]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1ab1457c | 2 | * Linux INET6 implementation |
1da177e4 LT |
3 | * Forwarding Information Database |
4 | * | |
5 | * Authors: | |
1ab1457c | 6 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 LT |
7 | * |
8 | * $Id: ip6_fib.c,v 1.25 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 | /* | |
17 | * Changes: | |
18 | * Yuji SEKIYA @USAGI: Support default route on router node; | |
19 | * remove ip6_null_entry from the top of | |
20 | * routing table. | |
c0bece9f | 21 | * Ville Nuorvala: Fixed routing subtrees. |
1da177e4 | 22 | */ |
1da177e4 LT |
23 | #include <linux/errno.h> |
24 | #include <linux/types.h> | |
25 | #include <linux/net.h> | |
26 | #include <linux/route.h> | |
27 | #include <linux/netdevice.h> | |
28 | #include <linux/in6.h> | |
29 | #include <linux/init.h> | |
c71099ac | 30 | #include <linux/list.h> |
1da177e4 LT |
31 | |
32 | #ifdef CONFIG_PROC_FS | |
33 | #include <linux/proc_fs.h> | |
34 | #endif | |
35 | ||
36 | #include <net/ipv6.h> | |
37 | #include <net/ndisc.h> | |
38 | #include <net/addrconf.h> | |
39 | ||
40 | #include <net/ip6_fib.h> | |
41 | #include <net/ip6_route.h> | |
42 | ||
43 | #define RT6_DEBUG 2 | |
44 | ||
45 | #if RT6_DEBUG >= 3 | |
46 | #define RT6_TRACE(x...) printk(KERN_DEBUG x) | |
47 | #else | |
48 | #define RT6_TRACE(x...) do { ; } while (0) | |
49 | #endif | |
50 | ||
51 | struct rt6_statistics rt6_stats; | |
52 | ||
e18b890b | 53 | static struct kmem_cache * fib6_node_kmem __read_mostly; |
1da177e4 LT |
54 | |
55 | enum fib_walk_state_t | |
56 | { | |
57 | #ifdef CONFIG_IPV6_SUBTREES | |
58 | FWS_S, | |
59 | #endif | |
60 | FWS_L, | |
61 | FWS_R, | |
62 | FWS_C, | |
63 | FWS_U | |
64 | }; | |
65 | ||
66 | struct fib6_cleaner_t | |
67 | { | |
68 | struct fib6_walker_t w; | |
69 | int (*func)(struct rt6_info *, void *arg); | |
70 | void *arg; | |
71 | }; | |
72 | ||
90d41122 | 73 | static DEFINE_RWLOCK(fib6_walker_lock); |
1da177e4 LT |
74 | |
75 | #ifdef CONFIG_IPV6_SUBTREES | |
76 | #define FWS_INIT FWS_S | |
1da177e4 LT |
77 | #else |
78 | #define FWS_INIT FWS_L | |
1da177e4 LT |
79 | #endif |
80 | ||
81 | static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt); | |
66729e18 | 82 | static struct rt6_info * fib6_find_prefix(struct fib6_node *fn); |
1da177e4 | 83 | static struct fib6_node * fib6_repair_tree(struct fib6_node *fn); |
90d41122 AB |
84 | static int fib6_walk(struct fib6_walker_t *w); |
85 | static int fib6_walk_continue(struct fib6_walker_t *w); | |
1da177e4 LT |
86 | |
87 | /* | |
88 | * A routing update causes an increase of the serial number on the | |
89 | * affected subtree. This allows for cached routes to be asynchronously | |
90 | * tested when modifications are made to the destination cache as a | |
91 | * result of redirects, path MTU changes, etc. | |
92 | */ | |
93 | ||
94 | static __u32 rt_sernum; | |
95 | ||
5b7c931d DL |
96 | static void fib6_gc_timer_cb(unsigned long arg); |
97 | ||
450d19f8 | 98 | static struct timer_list *ip6_fib_timer; |
1da177e4 | 99 | |
90d41122 | 100 | static struct fib6_walker_t fib6_walker_list = { |
1da177e4 | 101 | .prev = &fib6_walker_list, |
1ab1457c | 102 | .next = &fib6_walker_list, |
1da177e4 LT |
103 | }; |
104 | ||
105 | #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next) | |
106 | ||
90d41122 AB |
107 | static inline void fib6_walker_link(struct fib6_walker_t *w) |
108 | { | |
109 | write_lock_bh(&fib6_walker_lock); | |
110 | w->next = fib6_walker_list.next; | |
111 | w->prev = &fib6_walker_list; | |
112 | w->next->prev = w; | |
113 | w->prev->next = w; | |
114 | write_unlock_bh(&fib6_walker_lock); | |
115 | } | |
116 | ||
117 | static inline void fib6_walker_unlink(struct fib6_walker_t *w) | |
118 | { | |
119 | write_lock_bh(&fib6_walker_lock); | |
120 | w->next->prev = w->prev; | |
121 | w->prev->next = w->next; | |
122 | w->prev = w->next = w; | |
123 | write_unlock_bh(&fib6_walker_lock); | |
124 | } | |
1da177e4 LT |
125 | static __inline__ u32 fib6_new_sernum(void) |
126 | { | |
127 | u32 n = ++rt_sernum; | |
128 | if ((__s32)n <= 0) | |
129 | rt_sernum = n = 1; | |
130 | return n; | |
131 | } | |
132 | ||
133 | /* | |
134 | * Auxiliary address test functions for the radix tree. | |
135 | * | |
1ab1457c | 136 | * These assume a 32bit processor (although it will work on |
1da177e4 LT |
137 | * 64bit processors) |
138 | */ | |
139 | ||
140 | /* | |
141 | * test bit | |
142 | */ | |
143 | ||
e69a4adc | 144 | static __inline__ __be32 addr_bit_set(void *token, int fn_bit) |
1da177e4 | 145 | { |
e69a4adc | 146 | __be32 *addr = token; |
1da177e4 LT |
147 | |
148 | return htonl(1 << ((~fn_bit)&0x1F)) & addr[fn_bit>>5]; | |
149 | } | |
150 | ||
1da177e4 LT |
151 | static __inline__ struct fib6_node * node_alloc(void) |
152 | { | |
153 | struct fib6_node *fn; | |
154 | ||
c3762229 | 155 | fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC); |
1da177e4 LT |
156 | |
157 | return fn; | |
158 | } | |
159 | ||
160 | static __inline__ void node_free(struct fib6_node * fn) | |
161 | { | |
162 | kmem_cache_free(fib6_node_kmem, fn); | |
163 | } | |
164 | ||
165 | static __inline__ void rt6_release(struct rt6_info *rt) | |
166 | { | |
167 | if (atomic_dec_and_test(&rt->rt6i_ref)) | |
168 | dst_free(&rt->u.dst); | |
169 | } | |
170 | ||
c71099ac | 171 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
1b43af54 PM |
172 | #define FIB_TABLE_HASHSZ 256 |
173 | #else | |
174 | #define FIB_TABLE_HASHSZ 1 | |
175 | #endif | |
1b43af54 | 176 | |
58f09b78 | 177 | static void fib6_link_table(struct net *net, struct fib6_table *tb) |
1b43af54 PM |
178 | { |
179 | unsigned int h; | |
180 | ||
375216ad TG |
181 | /* |
182 | * Initialize table lock at a single place to give lockdep a key, | |
183 | * tables aren't visible prior to being linked to the list. | |
184 | */ | |
185 | rwlock_init(&tb->tb6_lock); | |
186 | ||
1b43af54 PM |
187 | h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1); |
188 | ||
189 | /* | |
190 | * No protection necessary, this is the only list mutatation | |
191 | * operation, tables never disappear once they exist. | |
