netfilter: nf_conntrack_ipv6: fix inconsistent lock state in nf_ct_frag6_gather()
[linux-2.6] / net / ipv6 / route.c
1 /*
2  *      Linux INET6 implementation
3  *      FIB front-end.
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      $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/mroute6.h>
40 #include <linux/init.h>
41 #include <linux/if_arp.h>
42 #include <linux/proc_fs.h>
43 #include <linux/seq_file.h>
44 #include <linux/nsproxy.h>
45 #include <net/net_namespace.h>
46 #include <net/snmp.h>
47 #include <net/ipv6.h>
48 #include <net/ip6_fib.h>
49 #include <net/ip6_route.h>
50 #include <net/ndisc.h>
51 #include <net/addrconf.h>
52 #include <net/tcp.h>
53 #include <linux/rtnetlink.h>
54 #include <net/dst.h>
55 #include <net/xfrm.h>
56 #include <net/netevent.h>
57 #include <net/netlink.h>
58
59 #include <asm/uaccess.h>
60
61 #ifdef CONFIG_SYSCTL
62 #include <linux/sysctl.h>
63 #endif
64
65 /* Set to 3 to get tracing. */
66 #define RT6_DEBUG 2
67
68 #if RT6_DEBUG >= 3
69 #define RDBG(x) printk x
70 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
71 #else
72 #define RDBG(x)
73 #define RT6_TRACE(x...) do { ; } while (0)
74 #endif
75
76 #define CLONE_OFFLINK_ROUTE 0
77
78 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
79 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
81 static void             ip6_dst_destroy(struct dst_entry *);
82 static void             ip6_dst_ifdown(struct dst_entry *,
83                                        struct net_device *dev, int how);
84 static int               ip6_dst_gc(struct dst_ops *ops);
85
86 static int              ip6_pkt_discard(struct sk_buff *skb);
87 static int              ip6_pkt_discard_out(struct sk_buff *skb);
88 static void             ip6_link_failure(struct sk_buff *skb);
89 static void             ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
90
91 #ifdef CONFIG_IPV6_ROUTE_INFO
92 static struct rt6_info *rt6_add_route_info(struct net *net,
93                                            struct in6_addr *prefix, int prefixlen,
94                                            struct in6_addr *gwaddr, int ifindex,
95                                            unsigned pref);
96 static struct rt6_info *rt6_get_route_info(struct net *net,
97                                            struct in6_addr *prefix, int prefixlen,
98                                            struct in6_addr *gwaddr, int ifindex);
99 #endif
100
101 static struct dst_ops ip6_dst_ops_template = {
102         .family                 =       AF_INET6,
103         .protocol               =       __constant_htons(ETH_P_IPV6),
104         .gc                     =       ip6_dst_gc,
105         .gc_thresh              =       1024,
106         .check                  =       ip6_dst_check,
107         .destroy                =       ip6_dst_destroy,
108         .ifdown                 =       ip6_dst_ifdown,
109         .negative_advice        =       ip6_negative_advice,
110         .link_failure           =       ip6_link_failure,
111         .update_pmtu            =       ip6_rt_update_pmtu,
112         .local_out              =       __ip6_local_out,
113         .entry_size             =       sizeof(struct rt6_info),
114         .entries                =       ATOMIC_INIT(0),
115 };
116
117 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
118 {
119 }
120
121 static struct dst_ops ip6_dst_blackhole_ops = {
122         .family                 =       AF_INET6,
123         .protocol               =       __constant_htons(ETH_P_IPV6),
124         .destroy                =       ip6_dst_destroy,
125         .check                  =       ip6_dst_check,
126         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
127         .entry_size             =       sizeof(struct rt6_info),
128         .entries                =       ATOMIC_INIT(0),
129 };
130
131 static struct rt6_info ip6_null_entry_template = {
132         .u = {
133                 .dst = {
134                         .__refcnt       = ATOMIC_INIT(1),
135                         .__use          = 1,
136                         .obsolete       = -1,
137                         .error          = -ENETUNREACH,
138                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
139                         .input          = ip6_pkt_discard,
140                         .output         = ip6_pkt_discard_out,
141                 }
142         },
143         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
144         .rt6i_metric    = ~(u32) 0,
145         .rt6i_ref       = ATOMIC_INIT(1),
146 };
147
148 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
149
150 static int ip6_pkt_prohibit(struct sk_buff *skb);
151 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
152
153 static struct rt6_info ip6_prohibit_entry_template = {
154         .u = {
155                 .dst = {
156                         .__refcnt       = ATOMIC_INIT(1),
157                         .__use          = 1,
158                         .obsolete       = -1,
159                         .error          = -EACCES,
160                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
161                         .input          = ip6_pkt_prohibit,
162                         .output         = ip6_pkt_prohibit_out,
163                 }
164         },
165         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
166         .rt6i_metric    = ~(u32) 0,
167         .rt6i_ref       = ATOMIC_INIT(1),
168 };
169
170 static struct rt6_info ip6_blk_hole_entry_template = {
171         .u = {
172                 .dst = {
173                         .__refcnt       = ATOMIC_INIT(1),
174                         .__use          = 1,
175                         .obsolete       = -1,
176                         .error          = -EINVAL,
177                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
178                         .input          = dst_discard,
179                         .output         = dst_discard,
180                 }
181         },
182         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
183         .rt6i_metric    = ~(u32) 0,
184         .rt6i_ref       = ATOMIC_INIT(1),
185 };
186
187 #endif
188
189 /* allocate dst with ip6_dst_ops */
190 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
191 {
192         return (struct rt6_info *)dst_alloc(ops);
193 }
194
195 static void ip6_dst_destroy(struct dst_entry *dst)
196 {
197         struct rt6_info *rt = (struct rt6_info *)dst;
198         struct inet6_dev *idev = rt->rt6i_idev;
199
200         if (idev != NULL) {
201                 rt->rt6i_idev = NULL;
202                 in6_dev_put(idev);
203         }
204 }
205
206 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
207                            int how)
208 {
209         struct rt6_info *rt = (struct rt6_info *)dst;
210         struct inet6_dev *idev = rt->rt6i_idev;
211         struct net_device *loopback_dev =
212                 dev_net(dev)->loopback_dev;
213
214         if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
215                 struct inet6_dev *loopback_idev =
216                         in6_dev_get(loopback_dev);
217                 if (loopback_idev != NULL) {
218                         rt->rt6i_idev = loopback_idev;
219                         in6_dev_put(idev);
220                 }
221         }
222 }
223
224 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
225 {
226         return (rt->rt6i_flags & RTF_EXPIRES &&
227                 time_after(jiffies, rt->rt6i_expires));
228 }
229
230 static inline int rt6_need_strict(struct in6_addr *daddr)
231 {
232         return (ipv6_addr_type(daddr) &
233                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
234 }
235
236 /*
237  *      Route lookup. Any table->tb6_lock is implied.
238  */
239
240 static inline struct rt6_info *rt6_device_match(struct net *net,
241                                                     struct rt6_info *rt,
242                                                     int oif,
243                                                     int strict)
244 {
245         struct rt6_info *local = NULL;
246         struct rt6_info *sprt;
247
248         if (oif) {
249                 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
250                         struct net_device *dev = sprt->rt6i_dev;
251                         if (dev->ifindex == oif)
252                                 return sprt;
253                         if (dev->flags & IFF_LOOPBACK) {
254                                 if (sprt->rt6i_idev == NULL ||
255                                     sprt->rt6i_idev->dev->ifindex != oif) {
256                                         if (strict && oif)
257                                                 continue;
258                                         if (local && (!oif ||
259                                                       local->rt6i_idev->dev->ifindex == oif))
260                                                 continue;
261                                 }
262                                 local = sprt;
263                         }
264                 }
265
266                 if (local)
267                         return local;
268
269                 if (strict)
270                         return net->ipv6.ip6_null_entry;
271         }
272         return rt;
273 }
274
275 #ifdef CONFIG_IPV6_ROUTER_PREF
276 static void rt6_probe(struct rt6_info *rt)
277 {
278         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
279         /*
280          * Okay, this does not seem to be appropriate
281          * for now, however, we need to check if it
282          * is really so; aka Router Reachability Probing.
283          *
284          * Router Reachability Probe MUST be rate-limited
285          * to no more than one per minute.
