Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Linux INET6 implementation | |
3 | * Forwarding Information Database | |
4 | * | |
5 | * Authors: | |
6 | * Pedro Roque <roque@di.fc.ul.pt> | |
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. | |
21 | */ | |
1da177e4 LT |
22 | #include <linux/errno.h> |
23 | #include <linux/types.h> | |
24 | #include <linux/net.h> | |
25 | #include <linux/route.h> | |
26 | #include <linux/netdevice.h> | |
27 | #include <linux/in6.h> | |
28 | #include <linux/init.h> | |
c71099ac | 29 | #include <linux/list.h> |
1da177e4 LT |
30 | |
31 | #ifdef CONFIG_PROC_FS | |
32 | #include <linux/proc_fs.h> | |
33 | #endif | |
34 | ||
35 | #include <net/ipv6.h> | |
36 | #include <net/ndisc.h> | |
37 | #include <net/addrconf.h> | |
38 | ||
39 | #include <net/ip6_fib.h> | |
40 | #include <net/ip6_route.h> | |
41 | ||
42 | #define RT6_DEBUG 2 | |
43 | ||
44 | #if RT6_DEBUG >= 3 | |
45 | #define RT6_TRACE(x...) printk(KERN_DEBUG x) | |
46 | #else | |
47 | #define RT6_TRACE(x...) do { ; } while (0) | |
48 | #endif | |
49 | ||
50 | struct rt6_statistics rt6_stats; | |
51 | ||
ba89966c | 52 | static kmem_cache_t * fib6_node_kmem __read_mostly; |
1da177e4 LT |
53 | |
54 | enum fib_walk_state_t | |
55 | { | |
56 | #ifdef CONFIG_IPV6_SUBTREES | |
57 | FWS_S, | |
58 | #endif | |
59 | FWS_L, | |
60 | FWS_R, | |
61 | FWS_C, | |
62 | FWS_U | |
63 | }; | |
64 | ||
65 | struct fib6_cleaner_t | |
66 | { | |
67 | struct fib6_walker_t w; | |
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 | |
76 | #define SUBTREE(fn) ((fn)->subtree) | |
77 | #else | |
78 | #define FWS_INIT FWS_L | |
79 | #define SUBTREE(fn) NULL | |
80 | #endif | |
81 | ||
82 | static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt); | |
66729e18 | 83 | static struct rt6_info * fib6_find_prefix(struct fib6_node *fn); |
1da177e4 | 84 | static struct fib6_node * fib6_repair_tree(struct fib6_node *fn); |
90d41122 AB |
85 | static int fib6_walk(struct fib6_walker_t *w); |
86 | static int fib6_walk_continue(struct fib6_walker_t *w); | |
1da177e4 LT |
87 | |
88 | /* | |
89 | * A routing update causes an increase of the serial number on the | |
90 | * affected subtree. This allows for cached routes to be asynchronously | |
91 | * tested when modifications are made to the destination cache as a | |
92 | * result of redirects, path MTU changes, etc. | |
93 | */ | |
94 | ||
95 | static __u32 rt_sernum; | |
96 | ||
8d06afab | 97 | static DEFINE_TIMER(ip6_fib_timer, fib6_run_gc, 0, 0); |
1da177e4 | 98 | |
90d41122 | 99 | static struct fib6_walker_t fib6_walker_list = { |
1da177e4 LT |
100 | .prev = &fib6_walker_list, |
101 | .next = &fib6_walker_list, | |
102 | }; | |
103 | ||
104 | #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next) | |
105 | ||
90d41122 AB |
106 | static inline void fib6_walker_link(struct fib6_walker_t *w) |
107 | { | |
108 | write_lock_bh(&fib6_walker_lock); | |
109 | w->next = fib6_walker_list.next; | |
110 | w->prev = &fib6_walker_list; | |
111 | w->next->prev = w; | |
112 | w->prev->next = w; | |
113 | write_unlock_bh(&fib6_walker_lock); | |
114 | } | |
115 | ||
116 | static inline void fib6_walker_unlink(struct fib6_walker_t *w) | |
117 | { | |
118 | write_lock_bh(&fib6_walker_lock); | |
119 | w->next->prev = w->prev; | |
120 | w->prev->next = w->next; | |
121 | w->prev = w->next = w; | |
122 | write_unlock_bh(&fib6_walker_lock); | |
123 | } | |
1da177e4 LT |
124 | static __inline__ u32 fib6_new_sernum(void) |
125 | { | |
126 | u32 n = ++rt_sernum; | |
127 | if ((__s32)n <= 0) | |
128 | rt_sernum = n = 1; | |
129 | return n; | |
130 | } | |
131 | ||
132 | /* | |
133 | * Auxiliary address test functions for the radix tree. | |
134 | * | |
135 | * These assume a 32bit processor (although it will work on | |
136 | * 64bit processors) | |
137 | */ | |
138 | ||
139 | /* | |
140 | * test bit | |
141 | */ | |
142 | ||
143 | static __inline__ int addr_bit_set(void *token, int fn_bit) | |
144 | { | |
145 | __u32 *addr = token; | |
146 | ||
147 | return htonl(1 << ((~fn_bit)&0x1F)) & addr[fn_bit>>5]; | |
148 | } | |
149 | ||
1da177e4 LT |
150 | static __inline__ struct fib6_node * node_alloc(void) |
151 | { | |
152 | struct fib6_node *fn; | |
153 | ||
154 | if ((fn = kmem_cache_alloc(fib6_node_kmem, SLAB_ATOMIC)) != NULL) | |
155 | memset(fn, 0, sizeof(struct fib6_node)); | |
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 TG |
171 | static struct fib6_table fib6_main_tbl = { |
172 | .tb6_id = RT6_TABLE_MAIN, | |
173 | .tb6_lock = RW_LOCK_UNLOCKED, | |
174 | .tb6_root = { | |
175 | .leaf = &ip6_null_entry, | |
176 | .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO, | |
177 | }, | |
178 | }; | |
179 | ||
180 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES | |
1b43af54 PM |
181 | #define FIB_TABLE_HASHSZ 256 |
182 | #else | |
183 | #define FIB_TABLE_HASHSZ 1 | |