192 | */ | |
58f09b78 | 193 | hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]); |
1b43af54 | 194 | } |
c71099ac | 195 | |
1b43af54 | 196 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 197 | |
c71099ac TG |
198 | static struct fib6_table *fib6_alloc_table(u32 id) |
199 | { | |
200 | struct fib6_table *table; | |
201 | ||
202 | table = kzalloc(sizeof(*table), GFP_ATOMIC); | |
203 | if (table != NULL) { | |
204 | table->tb6_id = id; | |
c71099ac TG |
205 | table->tb6_root.leaf = &ip6_null_entry; |
206 | table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
207 | } | |
208 | ||
209 | return table; | |
210 | } | |
211 | ||
58f09b78 | 212 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac TG |
213 | { |
214 | struct fib6_table *tb; | |
215 | ||
216 | if (id == 0) | |
217 | id = RT6_TABLE_MAIN; | |
58f09b78 | 218 | tb = fib6_get_table(net, id); |
c71099ac TG |
219 | if (tb) |
220 | return tb; | |
221 | ||
222 | tb = fib6_alloc_table(id); | |
223 | if (tb != NULL) | |
58f09b78 | 224 | fib6_link_table(net, tb); |
c71099ac TG |
225 | |
226 | return tb; | |
227 | } | |
228 | ||
58f09b78 | 229 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac TG |
230 | { |
231 | struct fib6_table *tb; | |
58f09b78 | 232 | struct hlist_head *head; |
c71099ac TG |
233 | struct hlist_node *node; |
234 | unsigned int h; | |
235 | ||
236 | if (id == 0) | |
237 | id = RT6_TABLE_MAIN; | |
238 | h = id & (FIB_TABLE_HASHSZ - 1); | |
239 | rcu_read_lock(); | |
58f09b78 DL |
240 | head = &net->ipv6.fib_table_hash[h]; |
241 | hlist_for_each_entry_rcu(tb, node, head, tb6_hlist) { | |
c71099ac TG |
242 | if (tb->tb6_id == id) { |
243 | rcu_read_unlock(); | |
244 | return tb; | |
245 | } | |
246 | } | |
247 | rcu_read_unlock(); | |
248 | ||
249 | return NULL; | |
250 | } | |
251 | ||
58f09b78 | 252 | static void fib6_tables_init(struct net *net) |
c71099ac | 253 | { |
58f09b78 DL |
254 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
255 | fib6_link_table(net, net->ipv6.fib6_local_tbl); | |
c71099ac | 256 | } |
c71099ac TG |
257 | #else |
258 | ||
58f09b78 | 259 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac | 260 | { |
58f09b78 | 261 | return fib6_get_table(net, id); |
c71099ac TG |
262 | } |
263 | ||
58f09b78 | 264 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac | 265 | { |
58f09b78 | 266 | return net->ipv6.fib6_main_tbl; |
c71099ac TG |
267 | } |
268 | ||
58f09b78 DL |
269 | struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi *fl, |
270 | int flags, pol_lookup_t lookup) | |
c71099ac | 271 | { |
58f09b78 | 272 | return (struct dst_entry *) lookup(net->ipv6.fib6_main_tbl, fl, flags); |
c71099ac TG |
273 | } |
274 | ||
58f09b78 | 275 | static void fib6_tables_init(struct net *net) |
c71099ac | 276 | { |
58f09b78 | 277 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
c71099ac TG |
278 | } |
279 | ||
280 | #endif | |
281 | ||
1b43af54 PM |
282 | static int fib6_dump_node(struct fib6_walker_t *w) |
283 | { | |
284 | int res; | |
285 | struct rt6_info *rt; | |
286 | ||
7cc48263 | 287 | for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) { |
1b43af54 PM |
288 | res = rt6_dump_route(rt, w->args); |
289 | if (res < 0) { | |
290 | /* Frame is full, suspend walking */ | |
291 | w->leaf = rt; | |
292 | return 1; | |
293 | } | |
294 | BUG_TRAP(res!=0); | |
295 | } | |
296 | w->leaf = NULL; | |
297 | return 0; | |
298 | } | |
299 | ||
300 | static void fib6_dump_end(struct netlink_callback *cb) | |
301 | { | |
302 | struct fib6_walker_t *w = (void*)cb->args[2]; | |
303 | ||
304 | if (w) { | |
305 | cb->args[2] = 0; | |
306 | kfree(w); | |
307 | } | |
308 | cb->done = (void*)cb->args[3]; | |
309 | cb->args[1] = 3; | |
310 | } | |
311 | ||
312 | static int fib6_dump_done(struct netlink_callback *cb) | |
313 | { | |
314 | fib6_dump_end(cb); | |
315 | return cb->done ? cb->done(cb) : 0; | |
316 | } | |
317 | ||
318 | static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb, | |
319 | struct netlink_callback *cb) | |
320 | { | |
321 | struct fib6_walker_t *w; | |
322 | int res; | |
323 | ||
324 | w = (void *)cb->args[2]; | |
325 | w->root = &table->tb6_root; | |
326 | ||
327 | if (cb->args[4] == 0) { | |
328 | read_lock_bh(&table->tb6_lock); | |
329 | res = fib6_walk(w); | |
330 | read_unlock_bh(&table->tb6_lock); | |
331 | if (res > 0) | |
332 | cb->args[4] = 1; | |
333 | } else { | |
334 | read_lock_bh(&table->tb6_lock); | |
335 | res = fib6_walk_continue(w); | |
336 | read_unlock_bh(&table->tb6_lock); | |
337 | if (res != 0) { | |
338 | if (res < 0) | |
339 | fib6_walker_unlink(w); | |
340 | goto end; | |
341 | } | |
342 | fib6_walker_unlink(w); | |
343 | cb->args[4] = 0; | |
344 | } | |
345 | end: | |
346 | return res; | |
347 | } | |
348 | ||
c127ea2c | 349 | static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) |
1b43af54 | 350 | { |
b854272b | 351 | struct net *net = skb->sk->sk_net; |
1b43af54 PM |
352 | unsigned int h, s_h; |
353 | unsigned int e = 0, s_e; | |
354 | struct rt6_rtnl_dump_arg arg; | |
355 | struct fib6_walker_t *w; | |
356 | struct fib6_table *tb; | |
357 | struct hlist_node *node; | |
58f09b78 | 358 | struct hlist_head *head; |
1b43af54 PM |
359 | int res = 0; |
360 | ||
361 | s_h = cb->args[0]; | |
362 | s_e = cb->args[1]; | |
363 | ||
364 | w = (void *)cb->args[2]; | |
365 | if (w == NULL) { | |
366 | /* New dump: | |
367 | * | |
368 | * 1. hook callback destructor. | |
369 | */ | |
370 | cb->args[3] = (long)cb->done; | |
371 | cb->done = fib6_dump_done; | |
372 | ||
373 | /* | |
374 | * 2. allocate and initialize walker. | |
375 | */ | |
376 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
377 | if (w == NULL) | |
378 | return -ENOMEM; | |
379 | w->func = fib6_dump_node; | |
380 | cb->args[2] = (long)w; | |
381 | } | |
382 | ||
383 | arg.skb = skb; | |
384 | arg.cb = cb; | |
385 | w->args = &arg; | |
386 | ||
387 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { | |
388 | e = 0; | |
58f09b78 DL |
389 | head = &net->ipv6.fib_table_hash[h]; |
390 | hlist_for_each_entry(tb, node, head, tb6_hlist) { | |