286          */
287         if (!neigh || (neigh->nud_state & NUD_VALID))
288                 return;
289         read_lock_bh(&neigh->lock);
290         if (!(neigh->nud_state & NUD_VALID) &&
291             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
292                 struct in6_addr mcaddr;
293                 struct in6_addr *target;
294
295                 neigh->updated = jiffies;
296                 read_unlock_bh(&neigh->lock);
297
298                 target = (struct in6_addr *)&neigh->primary_key;
299                 addrconf_addr_solict_mult(target, &mcaddr);
300                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
301         } else
302                 read_unlock_bh(&neigh->lock);
303 }
304 #else
305 static inline void rt6_probe(struct rt6_info *rt)
306 {
307         return;
308 }
309 #endif
310
311 /*
312  * Default Router Selection (RFC 2461 6.3.6)
313  */
314 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
315 {
316         struct net_device *dev = rt->rt6i_dev;
317         if (!oif || dev->ifindex == oif)
318                 return 2;
319         if ((dev->flags & IFF_LOOPBACK) &&
320             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
321                 return 1;
322         return 0;
323 }
324
325 static inline int rt6_check_neigh(struct rt6_info *rt)
326 {
327         struct neighbour *neigh = rt->rt6i_nexthop;
328         int m;
329         if (rt->rt6i_flags & RTF_NONEXTHOP ||
330             !(rt->rt6i_flags & RTF_GATEWAY))
331                 m = 1;
332         else if (neigh) {
333                 read_lock_bh(&neigh->lock);
334                 if (neigh->nud_state & NUD_VALID)
335                         m = 2;
336 #ifdef CONFIG_IPV6_ROUTER_PREF
337                 else if (neigh->nud_state & NUD_FAILED)
338                         m = 0;
339 #endif
340                 else
341                         m = 1;
342                 read_unlock_bh(&neigh->lock);
343         } else
344                 m = 0;
345         return m;
346 }
347
348 static int rt6_score_route(struct rt6_info *rt, int oif,
349                            int strict)
350 {
351         int m, n;
352
353         m = rt6_check_dev(rt, oif);
354         if (!m && (strict & RT6_LOOKUP_F_IFACE))
355                 return -1;
356 #ifdef CONFIG_IPV6_ROUTER_PREF
357         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
358 #endif
359         n = rt6_check_neigh(rt);
360         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
361                 return -1;
362         return m;
363 }
364
365 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
366                                    int *mpri, struct rt6_info *match)
367 {
368         int m;
369
370         if (rt6_check_expired(rt))
371                 goto out;
372
373         m = rt6_score_route(rt, oif, strict);
374         if (m < 0)
375                 goto out;
376
377         if (m > *mpri) {
378                 if (strict & RT6_LOOKUP_F_REACHABLE)
379                         rt6_probe(match);
380                 *mpri = m;
381                 match = rt;
382         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
383                 rt6_probe(rt);
384         }
385
386 out:
387         return match;
388 }
389
390 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
391                                      struct rt6_info *rr_head,
392                                      u32 metric, int oif, int strict)
393 {
394         struct rt6_info *rt, *match;
395         int mpri = -1;
396
397         match = NULL;
398         for (rt = rr_head; rt && rt->rt6i_metric == metric;
399              rt = rt->u.dst.rt6_next)
400                 match = find_match(rt, oif, strict, &mpri, match);
401         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
402              rt = rt->u.dst.rt6_next)
403                 match = find_match(rt, oif, strict, &mpri, match);
404
405         return match;
406 }
407
408 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
409 {
410         struct rt6_info *match, *rt0;
411         struct net *net;
412
413         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
414                   __func__, fn->leaf, oif);
415
416         rt0 = fn->rr_ptr;
417         if (!rt0)
418                 fn->rr_ptr = rt0 = fn->leaf;
419
420         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
421
422         if (!match &&
423             (strict & RT6_LOOKUP_F_REACHABLE)) {
424                 struct rt6_info *next = rt0->u.dst.rt6_next;
425
426                 /* no entries matched; do round-robin */
427                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
428                         next = fn->leaf;
429
430                 if (next != rt0)
431                         fn->rr_ptr = next;
432         }
433
434         RT6_TRACE("%s() => %p\n",
435                   __func__, match);
436
437         net = dev_net(rt0->rt6i_dev);
438         return (match ? match : net->ipv6.ip6_null_entry);
439 }
440
441 #ifdef CONFIG_IPV6_ROUTE_INFO
442 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
443                   struct in6_addr *gwaddr)
444 {
445         struct net *net = dev_net(dev);
446         struct route_info *rinfo = (struct route_info *) opt;
447         struct in6_addr prefix_buf, *prefix;
448         unsigned int pref;
449         u32 lifetime;
450         struct rt6_info *rt;
451
452         if (len < sizeof(struct route_info)) {
453                 return -EINVAL;
454         }
455
456         /* Sanity check for prefix_len and length */
457         if (rinfo->length > 3) {
458                 return -EINVAL;
459         } else if (rinfo->prefix_len > 128) {
460                 return -EINVAL;
461         } else if (rinfo->prefix_len > 64) {
462                 if (rinfo->length < 2) {
463                         return -EINVAL;
464                 }
465         } else if (rinfo->prefix_len > 0) {
466                 if (rinfo->length < 1) {
467                         return -EINVAL;
468                 }
469         }
470
471         pref = rinfo->route_pref;
472         if (pref == ICMPV6_ROUTER_PREF_INVALID)
473                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
474
475         lifetime = ntohl(rinfo->lifetime);
476         if (lifetime == 0xffffffff) {
477                 /* infinity */
478         } else if (lifetime > 0x7fffffff/HZ - 1) {
479                 /* Avoid arithmetic overflow */
480                 lifetime = 0x7fffffff/HZ - 1;
481         }
482
483         if (rinfo->length == 3)
484                 prefix = (struct in6_addr *)rinfo->prefix;
485         else {
486                 /* this function is safe */
487                 ipv6_addr_prefix(&prefix_buf,
488                                  (struct in6_addr *)rinfo->prefix,
489                                  rinfo->prefix_len);
490                 prefix = &prefix_buf;
491         }
492
493         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
494                                 dev->ifindex);
495
496         if (rt && !lifetime) {
497                 ip6_del_rt(rt);
498                 rt = NULL;
499         }
500
501         if (!rt && lifetime)
502                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
503                                         pref);
504         else if (rt)
505                 rt->rt6i_flags = RTF_ROUTEINFO |
506                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
507
508         if (rt) {
509                 if (lifetime == 0xffffffff) {
510                         rt->rt6i_flags &= ~RTF_EXPIRES;
511                 } else {
512                         rt->rt6i_expires = jiffies + HZ * lifetime;
513                         rt->rt6i_flags |= RTF_EXPIRES;
514                 }
515                 dst_release(&rt->u.dst);
516         }
517         return 0;
518 }
519 #endif
520
521 #define BACKTRACK(__net, saddr)                 \
522 do { \
523         if (rt == __net->ipv6.ip6_null_entry) { \
524                 struct fib6_node *pn; \
525                 while (1) { \
526                         if (fn->fn_flags & RTN_TL_ROOT) \
527                                 goto out; \
528                         pn = fn->parent; \
529                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
530                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
531                         else \
532                                 fn = pn; \
533                         if (fn->fn_flags & RTN_RTINFO) \
534                                 goto restart; \
535                 } \
536         } \
537 } while(0)
538
539 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
540                                              struct fib6_table *table,
541                                              struct flowi *fl, int flags)
542 {
543         struct fib6_node *fn;
544         struct rt6_info *rt;
545
546         read_lock_bh(&table->tb6_lock);
547         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
548 restart:
549         rt = fn->leaf;
550         rt = rt6_device_match(net, rt, fl->oif, flags);
551         BACKTRACK(net, &fl->fl6_src);
552 out:
553         dst_use(&rt->u.dst, jiffies);
554         read_unlock_bh(&table->tb6_lock);
555         return rt;
556
557 }
558
559 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
560                             const struct in6_addr *saddr, int oif, int strict)
561 {
562         struct flowi fl = {
563                 .oif = oif,
564                 .nl_u = {
565                         .ip6_u = {
566                                 .daddr = *daddr,
567                         },
568                 },
569         };
570         struct dst_entry *dst;
571         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
572
573         if (saddr) {
574                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
575                 flags |= RT6_LOOKUP_F_HAS_SADDR;
576         }
577
578         dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
579         if (dst->error == 0)
580                 return (struct rt6_info *) dst;
581
582         dst_release(dst);
583
584         return NULL;
585 }
586
587 EXPORT_SYMBOL(rt6_lookup);
588
589 /* ip6_ins_rt is called with FREE table->tb6_lock.
590    It takes new route entry, the addition fails by any reason the
591    route is freed. In any case, if caller does not hold it, it may
592    be destroyed.
593  */
594
595 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
596 {
597         int err;
598         struct fib6_table *table;
599
600         table = rt->rt6i_table;
601         write_lock_bh(&table->tb6_lock);
602         err = fib6_add(&table->tb6_root, rt, info);
603         write_unlock_bh(&table->tb6_lock);
604
605         return err;
606 }
607
608 int ip6_ins_rt(struct rt6_info *rt)
609 {
610         struct nl_info info = {
611                 .nl_net = dev_net(rt->rt6i_dev),
612         };
613         return __ip6_ins_rt(rt, &info);
614 }
615
616 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
617                                       struct in6_addr *saddr)
618 {
619         struct rt6_info *rt;
620
621         /*
622          *      Clone the route.
623          */
624
625         rt = ip6_rt_copy(ort);
626
627         if (rt) {
628                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
629                         if (rt->rt6i_dst.plen != 128 &&
630                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
631                                 rt->rt6i_flags |= RTF_ANYCAST;
632                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
633                 }
634
635                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
636                 rt->rt6i_dst.plen = 128;
637                 rt->rt6i_flags |= RTF_CACHE;
638                 rt->u.dst.flags |= DST_HOST;
639
640 #ifdef CONFIG_IPV6_SUBTREES
641                 if (rt->rt6i_src.plen && saddr) {
642                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
643                         rt->rt6i_src.plen = 128;
644                 }
645 #endif
646
647                 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
648
649         }
650
651         return rt;
652 }
653
654 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
655 {
656         struct rt6_info *rt = ip6_rt_copy(ort);
657         if (rt) {
658                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
659                 rt->rt6i_dst.plen = 128;
660                 rt->rt6i_flags |= RTF_CACHE;
661                 rt->u.dst.flags |= DST_HOST;
662                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
663         }
664         return rt;
665 }
666
667 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
668                                       struct flowi *fl, int flags)
669 {
670         struct fib6_node *fn;
671         struct rt6_info *rt, *nrt;
672         int strict = 0;
673         int attempts = 3;
674         int err;
675         int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
676
677         strict |= flags & RT6_LOOKUP_F_IFACE;
678
679 relookup:
680         read_lock_bh(&table->tb6_lock);
681
682 restart_2:
683         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
684
685 restart:
686         rt = rt6_select(fn, oif, strict | reachable);
687
688         BACKTRACK(net, &fl->fl6_src);
689         if (rt == net->ipv6.ip6_null_entry ||
690             rt->rt6i_flags & RTF_CACHE)
691                 goto out;
692
693         dst_hold(&rt->u.dst);
694         read_unlock_bh(&table->tb6_lock);
695
696         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
697                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
698         else {
699 #if CLONE_OFFLINK_ROUTE
700                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
701 #else
702                 goto out2;
703 #endif
704         }
705
706         dst_release(&rt->u.dst);
707         rt = nrt ? : net->ipv6.ip6_null_entry;
708
709         dst_hold(&rt->u.dst);
710         if (nrt) {
711                 err = ip6_ins_rt(nrt);
712                 if (!err)
713                         goto out2;
714         }
715
716         if (--attempts <= 0)
717                 goto out2;
718
719         /*
720          * Race condition! In the gap, when table->tb6_lock was
721          * released someone could insert this route.  Relookup.