184 | #endif | |
185 | static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ]; | |
186 | ||
187 | static void fib6_link_table(struct fib6_table *tb) | |
188 | { | |
189 | unsigned int h; | |
190 | ||
191 | h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1); | |
192 | ||
193 | /* | |
194 | * No protection necessary, this is the only list mutatation | |
195 | * operation, tables never disappear once they exist. | |
196 | */ | |
197 | hlist_add_head_rcu(&tb->tb6_hlist, &fib_table_hash[h]); | |
198 | } | |
c71099ac | 199 | |
1b43af54 | 200 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
101367c2 TG |
201 | static struct fib6_table fib6_local_tbl = { |
202 | .tb6_id = RT6_TABLE_LOCAL, | |
203 | .tb6_lock = RW_LOCK_UNLOCKED, | |
204 | .tb6_root = { | |
205 | .leaf = &ip6_null_entry, | |
206 | .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO, | |
207 | }, | |
208 | }; | |
209 | ||
c71099ac TG |
210 | static struct fib6_table *fib6_alloc_table(u32 id) |
211 | { | |
212 | struct fib6_table *table; | |
213 | ||
214 | table = kzalloc(sizeof(*table), GFP_ATOMIC); | |
215 | if (table != NULL) { | |
216 | table->tb6_id = id; | |
217 | table->tb6_lock = RW_LOCK_UNLOCKED; | |
218 | table->tb6_root.leaf = &ip6_null_entry; | |
219 | table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
220 | } | |
221 | ||
222 | return table; | |
223 | } | |
224 | ||
c71099ac TG |
225 | struct fib6_table *fib6_new_table(u32 id) |
226 | { | |
227 | struct fib6_table *tb; | |
228 | ||
229 | if (id == 0) | |
230 | id = RT6_TABLE_MAIN; | |
231 | tb = fib6_get_table(id); | |
232 | if (tb) | |
233 | return tb; | |
234 | ||
235 | tb = fib6_alloc_table(id); | |
236 | if (tb != NULL) | |
237 | fib6_link_table(tb); | |
238 | ||
239 | return tb; | |
240 | } | |
241 | ||
242 | struct fib6_table *fib6_get_table(u32 id) | |
243 | { | |
244 | struct fib6_table *tb; | |
245 | struct hlist_node *node; | |
246 | unsigned int h; | |
247 | ||
248 | if (id == 0) | |
249 | id = RT6_TABLE_MAIN; | |
250 | h = id & (FIB_TABLE_HASHSZ - 1); | |
251 | rcu_read_lock(); | |
252 | hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb6_hlist) { | |
253 | if (tb->tb6_id == id) { | |
254 | rcu_read_unlock(); | |
255 | return tb; | |
256 | } | |
257 | } | |
258 | rcu_read_unlock(); | |
259 | ||
260 | return NULL; | |
261 | } | |
262 | ||
c71099ac TG |
263 | static void __init fib6_tables_init(void) |
264 | { | |
265 | fib6_link_table(&fib6_main_tbl); | |
101367c2 | 266 | fib6_link_table(&fib6_local_tbl); |
c71099ac TG |
267 | } |
268 | ||
269 | #else | |
270 | ||
271 | struct fib6_table *fib6_new_table(u32 id) | |
272 | { | |
273 | return fib6_get_table(id); | |
274 | } | |
275 | ||
276 | struct fib6_table *fib6_get_table(u32 id) | |
277 | { | |
278 | return &fib6_main_tbl; | |
279 | } | |
280 | ||
281 | struct dst_entry *fib6_rule_lookup(struct flowi *fl, int flags, | |
282 | pol_lookup_t lookup) | |
283 | { | |
284 | return (struct dst_entry *) lookup(&fib6_main_tbl, fl, flags); | |
285 | } | |
286 | ||
287 | static void __init fib6_tables_init(void) | |
288 | { | |
1b43af54 | 289 | fib6_link_table(&fib6_main_tbl); |
c71099ac TG |
290 | } |
291 | ||
292 | #endif | |
293 | ||
1b43af54 PM |
294 | static int fib6_dump_node(struct fib6_walker_t *w) |
295 | { | |
296 | int res; | |
297 | struct rt6_info *rt; | |
298 | ||
299 | for (rt = w->leaf; rt; rt = rt->u.next) { | |
300 | res = rt6_dump_route(rt, w->args); | |
301 | if (res < 0) { | |
302 | /* Frame is full, suspend walking */ | |
303 | w->leaf = rt; | |
304 | return 1; | |
305 | } | |
306 | BUG_TRAP(res!=0); | |
307 | } | |
308 | w->leaf = NULL; | |
309 | return 0; | |
310 | } | |
311 | ||
312 | static void fib6_dump_end(struct netlink_callback *cb) | |
313 | { | |
314 | struct fib6_walker_t *w = (void*)cb->args[2]; | |
315 | ||
316 | if (w) { | |
317 | cb->args[2] = 0; | |
318 | kfree(w); | |
319 | } | |
320 | cb->done = (void*)cb->args[3]; | |
321 | cb->args[1] = 3; | |
322 | } | |
323 | ||
324 | static int fib6_dump_done(struct netlink_callback *cb) | |
325 | { | |
326 | fib6_dump_end(cb); | |
327 | return cb->done ? cb->done(cb) : 0; | |
328 | } | |
329 | ||
330 | static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb, | |
331 | struct netlink_callback *cb) | |
332 | { | |
333 | struct fib6_walker_t *w; | |
334 | int res; | |
335 | ||
336 | w = (void *)cb->args[2]; | |
337 | w->root = &table->tb6_root; | |
338 | ||
339 | if (cb->args[4] == 0) { | |
340 | read_lock_bh(&table->tb6_lock); | |
341 | res = fib6_walk(w); | |
342 | read_unlock_bh(&table->tb6_lock); | |
343 | if (res > 0) | |
344 | cb->args[4] = 1; | |
345 | } else { | |
346 | read_lock_bh(&table->tb6_lock); | |
347 | res = fib6_walk_continue(w); | |
348 | read_unlock_bh(&table->tb6_lock); | |
349 | if (res != 0) { | |
350 | if (res < 0) | |
351 | fib6_walker_unlink(w); | |
