[IPV6]: Decentralize EXPORT_SYMBOLs.
[linux-2.6.git] / net / ipv6 / route.c
1 /*
2  *      Linux INET6 implementation
3  *      FIB front-end.
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
9  *
10  *      This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  */
15
16 /*      Changes:
17  *
18  *      YOSHIFUJI Hideaki @USAGI
19  *              reworked default router selection.
20  *              - respect outgoing interface
21  *              - select from (probably) reachable routers (i.e.
22  *              routers in REACHABLE, STALE, DELAY or PROBE states).
23  *              - always select the same router if it is (probably)
24  *              reachable.  otherwise, round-robin the list.
25  *      Ville Nuorvala
26  *              Fixed routing subtrees.
27  */
28
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/times.h>
33 #include <linux/socket.h>
34 #include <linux/sockios.h>
35 #include <linux/net.h>
36 #include <linux/route.h>
37 #include <linux/netdevice.h>
38 #include <linux/in6.h>
39 #include <linux/init.h>
40 #include <linux/if_arp.h>
41
42 #ifdef  CONFIG_PROC_FS
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #endif
46
47 #include <net/snmp.h>
48 #include <net/ipv6.h>
49 #include <net/ip6_fib.h>
50 #include <net/ip6_route.h>
51 #include <net/ndisc.h>
52 #include <net/addrconf.h>
53 #include <net/tcp.h>
54 #include <linux/rtnetlink.h>
55 #include <net/dst.h>
56 #include <net/xfrm.h>
57 #include <net/netevent.h>
58 #include <net/netlink.h>
59
60 #include <asm/uaccess.h>
61
62 #ifdef CONFIG_SYSCTL
63 #include <linux/sysctl.h>
64 #endif
65
66 /* Set to 3 to get tracing. */
67 #define RT6_DEBUG 2
68
69 #if RT6_DEBUG >= 3
70 #define RDBG(x) printk x
71 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
72 #else
73 #define RDBG(x)
74 #define RT6_TRACE(x...) do { ; } while (0)
75 #endif
76
77 #define CLONE_OFFLINK_ROUTE 0
78
79 static int ip6_rt_max_size = 4096;
80 static int ip6_rt_gc_min_interval = HZ / 2;
81 static int ip6_rt_gc_timeout = 60*HZ;
82 int ip6_rt_gc_interval = 30*HZ;
83 static int ip6_rt_gc_elasticity = 9;
84 static int ip6_rt_mtu_expires = 10*60*HZ;
85 static int ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
86
87 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
88 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
89 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
90 static void             ip6_dst_destroy(struct dst_entry *);
91 static void             ip6_dst_ifdown(struct dst_entry *,
92                                        struct net_device *dev, int how);
93 static int               ip6_dst_gc(void);
94
95 static int              ip6_pkt_discard(struct sk_buff *skb);
96 static int              ip6_pkt_discard_out(struct sk_buff *skb);
97 static void             ip6_link_failure(struct sk_buff *skb);
98 static void             ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
99
100 #ifdef CONFIG_IPV6_ROUTE_INFO
101 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
102                                            struct in6_addr *gwaddr, int ifindex,
103                                            unsigned pref);
104 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
105                                            struct in6_addr *gwaddr, int ifindex);
106 #endif
107
108 static struct dst_ops ip6_dst_ops = {
109         .family                 =       AF_INET6,
110         .protocol               =       __constant_htons(ETH_P_IPV6),
111         .gc                     =       ip6_dst_gc,
112         .gc_thresh              =       1024,
113         .check                  =       ip6_dst_check,
114         .destroy                =       ip6_dst_destroy,
115         .ifdown                 =       ip6_dst_ifdown,
116         .negative_advice        =       ip6_negative_advice,
117         .link_failure           =       ip6_link_failure,
118         .update_pmtu            =       ip6_rt_update_pmtu,
119         .entry_size             =       sizeof(struct rt6_info),
120 };
121
122 struct rt6_info ip6_null_entry = {
123         .u = {
124                 .dst = {
125                         .__refcnt       = ATOMIC_INIT(1),
126                         .__use          = 1,
127                         .dev            = &loopback_dev,
128                         .obsolete       = -1,
129                         .error          = -ENETUNREACH,
130                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
131                         .input          = ip6_pkt_discard,
132                         .output         = ip6_pkt_discard_out,
133                         .ops            = &ip6_dst_ops,
134                         .path           = (struct dst_entry*)&ip6_null_entry,
135                 }
136         },
137         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
138         .rt6i_metric    = ~(u32) 0,
139         .rt6i_ref       = ATOMIC_INIT(1),
140 };
141
142 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
143
144 static int ip6_pkt_prohibit(struct sk_buff *skb);
145 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
146 static int ip6_pkt_blk_hole(struct sk_buff *skb);
147
148 struct rt6_info ip6_prohibit_entry = {
149         .u = {
150                 .dst = {
151                         .__refcnt       = ATOMIC_INIT(1),
152                         .__use          = 1,
153                         .dev            = &loopback_dev,
154                         .obsolete       = -1,
155                         .error          = -EACCES,
156                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
157                         .input          = ip6_pkt_prohibit,
158                         .output         = ip6_pkt_prohibit_out,
159                         .ops            = &ip6_dst_ops,
160                         .path           = (struct dst_entry*)&ip6_prohibit_entry,
161                 }
162         },
163         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
164         .rt6i_metric    = ~(u32) 0,
165         .rt6i_ref       = ATOMIC_INIT(1),
166 };
167
168 struct rt6_info ip6_blk_hole_entry = {
169         .u = {
170                 .dst = {
171                         .__refcnt       = ATOMIC_INIT(1),
172                         .__use          = 1,
173                         .dev            = &loopback_dev,
174                         .obsolete       = -1,
175                         .error          = -EINVAL,
176                         .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
177                         .input          = ip6_pkt_blk_hole,
178                         .output         = ip6_pkt_blk_hole,
179                         .ops            = &ip6_dst_ops,
180                         .path           = (struct dst_entry*)&ip6_blk_hole_entry,
181                 }
182         },
183         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
184         .rt6i_metric    = ~(u32) 0,
185         .rt6i_ref       = ATOMIC_INIT(1),
186 };
187
188 #endif
189
190 /* allocate dst with ip6_dst_ops */
191 static __inline__ struct rt6_info *ip6_dst_alloc(void)
192 {
193         return (struct rt6_info *)dst_alloc(&ip6_dst_ops);
194 }
195
196 static void ip6_dst_destroy(struct dst_entry *dst)
197 {
198         struct rt6_info *rt = (struct rt6_info *)dst;
199         struct inet6_dev *idev = rt->rt6i_idev;
200
201         if (idev != NULL) {
202                 rt->rt6i_idev = NULL;
203                 in6_dev_put(idev);
204         }
205 }
206
207 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
208                            int how)
209 {
210         struct rt6_info *rt = (struct rt6_info *)dst;
211         struct inet6_dev *idev = rt->rt6i_idev;
212
213         if (dev != &loopback_dev && idev != NULL && idev->dev == dev) {
214                 struct inet6_dev *loopback_idev = in6_dev_get(&loopback_dev);
215                 if (loopback_idev != NULL) {
216                         rt->rt6i_idev = loopback_idev;
217                         in6_dev_put(idev);
218                 }
219         }
220 }
221
222 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
223 {
224         return (rt->rt6i_flags & RTF_EXPIRES &&
225                 time_after(jiffies, rt->rt6i_expires));
226 }
227
228 static inline int rt6_need_strict(struct in6_addr *daddr)
229 {
230         return (ipv6_addr_type(daddr) &
231                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
232 }
233
234 /*
235  *      Route lookup. Any table->tb6_lock is implied.
236  */
237
238 static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
239                                                     int oif,
240                                                     int strict)
241 {
242         struct rt6_info *local = NULL;
243         struct rt6_info *sprt;
244
245         if (oif) {
246                 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
247                         struct net_device *dev = sprt->rt6i_dev;
248                         if (dev->ifindex == oif)
249                                 return sprt;
250                         if (dev->flags & IFF_LOOPBACK) {
251                                 if (sprt->rt6i_idev == NULL ||
252                                     sprt->rt6i_idev->dev->ifindex != oif) {
253                                         if (strict && oif)
254                                                 continue;
255                                         if (local && (!oif ||
256                                                       local->rt6i_idev->dev->ifindex == oif))
257                                                 continue;
258                                 }
259                                 local = sprt;
260                         }
261                 }
262
263                 if (local)
264                         return local;
265
266                 if (strict)
267                         return &ip6_null_entry;
268         }
269         return rt;
270 }
271
272 #ifdef CONFIG_IPV6_ROUTER_PREF
273 static void rt6_probe(struct rt6_info *rt)
274 {
275         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
276         /*
277          * Okay, this does not seem to be appropriate
278          * for now, however, we need to check if it
279          * is really so; aka Router Reachability Probing.
280          *
281          * Router Reachability Probe MUST be rate-limited
282          * to no more than one per minute.
