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); | |
83 | static struct fib6_node * fib6_repair_tree(struct fib6_node *fn); | |
90d41122 AB |
84 | static int fib6_walk(struct fib6_walker_t *w); |
85 | static int fib6_walk_continue(struct fib6_walker_t *w); | |
1da177e4 LT |
86 | |
87 | /* | |
88 | * A routing update causes an increase of the serial number on the | |
89 | * affected subtree. This allows for cached routes to be asynchronously | |
90 | * tested when modifications are made to the destination cache as a | |
91 | * result of redirects, path MTU changes, etc. | |
92 | */ | |
93 | ||
94 | static __u32 rt_sernum; | |
95 | ||
8d06afab | 96 | static DEFINE_TIMER(ip6_fib_timer, fib6_run_gc, 0, 0); |
1da177e4 | 97 | |
90d41122 | 98 | static struct fib6_walker_t fib6_walker_list = { |
1da177e4 LT |
99 | .prev = &fib6_walker_list, |
100 | .next = &fib6_walker_list, | |
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 | * | |
134 | * These assume a 32bit processor (although it will work on | |
135 | * 64bit processors) | |
136 | */ | |
137 | ||
138 | /* | |
139 | * test bit | |
140 | */ | |
141 | ||
142 | static __inline__ int addr_bit_set(void *token, int fn_bit) | |
143 | { | |
144 | __u32 *addr = token; | |
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 | ||
153 | if ((fn = kmem_cache_alloc(fib6_node_kmem, SLAB_ATOMIC)) != NULL) | |
154 | memset(fn, 0, sizeof(struct fib6_node)); | |
155 | ||
156 | return fn; | |
157 | } | |
158 | ||
159 | static __inline__ void node_free(struct fib6_node * fn) | |
160 | { | |
161 | kmem_cache_free(fib6_node_kmem, fn); | |
162 | } | |
163 | ||
164 | static __inline__ void rt6_release(struct rt6_info *rt) | |
165 | { | |
166 | if (atomic_dec_and_test(&rt->rt6i_ref)) | |
167 | dst_free(&rt->u.dst); | |
168 | } | |
169 | ||
c71099ac TG |
170 | static struct fib6_table fib6_main_tbl = { |
171 | .tb6_id = RT6_TABLE_MAIN, | |
172 | .tb6_lock = RW_LOCK_UNLOCKED, | |
173 | .tb6_root = { | |
174 | .leaf = &ip6_null_entry, | |
175 | .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO, | |
176 | }, | |
177 | }; | |
178 | ||
179 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES | |
1b43af54 PM |
180 | #define FIB_TABLE_HASHSZ 256 |
181 | #else | |
182 | #define FIB_TABLE_HASHSZ 1 | |
183 | #endif | |
184 | static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ]; | |
185 | ||
186 | static void fib6_link_table(struct fib6_table *tb) | |
187 | { | |
188 | unsigned int h; | |
189 | ||
190 | h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1); | |
191 | ||
192 | /* | |
193 | * No protection necessary, this is the only list mutatation | |
194 | * operation, tables never disappear once they exist. | |
195 | */ | |
196 | hlist_add_head_rcu(&tb->tb6_hlist, &fib_table_hash[h]); | |
197 | } | |
c71099ac | 198 | |
1b43af54 | 199 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
101367c2 TG |
200 | static struct fib6_table fib6_local_tbl = { |
201 | .tb6_id = RT6_TABLE_LOCAL, | |
202 | .tb6_lock = RW_LOCK_UNLOCKED, | |
203 | .tb6_root = { | |
204 | .leaf = &ip6_null_entry, | |
205 | .fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO, | |
206 | }, | |
207 | }; | |
208 | ||
c71099ac TG |
209 | static struct fib6_table *fib6_alloc_table(u32 id) |
210 | { | |
211 | struct fib6_table *table; | |
212 | ||
213 | table = kzalloc(sizeof(*table), GFP_ATOMIC); | |
214 | if (table != NULL) { | |
215 | table->tb6_id = id; | |
216 | table->tb6_lock = RW_LOCK_UNLOCKED; | |
217 | table->tb6_root.leaf = &ip6_null_entry; | |
218 | table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
219 | } | |
220 | ||
221 | return table; | |
222 | } | |
223 | ||
c71099ac TG |
224 | struct fib6_table *fib6_new_table(u32 id) |
225 | { | |
226 | struct fib6_table *tb; | |
227 | ||
228 | if (id == 0) | |
229 | id = RT6_TABLE_MAIN; | |
230 | tb = fib6_get_table(id); | |
231 | if (tb) | |
232 | return tb; | |
233 | ||
234 | tb = fib6_alloc_table(id); | |
235 | if (tb != NULL) | |
236 | fib6_link_table(tb); | |
237 | ||
238 | return tb; | |
239 | } | |
240 | ||
241 | struct fib6_table *fib6_get_table(u32 id) | |
242 | { | |
243 | struct fib6_table *tb; | |
244 | struct hlist_node *node; | |
245 | unsigned int h; | |
246 | ||
247 | if (id == 0) | |
248 | id = RT6_TABLE_MAIN; | |