722          */
723         dst_release(&rt->u.dst);
724         goto relookup;
725
726 out:
727         if (reachable) {
728                 reachable = 0;
729                 goto restart_2;
730         }
731         dst_hold(&rt->u.dst);
732         read_unlock_bh(&table->tb6_lock);
733 out2:
734         rt->u.dst.lastuse = jiffies;
735         rt->u.dst.__use++;
736
737         return rt;
738 }
739
740 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
741                                             struct flowi *fl, int flags)
742 {
743         return ip6_pol_route(net, table, fl->iif, fl, flags);
744 }
745
746 void ip6_route_input(struct sk_buff *skb)
747 {
748         struct ipv6hdr *iph = ipv6_hdr(skb);
749         struct net *net = dev_net(skb->dev);
750         int flags = RT6_LOOKUP_F_HAS_SADDR;
751         struct flowi fl = {
752                 .iif = skb->dev->ifindex,
753                 .nl_u = {
754                         .ip6_u = {
755                                 .daddr = iph->daddr,
756                                 .saddr = iph->saddr,
757                                 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
758                         },
759                 },
760                 .mark = skb->mark,
761                 .proto = iph->nexthdr,
762         };
763
764         if (rt6_need_strict(&iph->daddr))
765                 flags |= RT6_LOOKUP_F_IFACE;
766
767         skb->dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input);
768 }
769
770 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
771                                              struct flowi *fl, int flags)
772 {
773         return ip6_pol_route(net, table, fl->oif, fl, flags);
774 }
775
776 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
777                                     struct flowi *fl)
778 {
779         int flags = 0;
780
781         if (rt6_need_strict(&fl->fl6_dst))
782                 flags |= RT6_LOOKUP_F_IFACE;
783
784         if (!ipv6_addr_any(&fl->fl6_src))
785                 flags |= RT6_LOOKUP_F_HAS_SADDR;
786         else if (sk) {
787                 unsigned int prefs = inet6_sk(sk)->srcprefs;
788                 if (prefs & IPV6_PREFER_SRC_TMP)
789                         flags |= RT6_LOOKUP_F_SRCPREF_TMP;
790                 if (prefs & IPV6_PREFER_SRC_PUBLIC)
791                         flags |= RT6_LOOKUP_F_SRCPREF_PUBLIC;
792                 if (prefs & IPV6_PREFER_SRC_COA)
793                         flags |= RT6_LOOKUP_F_SRCPREF_COA;
794         }
795
796         return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
797 }
798
799 EXPORT_SYMBOL(ip6_route_output);
800
801 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
802 {
803         struct rt6_info *ort = (struct rt6_info *) *dstp;
804         struct rt6_info *rt = (struct rt6_info *)
805                 dst_alloc(&ip6_dst_blackhole_ops);
806         struct dst_entry *new = NULL;
807
808         if (rt) {
809                 new = &rt->u.dst;
810
811                 atomic_set(&new->__refcnt, 1);
812                 new->__use = 1;
813                 new->input = dst_discard;
814                 new->output = dst_discard;
815
816                 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
817                 new->dev = ort->u.dst.dev;
818                 if (new->dev)
819                         dev_hold(new->dev);
820                 rt->rt6i_idev = ort->rt6i_idev;
821                 if (rt->rt6i_idev)
822                         in6_dev_hold(rt->rt6i_idev);
823                 rt->rt6i_expires = 0;
824
825                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
826                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
827                 rt->rt6i_metric = 0;
828
829                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
830 #ifdef CONFIG_IPV6_SUBTREES
831                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
832 #endif
833
834                 dst_free(new);
835         }
836
837         dst_release(*dstp);
838         *dstp = new;
839         return (new ? 0 : -ENOMEM);
840 }
841 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
842
843 /*
844  *      Destination cache support functions
845  */
846
847 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
848 {
849         struct rt6_info *rt;
850
851         rt = (struct rt6_info *) dst;
852
853         if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
854                 return dst;
855
856         return NULL;
857 }
858
859 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
860 {
861         struct rt6_info *rt = (struct rt6_info *) dst;
862
863         if (rt) {
864                 if (rt->rt6i_flags & RTF_CACHE)
865                         ip6_del_rt(rt);
866                 else
867                         dst_release(dst);
868         }
869         return NULL;
870 }
871
872 static void ip6_link_failure(struct sk_buff *skb)
873 {
874         struct rt6_info *rt;
875
876         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
877
878         rt = (struct rt6_info *) skb->dst;
879         if (rt) {
880                 if (rt->rt6i_flags&RTF_CACHE) {
881                         dst_set_expires(&rt->u.dst, 0);
882                         rt->rt6i_flags |= RTF_EXPIRES;
883                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
884                         rt->rt6i_node->fn_sernum = -1;
885         }
886 }
887
888 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
889 {
890         struct rt6_info *rt6 = (struct rt6_info*)dst;
891
892         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
893                 rt6->rt6i_flags |= RTF_MODIFIED;
894                 if (mtu < IPV6_MIN_MTU) {
895                         mtu = IPV6_MIN_MTU;
896                         dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
897                 }
898                 dst->metrics[RTAX_MTU-1] = mtu;
899                 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
900         }
901 }
902
903 static int ipv6_get_mtu(struct net_device *dev);
904
905 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
906 {
907         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
908
909         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
910                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
911
912         /*
913          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
914          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
915          * IPV6_MAXPLEN is also valid and means: "any MSS,
916          * rely only on pmtu discovery"
917          */
918         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
919                 mtu = IPV6_MAXPLEN;
920         return mtu;
921 }
922
923 static struct dst_entry *icmp6_dst_gc_list;
924 static DEFINE_SPINLOCK(icmp6_dst_lock);
925
926 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
927                                   struct neighbour *neigh,
928                                   const struct in6_addr *addr)
929 {
930         struct rt6_info *rt;
931         struct inet6_dev *idev = in6_dev_get(dev);
932         struct net *net = dev_net(dev);
933
934         if (unlikely(idev == NULL))
935                 return NULL;
936
937         rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
938         if (unlikely(rt == NULL)) {
939                 in6_dev_put(idev);
940                 goto out;
941         }
942
943         dev_hold(dev);
944         if (neigh)
945                 neigh_hold(neigh);
946         else
947                 neigh = ndisc_get_neigh(dev, addr);
948
949         rt->rt6i_dev      = dev;
950         rt->rt6i_idev     = idev;
951         rt->rt6i_nexthop  = neigh;
952         atomic_set(&rt->u.dst.__refcnt, 1);
953         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
954         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
955         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
956         rt->u.dst.output  = ip6_output;
957
958 #if 0   /* there's no chance to use these for ndisc */
959         rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
960                                 ? DST_HOST
961                                 : 0;
962         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
963         rt->rt6i_dst.plen = 128;
964 #endif
965
966         spin_lock_bh(&icmp6_dst_lock);
967         rt->u.dst.next = icmp6_dst_gc_list;
968         icmp6_dst_gc_list = &rt->u.dst;
969         spin_unlock_bh(&icmp6_dst_lock);
970
971         fib6_force_start_gc(net);
972
973 out:
974         return &rt->u.dst;
975 }
976
977 int icmp6_dst_gc(int *more)
978 {
979         struct dst_entry *dst, *next, **pprev;
980         int freed;
981
982         next = NULL;
983         freed = 0;
984
985         spin_lock_bh(&icmp6_dst_lock);
986         pprev = &icmp6_dst_gc_list;
987
988         while ((dst = *pprev) != NULL) {
989                 if (!atomic_read(&dst->__refcnt)) {
990                         *pprev = dst->next;
991                         dst_free(dst);
992                         freed++;
993                 } else {
994                         pprev = &dst->next;
995                         (*more)++;
996                 }
997         }
998
999         spin_unlock_bh(&icmp6_dst_lock);
1000
1001         return freed;
1002 }
1003
1004 static int ip6_dst_gc(struct dst_ops *ops)
1005 {
1006         unsigned long now = jiffies;
1007         struct net *net = ops->dst_net;
1008         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1009         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1010         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1011         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1012         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1013
1014         if (time_after(rt_last_gc + rt_min_interval, now) &&
1015             atomic_read(&ops->entries) <= rt_max_size)
1016                 goto out;
1017
1018         net->ipv6.ip6_rt_gc_expire++;
1019         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1020         net->ipv6.ip6_rt_last_gc = now;
1021         if (atomic_read(&ops->entries) < ops->gc_thresh)
1022                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1023 out:
1024         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1025         return (atomic_read(&ops->entries) > rt_max_size);
1026 }
1027
1028 /* Clean host part of a prefix. Not necessary in radix tree,
1029    but results in cleaner routing tables.