283          */
284         if (!neigh || (neigh->nud_state & NUD_VALID))
285                 return;
286         read_lock_bh(&neigh->lock);
287         if (!(neigh->nud_state & NUD_VALID) &&
288             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
289                 struct in6_addr mcaddr;
290                 struct in6_addr *target;
291
292                 neigh->updated = jiffies;
293                 read_unlock_bh(&neigh->lock);
294
295                 target = (struct in6_addr *)&neigh->primary_key;
296                 addrconf_addr_solict_mult(target, &mcaddr);
297                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
298         } else
299                 read_unlock_bh(&neigh->lock);
300 }
301 #else
302 static inline void rt6_probe(struct rt6_info *rt)
303 {
304         return;
305 }
306 #endif
307
308 /*
309  * Default Router Selection (RFC 2461 6.3.6)
310  */
311 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
312 {
313         struct net_device *dev = rt->rt6i_dev;
314         if (!oif || dev->ifindex == oif)
315                 return 2;
316         if ((dev->flags & IFF_LOOPBACK) &&
317             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
318                 return 1;
319         return 0;
320 }
321
322 static inline int rt6_check_neigh(struct rt6_info *rt)
323 {
324         struct neighbour *neigh = rt->rt6i_nexthop;
325         int m = 0;
326         if (rt->rt6i_flags & RTF_NONEXTHOP ||
327             !(rt->rt6i_flags & RTF_GATEWAY))
328                 m = 1;
329         else if (neigh) {
330                 read_lock_bh(&neigh->lock);
331                 if (neigh->nud_state & NUD_VALID)
332                         m = 2;
333                 else if (!(neigh->nud_state & NUD_FAILED))
334                         m = 1;
335                 read_unlock_bh(&neigh->lock);
336         }
337         return m;
338 }
339
340 static int rt6_score_route(struct rt6_info *rt, int oif,
341                            int strict)
342 {
343         int m, n;
344
345         m = rt6_check_dev(rt, oif);
346         if (!m && (strict & RT6_LOOKUP_F_IFACE))
347                 return -1;
348 #ifdef CONFIG_IPV6_ROUTER_PREF
349         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
350 #endif
351         n = rt6_check_neigh(rt);
352         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
353                 return -1;
354         return m;
355 }
356
357 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
358                                    int *mpri, struct rt6_info *match)
359 {
360         int m;
361
362         if (rt6_check_expired(rt))
363                 goto out;
364
365         m = rt6_score_route(rt, oif, strict);
366         if (m < 0)
367                 goto out;
368
369         if (m > *mpri) {
370                 if (strict & RT6_LOOKUP_F_REACHABLE)
371                         rt6_probe(match);
372                 *mpri = m;
373                 match = rt;
374         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
375                 rt6_probe(rt);
376         }
377
378 out:
379         return match;
380 }
381
382 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
383                                      struct rt6_info *rr_head,
384                                      u32 metric, int oif, int strict)
385 {
386         struct rt6_info *rt, *match;
387         int mpri = -1;
388
389         match = NULL;
390         for (rt = rr_head; rt && rt->rt6i_metric == metric;
391              rt = rt->u.dst.rt6_next)
392                 match = find_match(rt, oif, strict, &mpri, match);
393         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
394              rt = rt->u.dst.rt6_next)
395                 match = find_match(rt, oif, strict, &mpri, match);
396
397         return match;
398 }
399
400 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
401 {
402         struct rt6_info *match, *rt0;
403
404         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
405                   __FUNCTION__, fn->leaf, oif);
406
407         rt0 = fn->rr_ptr;
408         if (!rt0)
409                 fn->rr_ptr = rt0 = fn->leaf;
410
411         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
412
413         if (!match &&
414             (strict & RT6_LOOKUP_F_REACHABLE)) {
415                 struct rt6_info *next = rt0->u.dst.rt6_next;
416
417                 /* no entries matched; do round-robin */
418                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
419                         next = fn->leaf;
420
421                 if (next != rt0)
422                         fn->rr_ptr = next;
423         }
424
425         RT6_TRACE("%s() => %p\n",
426                   __FUNCTION__, match);
427
428         return (match ? match : &ip6_null_entry);
429 }
430
431 #ifdef CONFIG_IPV6_ROUTE_INFO
432 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
433                   struct in6_addr *gwaddr)
434 {
435         struct route_info *rinfo = (struct route_info *) opt;
436         struct in6_addr prefix_buf, *prefix;
437         unsigned int pref;
438         u32 lifetime;
439         struct rt6_info *rt;
440
441         if (len < sizeof(struct route_info)) {
442                 return -EINVAL;
443         }
444
445         /* Sanity check for prefix_len and length */
446         if (rinfo->length > 3) {
447                 return -EINVAL;
448         } else if (rinfo->prefix_len > 128) {
449                 return -EINVAL;
450         } else if (rinfo->prefix_len > 64) {
451                 if (rinfo->length < 2) {
452                         return -EINVAL;
453                 }
454         } else if (rinfo->prefix_len > 0) {
455                 if (rinfo->length < 1) {
456                         return -EINVAL;
457                 }
458         }
459
460         pref = rinfo->route_pref;
461         if (pref == ICMPV6_ROUTER_PREF_INVALID)
462                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
463
464         lifetime = ntohl(rinfo->lifetime);
465         if (lifetime == 0xffffffff) {
466                 /* infinity */
467         } else if (lifetime > 0x7fffffff/HZ) {
468                 /* Avoid arithmetic overflow */
469                 lifetime = 0x7fffffff/HZ - 1;
470         }
471
472         if (rinfo->length == 3)
473                 prefix = (struct in6_addr *)rinfo->prefix;
474         else {
475                 /* this function is safe */
476                 ipv6_addr_prefix(&prefix_buf,
477                                  (struct in6_addr *)rinfo->prefix,
478                                  rinfo->prefix_len);
479                 prefix = &prefix_buf;
480         }
481
482         rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
483
484         if (rt && !lifetime) {
485                 ip6_del_rt(rt);
486                 rt = NULL;
487         }
488
489         if (!rt && lifetime)
490                 rt = rt6_add_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
491                                         pref);
492         else if (rt)
493                 rt->rt6i_flags = RTF_ROUTEINFO |
494                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
495
496         if (rt) {
497                 if (lifetime == 0xffffffff) {
498                         rt->rt6i_flags &= ~RTF_EXPIRES;
499                 } else {
500                         rt->rt6i_expires = jiffies + HZ * lifetime;
501                         rt->rt6i_flags |= RTF_EXPIRES;
502                 }
503                 dst_release(&rt->u.dst);
504         }
505         return 0;
506 }
507 #endif
508
509 #define BACKTRACK(saddr) \
510 do { \
511         if (rt == &ip6_null_entry) { \
512                 struct fib6_node *pn; \
513                 while (1) { \
514                         if (fn->fn_flags & RTN_TL_ROOT) \
515                                 goto out; \
516                         pn = fn->parent; \
517                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
518                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
519                         else \
520                                 fn = pn; \
521                         if (fn->fn_flags & RTN_RTINFO) \
522                                 goto restart; \
523                 } \
524         } \
525 } while(0)
526
527 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
528                                              struct flowi *fl, int flags)
529 {
530         struct fib6_node *fn;
531         struct rt6_info *rt;
532
533         read_lock_bh(&table->tb6_lock);
534         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
535 restart:
536         rt = fn->leaf;
537         rt = rt6_device_match(rt, fl->oif, flags);
538         BACKTRACK(&fl->fl6_src);
539 out:
540         dst_hold(&rt->u.dst);
541         read_unlock_bh(&table->tb6_lock);
542
543         rt->u.dst.lastuse = jiffies;
544         rt->u.dst.__use++;
545
546         return rt;
547
548 }
549
550 struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
551                             int oif, int strict)
552 {
553         struct flowi fl = {
554                 .oif = oif,
555                 .nl_u = {
556                         .ip6_u = {
557                                 .daddr = *daddr,
558                         },
559                 },
560         };
561         struct dst_entry *dst;
562         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
563
564         if (saddr) {
565                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
566                 flags |= RT6_LOOKUP_F_HAS_SADDR;
567         }
568
569         dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_lookup);
570         if (dst->error == 0)
571                 return (struct rt6_info *) dst;
572
573         dst_release(dst);
574
575         return NULL;
576 }
577
578 EXPORT_SYMBOL(rt6_lookup);
579
580 /* ip6_ins_rt is called with FREE table->tb6_lock.
581    It takes new route entry, the addition fails by any reason the
582    route is freed. In any case, if caller does not hold it, it may
583    be destroyed.
584  */
585
586 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
587 {
588         int err;
589         struct fib6_table *table;
590
591         table = rt->rt6i_table;
592         write_lock_bh(&table->tb6_lock);
593         err = fib6_add(&table->tb6_root, rt, info);
594         write_unlock_bh(&table->tb6_lock);
595
596         return err;
597 }
598
599 int ip6_ins_rt(struct rt6_info *rt)
600 {
601         return __ip6_ins_rt(rt, NULL);
602 }
603
604 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
605                                       struct in6_addr *saddr)
606 {
607         struct rt6_info *rt;
608
609         /*
610          *      Clone the route.