249 | h = id & (FIB_TABLE_HASHSZ - 1); | |
250 | rcu_read_lock(); | |
251 | hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb6_hlist) { | |
252 | if (tb->tb6_id == id) { | |
253 | rcu_read_unlock(); | |
254 | return tb; | |
255 | } | |
256 | } | |
257 | rcu_read_unlock(); | |
258 | ||
259 | return NULL; | |
260 | } | |
261 | ||
c71099ac TG |
262 | static void __init fib6_tables_init(void) |
263 | { | |
264 | fib6_link_table(&fib6_main_tbl); | |
101367c2 | 265 | fib6_link_table(&fib6_local_tbl); |
c71099ac TG |
266 | } |
267 | ||
268 | #else | |
269 | ||
270 | struct fib6_table *fib6_new_table(u32 id) | |
271 | { | |
272 | return fib6_get_table(id); | |
273 | } | |
274 | ||
275 | struct fib6_table *fib6_get_table(u32 id) | |
276 | { | |
277 | return &fib6_main_tbl; | |
278 | } | |
279 | ||
280 | struct dst_entry *fib6_rule_lookup(struct flowi *fl, int flags, | |
281 | pol_lookup_t lookup) | |
282 | { | |
283 | return (struct dst_entry *) lookup(&fib6_main_tbl, fl, flags); | |
284 | } | |
285 | ||
286 | static void __init fib6_tables_init(void) | |
287 | { | |
1b43af54 | 288 | fib6_link_table(&fib6_main_tbl); |
c71099ac TG |
289 | } |
290 | ||
291 | #endif | |
292 | ||
1b43af54 PM |
293 | static int fib6_dump_node(struct fib6_walker_t *w) |
294 | { | |
295 | int res; | |
296 | struct rt6_info *rt; | |
297 | ||
298 | for (rt = w->leaf; rt; rt = rt->u.next) { | |
299 | res = rt6_dump_route(rt, w->args); | |
300 | if (res < 0) { | |
301 | /* Frame is full, suspend walking */ | |
302 | w->leaf = rt; | |
303 | return 1; | |
304 | } | |
305 | BUG_TRAP(res!=0); | |
306 | } | |
307 | w->leaf = NULL; | |
308 | return 0; | |
309 | } | |
310 | ||
311 | static void fib6_dump_end(struct netlink_callback *cb) | |
312 | { | |
313 | struct fib6_walker_t *w = (void*)cb->args[2]; | |
314 | ||
315 | if (w) { | |
316 | cb->args[2] = 0; | |
317 | kfree(w); | |
318 | } | |
319 | cb->done = (void*)cb->args[3]; | |
320 | cb->args[1] = 3; | |
321 | } | |
322 | ||
323 | static int fib6_dump_done(struct netlink_callback *cb) | |
324 | { | |
325 | fib6_dump_end(cb); | |
326 | return cb->done ? cb->done(cb) : 0; | |
327 | } | |
328 | ||
329 | static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb, | |
330 | struct netlink_callback *cb) | |
331 | { | |
332 | struct fib6_walker_t *w; | |
333 | int res; | |
334 | ||
335 | w = (void *)cb->args[2]; | |
336 | w->root = &table->tb6_root; | |
337 | ||
338 | if (cb->args[4] == 0) { | |
339 | read_lock_bh(&table->tb6_lock); | |
340 | res = fib6_walk(w); | |
341 | read_unlock_bh(&table->tb6_lock); | |
342 | if (res > 0) | |
343 | cb->args[4] = 1; | |
344 | } else { | |
345 | read_lock_bh(&table->tb6_lock); | |
346 | res = fib6_walk_continue(w); | |
347 | read_unlock_bh(&table->tb6_lock); | |
348 | if (res != 0) { | |
349 | if (res < 0) | |
350 | fib6_walker_unlink(w); | |
351 | goto end; | |
352 | } | |
353 | fib6_walker_unlink(w); | |
354 | cb->args[4] = 0; | |
355 | } | |
356 | end: | |
357 | return res; | |
358 | } | |
359 | ||
360 | int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) | |
361 | { | |
362 | unsigned int h, s_h; | |
363 | unsigned int e = 0, s_e; | |
364 | struct rt6_rtnl_dump_arg arg; | |
365 | struct fib6_walker_t *w; | |
366 | struct fib6_table *tb; | |
367 | struct hlist_node *node; | |
368 | int res = 0; | |
369 | ||
370 | s_h = cb->args[0]; | |
371 | s_e = cb->args[1]; | |
372 | ||
373 | w = (void *)cb->args[2]; | |
374 | if (w == NULL) { | |
375 | /* New dump: | |
376 | * | |
377 | * 1. hook callback destructor. | |
378 | */ | |
379 | cb->args[3] = (long)cb->done; | |
380 | cb->done = fib6_dump_done; | |
381 | ||
382 | /* | |
383 | * 2. allocate and initialize walker. | |
384 | */ | |
385 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
386 | if (w == NULL) | |
387 | return -ENOMEM; | |
388 | w->func = fib6_dump_node; | |
389 | cb->args[2] = (long)w; | |
390 | } | |
391 | ||
392 | arg.skb = skb; | |
393 | arg.cb = cb; | |
394 | w->args = &arg; | |
395 | ||
396 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { | |
397 | e = 0; | |
398 | hlist_for_each_entry(tb, node, &fib_table_hash[h], tb6_hlist) { | |
399 | if (e < s_e) | |
400 | goto next; | |
401 | res = fib6_dump_table(tb, skb, cb); | |
402 | if (res != 0) | |
403 | goto out; | |
