int anycast)
1864 {
1865         struct net *net = dev_net(idev->dev);
1866         struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1867
1868         if (rt == NULL)
1869                 return ERR_PTR(-ENOMEM);
1870
1871         dev_hold(net->loopback_dev);
1872         in6_dev_hold(idev);
1873
1874         rt->u.dst.flags = DST_HOST;
1875         rt->u.dst.input = ip6_input;
1876         rt->u.dst.output = ip6_output;
1877         rt->rt6i_dev = net->loopback_dev;
1878         rt->rt6i_idev = idev;
1879         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1880         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1881         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1882         rt->u.dst.obsolete = -1;
1883
1884         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1885         if (anycast)
1886                 rt->rt6i_flags |= RTF_ANYCAST;
1887         else
1888                 rt->rt6i_flags |= RTF_LOCAL;
1889         rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1890         if (rt->rt6i_nexthop == NULL) {
1891                 dst_free(&rt->u.dst);
1892                 return ERR_PTR(-ENOMEM);
1893         }
1894
1895         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1896         rt->rt6i_dst.plen = 128;
1897         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1898
1899         atomic_set(&rt->u.dst.__refcnt, 1);
1900
1901         return rt;
1902 }
1903
1904 struct arg_dev_net {
1905         struct net_device *dev;
1906         struct net *net;
1907 };
1908
1909 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1910 {
1911         struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1912         struct net *net = ((struct arg_dev_net *)arg)->net;
1913
1914         if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1915             rt != net->ipv6.ip6_null_entry) {
1916                 RT6_TRACE("deleted by ifdown %p\n", rt);
1917                 return -1;
1918         }
1919         return 0;
1920 }
1921
1922 void rt6_ifdown(struct net *net, struct net_device *dev)
1923 {
1924         struct arg_dev_net adn = {
1925                 .dev = dev,
1926                 .net = net,
1927         };
1928
1929         fib6_clean_all(net, fib6_ifdown, 0, &adn);
1930 }
1931
1932 struct rt6_mtu_change_arg
1933 {
1934         struct net_device *dev;
1935         unsigned mtu;
1936 };
1937
1938 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1939 {
1940         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1941         struct inet6_dev *idev;
1942         struct net *net = dev_net(arg->dev);
1943
1944         /* In IPv6 pmtu discovery is not optional,
1945            so that RTAX_MTU lock cannot disable it.
1946            We still use this lock to block changes
1947            caused by addrconf/ndisc.
1948         */
1949
1950         idev = __in6_dev_get(arg->dev);
1951         if (idev == NULL)
1952                 return 0;
1953
1954         /* For administrative MTU increase, there is no way to discover
1955            IPv6 PMTU increase, so PMTU increase should be updated here.
1956            Since RFC 1981 doesn't include administrative MTU increase
1957            update PMTU increase is a MUST. (i.e. jumbo frame)
