blob: de382114609b7de8d0cbf5c5447a18d4f5a89a2d [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09002 * Linux INET6 implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07003 * Forwarding Information Database
4 *
5 * Authors:
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09006 * Pedro Roque <roque@di.fc.ul.pt>
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14/*
15 * Changes:
16 * Yuji SEKIYA @USAGI: Support default route on router node;
17 * remove ip6_null_entry from the top of
18 * routing table.
YOSHIFUJI Hideakic0bece92006-08-23 17:23:25 -070019 * Ville Nuorvala: Fixed routing subtrees.
Linus Torvalds1da177e2005-04-16 15:20:36 -070020 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/errno.h>
22#include <linux/types.h>
23#include <linux/net.h>
24#include <linux/route.h>
25#include <linux/netdevice.h>
26#include <linux/in6.h>
27#include <linux/init.h>
Thomas Grafc71099a2006-08-04 23:20:06 -070028#include <linux/list.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090029#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030
31#ifdef CONFIG_PROC_FS
32#include <linux/proc_fs.h>
33#endif
34
35#include <net/ipv6.h>
36#include <net/ndisc.h>
37#include <net/addrconf.h>
38
39#include <net/ip6_fib.h>
40#include <net/ip6_route.h>
41
42#define RT6_DEBUG 2
43
44#if RT6_DEBUG >= 3
45#define RT6_TRACE(x...) printk(KERN_DEBUG x)
46#else
47#define RT6_TRACE(x...) do { ; } while (0)
48#endif
49
Christoph Lametere18b8902006-12-06 20:33:20 -080050static struct kmem_cache * fib6_node_kmem __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070051
52enum fib_walk_state_t
53{
54#ifdef CONFIG_IPV6_SUBTREES
55 FWS_S,
56#endif
57 FWS_L,
58 FWS_R,
59 FWS_C,
60 FWS_U
61};
62
63struct fib6_cleaner_t
64{
65 struct fib6_walker_t w;
Benjamin Theryec7d43c2008-03-03 23:31:57 -080066 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 int (*func)(struct rt6_info *, void *arg);
68 void *arg;
69};
70
Adrian Bunk90d41122006-08-14 23:49:16 -070071static DEFINE_RWLOCK(fib6_walker_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
73#ifdef CONFIG_IPV6_SUBTREES
74#define FWS_INIT FWS_S
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#else
76#define FWS_INIT FWS_L
Linus Torvalds1da177e2005-04-16 15:20:36 -070077#endif
78
Benjamin Theryec7d43c2008-03-03 23:31:57 -080079static void fib6_prune_clones(struct net *net, struct fib6_node *fn,
80 struct rt6_info *rt);
Daniel Lezcano8ed67782008-03-04 13:48:30 -080081static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn);
82static struct fib6_node *fib6_repair_tree(struct net *net, struct fib6_node *fn);
Adrian Bunk90d41122006-08-14 23:49:16 -070083static int fib6_walk(struct fib6_walker_t *w);
84static int fib6_walk_continue(struct fib6_walker_t *w);
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
86/*
87 * A routing update causes an increase of the serial number on the
88 * affected subtree. This allows for cached routes to be asynchronously
89 * tested when modifications are made to the destination cache as a
90 * result of redirects, path MTU changes, etc.
91 */
92
93static __u32 rt_sernum;
94
Daniel Lezcano5b7c9312008-03-03 23:28:58 -080095static void fib6_gc_timer_cb(unsigned long arg);
96
Alexey Dobriyanbbef49d2010-02-18 08:13:30 +000097static LIST_HEAD(fib6_walkers);
98#define FOR_WALKERS(w) list_for_each_entry(w, &fib6_walkers, lh)
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Adrian Bunk90d41122006-08-14 23:49:16 -0700100static inline void fib6_walker_link(struct fib6_walker_t *w)
101{
102 write_lock_bh(&fib6_walker_lock);
Alexey Dobriyanbbef49d2010-02-18 08:13:30 +0000103 list_add(&w->lh, &fib6_walkers);
Adrian Bunk90d41122006-08-14 23:49:16 -0700104 write_unlock_bh(&fib6_walker_lock);
105}
106
107static inline void fib6_walker_unlink(struct fib6_walker_t *w)
108{
109 write_lock_bh(&fib6_walker_lock);
Alexey Dobriyanbbef49d2010-02-18 08:13:30 +0000110 list_del(&w->lh);
Adrian Bunk90d41122006-08-14 23:49:16 -0700111 write_unlock_bh(&fib6_walker_lock);
112}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113static __inline__ u32 fib6_new_sernum(void)
114{
115 u32 n = ++rt_sernum;
116 if ((__s32)n <= 0)
117 rt_sernum = n = 1;
118 return n;
119}
120
121/*
122 * Auxiliary address test functions for the radix tree.
123 *
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900124 * These assume a 32bit processor (although it will work on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 * 64bit processors)
126 */
127
128/*
129 * test bit
130 */
YOSHIFUJI Hideaki / 吉藤英明02cdce52010-03-27 01:24:16 +0000131#if defined(__LITTLE_ENDIAN)
132# define BITOP_BE32_SWIZZLE (0x1F & ~7)
133#else
134# define BITOP_BE32_SWIZZLE 0
135#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136
Al Viroe69a4adc2006-11-14 20:56:00 -0800137static __inline__ __be32 addr_bit_set(void *token, int fn_bit)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138{
Al Viroe69a4adc2006-11-14 20:56:00 -0800139 __be32 *addr = token;
YOSHIFUJI Hideaki / 吉藤英明02cdce52010-03-27 01:24:16 +0000140 /*
141 * Here,
142 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
143 * is optimized version of
144 * htonl(1 << ((~fn_bit)&0x1F))
145 * See include/asm-generic/bitops/le.h.
146 */
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700147 return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) &
148 addr[fn_bit >> 5];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149}
150
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151static __inline__ struct fib6_node * node_alloc(void)
152{
153 struct fib6_node *fn;
154
Robert P. J. Dayc3762222007-02-10 01:45:03 -0800155 fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156
157 return fn;
158}
159
160static __inline__ void node_free(struct fib6_node * fn)
161{
162 kmem_cache_free(fib6_node_kmem, fn);
163}
164
165static __inline__ void rt6_release(struct rt6_info *rt)
166{
167 if (atomic_dec_and_test(&rt->rt6i_ref))
Changli Gaod8d1f302010-06-10 23:31:35 -0700168 dst_free(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169}
170
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800171static void fib6_link_table(struct net *net, struct fib6_table *tb)
Patrick McHardy1b43af52006-08-10 23:11:17 -0700172{
173 unsigned int h;
174
Thomas Graf375216a2006-10-21 20:20:54 -0700175 /*
176 * Initialize table lock at a single place to give lockdep a key,
177 * tables aren't visible prior to being linked to the list.
178 */
179 rwlock_init(&tb->tb6_lock);
180
Neil Hormana33bc5c2009-07-30 18:52:15 -0700181 h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1);
Patrick McHardy1b43af52006-08-10 23:11:17 -0700182
183 /*
184 * No protection necessary, this is the only list mutatation
185 * operation, tables never disappear once they exist.
