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 | ||
e18b890b | 51 | static struct kmem_cache * fib6_node_kmem __read_mostly; |
1da177e4 LT |
52 | |
53 | enum fib_walk_state_t | |
54 | { | |
55 | #ifdef CONFIG_IPV6_SUBTREES | |
56 | FWS_S, | |
57 | #endif | |
58 | FWS_L, | |
59 | FWS_R, | |
60 | FWS_C, | |
61 | FWS_U | |
62 | }; | |
63 | ||
64 | struct fib6_cleaner_t | |
65 | { | |
66 | struct fib6_walker_t w; | |
ec7d43c2 | 67 | struct net *net; |
1da177e4 LT |
68 | int (*func)(struct rt6_info *, void *arg); |
69 | void *arg; | |
70 | }; | |
71 | ||
90d41122 | 72 | static DEFINE_RWLOCK(fib6_walker_lock); |
1da177e4 LT |
73 | |
74 | #ifdef CONFIG_IPV6_SUBTREES | |
75 | #define FWS_INIT FWS_S | |
1da177e4 LT |
76 | #else |
77 | #define FWS_INIT FWS_L | |
1da177e4 LT |
78 | #endif |
79 | ||
ec7d43c2 BT |
80 | static void fib6_prune_clones(struct net *net, struct fib6_node *fn, |
81 | struct rt6_info *rt); | |
8ed67789 DL |
82 | static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn); |
83 | static struct fib6_node *fib6_repair_tree(struct net *net, 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 | ||
90d41122 | 98 | static struct fib6_walker_t fib6_walker_list = { |
1da177e4 | 99 | .prev = &fib6_walker_list, |
1ab1457c | 100 | .next = &fib6_walker_list, |
1da177e4 LT |
101 | }; |
102 | ||
103 | #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next) | |
104 | ||
90d41122 AB |
105 | static inline void fib6_walker_link(struct fib6_walker_t *w) |
106 | { | |
107 | write_lock_bh(&fib6_walker_lock); | |
108 | w->next = fib6_walker_list.next; | |
109 | w->prev = &fib6_walker_list; | |
110 | w->next->prev = w; | |
111 | w->prev->next = w; | |
112 | write_unlock_bh(&fib6_walker_lock); | |
113 | } | |
114 | ||
115 | static inline void fib6_walker_unlink(struct fib6_walker_t *w) | |
116 | { | |
117 | write_lock_bh(&fib6_walker_lock); | |
118 | w->next->prev = w->prev; | |
119 | w->prev->next = w->next; | |
120 | w->prev = w->next = w; | |
121 | write_unlock_bh(&fib6_walker_lock); | |
122 | } | |
1da177e4 LT |
123 | static __inline__ u32 fib6_new_sernum(void) |
124 | { | |
125 | u32 n = ++rt_sernum; | |
126 | if ((__s32)n <= 0) | |
127 | rt_sernum = n = 1; | |
128 | return n; | |
129 | } | |
130 | ||
131 | /* | |
132 | * Auxiliary address test functions for the radix tree. | |
133 | * | |
1ab1457c | 134 | * These assume a 32bit processor (although it will work on |
1da177e4 LT |
135 | * 64bit processors) |
136 | */ | |
137 | ||
138 | /* | |
139 | * test bit | |
140 | */ | |
141 | ||
e69a4adc | 142 | static __inline__ __be32 addr_bit_set(void *token, int fn_bit) |
1da177e4 | 143 | { |
e69a4adc | 144 | __be32 *addr = token; |
1da177e4 LT |
145 | |
146 | return htonl(1 << ((~fn_bit)&0x1F)) & addr[fn_bit>>5]; | |
147 | } | |
148 | ||
1da177e4 LT |
149 | static __inline__ struct fib6_node * node_alloc(void) |
150 | { | |
151 | struct fib6_node *fn; | |
152 | ||
c3762229 | 153 | fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC); |
1da177e4 LT |
154 | |
155 | return fn; | |
156 | } | |
157 | ||
158 | static __inline__ void node_free(struct fib6_node * fn) | |
159 | { | |
160 | kmem_cache_free(fib6_node_kmem, fn); | |
161 | } | |
162 | ||
163 | static __inline__ void rt6_release(struct rt6_info *rt) | |
164 | { | |
165 | if (atomic_dec_and_test(&rt->rt6i_ref)) | |
166 | dst_free(&rt->u.dst); | |
167 | } | |
168 | ||
c71099ac | 169 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
1b43af54 PM |
170 | #define FIB_TABLE_HASHSZ 256 |
171 | #else | |
172 | #define FIB_TABLE_HASHSZ 1 | |
173 | #endif | |
1b43af54 | 174 | |
58f09b78 | 175 | static void fib6_link_table(struct net *net, struct fib6_table *tb) |
1b43af54 PM |
176 | { |
177 | unsigned int h; | |
178 | ||
375216ad TG |
179 | /* |
180 | * Initialize table lock at a single place to give lockdep a key, | |
181 | * tables aren't visible prior to being linked to the list. | |
182 | */ | |
183 | rwlock_init(&tb->tb6_lock); | |
184 | ||
1b43af54 PM |
185 | h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1); |
186 | ||
187 | /* | |
188 | * No protection necessary, this is the only list mutatation | |
189 | * operation, tables never disappear once they exist. | |
190 | */ | |
58f09b78 | 191 | hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]); |
1b43af54 | 192 | } |
c71099ac | 193 | |
1b43af54 | 194 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 195 | |
8ed67789 | 196 | static struct fib6_table *fib6_alloc_table(struct net *net, u32 id) |
c71099ac TG |
197 | { |
198 | struct fib6_table *table; | |
199 | ||
200 | table = kzalloc(sizeof(*table), GFP_ATOMIC); | |
201 | if (table != NULL) { | |
202 | table->tb6_id = id; | |
8ed67789 | 203 | table->tb6_root.leaf = net->ipv6.ip6_null_entry; |
c71099ac TG |
204 | table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; |
205 | } | |
206 | ||
207 | return table; | |
208 | } | |
209 | ||
58f09b78 | 210 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac TG |
211 | { |
212 | struct fib6_table *tb; | |
213 | ||
214 | if (id == 0) | |
215 | id = RT6_TABLE_MAIN; | |
58f09b78 | 216 | tb = fib6_get_table(net, id); |
c71099ac TG |
217 | if (tb) |
218 | return tb; | |
219 | ||
8ed67789 | 220 | tb = fib6_alloc_table(net, id); |
c71099ac | 221 | if (tb != NULL) |
58f09b78 | 222 | fib6_link_table(net, tb); |
c71099ac TG |
223 | |
224 | return tb; | |
225 | } | |
226 | ||
58f09b78 | 227 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac TG |
228 | { |
229 | struct fib6_table *tb; | |
58f09b78 | 230 | struct hlist_head *head; |
c71099ac TG |
231 | struct hlist_node *node; |
232 | unsigned int h; | |
233 | ||
234 | if (id == 0) | |
235 | id = RT6_TABLE_MAIN; | |
236 | h = id & (FIB_TABLE_HASHSZ - 1); | |
237 | rcu_read_lock(); | |
58f09b78 DL |
238 | head = &net->ipv6.fib_table_hash[h]; |