404 | next: | |
405 | e++; | |
406 | } | |
407 | } | |
408 | out: | |
409 | cb->args[1] = e; | |
410 | cb->args[0] = h; | |
411 | ||
412 | res = res < 0 ? res : skb->len; | |
413 | if (res <= 0) | |
414 | fib6_dump_end(cb); | |
415 | return res; | |
416 | } | |
1da177e4 LT |
417 | |
418 | /* | |
419 | * Routing Table | |
420 | * | |
421 | * return the appropriate node for a routing tree "add" operation | |
422 | * by either creating and inserting or by returning an existing | |
423 | * node. | |
424 | */ | |
425 | ||
426 | static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr, | |
427 | int addrlen, int plen, | |
428 | int offset) | |
429 | { | |
430 | struct fib6_node *fn, *in, *ln; | |
431 | struct fib6_node *pn = NULL; | |
432 | struct rt6key *key; | |
433 | int bit; | |
434 | int dir = 0; | |
435 | __u32 sernum = fib6_new_sernum(); | |
436 | ||
437 | RT6_TRACE("fib6_add_1\n"); | |
438 | ||
439 | /* insert node in tree */ | |
440 | ||
441 | fn = root; | |
442 | ||
443 | do { | |
444 | key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
445 | ||
446 | /* | |
447 | * Prefix match | |
448 | */ | |
449 | if (plen < fn->fn_bit || | |
450 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
451 | goto insert_above; | |
452 | ||
453 | /* | |
454 | * Exact match ? | |
455 | */ | |
456 | ||
457 | if (plen == fn->fn_bit) { | |
458 | /* clean up an intermediate node */ | |
459 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
460 | rt6_release(fn->leaf); | |
461 | fn->leaf = NULL; | |
462 | } | |
463 | ||
464 | fn->fn_sernum = sernum; | |
465 | ||
466 | return fn; | |
467 | } | |
468 | ||
469 | /* | |
470 | * We have more bits to go | |
471 | */ | |
472 | ||
473 | /* Try to walk down on tree. */ | |
474 | fn->fn_sernum = sernum; | |
475 | dir = addr_bit_set(addr, fn->fn_bit); | |
476 | pn = fn; | |
477 | fn = dir ? fn->right: fn->left; | |
478 | } while (fn); | |
479 | ||
480 | /* | |
481 | * We walked to the bottom of tree. | |
482 | * Create new leaf node without children. | |
483 | */ | |
484 | ||
485 | ln = node_alloc(); | |
486 | ||
487 | if (ln == NULL) | |
488 | return NULL; | |
489 | ln->fn_bit = plen; | |
490 | ||
491 | ln->parent = pn; | |
492 | ln->fn_sernum = sernum; | |
493 | ||
494 | if (dir) | |
495 | pn->right = ln; | |
496 | else | |
497 | pn->left = ln; | |
498 | ||
499 | return ln; | |
500 | ||
501 | ||
502 | insert_above: | |
503 | /* | |
504 | * split since we don't have a common prefix anymore or | |
505 | * we have a less significant route. | |
506 | * we've to insert an intermediate node on the list | |
507 | * this new node will point to the one we need to create | |
508 | * and the current | |
509 | */ | |
510 | ||
511 | pn = fn->parent; | |
512 | ||
513 | /* find 1st bit in difference between the 2 addrs. | |
514 | ||
971f359d | 515 | See comment in __ipv6_addr_diff: bit may be an invalid value, |
1da177e4 LT |
516 | but if it is >= plen, the value is ignored in any case. |
517 | */ | |
518 | ||
971f359d | 519 | bit = __ipv6_addr_diff(addr, &key->addr, addrlen); |
1da177e4 LT |
520 | |
521 | /* | |
522 | * (intermediate)[in] | |
523 | * / \ | |
524 | * (new leaf node)[ln] (old node)[fn] | |
525 | */ | |
526 | if (plen > bit) { | |
527 | in = node_alloc(); | |
528 | ln = node_alloc(); | |
529 | ||
530 | if (in == NULL || ln == NULL) { | |
531 | if (in) | |
532 | node_free(in); | |
533 | if (ln) | |
534 | node_free(ln); | |
535 | return NULL; | |
536 | } | |
537 | ||
538 | /* | |
539 | * new intermediate node. | |
540 | * RTN_RTINFO will | |
541 | * be off since that an address that chooses one of | |
542 | * the branches would not match less specific routes | |
543 | * in the other branch | |
544 | */ | |
545 | ||
546 | in->fn_bit = bit; | |
547 | ||
548 | in->parent = pn; | |
549 | in->leaf = fn->leaf; | |
550 | atomic_inc(&in->leaf->rt6i_ref); | |
551 | ||
552 | in->fn_sernum = sernum; | |
553 | ||
554 | /* update parent pointer */ | |
555 | if (dir) | |
556 | pn->right = in; | |
557 | else | |
558 | pn->left = in; | |
559 | ||
560 | ln->fn_bit = plen; | |
561 | ||
562 | ln->parent = in; | |
563 | fn->parent = in; | |
564 | ||
565 | ln->fn_sernum = sernum; | |
566 | ||
567 | if (addr_bit_set(addr, bit)) { | |
568 | in->right = ln; | |
569 | in->left = fn; | |