1b43af54 PM |
391 | if (e < s_e) |
392 | goto next; | |
393 | res = fib6_dump_table(tb, skb, cb); | |
394 | if (res != 0) | |
395 | goto out; | |
396 | next: | |
397 | e++; | |
398 | } | |
399 | } | |
400 | out: | |
401 | cb->args[1] = e; | |
402 | cb->args[0] = h; | |
403 | ||
404 | res = res < 0 ? res : skb->len; | |
405 | if (res <= 0) | |
406 | fib6_dump_end(cb); | |
407 | return res; | |
408 | } | |
1da177e4 LT |
409 | |
410 | /* | |
411 | * Routing Table | |
412 | * | |
413 | * return the appropriate node for a routing tree "add" operation | |
414 | * by either creating and inserting or by returning an existing | |
415 | * node. | |
416 | */ | |
417 | ||
418 | static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr, | |
419 | int addrlen, int plen, | |
420 | int offset) | |
421 | { | |
422 | struct fib6_node *fn, *in, *ln; | |
423 | struct fib6_node *pn = NULL; | |
424 | struct rt6key *key; | |
425 | int bit; | |
1ab1457c | 426 | __be32 dir = 0; |
1da177e4 LT |
427 | __u32 sernum = fib6_new_sernum(); |
428 | ||
429 | RT6_TRACE("fib6_add_1\n"); | |
430 | ||
431 | /* insert node in tree */ | |
432 | ||
433 | fn = root; | |
434 | ||
435 | do { | |
436 | key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
437 | ||
438 | /* | |
439 | * Prefix match | |
440 | */ | |
441 | if (plen < fn->fn_bit || | |
442 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
443 | goto insert_above; | |
1ab1457c | 444 | |
1da177e4 LT |
445 | /* |
446 | * Exact match ? | |
447 | */ | |
1ab1457c | 448 | |
1da177e4 LT |
449 | if (plen == fn->fn_bit) { |
450 | /* clean up an intermediate node */ | |
451 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
452 | rt6_release(fn->leaf); | |
453 | fn->leaf = NULL; | |
454 | } | |
1ab1457c | 455 | |
1da177e4 | 456 | fn->fn_sernum = sernum; |
1ab1457c | 457 | |
1da177e4 LT |
458 | return fn; |
459 | } | |
460 | ||
461 | /* | |
462 | * We have more bits to go | |
463 | */ | |
1ab1457c | 464 | |
1da177e4 LT |
465 | /* Try to walk down on tree. */ |
466 | fn->fn_sernum = sernum; | |
467 | dir = addr_bit_set(addr, fn->fn_bit); | |
468 | pn = fn; | |
469 | fn = dir ? fn->right: fn->left; | |
470 | } while (fn); | |
471 | ||
472 | /* | |
473 | * We walked to the bottom of tree. | |
474 | * Create new leaf node without children. | |
475 | */ | |
476 | ||
477 | ln = node_alloc(); | |
478 | ||
479 | if (ln == NULL) | |
480 | return NULL; | |
481 | ln->fn_bit = plen; | |
1ab1457c | 482 | |
1da177e4 LT |
483 | ln->parent = pn; |
484 | ln->fn_sernum = sernum; | |
485 | ||
486 | if (dir) | |
487 | pn->right = ln; | |
488 | else | |
489 | pn->left = ln; | |
490 | ||
491 | return ln; | |
492 | ||
493 | ||
494 | insert_above: | |
495 | /* | |
1ab1457c | 496 | * split since we don't have a common prefix anymore or |
1da177e4 LT |
497 | * we have a less significant route. |
498 | * we've to insert an intermediate node on the list | |
499 | * this new node will point to the one we need to create | |
500 | * and the current | |
501 | */ | |
502 | ||
503 | pn = fn->parent; | |
504 | ||
505 | /* find 1st bit in difference between the 2 addrs. | |
506 | ||
971f359d | 507 | See comment in __ipv6_addr_diff: bit may be an invalid value, |
1da177e4 LT |
508 | but if it is >= plen, the value is ignored in any case. |
509 | */ | |
1ab1457c | 510 | |
971f359d | 511 | bit = __ipv6_addr_diff(addr, &key->addr, addrlen); |
1da177e4 | 512 | |
1ab1457c YH |
513 | /* |
514 | * (intermediate)[in] | |
1da177e4 LT |
515 | * / \ |
516 | * (new leaf node)[ln] (old node)[fn] | |
517 | */ | |
518 | if (plen > bit) { | |
519 | in = node_alloc(); | |
520 | ln = node_alloc(); | |
1ab1457c | 521 | |
1da177e4 LT |
522 | if (in == NULL || ln == NULL) { |
523 | if (in) | |
524 | node_free(in); | |
525 | if (ln) | |
526 | node_free(ln); | |
527 | return NULL; | |
528 | } | |
529 | ||
1ab1457c YH |
530 | /* |
531 | * new intermediate node. | |
1da177e4 LT |
532 | * RTN_RTINFO will |
533 | * be off since that an address that chooses one of | |
534 | * the branches would not match less specific routes | |
535 | * in the other branch | |
536 | */ | |
537 | ||
538 | in->fn_bit = bit; | |
539 | ||
540 | in->parent = pn; | |
541 | in->leaf = fn->leaf; | |
542 | atomic_inc(&in->leaf->rt6i_ref); | |
543 | ||
544 | in->fn_sernum = sernum; | |
545 | ||
546 | /* update parent pointer */ | |
547 | if (dir) | |
548 | pn->right = in; | |
549 | else | |
550 | pn->left = in; | |
551 | ||
552 | ln->fn_bit = plen; | |
553 | ||
554 | ln->parent = in; | |
555 | fn->parent = in; | |
556 | ||
557 | ln->fn_sernum = sernum; | |
558 | ||
559 | if (addr_bit_set(addr, bit)) { | |
560 | in->right = ln; | |
561 | in->left = fn; | |
562 | } else { | |
563 | in->left = ln; | |
564 | in->right = fn; | |
565 | } | |
566 | } else { /* plen <= bit */ | |
567 | ||
1ab1457c | 568 | /* |
1da177e4 LT |
569 | * (new leaf node)[ln] |
570 | * / \ | |
571 | * (old node)[fn] NULL | |
572 | */ | |
573 | ||
574 | ln = node_alloc(); | |
575 | ||
576 | if (ln == NULL) | |
577 | return NULL; | |
578 | ||
579 | ln->fn_bit = plen; | |
580 | ||
581 | ln->parent = pn; | |
582 | ||
583 | ln->fn_sernum = sernum; | |
1ab1457c | 584 | |
1da177e4 LT |
585 | if (dir) |
586 | pn->right = ln; | |
587 | else | |
588 | pn->left = ln; | |
589 | ||
590 | if (addr_bit_set(&key->addr, plen)) | |
591 | ln->right = fn; | |
592 | else | |
593 | ln->left = fn; | |
594 | ||
595 | fn->parent = ln; | |
596 | } | |
597 | return ln; | |
598 | } | |
599 | ||
600 | /* | |
601 | * Insert routing information in a node. | |
602 | */ | |
603 | ||
604 | static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt, | |
86872cb5 | 605 | struct nl_info *info) |
1da177e4 LT |
606 | { |
607 | struct rt6_info *iter = NULL; | |
608 | struct rt6_info **ins; | |
609 | ||
610 | ins = &fn->leaf; | |
611 | ||
7cc48263 | 612 | for (iter = fn->leaf; iter; iter=iter->u.dst.rt6_next) { |
1da177e4 LT |
613 | /* |
614 | * Search for duplicates | |
615 | */ | |
616 | ||
617 | if (iter->rt6i_metric == rt->rt6i_metric) { | |
618 | /* | |
619 | * Same priority level | |
620 | */ | |
621 | ||