239 | hlist_for_each_entry_rcu(tb, node, head, tb6_hlist) { | |
c71099ac TG |
240 | if (tb->tb6_id == id) { |
241 | rcu_read_unlock(); | |
242 | return tb; | |
243 | } | |
244 | } | |
245 | rcu_read_unlock(); | |
246 | ||
247 | return NULL; | |
248 | } | |
249 | ||
58f09b78 | 250 | static void fib6_tables_init(struct net *net) |
c71099ac | 251 | { |
58f09b78 DL |
252 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
253 | fib6_link_table(net, net->ipv6.fib6_local_tbl); | |
c71099ac | 254 | } |
c71099ac TG |
255 | #else |
256 | ||
58f09b78 | 257 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac | 258 | { |
58f09b78 | 259 | return fib6_get_table(net, id); |
c71099ac TG |
260 | } |
261 | ||
58f09b78 | 262 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac | 263 | { |
58f09b78 | 264 | return net->ipv6.fib6_main_tbl; |
c71099ac TG |
265 | } |
266 | ||
58f09b78 DL |
267 | struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi *fl, |
268 | int flags, pol_lookup_t lookup) | |
c71099ac | 269 | { |
8ed67789 | 270 | return (struct dst_entry *) lookup(net, net->ipv6.fib6_main_tbl, fl, flags); |
c71099ac TG |
271 | } |
272 | ||
58f09b78 | 273 | static void fib6_tables_init(struct net *net) |
c71099ac | 274 | { |
58f09b78 | 275 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
c71099ac TG |
276 | } |
277 | ||
278 | #endif | |
279 | ||
1b43af54 PM |
280 | static int fib6_dump_node(struct fib6_walker_t *w) |
281 | { | |
282 | int res; | |
283 | struct rt6_info *rt; | |
284 | ||
7cc48263 | 285 | for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) { |
1b43af54 PM |
286 | res = rt6_dump_route(rt, w->args); |
287 | if (res < 0) { | |
288 | /* Frame is full, suspend walking */ | |
289 | w->leaf = rt; | |
290 | return 1; | |
291 | } | |
292 | BUG_TRAP(res!=0); | |
293 | } | |
294 | w->leaf = NULL; | |
295 | return 0; | |
296 | } | |
297 | ||
298 | static void fib6_dump_end(struct netlink_callback *cb) | |
299 | { | |
300 | struct fib6_walker_t *w = (void*)cb->args[2]; | |
301 | ||
302 | if (w) { | |
303 | cb->args[2] = 0; | |
304 | kfree(w); | |
305 | } | |
306 | cb->done = (void*)cb->args[3]; | |
307 | cb->args[1] = 3; | |
308 | } | |
309 | ||
310 | static int fib6_dump_done(struct netlink_callback *cb) | |
311 | { | |
312 | fib6_dump_end(cb); | |
313 | return cb->done ? cb->done(cb) : 0; | |
314 | } | |
315 | ||
316 | static 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) { | |
326 | read_lock_bh(&table->tb6_lock); | |
327 | res = fib6_walk(w); | |
328 | read_unlock_bh(&table->tb6_lock); | |
329 | if (res > 0) | |
330 | cb->args[4] = 1; | |
331 | } else { | |
332 | read_lock_bh(&table->tb6_lock); | |
333 | res = fib6_walk_continue(w); | |
334 | read_unlock_bh(&table->tb6_lock); | |
335 | if (res != 0) { | |
336 | if (res < 0) | |
337 | fib6_walker_unlink(w); | |
338 | goto end; | |
339 | } | |
340 | fib6_walker_unlink(w); | |
341 | cb->args[4] = 0; | |
342 | } | |
343 | end: | |
344 | return res; | |
345 | } | |
346 | ||
c127ea2c | 347 | static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) |
1b43af54 | 348 | { |
b854272b | 349 | struct net *net = skb->sk->sk_net; |
1b43af54 PM |
350 | unsigned int h, s_h; |
351 | unsigned int e = 0, s_e; | |
352 | struct rt6_rtnl_dump_arg arg; | |
353 | struct fib6_walker_t *w; | |
354 | struct fib6_table *tb; | |
355 | struct hlist_node *node; | |
58f09b78 | 356 | struct hlist_head *head; |
1b43af54 PM |
357 | int res = 0; |
358 | ||
359 | s_h = cb->args[0]; | |
360 | s_e = cb->args[1]; | |
361 | ||
362 | w = (void *)cb->args[2]; | |
363 | if (w == NULL) { | |
364 | /* New dump: | |
365 | * | |
366 | * 1. hook callback destructor. | |
367 | */ | |
368 | cb->args[3] = (long)cb->done; | |
369 | cb->done = fib6_dump_done; | |
370 | ||
371 | /* | |
372 | * 2. allocate and initialize walker. | |
373 | */ | |
374 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
375 | if (w == NULL) | |
376 | return -ENOMEM; | |
377 | w->func = fib6_dump_node; | |
378 | cb->args[2] = (long)w; | |
379 | } | |
380 | ||
381 | arg.skb = skb; | |
382 | arg.cb = cb; | |
383 | w->args = &arg; | |
384 | ||
385 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { | |
386 | e = 0; | |
58f09b78 DL |
387 | head = &net->ipv6.fib_table_hash[h]; |
388 | hlist_for_each_entry(tb, node, head, tb6_hlist) { | |
1b43af54 PM |
389 | if (e < s_e) |
390 | goto next; | |
391 | res = fib6_dump_table(tb, skb, cb); | |
392 | if (res != 0) | |
393 | goto out; | |
394 | next: | |
395 | e++; | |
396 | } | |
397 | } | |
398 | out: | |
399 | cb->args[1] = e; | |
400 | cb->args[0] = h; | |
401 | ||
402 | res = res < 0 ? res : skb->len; | |
403 | if (res <= 0) | |
404 | fib6_dump_end(cb); | |
405 | return res; | |
406 | } | |
1da177e4 LT |
407 | |
408 | /* | |
409 | * Routing Table | |
410 | * | |
411 | * return the appropriate node for a routing tree "add" operation | |
412 | * by either creating and inserting or by returning an existing | |
413 | * node. | |
414 | */ | |
415 | ||
416 | static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr, | |
417 | int addrlen, int plen, | |
418 | int offset) | |
419 | { | |
420 | struct fib6_node *fn, *in, *ln; | |
421 | struct fib6_node *pn = NULL; | |
422 | struct rt6key *key; | |
423 | int bit; | |
1ab1457c | 424 | __be32 dir = 0; |
1da177e4 LT |
425 | __u32 sernum = fib6_new_sernum(); |
426 | ||
427 | RT6_TRACE("fib6_add_1\n"); | |
428 | ||
429 | /* insert node in tree */ | |
430 | ||
431 | fn = root; | |
432 | ||
433 | do { | |
434 | key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
435 | ||
436 | /* | |
437 | * Prefix match | |
438 | */ | |
439 | if (plen < fn->fn_bit || | |
440 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
441 | goto insert_above; | |
1ab1457c | 442 | |
1da177e4 LT |
443 | /* |
444 | * Exact match ? | |
445 | */ | |
1ab1457c | 446 | |