1030
1031    Remove it only when all the things will work!
1032  */
1033
1034 static int ipv6_get_mtu(struct net_device *dev)
1035 {
1036         int mtu = IPV6_MIN_MTU;
1037         struct inet6_dev *idev;
1038
1039         idev = in6_dev_get(dev);
1040         if (idev) {
1041                 mtu = idev->cnf.mtu6;
1042                 in6_dev_put(idev);
1043         }
1044         return mtu;
1045 }
1046
1047 int ip6_dst_hoplimit(struct dst_entry *dst)
1048 {
1049         int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1050         if (hoplimit < 0) {
1051                 struct net_device *dev = dst->dev;
1052                 struct inet6_dev *idev = in6_dev_get(dev);
1053                 if (idev) {
1054                         hoplimit = idev->cnf.hop_limit;
1055                         in6_dev_put(idev);
1056                 } else
1057                         hoplimit = ipv6_devconf.hop_limit;
1058         }
1059         return hoplimit;
1060 }
1061
1062 /*
1063  *
1064  */
1065
1066 int ip6_route_add(struct fib6_config *cfg)
1067 {
1068         int err;
1069         struct net *net = cfg->fc_nlinfo.nl_net;
1070         struct rt6_info *rt = NULL;
1071         struct net_device *dev = NULL;
1072         struct inet6_dev *idev = NULL;
1073         struct fib6_table *table;
1074         int addr_type;
1075
1076         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1077                 return -EINVAL;
1078 #ifndef CONFIG_IPV6_SUBTREES
1079         if (cfg->fc_src_len)
1080                 return -EINVAL;
1081 #endif
1082         if (cfg->fc_ifindex) {
1083                 err = -ENODEV;
1084                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1085                 if (!dev)
1086                         goto out;
1087                 idev = in6_dev_get(dev);
1088                 if (!idev)
1089                         goto out;
1090         }
1091
1092         if (cfg->fc_metric == 0)
1093                 cfg->fc_metric = IP6_RT_PRIO_USER;
1094
1095         table = fib6_new_table(net, cfg->fc_table);
1096         if (table == NULL) {
1097                 err = -ENOBUFS;
1098                 goto out;
1099         }
1100
1101         rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1102
1103         if (rt == NULL) {
1104                 err = -ENOMEM;
1105                 goto out;
1106         }
1107
1108         rt->u.dst.obsolete = -1;
1109         rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1110                                 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1111                                 0;
1112
1113         if (cfg->fc_protocol == RTPROT_UNSPEC)
1114                 cfg->fc_protocol = RTPROT_BOOT;
1115         rt->rt6i_protocol = cfg->fc_protocol;
1116
1117         addr_type = ipv6_addr_type(&cfg->fc_dst);
1118
1119         if (addr_type & IPV6_ADDR_MULTICAST)
1120                 rt->u.dst.input = ip6_mc_input;
1121         else
1122                 rt->u.dst.input = ip6_forward;
1123
1124         rt->u.dst.output = ip6_output;
1125
1126         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1127         rt->rt6i_dst.plen = cfg->fc_dst_len;
1128         if (rt->rt6i_dst.plen == 128)
1129                rt->u.dst.flags = DST_HOST;
1130
1131 #ifdef CONFIG_IPV6_SUBTREES
1132         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1133         rt->rt6i_src.plen = cfg->fc_src_len;
1134 #endif
1135
1136         rt->rt6i_metric = cfg->fc_metric;
1137
1138         /* We cannot add true routes via loopback here,
1139            they would result in kernel looping; promote them to reject routes
1140          */
1141         if ((cfg->fc_flags & RTF_REJECT) ||
1142             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1143                 /* hold loopback dev/idev if we haven't done so. */
1144                 if (dev != net->loopback_dev) {
1145                         if (dev) {
1146                                 dev_put(dev);
1147                                 in6_dev_put(idev);
1148                         }
1149                         dev = net->loopback_dev;
1150                         dev_hold(dev);
1151                         idev = in6_dev_get(dev);
1152                         if (!idev) {
1153                                 err = -ENODEV;
1154                                 goto out;
1155                         }
1156                 }
1157                 rt->u.dst.output = ip6_pkt_discard_out;
1158                 rt->u.dst.input = ip6_pkt_discard;
1159                 rt->u.dst.error = -ENETUNREACH;
1160                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1161                 goto install_route;
1162         }
1163
1164         if (cfg->fc_flags & RTF_GATEWAY) {
1165                 struct in6_addr *gw_addr;
1166                 int gwa_type;
1167
1168                 gw_addr = &cfg->fc_gateway;
1169                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1170                 gwa_type = ipv6_addr_type(gw_addr);
1171
1172                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1173                         struct rt6_info *grt;
1174
1175                         /* IPv6 strictly inhibits using not link-local
1176                            addresses as nexthop address.
1177                            Otherwise, router will not able to send redirects.
1178                            It is very good, but in some (rare!) circumstances
1179                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1180                            some exceptions. --ANK
1181                          */
1182                         err = -EINVAL;
1183                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1184                                 goto out;
1185
1186                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1187
1188                         err = -EHOSTUNREACH;
1189                         if (grt == NULL)
1190                                 goto out;
1191                         if (dev) {
1192                                 if (dev != grt->rt6i_dev) {
1193                                         dst_release(&grt->u.dst);
1194                                         goto out;
1195                                 }
1196                         } else {
1197                                 dev = grt->rt6i_dev;
1198                                 idev = grt->rt6i_idev;
1199                                 dev_hold(dev);
1200                                 in6_dev_hold(grt->rt6i_idev);
1201                         }
1202                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1203                                 err = 0;
1204                         dst_release(&grt->u.dst);
1205
1206                         if (err)
1207                                 goto out;
1208                 }
1209                 err = -EINVAL;
1210                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1211                         goto out;
1212         }
1213
1214         err = -ENODEV;
1215         if (dev == NULL)
1216                 goto out;
1217
1218         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1219                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1220                 if (IS_ERR(rt->rt6i_nexthop)) {
1221                         err = PTR_ERR(rt->rt6i_nexthop);
1222                         rt->rt6i_nexthop = NULL;
1223                         goto out;
1224                 }
1225         }
1226
1227         rt->rt6i_flags = cfg->fc_flags;
1228
1229 install_route:
1230         if (cfg->fc_mx) {
1231                 struct nlattr *nla;
1232                 int remaining;
1233
1234                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1235                         int type = nla_type(nla);
1236
1237                         if (type) {
1238                                 if (type > RTAX_MAX) {
1239                                         err = -EINVAL;
1240                                         goto out;
1241                                 }
1242
1243                                 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1244                         }
1245                 }
1246         }
1247
1248         if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0)
1249                 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1250         if (!dst_metric(&rt->u.dst, RTAX_MTU))
1251                 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1252         if (!dst_metric(&rt->u.dst, RTAX_ADVMSS))
1253                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1254         rt->u.dst.dev = dev;
1255         rt->rt6i_idev = idev;
1256         rt->rt6i_table = table;
1257
1258         cfg->fc_nlinfo.nl_net = dev_net(dev);
1259
1260         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1261
1262 out:
1263         if (dev)
1264                 dev_put(dev);
1265         if (idev)
1266                 in6_dev_put(idev);
1267         if (rt)
1268                 dst_free(&rt->u.dst);
1269         return err;
1270 }
1271
1272 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1273 {
1274         int err;
1275         struct fib6_table *table;
1276         struct net *net = dev_net(rt->rt6i_dev);
1277
1278         if (rt == net->ipv6.ip6_null_entry)
1279                 return -ENOENT;
1280
1281         table = rt->rt6i_table;
1282         write_lock_bh(&table->tb6_lock);
1283
1284         err = fib6_del(rt, info);
1285         dst_release(&rt->u.dst);
1286
1287         write_unlock_bh(&table->tb6_lock);
1288
1289         return err;
1290 }
1291
1292 int ip6_del_rt(struct rt6_info *rt)
1293 {
1294         struct nl_info info = {
1295                 .nl_net = dev_net(rt->rt6i_dev),
1296         };
1297         return __ip6_del_rt(rt, &info);
1298 }
1299
1300 static int ip6_route_del(struct fib6_config *cfg)
1301 {
1302         struct fib6_table *table;
1303         struct fib6_node *fn;
1304         struct rt6_info *rt;
1305         int err = -ESRCH;
1306
1307         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1308         if (table == NULL)
1309                 return err;
1310
1311         read_lock_bh(&table->tb6_lock);
1312
1313         fn = fib6_locate(&table->tb6_root,
1314                          &cfg->fc_dst, cfg->fc_dst_len,
1315                          &cfg->fc_src, cfg->fc_src_len);
1316
1317         if (fn) {
1318                 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1319                         if (cfg->fc_ifindex &&
1320                             (rt->rt6i_dev == NULL ||
1321                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1322                                 continue;
1323                         if (cfg->fc_flags & RTF_GATEWAY &&
1324                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1325                                 continue;
1326                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1327                                 continue;
1328                         dst_hold(&rt->u.dst);
1329                         read_unlock_bh(&table->tb6_lock);
1330
1331                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1332                 }
1333         }
1334         read_unlock_bh(&table->tb6_lock);
1335
1336         return err;
1337 }
1338
1339 /*
1340  *      Handle redirects
1341  */
1342 struct ip6rd_flowi {
1343         struct flowi fl;
1344         struct in6_addr gateway;
1345 };
1346
1347 static struct rt6_info *__ip6_route_redirect(struct net *net,
1348                                              struct fib6_table *table,
1349                                              struct flowi *fl,
1350                                              int flags)
1351 {
1352         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1353         struct rt6_info *rt;
1354         struct fib6_node *fn;
1355
1356         /*
1357          * Get the "current" route for this destination and
1358          * check if the redirect has come from approriate router.
1359          *
1360          * RFC 2461 specifies that redirects should only be
1361          * accepted if they come from the nexthop to the target.
1362          * Due to the way the routes are chosen, this notion
1363          * is a bit fuzzy and one might need to check all possible
1364          * routes.
1365          */
1366
1367         read_lock_bh(&table->tb6_lock);
1368         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1369 restart:
1370         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1371                 /*
1372                  * Current route is on-link; redirect is always invalid.
1373                  *
1374                  * Seems, previous statement is not true. It could
1375                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1376                  * But then router serving it might decide, that we should
1377                  * know truth 8)8) --ANK (980726).
1378                  */
1379                 if (rt6_check_expired(rt))
1380                         continue;
1381                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1382                         continue;
1383                 if (fl->oif != rt->rt6i_dev->ifindex)
1384                         continue;
1385                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1386                         continue;
1387                 break;
1388         }
1389
1390         if (!rt)
1391                 rt = net->ipv6.ip6_null_entry;
1392         BACKTRACK(net, &fl->fl6_src);
1393 out:
1394         dst_hold(&rt->u.dst);
1395
1396         read_unlock_bh(&table->tb6_lock);
1397
1398         return rt;
1399 };
1400
1401 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1402                                            struct in6_addr *src,
1403                                            struct in6_addr *gateway,
1404                                            struct net_device *dev)
1405 {
1406         int flags = RT6_LOOKUP_F_HAS_SADDR;
1407         struct net *net = dev_net(dev);
1408         struct ip6rd_flowi rdfl = {
1409                 .fl = {
1410                         .oif = dev->ifindex,
1411                         .nl_u = {
1412                                 .ip6_u = {
1413                                         .daddr = *dest,
1414                                         .saddr = *src,
1415                                 },
1416                         },
1417                 },
1418                 .gateway = *gateway,
1419         };
1420
1421         if (rt6_need_strict(dest))
1422                 flags |= RT6_LOOKUP_F_IFACE;
1423
1424         return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1425                                                    flags, __ip6_route_redirect);
1426 }
1427
1428 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1429                   struct in6_addr *saddr,
1430                   struct neighbour *neigh, u8 *lladdr, int on_link)
1431 {
1432         struct rt6_info *rt, *nrt = NULL;
1433         struct netevent_redirect netevent;
1434         struct net *net = dev_net(neigh->dev);
1435
1436         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1437
1438         if (rt == net->ipv6.ip6_null_entry) {
1439                 if (net_ratelimit())
1440                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1441                                "for redirect target\n");
1442                 goto out;
1443         }
1444
1445         /*
1446          *      We have finally decided to accept it.