622 | if (iter->rt6i_dev == rt->rt6i_dev && | |
623 | iter->rt6i_idev == rt->rt6i_idev && | |
624 | ipv6_addr_equal(&iter->rt6i_gateway, | |
625 | &rt->rt6i_gateway)) { | |
626 | if (!(iter->rt6i_flags&RTF_EXPIRES)) | |
627 | return -EEXIST; | |
628 | iter->rt6i_expires = rt->rt6i_expires; | |
629 | if (!(rt->rt6i_flags&RTF_EXPIRES)) { | |
630 | iter->rt6i_flags &= ~RTF_EXPIRES; | |
631 | iter->rt6i_expires = 0; | |
632 | } | |
633 | return -EEXIST; | |
634 | } | |
635 | } | |
636 | ||
637 | if (iter->rt6i_metric > rt->rt6i_metric) | |
638 | break; | |
639 | ||
7cc48263 | 640 | ins = &iter->u.dst.rt6_next; |
1da177e4 LT |
641 | } |
642 | ||
f11e6659 DM |
643 | /* Reset round-robin state, if necessary */ |
644 | if (ins == &fn->leaf) | |
645 | fn->rr_ptr = NULL; | |
646 | ||
1da177e4 LT |
647 | /* |
648 | * insert node | |
649 | */ | |
650 | ||
7cc48263 | 651 | rt->u.dst.rt6_next = iter; |
1da177e4 LT |
652 | *ins = rt; |
653 | rt->rt6i_node = fn; | |
654 | atomic_inc(&rt->rt6i_ref); | |
86872cb5 | 655 | inet6_rt_notify(RTM_NEWROUTE, rt, info); |
1da177e4 LT |
656 | rt6_stats.fib_rt_entries++; |
657 | ||
658 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
659 | rt6_stats.fib_route_nodes++; | |
660 | fn->fn_flags |= RTN_RTINFO; | |
661 | } | |
662 | ||
663 | return 0; | |
664 | } | |
665 | ||
666 | static __inline__ void fib6_start_gc(struct rt6_info *rt) | |
667 | { | |
450d19f8 | 668 | if (ip6_fib_timer->expires == 0 && |
1da177e4 | 669 | (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE))) |
450d19f8 | 670 | mod_timer(ip6_fib_timer, jiffies + |
4990509f | 671 | init_net.ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 LT |
672 | } |
673 | ||
674 | void fib6_force_start_gc(void) | |
675 | { | |
450d19f8 DL |
676 | if (ip6_fib_timer->expires == 0) |
677 | mod_timer(ip6_fib_timer, jiffies + | |
4990509f | 678 | init_net.ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 LT |
679 | } |
680 | ||
681 | /* | |
682 | * Add routing information to the routing tree. | |
683 | * <destination addr>/<source addr> | |
684 | * with source addr info in sub-trees | |
685 | */ | |
686 | ||
86872cb5 | 687 | int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info) |
1da177e4 | 688 | { |
66729e18 | 689 | struct fib6_node *fn, *pn = NULL; |
1da177e4 LT |
690 | int err = -ENOMEM; |
691 | ||
692 | fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr), | |
693 | rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst)); | |
694 | ||
695 | if (fn == NULL) | |
696 | goto out; | |
697 | ||
66729e18 YH |
698 | pn = fn; |
699 | ||
1da177e4 LT |
700 | #ifdef CONFIG_IPV6_SUBTREES |
701 | if (rt->rt6i_src.plen) { | |
702 | struct fib6_node *sn; | |
703 | ||
704 | if (fn->subtree == NULL) { | |
705 | struct fib6_node *sfn; | |
706 | ||
707 | /* | |
708 | * Create subtree. | |
709 | * | |
710 | * fn[main tree] | |
711 | * | | |
712 | * sfn[subtree root] | |
713 | * \ | |
714 | * sn[new leaf node] | |
715 | */ | |
716 | ||
717 | /* Create subtree root node */ | |
718 | sfn = node_alloc(); | |
719 | if (sfn == NULL) | |
720 | goto st_failure; | |
721 | ||
722 | sfn->leaf = &ip6_null_entry; | |
723 | atomic_inc(&ip6_null_entry.rt6i_ref); | |
724 | sfn->fn_flags = RTN_ROOT; | |
725 | sfn->fn_sernum = fib6_new_sernum(); | |
726 | ||
727 | /* Now add the first leaf node to new subtree */ | |
728 | ||
729 | sn = fib6_add_1(sfn, &rt->rt6i_src.addr, | |
730 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
731 | offsetof(struct rt6_info, rt6i_src)); | |
732 | ||
733 | if (sn == NULL) { | |
734 | /* If it is failed, discard just allocated | |
735 | root, and then (in st_failure) stale node | |
736 | in main tree. | |
737 | */ | |
738 | node_free(sfn); | |
739 | goto st_failure; | |
740 | } | |
741 | ||
742 | /* Now link new subtree to main tree */ | |
743 | sfn->parent = fn; | |
744 | fn->subtree = sfn; | |
1da177e4 LT |
745 | } else { |
746 | sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr, | |
747 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
748 | offsetof(struct rt6_info, rt6i_src)); | |
749 | ||
750 | if (sn == NULL) | |
751 | goto st_failure; | |
752 | } | |
753 | ||
66729e18 YH |
754 | if (fn->leaf == NULL) { |
755 | fn->leaf = rt; | |
756 | atomic_inc(&rt->rt6i_ref); | |
757 | } | |
1da177e4 LT |
758 | fn = sn; |
759 | } | |
760 | #endif | |
761 | ||
86872cb5 | 762 | err = fib6_add_rt2node(fn, rt, info); |
1da177e4 LT |
763 | |
764 | if (err == 0) { | |
765 | fib6_start_gc(rt); | |
766 | if (!(rt->rt6i_flags&RTF_CACHE)) | |
2285adc1 | 767 | fib6_prune_clones(pn, rt); |
1da177e4 LT |
768 | } |
769 | ||
770 | out: | |
66729e18 YH |
771 | if (err) { |
772 | #ifdef CONFIG_IPV6_SUBTREES | |
773 | /* | |
774 | * If fib6_add_1 has cleared the old leaf pointer in the | |
775 | * super-tree leaf node we have to find a new one for it. | |
776 | */ | |
777 | if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) { | |
778 | pn->leaf = fib6_find_prefix(pn); | |
779 | #if RT6_DEBUG >= 2 | |
780 | if (!pn->leaf) { | |
781 | BUG_TRAP(pn->leaf != NULL); | |
782 | pn->leaf = &ip6_null_entry; | |
783 | } | |
784 | #endif | |
785 | atomic_inc(&pn->leaf->rt6i_ref); | |
786 | } | |
787 | #endif | |
1da177e4 | 788 | dst_free(&rt->u.dst); |
66729e18 | 789 | } |
1da177e4 LT |
790 | return err; |
791 | ||
792 | #ifdef CONFIG_IPV6_SUBTREES | |
793 | /* Subtree creation failed, probably main tree node | |
794 | is orphan. If it is, shoot it. | |
795 | */ | |
796 | st_failure: | |
797 | if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT))) | |
798 | fib6_repair_tree(fn); | |
799 | dst_free(&rt->u.dst); | |
800 | return err; | |
801 | #endif | |
802 | } | |
803 | ||
804 | /* | |
805 | * Routing tree lookup | |
806 | * | |
807 | */ | |
808 | ||
809 | struct lookup_args { | |
810 | int offset; /* key offset on rt6_info */ | |
811 | struct in6_addr *addr; /* search key */ | |
812 | }; | |
813 | ||
814 | static struct fib6_node * fib6_lookup_1(struct fib6_node *root, | |
815 | struct lookup_args *args) | |
816 | { | |
817 | struct fib6_node *fn; | |
e69a4adc | 818 | __be32 dir; |
1da177e4 | 819 | |