1958          */
1959         /*
1960            If new MTU is less than route PMTU, this new MTU will be the
1961            lowest MTU in the path, update the route PMTU to reflect PMTU
1962            decreases; if new MTU is greater than route PMTU, and the
1963            old MTU is the lowest MTU in the path, update the route PMTU
1964            to reflect the increase. In this case if the other nodes' MTU
1965            also have the lowest MTU, TOO BIG MESSAGE will be lead to
1966            PMTU discouvery.
1967          */
1968         if (rt->rt6i_dev == arg->dev &&
1969             !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1970             (dst_mtu(&rt->u.dst) >= arg->mtu ||
1971              (dst_mtu(&rt->u.dst) < arg->mtu &&
1972               dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1973                 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1974                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
1975         }
1976         return 0;
1977 }
1978
1979 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1980 {
1981         struct rt6_mtu_change_arg arg = {
1982                 .dev = dev,
1983                 .mtu = mtu,
1984         };
1985
1986         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
1987 }
1988
1989 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1990         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
1991         [RTA_OIF]               = { .type = NLA_U32 },
1992         [RTA_IIF]               = { .type = NLA_U32 },
1993         [RTA_PRIORITY]          = { .type = NLA_U32 },
1994         [RTA_METRICS]           = { .type = NLA_NESTED },
1995 };
1996
1997 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1998                               struct fib6_config *cfg)
1999 {
2000         struct rtmsg *rtm;
2001         struct nlattr *tb[RTA_MAX+1];
2002         int err;
2003
2004         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2005         if (err < 0)
2006                 goto errout;
2007
2008         err = -EINVAL;
2009         rtm = nlmsg_data(nlh);
2010         memset(cfg, 0, sizeof(*cfg));
2011
2012         cfg->fc_table = rtm->rtm_table;
2013         cfg->fc_dst_len = rtm->rtm_dst_len;
2014         cfg->fc_src_len = rtm->rtm_src_len;
2015         cfg->fc_flags = RTF_UP;
2016         cfg->fc_protocol = rtm->rtm_protocol;
2017
2018         if (rtm->rtm_type == RTN_UNREACHABLE)
2019                 cfg->fc_flags |= RTF_REJECT;
2020
2021         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2022         cfg->fc_nlinfo.nlh = nlh;
2023         cfg->fc_nlinfo.nl_net = skb->sk->sk_net;
2024
2025         if (tb[RTA_GATEWAY]) {
2026                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2027                 cfg->fc_flags |= RTF_GATEWAY;
2028         }
2029
2030         if (tb[RTA_DST]) {
2031                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2032
2033                 if (nla_len(tb[RTA_DST]) < plen)
2034                         goto errout;
2035
2036                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2037         }
2038
2039         if (tb[RTA_SRC]) {
2040                 int plen = (rtm->rtm_src_len + 7) >> 3;
2041
2042                 if (nla_len(tb[RTA_SRC]) < plen)
2043                         goto errout;
2044
2045                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2046         }
2047
2048         if (tb[RTA_OIF])
2049                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2050
2051         if (tb[RTA_PRIORITY])
2052                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2053
2054         if (tb[RTA_METRICS]) {
2055                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2056                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2057         }
2058
2059         if (tb[RTA_TABLE])
2060                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2061
2062         err = 0;
2063 errout:
2064         return err;
2065 }
2066
2067 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2068 {
2069         struct fib6_config cfg;
2070         int err;
2071
2072         err = rtm_to_fib6_config(skb, nlh, &cfg);
2073         if (err < 0)
2074                 return err;
2075
2076         return ip6_route_del(&cfg);
2077 }
2078
2079 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2080 {
2081         struct fib6_config cfg;
2082         int err;
2083
2084         err = rtm_to_fib6_config(skb, nlh, &cfg);
2085         if (err < 0)
2086                 return err;
2087
2088         return ip6_route_add(&cfg);
2089 }
2090
2091 static inline size_t rt6_nlmsg_size(void)
2092 {
2093         return NLMSG_ALIGN(sizeof(struct rtmsg))
2094                + nla_total_size(16) /* RTA_SRC */
2095                + nla_total_size(16) /* RTA_DST */
2096                + nla_total_size(16) /* RTA_GATEWAY */
2097                + nla_total_size(16) /* RTA_PREFSRC */
2098                + nla_total_size(4) /* RTA_TABLE */
2099                + nla_total_size(4) /* RTA_IIF */
2100                + nla_total_size(4) /* RTA_OIF */
2101                + nla_total_size(4) /* RTA_PRIORITY */
2102                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2103                + nla_total_size(sizeof(struct rta_cacheinfo));
2104 }
2105
2106 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2107                          struct in6_addr *dst, struct in6_addr *src,
2108                          int iif, int type, u32 pid, u32 seq,
2109                          int prefix, unsigned int flags)
2110 {
2111         struct rtmsg *rtm;
2112         struct nlmsghdr *nlh;
2113         long expires;
2114         u32 table;
2115
2116         if (prefix) {   /* user wants prefix routes only */