352 | goto end; | |
353 | } | |
354 | fib6_walker_unlink(w); | |
355 | cb->args[4] = 0; | |
356 | } | |
357 | end: | |
358 | return res; | |
359 | } | |
360 | ||
361 | int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) | |
362 | { | |
363 | unsigned int h, s_h; | |
364 | unsigned int e = 0, s_e; | |
365 | struct rt6_rtnl_dump_arg arg; | |
366 | struct fib6_walker_t *w; | |
367 | struct fib6_table *tb; | |
368 | struct hlist_node *node; | |
369 | int res = 0; | |
370 | ||
371 | s_h = cb->args[0]; | |
372 | s_e = cb->args[1]; | |
373 | ||
374 | w = (void *)cb->args[2]; | |
375 | if (w == NULL) { | |
376 | /* New dump: | |
377 | * | |
378 | * 1. hook callback destructor. | |
379 | */ | |
380 | cb->args[3] = (long)cb->done; | |
381 | cb->done = fib6_dump_done; | |
382 | ||
383 | /* | |
384 | * 2. allocate and initialize walker. | |
385 | */ | |
386 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
387 | if (w == NULL) | |
388 | return -ENOMEM; | |
389 | w->func = fib6_dump_node; | |
390 | cb->args[2] = (long)w; | |
391 | } | |
392 | ||
393 | arg.skb = skb; | |
394 | arg.cb = cb; | |
395 | w->args = &arg; | |
396 | ||
397 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { | |
398 | e = 0; | |
399 | hlist_for_each_entry(tb, node, &fib_table_hash[h], tb6_hlist) { | |
400 | if (e < s_e) | |
401 | goto next; | |
402 | res = fib6_dump_table(tb, skb, cb); | |
403 | if (res != 0) | |
404 | goto out; | |
405 | next: | |
406 | e++; | |
407 | } | |
408 | } | |
409 | out: | |
410 | cb->args[1] = e; | |
411 | cb->args[0] = h; | |
412 | ||
413 | res = res < 0 ? res : skb->len; | |
414 | if (res <= 0) | |
415 | fib6_dump_end(cb); | |
416 | return res; | |
417 | } | |
1da177e4 LT |
418 | |
419 | /* | |
420 | * Routing Table | |
421 | * | |
422 | * return the appropriate node for a routing tree "add" operation | |
423 | * by either creating and inserting or by returning an existing | |
424 | * node. | |
425 | */ | |
426 | ||
427 | static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr, | |
428 | int addrlen, int plen, | |
429 | int offset) | |
430 | { | |
431 | struct fib6_node *fn, *in, *ln; | |
432 | struct fib6_node *pn = NULL; | |
433 | struct rt6key *key; | |
434 | int bit; | |
435 | int dir = 0; | |
436 | __u32 sernum = fib6_new_sernum(); | |
437 | ||
438 | RT6_TRACE("fib6_add_1\n"); | |
439 | ||
440 | /* insert node in tree */ | |
441 | ||
442 | fn = root; | |
443 | ||
444 | do { | |
445 | key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
446 | ||
447 | /* | |
448 | * Prefix match | |
449 | */ | |
450 | if (plen < fn->fn_bit || | |
451 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
452 | goto insert_above; | |
453 | ||
454 | /* | |
455 | * Exact match ? | |
456 | */ | |
457 | ||
458 | if (plen == fn->fn_bit) { | |
459 | /* clean up an intermediate node */ | |
460 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
461 | rt6_release(fn->leaf); | |
462 | fn->leaf = NULL; | |
463 | } | |
464 | ||
465 | fn->fn_sernum = sernum; | |
466 | ||
467 | return fn; | |
468 | } | |
469 | ||
470 | /* | |
471 | * We have more bits to go | |
472 | */ | |
473 | ||
474 | /* Try to walk down on tree. */ | |
475 | fn->fn_sernum = sernum; | |
476 | dir = addr_bit_set(addr, fn->fn_bit); | |
477 | pn = fn; | |
478 | fn = dir ? fn->right: fn->left; | |
479 | } while (fn); | |
480 | ||
481 | /* | |
482 | * We walked to the bottom of tree. | |
483 | * Create new leaf node without children. | |
484 | */ | |
485 | ||
486 | ln = node_alloc(); | |
487 | ||
488 | if (ln == NULL) | |
489 | return NULL; | |
490 | ln->fn_bit = plen; | |
491 | ||
492 | ln->parent = pn; | |
493 | ln->fn_sernum = sernum; | |
494 | ||
495 | if (dir) | |
496 | pn->right = ln; | |
497 | else | |
498 | pn->left = ln; | |
499 | ||
500 | return ln; | |
501 | ||
502 | ||
503 | insert_above: | |
504 | /* | |
505 | * split since we don't have a common prefix anymore or | |
506 | * we have a less significant route. | |
507 | * we've to insert an intermediate node on the list | |
508 | * this new node will point to the one we need to create | |
509 | * and the current | |
510 | */ | |
511 | ||
512 | pn = fn->parent; | |
513 | ||
514 | /* find 1st bit in difference between the 2 addrs. | |
515 | ||
971f359d | 516 | See comment in __ipv6_addr_diff: bit may be an invalid value, |
1da177e4 LT |
517 | but if it is >= plen, the value is ignored in any case. |
518 | */ | |
519 | ||
971f359d | 520 | bit = __ipv6_addr_diff(addr, &key->addr, addrlen); |
1da177e4 LT |
521 | |
522 | /* | |
523 | * (intermediate)[in] | |
524 | * / \ | |
525 | * (new leaf node)[ln] (old node)[fn] | |
526 | */ | |
527 | if (plen > bit) { | |
528 | in = node_alloc(); | |
529 | ln = node_alloc(); | |
530 | ||
531 | if (in == NULL || ln == NULL) { | |
532 | if (in) | |
533 | node_free(in); | |
534 | if (ln) | |
535 | node_free(ln); | |
536 | return NULL; | |