1447          */
1448
1449         neigh_update(neigh, lladdr, NUD_STALE,
1450                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1451                      NEIGH_UPDATE_F_OVERRIDE|
1452                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1453                                      NEIGH_UPDATE_F_ISROUTER))
1454                      );
1455
1456         /*
1457          * Redirect received -> path was valid.
1458          * Look, redirects are sent only in response to data packets,
1459          * so that this nexthop apparently is reachable. --ANK
1460          */
1461         dst_confirm(&rt->u.dst);
1462
1463         /* Duplicate redirect: silently ignore. */
1464         if (neigh == rt->u.dst.neighbour)
1465                 goto out;
1466
1467         nrt = ip6_rt_copy(rt);
1468         if (nrt == NULL)
1469                 goto out;
1470
1471         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1472         if (on_link)
1473                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1474
1475         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1476         nrt->rt6i_dst.plen = 128;
1477         nrt->u.dst.flags |= DST_HOST;
1478
1479         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1480         nrt->rt6i_nexthop = neigh_clone(neigh);
1481         /* Reset pmtu, it may be better */
1482         nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1483         nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1484                                                         dst_mtu(&nrt->u.dst));
1485
1486         if (ip6_ins_rt(nrt))
1487                 goto out;
1488
1489         netevent.old = &rt->u.dst;
1490         netevent.new = &nrt->u.dst;
1491         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1492
1493         if (rt->rt6i_flags&RTF_CACHE) {
1494                 ip6_del_rt(rt);
1495                 return;
1496         }
1497
1498 out:
1499         dst_release(&rt->u.dst);
1500         return;
1501 }
1502
1503 /*
1504  *      Handle ICMP "packet too big" messages
1505  *      i.e. Path MTU discovery
1506  */
1507
1508 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1509                         struct net_device *dev, u32 pmtu)
1510 {
1511         struct rt6_info *rt, *nrt;
1512         struct net *net = dev_net(dev);
1513         int allfrag = 0;
1514
1515         rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1516         if (rt == NULL)
1517                 return;
1518
1519         if (pmtu >= dst_mtu(&rt->u.dst))
1520                 goto out;
1521
1522         if (pmtu < IPV6_MIN_MTU) {
1523                 /*
1524                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1525                  * MTU (1280) and a fragment header should always be included
1526                  * after a node receiving Too Big message reporting PMTU is
1527                  * less than the IPv6 Minimum Link MTU.
1528                  */
1529                 pmtu = IPV6_MIN_MTU;
1530                 allfrag = 1;
1531         }
1532
1533         /* New mtu received -> path was valid.
1534            They are sent only in response to data packets,
1535            so that this nexthop apparently is reachable. --ANK
1536          */
1537         dst_confirm(&rt->u.dst);
1538
1539         /* Host route. If it is static, it would be better
1540            not to override it, but add new one, so that
1541            when cache entry will expire old pmtu
1542            would return automatically.
1543          */
1544         if (rt->rt6i_flags & RTF_CACHE) {
1545                 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1546                 if (allfrag)
1547                         rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1548                 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1549                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1550                 goto out;
1551         }
1552
1553         /* Network route.
1554            Two cases are possible:
1555            1. It is connected route. Action: COW
1556            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1557          */
1558         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1559                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1560         else
1561                 nrt = rt6_alloc_clone(rt, daddr);
1562
1563         if (nrt) {
1564                 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1565                 if (allfrag)
1566                         nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1567
1568                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1569                  * happened within 5 mins, the recommended timer is 10 mins.
1570                  * Here this route expiration time is set to ip6_rt_mtu_expires
1571                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1572                  * and detecting PMTU increase will be automatically happened.
1573                  */
1574                 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1575                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1576
1577                 ip6_ins_rt(nrt);
1578         }
1579 out:
1580         dst_release(&rt->u.dst);
1581 }
1582
1583 /*
1584  *      Misc support functions
1585  */
1586
1587 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1588 {
1589         struct net *net = dev_net(ort->rt6i_dev);
1590         struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1591
1592         if (rt) {
1593                 rt->u.dst.input = ort->u.dst.input;
1594                 rt->u.dst.output = ort->u.dst.output;
1595
1596                 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1597                 rt->u.dst.error = ort->u.dst.error;
1598                 rt->u.dst.dev = ort->u.dst.dev;
1599                 if (rt->u.dst.dev)
1600                         dev_hold(rt->u.dst.dev);
1601                 rt->rt6i_idev = ort->rt6i_idev;
1602                 if (rt->rt6i_idev)
1603                         in6_dev_hold(rt->rt6i_idev);
1604                 rt->u.dst.lastuse = jiffies;
1605                 rt->rt6i_expires = 0;
1606
1607                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1608                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1609                 rt->rt6i_metric = 0;
1610
1611                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1612 #ifdef CONFIG_IPV6_SUBTREES
1613                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1614 #endif
1615                 rt->rt6i_table = ort->rt6i_table;
1616         }
1617         return rt;
1618 }
1619
1620 #ifdef CONFIG_IPV6_ROUTE_INFO
1621 static struct rt6_info *rt6_get_route_info(struct net *net,
1622                                            struct in6_addr *prefix, int prefixlen,
1623                                            struct in6_addr *gwaddr, int ifindex)
1624 {
1625         struct fib6_node *fn;
1626         struct rt6_info *rt = NULL;
1627         struct fib6_table *table;
1628
1629         table = fib6_get_table(net, RT6_TABLE_INFO);
1630         if (table == NULL)
1631                 return NULL;
1632
1633         write_lock_bh(&table->tb6_lock);
1634         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1635         if (!fn)
1636                 goto out;
1637
1638         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1639                 if (rt->rt6i_dev->ifindex != ifindex)
1640                         continue;
1641                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1642                         continue;
1643                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1644                         continue;
1645                 dst_hold(&rt->u.dst);
1646                 break;
1647         }
1648 out:
1649         write_unlock_bh(&table->tb6_lock);
1650         return rt;
1651 }
1652
1653 static struct rt6_info *rt6_add_route_info(struct net *net,
1654                                            struct in6_addr *prefix, int prefixlen,
1655                                            struct in6_addr *gwaddr, int ifindex,
1656                                            unsigned pref)
1657 {
1658         struct fib6_config cfg = {
1659                 .fc_table       = RT6_TABLE_INFO,
1660                 .fc_metric      = IP6_RT_PRIO_USER,
1661                 .fc_ifindex     = ifindex,
1662                 .fc_dst_len     = prefixlen,
1663                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1664                                   RTF_UP | RTF_PREF(pref),
1665                 .fc_nlinfo.pid = 0,
1666                 .fc_nlinfo.nlh = NULL,
1667                 .fc_nlinfo.nl_net = net,
1668         };
1669
1670         ipv6_addr_copy(&cfg.fc_dst, prefix);
1671         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1672
1673         /* We should treat it as a default route if prefix length is 0. */
1674         if (!prefixlen)
1675                 cfg.fc_flags |= RTF_DEFAULT;
1676
1677         ip6_route_add(&cfg);
1678
1679         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1680 }
1681 #endif
1682
1683 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1684 {
1685         struct rt6_info *rt;
1686         struct fib6_table *table;
1687
1688         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1689         if (table == NULL)
1690                 return NULL;
1691
1692         write_lock_bh(&table->tb6_lock);
1693         for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1694                 if (dev == rt->rt6i_dev &&
1695                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1696                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1697                         break;
1698         }
1699         if (rt)
1700                 dst_hold(&rt->u.dst);
1701         write_unlock_bh(&table->tb6_lock);
1702         return rt;
1703 }
1704
1705 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1706                                      struct net_device *dev,
1707                                      unsigned int pref)
1708 {
1709         struct fib6_config cfg = {
1710                 .fc_table       = RT6_TABLE_DFLT,
1711                 .fc_metric      = IP6_RT_PRIO_USER,
1712                 .fc_ifindex     = dev->ifindex,
1713                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1714                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1715                 .fc_nlinfo.pid = 0,
1716                 .fc_nlinfo.nlh = NULL,
1717                 .fc_nlinfo.nl_net = dev_net(dev),
1718         };
1719
1720         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1721
1722         ip6_route_add(&cfg);
1723
1724         return rt6_get_dflt_router(gwaddr, dev);
1725 }
1726
1727 void rt6_purge_dflt_routers(struct net *net)
1728 {
1729         struct rt6_info *rt;
1730         struct fib6_table *table;
1731
1732         /* NOTE: Keep consistent with rt6_get_dflt_router */
1733         table = fib6_get_table(net, RT6_TABLE_DFLT);
1734         if (table == NULL)
1735                 return;
1736
1737 restart:
1738         read_lock_bh(&table->tb6_lock);
1739         for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1740                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1741                         dst_hold(&rt->u.dst);
1742                         read_unlock_bh(&table->tb6_lock);
1743                         ip6_del_rt(rt);
1744                         goto restart;
1745                 }
1746         }
1747         read_unlock_bh(&table->tb6_lock);
1748 }
1749
1750 static void rtmsg_to_fib6_config(struct net *net,
1751                                  struct in6_rtmsg *rtmsg,
1752                                  struct fib6_config *cfg)
1753 {
1754         memset(cfg, 0, sizeof(*cfg));
1755
1756         cfg->fc_table = RT6_TABLE_MAIN;
1757         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1758         cfg->fc_metric = rtmsg->rtmsg_metric;
1759         cfg->fc_expires = rtmsg->rtmsg_info;
1760         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1761         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1762         cfg->fc_flags = rtmsg->rtmsg_flags;
1763
1764         cfg->fc_nlinfo.nl_net = net;
1765
1766         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1767         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1768         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1769 }
1770
1771 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1772 {
1773         struct fib6_config cfg;
1774         struct in6_rtmsg rtmsg;
1775         int err;
1776
1777         switch(cmd) {
1778         case SIOCADDRT:         /* Add a route */
1779         case SIOCDELRT:         /* Delete a route */
1780                 if (!capable(CAP_NET_ADMIN))
1781                         return -EPERM;
1782                 err = copy_from_user(&rtmsg, arg,
1783                                      sizeof(struct in6_rtmsg));
1784                 if (err)
1785                         return -EFAULT;
1786
1787                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1788
1789                 rtnl_lock();
1790                 switch (cmd) {
1791                 case SIOCADDRT:
1792                         err = ip6_route_add(&cfg);
1793                         break;
1794                 case SIOCDELRT:
1795                         err = ip6_route_del(&cfg);
1796                         break;
1797                 default:
1798                         err = -EINVAL;
1799                 }
1800                 rtnl_unlock();
1801
1802                 return err;
1803         }
1804
1805         return -EINVAL;
1806 }
1807
1808 /*
1809  *      Drop the packet on the floor
1810  */
1811
1812 static int ip6_pkt_drop(struct sk_buff *skb, int code, int ipstats_mib_noroutes)
1813 {
1814         int type;
1815         switch (ipstats_mib_noroutes) {
1816         case IPSTATS_MIB_INNOROUTES:
1817                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1818                 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1819                         IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1820                         break;
1821                 }
1822                 /* FALLTHROUGH */
1823         case IPSTATS_MIB_OUTNOROUTES:
1824                 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1825                 break;
1826         }
1827         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1828         kfree_skb(skb);
1829         return 0;
1830 }
1831
1832 static int ip6_pkt_discard(struct sk_buff *skb)
1833 {
1834         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1835 }
1836
1837 static int ip6_pkt_discard_out(struct sk_buff *skb)
1838 {
1839         skb->dev = skb->dst->dev;
1840         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1841 }
1842
1843 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1844
1845 static int ip6_pkt_prohibit(struct sk_buff *skb)
1846 {
1847         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1848 }
1849
1850 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1851 {
1852         skb->dev = skb->dst->dev;
1853         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1854 }
1855
1856 #endif
1857
1858 /*
1859  *      Allocate a dst for local (unicast / anycast) address.