611          */
612
613         rt = ip6_rt_copy(ort);
614
615         if (rt) {
616                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
617                         if (rt->rt6i_dst.plen != 128 &&
618                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
619                                 rt->rt6i_flags |= RTF_ANYCAST;
620                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
621                 }
622
623                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
624                 rt->rt6i_dst.plen = 128;
625                 rt->rt6i_flags |= RTF_CACHE;
626                 rt->u.dst.flags |= DST_HOST;
627
628 #ifdef CONFIG_IPV6_SUBTREES
629                 if (rt->rt6i_src.plen && saddr) {
630                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
631                         rt->rt6i_src.plen = 128;
632                 }
633 #endif
634
635                 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
636
637         }
638
639         return rt;
640 }
641
642 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
643 {
644         struct rt6_info *rt = ip6_rt_copy(ort);
645         if (rt) {
646                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
647                 rt->rt6i_dst.plen = 128;
648                 rt->rt6i_flags |= RTF_CACHE;
649                 rt->u.dst.flags |= DST_HOST;
650                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
651         }
652         return rt;
653 }
654
655 static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
656                                             struct flowi *fl, int flags)
657 {
658         struct fib6_node *fn;
659         struct rt6_info *rt, *nrt;
660         int strict = 0;
661         int attempts = 3;
662         int err;
663         int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
664
665         strict |= flags & RT6_LOOKUP_F_IFACE;
666
667 relookup:
668         read_lock_bh(&table->tb6_lock);
669
670 restart_2:
671         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
672
673 restart:
674         rt = rt6_select(fn, fl->iif, strict | reachable);
675         BACKTRACK(&fl->fl6_src);
676         if (rt == &ip6_null_entry ||
677             rt->rt6i_flags & RTF_CACHE)
678                 goto out;
679
680         dst_hold(&rt->u.dst);
681         read_unlock_bh(&table->tb6_lock);
682
683         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
684                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
685         else {
686 #if CLONE_OFFLINK_ROUTE
687                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
688 #else
689                 goto out2;
690 #endif
691         }
692
693         dst_release(&rt->u.dst);
694         rt = nrt ? : &ip6_null_entry;
695
696         dst_hold(&rt->u.dst);
697         if (nrt) {
698                 err = ip6_ins_rt(nrt);
699                 if (!err)
700                         goto out2;
701         }
702
703         if (--attempts <= 0)
704                 goto out2;
705
706         /*
707          * Race condition! In the gap, when table->tb6_lock was
708          * released someone could insert this route.  Relookup.
709          */
710         dst_release(&rt->u.dst);
711         goto relookup;
712
713 out:
714         if (reachable) {
715                 reachable = 0;
716                 goto restart_2;
717         }
718         dst_hold(&rt->u.dst);
719         read_unlock_bh(&table->tb6_lock);
720 out2:
721         rt->u.dst.lastuse = jiffies;
722         rt->u.dst.__use++;
723
724         return rt;
725 }
726
727 void ip6_route_input(struct sk_buff *skb)
728 {
729         struct ipv6hdr *iph = skb->nh.ipv6h;
730         int flags = RT6_LOOKUP_F_HAS_SADDR;
731         struct flowi fl = {
732                 .iif = skb->dev->ifindex,
733                 .nl_u = {
734                         .ip6_u = {
735                                 .daddr = iph->daddr,
736                                 .saddr = iph->saddr,
737                                 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
738                         },
739                 },
740                 .mark = skb->mark,
741                 .proto = iph->nexthdr,
742         };
743
744         if (rt6_need_strict(&iph->daddr))
745                 flags |= RT6_LOOKUP_F_IFACE;
746
747         skb->dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_input);
748 }
749
750 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
751                                              struct flowi *fl, int flags)
752 {
753         struct fib6_node *fn;
754         struct rt6_info *rt, *nrt;
755         int strict = 0;
756         int attempts = 3;
757         int err;
758         int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
759
760         strict |= flags & RT6_LOOKUP_F_IFACE;
761
762 relookup:
763         read_lock_bh(&table->tb6_lock);
764
765 restart_2:
766         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
767
768 restart:
769         rt = rt6_select(fn, fl->oif, strict | reachable);
770         BACKTRACK(&fl->fl6_src);
771         if (rt == &ip6_null_entry ||
772             rt->rt6i_flags & RTF_CACHE)
773                 goto out;
774
775         dst_hold(&rt->u.dst);
776         read_unlock_bh(&table->tb6_lock);
777
778         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
779                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
780         else {
781 #if CLONE_OFFLINK_ROUTE
782                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
783 #else
784                 goto out2;
785 #endif
786         }
787
788         dst_release(&rt->u.dst);
789         rt = nrt ? : &ip6_null_entry;
790
791         dst_hold(&rt->u.dst);
792         if (nrt) {
793                 err = ip6_ins_rt(nrt);
794                 if (!err)
795                         goto out2;
796         }
797
798         if (--attempts <= 0)
799                 goto out2;
800
801         /*
802          * Race condition! In the gap, when table->tb6_lock was
803          * released someone could insert this route.  Relookup.
804          */
805         dst_release(&rt->u.dst);
806         goto relookup;
807
808 out:
809         if (reachable) {
810                 reachable = 0;
811                 goto restart_2;
812         }
813         dst_hold(&rt->u.dst);
814         read_unlock_bh(&table->tb6_lock);
815 out2:
816         rt->u.dst.lastuse = jiffies;
817         rt->u.dst.__use++;
818         return rt;
819 }
820
821 struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
822 {
823         int flags = 0;
824
825         if (rt6_need_strict(&fl->fl6_dst))
826                 flags |= RT6_LOOKUP_F_IFACE;
827
828         if (!ipv6_addr_any(&fl->fl6_src))
829                 flags |= RT6_LOOKUP_F_HAS_SADDR;
830
831         return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
832 }
833
834 EXPORT_SYMBOL(ip6_route_output);
835
836 /*
837  *      Destination cache support functions
838  */
839
840 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
841 {
842         struct rt6_info *rt;
843
844         rt = (struct rt6_info *) dst;
845
846         if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
847                 return dst;
848
849         return NULL;
850 }
851
852 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
853 {
854         struct rt6_info *rt = (struct rt6_info *) dst;
855
856         if (rt) {
857                 if (rt->rt6i_flags & RTF_CACHE)
858                         ip6_del_rt(rt);
859                 else
860                         dst_release(dst);
861         }
862         return NULL;
863 }
864
865 static void ip6_link_failure(struct sk_buff *skb)
866 {
867         struct rt6_info *rt;
868
869         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
870
871         rt = (struct rt6_info *) skb->dst;
872         if (rt) {
873                 if (rt->rt6i_flags&RTF_CACHE) {
874                         dst_set_expires(&rt->u.dst, 0);
875                         rt->rt6i_flags |= RTF_EXPIRES;
876                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
877                         rt->rt6i_node->fn_sernum = -1;
878         }
879 }
880
881 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
882 {
883         struct rt6_info *rt6 = (struct rt6_info*)dst;
884
885         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
886                 rt6->rt6i_flags |= RTF_MODIFIED;
887                 if (mtu < IPV6_MIN_MTU) {
888                         mtu = IPV6_MIN_MTU;
889                         dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
890                 }
891                 dst->metrics[RTAX_MTU-1] = mtu;
892                 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
893         }
894 }
895
896 static int ipv6_get_mtu(struct net_device *dev);
897
898 static inline unsigned int ipv6_advmss(unsigned int mtu)
899 {
900         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
901
902         if (mtu < ip6_rt_min_advmss)
903                 mtu = ip6_rt_min_advmss;
904
905         /*
906          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
907          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
908          * IPV6_MAXPLEN is also valid and means: "any MSS,
909          * rely only on pmtu discovery"
910          */
911         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
912                 mtu = IPV6_MAXPLEN;
913         return mtu;
914 }
915
916 static struct dst_entry *ndisc_dst_gc_list;
917 static DEFINE_SPINLOCK(ndisc_lock);
918
919 struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
920                                   struct neighbour *neigh,
921                                   struct in6_addr *addr,
922                                   int (*output)(struct sk_buff *))
923 {
924         struct rt6_info *rt;
925         struct inet6_dev *idev = in6_dev_get(dev);
926
927         if (unlikely(idev == NULL))
928                 return NULL;
929
930         rt = ip6_dst_alloc();
931         if (unlikely(rt == NULL)) {
932                 in6_dev_put(idev);
933                 goto out;
934         }
935
936         dev_hold(dev);
937         if (neigh)
938                 neigh_hold(neigh);
939         else
940                 neigh = ndisc_get_neigh(dev, addr);
941
942         rt->rt6i_dev      = dev;
943         rt->rt6i_idev     = idev;
944         rt->rt6i_nexthop  = neigh;
945         atomic_set(&rt->u.dst.__refcnt, 1);
946         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
947         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
948         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
949         rt->u.dst.output  = output;
950
951 #if 0   /* there's no chance to use these for ndisc */
952         rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
953                                 ? DST_HOST
954                                 : 0;
955         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
956         rt->rt6i_dst.plen = 128;
957 #endif
958
959         spin_lock_bh(&ndisc_lock);
960         rt->u.dst.next = ndisc_dst_gc_list;
961         ndisc_dst_gc_list = &rt->u.dst;
962         spin_unlock_bh(&ndisc_lock);
963
964         fib6_force_start_gc();
965
966 out:
967         return &rt->u.dst;
968 }
969
970 int ndisc_dst_gc(int *more)
971 {
972         struct dst_entry *dst, *next, **pprev;
973         int freed;
974
975         next = NULL;
976         freed = 0;
977
978         spin_lock_bh(&ndisc_lock);
979         pprev = &ndisc_dst_gc_list;
980
981         while ((dst = *pprev) != NULL) {
982                 if (!atomic_read(&dst->__refcnt)) {
983                         *pprev = dst->next;
984                         dst_free(dst);
985                         freed++;
986                 } else {
987                         pprev = &dst->next;
988                         (*more)++;
989                 }
990         }
991
992         spin_unlock_bh(&ndisc_lock);
993
994         return freed;
995 }
996
997 static int ip6_dst_gc(void)
998 {
999         static unsigned expire = 30*HZ;
1000         static unsigned long last_gc;
1001         unsigned long now = jiffies;
1002
1003         if (time_after(last_gc + ip6_rt_gc_min_interval, now) &&
1004             atomic_read(&ip6_dst_ops.entries) <= ip6_rt_max_size)
1005                 goto out;
1006
1007         expire++;
1008         fib6_run_gc(expire);
1009         last_gc = now;
1010         if (atomic_read(&ip6_dst_ops.entries) < ip6_dst_ops.gc_thresh)
1011                 expire = ip6_rt_gc_timeout>>1;
1012
1013 out:
1014         expire -= expire>>ip6_rt_gc_elasticity;
1015         return (atomic_read(&ip6_dst_ops.entries) > ip6_rt_max_size);
1016 }
1017
1018 /* Clean host part of a prefix. Not necessary in radix tree,
1019    but results in cleaner routing tables.