1da177e4 LT |
447 | if (plen == fn->fn_bit) { |
448 | /* clean up an intermediate node */ | |
449 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
450 | rt6_release(fn->leaf); | |
451 | fn->leaf = NULL; | |
452 | } | |
1ab1457c | 453 | |
1da177e4 | 454 | fn->fn_sernum = sernum; |
1ab1457c | 455 | |
1da177e4 LT |
456 | return fn; |
457 | } | |
458 | ||
459 | /* | |
460 | * We have more bits to go | |
461 | */ | |
1ab1457c | 462 | |
1da177e4 LT |
463 | /* Try to walk down on tree. */ |
464 | fn->fn_sernum = sernum; | |
465 | dir = addr_bit_set(addr, fn->fn_bit); | |
466 | pn = fn; | |
467 | fn = dir ? fn->right: fn->left; | |
468 | } while (fn); | |
469 | ||
470 | /* | |
471 | * We walked to the bottom of tree. | |
472 | * Create new leaf node without children. | |
473 | */ | |
474 | ||
475 | ln = node_alloc(); | |
476 | ||
477 | if (ln == NULL) | |
478 | return NULL; | |
479 | ln->fn_bit = plen; | |
1ab1457c | 480 | |
1da177e4 LT |
481 | ln->parent = pn; |
482 | ln->fn_sernum = sernum; | |
483 | ||
484 | if (dir) | |
485 | pn->right = ln; | |
486 | else | |
487 | pn->left = ln; | |
488 | ||
489 | return ln; | |
490 | ||
491 | ||
492 | insert_above: | |
493 | /* | |
1ab1457c | 494 | * split since we don't have a common prefix anymore or |
1da177e4 LT |
495 | * we have a less significant route. |
496 | * we've to insert an intermediate node on the list | |
497 | * this new node will point to the one we need to create | |
498 | * and the current | |
499 | */ | |
500 | ||
501 | pn = fn->parent; | |
502 | ||
503 | /* find 1st bit in difference between the 2 addrs. | |
504 | ||
971f359d | 505 | See comment in __ipv6_addr_diff: bit may be an invalid value, |
1da177e4 LT |
506 | but if it is >= plen, the value is ignored in any case. |
507 | */ | |
1ab1457c | 508 | |
971f359d | 509 | bit = __ipv6_addr_diff(addr, &key->addr, addrlen); |
1da177e4 | 510 | |
1ab1457c YH |
511 | /* |
512 | * (intermediate)[in] | |
1da177e4 LT |
513 | * / \ |
514 | * (new leaf node)[ln] (old node)[fn] | |
515 | */ | |
516 | if (plen > bit) { | |
517 | in = node_alloc(); | |
518 | ln = node_alloc(); | |
1ab1457c | 519 | |
1da177e4 LT |
520 | if (in == NULL || ln == NULL) { |
521 | if (in) | |
522 | node_free(in); | |
523 | if (ln) | |
524 | node_free(ln); | |
525 | return NULL; | |
526 | } | |
527 | ||
1ab1457c YH |
528 | /* |
529 | * new intermediate node. | |
1da177e4 LT |
530 | * RTN_RTINFO will |
531 | * be off since that an address that chooses one of | |
532 | * the branches would not match less specific routes | |
533 | * in the other branch | |
534 | */ | |
535 | ||
536 | in->fn_bit = bit; | |
537 | ||
538 | in->parent = pn; | |
539 | in->leaf = fn->leaf; | |
540 | atomic_inc(&in->leaf->rt6i_ref); | |
541 | ||
542 | in->fn_sernum = sernum; | |
543 | ||
544 | /* update parent pointer */ | |
545 | if (dir) | |
546 | pn->right = in; | |
547 | else | |
548 | pn->left = in; | |
549 | ||
550 | ln->fn_bit = plen; | |
551 | ||
552 | ln->parent = in; | |
553 | fn->parent = in; | |
554 | ||
555 | ln->fn_sernum = sernum; | |
556 | ||
557 | if (addr_bit_set(addr, bit)) { | |
558 | in->right = ln; | |
559 | in->left = fn; | |
560 | } else { | |
561 | in->left = ln; | |
562 | in->right = fn; | |
563 | } | |
564 | } else { /* plen <= bit */ | |
565 | ||
1ab1457c | 566 | /* |
1da177e4 LT |
567 | * (new leaf node)[ln] |
568 | * / \ | |
569 | * (old node)[fn] NULL | |
570 | */ | |
571 | ||
572 | ln = node_alloc(); | |
573 | ||
574 | if (ln == NULL) | |
575 | return NULL; | |
576 | ||
577 | ln->fn_bit = plen; | |
578 | ||
579 | ln->parent = pn; | |
580 | ||
581 | ln->fn_sernum = sernum; | |
1ab1457c | 582 | |
1da177e4 LT |
583 | if (dir) |
584 | pn->right = ln; | |
585 | else | |
586 | pn->left = ln; | |
587 | ||
588 | if (addr_bit_set(&key->addr, plen)) | |
589 | ln->right = fn; | |
590 | else | |
591 | ln->left = fn; | |
592 | ||
593 | fn->parent = ln; | |
594 | } | |
595 | return ln; | |
596 | } | |
597 | ||
598 | /* | |
599 | * Insert routing information in a node. | |
600 | */ | |
601 | ||
602 | static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt, | |
86872cb5 | 603 | struct nl_info *info) |
1da177e4 LT |
604 | { |
605 | struct rt6_info *iter = NULL; | |
606 | struct rt6_info **ins; | |
607 | ||
608 | ins = &fn->leaf; | |
609 | ||
7cc48263 | 610 | for (iter = fn->leaf; iter; iter=iter->u.dst.rt6_next) { |
1da177e4 LT |
611 | /* |
612 | * Search for duplicates | |
613 | */ | |
614 | ||
615 | if (iter->rt6i_metric == rt->rt6i_metric) { | |
616 | /* | |
617 | * Same priority level | |
618 | */ | |
619 | ||
620 | if (iter->rt6i_dev == rt->rt6i_dev && | |
621 | iter->rt6i_idev == rt->rt6i_idev && | |
622 | ipv6_addr_equal(&iter->rt6i_gateway, | |
623 | &rt->rt6i_gateway)) { | |
624 | if (!(iter->rt6i_flags&RTF_EXPIRES)) | |
625 | return -EEXIST; | |
626 | iter->rt6i_expires = rt->rt6i_expires; | |
627 | if (!(rt->rt6i_flags&RTF_EXPIRES)) { | |
628 | iter->rt6i_flags &= ~RTF_EXPIRES; | |
629 | iter->rt6i_expires = 0; | |
630 | } | |
631 | return -EEXIST; | |
632 | } | |
633 | } | |
634 | ||
635 | if (iter->rt6i_metric > rt->rt6i_metric) | |
636 | break; | |
637 | ||
7cc48263 | 638 | ins = &iter->u.dst.rt6_next; |
1da177e4 LT |
639 | } |
640 | ||
f11e6659 DM |
641 | /* Reset round-robin state, if necessary */ |
642 | if (ins == &fn->leaf) | |
643 | fn->rr_ptr = NULL; | |
644 | ||
1da177e4 LT |
645 | /* |
646 | * insert node | |
647 | */ | |
648 | ||
7cc48263 | 649 | rt->u.dst.rt6_next = iter; |
1da177e4 LT |
650 | *ins = rt; |
651 | rt->rt6i_node = fn; | |
652 | atomic_inc(&rt->rt6i_ref); | |
86872cb5 | 653 | inet6_rt_notify(RTM_NEWROUTE, rt, info); |
c572872f | 654 | info->nl_net->ipv6.rt6_stats->fib_rt_entries++; |
1da177e4 LT |
655 | |
656 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
c572872f | 657 | info->nl_net->ipv6.rt6_stats->fib_route_nodes++; |