186 */
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800187 hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]);
Patrick McHardy1b43af52006-08-10 23:11:17 -0700188}
189
190#ifdef CONFIG_IPV6_MULTIPLE_TABLES
Daniel Lezcanoe0b855902008-03-03 23:24:31 -0800191
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800192static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
Thomas Grafc71099a2006-08-04 23:20:06 -0700193{
194 struct fib6_table *table;
195
196 table = kzalloc(sizeof(*table), GFP_ATOMIC);
197 if (table != NULL) {
198 table->tb6_id = id;
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800199 table->tb6_root.leaf = net->ipv6.ip6_null_entry;
Thomas Grafc71099a2006-08-04 23:20:06 -0700200 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
201 }
202
203 return table;
204}
205
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800206struct fib6_table *fib6_new_table(struct net *net, u32 id)
Thomas Grafc71099a2006-08-04 23:20:06 -0700207{
208 struct fib6_table *tb;
209
210 if (id == 0)
211 id = RT6_TABLE_MAIN;
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800212 tb = fib6_get_table(net, id);
Thomas Grafc71099a2006-08-04 23:20:06 -0700213 if (tb)
214 return tb;
215
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800216 tb = fib6_alloc_table(net, id);
Thomas Grafc71099a2006-08-04 23:20:06 -0700217 if (tb != NULL)
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800218 fib6_link_table(net, tb);
Thomas Grafc71099a2006-08-04 23:20:06 -0700219
220 return tb;
221}
222
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800223struct fib6_table *fib6_get_table(struct net *net, u32 id)
Thomas Grafc71099a2006-08-04 23:20:06 -0700224{
225 struct fib6_table *tb;
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800226 struct hlist_head *head;
Thomas Grafc71099a2006-08-04 23:20:06 -0700227 struct hlist_node *node;
228 unsigned int h;
229
230 if (id == 0)
231 id = RT6_TABLE_MAIN;
Neil Hormana33bc5c2009-07-30 18:52:15 -0700232 h = id & (FIB6_TABLE_HASHSZ - 1);
Thomas Grafc71099a2006-08-04 23:20:06 -0700233 rcu_read_lock();
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800234 head = &net->ipv6.fib_table_hash[h];
235 hlist_for_each_entry_rcu(tb, node, head, tb6_hlist) {
Thomas Grafc71099a2006-08-04 23:20:06 -0700236 if (tb->tb6_id == id) {
237 rcu_read_unlock();
238 return tb;
239 }
240 }
241 rcu_read_unlock();
242
243 return NULL;
244}
245
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000246static void __net_init fib6_tables_init(struct net *net)
Thomas Grafc71099a2006-08-04 23:20:06 -0700247{
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800248 fib6_link_table(net, net->ipv6.fib6_main_tbl);
249 fib6_link_table(net, net->ipv6.fib6_local_tbl);
Thomas Grafc71099a2006-08-04 23:20:06 -0700250}
Thomas Grafc71099a2006-08-04 23:20:06 -0700251#else
252
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800253struct fib6_table *fib6_new_table(struct net *net, u32 id)
Thomas Grafc71099a2006-08-04 23:20:06 -0700254{
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800255 return fib6_get_table(net, id);
Thomas Grafc71099a2006-08-04 23:20:06 -0700256}
257
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800258struct fib6_table *fib6_get_table(struct net *net, u32 id)
Thomas Grafc71099a2006-08-04 23:20:06 -0700259{
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800260 return net->ipv6.fib6_main_tbl;
Thomas Grafc71099a2006-08-04 23:20:06 -0700261}
262
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800263struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi *fl,
264 int flags, pol_lookup_t lookup)
Thomas Grafc71099a2006-08-04 23:20:06 -0700265{
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800266 return (struct dst_entry *) lookup(net, net->ipv6.fib6_main_tbl, fl, flags);
Thomas Grafc71099a2006-08-04 23:20:06 -0700267}
268
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000269static void __net_init fib6_tables_init(struct net *net)
Thomas Grafc71099a2006-08-04 23:20:06 -0700270{
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800271 fib6_link_table(net, net->ipv6.fib6_main_tbl);
Thomas Grafc71099a2006-08-04 23:20:06 -0700272}
273
274#endif
275
Patrick McHardy1b43af52006-08-10 23:11:17 -0700276static int fib6_dump_node(struct fib6_walker_t *w)
277{
278 int res;
279 struct rt6_info *rt;
280
Changli Gaod8d1f302010-06-10 23:31:35 -0700281 for (rt = w->leaf; rt; rt = rt->dst.rt6_next) {
Patrick McHardy1b43af52006-08-10 23:11:17 -0700282 res = rt6_dump_route(rt, w->args);
283 if (res < 0) {
284 /* Frame is full, suspend walking */
285 w->leaf = rt;
286 return 1;
287 }
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700288 WARN_ON(res == 0);
Patrick McHardy1b43af52006-08-10 23:11:17 -0700289 }
290 w->leaf = NULL;
291 return 0;
292}
293
294static void fib6_dump_end(struct netlink_callback *cb)
295{
296 struct fib6_walker_t *w = (void*)cb->args[2];
297
298 if (w) {
Herbert Xu7891cc82009-01-13 22:17:51 -0800299 if (cb->args[4]) {
300 cb->args[4] = 0;
301 fib6_walker_unlink(w);
302 }
Patrick McHardy1b43af52006-08-10 23:11:17 -0700303 cb->args[2] = 0;
304 kfree(w);
305 }
306 cb->done = (void*)cb->args[3];
307 cb->args[1] = 3;
308}
309
310static int fib6_dump_done(struct netlink_callback *cb)
311{
312 fib6_dump_end(cb);
313 return cb->done ? cb->done(cb) : 0;
314}
315
316static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
317 struct netlink_callback *cb)
318{
319 struct fib6_walker_t *w;
320 int res;
321
322 w = (void *)cb->args[2];
323 w->root = &table->tb6_root;
324
325 if (cb->args[4] == 0) {
Patrick McHardy2bec5a32010-02-08 05:19:03 +0000326 w->count = 0;
327 w->skip = 0;
328
Patrick McHardy1b43af52006-08-10 23:11:17 -0700329 read_lock_bh(&table->tb6_lock);
330 res = fib6_walk(w);
331 read_unlock_bh(&table->tb6_lock);
Patrick McHardy2bec5a32010-02-08 05:19:03 +0000332 if (res > 0) {
Patrick McHardy1b43af52006-08-10 23:11:17 -0700333 cb->args[4] = 1;
Patrick McHardy2bec5a32010-02-08 05:19:03 +0000334 cb->args[5] = w->root->fn_sernum;
335 }
Patrick McHardy1b43af52006-08-10 23:11:17 -0700336 } else {
Patrick McHardy2bec5a32010-02-08 05:19:03 +0000337 if (cb->args[5] != w->root->fn_sernum) {
338 /* Begin at the root if the tree changed */
339 cb->args[5] = w->root->fn_sernum;
340 w->state = FWS_INIT;
341 w->node = w->root;
342 w->skip = w->count;
343 } else
344 w->skip = 0;
345
Patrick McHardy1b43af52006-08-10 23:11:17 -0700346 read_lock_bh(&table->tb6_lock);
347 res = fib6_walk_continue(w);
348 read_unlock_bh(&table->tb6_lock);
Herbert Xu7891cc82009-01-13 22:17:51 -0800349 if (res <= 0) {
350 fib6_walker_unlink(w);
351 cb->args[4] = 0;
Patrick McHardy1b43af52006-08-10 23:11:17 -0700352 }
Patrick McHardy1b43af52006-08-10 23:11:17 -0700353 }
Herbert Xu7891cc82009-01-13 22:17:51 -0800354
Patrick McHardy1b43af52006-08-10 23:11:17 -0700355 return res;
356}
357
Thomas Grafc127ea22007-03-22 11:58:32 -0700358static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
Patrick McHardy1b43af52006-08-10 23:11:17 -0700359{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900360 struct net *net = sock_net(skb->sk);
Patrick McHardy1b43af52006-08-10 23:11:17 -0700361 unsigned int h, s_h;
362 unsigned int e = 0, s_e;
363 struct rt6_rtnl_dump_arg arg;
364 struct fib6_walker_t *w;
365 struct fib6_table *tb;
366 struct hlist_node *node;
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800367 struct hlist_head *head;
Patrick McHardy1b43af52006-08-10 23:11:17 -0700368 int res = 0;
369
370 s_h = cb->args[0];
371 s_e = cb->args[1];
372
373 w = (void *)cb->args[2];
374 if (w == NULL) {
375 /* New dump:
376 *
377 * 1. hook callback destructor.