1020
1021    Remove it only when all the things will work!
1022  */
1023
1024 static int ipv6_get_mtu(struct net_device *dev)
1025 {
1026         int mtu = IPV6_MIN_MTU;
1027         struct inet6_dev *idev;
1028
1029         idev = in6_dev_get(dev);
1030         if (idev) {
1031                 mtu = idev->cnf.mtu6;
1032                 in6_dev_put(idev);
1033         }
1034         return mtu;
1035 }
1036
1037 int ipv6_get_hoplimit(struct net_device *dev)
1038 {
1039         int hoplimit = ipv6_devconf.hop_limit;
1040         struct inet6_dev *idev;
1041
1042         idev = in6_dev_get(dev);
1043         if (idev) {
1044                 hoplimit = idev->cnf.hop_limit;
1045                 in6_dev_put(idev);
1046         }
1047         return hoplimit;
1048 }
1049
1050 /*
1051  *
1052  */
1053
1054 int ip6_route_add(struct fib6_config *cfg)
1055 {
1056         int err;
1057         struct rt6_info *rt = NULL;
1058         struct net_device *dev = NULL;
1059         struct inet6_dev *idev = NULL;
1060         struct fib6_table *table;
1061         int addr_type;
1062
1063         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1064                 return -EINVAL;
1065 #ifndef CONFIG_IPV6_SUBTREES
1066         if (cfg->fc_src_len)
1067                 return -EINVAL;
1068 #endif
1069         if (cfg->fc_ifindex) {
1070                 err = -ENODEV;
1071                 dev = dev_get_by_index(cfg->fc_ifindex);
1072                 if (!dev)
1073                         goto out;
1074                 idev = in6_dev_get(dev);
1075                 if (!idev)
1076                         goto out;
1077         }
1078
1079         if (cfg->fc_metric == 0)
1080                 cfg->fc_metric = IP6_RT_PRIO_USER;
1081
1082         table = fib6_new_table(cfg->fc_table);
1083         if (table == NULL) {
1084                 err = -ENOBUFS;
1085                 goto out;
1086         }
1087
1088         rt = ip6_dst_alloc();
1089
1090         if (rt == NULL) {
1091                 err = -ENOMEM;
1092                 goto out;
1093         }
1094
1095         rt->u.dst.obsolete = -1;
1096         rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1097
1098         if (cfg->fc_protocol == RTPROT_UNSPEC)
1099                 cfg->fc_protocol = RTPROT_BOOT;
1100         rt->rt6i_protocol = cfg->fc_protocol;
1101
1102         addr_type = ipv6_addr_type(&cfg->fc_dst);
1103
1104         if (addr_type & IPV6_ADDR_MULTICAST)
1105                 rt->u.dst.input = ip6_mc_input;
1106         else
1107                 rt->u.dst.input = ip6_forward;
1108
1109         rt->u.dst.output = ip6_output;
1110
1111         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1112         rt->rt6i_dst.plen = cfg->fc_dst_len;
1113         if (rt->rt6i_dst.plen == 128)
1114                rt->u.dst.flags = DST_HOST;
1115
1116 #ifdef CONFIG_IPV6_SUBTREES
1117         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1118         rt->rt6i_src.plen = cfg->fc_src_len;
1119 #endif
1120
1121         rt->rt6i_metric = cfg->fc_metric;
1122
1123         /* We cannot add true routes via loopback here,
1124            they would result in kernel looping; promote them to reject routes
1125          */
1126         if ((cfg->fc_flags & RTF_REJECT) ||
1127             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1128                 /* hold loopback dev/idev if we haven't done so. */
1129                 if (dev != &loopback_dev) {
1130                         if (dev) {
1131                                 dev_put(dev);
1132                                 in6_dev_put(idev);
1133                         }
1134                         dev = &loopback_dev;
1135                         dev_hold(dev);
1136                         idev = in6_dev_get(dev);
1137                         if (!idev) {
1138                                 err = -ENODEV;
1139                                 goto out;
1140                         }
1141                 }
1142                 rt->u.dst.output = ip6_pkt_discard_out;
1143                 rt->u.dst.input = ip6_pkt_discard;
1144                 rt->u.dst.error = -ENETUNREACH;
1145                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1146                 goto install_route;
1147         }
1148
1149         if (cfg->fc_flags & RTF_GATEWAY) {
1150                 struct in6_addr *gw_addr;
1151                 int gwa_type;
1152
1153                 gw_addr = &cfg->fc_gateway;
1154                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1155                 gwa_type = ipv6_addr_type(gw_addr);
1156
1157                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1158                         struct rt6_info *grt;
1159
1160                         /* IPv6 strictly inhibits using not link-local
1161                            addresses as nexthop address.
1162                            Otherwise, router will not able to send redirects.
1163                            It is very good, but in some (rare!) circumstances
1164                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1165                            some exceptions. --ANK
1166                          */
1167                         err = -EINVAL;
1168                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1169                                 goto out;
1170
1171                         grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
1172
1173                         err = -EHOSTUNREACH;
1174                         if (grt == NULL)
1175                                 goto out;
1176                         if (dev) {
1177                                 if (dev != grt->rt6i_dev) {
1178                                         dst_release(&grt->u.dst);
1179                                         goto out;
1180                                 }
1181                         } else {
1182                                 dev = grt->rt6i_dev;
1183                                 idev = grt->rt6i_idev;
1184                                 dev_hold(dev);
1185                                 in6_dev_hold(grt->rt6i_idev);
1186                         }
1187                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1188                                 err = 0;
1189                         dst_release(&grt->u.dst);
1190
1191                         if (err)
1192                                 goto out;
1193                 }
1194                 err = -EINVAL;
1195                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1196                         goto out;
1197         }
1198
1199         err = -ENODEV;
1200         if (dev == NULL)
1201                 goto out;
1202
1203         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1204                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1205                 if (IS_ERR(rt->rt6i_nexthop)) {
1206                         err = PTR_ERR(rt->rt6i_nexthop);
1207                         rt->rt6i_nexthop = NULL;
1208                         goto out;
1209                 }
1210         }
1211
1212         rt->rt6i_flags = cfg->fc_flags;
1213
1214 install_route:
1215         if (cfg->fc_mx) {
1216                 struct nlattr *nla;
1217                 int remaining;
1218
1219                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1220                         int type = nla->nla_type;
1221
1222                         if (type) {
1223                                 if (type > RTAX_MAX) {
1224                                         err = -EINVAL;
1225                                         goto out;
1226                                 }
1227
1228                                 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1229                         }
1230                 }
1231         }
1232
1233         if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1234                 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1235         if (!rt->u.dst.metrics[RTAX_MTU-1])
1236                 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1237         if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1238                 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1239         rt->u.dst.dev = dev;
1240         rt->rt6i_idev = idev;
1241         rt->rt6i_table = table;
1242         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1243
1244 out:
1245         if (dev)
1246                 dev_put(dev);
1247         if (idev)
1248                 in6_dev_put(idev);
1249         if (rt)
1250                 dst_free(&rt->u.dst);
1251         return err;
1252 }
1253
1254 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1255 {
1256         int err;
1257         struct fib6_table *table;
1258
1259         if (rt == &ip6_null_entry)
1260                 return -ENOENT;
1261
1262         table = rt->rt6i_table;
1263         write_lock_bh(&table->tb6_lock);
1264
1265         err = fib6_del(rt, info);
1266         dst_release(&rt->u.dst);
1267
1268         write_unlock_bh(&table->tb6_lock);
1269
1270         return err;
1271 }
1272
1273 int ip6_del_rt(struct rt6_info *rt)
1274 {
1275         return __ip6_del_rt(rt, NULL);
1276 }
1277
1278 static int ip6_route_del(struct fib6_config *cfg)
1279 {
1280         struct fib6_table *table;
1281         struct fib6_node *fn;
1282         struct rt6_info *rt;
1283         int err = -ESRCH;
1284
1285         table = fib6_get_table(cfg->fc_table);
1286         if (table == NULL)
1287                 return err;
1288
1289         read_lock_bh(&table->tb6_lock);
1290
1291         fn = fib6_locate(&table->tb6_root,
1292                          &cfg->fc_dst, cfg->fc_dst_len,
1293                          &cfg->fc_src, cfg->fc_src_len);
1294
1295         if (fn) {
1296                 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1297                         if (cfg->fc_ifindex &&
1298                             (rt->rt6i_dev == NULL ||
1299                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1300                                 continue;
1301                         if (cfg->fc_flags & RTF_GATEWAY &&
1302                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1303                                 continue;
1304                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1305                                 continue;
1306                         dst_hold(&rt->u.dst);
1307                         read_unlock_bh(&table->tb6_lock);
1308
1309                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1310                 }
1311         }
1312         read_unlock_bh(&table->tb6_lock);
1313
1314         return err;
1315 }
1316
1317 /*
1318  *      Handle redirects
1319  */
1320 struct ip6rd_flowi {
1321         struct flowi fl;
1322         struct in6_addr gateway;
1323 };
1324
1325 static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
1326                                              struct flowi *fl,
1327                                              int flags)
1328 {
1329         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1330         struct rt6_info *rt;
1331         struct fib6_node *fn;
1332
1333         /*
1334          * Get the "current" route for this destination and
1335          * check if the redirect has come from approriate router.