1da177e4 LT |
658 | fn->fn_flags |= RTN_RTINFO; |
659 | } | |
660 | ||
661 | return 0; | |
662 | } | |
663 | ||
63152fc0 | 664 | static __inline__ void fib6_start_gc(struct net *net, struct rt6_info *rt) |
1da177e4 | 665 | { |
63152fc0 | 666 | if (net->ipv6.ip6_fib_timer->expires == 0 && |
1da177e4 | 667 | (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE))) |
63152fc0 DL |
668 | mod_timer(net->ipv6.ip6_fib_timer, jiffies + |
669 | net->ipv6.sysctl.ip6_rt_gc_interval); | |
1da177e4 LT |
670 | } |
671 | ||
63152fc0 | 672 | void fib6_force_start_gc(struct net *net) |
1da177e4 | 673 | { |
63152fc0 DL |
674 | if (net->ipv6.ip6_fib_timer->expires == 0) |
675 | mod_timer(net->ipv6.ip6_fib_timer, jiffies + | |
676 | net->ipv6.sysctl.ip6_rt_gc_interval); | |
1da177e4 LT |
677 | } |
678 | ||
679 | /* | |
680 | * Add routing information to the routing tree. | |
681 | * <destination addr>/<source addr> | |
682 | * with source addr info in sub-trees | |
683 | */ | |
684 | ||
86872cb5 | 685 | int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info) |
1da177e4 | 686 | { |
66729e18 | 687 | struct fib6_node *fn, *pn = NULL; |
1da177e4 LT |
688 | int err = -ENOMEM; |
689 | ||
690 | fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr), | |
691 | rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst)); | |
692 | ||
693 | if (fn == NULL) | |
694 | goto out; | |
695 | ||
66729e18 YH |
696 | pn = fn; |
697 | ||
1da177e4 LT |
698 | #ifdef CONFIG_IPV6_SUBTREES |
699 | if (rt->rt6i_src.plen) { | |
700 | struct fib6_node *sn; | |
701 | ||
702 | if (fn->subtree == NULL) { | |
703 | struct fib6_node *sfn; | |
704 | ||
705 | /* | |
706 | * Create subtree. | |
707 | * | |
708 | * fn[main tree] | |
709 | * | | |
710 | * sfn[subtree root] | |
711 | * \ | |
712 | * sn[new leaf node] | |
713 | */ | |
714 | ||
715 | /* Create subtree root node */ | |
716 | sfn = node_alloc(); | |
717 | if (sfn == NULL) | |
718 | goto st_failure; | |
719 | ||
8ed67789 DL |
720 | sfn->leaf = info->nl_net->ipv6.ip6_null_entry; |
721 | atomic_inc(&info->nl_net->ipv6.ip6_null_entry->rt6i_ref); | |
1da177e4 LT |
722 | sfn->fn_flags = RTN_ROOT; |
723 | sfn->fn_sernum = fib6_new_sernum(); | |
724 | ||
725 | /* Now add the first leaf node to new subtree */ | |
726 | ||
727 | sn = fib6_add_1(sfn, &rt->rt6i_src.addr, | |
728 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
729 | offsetof(struct rt6_info, rt6i_src)); | |
730 | ||
731 | if (sn == NULL) { | |
732 | /* If it is failed, discard just allocated | |
733 | root, and then (in st_failure) stale node | |
734 | in main tree. | |
735 | */ | |
736 | node_free(sfn); | |
737 | goto st_failure; | |
738 | } | |
739 | ||
740 | /* Now link new subtree to main tree */ | |
741 | sfn->parent = fn; | |
742 | fn->subtree = sfn; | |
1da177e4 LT |
743 | } else { |
744 | sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr, | |
745 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
746 | offsetof(struct rt6_info, rt6i_src)); | |
747 | ||
748 | if (sn == NULL) | |
749 | goto st_failure; | |
750 | } | |
751 | ||
66729e18 YH |
752 | if (fn->leaf == NULL) { |
753 | fn->leaf = rt; | |
754 | atomic_inc(&rt->rt6i_ref); | |
755 | } | |
1da177e4 LT |
756 | fn = sn; |
757 | } | |
758 | #endif | |
759 | ||
86872cb5 | 760 | err = fib6_add_rt2node(fn, rt, info); |
1da177e4 LT |
761 | |
762 | if (err == 0) { | |
63152fc0 | 763 | fib6_start_gc(info->nl_net, rt); |
1da177e4 | 764 | if (!(rt->rt6i_flags&RTF_CACHE)) |
ec7d43c2 | 765 | fib6_prune_clones(info->nl_net, pn, rt); |
1da177e4 LT |
766 | } |
767 | ||
768 | out: | |
66729e18 YH |
769 | if (err) { |
770 | #ifdef CONFIG_IPV6_SUBTREES | |
771 | /* | |
772 | * If fib6_add_1 has cleared the old leaf pointer in the | |
773 | * super-tree leaf node we have to find a new one for it. | |
774 | */ | |
775 | if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) { | |
8ed67789 | 776 | pn->leaf = fib6_find_prefix(info->nl_net, pn); |
66729e18 YH |
777 | #if RT6_DEBUG >= 2 |
778 | if (!pn->leaf) { | |
779 | BUG_TRAP(pn->leaf != NULL); | |
8ed67789 | 780 | pn->leaf = info->nl_net->ipv6.ip6_null_entry; |
66729e18 YH |
781 | } |
782 | #endif | |
783 | atomic_inc(&pn->leaf->rt6i_ref); | |
784 | } | |
785 | #endif | |
1da177e4 | 786 | dst_free(&rt->u.dst); |
66729e18 | 787 | } |
1da177e4 LT |
788 | return err; |
789 | ||
790 | #ifdef CONFIG_IPV6_SUBTREES | |
791 | /* Subtree creation failed, probably main tree node | |
792 | is orphan. If it is, shoot it. | |
793 | */ | |
794 | st_failure: | |
795 | if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT))) | |
8ed67789 | 796 | fib6_repair_tree(info->nl_net, fn); |
1da177e4 LT |
797 | dst_free(&rt->u.dst); |
798 | return err; | |
799 | #endif | |
800 | } | |
801 | ||
802 | /* | |
803 | * Routing tree lookup | |
804 | * | |
805 | */ | |
806 | ||
807 | struct lookup_args { | |
808 | int offset; /* key offset on rt6_info */ | |
809 | struct in6_addr *addr; /* search key */ | |
810 | }; | |
811 | ||
812 | static struct fib6_node * fib6_lookup_1(struct fib6_node *root, | |
813 | struct lookup_args *args) | |
814 | { | |
815 | struct fib6_node *fn; | |
e69a4adc | 816 | __be32 dir; |
1da177e4 | 817 | |
825e288e YH |
818 | if (unlikely(args->offset == 0)) |
819 | return NULL; | |
820 | ||
1da177e4 LT |
821 | /* |
822 | * Descend on a tree | |
823 | */ | |
824 | ||
825 | fn = root; | |
826 | ||
827 | for (;;) { | |
828 | struct fib6_node *next; | |
829 | ||
830 | dir = addr_bit_set(args->addr, fn->fn_bit); | |
831 | ||
832 | next = dir ? fn->right : fn->left; | |
833 | ||
834 | if (next) { | |
835 | fn = next; | |
836 | continue; | |
837 | } | |
838 | ||
839 | break; | |
840 | } | |
841 | ||
3fc5e044 | 842 | while(fn) { |
7fc33165 | 843 | if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) { |
1da177e4 LT |
844 | struct rt6key *key; |