378 */
379 cb->args[3] = (long)cb->done;
380 cb->done = fib6_dump_done;
381
382 /*
383 * 2. allocate and initialize walker.
384 */
385 w = kzalloc(sizeof(*w), GFP_ATOMIC);
386 if (w == NULL)
387 return -ENOMEM;
388 w->func = fib6_dump_node;
389 cb->args[2] = (long)w;
390 }
391
392 arg.skb = skb;
393 arg.cb = cb;
Brian Haley191cd582008-08-14 15:33:21 -0700394 arg.net = net;
Patrick McHardy1b43af52006-08-10 23:11:17 -0700395 w->args = &arg;
396
Neil Hormana33bc5c2009-07-30 18:52:15 -0700397 for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) {
Patrick McHardy1b43af52006-08-10 23:11:17 -0700398 e = 0;
Daniel Lezcano58f09b72008-03-03 23:25:27 -0800399 head = &net->ipv6.fib_table_hash[h];
400 hlist_for_each_entry(tb, node, head, tb6_hlist) {
Patrick McHardy1b43af52006-08-10 23:11:17 -0700401 if (e < s_e)
402 goto next;
403 res = fib6_dump_table(tb, skb, cb);
404 if (res != 0)
405 goto out;
406next:
407 e++;
408 }
409 }
410out:
411 cb->args[1] = e;
412 cb->args[0] = h;
413
414 res = res < 0 ? res : skb->len;
415 if (res <= 0)
416 fib6_dump_end(cb);
417 return res;
418}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419
420/*
421 * Routing Table
422 *
423 * return the appropriate node for a routing tree "add" operation
424 * by either creating and inserting or by returning an existing
425 * node.
426 */
427
428static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr,
429 int addrlen, int plen,
430 int offset)
431{
432 struct fib6_node *fn, *in, *ln;
433 struct fib6_node *pn = NULL;
434 struct rt6key *key;
435 int bit;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900436 __be32 dir = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 __u32 sernum = fib6_new_sernum();
438
439 RT6_TRACE("fib6_add_1\n");
440
441 /* insert node in tree */
442
443 fn = root;
444
445 do {
446 key = (struct rt6key *)((u8 *)fn->leaf + offset);
447
448 /*
449 * Prefix match
450 */
451 if (plen < fn->fn_bit ||
452 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
453 goto insert_above;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900454
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 /*
456 * Exact match ?
457 */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900458
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 if (plen == fn->fn_bit) {
460 /* clean up an intermediate node */
461 if ((fn->fn_flags & RTN_RTINFO) == 0) {
462 rt6_release(fn->leaf);
463 fn->leaf = NULL;
464 }
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 fn->fn_sernum = sernum;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900467
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 return fn;
469 }
470
471 /*
472 * We have more bits to go
473 */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900474
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 /* Try to walk down on tree. */
476 fn->fn_sernum = sernum;
477 dir = addr_bit_set(addr, fn->fn_bit);
478 pn = fn;
479 fn = dir ? fn->right: fn->left;
480 } while (fn);
481
482 /*
483 * We walked to the bottom of tree.
484 * Create new leaf node without children.
485 */
486
487 ln = node_alloc();
488
489 if (ln == NULL)
490 return NULL;
491 ln->fn_bit = plen;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900492
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 ln->parent = pn;
494 ln->fn_sernum = sernum;
495
496 if (dir)
497 pn->right = ln;
498 else
499 pn->left = ln;
500
501 return ln;
502
503
504insert_above:
505 /*
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900506 * split since we don't have a common prefix anymore or
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 * we have a less significant route.
508 * we've to insert an intermediate node on the list
509 * this new node will point to the one we need to create
510 * and the current
511 */
512
513 pn = fn->parent;
514
515 /* find 1st bit in difference between the 2 addrs.
516
YOSHIFUJI Hideaki971f3592005-11-08 09:37:56 -0800517 See comment in __ipv6_addr_diff: bit may be an invalid value,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 but if it is >= plen, the value is ignored in any case.
519 */
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900520
YOSHIFUJI Hideaki971f3592005-11-08 09:37:56 -0800521 bit = __ipv6_addr_diff(addr, &key->addr, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900523 /*
524 * (intermediate)[in]
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 * / \
526 * (new leaf node)[ln] (old node)[fn]
527 */
528 if (plen > bit) {
529 in = node_alloc();
530 ln = node_alloc();
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900531
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 if (in == NULL || ln == NULL) {
533 if (in)
534 node_free(in);
535 if (ln)
536 node_free(ln);
537 return NULL;
538 }
539
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900540 /*
541 * new intermediate node.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 * RTN_RTINFO will
543 * be off since that an address that chooses one of
544 * the branches would not match less specific routes
545 * in the other branch
546 */
547
548 in->fn_bit = bit;
549
550 in->parent = pn;
551 in->leaf = fn->leaf;
552 atomic_inc(&in->leaf->rt6i_ref);
553
554 in->fn_sernum = sernum;
555
556 /* update parent pointer */
557 if (dir)
558 pn->right = in;
559 else
560 pn->left = in;
561
562 ln->fn_bit = plen;
563
564 ln->parent = in;
565 fn->parent = in;
566
567 ln->fn_sernum = sernum;
568
569 if (addr_bit_set(addr, bit)) {
570 in->right = ln;
571 in->left = fn;
572 } else {
573 in->left = ln;
574 in->right = fn;
575 }
576 } else { /* plen <= bit */
577
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900578 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 * (new leaf node)[ln]
580 * / \
581 * (old node)[fn] NULL
582 */
583
584 ln = node_alloc();
585
586 if (ln == NULL)
587 return NULL;
588
589 ln->fn_bit = plen;
590
591 ln->parent = pn;
592
593 ln->fn_sernum = sernum;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900594
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 if (dir)
596 pn->right = ln;
597 else
598 pn->left = ln;
599
600 if (addr_bit_set(&key->addr, plen))
601 ln->right = fn;
602 else
603 ln->left = fn;
604
605 fn->parent = ln;
606 }
607 return ln;
608}
609
610/*
611 * Insert routing information in a node.