2117                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2118                         /* success since this is not a prefix route */
2119                         return 1;
2120                 }
2121         }
2122
2123         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2124         if (nlh == NULL)
2125                 return -EMSGSIZE;
2126
2127         rtm = nlmsg_data(nlh);
2128         rtm->rtm_family = AF_INET6;
2129         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2130         rtm->rtm_src_len = rt->rt6i_src.plen;
2131         rtm->rtm_tos = 0;
2132         if (rt->rt6i_table)
2133                 table = rt->rt6i_table->tb6_id;
2134         else
2135                 table = RT6_TABLE_UNSPEC;
2136         rtm->rtm_table = table;
2137         NLA_PUT_U32(skb, RTA_TABLE, table);
2138         if (rt->rt6i_flags&RTF_REJECT)
2139                 rtm->rtm_type = RTN_UNREACHABLE;
2140         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2141                 rtm->rtm_type = RTN_LOCAL;
2142         else
2143                 rtm->rtm_type = RTN_UNICAST;
2144         rtm->rtm_flags = 0;
2145         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2146         rtm->rtm_protocol = rt->rt6i_protocol;
2147         if (rt->rt6i_flags&RTF_DYNAMIC)
2148                 rtm->rtm_protocol = RTPROT_REDIRECT;
2149         else if (rt->rt6i_flags & RTF_ADDRCONF)
2150                 rtm->rtm_protocol = RTPROT_KERNEL;
2151         else if (rt->rt6i_flags&RTF_DEFAULT)
2152                 rtm->rtm_protocol = RTPROT_RA;
2153
2154         if (rt->rt6i_flags&RTF_CACHE)
2155                 rtm->rtm_flags |= RTM_F_CLONED;
2156
2157         if (dst) {
2158                 NLA_PUT(skb, RTA_DST, 16, dst);
2159                 rtm->rtm_dst_len = 128;
2160         } else if (rtm->rtm_dst_len)
2161                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2162 #ifdef CONFIG_IPV6_SUBTREES
2163         if (src) {
2164                 NLA_PUT(skb, RTA_SRC, 16, src);
2165                 rtm->rtm_src_len = 128;
2166         } else if (rtm->rtm_src_len)
2167                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2168 #endif
2169         if (iif)
2170                 NLA_PUT_U32(skb, RTA_IIF, iif);
2171         else if (dst) {
2172                 struct in6_addr saddr_buf;
2173                 if (ipv6_dev_get_saddr(ip6_dst_idev(&rt->u.dst)->dev,
2174                                        dst, 0, &saddr_buf) == 0)
2175                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2176         }
2177
2178         if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2179                 goto nla_put_failure;
2180
2181         if (rt->u.dst.neighbour)
2182                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2183
2184         if (rt->u.dst.dev)
2185                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2186
2187         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2188
2189         expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2190         if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2191                                expires, rt->u.dst.error) < 0)
2192                 goto nla_put_failure;
2193
2194         return nlmsg_end(skb, nlh);
2195
2196 nla_put_failure:
2197         nlmsg_cancel(skb, nlh);
2198         return -EMSGSIZE;
2199 }
2200
2201 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2202 {
2203         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2204         int prefix;
2205
2206         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2207                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2208                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2209         } else
2210                 prefix = 0;
2211
2212         return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2213                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2214                      prefix, NLM_F_MULTI);
2215 }
2216
2217 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2218 {
2219         struct net *net = in_skb->sk->sk_net;
2220         struct nlattr *tb[RTA_MAX+1];
2221         struct rt6_info *rt;
2222         struct sk_buff *skb;
2223         struct rtmsg *rtm;
2224         struct flowi fl;
2225         int err, iif = 0;
2226
2227         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2228         if (err < 0)
2229                 goto errout;
2230
2231         err = -EINVAL;
2232         memset(&fl, 0, sizeof(fl));
2233
2234         if (tb[RTA_SRC]) {
2235                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2236                         goto errout;
2237
2238                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2239         }
2240
2241         if (tb[RTA_DST]) {
2242                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2243                         goto errout;
2244
2245                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2246         }
2247
2248         if (tb[RTA_IIF])
2249                 iif = nla_get_u32(tb[RTA_IIF]);
2250
2251         if (tb[RTA_OIF])
2252                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2253
2254         if (iif) {
2255                 struct net_device *dev;
2256                 dev = __dev_get_by_index(net, iif);
2257                 if (!dev) {
2258                         err = -ENODEV;
2259                         goto errout;
2260                 }
2261         }
2262
2263         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2264         if (skb == NULL) {
2265                 err = -ENOBUFS;
2266                 goto errout;
2267         }
2268
2269         /* Reserve room for dummy headers, this skb can pass
2270            through good chunk of routing engine.