570 | } else { | |
571 | in->left = ln; | |
572 | in->right = fn; | |
573 | } | |
574 | } else { /* plen <= bit */ | |
575 | ||
576 | /* | |
577 | * (new leaf node)[ln] | |
578 | * / \ | |
579 | * (old node)[fn] NULL | |
580 | */ | |
581 | ||
582 | ln = node_alloc(); | |
583 | ||
584 | if (ln == NULL) | |
585 | return NULL; | |
586 | ||
587 | ln->fn_bit = plen; | |
588 | ||
589 | ln->parent = pn; | |
590 | ||
591 | ln->fn_sernum = sernum; | |
592 | ||
593 | if (dir) | |
594 | pn->right = ln; | |
595 | else | |
596 | pn->left = ln; | |
597 | ||
598 | if (addr_bit_set(&key->addr, plen)) | |
599 | ln->right = fn; | |
600 | else | |
601 | ln->left = fn; | |
602 | ||
603 | fn->parent = ln; | |
604 | } | |
605 | return ln; | |
606 | } | |
607 | ||
608 | /* | |
609 | * Insert routing information in a node. | |
610 | */ | |
611 | ||
612 | static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt, | |
0d51aa80 | 613 | struct nlmsghdr *nlh, struct netlink_skb_parms *req) |
1da177e4 LT |
614 | { |
615 | struct rt6_info *iter = NULL; | |
616 | struct rt6_info **ins; | |
617 | ||
618 | ins = &fn->leaf; | |
619 | ||
620 | if (fn->fn_flags&RTN_TL_ROOT && | |
621 | fn->leaf == &ip6_null_entry && | |
622 | !(rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ){ | |
623 | fn->leaf = rt; | |
624 | rt->u.next = NULL; | |
625 | goto out; | |
626 | } | |
627 | ||
628 | for (iter = fn->leaf; iter; iter=iter->u.next) { | |
629 | /* | |
630 | * Search for duplicates | |
631 | */ | |
632 | ||
633 | if (iter->rt6i_metric == rt->rt6i_metric) { | |
634 | /* | |
635 | * Same priority level | |
636 | */ | |
637 | ||
638 | if (iter->rt6i_dev == rt->rt6i_dev && | |
639 | iter->rt6i_idev == rt->rt6i_idev && | |
640 | ipv6_addr_equal(&iter->rt6i_gateway, | |
641 | &rt->rt6i_gateway)) { | |
642 | if (!(iter->rt6i_flags&RTF_EXPIRES)) | |
643 | return -EEXIST; | |
644 | iter->rt6i_expires = rt->rt6i_expires; | |
645 | if (!(rt->rt6i_flags&RTF_EXPIRES)) { | |
646 | iter->rt6i_flags &= ~RTF_EXPIRES; | |
647 | iter->rt6i_expires = 0; | |
648 | } | |
649 | return -EEXIST; | |
650 | } | |
651 | } | |
652 | ||
653 | if (iter->rt6i_metric > rt->rt6i_metric) | |
654 | break; | |
655 | ||
656 | ins = &iter->u.next; | |
657 | } | |
658 | ||
659 | /* | |
660 | * insert node | |
661 | */ | |
662 | ||
663 | out: | |
664 | rt->u.next = iter; | |
665 | *ins = rt; | |
666 | rt->rt6i_node = fn; | |
667 | atomic_inc(&rt->rt6i_ref); | |
0d51aa80 | 668 | inet6_rt_notify(RTM_NEWROUTE, rt, nlh, req); |
1da177e4 LT |
669 | rt6_stats.fib_rt_entries++; |
670 | ||
671 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
672 | rt6_stats.fib_route_nodes++; | |
673 | fn->fn_flags |= RTN_RTINFO; | |
674 | } | |
675 | ||
676 | return 0; | |
677 | } | |
678 | ||
679 | static __inline__ void fib6_start_gc(struct rt6_info *rt) | |
680 | { | |
681 | if (ip6_fib_timer.expires == 0 && | |
682 | (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE))) | |
683 | mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval); | |
684 | } | |
685 | ||
686 | void fib6_force_start_gc(void) | |
687 | { | |
688 | if (ip6_fib_timer.expires == 0) | |
689 | mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval); | |
690 | } | |
691 | ||
692 | /* | |
693 | * Add routing information to the routing tree. | |
694 | * <destination addr>/<source addr> | |
695 | * with source addr info in sub-trees | |
696 | */ | |
697 | ||
0d51aa80 JHS |
698 | int fib6_add(struct fib6_node *root, struct rt6_info *rt, |
699 | struct nlmsghdr *nlh, void *_rtattr, struct netlink_skb_parms *req) | |
1da177e4 LT |
700 | { |
701 | struct fib6_node *fn; | |
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 | ||
710 | #ifdef CONFIG_IPV6_SUBTREES | |
711 | if (rt->rt6i_src.plen) { | |
712 | struct fib6_node *sn; | |
713 | ||
714 | if (fn->subtree == NULL) { | |
715 | struct fib6_node *sfn; | |
716 | ||
717 | /* | |
718 | * Create subtree. | |
719 | * | |
720 | * fn[main tree] | |
721 | * | | |
722 | * sfn[subtree root] | |
723 | * \ | |
724 | * sn[new leaf node] | |
725 | */ | |
726 | ||
727 | /* Create subtree root node */ | |
728 | sfn = node_alloc(); | |
729 | if (sfn == NULL) | |
730 | goto st_failure; | |
731 | ||
732 | sfn->leaf = &ip6_null_entry; | |
733 | atomic_inc(&ip6_null_entry.rt6i_ref); | |