612 */
613
614static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt,
Thomas Graf86872cb2006-08-22 00:01:08 -0700615 struct nl_info *info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616{
617 struct rt6_info *iter = NULL;
618 struct rt6_info **ins;
619
620 ins = &fn->leaf;
621
Changli Gaod8d1f302010-06-10 23:31:35 -0700622 for (iter = fn->leaf; iter; iter=iter->dst.rt6_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 /*
624 * Search for duplicates
625 */
626
627 if (iter->rt6i_metric == rt->rt6i_metric) {
628 /*
629 * Same priority level
630 */
631
632 if (iter->rt6i_dev == rt->rt6i_dev &&
633 iter->rt6i_idev == rt->rt6i_idev &&
634 ipv6_addr_equal(&iter->rt6i_gateway,
635 &rt->rt6i_gateway)) {
636 if (!(iter->rt6i_flags&RTF_EXPIRES))
637 return -EEXIST;
638 iter->rt6i_expires = rt->rt6i_expires;
639 if (!(rt->rt6i_flags&RTF_EXPIRES)) {
640 iter->rt6i_flags &= ~RTF_EXPIRES;
641 iter->rt6i_expires = 0;
642 }
643 return -EEXIST;
644 }
645 }
646
647 if (iter->rt6i_metric > rt->rt6i_metric)
648 break;
649
Changli Gaod8d1f302010-06-10 23:31:35 -0700650 ins = &iter->dst.rt6_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 }
652
David S. Millerf11e6652007-03-24 20:36:25 -0700653 /* Reset round-robin state, if necessary */
654 if (ins == &fn->leaf)
655 fn->rr_ptr = NULL;
656
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 /*
658 * insert node
659 */
660
Changli Gaod8d1f302010-06-10 23:31:35 -0700661 rt->dst.rt6_next = iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 *ins = rt;
663 rt->rt6i_node = fn;
664 atomic_inc(&rt->rt6i_ref);
Thomas Graf86872cb2006-08-22 00:01:08 -0700665 inet6_rt_notify(RTM_NEWROUTE, rt, info);
Benjamin Theryc5728722008-03-03 23:34:17 -0800666 info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
668 if ((fn->fn_flags & RTN_RTINFO) == 0) {
Benjamin Theryc5728722008-03-03 23:34:17 -0800669 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 fn->fn_flags |= RTN_RTINFO;
671 }
672
673 return 0;
674}
675
Daniel Lezcano63152fc2008-03-03 23:31:11 -0800676static __inline__ void fib6_start_gc(struct net *net, struct rt6_info *rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677{
Stephen Hemminger417f28b2008-07-22 14:33:45 -0700678 if (!timer_pending(&net->ipv6.ip6_fib_timer) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE)))
Stephen Hemminger417f28b2008-07-22 14:33:45 -0700680 mod_timer(&net->ipv6.ip6_fib_timer,
Stephen Hemminger847499c2008-07-21 13:21:35 -0700681 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682}
683
Daniel Lezcano63152fc2008-03-03 23:31:11 -0800684void fib6_force_start_gc(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685{
Stephen Hemminger417f28b2008-07-22 14:33:45 -0700686 if (!timer_pending(&net->ipv6.ip6_fib_timer))
687 mod_timer(&net->ipv6.ip6_fib_timer,
Stephen Hemminger847499c2008-07-21 13:21:35 -0700688 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689}
690
691/*
692 * Add routing information to the routing tree.
693 * <destination addr>/<source addr>
694 * with source addr info in sub-trees
695 */
696
Thomas Graf86872cb2006-08-22 00:01:08 -0700697int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698{
YOSHIFUJI Hideaki66729e12006-08-23 17:20:34 -0700699 struct fib6_node *fn, *pn = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 int err = -ENOMEM;
701
702 fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr),
703 rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst));
704
705 if (fn == NULL)
706 goto out;
707
YOSHIFUJI Hideaki66729e12006-08-23 17:20:34 -0700708 pn = fn;
709
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710#ifdef CONFIG_IPV6_SUBTREES
711 if (rt->rt6i_src.plen) {
712 struct fib6_node *sn;
713
714 if (fn->subtree == NULL) {
715 struct fib6_node *sfn;
716
717 /*
718 * Create subtree.
719 *
720 * fn[main tree]
721 * |
722 * sfn[subtree root]
723 * \
724 * sn[new leaf node]
725 */
726
727 /* Create subtree root node */
728 sfn = node_alloc();
729 if (sfn == NULL)
730 goto st_failure;
731
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800732 sfn->leaf = info->nl_net->ipv6.ip6_null_entry;
733 atomic_inc(&info->nl_net->ipv6.ip6_null_entry->rt6i_ref);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 sfn->fn_flags = RTN_ROOT;
735 sfn->fn_sernum = fib6_new_sernum();
736
737 /* Now add the first leaf node to new subtree */
738
739 sn = fib6_add_1(sfn, &rt->rt6i_src.addr,
740 sizeof(struct in6_addr), rt->rt6i_src.plen,
741 offsetof(struct rt6_info, rt6i_src));
742
743 if (sn == NULL) {
744 /* If it is failed, discard just allocated
745 root, and then (in st_failure) stale node
746 in main tree.
747 */
748 node_free(sfn);
749 goto st_failure;
750 }
751
752 /* Now link new subtree to main tree */
753 sfn->parent = fn;
754 fn->subtree = sfn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 } else {
756 sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr,
757 sizeof(struct in6_addr), rt->rt6i_src.plen,
758 offsetof(struct rt6_info, rt6i_src));
759
760 if (sn == NULL)
761 goto st_failure;
762 }
763
YOSHIFUJI Hideaki66729e12006-08-23 17:20:34 -0700764 if (fn->leaf == NULL) {
765 fn->leaf = rt;
766 atomic_inc(&rt->rt6i_ref);
767 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 fn = sn;
769 }
770#endif
771
Thomas Graf86872cb2006-08-22 00:01:08 -0700772 err = fib6_add_rt2node(fn, rt, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
774 if (err == 0) {
Daniel Lezcano63152fc2008-03-03 23:31:11 -0800775 fib6_start_gc(info->nl_net, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 if (!(rt->rt6i_flags&RTF_CACHE))
Benjamin Theryec7d43c2008-03-03 23:31:57 -0800777 fib6_prune_clones(info->nl_net, pn, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 }
779
780out:
YOSHIFUJI Hideaki66729e12006-08-23 17:20:34 -0700781 if (err) {
782#ifdef CONFIG_IPV6_SUBTREES
783 /*
784 * If fib6_add_1 has cleared the old leaf pointer in the
785 * super-tree leaf node we have to find a new one for it.
786 */
David S. Miller3c051232008-04-18 01:46:19 -0700787 if (pn != fn && pn->leaf == rt) {
788 pn->leaf = NULL;
789 atomic_dec(&rt->rt6i_ref);
790 }
YOSHIFUJI Hideaki66729e12006-08-23 17:20:34 -0700791 if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) {
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800792 pn->leaf = fib6_find_prefix(info->nl_net, pn);
YOSHIFUJI Hideaki66729e12006-08-23 17:20:34 -0700793#if RT6_DEBUG >= 2
794 if (!pn->leaf) {
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700795 WARN_ON(pn->leaf == NULL);
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800796 pn->leaf = info->nl_net->ipv6.ip6_null_entry;
YOSHIFUJI Hideaki66729e12006-08-23 17:20:34 -0700797 }
798#endif
799 atomic_inc(&pn->leaf->rt6i_ref);
800 }
801#endif
Changli Gaod8d1f302010-06-10 23:31:35 -0700802 dst_free(&rt->dst);
YOSHIFUJI Hideaki66729e12006-08-23 17:20:34 -0700803 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 return err;
805
806#ifdef CONFIG_IPV6_SUBTREES
807 /* Subtree creation failed, probably main tree node
808 is orphan. If it is, shoot it.