825e288e YH |
820 | if (unlikely(args->offset == 0)) |
821 | return NULL; | |
822 | ||
1da177e4 LT |
823 | /* |
824 | * Descend on a tree | |
825 | */ | |
826 | ||
827 | fn = root; | |
828 | ||
829 | for (;;) { | |
830 | struct fib6_node *next; | |
831 | ||
832 | dir = addr_bit_set(args->addr, fn->fn_bit); | |
833 | ||
834 | next = dir ? fn->right : fn->left; | |
835 | ||
836 | if (next) { | |
837 | fn = next; | |
838 | continue; | |
839 | } | |
840 | ||
841 | break; | |
842 | } | |
843 | ||
3fc5e044 | 844 | while(fn) { |
7fc33165 | 845 | if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) { |
1da177e4 LT |
846 | struct rt6key *key; |
847 | ||
848 | key = (struct rt6key *) ((u8 *) fn->leaf + | |
849 | args->offset); | |
850 | ||
3fc5e044 YH |
851 | if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) { |
852 | #ifdef CONFIG_IPV6_SUBTREES | |
853 | if (fn->subtree) | |
854 | fn = fib6_lookup_1(fn->subtree, args + 1); | |
855 | #endif | |
856 | if (!fn || fn->fn_flags & RTN_RTINFO) | |
857 | return fn; | |
858 | } | |
1da177e4 LT |
859 | } |
860 | ||
3fc5e044 YH |
861 | if (fn->fn_flags & RTN_ROOT) |
862 | break; | |
863 | ||
1da177e4 LT |
864 | fn = fn->parent; |
865 | } | |
866 | ||
867 | return NULL; | |
868 | } | |
869 | ||
870 | struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr, | |
871 | struct in6_addr *saddr) | |
872 | { | |
1da177e4 | 873 | struct fib6_node *fn; |
825e288e YH |
874 | struct lookup_args args[] = { |
875 | { | |
876 | .offset = offsetof(struct rt6_info, rt6i_dst), | |
877 | .addr = daddr, | |
878 | }, | |
1da177e4 | 879 | #ifdef CONFIG_IPV6_SUBTREES |
825e288e YH |
880 | { |
881 | .offset = offsetof(struct rt6_info, rt6i_src), | |
882 | .addr = saddr, | |
883 | }, | |
1da177e4 | 884 | #endif |
825e288e YH |
885 | { |
886 | .offset = 0, /* sentinel */ | |
887 | } | |
888 | }; | |
1da177e4 | 889 | |
fefc2a6c | 890 | fn = fib6_lookup_1(root, daddr ? args : args + 1); |
1da177e4 LT |
891 | |
892 | if (fn == NULL || fn->fn_flags & RTN_TL_ROOT) | |
893 | fn = root; | |
894 | ||
895 | return fn; | |
896 | } | |
897 | ||
898 | /* | |
899 | * Get node with specified destination prefix (and source prefix, | |
900 | * if subtrees are used) | |
901 | */ | |
902 | ||
903 | ||
904 | static struct fib6_node * fib6_locate_1(struct fib6_node *root, | |
905 | struct in6_addr *addr, | |
906 | int plen, int offset) | |
907 | { | |
908 | struct fib6_node *fn; | |
909 | ||
910 | for (fn = root; fn ; ) { | |
911 | struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
912 | ||
913 | /* | |
914 | * Prefix match | |
915 | */ | |
916 | if (plen < fn->fn_bit || | |
917 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
918 | return NULL; | |
919 | ||
920 | if (plen == fn->fn_bit) | |
921 | return fn; | |
922 | ||
923 | /* | |
924 | * We have more bits to go | |
925 | */ | |
926 | if (addr_bit_set(addr, fn->fn_bit)) | |
927 | fn = fn->right; | |
928 | else | |
929 | fn = fn->left; | |
930 | } | |
931 | return NULL; | |
932 | } | |
933 | ||
934 | struct fib6_node * fib6_locate(struct fib6_node *root, | |
935 | struct in6_addr *daddr, int dst_len, | |
936 | struct in6_addr *saddr, int src_len) | |
937 | { | |
938 | struct fib6_node *fn; | |
939 | ||
940 | fn = fib6_locate_1(root, daddr, dst_len, | |
941 | offsetof(struct rt6_info, rt6i_dst)); | |
942 | ||
943 | #ifdef CONFIG_IPV6_SUBTREES | |
944 | if (src_len) { | |
945 | BUG_TRAP(saddr!=NULL); | |
3fc5e044 YH |
946 | if (fn && fn->subtree) |
947 | fn = fib6_locate_1(fn->subtree, saddr, src_len, | |
1da177e4 LT |
948 | offsetof(struct rt6_info, rt6i_src)); |
949 | } | |
950 | #endif | |
951 | ||
952 | if (fn && fn->fn_flags&RTN_RTINFO) | |
953 | return fn; | |
954 | ||
955 | return NULL; | |
956 | } | |
957 | ||
958 | ||
959 | /* | |
960 | * Deletion | |
961 | * | |
962 | */ | |
963 | ||
964 | static struct rt6_info * fib6_find_prefix(struct fib6_node *fn) | |
965 | { | |
966 | if (fn->fn_flags&RTN_ROOT) | |
967 | return &ip6_null_entry; | |
968 | ||
969 | while(fn) { | |
970 | if(fn->left) | |
971 | return fn->left->leaf; | |
972 | ||
973 | if(fn->right) | |
974 | return fn->right->leaf; | |
975 | ||
7fc33165 | 976 | fn = FIB6_SUBTREE(fn); |
1da177e4 LT |
977 | } |
978 | return NULL; | |
979 | } | |
980 | ||
981 | /* | |
982 | * Called to trim the tree of intermediate nodes when possible. "fn" | |
983 | * is the node we want to try and remove. | |
984 | */ | |
985 | ||
986 | static struct fib6_node * fib6_repair_tree(struct fib6_node *fn) | |
987 | { | |
988 | int children; | |
989 | int nstate; | |
990 | struct fib6_node *child, *pn; | |
991 | struct fib6_walker_t *w; | |
992 | int iter = 0; | |
993 | ||
994 | for (;;) { | |
995 | RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter); | |
996 | iter++; | |
997 | ||
998 | BUG_TRAP(!(fn->fn_flags&RTN_RTINFO)); | |
999 | BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT)); | |
1000 | BUG_TRAP(fn->leaf==NULL); | |
1001 | ||
1002 | children = 0; | |
1003 | child = NULL; | |
1004 | if (fn->right) child = fn->right, children |= 1; | |
1005 | if (fn->left) child = fn->left, children |= 2; | |
1006 | ||
7fc33165 | 1007 | if (children == 3 || FIB6_SUBTREE(fn) |
1da177e4 LT |
1008 | #ifdef CONFIG_IPV6_SUBTREES |
1009 | /* Subtree root (i.e. fn) may have one child */ | |
1010 | || (children && fn->fn_flags&RTN_ROOT) | |
1011 | #endif | |
1012 | ) { | |
1013 | fn->leaf = fib6_find_prefix(fn); | |
1014 | #if RT6_DEBUG >= 2 | |
1015 | if (fn->leaf==NULL) { | |
1016 | BUG_TRAP(fn->leaf); | |
1017 | fn->leaf = &ip6_null_entry; | |
1018 | } | |
1019 | #endif | |
1020 | atomic_inc(&fn->leaf->rt6i_ref); | |
1021 | return fn->parent; | |
1022 | } | |
1023 | ||
1024 | pn = fn->parent; | |
1025 | #ifdef CONFIG_IPV6_SUBTREES | |
7fc33165 | 1026 | if (FIB6_SUBTREE(pn) == fn) { |
1da177e4 | 1027 | BUG_TRAP(fn->fn_flags&RTN_ROOT); |
7fc33165 | 1028 | FIB6_SUBTREE(pn) = NULL; |
1da177e4 LT |
1029 | nstate = FWS_L; |
1030 | } else { | |
1031 | BUG_TRAP(!(fn->fn_flags&RTN_ROOT)); | |
1032 | #endif | |
1033 | if (pn->right == fn) pn->right = child; | |