1860  */
1861
1862 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1863                                     const struct in6_addr *addr,
1864                                     int anycast)
1865 {
1866         struct net *net = dev_net(idev->dev);
1867         struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1868
1869         if (rt == NULL)
1870                 return ERR_PTR(-ENOMEM);
1871
1872         dev_hold(net->loopback_dev);
1873         in6_dev_hold(idev);
1874
1875         rt->u.dst.flags = DST_HOST;
1876         rt->u.dst.input = ip6_input;
1877         rt->u.dst.output = ip6_output;
1878         rt->rt6i_dev = net->loopback_dev;
1879         rt->rt6i_idev = idev;
1880         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1881         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1882         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1883         rt->u.dst.obsolete = -1;
1884
1885         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1886         if (anycast)
1887                 rt->rt6i_flags |= RTF_ANYCAST;
1888         else
1889                 rt->rt6i_flags |= RTF_LOCAL;
1890         rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1891         if (rt->rt6i_nexthop == NULL) {
1892                 dst_free(&rt->u.dst);
1893                 return ERR_PTR(-ENOMEM);
1894         }
1895
1896         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1897         rt->rt6i_dst.plen = 128;
1898         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1899
1900         atomic_set(&rt->u.dst.__refcnt, 1);
1901
1902         return rt;
1903 }
1904
1905 struct arg_dev_net {
1906         struct net_device *dev;
1907         struct net *net;
1908 };
1909
1910 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1911 {
1912         struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1913         struct net *net = ((struct arg_dev_net *)arg)->net;
1914
1915         if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1916             rt != net->ipv6.ip6_null_entry) {
1917                 RT6_TRACE("deleted by ifdown %p\n", rt);
1918                 return -1;
1919         }
1920         return 0;
1921 }
1922
1923 void rt6_ifdown(struct net *net, struct net_device *dev)
1924 {
1925         struct arg_dev_net adn = {
1926                 .dev = dev,
1927                 .net = net,
1928         };
1929
1930         fib6_clean_all(net, fib6_ifdown, 0, &adn);
1931 }
1932
1933 struct rt6_mtu_change_arg
1934 {
1935         struct net_device *dev;
1936         unsigned mtu;
1937 };
1938
1939 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1940 {
1941         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1942         struct inet6_dev *idev;
1943         struct net *net = dev_net(arg->dev);
1944
1945         /* In IPv6 pmtu discovery is not optional,
1946            so that RTAX_MTU lock cannot disable it.
1947            We still use this lock to block changes
1948            caused by addrconf/ndisc.
1949         */
1950
1951         idev = __in6_dev_get(arg->dev);
1952         if (idev == NULL)
1953                 return 0;
1954
1955         /* For administrative MTU increase, there is no way to discover
1956            IPv6 PMTU increase, so PMTU increase should be updated here.
1957            Since RFC 1981 doesn't include administrative MTU increase
1958            update PMTU increase is a MUST. (i.e. jumbo frame)
1959          */
1960         /*
1961            If new MTU is less than route PMTU, this new MTU will be the
1962            lowest MTU in the path, update the route PMTU to reflect PMTU
1963            decreases; if new MTU is greater than route PMTU, and the
1964            old MTU is the lowest MTU in the path, update the route PMTU
1965            to reflect the increase. In this case if the other nodes' MTU
1966            also have the lowest MTU, TOO BIG MESSAGE will be lead to
1967            PMTU discouvery.
1968          */
1969         if (rt->rt6i_dev == arg->dev &&
1970             !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1971             (dst_mtu(&rt->u.dst) >= arg->mtu ||
1972              (dst_mtu(&rt->u.dst) < arg->mtu &&
1973               dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1974                 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1975                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
1976         }
1977         return 0;
1978 }
1979
1980 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1981 {
1982         struct rt6_mtu_change_arg arg = {
1983                 .dev = dev,
1984                 .mtu = mtu,
1985         };
1986
1987         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
1988 }
1989
1990 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1991         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
1992         [RTA_OIF]               = { .type = NLA_U32 },
1993         [RTA_IIF]               = { .type = NLA_U32 },
1994         [RTA_PRIORITY]          = { .type = NLA_U32 },
1995         [RTA_METRICS]           = { .type = NLA_NESTED },
1996 };
1997
1998 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1999                               struct fib6_config *cfg)
2000 {
2001         struct rtmsg *rtm;
2002         struct nlattr *tb[RTA_MAX+1];
2003         int err;
2004
2005         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2006         if (err < 0)
2007                 goto errout;
2008
2009         err = -EINVAL;
2010         rtm = nlmsg_data(nlh);
2011         memset(cfg, 0, sizeof(*cfg));
2012
2013         cfg->fc_table = rtm->rtm_table;
2014         cfg->fc_dst_len = rtm->rtm_dst_len;
2015         cfg->fc_src_len = rtm->rtm_src_len;
2016         cfg->fc_flags = RTF_UP;
2017         cfg->fc_protocol = rtm->rtm_protocol;
2018
2019         if (rtm->rtm_type == RTN_UNREACHABLE)
2020                 cfg->fc_flags |= RTF_REJECT;
2021
2022         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2023         cfg->fc_nlinfo.nlh = nlh;
2024         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2025
2026         if (tb[RTA_GATEWAY]) {
2027                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2028                 cfg->fc_flags |= RTF_GATEWAY;
2029         }
2030
2031         if (tb[RTA_DST]) {
2032                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2033
2034                 if (nla_len(tb[RTA_DST]) < plen)
2035                         goto errout;
2036
2037                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2038         }
2039
2040         if (tb[RTA_SRC]) {
2041                 int plen = (rtm->rtm_src_len + 7) >> 3;
2042
2043                 if (nla_len(tb[RTA_SRC]) < plen)
2044                         goto errout;
2045
2046                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2047         }
2048
2049         if (tb[RTA_OIF])
2050                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2051
2052         if (tb[RTA_PRIORITY])
2053                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2054
2055         if (tb[RTA_METRICS]) {
2056                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2057                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2058         }
2059
2060         if (tb[RTA_TABLE])
2061                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2062
2063         err = 0;
2064 errout:
2065         return err;
2066 }
2067
2068 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2069 {
2070         struct fib6_config cfg;
2071         int err;
2072
2073         err = rtm_to_fib6_config(skb, nlh, &cfg);
2074         if (err < 0)
2075                 return err;
2076
2077         return ip6_route_del(&cfg);
2078 }
2079
2080 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2081 {
2082         struct fib6_config cfg;
2083         int err;
2084
2085         err = rtm_to_fib6_config(skb, nlh, &cfg);
2086         if (err < 0)
2087                 return err;
2088
2089         return ip6_route_add(&cfg);
2090 }
2091
2092 static inline size_t rt6_nlmsg_size(void)
2093 {
2094         return NLMSG_ALIGN(sizeof(struct rtmsg))
2095                + nla_total_size(16) /* RTA_SRC */
2096                + nla_total_size(16) /* RTA_DST */
2097                + nla_total_size(16) /* RTA_GATEWAY */
2098                + nla_total_size(16) /* RTA_PREFSRC */
2099                + nla_total_size(4) /* RTA_TABLE */
2100                + nla_total_size(4) /* RTA_IIF */
2101                + nla_total_size(4) /* RTA_OIF */
2102                + nla_total_size(4) /* RTA_PRIORITY */
2103                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2104                + nla_total_size(sizeof(struct rta_cacheinfo));
2105 }
2106
2107 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2108                          struct in6_addr *dst, struct in6_addr *src,
2109                          int iif, int type, u32 pid, u32 seq,
2110                          int prefix, int nowait, unsigned int flags)
2111 {
2112         struct rtmsg *rtm;
2113         struct nlmsghdr *nlh;
2114         long expires;
2115         u32 table;
2116
2117         if (prefix) {   /* user wants prefix routes only */
2118                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2119                         /* success since this is not a prefix route */
2120                         return 1;
2121                 }
2122         }
2123
2124         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2125         if (nlh == NULL)
2126                 return -EMSGSIZE;
2127
2128         rtm = nlmsg_data(nlh);
2129         rtm->rtm_family = AF_INET6;
2130         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2131         rtm->rtm_src_len = rt->rt6i_src.plen;
2132         rtm->rtm_tos = 0;
2133         if (rt->rt6i_table)
2134                 table = rt->rt6i_table->tb6_id;
2135         else
2136                 table = RT6_TABLE_UNSPEC;
2137         rtm->rtm_table = table;
2138         NLA_PUT_U32(skb, RTA_TABLE, table);
2139         if (rt->rt6i_flags&RTF_REJECT)
2140                 rtm->rtm_type = RTN_UNREACHABLE;
2141         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2142                 rtm->rtm_type = RTN_LOCAL;
2143         else
2144                 rtm->rtm_type = RTN_UNICAST;
2145         rtm->rtm_flags = 0;
2146         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2147         rtm->rtm_protocol = rt->rt6i_protocol;
2148         if (rt->rt6i_flags&RTF_DYNAMIC)
2149                 rtm->rtm_protocol = RTPROT_REDIRECT;
2150         else if (rt->rt6i_flags & RTF_ADDRCONF)
2151                 rtm->rtm_protocol = RTPROT_KERNEL;
2152         else if (rt->rt6i_flags&RTF_DEFAULT)
2153                 rtm->rtm_protocol = RTPROT_RA;
2154
2155         if (rt->rt6i_flags&RTF_CACHE)
2156                 rtm->rtm_flags |= RTM_F_CLONED;
2157
2158         if (dst) {
2159                 NLA_PUT(skb, RTA_DST, 16, dst);
2160                 rtm->rtm_dst_len = 128;
2161         } else if (rtm->rtm_dst_len)
2162                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2163 #ifdef CONFIG_IPV6_SUBTREES
2164         if (src) {
2165                 NLA_PUT(skb, RTA_SRC, 16, src);
2166                 rtm->rtm_src_len = 128;
2167         } else if (rtm->rtm_src_len)
2168                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2169 #endif
2170         if (iif) {
2171 #ifdef CONFIG_IPV6_MROUTE