1336          *
1337          * RFC 2461 specifies that redirects should only be
1338          * accepted if they come from the nexthop to the target.
1339          * Due to the way the routes are chosen, this notion
1340          * is a bit fuzzy and one might need to check all possible
1341          * routes.
1342          */
1343
1344         read_lock_bh(&table->tb6_lock);
1345         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1346 restart:
1347         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1348                 /*
1349                  * Current route is on-link; redirect is always invalid.
1350                  *
1351                  * Seems, previous statement is not true. It could
1352                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1353                  * But then router serving it might decide, that we should
1354                  * know truth 8)8) --ANK (980726).
1355                  */
1356                 if (rt6_check_expired(rt))
1357                         continue;
1358                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1359                         continue;
1360                 if (fl->oif != rt->rt6i_dev->ifindex)
1361                         continue;
1362                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1363                         continue;
1364                 break;
1365         }
1366
1367         if (!rt)
1368                 rt = &ip6_null_entry;
1369         BACKTRACK(&fl->fl6_src);
1370 out:
1371         dst_hold(&rt->u.dst);
1372
1373         read_unlock_bh(&table->tb6_lock);
1374
1375         return rt;
1376 };
1377
1378 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1379                                            struct in6_addr *src,
1380                                            struct in6_addr *gateway,
1381                                            struct net_device *dev)
1382 {
1383         int flags = RT6_LOOKUP_F_HAS_SADDR;
1384         struct ip6rd_flowi rdfl = {
1385                 .fl = {
1386                         .oif = dev->ifindex,
1387                         .nl_u = {
1388                                 .ip6_u = {
1389                                         .daddr = *dest,
1390                                         .saddr = *src,
1391                                 },
1392                         },
1393                 },
1394                 .gateway = *gateway,
1395         };
1396
1397         if (rt6_need_strict(dest))
1398                 flags |= RT6_LOOKUP_F_IFACE;
1399
1400         return (struct rt6_info *)fib6_rule_lookup((struct flowi *)&rdfl, flags, __ip6_route_redirect);
1401 }
1402
1403 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1404                   struct in6_addr *saddr,
1405                   struct neighbour *neigh, u8 *lladdr, int on_link)
1406 {
1407         struct rt6_info *rt, *nrt = NULL;
1408         struct netevent_redirect netevent;
1409
1410         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1411
1412         if (rt == &ip6_null_entry) {
1413                 if (net_ratelimit())
1414                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1415                                "for redirect target\n");
1416                 goto out;
1417         }
1418
1419         /*
1420          *      We have finally decided to accept it.
1421          */
1422
1423         neigh_update(neigh, lladdr, NUD_STALE,
1424                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1425                      NEIGH_UPDATE_F_OVERRIDE|
1426                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1427                                      NEIGH_UPDATE_F_ISROUTER))
1428                      );
1429
1430         /*
1431          * Redirect received -> path was valid.
1432          * Look, redirects are sent only in response to data packets,
1433          * so that this nexthop apparently is reachable. --ANK
1434          */
1435         dst_confirm(&rt->u.dst);
1436
1437         /* Duplicate redirect: silently ignore. */
1438         if (neigh == rt->u.dst.neighbour)
1439                 goto out;
1440
1441         nrt = ip6_rt_copy(rt);
1442         if (nrt == NULL)
1443                 goto out;
1444
1445         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1446         if (on_link)
1447                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1448
1449         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1450         nrt->rt6i_dst.plen = 128;
1451         nrt->u.dst.flags |= DST_HOST;
1452
1453         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1454         nrt->rt6i_nexthop = neigh_clone(neigh);
1455         /* Reset pmtu, it may be better */
1456         nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1457         nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&nrt->u.dst));
1458
1459         if (ip6_ins_rt(nrt))
1460                 goto out;
1461
1462         netevent.old = &rt->u.dst;
1463         netevent.new = &nrt->u.dst;
1464         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1465
1466         if (rt->rt6i_flags&RTF_CACHE) {
1467                 ip6_del_rt(rt);
1468                 return;
1469         }
1470
1471 out:
1472         dst_release(&rt->u.dst);
1473         return;
1474 }
1475
1476 /*
1477  *      Handle ICMP "packet too big" messages
1478  *      i.e. Path MTU discovery
1479  */
1480
1481 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1482                         struct net_device *dev, u32 pmtu)
1483 {
1484         struct rt6_info *rt, *nrt;
1485         int allfrag = 0;
1486
1487         rt = rt6_lookup(daddr, saddr, dev->ifindex, 0);
1488         if (rt == NULL)
1489                 return;
1490
1491         if (pmtu >= dst_mtu(&rt->u.dst))
1492                 goto out;
1493
1494         if (pmtu < IPV6_MIN_MTU) {
1495                 /*
1496                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1497                  * MTU (1280) and a fragment header should always be included
1498                  * after a node receiving Too Big message reporting PMTU is
1499                  * less than the IPv6 Minimum Link MTU.
1500                  */
1501                 pmtu = IPV6_MIN_MTU;
1502                 allfrag = 1;
1503         }
1504
1505         /* New mtu received -> path was valid.
1506            They are sent only in response to data packets,
1507            so that this nexthop apparently is reachable. --ANK
1508          */
1509         dst_confirm(&rt->u.dst);
1510
1511         /* Host route. If it is static, it would be better
1512            not to override it, but add new one, so that
1513            when cache entry will expire old pmtu
1514            would return automatically.
1515          */
1516         if (rt->rt6i_flags & RTF_CACHE) {
1517                 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1518                 if (allfrag)
1519                         rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1520                 dst_set_expires(&rt->u.dst, ip6_rt_mtu_expires);
1521                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1522                 goto out;
1523         }
1524
1525         /* Network route.
1526            Two cases are possible:
1527            1. It is connected route. Action: COW
1528            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1529          */
1530         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1531                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1532         else
1533                 nrt = rt6_alloc_clone(rt, daddr);
1534
1535         if (nrt) {
1536                 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1537                 if (allfrag)
1538                         nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1539
1540                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1541                  * happened within 5 mins, the recommended timer is 10 mins.
1542                  * Here this route expiration time is set to ip6_rt_mtu_expires
1543                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1544                  * and detecting PMTU increase will be automatically happened.