845 | ||
846 | key = (struct rt6key *) ((u8 *) fn->leaf + | |
847 | args->offset); | |
848 | ||
3fc5e044 YH |
849 | if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) { |
850 | #ifdef CONFIG_IPV6_SUBTREES | |
851 | if (fn->subtree) | |
852 | fn = fib6_lookup_1(fn->subtree, args + 1); | |
853 | #endif | |
854 | if (!fn || fn->fn_flags & RTN_RTINFO) | |
855 | return fn; | |
856 | } | |
1da177e4 LT |
857 | } |
858 | ||
3fc5e044 YH |
859 | if (fn->fn_flags & RTN_ROOT) |
860 | break; | |
861 | ||
1da177e4 LT |
862 | fn = fn->parent; |
863 | } | |
864 | ||
865 | return NULL; | |
866 | } | |
867 | ||
868 | struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr, | |
869 | struct in6_addr *saddr) | |
870 | { | |
1da177e4 | 871 | struct fib6_node *fn; |
825e288e YH |
872 | struct lookup_args args[] = { |
873 | { | |
874 | .offset = offsetof(struct rt6_info, rt6i_dst), | |
875 | .addr = daddr, | |
876 | }, | |
1da177e4 | 877 | #ifdef CONFIG_IPV6_SUBTREES |
825e288e YH |
878 | { |
879 | .offset = offsetof(struct rt6_info, rt6i_src), | |
880 | .addr = saddr, | |
881 | }, | |
1da177e4 | 882 | #endif |
825e288e YH |
883 | { |
884 | .offset = 0, /* sentinel */ | |
885 | } | |
886 | }; | |
1da177e4 | 887 | |
fefc2a6c | 888 | fn = fib6_lookup_1(root, daddr ? args : args + 1); |
1da177e4 LT |
889 | |
890 | if (fn == NULL || fn->fn_flags & RTN_TL_ROOT) | |
891 | fn = root; | |
892 | ||
893 | return fn; | |
894 | } | |
895 | ||
896 | /* | |
897 | * Get node with specified destination prefix (and source prefix, | |
898 | * if subtrees are used) | |
899 | */ | |
900 | ||
901 | ||
902 | static struct fib6_node * fib6_locate_1(struct fib6_node *root, | |
903 | struct in6_addr *addr, | |
904 | int plen, int offset) | |
905 | { | |
906 | struct fib6_node *fn; | |
907 | ||
908 | for (fn = root; fn ; ) { | |
909 | struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
910 | ||
911 | /* | |
912 | * Prefix match | |
913 | */ | |
914 | if (plen < fn->fn_bit || | |
915 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
916 | return NULL; | |
917 | ||
918 | if (plen == fn->fn_bit) | |
919 | return fn; | |
920 | ||
921 | /* | |
922 | * We have more bits to go | |
923 | */ | |
924 | if (addr_bit_set(addr, fn->fn_bit)) | |
925 | fn = fn->right; | |
926 | else | |
927 | fn = fn->left; | |
928 | } | |
929 | return NULL; | |
930 | } | |
931 | ||
932 | struct fib6_node * fib6_locate(struct fib6_node *root, | |
933 | struct in6_addr *daddr, int dst_len, | |
934 | struct in6_addr *saddr, int src_len) | |
935 | { | |
936 | struct fib6_node *fn; | |
937 | ||
938 | fn = fib6_locate_1(root, daddr, dst_len, | |
939 | offsetof(struct rt6_info, rt6i_dst)); | |
940 | ||
941 | #ifdef CONFIG_IPV6_SUBTREES | |
942 | if (src_len) { | |
943 | BUG_TRAP(saddr!=NULL); | |
3fc5e044 YH |
944 | if (fn && fn->subtree) |
945 | fn = fib6_locate_1(fn->subtree, saddr, src_len, | |
1da177e4 LT |
946 | offsetof(struct rt6_info, rt6i_src)); |
947 | } | |
948 | #endif | |
949 | ||
950 | if (fn && fn->fn_flags&RTN_RTINFO) | |
951 | return fn; | |
952 | ||
953 | return NULL; | |
954 | } | |
955 | ||
956 | ||
957 | /* | |
958 | * Deletion | |
959 | * | |
960 | */ | |
961 | ||
8ed67789 | 962 | static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn) |
1da177e4 LT |
963 | { |
964 | if (fn->fn_flags&RTN_ROOT) | |
8ed67789 | 965 | return net->ipv6.ip6_null_entry; |
1da177e4 LT |
966 | |
967 | while(fn) { | |
968 | if(fn->left) | |
969 | return fn->left->leaf; | |
970 | ||
971 | if(fn->right) | |
972 | return fn->right->leaf; | |
973 | ||
7fc33165 | 974 | fn = FIB6_SUBTREE(fn); |
1da177e4 LT |
975 | } |
976 | return NULL; | |
977 | } | |
978 | ||
979 | /* | |
980 | * Called to trim the tree of intermediate nodes when possible. "fn" | |
981 | * is the node we want to try and remove. | |
982 | */ | |
983 | ||
8ed67789 DL |
984 | static struct fib6_node *fib6_repair_tree(struct net *net, |
985 | struct fib6_node *fn) | |
1da177e4 LT |
986 | { |
987 | int children; | |
988 | int nstate; | |
989 | struct fib6_node *child, *pn; | |
990 | struct fib6_walker_t *w; | |
991 | int iter = 0; | |
992 | ||
993 | for (;;) { | |
994 | RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter); | |
995 | iter++; | |
996 | ||
997 | BUG_TRAP(!(fn->fn_flags&RTN_RTINFO)); | |
998 | BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT)); | |
999 | BUG_TRAP(fn->leaf==NULL); | |
1000 | ||
1001 | children = 0; | |
1002 | child = NULL; | |
1003 | if (fn->right) child = fn->right, children |= 1; | |
1004 | if (fn->left) child = fn->left, children |= 2; | |
1005 | ||
7fc33165 | 1006 | if (children == 3 || FIB6_SUBTREE(fn) |
1da177e4 LT |
1007 | #ifdef CONFIG_IPV6_SUBTREES |
1008 | /* Subtree root (i.e. fn) may have one child */ | |
1009 | || (children && fn->fn_flags&RTN_ROOT) | |
1010 | #endif | |
1011 | ) { | |
8ed67789 | 1012 | fn->leaf = fib6_find_prefix(net, fn); |
1da177e4 LT |
1013 | #if RT6_DEBUG >= 2 |
1014 | if (fn->leaf==NULL) { | |
1015 | BUG_TRAP(fn->leaf); | |
8ed67789 | 1016 | fn->leaf = net->ipv6.ip6_null_entry; |
1da177e4 LT |
1017 | } |
1018 | #endif | |
1019 | atomic_inc(&fn->leaf->rt6i_ref); | |
1020 | return fn->parent; | |
1021 | } | |
1022 | ||
1023 | pn = fn->parent; | |
1024 | #ifdef CONFIG_IPV6_SUBTREES | |
7fc33165 | 1025 | if (FIB6_SUBTREE(pn) == fn) { |
1da177e4 | 1026 | BUG_TRAP(fn->fn_flags&RTN_ROOT); |
7fc33165 | 1027 | FIB6_SUBTREE(pn) = NULL; |
1da177e4 LT |
1028 | nstate = FWS_L; |
1029 | } else { | |
1030 | BUG_TRAP(!(fn->fn_flags&RTN_ROOT)); | |
1031 | #endif | |
1032 | if (pn->right == fn) pn->right = child; | |
1033 | else if (pn->left == fn) pn->left = child; | |
1034 | #if RT6_DEBUG >= 2 | |
1035 | else BUG_TRAP(0); | |
1036 | #endif | |
1037 | if (child) | |
1038 | child->parent = pn; | |
1039 | nstate = FWS_R; | |
1040 | #ifdef CONFIG_IPV6_SUBTREES | |