2172                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2173                         int err = ip6mr_get_route(skb, rtm, nowait);
2174                         if (err <= 0) {
2175                                 if (!nowait) {
2176                                         if (err == 0)
2177                                                 return 0;
2178                                         goto nla_put_failure;
2179                                 } else {
2180                                         if (err == -EMSGSIZE)
2181                                                 goto nla_put_failure;
2182                                 }
2183                         }
2184                 } else
2185 #endif
2186                         NLA_PUT_U32(skb, RTA_IIF, iif);
2187         } else if (dst) {
2188                 struct in6_addr saddr_buf;
2189                 if (ipv6_dev_get_saddr(ip6_dst_idev(&rt->u.dst)->dev,
2190                                        dst, 0, &saddr_buf) == 0)
2191                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2192         }
2193
2194         if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2195                 goto nla_put_failure;
2196
2197         if (rt->u.dst.neighbour)
2198                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2199
2200         if (rt->u.dst.dev)
2201                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2202
2203         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2204
2205         expires = (rt->rt6i_flags & RTF_EXPIRES) ?
2206                         rt->rt6i_expires - jiffies : 0;
2207
2208         if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2209                                expires, rt->u.dst.error) < 0)
2210                 goto nla_put_failure;
2211
2212         return nlmsg_end(skb, nlh);
2213
2214 nla_put_failure:
2215         nlmsg_cancel(skb, nlh);
2216         return -EMSGSIZE;
2217 }
2218
2219 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2220 {
2221         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2222         int prefix;
2223
2224         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2225                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2226                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2227         } else
2228                 prefix = 0;
2229
2230         return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2231                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2232                      prefix, 0, NLM_F_MULTI);
2233 }
2234
2235 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2236 {
2237         struct net *net = sock_net(in_skb->sk);
2238         struct nlattr *tb[RTA_MAX+1];
2239         struct rt6_info *rt;
2240         struct sk_buff *skb;
2241         struct rtmsg *rtm;
2242         struct flowi fl;
2243         int err, iif = 0;
2244
2245         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2246         if (err < 0)
2247                 goto errout;
2248
2249         err = -EINVAL;
2250         memset(&fl, 0, sizeof(fl));
2251
2252         if (tb[RTA_SRC]) {
2253                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2254                         goto errout;
2255
2256                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2257         }
2258
2259         if (tb[RTA_DST]) {
2260                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2261                         goto errout;
2262
2263                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2264         }
2265
2266         if (tb[RTA_IIF])
2267                 iif = nla_get_u32(tb[RTA_IIF]);
2268
2269         if (tb[RTA_OIF])
2270                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2271
2272         if (iif) {
2273                 struct net_device *dev;
2274                 dev = __dev_get_by_index(net, iif);
2275                 if (!dev) {
2276                         err = -ENODEV;
2277                         goto errout;
2278                 }
2279         }
2280
2281         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2282         if (skb == NULL) {
2283                 err = -ENOBUFS;
2284                 goto errout;
2285         }
2286
2287         /* Reserve room for dummy headers, this skb can pass
2288            through good chunk of routing engine.
2289          */
2290         skb_reset_mac_header(skb);
2291         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2292
2293         rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2294         skb->dst = &rt->u.dst;
2295
2296         err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2297                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2298                             nlh->nlmsg_seq, 0, 0, 0);
2299         if (err < 0) {
2300                 kfree_skb(skb);
2301                 goto errout;
2302         }
2303
2304         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2305 errout:
2306         return err;
2307 }
2308
2309 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2310 {
2311         struct sk_buff *skb;
2312         struct net *net = info->nl_net;
2313         u32 seq;
2314         int err;
2315
2316         err = -ENOBUFS;
2317         seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2318
2319         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2320         if (skb == NULL)
2321                 goto errout;
2322
2323         err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2324                                 event, info->pid, seq, 0, 0, 0);
2325         if (err < 0) {
2326                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2327                 WARN_ON(err == -EMSGSIZE);
2328                 kfree_skb(skb);
2329                 goto errout;
2330         }
2331         err = rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2332                           info->nlh, gfp_any());
2333 errout:
2334         if (err < 0)
2335                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2336 }
2337
2338 static int ip6_route_dev_notify(struct notifier_block *this,
2339                                 unsigned long event, void *data)
2340 {
2341         struct net_device *dev = (struct net_device *)data;
2342         struct net *net = dev_net(dev);
2343
2344         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2345                 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2346                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2347 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2348                 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2349                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2350                 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2351                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2352 #endif
2353         }
2354
2355         return NOTIFY_OK;
2356 }
2357
2358 /*
2359  *      /proc
2360  */
2361
2362 #ifdef CONFIG_PROC_FS
2363
2364 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2365
2366 struct rt6_proc_arg
2367 {
2368         char *buffer;
2369         int offset;
2370         int length;
2371         int skip;
2372         int len;
2373 };
2374
2375 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2376 {
2377         struct seq_file *m = p_arg;
2378
2379         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2380                    rt->rt6i_dst.plen);
2381
2382 #ifdef CONFIG_IPV6_SUBTREES
2383         seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2384                    rt->rt6i_src.plen);
2385 #else
2386         seq_puts(m, "00000000000000000000000000000000 00 ");
2387 #endif
2388
2389         if (rt->rt6i_nexthop) {
2390                 seq_printf(m, NIP6_SEQFMT,
2391                            NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2392         } else {
2393                 seq_puts(m, "00000000000000000000000000000000");
2394         }
2395         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2396                    rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2397                    rt->u.dst.__use, rt->rt6i_flags,
2398                    rt->rt6i_dev ? rt->rt6i_dev->name : "");
2399         return 0;
2400 }
2401
2402 static int ipv6_route_show(struct seq_file *m, void *v)
2403 {
2404         struct net *net = (struct net *)m->private;
2405         fib6_clean_all(net, rt6_info_route, 0, m);
2406         return 0;
2407 }
2408
2409 static int ipv6_route_open(struct inode *inode, struct file *file)
2410 {
2411         int err;
2412         struct net *net = get_proc_net(inode);
2413         if (!net)
2414                 return -ENXIO;
2415
2416         err = single_open(file, ipv6_route_show, net);
2417         if (err < 0) {
2418                 put_net(net);
2419                 return err;
2420         }
2421
2422         return 0;
2423 }
2424
2425 static int ipv6_route_release(struct inode *inode, struct file *file)
2426 {
2427         struct seq_file *seq = file->private_data;
2428         struct net *net = seq->private;
2429         put_net(net);
2430         return single_release(inode, file);
2431 }
2432
2433 static const struct file_operations ipv6_route_proc_fops = {
2434         .owner          = THIS_MODULE,
2435         .open           = ipv6_route_open,
2436         .read           = seq_read,
2437         .llseek         = seq_lseek,
2438         .release        = ipv6_route_release,
2439 };
2440
2441 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2442 {
2443         struct net *net = (struct net *)seq->private;
2444         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2445                    net->ipv6.rt6_stats->fib_nodes,
2446                    net->ipv6.rt6_stats->fib_route_nodes,
2447                    net->ipv6.rt6_stats->fib_rt_alloc,
2448                    net->ipv6.rt6_stats->fib_rt_entries,
2449                    net->ipv6.rt6_stats->fib_rt_cache,
2450                    atomic_read(&net->ipv6.ip6_dst_ops->entries),
2451                    net->ipv6.rt6_stats->fib_discarded_routes);
2452
2453         return 0;
2454 }
2455
2456 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2457 {
2458         int err;
2459         struct net *net = get_proc_net(inode);
2460         if (!net)
2461                 return -ENXIO;
2462
2463         err = single_open(file, rt6_stats_seq_show, net);
2464         if (err < 0) {
2465                 put_net(net);
2466                 return err;
2467         }
2468
2469         return 0;
2470 }
2471
2472 static int rt6_stats_seq_release(struct inode *inode, struct file *file)
2473 {
2474         struct seq_file *seq = file->private_data;
2475         struct net *net = (struct net *)seq->private;
2476         put_net(net);
2477         return single_release(inode, file);
2478 }
2479
2480 static const struct file_operations rt6_stats_seq_fops = {
2481         .owner   = THIS_MODULE,
2482         .open    = rt6_stats_seq_open,
2483         .read    = seq_read,
2484         .llseek  = seq_lseek,
2485         .release = rt6_stats_seq_release,
2486 };
2487 #endif  /* CONFIG_PROC_FS */
2488
2489 #ifdef CONFIG_SYSCTL
2490
2491 static
2492 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2493                               void __user *buffer, size_t *lenp, loff_t *ppos)
2494 {
2495         struct net *net = current->nsproxy->net_ns;