537 | } | |
538 | ||
539 | /* | |
540 | * new intermediate node. | |
541 | * RTN_RTINFO will | |
542 | * be off since that an address that chooses one of | |
543 | * the branches would not match less specific routes | |
544 | * in the other branch | |
545 | */ | |
546 | ||
547 | in->fn_bit = bit; | |
548 | ||
549 | in->parent = pn; | |
550 | in->leaf = fn->leaf; | |
551 | atomic_inc(&in->leaf->rt6i_ref); | |
552 | ||
553 | in->fn_sernum = sernum; | |
554 | ||
555 | /* update parent pointer */ | |
556 | if (dir) | |
557 | pn->right = in; | |
558 | else | |
559 | pn->left = in; | |
560 | ||
561 | ln->fn_bit = plen; | |
562 | ||
563 | ln->parent = in; | |
564 | fn->parent = in; | |
565 | ||
566 | ln->fn_sernum = sernum; | |
567 | ||
568 | if (addr_bit_set(addr, bit)) { | |
569 | in->right = ln; | |
570 | in->left = fn; | |
571 | } else { | |
572 | in->left = ln; | |
573 | in->right = fn; | |
574 | } | |
575 | } else { /* plen <= bit */ | |
576 | ||
577 | /* | |
578 | * (new leaf node)[ln] | |
579 | * / \ | |
580 | * (old node)[fn] NULL | |
581 | */ | |
582 | ||
583 | ln = node_alloc(); | |
584 | ||
585 | if (ln == NULL) | |
586 | return NULL; | |
587 | ||
588 | ln->fn_bit = plen; | |
589 | ||
590 | ln->parent = pn; | |
591 | ||
592 | ln->fn_sernum = sernum; | |
593 | ||
594 | if (dir) | |
595 | pn->right = ln; | |
596 | else | |
597 | pn->left = ln; | |
598 | ||
599 | if (addr_bit_set(&key->addr, plen)) | |
600 | ln->right = fn; | |
601 | else | |
602 | ln->left = fn; | |
603 | ||
604 | fn->parent = ln; | |
605 | } | |
606 | return ln; | |
607 | } | |
608 | ||
609 | /* | |
610 | * Insert routing information in a node. | |
611 | */ | |
612 | ||
613 | static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt, | |
86872cb5 | 614 | struct nl_info *info) |
1da177e4 LT |
615 | { |
616 | struct rt6_info *iter = NULL; | |
617 | struct rt6_info **ins; | |
618 | ||
619 | ins = &fn->leaf; | |
620 | ||
621 | if (fn->fn_flags&RTN_TL_ROOT && | |
622 | fn->leaf == &ip6_null_entry && | |
623 | !(rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ){ | |
624 | fn->leaf = rt; | |
625 | rt->u.next = NULL; | |
626 | goto out; | |
627 | } | |
628 | ||
629 | for (iter = fn->leaf; iter; iter=iter->u.next) { | |
630 | /* | |
631 | * Search for duplicates | |
632 | */ | |
633 | ||
634 | if (iter->rt6i_metric == rt->rt6i_metric) { | |
635 | /* | |
636 | * Same priority level | |
637 | */ | |
638 | ||
639 | if (iter->rt6i_dev == rt->rt6i_dev && | |
640 | iter->rt6i_idev == rt->rt6i_idev && | |
641 | ipv6_addr_equal(&iter->rt6i_gateway, | |
642 | &rt->rt6i_gateway)) { | |
643 | if (!(iter->rt6i_flags&RTF_EXPIRES)) | |
644 | return -EEXIST; | |
645 | iter->rt6i_expires = rt->rt6i_expires; | |
646 | if (!(rt->rt6i_flags&RTF_EXPIRES)) { | |
647 | iter->rt6i_flags &= ~RTF_EXPIRES; | |
648 | iter->rt6i_expires = 0; | |
649 | } | |
650 | return -EEXIST; | |
651 | } | |
652 | } | |
653 | ||
654 | if (iter->rt6i_metric > rt->rt6i_metric) | |
655 | break; | |
656 | ||
657 | ins = &iter->u.next; | |
658 | } | |
659 | ||
660 | /* | |
661 | * insert node | |
662 | */ | |
663 | ||
664 | out: | |
665 | rt->u.next = iter; | |
666 | *ins = rt; | |
667 | rt->rt6i_node = fn; | |
668 | atomic_inc(&rt->rt6i_ref); | |
86872cb5 | 669 | inet6_rt_notify(RTM_NEWROUTE, rt, info); |
1da177e4 LT |
670 | rt6_stats.fib_rt_entries++; |
671 | ||
672 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
673 | rt6_stats.fib_route_nodes++; | |
674 | fn->fn_flags |= RTN_RTINFO; | |
675 | } | |
676 | ||
677 | return 0; | |
678 | } | |
679 | ||
680 | static __inline__ void fib6_start_gc(struct rt6_info *rt) | |
681 | { | |
682 | if (ip6_fib_timer.expires == 0 && | |
683 | (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE))) | |
684 | mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval); | |
685 | } | |
686 | ||
687 | void fib6_force_start_gc(void) | |
688 | { | |
689 | if (ip6_fib_timer.expires == 0) | |
690 | mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval); | |
691 | } | |
692 | ||
693 | /* | |
694 | * Add routing information to the routing tree. | |
695 | * <destination addr>/<source addr> | |
696 | * with source addr info in sub-trees | |
697 | */ | |
698 | ||
86872cb5 | 699 | int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info) |
1da177e4 | 700 | { |
66729e18 | 701 | struct fib6_node *fn, *pn = NULL; |
1da177e4 LT |
702 | int err = -ENOMEM; |
703 | ||
704 | fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr), | |
705 | rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst)); | |
706 | ||
707 | if (fn == NULL) | |
708 | goto out; | |
709 | ||
66729e18 YH |
710 | pn = fn; |
711 | ||
1da177e4 LT |
712 | #ifdef CONFIG_IPV6_SUBTREES |
713 | if (rt->rt6i_src.plen) { | |
714 | struct fib6_node *sn; | |
715 | ||
716 | if (fn->subtree == NULL) { | |
717 | struct fib6_node *sfn; | |
718 | ||