1545                  */
1546                 dst_set_expires(&nrt->u.dst, ip6_rt_mtu_expires);
1547                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1548
1549                 ip6_ins_rt(nrt);
1550         }
1551 out:
1552         dst_release(&rt->u.dst);
1553 }
1554
1555 /*
1556  *      Misc support functions
1557  */
1558
1559 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1560 {
1561         struct rt6_info *rt = ip6_dst_alloc();
1562
1563         if (rt) {
1564                 rt->u.dst.input = ort->u.dst.input;
1565                 rt->u.dst.output = ort->u.dst.output;
1566
1567                 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1568                 rt->u.dst.error = ort->u.dst.error;
1569                 rt->u.dst.dev = ort->u.dst.dev;
1570                 if (rt->u.dst.dev)
1571                         dev_hold(rt->u.dst.dev);
1572                 rt->rt6i_idev = ort->rt6i_idev;
1573                 if (rt->rt6i_idev)
1574                         in6_dev_hold(rt->rt6i_idev);
1575                 rt->u.dst.lastuse = jiffies;
1576                 rt->rt6i_expires = 0;
1577
1578                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1579                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1580                 rt->rt6i_metric = 0;
1581
1582                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1583 #ifdef CONFIG_IPV6_SUBTREES
1584                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1585 #endif
1586                 rt->rt6i_table = ort->rt6i_table;
1587         }
1588         return rt;
1589 }
1590
1591 #ifdef CONFIG_IPV6_ROUTE_INFO
1592 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
1593                                            struct in6_addr *gwaddr, int ifindex)
1594 {
1595         struct fib6_node *fn;
1596         struct rt6_info *rt = NULL;
1597         struct fib6_table *table;
1598
1599         table = fib6_get_table(RT6_TABLE_INFO);
1600         if (table == NULL)
1601                 return NULL;
1602
1603         write_lock_bh(&table->tb6_lock);
1604         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1605         if (!fn)
1606                 goto out;
1607
1608         for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1609                 if (rt->rt6i_dev->ifindex != ifindex)
1610                         continue;
1611                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1612                         continue;
1613                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1614                         continue;
1615                 dst_hold(&rt->u.dst);
1616                 break;
1617         }
1618 out:
1619         write_unlock_bh(&table->tb6_lock);
1620         return rt;
1621 }
1622
1623 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
1624                                            struct in6_addr *gwaddr, int ifindex,
1625                                            unsigned pref)
1626 {
1627         struct fib6_config cfg = {
1628                 .fc_table       = RT6_TABLE_INFO,
1629                 .fc_metric      = 1024,
1630                 .fc_ifindex     = ifindex,
1631                 .fc_dst_len     = prefixlen,
1632                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1633                                   RTF_UP | RTF_PREF(pref),
1634         };
1635
1636         ipv6_addr_copy(&cfg.fc_dst, prefix);
1637         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1638
1639         /* We should treat it as a default route if prefix length is 0. */
1640         if (!prefixlen)
1641                 cfg.fc_flags |= RTF_DEFAULT;
1642
1643         ip6_route_add(&cfg);
1644
1645         return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
1646 }
1647 #endif
1648
1649 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1650 {
1651         struct rt6_info *rt;
1652         struct fib6_table *table;
1653
1654         table = fib6_get_table(RT6_TABLE_DFLT);
1655         if (table == NULL)
1656                 return NULL;
1657
1658         write_lock_bh(&table->tb6_lock);
1659         for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1660                 if (dev == rt->rt6i_dev &&
1661                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1662                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1663                         break;
1664         }
1665         if (rt)
1666                 dst_hold(&rt->u.dst);
1667         write_unlock_bh(&table->tb6_lock);
1668         return rt;
1669 }
1670
1671 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1672                                      struct net_device *dev,
1673                                      unsigned int pref)
1674 {
1675         struct fib6_config cfg = {
1676                 .fc_table       = RT6_TABLE_DFLT,
1677                 .fc_metric      = 1024,
1678                 .fc_ifindex     = dev->ifindex,
1679                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1680                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1681         };
1682
1683         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1684
1685         ip6_route_add(&cfg);
1686
1687         return rt6_get_dflt_router(gwaddr, dev);
1688 }
1689
1690 void rt6_purge_dflt_routers(void)
1691 {
1692         struct rt6_info *rt;
1693         struct fib6_table *table;
1694
1695         /* NOTE: Keep consistent with rt6_get_dflt_router */
1696         table = fib6_get_table(RT6_TABLE_DFLT);
1697         if (table == NULL)
1698                 return;
1699
1700 restart:
1701         read_lock_bh(&table->tb6_lock);
1702         for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1703                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1704                         dst_hold(&rt->u.dst);
1705                         read_unlock_bh(&table->tb6_lock);
1706                         ip6_del_rt(rt);
1707                         goto restart;
1708                 }
1709         }
1710         read_unlock_bh(&table->tb6_lock);
1711 }
1712
1713 static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
1714                                  struct fib6_config *cfg)
1715 {
1716         memset(cfg, 0, sizeof(*cfg));
1717
1718         cfg->fc_table = RT6_TABLE_MAIN;
1719         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1720         cfg->fc_metric = rtmsg->rtmsg_metric;
1721         cfg->fc_expires = rtmsg->rtmsg_info;
1722         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1723         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1724         cfg->fc_flags = rtmsg->rtmsg_flags;
1725
1726         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1727         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1728         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1729 }
1730
1731 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1732 {
1733         struct fib6_config cfg;
1734         struct in6_rtmsg rtmsg;
1735         int err;
1736
1737         switch(cmd) {
1738         case SIOCADDRT:         /* Add a route */
1739         case SIOCDELRT:         /* Delete a route */
1740                 if (!capable(CAP_NET_ADMIN))
1741                         return -EPERM;
1742                 err = copy_from_user(&rtmsg, arg,
1743                                      sizeof(struct in6_rtmsg));
1744                 if (err)
1745                         return -EFAULT;
1746
1747                 rtmsg_to_fib6_config(&rtmsg, &cfg);
1748
1749                 rtnl_lock();
1750                 switch (cmd) {
1751                 case SIOCADDRT:
1752                         err = ip6_route_add(&cfg);
1753                         break;
1754                 case SIOCDELRT:
1755                         err = ip6_route_del(&cfg);
1756                         break;
1757                 default:
1758                         err = -EINVAL;
1759                 }
1760                 rtnl_unlock();
1761
1762                 return err;
1763         };
1764
1765         return -EINVAL;
1766 }
1767
1768 /*
1769  *      Drop the packet on the floor
1770  */
1771
1772 static inline int ip6_pkt_drop(struct sk_buff *skb, int code,
1773                                int ipstats_mib_noroutes)
1774 {
1775         int type;
1776         switch (ipstats_mib_noroutes) {
1777         case IPSTATS_MIB_INNOROUTES:
1778                 type = ipv6_addr_type(&skb->nh.ipv6h->daddr);
1779                 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1780                         IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1781                         break;
1782                 }
1783                 /* FALLTHROUGH */
1784         case IPSTATS_MIB_OUTNOROUTES:
1785                 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1786                 break;
1787         }
1788         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1789         kfree_skb(skb);
1790         return 0;
1791 }
1792
1793 static int ip6_pkt_discard(struct sk_buff *skb)
1794 {
1795         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1796 }
1797
1798 static int ip6_pkt_discard_out(struct sk_buff *skb)
1799 {
1800         skb->dev = skb->dst->dev;
1801         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1802 }
1803
1804 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1805
1806 static int ip6_pkt_prohibit(struct sk_buff *skb)
1807 {
1808         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1809 }
1810
1811 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1812 {
1813         skb->dev = skb->dst->dev;
1814         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1815 }
1816
1817 static int ip6_pkt_blk_hole(struct sk_buff *skb)
1818 {
1819         kfree_skb(skb);
1820         return 0;
1821 }
1822
1823 #endif
1824
1825 /*
1826  *      Allocate a dst for local (unicast / anycast) address.
1827  */
1828
1829 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1830                                     const struct in6_addr *addr,
1831                                     int anycast)
1832 {
1833         struct rt6_info *rt = ip6_dst_alloc();
1834
1835         if (rt == NULL)
1836                 return ERR_PTR(-ENOMEM);
1837
1838         dev_hold(&loopback_dev);
1839         in6_dev_hold(idev);
1840
1841         rt->u.dst.flags = DST_HOST;
1842         rt->u.dst.input = ip6_input;
1843         rt->u.dst.output = ip6_output;
1844         rt->rt6i_dev = &loopback_dev;
1845         rt->rt6i_idev = idev;
1846         rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1847         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1848         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1849         rt->u.dst.obsolete = -1;
1850
1851         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1852         if (anycast)
1853                 rt->rt6i_flags |= RTF_ANYCAST;
1854         else
1855                 rt->rt6i_flags |= RTF_LOCAL;
1856         rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1857         if (rt->rt6i_nexthop == NULL) {
1858                 dst_free(&rt->u.dst);
1859                 return ERR_PTR(-ENOMEM);
1860         }
1861
1862         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1863         rt->rt6i_dst.plen = 128;
1864         rt->rt6i_table = fib6_get_table(RT6_TABLE_LOCAL);
1865
1866         atomic_set(&rt->u.dst.__refcnt, 1);
1867
1868         return rt;
1869 }
1870
1871 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1872 {
1873         if (((void*)rt->rt6i_dev == arg || arg == NULL) &&
1874             rt != &ip6_null_entry) {
1875                 RT6_TRACE("deleted by ifdown %p\n", rt);
1876                 return -1;
1877         }
1878         return 0;
1879 }
1880
1881 void rt6_ifdown(struct net_device *dev)
1882 {
1883         fib6_clean_all(fib6_ifdown, 0, dev);
1884 }
1885
1886 struct rt6_mtu_change_arg
1887 {
1888         struct net_device *dev;
1889         unsigned mtu;
1890 };
1891
1892 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1893 {
1894         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1895         struct inet6_dev *idev;
1896
1897         /* In IPv6 pmtu discovery is not optional,
1898            so that RTAX_MTU lock cannot disable it.
1899            We still use this lock to block changes
1900            caused by addrconf/ndisc.
1901         */
1902
1903         idev = __in6_dev_get(arg->dev);
1904         if (idev == NULL)
1905                 return 0;
1906
1907         /* For administrative MTU increase, there is no way to discover
1908            IPv6 PMTU increase, so PMTU increase should be updated here.
1909            Since RFC 1981 doesn't include administrative MTU increase
1910            update PMTU increase is a MUST. (i.e. jumbo frame)