734 | sfn->fn_flags = RTN_ROOT; | |
735 | sfn->fn_sernum = fib6_new_sernum(); | |
736 | ||
737 | /* Now add the first leaf node to new subtree */ | |
738 | ||
739 | sn = fib6_add_1(sfn, &rt->rt6i_src.addr, | |
740 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
741 | offsetof(struct rt6_info, rt6i_src)); | |
742 | ||
743 | if (sn == NULL) { | |
744 | /* If it is failed, discard just allocated | |
745 | root, and then (in st_failure) stale node | |
746 | in main tree. | |
747 | */ | |
748 | node_free(sfn); | |
749 | goto st_failure; | |
750 | } | |
751 | ||
752 | /* Now link new subtree to main tree */ | |
753 | sfn->parent = fn; | |
754 | fn->subtree = sfn; | |
755 | if (fn->leaf == NULL) { | |
756 | fn->leaf = rt; | |
757 | atomic_inc(&rt->rt6i_ref); | |
758 | } | |
759 | } else { | |
760 | sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr, | |
761 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
762 | offsetof(struct rt6_info, rt6i_src)); | |
763 | ||
764 | if (sn == NULL) | |
765 | goto st_failure; | |
766 | } | |
767 | ||
768 | fn = sn; | |
769 | } | |
770 | #endif | |
771 | ||
0d51aa80 | 772 | err = fib6_add_rt2node(fn, rt, nlh, req); |
1da177e4 LT |
773 | |
774 | if (err == 0) { | |
775 | fib6_start_gc(rt); | |
776 | if (!(rt->rt6i_flags&RTF_CACHE)) | |
777 | fib6_prune_clones(fn, rt); | |
778 | } | |
779 | ||
780 | out: | |
781 | if (err) | |
782 | dst_free(&rt->u.dst); | |
783 | return err; | |
784 | ||
785 | #ifdef CONFIG_IPV6_SUBTREES | |
786 | /* Subtree creation failed, probably main tree node | |
787 | is orphan. If it is, shoot it. | |
788 | */ | |
789 | st_failure: | |
790 | if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT))) | |
791 | fib6_repair_tree(fn); | |
792 | dst_free(&rt->u.dst); | |
793 | return err; | |
794 | #endif | |
795 | } | |
796 | ||
797 | /* | |
798 | * Routing tree lookup | |
799 | * | |
800 | */ | |
801 | ||
802 | struct lookup_args { | |
803 | int offset; /* key offset on rt6_info */ | |
804 | struct in6_addr *addr; /* search key */ | |
805 | }; | |
806 | ||
807 | static struct fib6_node * fib6_lookup_1(struct fib6_node *root, | |
808 | struct lookup_args *args) | |
809 | { | |
810 | struct fib6_node *fn; | |
811 | int dir; | |
812 | ||
813 | /* | |
814 | * Descend on a tree | |
815 | */ | |
816 | ||
817 | fn = root; | |
818 | ||
819 | for (;;) { | |
820 | struct fib6_node *next; | |
821 | ||
822 | dir = addr_bit_set(args->addr, fn->fn_bit); | |
823 | ||
824 | next = dir ? fn->right : fn->left; | |
825 | ||
826 | if (next) { | |
827 | fn = next; | |
828 | continue; | |
829 | } | |
830 | ||
831 | break; | |
832 | } | |
833 | ||
834 | while ((fn->fn_flags & RTN_ROOT) == 0) { | |
835 | #ifdef CONFIG_IPV6_SUBTREES | |
836 | if (fn->subtree) { | |
837 | struct fib6_node *st; | |
838 | struct lookup_args *narg; | |
839 | ||
840 | narg = args + 1; | |
841 | ||
842 | if (narg->addr) { | |
843 | st = fib6_lookup_1(fn->subtree, narg); | |
844 | ||
845 | if (st && !(st->fn_flags & RTN_ROOT)) | |
846 | return st; | |
847 | } | |
848 | } | |
849 | #endif | |
850 | ||
851 | if (fn->fn_flags & RTN_RTINFO) { | |
852 | struct rt6key *key; | |
853 | ||
854 | key = (struct rt6key *) ((u8 *) fn->leaf + | |
855 | args->offset); | |
856 | ||
857 | if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) | |
858 | return fn; | |
859 | } | |
860 | ||
861 | fn = fn->parent; | |
862 | } | |
863 | ||
864 | return NULL; | |
865 | } | |
866 | ||
867 | struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr, | |
868 | struct in6_addr *saddr) | |
869 | { | |
870 | struct lookup_args args[2]; | |
871 | struct fib6_node *fn; | |
872 | ||
873 | args[0].offset = offsetof(struct rt6_info, rt6i_dst); | |
874 | args[0].addr = daddr; | |
875 | ||
876 | #ifdef CONFIG_IPV6_SUBTREES | |
877 | args[1].offset = offsetof(struct rt6_info, rt6i_src); | |
878 | args[1].addr = saddr; | |
879 | #endif | |
880 | ||
881 | fn = fib6_lookup_1(root, args); | |
882 | ||
883 | if (fn == NULL || fn->fn_flags & RTN_TL_ROOT) | |
884 | fn = root; | |
885 | ||
886 | return fn; | |
887 | } | |
888 | ||
889 | /* | |
890 | * Get node with specified destination prefix (and source prefix, | |
891 | * if subtrees are used) | |
892 | */ | |
893 | ||
894 | ||