2496         int delay = net->ipv6.sysctl.flush_delay;
2497         if (write) {
2498                 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2499                 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2500                 return 0;
2501         } else
2502                 return -EINVAL;
2503 }
2504
2505 ctl_table ipv6_route_table_template[] = {
2506         {
2507                 .procname       =       "flush",
2508                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2509                 .maxlen         =       sizeof(int),
2510                 .mode           =       0200,
2511                 .proc_handler   =       &ipv6_sysctl_rtcache_flush
2512         },
2513         {
2514                 .ctl_name       =       NET_IPV6_ROUTE_GC_THRESH,
2515                 .procname       =       "gc_thresh",
2516                 .data           =       &ip6_dst_ops_template.gc_thresh,
2517                 .maxlen         =       sizeof(int),
2518                 .mode           =       0644,
2519                 .proc_handler   =       &proc_dointvec,
2520         },
2521         {
2522                 .ctl_name       =       NET_IPV6_ROUTE_MAX_SIZE,
2523                 .procname       =       "max_size",
2524                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2525                 .maxlen         =       sizeof(int),
2526                 .mode           =       0644,
2527                 .proc_handler   =       &proc_dointvec,
2528         },
2529         {
2530                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2531                 .procname       =       "gc_min_interval",
2532                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
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_GC_TIMEOUT,
2540                 .procname       =       "gc_timeout",
2541                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
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_INTERVAL,
2549                 .procname       =       "gc_interval",
2550                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2551                 .maxlen         =       sizeof(int),
2552                 .mode           =       0644,
2553                 .proc_handler   =       &proc_dointvec_jiffies,
2554                 .strategy       =       &sysctl_jiffies,
2555         },
2556         {
2557                 .ctl_name       =       NET_IPV6_ROUTE_GC_ELASTICITY,
2558                 .procname       =       "gc_elasticity",
2559                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2560                 .maxlen         =       sizeof(int),
2561                 .mode           =       0644,
2562                 .proc_handler   =       &proc_dointvec_jiffies,
2563                 .strategy       =       &sysctl_jiffies,
2564         },
2565         {
2566                 .ctl_name       =       NET_IPV6_ROUTE_MTU_EXPIRES,
2567                 .procname       =       "mtu_expires",
2568                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2569                 .maxlen         =       sizeof(int),
2570                 .mode           =       0644,
2571                 .proc_handler   =       &proc_dointvec_jiffies,
2572                 .strategy       =       &sysctl_jiffies,
2573         },
2574         {
2575                 .ctl_name       =       NET_IPV6_ROUTE_MIN_ADVMSS,
2576                 .procname       =       "min_adv_mss",
2577                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2578                 .maxlen         =       sizeof(int),
2579                 .mode           =       0644,
2580                 .proc_handler   =       &proc_dointvec_jiffies,
2581                 .strategy       =       &sysctl_jiffies,
2582         },
2583         {
2584                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2585                 .procname       =       "gc_min_interval_ms",
2586                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2587                 .maxlen         =       sizeof(int),
2588                 .mode           =       0644,
2589                 .proc_handler   =       &proc_dointvec_ms_jiffies,
2590                 .strategy       =       &sysctl_ms_jiffies,
2591         },
2592         { .ctl_name = 0 }
2593 };
2594
2595 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2596 {
2597         struct ctl_table *table;
2598
2599         table = kmemdup(ipv6_route_table_template,
2600                         sizeof(ipv6_route_table_template),
2601                         GFP_KERNEL);
2602
2603         if (table) {
2604                 table[0].data = &net->ipv6.sysctl.flush_delay;
2605                 table[1].data = &net->ipv6.ip6_dst_ops->gc_thresh;
2606                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2607                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2608                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2609                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2610                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2611                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2612                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2613         }
2614
2615         return table;
2616 }
2617 #endif
2618
2619 static int ip6_route_net_init(struct net *net)
2620 {
2621         int ret = -ENOMEM;
2622
2623         net->ipv6.ip6_dst_ops = kmemdup(&ip6_dst_ops_template,
2624                                         sizeof(*net->ipv6.ip6_dst_ops),
2625                                         GFP_KERNEL);
2626         if (!net->ipv6.ip6_dst_ops)
2627                 goto out;
2628         net->ipv6.ip6_dst_ops->dst_net = hold_net(net);
2629
2630         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2631                                            sizeof(*net->ipv6.ip6_null_entry),
2632                                            GFP_KERNEL);
2633         if (!net->ipv6.ip6_null_entry)
2634                 goto out_ip6_dst_ops;
2635         net->ipv6.ip6_null_entry->u.dst.path =
2636                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2637         net->ipv6.ip6_null_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2638
2639 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2640         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2641                                                sizeof(*net->ipv6.ip6_prohibit_entry),
2642                                                GFP_KERNEL);
2643         if (!net->ipv6.ip6_prohibit_entry) {
2644                 kfree(net->ipv6.ip6_null_entry);
2645                 goto out;
2646         }
2647         net->ipv6.ip6_prohibit_entry->u.dst.path =
2648                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2649         net->ipv6.ip6_prohibit_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2650
2651         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2652                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
2653                                                GFP_KERNEL);
2654         if (!net->ipv6.ip6_blk_hole_entry) {
2655                 kfree(net->ipv6.ip6_null_entry);
2656                 kfree(net->ipv6.ip6_prohibit_entry);
2657                 goto out;
2658         }
2659         net->ipv6.ip6_blk_hole_entry->u.dst.path =
2660                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2661         net->ipv6.ip6_blk_hole_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2662 #endif
2663
2664 #ifdef CONFIG_PROC_FS
2665         proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2666         proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2667 #endif
2668         net->ipv6.ip6_rt_gc_expire = 30*HZ;
2669
2670         ret = 0;
2671 out:
2672         return ret;
2673
2674 out_ip6_dst_ops:
2675         release_net(net->ipv6.ip6_dst_ops->dst_net);
2676         kfree(net->ipv6.ip6_dst_ops);
2677         goto out;
2678 }
2679
2680 static void ip6_route_net_exit(struct net *net)
2681 {
2682 #ifdef CONFIG_PROC_FS
2683         proc_net_remove(net, "ipv6_route");
2684         proc_net_remove(net, "rt6_stats");
2685 #endif
2686         kfree(net->ipv6.ip6_null_entry);
2687 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2688         kfree(net->ipv6.ip6_prohibit_entry);
2689         kfree(net->ipv6.ip6_blk_hole_entry);
2690 #endif
2691         release_net(net->ipv6.ip6_dst_ops->dst_net);
2692         kfree(net->ipv6.ip6_dst_ops);
2693 }
2694
2695 static struct pernet_operations ip6_route_net_ops = {
2696         .init = ip6_route_net_init,
2697         .exit = ip6_route_net_exit,
2698 };
2699
2700 static struct notifier_block ip6_route_dev_notifier = {
2701         .notifier_call = ip6_route_dev_notify,
2702         .priority = 0,
2703 };
2704
2705 int __init ip6_route_init(void)
2706 {
2707         int ret;
2708
2709         ret = -ENOMEM;
2710         ip6_dst_ops_template.kmem_cachep =
2711                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2712                                   SLAB_HWCACHE_ALIGN, NULL);
2713         if (!ip6_dst_ops_template.kmem_cachep)
2714                 goto out;;
2715
2716         ret = register_pernet_subsys(&ip6_route_net_ops);
2717         if (ret)
2718                 goto out_kmem_cache;
2719
2720         /* Registering of the loopback is done before this portion of code,
2721          * the loopback reference in rt6_info will not be taken, do it
2722          * manually for init_net */
2723         init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2724         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2725   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2726         init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2727         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2728         init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2729         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2730   #endif
2731         ret = fib6_init();
2732         if (ret)
2733                 goto out_register_subsys;
2734
2735         ret = xfrm6_init();
2736         if (ret)
2737                 goto out_fib6_init;
2738
2739         ret = fib6_rules_init();
2740         if (ret)
2741                 goto xfrm6_init;
2742
2743         ret = -ENOBUFS;
2744         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2745             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2746             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2747                 goto fib6_rules_init;
2748
2749         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2750         if (ret)
2751                 goto fib6_rules_init;
2752
2753 out:
2754         return ret;
2755
2756 fib6_rules_init:
2757         fib6_rules_cleanup();
2758 xfrm6_init:
2759         xfrm6_fini();
2760 out_fib6_init:
2761         fib6_gc_cleanup();
2762 out_register_subsys:
2763         unregister_pernet_subsys(&ip6_route_net_ops);
2764 out_kmem_cache:
2765         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2766         goto out;
2767 }
2768
2769 void ip6_route_cleanup(void)
2770 {
2771         unregister_netdevice_notifier(&ip6_route_dev_notifier);
2772         fib6_rules_cleanup();
2773         xfrm6_fini();
2774         fib6_gc_cleanup();
2775         unregister_pernet_subsys(&ip6_route_net_ops);
2776         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2777 }