1034 | else if (pn->left == fn) pn->left = child; | |
1035 | #if RT6_DEBUG >= 2 | |
1036 | else BUG_TRAP(0); | |
1037 | #endif | |
1038 | if (child) | |
1039 | child->parent = pn; | |
1040 | nstate = FWS_R; | |
1041 | #ifdef CONFIG_IPV6_SUBTREES | |
1042 | } | |
1043 | #endif | |
1044 | ||
1045 | read_lock(&fib6_walker_lock); | |
1046 | FOR_WALKERS(w) { | |
1047 | if (child == NULL) { | |
1048 | if (w->root == fn) { | |
1049 | w->root = w->node = NULL; | |
1050 | RT6_TRACE("W %p adjusted by delroot 1\n", w); | |
1051 | } else if (w->node == fn) { | |
1052 | RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate); | |
1053 | w->node = pn; | |
1054 | w->state = nstate; | |
1055 | } | |
1056 | } else { | |
1057 | if (w->root == fn) { | |
1058 | w->root = child; | |
1059 | RT6_TRACE("W %p adjusted by delroot 2\n", w); | |
1060 | } | |
1061 | if (w->node == fn) { | |
1062 | w->node = child; | |
1063 | if (children&2) { | |
1064 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1065 | w->state = w->state>=FWS_R ? FWS_U : FWS_INIT; | |
1066 | } else { | |
1067 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1068 | w->state = w->state>=FWS_C ? FWS_U : FWS_INIT; | |
1069 | } | |
1070 | } | |
1071 | } | |
1072 | } | |
1073 | read_unlock(&fib6_walker_lock); | |
1074 | ||
1075 | node_free(fn); | |
7fc33165 | 1076 | if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn)) |
1da177e4 LT |
1077 | return pn; |
1078 | ||
1079 | rt6_release(pn->leaf); | |
1080 | pn->leaf = NULL; | |
1081 | fn = pn; | |
1082 | } | |
1083 | } | |
1084 | ||
1085 | static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp, | |
86872cb5 | 1086 | struct nl_info *info) |
1da177e4 LT |
1087 | { |
1088 | struct fib6_walker_t *w; | |
1089 | struct rt6_info *rt = *rtp; | |
1090 | ||
1091 | RT6_TRACE("fib6_del_route\n"); | |
1092 | ||
1093 | /* Unlink it */ | |
7cc48263 | 1094 | *rtp = rt->u.dst.rt6_next; |
1da177e4 LT |
1095 | rt->rt6i_node = NULL; |
1096 | rt6_stats.fib_rt_entries--; | |
1097 | rt6_stats.fib_discarded_routes++; | |
1098 | ||
f11e6659 DM |
1099 | /* Reset round-robin state, if necessary */ |
1100 | if (fn->rr_ptr == rt) | |
1101 | fn->rr_ptr = NULL; | |
1102 | ||
1da177e4 LT |
1103 | /* Adjust walkers */ |
1104 | read_lock(&fib6_walker_lock); | |
1105 | FOR_WALKERS(w) { | |
1106 | if (w->state == FWS_C && w->leaf == rt) { | |
1107 | RT6_TRACE("walker %p adjusted by delroute\n", w); | |
7cc48263 | 1108 | w->leaf = rt->u.dst.rt6_next; |
1da177e4 LT |
1109 | if (w->leaf == NULL) |
1110 | w->state = FWS_U; | |
1111 | } | |
1112 | } | |
1113 | read_unlock(&fib6_walker_lock); | |
1114 | ||
7cc48263 | 1115 | rt->u.dst.rt6_next = NULL; |
1da177e4 | 1116 | |
1da177e4 LT |
1117 | /* If it was last route, expunge its radix tree node */ |
1118 | if (fn->leaf == NULL) { | |
1119 | fn->fn_flags &= ~RTN_RTINFO; | |
1120 | rt6_stats.fib_route_nodes--; | |
1121 | fn = fib6_repair_tree(fn); | |
1122 | } | |
1123 | ||
1124 | if (atomic_read(&rt->rt6i_ref) != 1) { | |
1125 | /* This route is used as dummy address holder in some split | |
1126 | * nodes. It is not leaked, but it still holds other resources, | |
1127 | * which must be released in time. So, scan ascendant nodes | |
1128 | * and replace dummy references to this route with references | |
1129 | * to still alive ones. | |
1130 | */ | |
1131 | while (fn) { | |
1132 | if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) { | |
1133 | fn->leaf = fib6_find_prefix(fn); | |
1134 | atomic_inc(&fn->leaf->rt6i_ref); | |
1135 | rt6_release(rt); | |
1136 | } | |
1137 | fn = fn->parent; | |
1138 | } | |
1139 | /* No more references are possible at this point. */ | |
2df96af0 | 1140 | BUG_ON(atomic_read(&rt->rt6i_ref) != 1); |
1da177e4 LT |
1141 | } |
1142 | ||
86872cb5 | 1143 | inet6_rt_notify(RTM_DELROUTE, rt, info); |
1da177e4 LT |
1144 | rt6_release(rt); |
1145 | } | |
1146 | ||
86872cb5 | 1147 | int fib6_del(struct rt6_info *rt, struct nl_info *info) |
1da177e4 LT |
1148 | { |
1149 | struct fib6_node *fn = rt->rt6i_node; | |
1150 | struct rt6_info **rtp; | |
1151 | ||
1152 | #if RT6_DEBUG >= 2 | |
1153 | if (rt->u.dst.obsolete>0) { | |
1154 | BUG_TRAP(fn==NULL); | |
1155 | return -ENOENT; | |
1156 | } | |
1157 | #endif | |
1158 | if (fn == NULL || rt == &ip6_null_entry) | |
1159 | return -ENOENT; | |
1160 | ||
1161 | BUG_TRAP(fn->fn_flags&RTN_RTINFO); | |
1162 | ||
150730d5 YH |
1163 | if (!(rt->rt6i_flags&RTF_CACHE)) { |
1164 | struct fib6_node *pn = fn; | |
1165 | #ifdef CONFIG_IPV6_SUBTREES | |
1166 | /* clones of this route might be in another subtree */ | |
1167 | if (rt->rt6i_src.plen) { | |
1168 | while (!(pn->fn_flags&RTN_ROOT)) | |
1169 | pn = pn->parent; | |
1170 | pn = pn->parent; | |
1171 | } | |
1172 | #endif | |
1173 | fib6_prune_clones(pn, rt); | |
1174 | } | |
1da177e4 LT |
1175 | |
1176 | /* | |
1177 | * Walk the leaf entries looking for ourself | |
1178 | */ | |
1179 | ||
7cc48263 | 1180 | for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.dst.rt6_next) { |
1da177e4 | 1181 | if (*rtp == rt) { |
86872cb5 | 1182 | fib6_del_route(fn, rtp, info); |
1da177e4 LT |
1183 | return 0; |
1184 | } | |
1185 | } | |
1186 | return -ENOENT; | |
1187 | } | |
1188 | ||
1189 | /* | |
1190 | * Tree traversal function. | |
1191 | * | |
1192 | * Certainly, it is not interrupt safe. | |
1193 | * However, it is internally reenterable wrt itself and fib6_add/fib6_del. | |
1194 | * It means, that we can modify tree during walking | |
1195 | * and use this function for garbage collection, clone pruning, | |
1ab1457c | 1196 | * cleaning tree when a device goes down etc. etc. |
1da177e4 LT |
1197 | * |
1198 | * It guarantees that every node will be traversed, | |
1199 | * and that it will be traversed only once. | |
1200 | * | |
1201 | * Callback function w->func may return: | |
1202 | * 0 -> continue walking. | |
1203 | * positive value -> walking is suspended (used by tree dumps, | |
1204 | * and probably by gc, if it will be split to several slices) | |
1205 | * negative value -> terminate walking. | |
1206 | * | |
1207 | * The function itself returns: | |
1208 | * 0 -> walk is complete. | |