1041 | } | |
1042 | #endif | |
1043 | ||
1044 | read_lock(&fib6_walker_lock); | |
1045 | FOR_WALKERS(w) { | |
1046 | if (child == NULL) { | |
1047 | if (w->root == fn) { | |
1048 | w->root = w->node = NULL; | |
1049 | RT6_TRACE("W %p adjusted by delroot 1\n", w); | |
1050 | } else if (w->node == fn) { | |
1051 | RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate); | |
1052 | w->node = pn; | |
1053 | w->state = nstate; | |
1054 | } | |
1055 | } else { | |
1056 | if (w->root == fn) { | |
1057 | w->root = child; | |
1058 | RT6_TRACE("W %p adjusted by delroot 2\n", w); | |
1059 | } | |
1060 | if (w->node == fn) { | |
1061 | w->node = child; | |
1062 | if (children&2) { | |
1063 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1064 | w->state = w->state>=FWS_R ? FWS_U : FWS_INIT; | |
1065 | } else { | |
1066 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1067 | w->state = w->state>=FWS_C ? FWS_U : FWS_INIT; | |
1068 | } | |
1069 | } | |
1070 | } | |
1071 | } | |
1072 | read_unlock(&fib6_walker_lock); | |
1073 | ||
1074 | node_free(fn); | |
7fc33165 | 1075 | if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn)) |
1da177e4 LT |
1076 | return pn; |
1077 | ||
1078 | rt6_release(pn->leaf); | |
1079 | pn->leaf = NULL; | |
1080 | fn = pn; | |
1081 | } | |
1082 | } | |
1083 | ||
1084 | static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp, | |
86872cb5 | 1085 | struct nl_info *info) |
1da177e4 LT |
1086 | { |
1087 | struct fib6_walker_t *w; | |
1088 | struct rt6_info *rt = *rtp; | |
c572872f | 1089 | struct net *net = info->nl_net; |
1da177e4 LT |
1090 | |
1091 | RT6_TRACE("fib6_del_route\n"); | |
1092 | ||
1093 | /* Unlink it */ | |
7cc48263 | 1094 | *rtp = rt->u.dst.rt6_next; |
1da177e4 | 1095 | rt->rt6i_node = NULL; |
c572872f BT |
1096 | net->ipv6.rt6_stats->fib_rt_entries--; |
1097 | net->ipv6.rt6_stats->fib_discarded_routes++; | |
1da177e4 | 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; | |
c572872f | 1120 | net->ipv6.rt6_stats->fib_route_nodes--; |
8ed67789 | 1121 | fn = fib6_repair_tree(net, fn); |
1da177e4 LT |
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) { | |
8ed67789 | 1133 | fn->leaf = fib6_find_prefix(net, fn); |
1da177e4 LT |
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 | 1148 | { |
8ed67789 | 1149 | struct net *net = info->nl_net; |
1da177e4 LT |
1150 | struct fib6_node *fn = rt->rt6i_node; |
1151 | struct rt6_info **rtp; | |
1152 | ||
1153 | #if RT6_DEBUG >= 2 | |
1154 | if (rt->u.dst.obsolete>0) { | |
1155 | BUG_TRAP(fn==NULL); | |
1156 | return -ENOENT; | |
1157 | } | |
1158 | #endif | |
8ed67789 | 1159 | if (fn == NULL || rt == net->ipv6.ip6_null_entry) |
1da177e4 LT |
1160 | return -ENOENT; |
1161 | ||
1162 | BUG_TRAP(fn->fn_flags&RTN_RTINFO); | |
1163 | ||
150730d5 YH |
1164 | if (!(rt->rt6i_flags&RTF_CACHE)) { |
1165 | struct fib6_node *pn = fn; | |
1166 | #ifdef CONFIG_IPV6_SUBTREES | |
1167 | /* clones of this route might be in another subtree */ | |
1168 | if (rt->rt6i_src.plen) { | |
1169 | while (!(pn->fn_flags&RTN_ROOT)) | |
1170 | pn = pn->parent; | |
1171 | pn = pn->parent; | |
1172 | } | |
1173 | #endif | |
ec7d43c2 | 1174 | fib6_prune_clones(info->nl_net, pn, rt); |
150730d5 | 1175 | } |
1da177e4 LT |
1176 | |
1177 | /* | |
1178 | * Walk the leaf entries looking for ourself | |
1179 | */ | |
1180 | ||
7cc48263 | 1181 | for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.dst.rt6_next) { |
1da177e4 | 1182 | if (*rtp == rt) { |
86872cb5 | 1183 | fib6_del_route(fn, rtp, info); |
1da177e4 LT |
1184 | return 0; |
1185 | } | |
1186 | } | |
1187 | return -ENOENT; | |
1188 | } | |
1189 | ||
1190 | /* | |
1191 | * Tree traversal function. | |
1192 | * | |
1193 | * Certainly, it is not interrupt safe. | |
1194 | * However, it is internally reenterable wrt itself and fib6_add/fib6_del. | |
1195 | * It means, that we can modify tree during walking | |
1196 | * and use this function for garbage collection, clone pruning, | |
1ab1457c | 1197 | * cleaning tree when a device goes down etc. etc. |
1da177e4 LT |
1198 | * |
1199 | * It guarantees that every node will be traversed, | |
1200 | * and that it will be traversed only once. | |
1201 | * | |
1202 | * Callback function w->func may return: | |
1203 | * 0 -> continue walking. | |
1204 | * positive value -> walking is suspended (used by tree dumps, | |
1205 | * and probably by gc, if it will be split to several slices) | |
1206 | * negative value -> terminate walking. | |
1207 | * | |
1208 | * The function itself returns: | |
1209 | * 0 -> walk is complete. | |
1210 | * >0 -> walk is incomplete (i.e. suspended) | |
1211 | * <0 -> walk is terminated by an error. | |
1212 | */ | |
1213 | ||
90d41122 | 1214 | static int fib6_walk_continue(struct fib6_walker_t *w) |
1da177e4 LT |
1215 | { |
1216 | struct fib6_node *fn, *pn; | |
1217 | ||
1218 | for (;;) { | |
1219 | fn = w->node; | |
1220 | if (fn == NULL) | |
1221 | return 0; | |
1222 | ||
1223 | if (w->prune && fn != w->root && | |
1224 | fn->fn_flags&RTN_RTINFO && w->state < FWS_C) { | |
1225 | w->state = FWS_C; | |
1226 | w->leaf = fn->leaf; | |
1227 | } | |
1228 | switch (w->state) { | |
1229 | #ifdef CONFIG_IPV6_SUBTREES | |
1230 | case FWS_S: | |
7fc33165 YH |
1231 | if (FIB6_SUBTREE(fn)) { |
1232 | w->node = FIB6_SUBTREE(fn); | |
1da177e4 LT |
1233 | continue; |
1234 | } | |
1235 | w->state = FWS_L; | |
1ab1457c | 1236 | #endif |
1da177e4 LT |
1237 | case FWS_L: |
1238 | if (fn->left) { | |
1239 | w->node = fn->left; | |
1240 | w->state = FWS_INIT; | |
1241 | continue; | |
1242 | } | |
1243 | w->state = FWS_R; | |
1244 | case FWS_R: | |
1245 | if (fn->right) { | |
1246 | w->node = fn->right; | |
1247 | w->state = FWS_INIT; | |
1248 | continue; | |
1249 | } | |
1250 | w->state = FWS_C; | |
1251 | w->leaf = fn->leaf; | |
1252 | case FWS_C: | |