719 | /* | |
720 | * Create subtree. | |
721 | * | |
722 | * fn[main tree] | |
723 | * | | |
724 | * sfn[subtree root] | |
725 | * \ | |
726 | * sn[new leaf node] | |
727 | */ | |
728 | ||
729 | /* Create subtree root node */ | |
730 | sfn = node_alloc(); | |
731 | if (sfn == NULL) | |
732 | goto st_failure; | |
733 | ||
734 | sfn->leaf = &ip6_null_entry; | |
735 | atomic_inc(&ip6_null_entry.rt6i_ref); | |
736 | sfn->fn_flags = RTN_ROOT; | |
737 | sfn->fn_sernum = fib6_new_sernum(); | |
738 | ||
739 | /* Now add the first leaf node to new subtree */ | |
740 | ||
741 | sn = fib6_add_1(sfn, &rt->rt6i_src.addr, | |
742 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
743 | offsetof(struct rt6_info, rt6i_src)); | |
744 | ||
745 | if (sn == NULL) { | |
746 | /* If it is failed, discard just allocated | |
747 | root, and then (in st_failure) stale node | |
748 | in main tree. | |
749 | */ | |
750 | node_free(sfn); | |
751 | goto st_failure; | |
752 | } | |
753 | ||
754 | /* Now link new subtree to main tree */ | |
755 | sfn->parent = fn; | |
756 | fn->subtree = sfn; | |
1da177e4 LT |
757 | } else { |
758 | sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr, | |
759 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
760 | offsetof(struct rt6_info, rt6i_src)); | |
761 | ||
762 | if (sn == NULL) | |
763 | goto st_failure; | |
764 | } | |
765 | ||
66729e18 YH |
766 | if (fn->leaf == NULL) { |
767 | fn->leaf = rt; | |
768 | atomic_inc(&rt->rt6i_ref); | |
769 | } | |
1da177e4 LT |
770 | fn = sn; |
771 | } | |
772 | #endif | |
773 | ||
86872cb5 | 774 | err = fib6_add_rt2node(fn, rt, info); |
1da177e4 LT |
775 | |
776 | if (err == 0) { | |
777 | fib6_start_gc(rt); | |
778 | if (!(rt->rt6i_flags&RTF_CACHE)) | |
2285adc1 | 779 | fib6_prune_clones(pn, rt); |
1da177e4 LT |
780 | } |
781 | ||
782 | out: | |
66729e18 YH |
783 | if (err) { |
784 | #ifdef CONFIG_IPV6_SUBTREES | |
785 | /* | |
786 | * If fib6_add_1 has cleared the old leaf pointer in the | |
787 | * super-tree leaf node we have to find a new one for it. | |
788 | */ | |
789 | if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) { | |
790 | pn->leaf = fib6_find_prefix(pn); | |
791 | #if RT6_DEBUG >= 2 | |
792 | if (!pn->leaf) { | |
793 | BUG_TRAP(pn->leaf != NULL); | |
794 | pn->leaf = &ip6_null_entry; | |
795 | } | |
796 | #endif | |
797 | atomic_inc(&pn->leaf->rt6i_ref); | |
798 | } | |
799 | #endif | |
1da177e4 | 800 | dst_free(&rt->u.dst); |
66729e18 | 801 | } |
1da177e4 LT |
802 | return err; |
803 | ||
804 | #ifdef CONFIG_IPV6_SUBTREES | |
805 | /* Subtree creation failed, probably main tree node | |
806 | is orphan. If it is, shoot it. | |
807 | */ | |
808 | st_failure: | |
809 | if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT))) | |
810 | fib6_repair_tree(fn); | |
811 | dst_free(&rt->u.dst); | |
812 | return err; | |
813 | #endif | |
814 | } | |
815 | ||
816 | /* | |
817 | * Routing tree lookup | |
818 | * | |
819 | */ | |
820 | ||
821 | struct lookup_args { | |
822 | int offset; /* key offset on rt6_info */ | |
823 | struct in6_addr *addr; /* search key */ | |
824 | }; | |
825 | ||
826 | static struct fib6_node * fib6_lookup_1(struct fib6_node *root, | |
827 | struct lookup_args *args) | |
828 | { | |
829 | struct fib6_node *fn; | |
830 | int dir; | |
831 | ||
832 | /* | |
833 | * Descend on a tree | |
834 | */ | |
835 | ||
836 | fn = root; | |
837 | ||
838 | for (;;) { | |
839 | struct fib6_node *next; | |
840 | ||
841 | dir = addr_bit_set(args->addr, fn->fn_bit); | |
842 | ||
843 | next = dir ? fn->right : fn->left; | |
844 | ||
845 | if (next) { | |
846 | fn = next; | |
847 | continue; | |
848 | } | |
849 | ||
850 | break; | |
851 | } | |
852 | ||
853 | while ((fn->fn_flags & RTN_ROOT) == 0) { | |
854 | #ifdef CONFIG_IPV6_SUBTREES | |
855 | if (fn->subtree) { | |
856 | struct fib6_node *st; | |
857 | struct lookup_args *narg; | |
858 | ||
859 | narg = args + 1; | |
860 | ||
861 | if (narg->addr) { | |
862 | st = fib6_lookup_1(fn->subtree, narg); | |
863 | ||
864 | if (st && !(st->fn_flags & RTN_ROOT)) | |
865 | return st; | |
866 | } | |
867 | } | |
868 | #endif | |
869 | ||
870 | if (fn->fn_flags & RTN_RTINFO) { | |
871 | struct rt6key *key; | |
872 | ||
873 | key = (struct rt6key *) ((u8 *) fn->leaf + | |
874 | args->offset); | |
875 | ||
876 | if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) | |
877 | return fn; | |
878 | } | |
879 | ||
880 | fn = fn->parent; | |
881 | } | |
882 | ||
883 | return NULL; | |
884 | } | |
885 | ||
886 | struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr, | |
887 | struct in6_addr *saddr) | |
888 | { | |
889 | struct lookup_args args[2]; | |
890 | struct fib6_node *fn; | |
891 | ||
892 | args[0].offset = offsetof(struct rt6_info, rt6i_dst); | |
893 | args[0].addr = daddr; | |
894 | ||
895 | #ifdef CONFIG_IPV6_SUBTREES | |