1209 | * >0 -> walk is incomplete (i.e. suspended) | |
1210 | * <0 -> walk is terminated by an error. | |
1211 | */ | |
1212 | ||
90d41122 | 1213 | static int fib6_walk_continue(struct fib6_walker_t *w) |
1da177e4 LT |
1214 | { |
1215 | struct fib6_node *fn, *pn; | |
1216 | ||
1217 | for (;;) { | |
1218 | fn = w->node; | |
1219 | if (fn == NULL) | |
1220 | return 0; | |
1221 | ||
1222 | if (w->prune && fn != w->root && | |
1223 | fn->fn_flags&RTN_RTINFO && w->state < FWS_C) { | |
1224 | w->state = FWS_C; | |
1225 | w->leaf = fn->leaf; | |
1226 | } | |
1227 | switch (w->state) { | |
1228 | #ifdef CONFIG_IPV6_SUBTREES | |
1229 | case FWS_S: | |
7fc33165 YH |
1230 | if (FIB6_SUBTREE(fn)) { |
1231 | w->node = FIB6_SUBTREE(fn); | |
1da177e4 LT |
1232 | continue; |
1233 | } | |
1234 | w->state = FWS_L; | |
1ab1457c | 1235 | #endif |
1da177e4 LT |
1236 | case FWS_L: |
1237 | if (fn->left) { | |
1238 | w->node = fn->left; | |
1239 | w->state = FWS_INIT; | |
1240 | continue; | |
1241 | } | |
1242 | w->state = FWS_R; | |
1243 | case FWS_R: | |
1244 | if (fn->right) { | |
1245 | w->node = fn->right; | |
1246 | w->state = FWS_INIT; | |
1247 | continue; | |
1248 | } | |
1249 | w->state = FWS_C; | |
1250 | w->leaf = fn->leaf; | |
1251 | case FWS_C: | |
1252 | if (w->leaf && fn->fn_flags&RTN_RTINFO) { | |
1253 | int err = w->func(w); | |
1254 | if (err) | |
1255 | return err; | |
1256 | continue; | |
1257 | } | |
1258 | w->state = FWS_U; | |
1259 | case FWS_U: | |
1260 | if (fn == w->root) | |
1261 | return 0; | |
1262 | pn = fn->parent; | |
1263 | w->node = pn; | |
1264 | #ifdef CONFIG_IPV6_SUBTREES | |
7fc33165 | 1265 | if (FIB6_SUBTREE(pn) == fn) { |
1da177e4 LT |
1266 | BUG_TRAP(fn->fn_flags&RTN_ROOT); |
1267 | w->state = FWS_L; | |
1268 | continue; | |
1269 | } | |
1270 | #endif | |
1271 | if (pn->left == fn) { | |
1272 | w->state = FWS_R; | |
1273 | continue; | |
1274 | } | |
1275 | if (pn->right == fn) { | |
1276 | w->state = FWS_C; | |
1277 | w->leaf = w->node->leaf; | |
1278 | continue; | |
1279 | } | |
1280 | #if RT6_DEBUG >= 2 | |
1281 | BUG_TRAP(0); | |
1282 | #endif | |
1283 | } | |
1284 | } | |
1285 | } | |
1286 | ||
90d41122 | 1287 | static int fib6_walk(struct fib6_walker_t *w) |
1da177e4 LT |
1288 | { |
1289 | int res; | |
1290 | ||
1291 | w->state = FWS_INIT; | |
1292 | w->node = w->root; | |
1293 | ||
1294 | fib6_walker_link(w); | |
1295 | res = fib6_walk_continue(w); | |
1296 | if (res <= 0) | |
1297 | fib6_walker_unlink(w); | |
1298 | return res; | |
1299 | } | |
1300 | ||
1301 | static int fib6_clean_node(struct fib6_walker_t *w) | |
1302 | { | |
4d1169c1 DL |
1303 | struct nl_info info = { |
1304 | .nl_net = &init_net, | |
1305 | }; | |
1da177e4 LT |
1306 | int res; |
1307 | struct rt6_info *rt; | |
0a8891a0 | 1308 | struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w); |
1da177e4 | 1309 | |
7cc48263 | 1310 | for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) { |
1da177e4 LT |
1311 | res = c->func(rt, c->arg); |
1312 | if (res < 0) { | |
1313 | w->leaf = rt; | |
528c4ceb | 1314 | res = fib6_del(rt, &info); |
1da177e4 LT |
1315 | if (res) { |
1316 | #if RT6_DEBUG >= 2 | |
1317 | printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res); | |
1318 | #endif | |
1319 | continue; | |
1320 | } | |
1321 | return 0; | |
1322 | } | |
1323 | BUG_TRAP(res==0); | |
1324 | } | |
1325 | w->leaf = rt; | |
1326 | return 0; | |
1327 | } | |
1328 | ||
1329 | /* | |
1330 | * Convenient frontend to tree walker. | |
1ab1457c | 1331 | * |
1da177e4 LT |
1332 | * func is called on each route. |
1333 | * It may return -1 -> delete this route. | |
1334 | * 0 -> continue walking | |
1335 | * | |
1336 | * prune==1 -> only immediate children of node (certainly, | |
1337 | * ignoring pure split nodes) will be scanned. | |
1338 | */ | |
1339 | ||
8ce11e6a AB |
1340 | static void fib6_clean_tree(struct fib6_node *root, |
1341 | int (*func)(struct rt6_info *, void *arg), | |
1342 | int prune, void *arg) | |
1da177e4 LT |
1343 | { |
1344 | struct fib6_cleaner_t c; | |
1345 | ||
1346 | c.w.root = root; | |
1347 | c.w.func = fib6_clean_node; | |
1348 | c.w.prune = prune; | |
1349 | c.func = func; | |
1350 | c.arg = arg; | |
1351 | ||
1352 | fib6_walk(&c.w); | |
1353 | } | |
1354 | ||
f3db4851 | 1355 | void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg), |
c71099ac TG |
1356 | int prune, void *arg) |
1357 | { | |
c71099ac | 1358 | struct fib6_table *table; |
1b43af54 | 1359 | struct hlist_node *node; |
58f09b78 | 1360 | struct hlist_head *head; |
1b43af54 | 1361 | unsigned int h; |
c71099ac | 1362 | |
1b43af54 PM |
1363 | rcu_read_lock(); |
1364 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { | |
f3db4851 | 1365 | head = &net->ipv6.fib_table_hash[h]; |
58f09b78 | 1366 | hlist_for_each_entry_rcu(table, node, head, tb6_hlist) { |
c71099ac TG |
1367 | write_lock_bh(&table->tb6_lock); |
1368 | fib6_clean_tree(&table->tb6_root, func, prune, arg); | |
1369 | write_unlock_bh(&table->tb6_lock); | |
1370 | } | |
1371 | } | |
1b43af54 | 1372 | rcu_read_unlock(); |
c71099ac TG |
1373 | } |
1374 | ||
1da177e4 LT |
1375 | static int fib6_prune_clone(struct rt6_info *rt, void *arg) |
1376 | { | |
1377 | if (rt->rt6i_flags & RTF_CACHE) { | |
1378 | RT6_TRACE("pruning clone %p\n", rt); | |
1379 | return -1; | |
1380 | } | |
1381 | ||
1382 | return 0; | |
1383 | } | |
1384 | ||
1385 | static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt) | |
1386 | { | |
1387 | fib6_clean_tree(fn, fib6_prune_clone, 1, rt); | |
1388 | } | |
1389 | ||
1390 | /* | |
1391 | * Garbage collection | |
1392 | */ | |
1393 | ||
1394 | static struct fib6_gc_args | |
1395 | { | |
1396 | int timeout; | |
1397 | int more; | |
1398 | } gc_args; | |
1399 | ||
1400 | static int fib6_age(struct rt6_info *rt, void *arg) | |
1401 | { | |
1402 | unsigned long now = jiffies; | |
1403 | ||
1404 | /* | |
1405 | * check addrconf expiration here. | |
1406 | * Routes are expired even if they are in use. | |
1407 | * | |
1408 | * Also age clones. Note, that clones are aged out | |
1409 | * only if they are not in use now. | |