1253 | if (w->leaf && fn->fn_flags&RTN_RTINFO) { | |
1254 | int err = w->func(w); | |
1255 | if (err) | |
1256 | return err; | |
1257 | continue; | |
1258 | } | |
1259 | w->state = FWS_U; | |
1260 | case FWS_U: | |
1261 | if (fn == w->root) | |
1262 | return 0; | |
1263 | pn = fn->parent; | |
1264 | w->node = pn; | |
1265 | #ifdef CONFIG_IPV6_SUBTREES | |
7fc33165 | 1266 | if (FIB6_SUBTREE(pn) == fn) { |
1da177e4 LT |
1267 | BUG_TRAP(fn->fn_flags&RTN_ROOT); |
1268 | w->state = FWS_L; | |
1269 | continue; | |
1270 | } | |
1271 | #endif | |
1272 | if (pn->left == fn) { | |
1273 | w->state = FWS_R; | |
1274 | continue; | |
1275 | } | |
1276 | if (pn->right == fn) { | |
1277 | w->state = FWS_C; | |
1278 | w->leaf = w->node->leaf; | |
1279 | continue; | |
1280 | } | |
1281 | #if RT6_DEBUG >= 2 | |
1282 | BUG_TRAP(0); | |
1283 | #endif | |
1284 | } | |
1285 | } | |
1286 | } | |
1287 | ||
90d41122 | 1288 | static int fib6_walk(struct fib6_walker_t *w) |
1da177e4 LT |
1289 | { |
1290 | int res; | |
1291 | ||
1292 | w->state = FWS_INIT; | |
1293 | w->node = w->root; | |
1294 | ||
1295 | fib6_walker_link(w); | |
1296 | res = fib6_walk_continue(w); | |
1297 | if (res <= 0) | |
1298 | fib6_walker_unlink(w); | |
1299 | return res; | |
1300 | } | |
1301 | ||
1302 | static int fib6_clean_node(struct fib6_walker_t *w) | |
1303 | { | |
1304 | int res; | |
1305 | struct rt6_info *rt; | |
0a8891a0 | 1306 | struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w); |
ec7d43c2 BT |
1307 | struct nl_info info = { |
1308 | .nl_net = c->net, | |
1309 | }; | |
1da177e4 | 1310 | |
7cc48263 | 1311 | for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) { |
1da177e4 LT |
1312 | res = c->func(rt, c->arg); |
1313 | if (res < 0) { | |
1314 | w->leaf = rt; | |
528c4ceb | 1315 | res = fib6_del(rt, &info); |
1da177e4 LT |
1316 | if (res) { |
1317 | #if RT6_DEBUG >= 2 | |
1318 | printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res); | |
1319 | #endif | |
1320 | continue; | |
1321 | } | |
1322 | return 0; | |
1323 | } | |
1324 | BUG_TRAP(res==0); | |
1325 | } | |
1326 | w->leaf = rt; | |
1327 | return 0; | |
1328 | } | |
1329 | ||
1330 | /* | |
1331 | * Convenient frontend to tree walker. | |
1ab1457c | 1332 | * |
1da177e4 LT |
1333 | * func is called on each route. |
1334 | * It may return -1 -> delete this route. | |
1335 | * 0 -> continue walking | |
1336 | * | |
1337 | * prune==1 -> only immediate children of node (certainly, | |
1338 | * ignoring pure split nodes) will be scanned. | |
1339 | */ | |
1340 | ||
ec7d43c2 | 1341 | static void fib6_clean_tree(struct net *net, struct fib6_node *root, |
8ce11e6a AB |
1342 | int (*func)(struct rt6_info *, void *arg), |
1343 | int prune, void *arg) | |
1da177e4 LT |
1344 | { |
1345 | struct fib6_cleaner_t c; | |
1346 | ||
1347 | c.w.root = root; | |
1348 | c.w.func = fib6_clean_node; | |
1349 | c.w.prune = prune; | |
1350 | c.func = func; | |
1351 | c.arg = arg; | |
ec7d43c2 | 1352 | c.net = net; |
1da177e4 LT |
1353 | |
1354 | fib6_walk(&c.w); | |
1355 | } | |
1356 | ||
f3db4851 | 1357 | void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg), |
c71099ac TG |
1358 | int prune, void *arg) |
1359 | { | |
c71099ac | 1360 | struct fib6_table *table; |
1b43af54 | 1361 | struct hlist_node *node; |
58f09b78 | 1362 | struct hlist_head *head; |
1b43af54 | 1363 | unsigned int h; |
c71099ac | 1364 | |
1b43af54 PM |
1365 | rcu_read_lock(); |
1366 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { | |
f3db4851 | 1367 | head = &net->ipv6.fib_table_hash[h]; |
58f09b78 | 1368 | hlist_for_each_entry_rcu(table, node, head, tb6_hlist) { |
c71099ac | 1369 | write_lock_bh(&table->tb6_lock); |
ec7d43c2 BT |
1370 | fib6_clean_tree(net, &table->tb6_root, |
1371 | func, prune, arg); | |
c71099ac TG |
1372 | write_unlock_bh(&table->tb6_lock); |
1373 | } | |
1374 | } | |
1b43af54 | 1375 | rcu_read_unlock(); |
c71099ac TG |
1376 | } |
1377 | ||
1da177e4 LT |
1378 | static int fib6_prune_clone(struct rt6_info *rt, void *arg) |
1379 | { | |
1380 | if (rt->rt6i_flags & RTF_CACHE) { | |
1381 | RT6_TRACE("pruning clone %p\n", rt); | |
1382 | return -1; | |
1383 | } | |
1384 | ||
1385 | return 0; | |
1386 | } | |
1387 | ||
ec7d43c2 BT |
1388 | static void fib6_prune_clones(struct net *net, struct fib6_node *fn, |
1389 | struct rt6_info *rt) | |
1da177e4 | 1390 | { |
ec7d43c2 | 1391 | fib6_clean_tree(net, fn, fib6_prune_clone, 1, rt); |
1da177e4 LT |
1392 | } |
1393 | ||
1394 | /* | |
1395 | * Garbage collection | |
1396 | */ | |
1397 | ||
1398 | static struct fib6_gc_args | |
1399 | { | |
1400 | int timeout; | |
1401 | int more; | |
1402 | } gc_args; | |
1403 | ||
1404 | static int fib6_age(struct rt6_info *rt, void *arg) | |
1405 | { | |
1406 | unsigned long now = jiffies; | |
1407 | ||
1408 | /* | |
1409 | * check addrconf expiration here. | |
1410 | * Routes are expired even if they are in use. | |
1411 | * | |
1412 | * Also age clones. Note, that clones are aged out | |
1413 | * only if they are not in use now. | |
1414 | */ | |
1415 | ||
1416 | if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) { | |
1417 | if (time_after(now, rt->rt6i_expires)) { | |
1418 | RT6_TRACE("expiring %p\n", rt); | |
1da177e4 LT |
1419 | return -1; |
1420 | } | |
1421 | gc_args.more++; | |
1422 | } else if (rt->rt6i_flags & RTF_CACHE) { | |
1423 | if (atomic_read(&rt->u.dst.__refcnt) == 0 && | |
1424 | time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) { | |
1425 | RT6_TRACE("aging clone %p\n", rt); | |
1426 | return -1; | |
1427 | } else if ((rt->rt6i_flags & RTF_GATEWAY) && | |
1428 | (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) { | |
1429 | RT6_TRACE("purging route %p via non-router but gateway\n", | |
1430 | rt); | |
1431 | return -1; | |
1432 | } | |
1433 | gc_args.more++; | |
1434 | } | |
1435 | ||