895 | static struct fib6_node * fib6_locate_1(struct fib6_node *root, | |
896 | struct in6_addr *addr, | |
897 | int plen, int offset) | |
898 | { | |
899 | struct fib6_node *fn; | |
900 | ||
901 | for (fn = root; fn ; ) { | |
902 | struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
903 | ||
904 | /* | |
905 | * Prefix match | |
906 | */ | |
907 | if (plen < fn->fn_bit || | |
908 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
909 | return NULL; | |
910 | ||
911 | if (plen == fn->fn_bit) | |
912 | return fn; | |
913 | ||
914 | /* | |
915 | * We have more bits to go | |
916 | */ | |
917 | if (addr_bit_set(addr, fn->fn_bit)) | |
918 | fn = fn->right; | |
919 | else | |
920 | fn = fn->left; | |
921 | } | |
922 | return NULL; | |
923 | } | |
924 | ||
925 | struct fib6_node * fib6_locate(struct fib6_node *root, | |
926 | struct in6_addr *daddr, int dst_len, | |
927 | struct in6_addr *saddr, int src_len) | |
928 | { | |
929 | struct fib6_node *fn; | |
930 | ||
931 | fn = fib6_locate_1(root, daddr, dst_len, | |
932 | offsetof(struct rt6_info, rt6i_dst)); | |
933 | ||
934 | #ifdef CONFIG_IPV6_SUBTREES | |
935 | if (src_len) { | |
936 | BUG_TRAP(saddr!=NULL); | |
937 | if (fn == NULL) | |
938 | fn = fn->subtree; | |
939 | if (fn) | |
940 | fn = fib6_locate_1(fn, saddr, src_len, | |
941 | offsetof(struct rt6_info, rt6i_src)); | |
942 | } | |
943 | #endif | |
944 | ||
945 | if (fn && fn->fn_flags&RTN_RTINFO) | |
946 | return fn; | |
947 | ||
948 | return NULL; | |
949 | } | |
950 | ||
951 | ||
952 | /* | |
953 | * Deletion | |
954 | * | |
955 | */ | |
956 | ||
957 | static struct rt6_info * fib6_find_prefix(struct fib6_node *fn) | |
958 | { | |
959 | if (fn->fn_flags&RTN_ROOT) | |
960 | return &ip6_null_entry; | |
961 | ||
962 | while(fn) { | |
963 | if(fn->left) | |
964 | return fn->left->leaf; | |
965 | ||
966 | if(fn->right) | |
967 | return fn->right->leaf; | |
968 | ||
969 | fn = SUBTREE(fn); | |
970 | } | |
971 | return NULL; | |
972 | } | |
973 | ||
974 | /* | |
975 | * Called to trim the tree of intermediate nodes when possible. "fn" | |
976 | * is the node we want to try and remove. | |
977 | */ | |
978 | ||
979 | static struct fib6_node * fib6_repair_tree(struct fib6_node *fn) | |
980 | { | |
981 | int children; | |
982 | int nstate; | |
983 | struct fib6_node *child, *pn; | |
984 | struct fib6_walker_t *w; | |
985 | int iter = 0; | |
986 | ||
987 | for (;;) { | |
988 | RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter); | |
989 | iter++; | |
990 | ||
991 | BUG_TRAP(!(fn->fn_flags&RTN_RTINFO)); | |
992 | BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT)); | |
993 | BUG_TRAP(fn->leaf==NULL); | |
994 | ||
995 | children = 0; | |
996 | child = NULL; | |
997 | if (fn->right) child = fn->right, children |= 1; | |
998 | if (fn->left) child = fn->left, children |= 2; | |
999 | ||
1000 | if (children == 3 || SUBTREE(fn) | |
1001 | #ifdef CONFIG_IPV6_SUBTREES | |
1002 | /* Subtree root (i.e. fn) may have one child */ | |
1003 | || (children && fn->fn_flags&RTN_ROOT) | |
1004 | #endif | |
1005 | ) { | |
1006 | fn->leaf = fib6_find_prefix(fn); | |
1007 | #if RT6_DEBUG >= 2 | |
1008 | if (fn->leaf==NULL) { | |
1009 | BUG_TRAP(fn->leaf); | |
1010 | fn->leaf = &ip6_null_entry; | |
1011 | } | |
1012 | #endif | |
1013 | atomic_inc(&fn->leaf->rt6i_ref); | |
1014 | return fn->parent; | |
1015 | } | |
1016 | ||
1017 | pn = fn->parent; | |
1018 | #ifdef CONFIG_IPV6_SUBTREES | |
1019 | if (SUBTREE(pn) == fn) { | |
1020 | BUG_TRAP(fn->fn_flags&RTN_ROOT); | |
1021 | SUBTREE(pn) = NULL; | |
1022 | nstate = FWS_L; | |
1023 | } else { | |
1024 | BUG_TRAP(!(fn->fn_flags&RTN_ROOT)); | |
1025 | #endif | |
1026 | if (pn->right == fn) pn->right = child; | |
1027 | else if (pn->left == fn) pn->left = child; | |
1028 | #if RT6_DEBUG >= 2 | |
1029 | else BUG_TRAP(0); | |
1030 | #endif | |
1031 | if (child) | |
1032 | child->parent = pn; | |
1033 | nstate = FWS_R; | |
1034 | #ifdef CONFIG_IPV6_SUBTREES | |
1035 | } | |
1036 | #endif | |
1037 | ||
1038 | read_lock(&fib6_walker_lock); | |
1039 | FOR_WALKERS(w) { | |
1040 | if (child == NULL) { | |
1041 | if (w->root == fn) { | |
1042 | w->root = w->node = NULL; | |
1043 | RT6_TRACE("W %p adjusted by delroot 1\n", w); | |
1044 | } else if (w->node == fn) { | |