2271          */
2272         skb_reset_mac_header(skb);
2273         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2274
2275         rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2276         skb->dst = &rt->u.dst;
2277
2278         err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2279                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2280                             nlh->nlmsg_seq, 0, 0);
2281         if (err < 0) {
2282                 kfree_skb(skb);
2283                 goto errout;
2284         }
2285
2286         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2287 errout:
2288         return err;
2289 }
2290
2291 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2292 {
2293         struct sk_buff *skb;
2294         struct net *net = info->nl_net;
2295         u32 seq;
2296         int err;
2297
2298         err = -ENOBUFS;
2299         seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2300
2301         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2302         if (skb == NULL)
2303                 goto errout;
2304
2305         err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2306                                 event, info->pid, seq, 0, 0);
2307         if (err < 0) {
2308                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2309                 WARN_ON(err == -EMSGSIZE);
2310                 kfree_skb(skb);
2311                 goto errout;
2312         }
2313         err = rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2314                           info->nlh, gfp_any());
2315 errout:
2316         if (err < 0)
2317                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2318 }
2319
2320 static int ip6_route_dev_notify(struct notifier_block *this,
2321                                 unsigned long event, void *data)
2322 {
2323         struct net_device *dev = (struct net_device *)data;
2324         struct net *net = dev_net(dev);
2325
2326         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2327                 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2328                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2329 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2330                 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2331                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2332                 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2333                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2334 #endif
2335         }
2336
2337         return NOTIFY_OK;
2338 }
2339
2340 /*
2341  *      /proc
2342  */
2343
2344 #ifdef CONFIG_PROC_FS
2345
2346 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2347
2348 struct rt6_proc_arg
2349 {
2350         char *buffer;
2351         int offset;
2352         int length;
2353         int skip;
2354         int len;
2355 };
2356
2357 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2358 {
2359         struct seq_file *m = p_arg;
2360
2361         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2362                    rt->rt6i_dst.plen);
2363
2364 #ifdef CONFIG_IPV6_SUBTREES
2365         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2366                    rt->rt6i_src.plen);
2367 #else
2368         seq_puts(m, "00000000000000000000000000000000 00 ");
2369 #endif
2370
2371         if (rt->rt6i_nexthop) {
2372                 seq_printf(m, NIP6_SEQFMT,
2373                            NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2374         } else {
2375                 seq_puts(m, "00000000000000000000000000000000");
2376         }
2377         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2378                    rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2379                    rt->u.dst.__use, rt->rt6i_flags,
2380                    rt->rt6i_dev ? rt->rt6i_dev->name : "");
2381         return 0;
2382 }
2383
2384 static int ipv6_route_show(struct seq_file *m, void *v)
2385 {
2386         struct net *net = (struct net *)m->private;
2387         fib6_clean_all(net, rt6_info_route, 0, m);
2388         return 0;
2389 }
2390
2391 static int ipv6_route_open(struct inode *inode, struct file *file)
2392 {
2393         struct net *net = get_proc_net(inode);
2394         if (!net)
2395                 return -ENXIO;
2396         return single_open(file, ipv6_route_show, net);
2397 }
2398
2399 static int ipv6_route_release(struct inode *inode, struct file *file)
2400 {
2401         struct seq_file *seq = file->private_data;
2402         struct net *net = seq->private;
2403         put_net(net);
2404         return single_release(inode, file);