809 */
810st_failure:
811 if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800812 fib6_repair_tree(info->nl_net, fn);
Changli Gaod8d1f302010-06-10 23:31:35 -0700813 dst_free(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 return err;
815#endif
816}
817
818/*
819 * Routing tree lookup
820 *
821 */
822
823struct lookup_args {
824 int offset; /* key offset on rt6_info */
825 struct in6_addr *addr; /* search key */
826};
827
828static struct fib6_node * fib6_lookup_1(struct fib6_node *root,
829 struct lookup_args *args)
830{
831 struct fib6_node *fn;
Al Viroe69a4adc2006-11-14 20:56:00 -0800832 __be32 dir;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
YOSHIFUJI Hideaki825e2882006-08-23 17:21:29 -0700834 if (unlikely(args->offset == 0))
835 return NULL;
836
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 /*
838 * Descend on a tree
839 */
840
841 fn = root;
842
843 for (;;) {
844 struct fib6_node *next;
845
846 dir = addr_bit_set(args->addr, fn->fn_bit);
847
848 next = dir ? fn->right : fn->left;
849
850 if (next) {
851 fn = next;
852 continue;
853 }
854
855 break;
856 }
857
YOSHIFUJI Hideaki3fc5e042006-08-23 17:21:12 -0700858 while(fn) {
YOSHIFUJI Hideaki7fc33162006-08-23 17:22:24 -0700859 if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 struct rt6key *key;
861
862 key = (struct rt6key *) ((u8 *) fn->leaf +
863 args->offset);
864
YOSHIFUJI Hideaki3fc5e042006-08-23 17:21:12 -0700865 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
866#ifdef CONFIG_IPV6_SUBTREES
867 if (fn->subtree)
868 fn = fib6_lookup_1(fn->subtree, args + 1);
869#endif
870 if (!fn || fn->fn_flags & RTN_RTINFO)
871 return fn;
872 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 }
874
YOSHIFUJI Hideaki3fc5e042006-08-23 17:21:12 -0700875 if (fn->fn_flags & RTN_ROOT)
876 break;
877
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 fn = fn->parent;
879 }
880
881 return NULL;
882}
883
884struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr,
885 struct in6_addr *saddr)
886{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 struct fib6_node *fn;
YOSHIFUJI Hideaki825e2882006-08-23 17:21:29 -0700888 struct lookup_args args[] = {
889 {
890 .offset = offsetof(struct rt6_info, rt6i_dst),
891 .addr = daddr,
892 },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893#ifdef CONFIG_IPV6_SUBTREES
YOSHIFUJI Hideaki825e2882006-08-23 17:21:29 -0700894 {
895 .offset = offsetof(struct rt6_info, rt6i_src),
896 .addr = saddr,
897 },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898#endif
YOSHIFUJI Hideaki825e2882006-08-23 17:21:29 -0700899 {
900 .offset = 0, /* sentinel */
901 }
902 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903
YOSHIFUJI Hideakifefc2a62006-08-23 17:21:50 -0700904 fn = fib6_lookup_1(root, daddr ? args : args + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
906 if (fn == NULL || fn->fn_flags & RTN_TL_ROOT)
907 fn = root;
908
909 return fn;
910}
911
912/*
913 * Get node with specified destination prefix (and source prefix,
914 * if subtrees are used)
915 */
916
917
918static struct fib6_node * fib6_locate_1(struct fib6_node *root,
919 struct in6_addr *addr,
920 int plen, int offset)
921{
922 struct fib6_node *fn;
923
924 for (fn = root; fn ; ) {
925 struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset);
926
927 /*
928 * Prefix match
929 */
930 if (plen < fn->fn_bit ||
931 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
932 return NULL;
933
934 if (plen == fn->fn_bit)
935 return fn;
936
937 /*
938 * We have more bits to go
939 */
940 if (addr_bit_set(addr, fn->fn_bit))
941 fn = fn->right;
942 else
943 fn = fn->left;
944 }
945 return NULL;
946}
947
948struct fib6_node * fib6_locate(struct fib6_node *root,
949 struct in6_addr *daddr, int dst_len,
950 struct in6_addr *saddr, int src_len)
951{
952 struct fib6_node *fn;
953
954 fn = fib6_locate_1(root, daddr, dst_len,
955 offsetof(struct rt6_info, rt6i_dst));
956
957#ifdef CONFIG_IPV6_SUBTREES
958 if (src_len) {
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700959 WARN_ON(saddr == NULL);
YOSHIFUJI Hideaki3fc5e042006-08-23 17:21:12 -0700960 if (fn && fn->subtree)
961 fn = fib6_locate_1(fn->subtree, saddr, src_len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 offsetof(struct rt6_info, rt6i_src));
963 }
964#endif
965
966 if (fn && fn->fn_flags&RTN_RTINFO)
967 return fn;
968
969 return NULL;
970}
971
972
973/*
974 * Deletion
975 *
976 */
977
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800978static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979{
980 if (fn->fn_flags&RTN_ROOT)
Daniel Lezcano8ed67782008-03-04 13:48:30 -0800981 return net->ipv6.ip6_null_entry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
983 while(fn) {
984 if(fn->left)
985 return fn->left->leaf;
986
987 if(fn->right)
988 return fn->right->leaf;
989
YOSHIFUJI Hideaki7fc33162006-08-23 17:22:24 -0700990 fn = FIB6_SUBTREE(fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 }
992 return NULL;
993}
994
995/*
996 * Called to trim the tree of intermediate nodes when possible. "fn"
997 * is the node we want to try and remove.
998 */
999
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001000static struct fib6_node *fib6_repair_tree(struct net *net,
1001 struct fib6_node *fn)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002{
1003 int children;
1004 int nstate;
1005 struct fib6_node *child, *pn;
1006 struct fib6_walker_t *w;
1007 int iter = 0;
1008
1009 for (;;) {
1010 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1011 iter++;
1012
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001013 WARN_ON(fn->fn_flags & RTN_RTINFO);
1014 WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1015 WARN_ON(fn->leaf != NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016
1017 children = 0;
1018 child = NULL;
1019 if (fn->right) child = fn->right, children |= 1;
1020 if (fn->left) child = fn->left, children |= 2;
1021
YOSHIFUJI Hideaki7fc33162006-08-23 17:22:24 -07001022 if (children == 3 || FIB6_SUBTREE(fn)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023#ifdef CONFIG_IPV6_SUBTREES
1024 /* Subtree root (i.e. fn) may have one child */
1025 || (children && fn->fn_flags&RTN_ROOT)
1026#endif
1027 ) {
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001028 fn->leaf = fib6_find_prefix(net, fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029#if RT6_DEBUG >= 2
1030 if (fn->leaf==NULL) {
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001031 WARN_ON(!fn->leaf);
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001032 fn->leaf = net->ipv6.ip6_null_entry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 }
1034#endif
1035 atomic_inc(&fn->leaf->rt6i_ref);
1036 return fn->parent;
1037 }
1038
1039 pn = fn->parent;
1040#ifdef CONFIG_IPV6_SUBTREES
YOSHIFUJI Hideaki7fc33162006-08-23 17:22:24 -07001041 if (FIB6_SUBTREE(pn) == fn) {
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001042 WARN_ON(!(fn->fn_flags & RTN_ROOT));
YOSHIFUJI Hideaki7fc33162006-08-23 17:22:24 -07001043 FIB6_SUBTREE(pn) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 nstate = FWS_L;
1045 } else {
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001046 WARN_ON(fn->fn_flags & RTN_ROOT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047#endif