896 | args[1].offset = offsetof(struct rt6_info, rt6i_src); | |
897 | args[1].addr = saddr; | |
898 | #endif | |
899 | ||
900 | fn = fib6_lookup_1(root, args); | |
901 | ||
902 | if (fn == NULL || fn->fn_flags & RTN_TL_ROOT) | |
903 | fn = root; | |
904 | ||
905 | return fn; | |
906 | } | |
907 | ||
908 | /* | |
909 | * Get node with specified destination prefix (and source prefix, | |
910 | * if subtrees are used) | |
911 | */ | |
912 | ||
913 | ||
914 | static struct fib6_node * fib6_locate_1(struct fib6_node *root, | |
915 | struct in6_addr *addr, | |
916 | int plen, int offset) | |
917 | { | |
918 | struct fib6_node *fn; | |
919 | ||
920 | for (fn = root; fn ; ) { | |
921 | struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
922 | ||
923 | /* | |
924 | * Prefix match | |
925 | */ | |
926 | if (plen < fn->fn_bit || | |
927 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
928 | return NULL; | |
929 | ||
930 | if (plen == fn->fn_bit) | |
931 | return fn; | |
932 | ||
933 | /* | |
934 | * We have more bits to go | |
935 | */ | |
936 | if (addr_bit_set(addr, fn->fn_bit)) | |
937 | fn = fn->right; | |
938 | else | |
939 | fn = fn->left; | |
940 | } | |
941 | return NULL; | |
942 | } | |
943 | ||
944 | struct fib6_node * fib6_locate(struct fib6_node *root, | |
945 | struct in6_addr *daddr, int dst_len, | |
946 | struct in6_addr *saddr, int src_len) | |
947 | { | |
948 | struct fib6_node *fn; | |
949 | ||
950 | fn = fib6_locate_1(root, daddr, dst_len, | |
951 | offsetof(struct rt6_info, rt6i_dst)); | |
952 | ||
953 | #ifdef CONFIG_IPV6_SUBTREES | |
954 | if (src_len) { | |
955 | BUG_TRAP(saddr!=NULL); | |
956 | if (fn == NULL) | |
957 | fn = fn->subtree; | |
958 | if (fn) | |
959 | fn = fib6_locate_1(fn, saddr, src_len, | |
960 | offsetof(struct rt6_info, rt6i_src)); | |
961 | } | |
962 | #endif | |
963 | ||
964 | if (fn && fn->fn_flags&RTN_RTINFO) | |
965 | return fn; | |
966 | ||
967 | return NULL; | |
968 | } | |
969 | ||
970 | ||
971 | /* | |
972 | * Deletion | |
973 | * | |
974 | */ | |
975 | ||
976 | static struct rt6_info * fib6_find_prefix(struct fib6_node *fn) | |
977 | { | |
978 | if (fn->fn_flags&RTN_ROOT) | |
979 | return &ip6_null_entry; | |
980 | ||
981 | while(fn) { | |
982 | if(fn->left) | |
983 | return fn->left->leaf; | |
984 | ||
985 | if(fn->right) | |
986 | return fn->right->leaf; | |
987 | ||
988 | fn = SUBTREE(fn); | |
989 | } | |
990 | return NULL; | |
991 | } | |
992 | ||
993 | /* | |
994 | * Called to trim the tree of intermediate nodes when possible. "fn" | |
995 | * is the node we want to try and remove. | |
996 | */ | |
997 | ||
998 | static struct fib6_node * fib6_repair_tree(struct fib6_node *fn) | |
999 | { | |
1000 | int children; | |
1001 | int nstate; | |
1002 | struct fib6_node *child, *pn; | |
1003 | struct fib6_walker_t *w; | |
1004 | int iter = 0; | |
1005 | ||
1006 | for (;;) { | |
1007 | RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter); | |
1008 | iter++; | |
1009 | ||
1010 | BUG_TRAP(!(fn->fn_flags&RTN_RTINFO)); | |
1011 | BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT)); | |
1012 | BUG_TRAP(fn->leaf==NULL); | |
1013 | ||
1014 | children = 0; | |
1015 | child = NULL; | |
1016 | if (fn->right) child = fn->right, children |= 1; | |
1017 | if (fn->left) child = fn->left, children |= 2; | |
1018 | ||
1019 | if (children == 3 || SUBTREE(fn) | |
1020 | #ifdef CONFIG_IPV6_SUBTREES | |
1021 | /* Subtree root (i.e. fn) may have one child */ | |
1022 | || (children && fn->fn_flags&RTN_ROOT) | |
1023 | #endif | |
1024 | ) { | |
1025 | fn->leaf = fib6_find_prefix(fn); | |
1026 | #if RT6_DEBUG >= 2 | |
1027 | if (fn->leaf==NULL) { | |
1028 | BUG_TRAP(fn->leaf); | |
1029 | fn->leaf = &ip6_null_entry; | |
1030 | } | |
1031 | #endif | |
1032 | atomic_inc(&fn->leaf->rt6i_ref); | |
1033 | return fn->parent; | |
1034 | } | |
1035 | ||
1036 | pn = fn->parent; | |
1037 | #ifdef CONFIG_IPV6_SUBTREES | |
1038 | if (SUBTREE(pn) == fn) { | |
1039 | BUG_TRAP(fn->fn_flags&RTN_ROOT); | |
1040 | SUBTREE(pn) = NULL; | |
1041 | nstate = FWS_L; | |
1042 | } else { | |
1043 | BUG_TRAP(!(fn->fn_flags&RTN_ROOT)); | |
1044 | #endif | |
1045 | if (pn->right == fn) pn->right = child; | |
1046 | else if (pn->left == fn) pn->left = child; | |
1047 | #if RT6_DEBUG >= 2 | |
1048 | else BUG_TRAP(0); | |
1049 | #endif | |
1050 | if (child) | |
1051 | child->parent = pn; | |
1052 | nstate = FWS_R; | |
1053 | #ifdef CONFIG_IPV6_SUBTREES | |
1054 | } | |
1055 | #endif | |
1056 | ||
1057 | read_lock(&fib6_walker_lock); | |
1058 | FOR_WALKERS(w) { | |
1059 | if (child == NULL) { | |
1060 | if (w->root == fn) { | |
1061 | w->root = w->node = NULL; | |
1062 | RT6_TRACE("W %p adjusted by delroot 1\n", w); | |
1063 | } else if (w->node == fn) { | |
1064 | RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate); | |
1065 | w->node = pn; | |
1066 | w->state = nstate; | |
1067 | } | |