2405 }
2406
2407 static const struct file_operations ipv6_route_proc_fops = {
2408         .owner          = THIS_MODULE,
2409         .open           = ipv6_route_open,
2410         .read           = seq_read,
2411         .llseek         = seq_lseek,
2412         .release        = ipv6_route_release,
2413 };
2414
2415 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2416 {
2417         struct net *net = (struct net *)seq->private;
2418         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2419                    net->ipv6.rt6_stats->fib_nodes,
2420                    net->ipv6.rt6_stats->fib_route_nodes,
2421                    net->ipv6.rt6_stats->fib_rt_alloc,
2422                    net->ipv6.rt6_stats->fib_rt_entries,
2423                    net->ipv6.rt6_stats->fib_rt_cache,
2424                    atomic_read(&net->ipv6.ip6_dst_ops->entries),
2425                    net->ipv6.rt6_stats->fib_discarded_routes);
2426
2427         return 0;
2428 }
2429
2430 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2431 {
2432         struct net *net = get_proc_net(inode);
2433         return single_open(file, rt6_stats_seq_show, net);
2434 }
2435
2436 static int rt6_stats_seq_release(struct inode *inode, struct file *file)
2437 {
2438         struct seq_file *seq = file->private_data;
2439         struct net *net = (struct net *)seq->private;
2440         put_net(net);
2441         return single_release(inode, file);
2442 }
2443
2444 static const struct file_operations rt6_stats_seq_fops = {
2445         .owner   = THIS_MODULE,
2446         .open    = rt6_stats_seq_open,
2447         .read    = seq_read,
2448         .llseek  = seq_lseek,
2449         .release = rt6_stats_seq_release,
2450 };
2451 #endif  /* CONFIG_PROC_FS */
2452
2453 #ifdef CONFIG_SYSCTL
2454
2455 static
2456 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2457                               void __user *buffer, size_t *lenp, loff_t *ppos)
2458 {
2459         struct net *net = current->nsproxy->net_ns;
2460         int delay = net->ipv6.sysctl.flush_delay;
2461         if (write) {
2462                 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2463                 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2464                 return 0;
2465         } else
2466                 return -EINVAL;
2467 }
2468
2469 ctl_table ipv6_route_table_template[] = {
2470         {
2471                 .procname       =       "flush",
2472                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2473                 .maxlen         =       sizeof(int),
2474                 .mode           =       0200,
2475                 .proc_handler   =       &ipv6_sysctl_rtcache_flush
2476         },
2477         {
2478                 .ctl_name       =       NET_IPV6_ROUTE_GC_THRESH,
2479                 .procname       =       "gc_thresh",
2480                 .data           =       &ip6_dst_ops_template.gc_thresh,
2481                 .maxlen         =       sizeof(int),
2482                 .mode           =       0644,
2483                 .proc_handler   =       &proc_dointvec,
2484         },
2485         {
2486                 .ctl_name       =       NET_IPV6_ROUTE_MAX_SIZE,
2487                 .procname       =       "max_size",
2488                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2489                 .maxlen         =       sizeof(int),
2490                 .mode           =       0644,
2491                 .proc_handler   =       &proc_dointvec,
2492         },
2493         {
2494                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2495                 .procname       =       "gc_min_interval",
2496                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2497                 .maxlen         =       sizeof(int),
2498                 .mode           =       0644,
2499                 .proc_handler   =       &proc_dointvec_jiffies,
2500                 .strategy       =       &sysctl_jiffies,
2501         },
2502         {
2503                 .ctl_name       =       NET_IPV6_ROUTE_GC_TIMEOUT,
2504                 .procname       =       "gc_timeout",
2505                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2506                 .maxlen         =       sizeof(int),
2507                 .mode           =       0644,
2508                 .proc_handler   =       &proc_dointvec_jiffies,
2509                 .strategy       =       &sysctl_jiffies,
2510         },
2511         {
2512                 .ctl_name       =       NET_IPV6_ROUTE_GC_INTERVAL,
2513                 .procname       =       "gc_interval",