1048 if (pn->right == fn) pn->right = child;
1049 else if (pn->left == fn) pn->left = child;
1050#if RT6_DEBUG >= 2
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001051 else
1052 WARN_ON(1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053#endif
1054 if (child)
1055 child->parent = pn;
1056 nstate = FWS_R;
1057#ifdef CONFIG_IPV6_SUBTREES
1058 }
1059#endif
1060
1061 read_lock(&fib6_walker_lock);
1062 FOR_WALKERS(w) {
1063 if (child == NULL) {
1064 if (w->root == fn) {
1065 w->root = w->node = NULL;
1066 RT6_TRACE("W %p adjusted by delroot 1\n", w);
1067 } else if (w->node == fn) {
1068 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1069 w->node = pn;
1070 w->state = nstate;
1071 }
1072 } else {
1073 if (w->root == fn) {
1074 w->root = child;
1075 RT6_TRACE("W %p adjusted by delroot 2\n", w);
1076 }
1077 if (w->node == fn) {
1078 w->node = child;
1079 if (children&2) {
1080 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1081 w->state = w->state>=FWS_R ? FWS_U : FWS_INIT;
1082 } else {
1083 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1084 w->state = w->state>=FWS_C ? FWS_U : FWS_INIT;
1085 }
1086 }
1087 }
1088 }
1089 read_unlock(&fib6_walker_lock);
1090
1091 node_free(fn);
YOSHIFUJI Hideaki7fc33162006-08-23 17:22:24 -07001092 if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 return pn;
1094
1095 rt6_release(pn->leaf);
1096 pn->leaf = NULL;
1097 fn = pn;
1098 }
1099}
1100
1101static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp,
Thomas Graf86872cb2006-08-22 00:01:08 -07001102 struct nl_info *info)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103{
1104 struct fib6_walker_t *w;
1105 struct rt6_info *rt = *rtp;
Benjamin Theryc5728722008-03-03 23:34:17 -08001106 struct net *net = info->nl_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
1108 RT6_TRACE("fib6_del_route\n");
1109
1110 /* Unlink it */
Changli Gaod8d1f302010-06-10 23:31:35 -07001111 *rtp = rt->dst.rt6_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 rt->rt6i_node = NULL;
Benjamin Theryc5728722008-03-03 23:34:17 -08001113 net->ipv6.rt6_stats->fib_rt_entries--;
1114 net->ipv6.rt6_stats->fib_discarded_routes++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115
David S. Millerf11e6652007-03-24 20:36:25 -07001116 /* Reset round-robin state, if necessary */
1117 if (fn->rr_ptr == rt)
1118 fn->rr_ptr = NULL;
1119
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 /* Adjust walkers */
1121 read_lock(&fib6_walker_lock);
1122 FOR_WALKERS(w) {
1123 if (w->state == FWS_C && w->leaf == rt) {
1124 RT6_TRACE("walker %p adjusted by delroute\n", w);
Changli Gaod8d1f302010-06-10 23:31:35 -07001125 w->leaf = rt->dst.rt6_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 if (w->leaf == NULL)
1127 w->state = FWS_U;
1128 }
1129 }
1130 read_unlock(&fib6_walker_lock);
1131
Changli Gaod8d1f302010-06-10 23:31:35 -07001132 rt->dst.rt6_next = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 /* If it was last route, expunge its radix tree node */
1135 if (fn->leaf == NULL) {
1136 fn->fn_flags &= ~RTN_RTINFO;
Benjamin Theryc5728722008-03-03 23:34:17 -08001137 net->ipv6.rt6_stats->fib_route_nodes--;
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001138 fn = fib6_repair_tree(net, fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 }
1140
1141 if (atomic_read(&rt->rt6i_ref) != 1) {
1142 /* This route is used as dummy address holder in some split
1143 * nodes. It is not leaked, but it still holds other resources,
1144 * which must be released in time. So, scan ascendant nodes
1145 * and replace dummy references to this route with references
1146 * to still alive ones.
1147 */
1148 while (fn) {
1149 if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) {
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001150 fn->leaf = fib6_find_prefix(net, fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 atomic_inc(&fn->leaf->rt6i_ref);
1152 rt6_release(rt);
1153 }
1154 fn = fn->parent;
1155 }
1156 /* No more references are possible at this point. */
Pavel Emelyanov2df96af2008-02-18 20:50:42 -08001157 BUG_ON(atomic_read(&rt->rt6i_ref) != 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 }
1159
Thomas Graf86872cb2006-08-22 00:01:08 -07001160 inet6_rt_notify(RTM_DELROUTE, rt, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 rt6_release(rt);
1162}
1163
Thomas Graf86872cb2006-08-22 00:01:08 -07001164int fib6_del(struct rt6_info *rt, struct nl_info *info)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165{
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001166 struct net *net = info->nl_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 struct fib6_node *fn = rt->rt6i_node;
1168 struct rt6_info **rtp;
1169
1170#if RT6_DEBUG >= 2
Changli Gaod8d1f302010-06-10 23:31:35 -07001171 if (rt->dst.obsolete>0) {
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001172 WARN_ON(fn != NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 return -ENOENT;
1174 }
1175#endif
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001176 if (fn == NULL || rt == net->ipv6.ip6_null_entry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 return -ENOENT;
1178
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001179 WARN_ON(!(fn->fn_flags & RTN_RTINFO));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
YOSHIFUJI Hideaki150730d2006-08-23 17:22:55 -07001181 if (!(rt->rt6i_flags&RTF_CACHE)) {
1182 struct fib6_node *pn = fn;
1183#ifdef CONFIG_IPV6_SUBTREES
1184 /* clones of this route might be in another subtree */
1185 if (rt->rt6i_src.plen) {
1186 while (!(pn->fn_flags&RTN_ROOT))
1187 pn = pn->parent;
1188 pn = pn->parent;
1189 }
1190#endif
Benjamin Theryec7d43c2008-03-03 23:31:57 -08001191 fib6_prune_clones(info->nl_net, pn, rt);
YOSHIFUJI Hideaki150730d2006-08-23 17:22:55 -07001192 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193
1194 /*
1195 * Walk the leaf entries looking for ourself
1196 */
1197
Changli Gaod8d1f302010-06-10 23:31:35 -07001198 for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->dst.rt6_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 if (*rtp == rt) {
Thomas Graf86872cb2006-08-22 00:01:08 -07001200 fib6_del_route(fn, rtp, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 return 0;
1202 }
1203 }
1204 return -ENOENT;
1205}
1206
1207/*
1208 * Tree traversal function.
1209 *
1210 * Certainly, it is not interrupt safe.
1211 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1212 * It means, that we can modify tree during walking
1213 * and use this function for garbage collection, clone pruning,
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001214 * cleaning tree when a device goes down etc. etc.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 *
1216 * It guarantees that every node will be traversed,
1217 * and that it will be traversed only once.
1218 *
1219 * Callback function w->func may return:
1220 * 0 -> continue walking.
1221 * positive value -> walking is suspended (used by tree dumps,
1222 * and probably by gc, if it will be split to several slices)
1223 * negative value -> terminate walking.
1224 *
1225 * The function itself returns:
1226 * 0 -> walk is complete.
1227 * >0 -> walk is incomplete (i.e. suspended)
1228 * <0 -> walk is terminated by an error.