1436 | return 0; | |
1437 | } | |
1438 | ||
1439 | static DEFINE_SPINLOCK(fib6_gc_lock); | |
1440 | ||
5b7c931d | 1441 | void fib6_run_gc(unsigned long expires, struct net *net) |
1da177e4 | 1442 | { |
5b7c931d | 1443 | if (expires != ~0UL) { |
1da177e4 | 1444 | spin_lock_bh(&fib6_gc_lock); |
5b7c931d DL |
1445 | gc_args.timeout = expires ? (int)expires : |
1446 | net->ipv6.sysctl.ip6_rt_gc_interval; | |
1da177e4 LT |
1447 | } else { |
1448 | local_bh_disable(); | |
1449 | if (!spin_trylock(&fib6_gc_lock)) { | |
63152fc0 | 1450 | mod_timer(net->ipv6.ip6_fib_timer, jiffies + HZ); |
1da177e4 LT |
1451 | local_bh_enable(); |
1452 | return; | |
1453 | } | |
5b7c931d | 1454 | gc_args.timeout = net->ipv6.sysctl.ip6_rt_gc_interval; |
1da177e4 LT |
1455 | } |
1456 | gc_args.more = 0; | |
1457 | ||
3b00944c | 1458 | icmp6_dst_gc(&gc_args.more); |
f3db4851 | 1459 | |
5b7c931d | 1460 | fib6_clean_all(net, fib6_age, 0, NULL); |
1da177e4 LT |
1461 | |
1462 | if (gc_args.more) | |
63152fc0 | 1463 | mod_timer(net->ipv6.ip6_fib_timer, jiffies + |
5b7c931d | 1464 | net->ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 | 1465 | else { |
63152fc0 DL |
1466 | del_timer(net->ipv6.ip6_fib_timer); |
1467 | net->ipv6.ip6_fib_timer->expires = 0; | |
1da177e4 LT |
1468 | } |
1469 | spin_unlock_bh(&fib6_gc_lock); | |
1470 | } | |
1471 | ||
5b7c931d DL |
1472 | static void fib6_gc_timer_cb(unsigned long arg) |
1473 | { | |
1474 | fib6_run_gc(0, (struct net *)arg); | |
1475 | } | |
1476 | ||
58f09b78 | 1477 | static int fib6_net_init(struct net *net) |
1da177e4 | 1478 | { |
58f09b78 | 1479 | int ret; |
63152fc0 | 1480 | struct timer_list *timer; |
e0b85590 | 1481 | |
58f09b78 | 1482 | ret = -ENOMEM; |
63152fc0 DL |
1483 | timer = kzalloc(sizeof(*timer), GFP_KERNEL); |
1484 | if (!timer) | |
1485 | goto out; | |
1486 | ||
1487 | setup_timer(timer, fib6_gc_timer_cb, (unsigned long)net); | |
1488 | net->ipv6.ip6_fib_timer = timer; | |
1489 | ||
c572872f BT |
1490 | net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL); |
1491 | if (!net->ipv6.rt6_stats) | |
1492 | goto out_timer; | |
1493 | ||
58f09b78 DL |
1494 | net->ipv6.fib_table_hash = |
1495 | kzalloc(sizeof(*net->ipv6.fib_table_hash)*FIB_TABLE_HASHSZ, | |
1496 | GFP_KERNEL); | |
1497 | if (!net->ipv6.fib_table_hash) | |
c572872f | 1498 | goto out_rt6_stats; |
e0b85590 | 1499 | |
58f09b78 DL |
1500 | net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl), |
1501 | GFP_KERNEL); | |
1502 | if (!net->ipv6.fib6_main_tbl) | |
e0b85590 DL |
1503 | goto out_fib_table_hash; |
1504 | ||
58f09b78 | 1505 | net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN; |
8ed67789 | 1506 | net->ipv6.fib6_main_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry; |
58f09b78 DL |
1507 | net->ipv6.fib6_main_tbl->tb6_root.fn_flags = |
1508 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
e0b85590 DL |
1509 | |
1510 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES | |
58f09b78 DL |
1511 | net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl), |
1512 | GFP_KERNEL); | |
1513 | if (!net->ipv6.fib6_local_tbl) | |
e0b85590 | 1514 | goto out_fib6_main_tbl; |
58f09b78 | 1515 | net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL; |
8ed67789 | 1516 | net->ipv6.fib6_local_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry; |
58f09b78 DL |
1517 | net->ipv6.fib6_local_tbl->tb6_root.fn_flags = |
1518 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
e0b85590 | 1519 | #endif |
58f09b78 | 1520 | fib6_tables_init(net); |
f845ab6b | 1521 | |
58f09b78 | 1522 | ret = 0; |
d63bddbe DL |
1523 | out: |
1524 | return ret; | |
1525 | ||
e0b85590 | 1526 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 1527 | out_fib6_main_tbl: |
58f09b78 | 1528 | kfree(net->ipv6.fib6_main_tbl); |
e0b85590 | 1529 | #endif |
e0b85590 | 1530 | out_fib_table_hash: |
58f09b78 | 1531 | kfree(net->ipv6.fib_table_hash); |
c572872f BT |
1532 | out_rt6_stats: |
1533 | kfree(net->ipv6.rt6_stats); | |
63152fc0 DL |
1534 | out_timer: |
1535 | kfree(timer); | |
58f09b78 DL |
1536 | goto out; |
1537 | } | |
1538 | ||
1539 | static void fib6_net_exit(struct net *net) | |
1540 | { | |
8ed67789 | 1541 | rt6_ifdown(net, NULL); |
63152fc0 DL |
1542 | del_timer(net->ipv6.ip6_fib_timer); |
1543 | kfree(net->ipv6.ip6_fib_timer); | |
58f09b78 DL |
1544 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
1545 | kfree(net->ipv6.fib6_local_tbl); | |
1546 | #endif | |
1547 | kfree(net->ipv6.fib6_main_tbl); | |
1548 | kfree(net->ipv6.fib_table_hash); | |
c572872f | 1549 | kfree(net->ipv6.rt6_stats); |
58f09b78 DL |
1550 | } |
1551 | ||
1552 | static struct pernet_operations fib6_net_ops = { | |
1553 | .init = fib6_net_init, | |
1554 | .exit = fib6_net_exit, | |
1555 | }; | |
1556 | ||
1557 | int __init fib6_init(void) | |
1558 | { | |
1559 | int ret = -ENOMEM; | |
63152fc0 | 1560 | |
58f09b78 DL |
1561 | fib6_node_kmem = kmem_cache_create("fib6_nodes", |
1562 | sizeof(struct fib6_node), | |
1563 | 0, SLAB_HWCACHE_ALIGN, | |
1564 | NULL); | |
1565 | if (!fib6_node_kmem) | |
1566 | goto out; | |
1567 | ||
1568 | ret = register_pernet_subsys(&fib6_net_ops); | |
1569 | if (ret) | |
c572872f | 1570 | goto out_kmem_cache_create; |
58f09b78 DL |
1571 | |
1572 | ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib); | |
1573 | if (ret) | |
1574 | goto out_unregister_subsys; | |
1575 | out: | |
1576 | return ret; | |
1577 | ||
1578 | out_unregister_subsys: | |
1579 | unregister_pernet_subsys(&fib6_net_ops); | |
d63bddbe DL |
1580 | out_kmem_cache_create: |
1581 | kmem_cache_destroy(fib6_node_kmem); | |
1582 | goto out; | |
1da177e4 LT |
1583 | } |
1584 | ||
1585 | void fib6_gc_cleanup(void) | |
1586 | { | |
58f09b78 | 1587 | unregister_pernet_subsys(&fib6_net_ops); |
1da177e4 LT |
1588 | kmem_cache_destroy(fib6_node_kmem); |
1589 | } |