1410 | */ | |
1411 | ||
1412 | if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) { | |
1413 | if (time_after(now, rt->rt6i_expires)) { | |
1414 | RT6_TRACE("expiring %p\n", rt); | |
1da177e4 LT |
1415 | return -1; |
1416 | } | |
1417 | gc_args.more++; | |
1418 | } else if (rt->rt6i_flags & RTF_CACHE) { | |
1419 | if (atomic_read(&rt->u.dst.__refcnt) == 0 && | |
1420 | time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) { | |
1421 | RT6_TRACE("aging clone %p\n", rt); | |
1422 | return -1; | |
1423 | } else if ((rt->rt6i_flags & RTF_GATEWAY) && | |
1424 | (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) { | |
1425 | RT6_TRACE("purging route %p via non-router but gateway\n", | |
1426 | rt); | |
1427 | return -1; | |
1428 | } | |
1429 | gc_args.more++; | |
1430 | } | |
1431 | ||
1432 | return 0; | |
1433 | } | |
1434 | ||
1435 | static DEFINE_SPINLOCK(fib6_gc_lock); | |
1436 | ||
5b7c931d | 1437 | void fib6_run_gc(unsigned long expires, struct net *net) |
1da177e4 | 1438 | { |
5b7c931d | 1439 | if (expires != ~0UL) { |
1da177e4 | 1440 | spin_lock_bh(&fib6_gc_lock); |
5b7c931d DL |
1441 | gc_args.timeout = expires ? (int)expires : |
1442 | net->ipv6.sysctl.ip6_rt_gc_interval; | |
1da177e4 LT |
1443 | } else { |
1444 | local_bh_disable(); | |
1445 | if (!spin_trylock(&fib6_gc_lock)) { | |
450d19f8 | 1446 | mod_timer(ip6_fib_timer, jiffies + HZ); |
1da177e4 LT |
1447 | local_bh_enable(); |
1448 | return; | |
1449 | } | |
5b7c931d | 1450 | gc_args.timeout = net->ipv6.sysctl.ip6_rt_gc_interval; |
1da177e4 LT |
1451 | } |
1452 | gc_args.more = 0; | |
1453 | ||
3b00944c | 1454 | icmp6_dst_gc(&gc_args.more); |
f3db4851 | 1455 | |
5b7c931d | 1456 | fib6_clean_all(net, fib6_age, 0, NULL); |
1da177e4 LT |
1457 | |
1458 | if (gc_args.more) | |
450d19f8 | 1459 | mod_timer(ip6_fib_timer, jiffies + |
5b7c931d | 1460 | net->ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 | 1461 | else { |
450d19f8 DL |
1462 | del_timer(ip6_fib_timer); |
1463 | ip6_fib_timer->expires = 0; | |
1da177e4 LT |
1464 | } |
1465 | spin_unlock_bh(&fib6_gc_lock); | |
1466 | } | |
1467 | ||
5b7c931d DL |
1468 | static void fib6_gc_timer_cb(unsigned long arg) |
1469 | { | |
1470 | fib6_run_gc(0, (struct net *)arg); | |
1471 | } | |
1472 | ||
58f09b78 | 1473 | static int fib6_net_init(struct net *net) |
1da177e4 | 1474 | { |
58f09b78 | 1475 | int ret; |
e0b85590 | 1476 | |
58f09b78 DL |
1477 | ret = -ENOMEM; |
1478 | net->ipv6.fib_table_hash = | |
1479 | kzalloc(sizeof(*net->ipv6.fib_table_hash)*FIB_TABLE_HASHSZ, | |
1480 | GFP_KERNEL); | |
1481 | if (!net->ipv6.fib_table_hash) | |
1482 | goto out; | |
e0b85590 | 1483 | |
58f09b78 DL |
1484 | net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl), |
1485 | GFP_KERNEL); | |
1486 | if (!net->ipv6.fib6_main_tbl) | |
e0b85590 DL |
1487 | goto out_fib_table_hash; |
1488 | ||
58f09b78 DL |
1489 | net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN; |
1490 | net->ipv6.fib6_main_tbl->tb6_root.leaf = &ip6_null_entry; | |
1491 | net->ipv6.fib6_main_tbl->tb6_root.fn_flags = | |
1492 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
e0b85590 DL |
1493 | |
1494 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES | |
58f09b78 DL |
1495 | net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl), |
1496 | GFP_KERNEL); | |
1497 | if (!net->ipv6.fib6_local_tbl) | |
e0b85590 | 1498 | goto out_fib6_main_tbl; |
58f09b78 DL |
1499 | net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL; |
1500 | net->ipv6.fib6_local_tbl->tb6_root.leaf = &ip6_null_entry; | |
1501 | net->ipv6.fib6_local_tbl->tb6_root.fn_flags = | |
1502 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
e0b85590 | 1503 | #endif |
58f09b78 | 1504 | fib6_tables_init(net); |
f845ab6b | 1505 | |
58f09b78 | 1506 | ret = 0; |
d63bddbe DL |
1507 | out: |
1508 | return ret; | |
1509 | ||
e0b85590 | 1510 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 1511 | out_fib6_main_tbl: |
58f09b78 | 1512 | kfree(net->ipv6.fib6_main_tbl); |
e0b85590 | 1513 | #endif |
e0b85590 | 1514 | out_fib_table_hash: |
58f09b78 DL |
1515 | kfree(net->ipv6.fib_table_hash); |
1516 | goto out; | |
1517 | } | |
1518 | ||
1519 | static void fib6_net_exit(struct net *net) | |
1520 | { | |
1521 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES | |
1522 | kfree(net->ipv6.fib6_local_tbl); | |
1523 | #endif | |
1524 | kfree(net->ipv6.fib6_main_tbl); | |
1525 | kfree(net->ipv6.fib_table_hash); | |
1526 | } | |
1527 | ||
1528 | static struct pernet_operations fib6_net_ops = { | |
1529 | .init = fib6_net_init, | |
1530 | .exit = fib6_net_exit, | |
1531 | }; | |
1532 | ||
1533 | int __init fib6_init(void) | |
1534 | { | |
1535 | int ret = -ENOMEM; | |
1536 | fib6_node_kmem = kmem_cache_create("fib6_nodes", | |
1537 | sizeof(struct fib6_node), | |
1538 | 0, SLAB_HWCACHE_ALIGN, | |
1539 | NULL); | |
1540 | if (!fib6_node_kmem) | |
1541 | goto out; | |
1542 | ||
450d19f8 DL |
1543 | ret = -ENOMEM; |
1544 | ip6_fib_timer = kzalloc(sizeof(*ip6_fib_timer), GFP_KERNEL); | |
1545 | if (!ip6_fib_timer) | |
1546 | goto out_kmem_cache_create; | |
1547 | ||
1548 | setup_timer(ip6_fib_timer, fib6_gc_timer_cb, (unsigned long)&init_net); | |
1549 | ||
58f09b78 DL |
1550 | ret = register_pernet_subsys(&fib6_net_ops); |
1551 | if (ret) | |
450d19f8 | 1552 | goto out_timer; |
58f09b78 DL |
1553 | |
1554 | ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib); | |
1555 | if (ret) | |
1556 | goto out_unregister_subsys; | |
1557 | out: | |
1558 | return ret; | |
1559 | ||
1560 | out_unregister_subsys: | |
1561 | unregister_pernet_subsys(&fib6_net_ops); | |
450d19f8 DL |
1562 | out_timer: |
1563 | kfree(ip6_fib_timer); | |
d63bddbe DL |
1564 | out_kmem_cache_create: |
1565 | kmem_cache_destroy(fib6_node_kmem); | |
1566 | goto out; | |
1da177e4 LT |
1567 | } |
1568 | ||
1569 | void fib6_gc_cleanup(void) | |
1570 | { | |
450d19f8 DL |
1571 | del_timer(ip6_fib_timer); |
1572 | kfree(ip6_fib_timer); | |
58f09b78 | 1573 | unregister_pernet_subsys(&fib6_net_ops); |
1da177e4 LT |
1574 | kmem_cache_destroy(fib6_node_kmem); |
1575 | } |