1229 */
1230
Adrian Bunk90d41122006-08-14 23:49:16 -07001231static int fib6_walk_continue(struct fib6_walker_t *w)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232{
1233 struct fib6_node *fn, *pn;
1234
1235 for (;;) {
1236 fn = w->node;
1237 if (fn == NULL)
1238 return 0;
1239
1240 if (w->prune && fn != w->root &&
1241 fn->fn_flags&RTN_RTINFO && w->state < FWS_C) {
1242 w->state = FWS_C;
1243 w->leaf = fn->leaf;
1244 }
1245 switch (w->state) {
1246#ifdef CONFIG_IPV6_SUBTREES
1247 case FWS_S:
YOSHIFUJI Hideaki7fc33162006-08-23 17:22:24 -07001248 if (FIB6_SUBTREE(fn)) {
1249 w->node = FIB6_SUBTREE(fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 continue;
1251 }
1252 w->state = FWS_L;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001253#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 case FWS_L:
1255 if (fn->left) {
1256 w->node = fn->left;
1257 w->state = FWS_INIT;
1258 continue;
1259 }
1260 w->state = FWS_R;
1261 case FWS_R:
1262 if (fn->right) {
1263 w->node = fn->right;
1264 w->state = FWS_INIT;
1265 continue;
1266 }
1267 w->state = FWS_C;
1268 w->leaf = fn->leaf;
1269 case FWS_C:
1270 if (w->leaf && fn->fn_flags&RTN_RTINFO) {
Patrick McHardy2bec5a32010-02-08 05:19:03 +00001271 int err;
1272
1273 if (w->count < w->skip) {
1274 w->count++;
1275 continue;
1276 }
1277
1278 err = w->func(w);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 if (err)
1280 return err;
Patrick McHardy2bec5a32010-02-08 05:19:03 +00001281
1282 w->count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 continue;
1284 }
1285 w->state = FWS_U;
1286 case FWS_U:
1287 if (fn == w->root)
1288 return 0;
1289 pn = fn->parent;
1290 w->node = pn;
1291#ifdef CONFIG_IPV6_SUBTREES
YOSHIFUJI Hideaki7fc33162006-08-23 17:22:24 -07001292 if (FIB6_SUBTREE(pn) == fn) {
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001293 WARN_ON(!(fn->fn_flags & RTN_ROOT));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 w->state = FWS_L;
1295 continue;
1296 }
1297#endif
1298 if (pn->left == fn) {
1299 w->state = FWS_R;
1300 continue;
1301 }
1302 if (pn->right == fn) {
1303 w->state = FWS_C;
1304 w->leaf = w->node->leaf;
1305 continue;
1306 }
1307#if RT6_DEBUG >= 2
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001308 WARN_ON(1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309#endif
1310 }
1311 }
1312}
1313
Adrian Bunk90d41122006-08-14 23:49:16 -07001314static int fib6_walk(struct fib6_walker_t *w)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315{
1316 int res;
1317
1318 w->state = FWS_INIT;
1319 w->node = w->root;
1320
1321 fib6_walker_link(w);
1322 res = fib6_walk_continue(w);
1323 if (res <= 0)
1324 fib6_walker_unlink(w);
1325 return res;
1326}
1327
1328static int fib6_clean_node(struct fib6_walker_t *w)
1329{
1330 int res;
1331 struct rt6_info *rt;
Benjamin Thery0a8891a2007-10-08 20:39:36 -07001332 struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w);
Benjamin Theryec7d43c2008-03-03 23:31:57 -08001333 struct nl_info info = {
1334 .nl_net = c->net,
1335 };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
Changli Gaod8d1f302010-06-10 23:31:35 -07001337 for (rt = w->leaf; rt; rt = rt->dst.rt6_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 res = c->func(rt, c->arg);
1339 if (res < 0) {
1340 w->leaf = rt;
Denis V. Lunev528c4ce2007-12-13 09:45:12 -08001341 res = fib6_del(rt, &info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 if (res) {
1343#if RT6_DEBUG >= 2
1344 printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res);
1345#endif
1346 continue;
1347 }
1348 return 0;
1349 }
Ilpo Järvinen547b7922008-07-25 21:43:18 -07001350 WARN_ON(res != 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 }
1352 w->leaf = rt;
1353 return 0;
1354}
1355
1356/*
1357 * Convenient frontend to tree walker.
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +09001358 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 * func is called on each route.
1360 * It may return -1 -> delete this route.
1361 * 0 -> continue walking
1362 *
1363 * prune==1 -> only immediate children of node (certainly,
1364 * ignoring pure split nodes) will be scanned.
1365 */
1366
Benjamin Theryec7d43c2008-03-03 23:31:57 -08001367static void fib6_clean_tree(struct net *net, struct fib6_node *root,
Adrian Bunk8ce11e62006-08-07 21:50:48 -07001368 int (*func)(struct rt6_info *, void *arg),
1369 int prune, void *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370{
1371 struct fib6_cleaner_t c;
1372
1373 c.w.root = root;
1374 c.w.func = fib6_clean_node;
1375 c.w.prune = prune;
Patrick McHardy2bec5a32010-02-08 05:19:03 +00001376 c.w.count = 0;
1377 c.w.skip = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 c.func = func;
1379 c.arg = arg;
Benjamin Theryec7d43c2008-03-03 23:31:57 -08001380 c.net = net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381
1382 fib6_walk(&c.w);
1383}
1384
Daniel Lezcanof3db4852008-03-03 23:27:06 -08001385void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg),
Thomas Grafc71099a2006-08-04 23:20:06 -07001386 int prune, void *arg)
1387{
Thomas Grafc71099a2006-08-04 23:20:06 -07001388 struct fib6_table *table;
Patrick McHardy1b43af52006-08-10 23:11:17 -07001389 struct hlist_node *node;
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001390 struct hlist_head *head;
Patrick McHardy1b43af52006-08-10 23:11:17 -07001391 unsigned int h;
Thomas Grafc71099a2006-08-04 23:20:06 -07001392
Patrick McHardy1b43af52006-08-10 23:11:17 -07001393 rcu_read_lock();
Neil Hormana33bc5c2009-07-30 18:52:15 -07001394 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
Daniel Lezcanof3db4852008-03-03 23:27:06 -08001395 head = &net->ipv6.fib_table_hash[h];
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001396 hlist_for_each_entry_rcu(table, node, head, tb6_hlist) {
Thomas Grafc71099a2006-08-04 23:20:06 -07001397 write_lock_bh(&table->tb6_lock);
Benjamin Theryec7d43c2008-03-03 23:31:57 -08001398 fib6_clean_tree(net, &table->tb6_root,
1399 func, prune, arg);
Thomas Grafc71099a2006-08-04 23:20:06 -07001400 write_unlock_bh(&table->tb6_lock);
1401 }
1402 }
Patrick McHardy1b43af52006-08-10 23:11:17 -07001403 rcu_read_unlock();
Thomas Grafc71099a2006-08-04 23:20:06 -07001404}
1405
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406static int fib6_prune_clone(struct rt6_info *rt, void *arg)
1407{
1408 if (rt->rt6i_flags & RTF_CACHE) {
1409 RT6_TRACE("pruning clone %p\n", rt);
1410 return -1;
1411 }
1412
1413 return 0;
1414}
1415
Benjamin Theryec7d43c2008-03-03 23:31:57 -08001416static void fib6_prune_clones(struct net *net, struct fib6_node *fn,
1417 struct rt6_info *rt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418{
Benjamin Theryec7d43c2008-03-03 23:31:57 -08001419 fib6_clean_tree(net, fn, fib6_prune_clone, 1, rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420}
1421
1422/*
1423 * Garbage collection
1424 */
1425
1426static struct fib6_gc_args
1427{
1428 int timeout;
1429 int more;
1430} gc_args;
1431
1432static int fib6_age(struct rt6_info *rt, void *arg)
1433{
1434 unsigned long now = jiffies;
1435
1436 /*
1437 * check addrconf expiration here.
1438 * Routes are expired even if they are in use.
1439 *
1440 * Also age clones. Note, that clones are aged out
1441 * only if they are not in use now.