2514                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2515                 .maxlen         =       sizeof(int),
2516                 .mode           =       0644,
2517                 .proc_handler   =       &proc_dointvec_jiffies,
2518                 .strategy       =       &sysctl_jiffies,
2519         },
2520         {
2521                 .ctl_name       =       NET_IPV6_ROUTE_GC_ELASTICITY,
2522                 .procname       =       "gc_elasticity",
2523                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2524                 .maxlen         =       sizeof(int),
2525                 .mode           =       0644,
2526                 .proc_handler   =       &proc_dointvec_jiffies,
2527                 .strategy       =       &sysctl_jiffies,
2528         },
2529         {
2530                 .ctl_name       =       NET_IPV6_ROUTE_MTU_EXPIRES,
2531                 .procname       =       "mtu_expires",
2532                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2533                 .maxlen         =       sizeof(int),
2534                 .mode           =       0644,
2535                 .proc_handler   =       &proc_dointvec_jiffies,
2536                 .strategy       =       &sysctl_jiffies,
2537         },
2538         {
2539                 .ctl_name       =       NET_IPV6_ROUTE_MIN_ADVMSS,
2540                 .procname       =       "min_adv_mss",
2541                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2542                 .maxlen         =       sizeof(int),
2543                 .mode           =       0644,
2544                 .proc_handler   =       &proc_dointvec_jiffies,
2545                 .strategy       =       &sysctl_jiffies,
2546         },
2547         {
2548                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2549                 .procname       =       "gc_min_interval_ms",
2550                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2551                 .maxlen         =       sizeof(int),
2552                 .mode           =       0644,
2553                 .proc_handler   =       &proc_dointvec_ms_jiffies,
2554                 .strategy       =       &sysctl_ms_jiffies,
2555         },
2556         { .ctl_name = 0 }
2557 };
2558
2559 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2560 {
2561         struct ctl_table *table;
2562
2563         table = kmemdup(ipv6_route_table_template,
2564                         sizeof(ipv6_route_table_template),
2565                         GFP_KERNEL);
2566
2567         if (table) {
2568                 table[0].data = &net->ipv6.sysctl.flush_delay;
2569                 table[1].data = &net->ipv6.ip6_dst_ops->gc_thresh;
2570                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2571                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2572                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2573                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2574                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2575                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2576                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2577         }
2578
2579         return table;
2580 }
2581 #endif
2582
2583 static int ip6_route_net_init(struct net *net)
2584 {
2585         int ret = 0;
2586
2587         ret = -ENOMEM;
2588         net->ipv6.ip6_dst_ops = kmemdup(&ip6_dst_ops_template,
2589                                         sizeof(*net->ipv6.ip6_dst_ops),
2590                                         GFP_KERNEL);
2591         if (!net->ipv6.ip6_dst_ops)
2592                 goto out;
2593         net->ipv6.ip6_dst_ops->dst_net = net;
2594
2595         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2596                                            sizeof(*net->ipv6.ip6_null_entry),
2597                                            GFP_KERNEL);
2598         if (!net->ipv6.ip6_null_entry)
2599                 goto out_ip6_dst_ops;
2600         net->ipv6.ip6_null_entry->u.dst.path =
2601                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2602         net->ipv6.ip6_null_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2603
2604 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2605         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2606                                                sizeof(*net->ipv6.ip6_prohibit_entry),
2607                                                GFP_KERNEL);
2608         if (!net->ipv6.ip6_prohibit_entry) {
2609                 kfree(net->ipv6.ip6_null_entry);
2610                 goto out;
2611         }