1045 | RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate); | |
1046 | w->node = pn; | |
1047 | w->state = nstate; | |
1048 | } | |
1049 | } else { | |
1050 | if (w->root == fn) { | |
1051 | w->root = child; | |
1052 | RT6_TRACE("W %p adjusted by delroot 2\n", w); | |
1053 | } | |
1054 | if (w->node == fn) { | |
1055 | w->node = child; | |
1056 | if (children&2) { | |
1057 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1058 | w->state = w->state>=FWS_R ? FWS_U : FWS_INIT; | |
1059 | } else { | |
1060 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1061 | w->state = w->state>=FWS_C ? FWS_U : FWS_INIT; | |
1062 | } | |
1063 | } | |
1064 | } | |
1065 | } | |
1066 | read_unlock(&fib6_walker_lock); | |
1067 | ||
1068 | node_free(fn); | |
1069 | if (pn->fn_flags&RTN_RTINFO || SUBTREE(pn)) | |
1070 | return pn; | |
1071 | ||
1072 | rt6_release(pn->leaf); | |
1073 | pn->leaf = NULL; | |
1074 | fn = pn; | |
1075 | } | |
1076 | } | |
1077 | ||
1078 | static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp, | |
0d51aa80 | 1079 | struct nlmsghdr *nlh, void *_rtattr, struct netlink_skb_parms *req) |
1da177e4 LT |
1080 | { |
1081 | struct fib6_walker_t *w; | |
1082 | struct rt6_info *rt = *rtp; | |
1083 | ||
1084 | RT6_TRACE("fib6_del_route\n"); | |
1085 | ||
1086 | /* Unlink it */ | |
1087 | *rtp = rt->u.next; | |
1088 | rt->rt6i_node = NULL; | |
1089 | rt6_stats.fib_rt_entries--; | |
1090 | rt6_stats.fib_discarded_routes++; | |
1091 | ||
1092 | /* Adjust walkers */ | |
1093 | read_lock(&fib6_walker_lock); | |
1094 | FOR_WALKERS(w) { | |
1095 | if (w->state == FWS_C && w->leaf == rt) { | |
1096 | RT6_TRACE("walker %p adjusted by delroute\n", w); | |
1097 | w->leaf = rt->u.next; | |
1098 | if (w->leaf == NULL) | |
1099 | w->state = FWS_U; | |
1100 | } | |
1101 | } | |
1102 | read_unlock(&fib6_walker_lock); | |
1103 | ||
1104 | rt->u.next = NULL; | |
1105 | ||
1106 | if (fn->leaf == NULL && fn->fn_flags&RTN_TL_ROOT) | |
1107 | fn->leaf = &ip6_null_entry; | |
1108 | ||
1109 | /* If it was last route, expunge its radix tree node */ | |
1110 | if (fn->leaf == NULL) { | |
1111 | fn->fn_flags &= ~RTN_RTINFO; | |
1112 | rt6_stats.fib_route_nodes--; | |
1113 | fn = fib6_repair_tree(fn); | |
1114 | } | |
1115 | ||
1116 | if (atomic_read(&rt->rt6i_ref) != 1) { | |
1117 | /* This route is used as dummy address holder in some split | |
1118 | * nodes. It is not leaked, but it still holds other resources, | |
1119 | * which must be released in time. So, scan ascendant nodes | |
1120 | * and replace dummy references to this route with references | |
1121 | * to still alive ones. | |
1122 | */ | |
1123 | while (fn) { | |
1124 | if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) { | |
1125 | fn->leaf = fib6_find_prefix(fn); | |
1126 | atomic_inc(&fn->leaf->rt6i_ref); | |
1127 | rt6_release(rt); | |
1128 | } | |
1129 | fn = fn->parent; | |
1130 | } | |
1131 | /* No more references are possible at this point. */ | |
1132 | if (atomic_read(&rt->rt6i_ref) != 1) BUG(); | |
1133 | } | |
1134 | ||
0d51aa80 | 1135 | inet6_rt_notify(RTM_DELROUTE, rt, nlh, req); |
1da177e4 LT |
1136 | rt6_release(rt); |
1137 | } | |
1138 | ||
0d51aa80 | 1139 | int fib6_del(struct rt6_info *rt, struct nlmsghdr *nlh, void *_rtattr, struct netlink_skb_parms *req) |
1da177e4 LT |
1140 | { |
1141 | struct fib6_node *fn = rt->rt6i_node; | |
1142 | struct rt6_info **rtp; | |
1143 | ||
1144 | #if RT6_DEBUG >= 2 | |
1145 | if (rt->u.dst.obsolete>0) { | |
1146 | BUG_TRAP(fn==NULL); | |
1147 | return -ENOENT; | |
1148 | } | |
1149 | #endif | |
1150 | if (fn == NULL || rt == &ip6_null_entry) | |
1151 | return -ENOENT; | |
1152 | ||
1153 | BUG_TRAP(fn->fn_flags&RTN_RTINFO); | |
1154 | ||
1155 | if (!(rt->rt6i_flags&RTF_CACHE)) | |
1156 | fib6_prune_clones(fn, rt); | |
1157 | ||
1158 | /* | |
1159 | * Walk the leaf entries looking for ourself | |
1160 | */ | |
1161 | ||
1162 | for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.next) { | |
1163 | if (*rtp == rt) { | |
0d51aa80 | 1164 | fib6_del_route(fn, rtp, nlh, _rtattr, req); |
1da177e4 LT |
1165 | return 0; |
1166 | } | |
1167 | } | |
1168 | return -ENOENT; | |
1169 | } | |
1170 | ||
1171 | /* | |
1172 | * Tree traversal function. | |
1173 | * | |
1174 | * Certainly, it is not interrupt safe. | |