1911          */
1912         /*
1913            If new MTU is less than route PMTU, this new MTU will be the
1914            lowest MTU in the path, update the route PMTU to reflect PMTU
1915            decreases; if new MTU is greater than route PMTU, and the
1916            old MTU is the lowest MTU in the path, update the route PMTU
1917            to reflect the increase. In this case if the other nodes' MTU
1918            also have the lowest MTU, TOO BIG MESSAGE will be lead to
1919            PMTU discouvery.
1920          */
1921         if (rt->rt6i_dev == arg->dev &&
1922             !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1923             (dst_mtu(&rt->u.dst) > arg->mtu ||
1924              (dst_mtu(&rt->u.dst) < arg->mtu &&
1925               dst_mtu(&rt->u.dst) == idev->cnf.mtu6)))
1926                 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1927         rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
1928         return 0;
1929 }
1930
1931 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1932 {
1933         struct rt6_mtu_change_arg arg = {
1934                 .dev = dev,
1935                 .mtu = mtu,
1936         };
1937
1938         fib6_clean_all(rt6_mtu_change_route, 0, &arg);
1939 }
1940
1941 static struct nla_policy rtm_ipv6_policy[RTA_MAX+1] __read_mostly = {
1942         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
1943         [RTA_OIF]               = { .type = NLA_U32 },
1944         [RTA_IIF]               = { .type = NLA_U32 },
1945         [RTA_PRIORITY]          = { .type = NLA_U32 },
1946         [RTA_METRICS]           = { .type = NLA_NESTED },
1947 };
1948
1949 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1950                               struct fib6_config *cfg)
1951 {
1952         struct rtmsg *rtm;
1953         struct nlattr *tb[RTA_MAX+1];
1954         int err;
1955
1956         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
1957         if (err < 0)
1958                 goto errout;
1959
1960         err = -EINVAL;
1961         rtm = nlmsg_data(nlh);
1962         memset(cfg, 0, sizeof(*cfg));
1963
1964         cfg->fc_table = rtm->rtm_table;
1965         cfg->fc_dst_len = rtm->rtm_dst_len;
1966         cfg->fc_src_len = rtm->rtm_src_len;
1967         cfg->fc_flags = RTF_UP;
1968         cfg->fc_protocol = rtm->rtm_protocol;
1969
1970         if (rtm->rtm_type == RTN_UNREACHABLE)
1971                 cfg->fc_flags |= RTF_REJECT;
1972
1973         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
1974         cfg->fc_nlinfo.nlh = nlh;
1975
1976         if (tb[RTA_GATEWAY]) {
1977                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
1978                 cfg->fc_flags |= RTF_GATEWAY;
1979         }
1980
1981         if (tb[RTA_DST]) {
1982                 int plen = (rtm->rtm_dst_len + 7) >> 3;
1983
1984                 if (nla_len(tb[RTA_DST]) < plen)
1985                         goto errout;
1986
1987                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1988         }
1989
1990         if (tb[RTA_SRC]) {
1991                 int plen = (rtm->rtm_src_len + 7) >> 3;
1992
1993                 if (nla_len(tb[RTA_SRC]) < plen)
1994                         goto errout;
1995
1996                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1997         }
1998
1999         if (tb[RTA_OIF])
2000                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2001
2002         if (tb[RTA_PRIORITY])
2003                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2004
2005         if (tb[RTA_METRICS]) {
2006                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2007                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2008         }
2009
2010         if (tb[RTA_TABLE])
2011                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2012
2013         err = 0;
2014 errout:
2015         return err;
2016 }
2017
2018 int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2019 {
2020         struct fib6_config cfg;
2021         int err;
2022
2023         err = rtm_to_fib6_config(skb, nlh, &cfg);
2024         if (err < 0)
2025                 return err;
2026
2027         return ip6_route_del(&cfg);
2028 }
2029
2030 int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2031 {
2032         struct fib6_config cfg;
2033         int err;
2034
2035         err = rtm_to_fib6_config(skb, nlh, &cfg);
2036         if (err < 0)
2037                 return err;
2038
2039         return ip6_route_add(&cfg);
2040 }
2041
2042 static inline size_t rt6_nlmsg_size(void)
2043 {
2044         return NLMSG_ALIGN(sizeof(struct rtmsg))
2045                + nla_total_size(16) /* RTA_SRC */
2046                + nla_total_size(16) /* RTA_DST */
2047                + nla_total_size(16) /* RTA_GATEWAY */
2048                + nla_total_size(16) /* RTA_PREFSRC */
2049                + nla_total_size(4) /* RTA_TABLE */
2050                + nla_total_size(4) /* RTA_IIF */
2051                + nla_total_size(4) /* RTA_OIF */
2052                + nla_total_size(4) /* RTA_PRIORITY */
2053                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2054                + nla_total_size(sizeof(struct rta_cacheinfo));
2055 }
2056
2057 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2058                          struct in6_addr *dst, struct in6_addr *src,
2059                          int iif, int type, u32 pid, u32 seq,
2060                          int prefix, unsigned int flags)
2061 {
2062         struct rtmsg *rtm;
2063         struct nlmsghdr *nlh;
2064         long expires;
2065         u32 table;
2066
2067         if (prefix) {   /* user wants prefix routes only */
2068                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2069                         /* success since this is not a prefix route */
2070                         return 1;
2071                 }
2072         }
2073
2074         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2075         if (nlh == NULL)
2076                 return -EMSGSIZE;
2077
2078         rtm = nlmsg_data(nlh);
2079         rtm->rtm_family = AF_INET6;
2080         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2081         rtm->rtm_src_len = rt->rt6i_src.plen;
2082         rtm->rtm_tos = 0;
2083         if (rt->rt6i_table)
2084                 table = rt->rt6i_table->tb6_id;
2085         else
2086                 table = RT6_TABLE_UNSPEC;
2087         rtm->rtm_table = table;
2088         NLA_PUT_U32(skb, RTA_TABLE, table);
2089         if (rt->rt6i_flags&RTF_REJECT)
2090                 rtm->rtm_type = RTN_UNREACHABLE;
2091         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2092                 rtm->rtm_type = RTN_LOCAL;
2093         else
2094                 rtm->rtm_type = RTN_UNICAST;
2095         rtm->rtm_flags = 0;
2096         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2097         rtm->rtm_protocol = rt->rt6i_protocol;
2098         if (rt->rt6i_flags&RTF_DYNAMIC)
2099                 rtm->rtm_protocol = RTPROT_REDIRECT;
2100         else if (rt->rt6i_flags & RTF_ADDRCONF)
2101                 rtm->rtm_protocol = RTPROT_KERNEL;
2102         else if (rt->rt6i_flags&RTF_DEFAULT)
2103                 rtm->rtm_protocol = RTPROT_RA;
2104
2105         if (rt->rt6i_flags&RTF_CACHE)
2106                 rtm->rtm_flags |= RTM_F_CLONED;
2107
2108         if (dst) {
2109                 NLA_PUT(skb, RTA_DST, 16, dst);
2110                 rtm->rtm_dst_len = 128;
2111         } else if (rtm->rtm_dst_len)
2112                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2113 #ifdef CONFIG_IPV6_SUBTREES
2114         if (src) {
2115                 NLA_PUT(skb, RTA_SRC, 16, src);
2116                 rtm->rtm_src_len = 128;
2117         } else if (rtm->rtm_src_len)
2118                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2119 #endif
2120         if (iif)
2121                 NLA_PUT_U32(skb, RTA_IIF, iif);
2122         else if (dst) {
2123                 struct in6_addr saddr_buf;
2124                 if (ipv6_get_saddr(&rt->u.dst, dst, &saddr_buf) == 0)
2125                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2126         }
2127
2128         if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2129                 goto nla_put_failure;
2130
2131         if (rt->u.dst.neighbour)
2132                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2133
2134         if (rt->u.dst.dev)
2135                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2136
2137         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2138
2139         expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2140         if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2141                                expires, rt->u.dst.error) < 0)
2142                 goto nla_put_failure;
2143
2144         return nlmsg_end(skb, nlh);
2145
2146 nla_put_failure:
2147         nlmsg_cancel(skb, nlh);
2148         return -EMSGSIZE;
2149 }
2150
2151 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2152 {
2153         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2154         int prefix;
2155
2156         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2157                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2158                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2159         } else
2160                 prefix = 0;
2161
2162         return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2163                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2164                      prefix, NLM_F_MULTI);
2165 }
2166
2167 int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2168 {
2169         struct nlattr *tb[RTA_MAX+1];
2170         struct rt6_info *rt;
2171         struct sk_buff *skb;
2172         struct rtmsg *rtm;
2173         struct flowi fl;
2174         int err, iif = 0;
2175
2176         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2177         if (err < 0)
2178                 goto errout;
2179
2180         err = -EINVAL;
2181         memset(&fl, 0, sizeof(fl));
2182
2183         if (tb[RTA_SRC]) {
2184                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2185                         goto errout;
2186
2187                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2188         }
2189
2190         if (tb[RTA_DST]) {
2191                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2192                         goto errout;
2193
2194                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2195         }
2196
2197         if (tb[RTA_IIF])
2198                 iif = nla_get_u32(tb[RTA_IIF]);
2199
2200         if (tb[RTA_OIF])
2201                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2202
2203         if (iif) {
2204                 struct net_device *dev;
2205                 dev = __dev_get_by_index(iif);
2206                 if (!dev) {
2207                         err = -ENODEV;
2208                         goto errout;
2209                 }
2210         }
2211
2212         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2213         if (skb == NULL) {
2214                 err = -ENOBUFS;
2215                 goto errout;
2216         }
2217
2218         /* Reserve room for dummy headers, this skb can pass
2219            through good chunk of routing engine.