1442 */
1443
1444 if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) {
1445 if (time_after(now, rt->rt6i_expires)) {
1446 RT6_TRACE("expiring %p\n", rt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 return -1;
1448 }
1449 gc_args.more++;
1450 } else if (rt->rt6i_flags & RTF_CACHE) {
Changli Gaod8d1f302010-06-10 23:31:35 -07001451 if (atomic_read(&rt->dst.__refcnt) == 0 &&
1452 time_after_eq(now, rt->dst.lastuse + gc_args.timeout)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 RT6_TRACE("aging clone %p\n", rt);
1454 return -1;
1455 } else if ((rt->rt6i_flags & RTF_GATEWAY) &&
1456 (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) {
1457 RT6_TRACE("purging route %p via non-router but gateway\n",
1458 rt);
1459 return -1;
1460 }
1461 gc_args.more++;
1462 }
1463
1464 return 0;
1465}
1466
1467static DEFINE_SPINLOCK(fib6_gc_lock);
1468
Daniel Lezcano5b7c9312008-03-03 23:28:58 -08001469void fib6_run_gc(unsigned long expires, struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470{
Daniel Lezcano5b7c9312008-03-03 23:28:58 -08001471 if (expires != ~0UL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 spin_lock_bh(&fib6_gc_lock);
Daniel Lezcano5b7c9312008-03-03 23:28:58 -08001473 gc_args.timeout = expires ? (int)expires :
1474 net->ipv6.sysctl.ip6_rt_gc_interval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 } else {
Stephen Hemmingera76d7342008-07-22 14:34:35 -07001476 if (!spin_trylock_bh(&fib6_gc_lock)) {
Stephen Hemminger417f28b2008-07-22 14:33:45 -07001477 mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 return;
1479 }
Daniel Lezcano5b7c9312008-03-03 23:28:58 -08001480 gc_args.timeout = net->ipv6.sysctl.ip6_rt_gc_interval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482
Stephen Hemminger3d0f24a2008-07-22 14:35:50 -07001483 gc_args.more = icmp6_dst_gc();
Daniel Lezcanof3db4852008-03-03 23:27:06 -08001484
Daniel Lezcano5b7c9312008-03-03 23:28:58 -08001485 fib6_clean_all(net, fib6_age, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486
1487 if (gc_args.more)
Stephen Hemmingerc8a45222008-07-22 14:34:09 -07001488 mod_timer(&net->ipv6.ip6_fib_timer,
1489 round_jiffies(jiffies
1490 + net->ipv6.sysctl.ip6_rt_gc_interval));
Stephen Hemminger417f28b2008-07-22 14:33:45 -07001491 else
1492 del_timer(&net->ipv6.ip6_fib_timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 spin_unlock_bh(&fib6_gc_lock);
1494}
1495
Daniel Lezcano5b7c9312008-03-03 23:28:58 -08001496static void fib6_gc_timer_cb(unsigned long arg)
1497{
1498 fib6_run_gc(0, (struct net *)arg);
1499}
1500
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001501static int __net_init fib6_net_init(struct net *net)
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001502{
Eric Dumazet10da66f2010-10-13 08:22:03 +00001503 size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ;
1504
Stephen Hemminger417f28b2008-07-22 14:33:45 -07001505 setup_timer(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, (unsigned long)net);
Daniel Lezcano63152fc2008-03-03 23:31:11 -08001506
Benjamin Theryc5728722008-03-03 23:34:17 -08001507 net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL);
1508 if (!net->ipv6.rt6_stats)
1509 goto out_timer;
1510
Eric Dumazet10da66f2010-10-13 08:22:03 +00001511 /* Avoid false sharing : Use at least a full cache line */
1512 size = max_t(size_t, size, L1_CACHE_BYTES);
1513
1514 net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL);
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001515 if (!net->ipv6.fib_table_hash)
Benjamin Theryc5728722008-03-03 23:34:17 -08001516 goto out_rt6_stats;
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001517
1518 net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
1519 GFP_KERNEL);
1520 if (!net->ipv6.fib6_main_tbl)
1521 goto out_fib_table_hash;
1522
1523 net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001524 net->ipv6.fib6_main_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001525 net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
1526 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1527
1528#ifdef CONFIG_IPV6_MULTIPLE_TABLES
1529 net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
1530 GFP_KERNEL);
1531 if (!net->ipv6.fib6_local_tbl)
1532 goto out_fib6_main_tbl;
1533 net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001534 net->ipv6.fib6_local_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry;
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001535 net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
1536 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
1537#endif
1538 fib6_tables_init(net);
1539
Stephen Hemminger417f28b2008-07-22 14:33:45 -07001540 return 0;
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001541
1542#ifdef CONFIG_IPV6_MULTIPLE_TABLES
1543out_fib6_main_tbl:
1544 kfree(net->ipv6.fib6_main_tbl);
1545#endif
1546out_fib_table_hash:
1547 kfree(net->ipv6.fib_table_hash);
Benjamin Theryc5728722008-03-03 23:34:17 -08001548out_rt6_stats:
1549 kfree(net->ipv6.rt6_stats);
Daniel Lezcano63152fc2008-03-03 23:31:11 -08001550out_timer:
Stephen Hemminger417f28b2008-07-22 14:33:45 -07001551 return -ENOMEM;
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001552 }
1553
1554static void fib6_net_exit(struct net *net)
1555{
Daniel Lezcano8ed67782008-03-04 13:48:30 -08001556 rt6_ifdown(net, NULL);
Stephen Hemminger417f28b2008-07-22 14:33:45 -07001557 del_timer_sync(&net->ipv6.ip6_fib_timer);
1558
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001559#ifdef CONFIG_IPV6_MULTIPLE_TABLES
1560 kfree(net->ipv6.fib6_local_tbl);
1561#endif
1562 kfree(net->ipv6.fib6_main_tbl);
1563 kfree(net->ipv6.fib_table_hash);
Benjamin Theryc5728722008-03-03 23:34:17 -08001564 kfree(net->ipv6.rt6_stats);
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001565}
1566
1567static struct pernet_operations fib6_net_ops = {
1568 .init = fib6_net_init,
1569 .exit = fib6_net_exit,
1570};
1571
Daniel Lezcanod63bddb2007-12-07 00:40:34 -08001572int __init fib6_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573{
Daniel Lezcanoe0b855902008-03-03 23:24:31 -08001574 int ret = -ENOMEM;
Daniel Lezcano63152fc2008-03-03 23:31:11 -08001575
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 fib6_node_kmem = kmem_cache_create("fib6_nodes",
1577 sizeof(struct fib6_node),
Daniel Lezcanof845ab62007-12-07 00:45:16 -08001578 0, SLAB_HWCACHE_ALIGN,
Paul Mundt20c2df82007-07-20 10:11:58 +09001579 NULL);
Daniel Lezcanof845ab62007-12-07 00:45:16 -08001580 if (!fib6_node_kmem)
Daniel Lezcanoe0b855902008-03-03 23:24:31 -08001581 goto out;
1582
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001583 ret = register_pernet_subsys(&fib6_net_ops);
1584 if (ret)
Benjamin Theryc5728722008-03-03 23:34:17 -08001585 goto out_kmem_cache_create;
Daniel Lezcanoe0b855902008-03-03 23:24:31 -08001586
Daniel Lezcanod63bddb2007-12-07 00:40:34 -08001587 ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib);
1588 if (ret)
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001589 goto out_unregister_subsys;
Daniel Lezcanod63bddb2007-12-07 00:40:34 -08001590out:
1591 return ret;
1592
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001593out_unregister_subsys:
1594 unregister_pernet_subsys(&fib6_net_ops);
Daniel Lezcanod63bddb2007-12-07 00:40:34 -08001595out_kmem_cache_create:
1596 kmem_cache_destroy(fib6_node_kmem);
1597 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598}
1599
1600void fib6_gc_cleanup(void)
1601{
Daniel Lezcano58f09b72008-03-03 23:25:27 -08001602 unregister_pernet_subsys(&fib6_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 kmem_cache_destroy(fib6_node_kmem);
1604}