1068 | } else { | |
1069 | if (w->root == fn) { | |
1070 | w->root = child; | |
1071 | RT6_TRACE("W %p adjusted by delroot 2\n", w); | |
1072 | } | |
1073 | if (w->node == fn) { | |
1074 | w->node = child; | |
1075 | if (children&2) { | |
1076 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1077 | w->state = w->state>=FWS_R ? FWS_U : FWS_INIT; | |
1078 | } else { | |
1079 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1080 | w->state = w->state>=FWS_C ? FWS_U : FWS_INIT; | |
1081 | } | |
1082 | } | |
1083 | } | |
1084 | } | |
1085 | read_unlock(&fib6_walker_lock); | |
1086 | ||
1087 | node_free(fn); | |
1088 | if (pn->fn_flags&RTN_RTINFO || SUBTREE(pn)) | |
1089 | return pn; | |
1090 | ||
1091 | rt6_release(pn->leaf); | |
1092 | pn->leaf = NULL; | |
1093 | fn = pn; | |
1094 | } | |
1095 | } | |
1096 | ||
1097 | static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp, | |
86872cb5 | 1098 | struct nl_info *info) |
1da177e4 LT |
1099 | { |
1100 | struct fib6_walker_t *w; | |
1101 | struct rt6_info *rt = *rtp; | |
1102 | ||
1103 | RT6_TRACE("fib6_del_route\n"); | |
1104 | ||
1105 | /* Unlink it */ | |
1106 | *rtp = rt->u.next; | |
1107 | rt->rt6i_node = NULL; | |
1108 | rt6_stats.fib_rt_entries--; | |
1109 | rt6_stats.fib_discarded_routes++; | |
1110 | ||
1111 | /* Adjust walkers */ | |
1112 | read_lock(&fib6_walker_lock); | |
1113 | FOR_WALKERS(w) { | |
1114 | if (w->state == FWS_C && w->leaf == rt) { | |
1115 | RT6_TRACE("walker %p adjusted by delroute\n", w); | |
1116 | w->leaf = rt->u.next; | |
1117 | if (w->leaf == NULL) | |
1118 | w->state = FWS_U; | |
1119 | } | |
1120 | } | |
1121 | read_unlock(&fib6_walker_lock); | |
1122 | ||
1123 | rt->u.next = NULL; | |
1124 | ||
1125 | if (fn->leaf == NULL && fn->fn_flags&RTN_TL_ROOT) | |
1126 | fn->leaf = &ip6_null_entry; | |
1127 | ||
1128 | /* If it was last route, expunge its radix tree node */ | |
1129 | if (fn->leaf == NULL) { | |
1130 | fn->fn_flags &= ~RTN_RTINFO; | |
1131 | rt6_stats.fib_route_nodes--; | |
1132 | fn = fib6_repair_tree(fn); | |
1133 | } | |
1134 | ||
1135 | if (atomic_read(&rt->rt6i_ref) != 1) { | |
1136 | /* This route is used as dummy address holder in some split | |
1137 | * nodes. It is not leaked, but it still holds other resources, | |
1138 | * which must be released in time. So, scan ascendant nodes | |
1139 | * and replace dummy references to this route with references | |
1140 | * to still alive ones. | |
1141 | */ | |
1142 | while (fn) { | |
1143 | if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) { | |
1144 | fn->leaf = fib6_find_prefix(fn); | |
1145 | atomic_inc(&fn->leaf->rt6i_ref); | |
1146 | rt6_release(rt); | |
1147 | } | |
1148 | fn = fn->parent; | |
1149 | } | |
1150 | /* No more references are possible at this point. */ | |
1151 | if (atomic_read(&rt->rt6i_ref) != 1) BUG(); | |
1152 | } | |
1153 | ||
86872cb5 | 1154 | inet6_rt_notify(RTM_DELROUTE, rt, info); |
1da177e4 LT |
1155 | rt6_release(rt); |
1156 | } | |
1157 | ||
86872cb5 | 1158 | int fib6_del(struct rt6_info *rt, struct nl_info *info) |
1da177e4 LT |
1159 | { |
1160 | struct fib6_node *fn = rt->rt6i_node; | |
1161 | struct rt6_info **rtp; | |
1162 | ||
1163 | #if RT6_DEBUG >= 2 | |
1164 | if (rt->u.dst.obsolete>0) { | |
1165 | BUG_TRAP(fn==NULL); | |
1166 | return -ENOENT; | |
1167 | } | |
1168 | #endif | |
1169 | if (fn == NULL || rt == &ip6_null_entry) | |
1170 | return -ENOENT; | |
1171 | ||
1172 | BUG_TRAP(fn->fn_flags&RTN_RTINFO); | |
1173 | ||
1174 | if (!(rt->rt6i_flags&RTF_CACHE)) | |
1175 | fib6_prune_clones(fn, rt); | |
1176 | ||
1177 | /* | |
1178 | * Walk the leaf entries looking for ourself | |
1179 | */ | |
1180 | ||
1181 | for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.next) { | |
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, | |
1197 | * cleaning tree when a device goes down etc. etc. | |
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: | |
1231 | if (SUBTREE(fn)) { | |
1232 | w->node = SUBTREE(fn); | |
1233 | continue; | |
1234 | } | |
1235 | w->state = FWS_L; | |
1236 | #endif | |
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 | |
1266 | if (SUBTREE(pn) == fn) { | |
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; | |
1306 | struct fib6_cleaner_t *c = (struct fib6_cleaner_t*)w; | |
1307 | ||
1308 | for (rt = w->leaf; rt; rt = rt->u.next) { | |
1309 | res = c->func(rt, c->arg); | |
1310 | if (res < 0) { | |
1311 | w->leaf = rt; | |
86872cb5 | 1312 | res = fib6_del(rt, NULL); |
1da177e4 LT |
1313 | if (res) { |
1314 | #if RT6_DEBUG >= 2 | |
1315 | printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res); | |
1316 | #endif | |
1317 | continue; | |
1318 | } | |
1319 | return 0; | |
1320 | } | |
1321 | BUG_TRAP(res==0); | |
1322 | } | |
1323 | w->leaf = rt; | |
1324 | return 0; | |
1325 | } | |
1326 | ||
1327 | /* | |
1328 | * Convenient frontend to tree walker. | |