2220          */
2221         skb->mac.raw = skb->data;
2222         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2223
2224         rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
2225         skb->dst = &rt->u.dst;
2226
2227         err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2228                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2229                             nlh->nlmsg_seq, 0, 0);
2230         if (err < 0) {
2231                 kfree_skb(skb);
2232                 goto errout;
2233         }
2234
2235         err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
2236 errout:
2237         return err;
2238 }
2239
2240 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2241 {
2242         struct sk_buff *skb;
2243         u32 pid = 0, seq = 0;
2244         struct nlmsghdr *nlh = NULL;
2245         int err = -ENOBUFS;
2246
2247         if (info) {
2248                 pid = info->pid;
2249                 nlh = info->nlh;
2250                 if (nlh)
2251                         seq = nlh->nlmsg_seq;
2252         }
2253
2254         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2255         if (skb == NULL)
2256                 goto errout;
2257
2258         err = rt6_fill_node(skb, rt, NULL, NULL, 0, event, pid, seq, 0, 0);
2259         if (err < 0) {
2260                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2261                 WARN_ON(err == -EMSGSIZE);
2262                 kfree_skb(skb);
2263                 goto errout;
2264         }
2265         err = rtnl_notify(skb, pid, RTNLGRP_IPV6_ROUTE, nlh, gfp_any());
2266 errout:
2267         if (err < 0)
2268                 rtnl_set_sk_err(RTNLGRP_IPV6_ROUTE, err);
2269 }
2270
2271 /*
2272  *      /proc
2273  */
2274
2275 #ifdef CONFIG_PROC_FS
2276
2277 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2278
2279 struct rt6_proc_arg
2280 {
2281         char *buffer;
2282         int offset;
2283         int length;
2284         int skip;
2285         int len;
2286 };
2287
2288 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2289 {
2290         struct rt6_proc_arg *arg = (struct rt6_proc_arg *) p_arg;
2291
2292         if (arg->skip < arg->offset / RT6_INFO_LEN) {
2293                 arg->skip++;
2294                 return 0;
2295         }
2296
2297         if (arg->len >= arg->length)
2298                 return 0;
2299
2300         arg->len += sprintf(arg->buffer + arg->len,
2301                             NIP6_SEQFMT " %02x ",
2302                             NIP6(rt->rt6i_dst.addr),
2303                             rt->rt6i_dst.plen);
2304
2305 #ifdef CONFIG_IPV6_SUBTREES
2306         arg->len += sprintf(arg->buffer + arg->len,
2307                             NIP6_SEQFMT " %02x ",
2308                             NIP6(rt->rt6i_src.addr),
2309                             rt->rt6i_src.plen);
2310 #else
2311         arg->len += sprintf(arg->buffer + arg->len,
2312                             "00000000000000000000000000000000 00 ");
2313 #endif
2314
2315         if (rt->rt6i_nexthop) {
2316                 arg->len += sprintf(arg->buffer + arg->len,
2317                                     NIP6_SEQFMT,
2318                                     NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2319         } else {
2320                 arg->len += sprintf(arg->buffer + arg->len,
2321                                     "00000000000000000000000000000000");
2322         }
2323         arg->len += sprintf(arg->buffer + arg->len,
2324                             " %08x %08x %08x %08x %8s\n",
2325                             rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2326                             rt->u.dst.__use, rt->rt6i_flags,
2327                             rt->rt6i_dev ? rt->rt6i_dev->name : "");
2328         return 0;
2329 }
2330
2331 static int rt6_proc_info(char *buffer, char **start, off_t offset, int length)
2332 {
2333         struct rt6_proc_arg arg = {
2334                 .buffer = buffer,
2335                 .offset = offset,
2336                 .length = length,
2337         };
2338
2339         fib6_clean_all(rt6_info_route, 0, &arg);
2340
2341         *start = buffer;
2342         if (offset)
2343                 *start += offset % RT6_INFO_LEN;
2344
2345         arg.len -= offset % RT6_INFO_LEN;
2346
2347         if (arg.len > length)
2348                 arg.len = length;
2349         if (arg.len < 0)
2350                 arg.len = 0;
2351
2352         return arg.len;
2353 }
2354
2355 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2356 {
2357         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2358                       rt6_stats.fib_nodes, rt6_stats.fib_route_nodes,
2359                       rt6_stats.fib_rt_alloc, rt6_stats.fib_rt_entries,
2360                       rt6_stats.fib_rt_cache,
2361                       atomic_read(&ip6_dst_ops.entries),
2362                       rt6_stats.fib_discarded_routes);
2363
2364         return 0;
2365 }
2366
2367 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2368 {
2369         return single_open(file, rt6_stats_seq_show, NULL);
2370 }
2371
2372 static const struct file_operations rt6_stats_seq_fops = {
2373         .owner   = THIS_MODULE,
2374         .open    = rt6_stats_seq_open,
2375         .read    = seq_read,
2376         .llseek  = seq_lseek,
2377         .release = single_release,
2378 };
2379 #endif  /* CONFIG_PROC_FS */
2380
2381 #ifdef CONFIG_SYSCTL
2382
2383 static int flush_delay;
2384
2385 static
2386 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2387                               void __user *buffer, size_t *lenp, loff_t *ppos)
2388 {
2389         if (write) {
2390                 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2391                 fib6_run_gc(flush_delay <= 0 ? ~0UL : (unsigned long)flush_delay);
2392                 return 0;
2393         } else
2394                 return -EINVAL;
2395 }
2396
2397 ctl_table ipv6_route_table[] = {
2398         {
2399                 .ctl_name       =       NET_IPV6_ROUTE_FLUSH,
2400                 .procname       =       "flush",
2401                 .data           =       &flush_delay,
2402                 .maxlen         =       sizeof(int),
2403                 .mode           =       0200,
2404                 .proc_handler   =       &ipv6_sysctl_rtcache_flush
2405         },
2406         {
2407                 .ctl_name       =       NET_IPV6_ROUTE_GC_THRESH,
2408                 .procname       =       "gc_thresh",
2409                 .data           =       &ip6_dst_ops.gc_thresh,
2410                 .maxlen         =       sizeof(int),
2411                 .mode           =       0644,
2412                 .proc_handler   =       &proc_dointvec,
2413         },
2414         {
2415                 .ctl_name       =       NET_IPV6_ROUTE_MAX_SIZE,
2416                 .procname       =       "max_size",
2417                 .data           =       &ip6_rt_max_size,
2418                 .maxlen         =       sizeof(int),
2419                 .mode           =       0644,
2420                 .proc_handler   =       &proc_dointvec,
2421         },
2422         {
2423                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2424                 .procname       =       "gc_min_interval",
2425                 .data           =       &ip6_rt_gc_min_interval,
2426                 .maxlen         =       sizeof(int),
2427                 .mode           =       0644,
2428                 .proc_handler   =       &proc_dointvec_jiffies,
2429                 .strategy       =       &sysctl_jiffies,
2430         },
2431         {
2432                 .ctl_name       =       NET_IPV6_ROUTE_GC_TIMEOUT,
2433                 .procname       =       "gc_timeout",
2434                 .data           =       &ip6_rt_gc_timeout,
2435                 .maxlen         =       sizeof(int),
2436                 .mode           =       0644,
2437                 .proc_handler   =       &proc_dointvec_jiffies,
2438                 .strategy       =       &sysctl_jiffies,
2439         },
2440         {
2441                 .ctl_name       =       NET_IPV6_ROUTE_GC_INTERVAL,
2442                 .procname       =       "gc_interval",
2443                 .data           =       &ip6_rt_gc_interval,
2444                 .maxlen         =       sizeof(int),
2445                 .mode           =       0644,
2446                 .proc_handler   =       &proc_dointvec_jiffies,
2447                 .strategy       =       &sysctl_jiffies,
2448         },
2449         {
2450                 .ctl_name       =       NET_IPV6_ROUTE_GC_ELASTICITY,
2451                 .procname       =       "gc_elasticity",
2452                 .data           =       &ip6_rt_gc_elasticity,
2453                 .maxlen         =       sizeof(int),
2454                 .mode           =       0644,
2455                 .proc_handler   =       &proc_dointvec_jiffies,
2456                 .strategy       =       &sysctl_jiffies,
2457         },
2458         {
2459                 .ctl_name       =       NET_IPV6_ROUTE_MTU_EXPIRES,
2460                 .procname       =       "mtu_expires",
2461                 .data           =       &ip6_rt_mtu_expires,
2462                 .maxlen         =       sizeof(int),
2463                 .mode           =       0644,
2464                 .proc_handler   =       &proc_dointvec_jiffies,
2465                 .strategy       =       &sysctl_jiffies,
2466         },
2467         {
2468                 .ctl_name       =       NET_IPV6_ROUTE_MIN_ADVMSS,
2469                 .procname       =       "min_adv_mss",
2470                 .data           =       &ip6_rt_min_advmss,
2471                 .maxlen         =       sizeof(int),
2472                 .mode           =       0644,
2473                 .proc_handler   =       &proc_dointvec_jiffies,
2474                 .strategy       =       &sysctl_jiffies,
2475         },
2476         {
2477                 .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2478                 .procname       =       "gc_min_interval_ms",
2479                 .data           =       &ip6_rt_gc_min_interval,
2480                 .maxlen         =       sizeof(int),
2481                 .mode           =       0644,
2482                 .proc_handler   =       &proc_dointvec_ms_jiffies,
2483                 .strategy       =       &sysctl_ms_jiffies,
2484         },
2485         { .ctl_name = 0 }
2486 };
2487
2488 #endif
2489
2490 void __init ip6_route_init(void)
2491 {
2492         struct proc_dir_entry *p;
2493
2494         ip6_dst_ops.kmem_cachep =
2495                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2496                                   SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
2497         fib6_init();
2498 #ifdef  CONFIG_PROC_FS
2499         p = proc_net_create("ipv6_route", 0, rt6_proc_info);
2500         if (p)
2501                 p->owner = THIS_MODULE;
2502
2503         proc_net_fops_create("rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2504 #endif
2505 #ifdef CONFIG_XFRM
2506         xfrm6_init();
2507 #endif
2508 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2509         fib6_rules_init();
2510 #endif
2511 }
2512
2513 void ip6_route_cleanup(void)
2514 {
2515 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2516         fib6_rules_cleanup();
2517 #endif
2518 #ifdef CONFIG_PROC_FS
2519         proc_net_remove("ipv6_route");
2520         proc_net_remove("rt6_stats");
2521 #endif
2522 #ifdef CONFIG_XFRM
2523         xfrm6_fini();
2524 #endif
2525         rt6_ifdown(NULL);
2526         fib6_gc_cleanup();
2527         kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2528 }