2612         net->ipv6.ip6_prohibit_entry->u.dst.path =
2613                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2614         net->ipv6.ip6_prohibit_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2615
2616         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2617                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
2618                                                GFP_KERNEL);
2619         if (!net->ipv6.ip6_blk_hole_entry) {
2620                 kfree(net->ipv6.ip6_null_entry);
2621                 kfree(net->ipv6.ip6_prohibit_entry);
2622                 goto out;
2623         }
2624         net->ipv6.ip6_blk_hole_entry->u.dst.path =
2625                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2626         net->ipv6.ip6_blk_hole_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2627 #endif
2628
2629 #ifdef CONFIG_PROC_FS
2630         proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2631         proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2632 #endif
2633         net->ipv6.ip6_rt_gc_expire = 30*HZ;
2634
2635         ret = 0;
2636 out:
2637         return ret;
2638
2639 out_ip6_dst_ops:
2640         kfree(net->ipv6.ip6_dst_ops);
2641         goto out;
2642 }
2643
2644 static void ip6_route_net_exit(struct net *net)
2645 {
2646 #ifdef CONFIG_PROC_FS
2647         proc_net_remove(net, "ipv6_route");
2648         proc_net_remove(net, "rt6_stats");
2649 #endif
2650         kfree(net->ipv6.ip6_null_entry);
2651 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2652         kfree(net->ipv6.ip6_prohibit_entry);
2653         kfree(net->ipv6.ip6_blk_hole_entry);
2654 #endif
2655         kfree(net->ipv6.ip6_dst_ops);
2656 }
2657
2658 static struct pernet_operations ip6_route_net_ops = {
2659         .init = ip6_route_net_init,
2660         .exit = ip6_route_net_exit,
2661 };
2662
2663 static struct notifier_block ip6_route_dev_notifier = {
2664         .notifier_call = ip6_route_dev_notify,
2665         .priority = 0,
2666 };
2667
2668 int __init ip6_route_init(void)
2669 {
2670         int ret;
2671
2672         ret = -ENOMEM;
2673         ip6_dst_ops_template.kmem_cachep =
2674                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2675                                   SLAB_HWCACHE_ALIGN, NULL);
2676         if (!ip6_dst_ops_template.kmem_cachep)
2677                 goto out;;
2678
2679         ret = register_pernet_subsys(&ip6_route_net_ops);
2680         if (ret)
2681                 goto out_kmem_cache;
2682
2683         /* Registering of the loopback is done before this portion of code,
2684          * the loopback reference in rt6_info will not be taken, do it
2685          * manually for init_net */
2686         init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2687         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2688   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2689         init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2690         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2691         init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2692         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2693   #endif
2694         ret = fib6_init();
2695         if (ret)
2696                 goto out_register_subsys;
2697
2698         ret = xfrm6_init();
2699         if (ret)
2700                 goto out_fib6_init;
2701
2702         ret = fib6_rules_init();
2703         if (ret)
2704                 goto xfrm6_init;
2705
2706         ret = -ENOBUFS;
2707         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2708             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2709             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2710                 goto fib6_rules_init;
2711
2712         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2713         if (ret)
2714                 goto fib6_rules_init;
2715
2716 out:
2717         return ret;
2718
2719 fib6_rules_init:
2720         fib6_rules_cleanup();
2721 xfrm6_init:
2722         xfrm6_fini();
2723 out_fib6_init:
2724         fib6_gc_cleanup();
2725 out_register_subsys:
2726         unregister_pernet_subsys(&ip6_route_net_ops);
2727 out_kmem_cache:
2728         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2729         goto out;
2730 }
2731
2732 void ip6_route_cleanup(void)
2733 {
2734         unregister_netdevice_notifier(&ip6_route_dev_notifier);
2735         fib6_rules_cleanup();
2736         xfrm6_fini();
2737         fib6_gc_cleanup();
2738         unregister_pernet_subsys(&ip6_route_net_ops);
2739         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2740 }