1329 | * | |
1330 | * func is called on each route. | |
1331 | * It may return -1 -> delete this route. | |
1332 | * 0 -> continue walking | |
1333 | * | |
1334 | * prune==1 -> only immediate children of node (certainly, | |
1335 | * ignoring pure split nodes) will be scanned. | |
1336 | */ | |
1337 | ||
8ce11e6a AB |
1338 | static void fib6_clean_tree(struct fib6_node *root, |
1339 | int (*func)(struct rt6_info *, void *arg), | |
1340 | int prune, void *arg) | |
1da177e4 LT |
1341 | { |
1342 | struct fib6_cleaner_t c; | |
1343 | ||
1344 | c.w.root = root; | |
1345 | c.w.func = fib6_clean_node; | |
1346 | c.w.prune = prune; | |
1347 | c.func = func; | |
1348 | c.arg = arg; | |
1349 | ||
1350 | fib6_walk(&c.w); | |
1351 | } | |
1352 | ||
c71099ac TG |
1353 | void fib6_clean_all(int (*func)(struct rt6_info *, void *arg), |
1354 | int prune, void *arg) | |
1355 | { | |
c71099ac | 1356 | struct fib6_table *table; |
1b43af54 PM |
1357 | struct hlist_node *node; |
1358 | unsigned int h; | |
c71099ac | 1359 | |
1b43af54 PM |
1360 | rcu_read_lock(); |
1361 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { | |
1362 | hlist_for_each_entry_rcu(table, node, &fib_table_hash[h], | |
1363 | tb6_hlist) { | |
c71099ac TG |
1364 | write_lock_bh(&table->tb6_lock); |
1365 | fib6_clean_tree(&table->tb6_root, func, prune, arg); | |
1366 | write_unlock_bh(&table->tb6_lock); | |
1367 | } | |
1368 | } | |
1b43af54 | 1369 | rcu_read_unlock(); |
c71099ac TG |
1370 | } |
1371 | ||
1da177e4 LT |
1372 | static int fib6_prune_clone(struct rt6_info *rt, void *arg) |
1373 | { | |
1374 | if (rt->rt6i_flags & RTF_CACHE) { | |
1375 | RT6_TRACE("pruning clone %p\n", rt); | |
1376 | return -1; | |
1377 | } | |
1378 | ||
1379 | return 0; | |
1380 | } | |
1381 | ||
1382 | static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt) | |
1383 | { | |
1384 | fib6_clean_tree(fn, fib6_prune_clone, 1, rt); | |
1385 | } | |
1386 | ||
1387 | /* | |
1388 | * Garbage collection | |
1389 | */ | |
1390 | ||
1391 | static struct fib6_gc_args | |
1392 | { | |
1393 | int timeout; | |
1394 | int more; | |
1395 | } gc_args; | |
1396 | ||
1397 | static int fib6_age(struct rt6_info *rt, void *arg) | |
1398 | { | |
1399 | unsigned long now = jiffies; | |
1400 | ||
1401 | /* | |
1402 | * check addrconf expiration here. | |
1403 | * Routes are expired even if they are in use. | |
1404 | * | |
1405 | * Also age clones. Note, that clones are aged out | |
1406 | * only if they are not in use now. | |
1407 | */ | |
1408 | ||
1409 | if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) { | |
1410 | if (time_after(now, rt->rt6i_expires)) { | |
1411 | RT6_TRACE("expiring %p\n", rt); | |
1da177e4 LT |
1412 | return -1; |
1413 | } | |
1414 | gc_args.more++; | |
1415 | } else if (rt->rt6i_flags & RTF_CACHE) { | |
1416 | if (atomic_read(&rt->u.dst.__refcnt) == 0 && | |
1417 | time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) { | |
1418 | RT6_TRACE("aging clone %p\n", rt); | |
1419 | return -1; | |
1420 | } else if ((rt->rt6i_flags & RTF_GATEWAY) && | |
1421 | (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) { | |
1422 | RT6_TRACE("purging route %p via non-router but gateway\n", | |
1423 | rt); | |
1424 | return -1; | |
1425 | } | |
1426 | gc_args.more++; | |
1427 | } | |
1428 | ||
1429 | return 0; | |
1430 | } | |
1431 | ||
1432 | static DEFINE_SPINLOCK(fib6_gc_lock); | |
1433 | ||
1434 | void fib6_run_gc(unsigned long dummy) | |
1435 | { | |
1436 | if (dummy != ~0UL) { | |
1437 | spin_lock_bh(&fib6_gc_lock); | |
1438 | gc_args.timeout = dummy ? (int)dummy : ip6_rt_gc_interval; | |
1439 | } else { | |
1440 | local_bh_disable(); | |
1441 | if (!spin_trylock(&fib6_gc_lock)) { | |
1442 | mod_timer(&ip6_fib_timer, jiffies + HZ); | |
1443 | local_bh_enable(); | |
1444 | return; | |
1445 | } | |
1446 | gc_args.timeout = ip6_rt_gc_interval; | |
1447 | } | |
1448 | gc_args.more = 0; | |
1449 | ||
1da177e4 | 1450 | ndisc_dst_gc(&gc_args.more); |
c71099ac | 1451 | fib6_clean_all(fib6_age, 0, NULL); |
1da177e4 LT |
1452 | |
1453 | if (gc_args.more) | |
1454 | mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval); | |
1455 | else { | |
1456 | del_timer(&ip6_fib_timer); | |
1457 | ip6_fib_timer.expires = 0; | |
1458 | } | |
1459 | spin_unlock_bh(&fib6_gc_lock); | |
1460 | } | |
1461 | ||
1462 | void __init fib6_init(void) | |
1463 | { | |
1464 | fib6_node_kmem = kmem_cache_create("fib6_nodes", | |
1465 | sizeof(struct fib6_node), | |
1466 | 0, SLAB_HWCACHE_ALIGN, | |
1467 | NULL, NULL); | |
1468 | if (!fib6_node_kmem) | |
1469 | panic("cannot create fib6_nodes cache"); | |
c71099ac TG |
1470 | |
1471 | fib6_tables_init(); | |
1da177e4 LT |
1472 | } |
1473 | ||
1474 | void fib6_gc_cleanup(void) | |
1475 | { | |
1476 | del_timer(&ip6_fib_timer); | |
1477 | kmem_cache_destroy(fib6_node_kmem); | |
1478 | } |