1175 | * However, it is internally reenterable wrt itself and fib6_add/fib6_del. | |
1176 | * It means, that we can modify tree during walking | |
1177 | * and use this function for garbage collection, clone pruning, | |
1178 | * cleaning tree when a device goes down etc. etc. | |
1179 | * | |
1180 | * It guarantees that every node will be traversed, | |
1181 | * and that it will be traversed only once. | |
1182 | * | |
1183 | * Callback function w->func may return: | |
1184 | * 0 -> continue walking. | |
1185 | * positive value -> walking is suspended (used by tree dumps, | |
1186 | * and probably by gc, if it will be split to several slices) | |
1187 | * negative value -> terminate walking. | |
1188 | * | |
1189 | * The function itself returns: | |
1190 | * 0 -> walk is complete. | |
1191 | * >0 -> walk is incomplete (i.e. suspended) | |
1192 | * <0 -> walk is terminated by an error. | |
1193 | */ | |
1194 | ||
90d41122 | 1195 | static int fib6_walk_continue(struct fib6_walker_t *w) |
1da177e4 LT |
1196 | { |
1197 | struct fib6_node *fn, *pn; | |
1198 | ||
1199 | for (;;) { | |
1200 | fn = w->node; | |
1201 | if (fn == NULL) | |
1202 | return 0; | |
1203 | ||
1204 | if (w->prune && fn != w->root && | |
1205 | fn->fn_flags&RTN_RTINFO && w->state < FWS_C) { | |
1206 | w->state = FWS_C; | |
1207 | w->leaf = fn->leaf; | |
1208 | } | |
1209 | switch (w->state) { | |
1210 | #ifdef CONFIG_IPV6_SUBTREES | |
1211 | case FWS_S: | |
1212 | if (SUBTREE(fn)) { | |
1213 | w->node = SUBTREE(fn); | |
1214 | continue; | |
1215 | } | |
1216 | w->state = FWS_L; | |
1217 | #endif | |
1218 | case FWS_L: | |
1219 | if (fn->left) { | |
1220 | w->node = fn->left; | |
1221 | w->state = FWS_INIT; | |
1222 | continue; | |
1223 | } | |
1224 | w->state = FWS_R; | |
1225 | case FWS_R: | |
1226 | if (fn->right) { | |
1227 | w->node = fn->right; | |
1228 | w->state = FWS_INIT; | |
1229 | continue; | |
1230 | } | |
1231 | w->state = FWS_C; | |
1232 | w->leaf = fn->leaf; | |
1233 | case FWS_C: | |
1234 | if (w->leaf && fn->fn_flags&RTN_RTINFO) { | |
1235 | int err = w->func(w); | |
1236 | if (err) | |
1237 | return err; | |
1238 | continue; | |
1239 | } | |
1240 | w->state = FWS_U; | |
1241 | case FWS_U: | |
1242 | if (fn == w->root) | |
1243 | return 0; | |
1244 | pn = fn->parent; | |
1245 | w->node = pn; | |
1246 | #ifdef CONFIG_IPV6_SUBTREES | |
1247 | if (SUBTREE(pn) == fn) { | |
1248 | BUG_TRAP(fn->fn_flags&RTN_ROOT); | |
1249 | w->state = FWS_L; | |
1250 | continue; | |
1251 | } | |
1252 | #endif | |
1253 | if (pn->left == fn) { | |
1254 | w->state = FWS_R; | |
1255 | continue; | |
1256 | } | |
1257 | if (pn->right == fn) { | |
1258 | w->state = FWS_C; | |
1259 | w->leaf = w->node->leaf; | |
1260 | continue; | |
1261 | } | |
1262 | #if RT6_DEBUG >= 2 | |
1263 | BUG_TRAP(0); | |
1264 | #endif | |
1265 | } | |
1266 | } | |
1267 | } | |
1268 | ||
90d41122 | 1269 | static int fib6_walk(struct fib6_walker_t *w) |
1da177e4 LT |
1270 | { |
1271 | int res; | |
1272 | ||
1273 | w->state = FWS_INIT; | |
1274 | w->node = w->root; | |
1275 | ||
1276 | fib6_walker_link(w); | |
1277 | res = fib6_walk_continue(w); | |
1278 | if (res <= 0) | |
1279 | fib6_walker_unlink(w); | |
1280 | return res; | |
1281 | } | |
1282 | ||
1283 | static int fib6_clean_node(struct fib6_walker_t *w) | |
1284 | { | |
1285 | int res; | |
1286 | struct rt6_info *rt; | |
1287 | struct fib6_cleaner_t *c = (struct fib6_cleaner_t*)w; | |
1288 | ||
1289 | for (rt = w->leaf; rt; rt = rt->u.next) { | |
1290 | res = c->func(rt, c->arg); | |
1291 | if (res < 0) { | |
1292 | w->leaf = rt; | |
0d51aa80 | 1293 | res = fib6_del(rt, NULL, NULL, NULL); |
1da177e4 LT |
1294 | if (res) { |
1295 | #if RT6_DEBUG >= 2 | |
1296 | printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res); | |
1297 | #endif | |
1298 | continue; | |
1299 | } | |
1300 | return 0; | |
1301 | } | |
1302 | BUG_TRAP(res==0); | |
1303 | } | |
1304 | w->leaf = rt; | |
1305 | return 0; | |
1306 | } | |
1307 | ||
1308 | /* | |
1309 | * Convenient frontend to tree walker. | |
1310 | * | |
1311 | * func is called on each route. | |
1312 | * It may return -1 -> delete this route. | |
1313 | * 0 -> continue walking | |
1314 | * | |
1315 | * prune==1 -> only immediate children of node (certainly, | |
1316 | * ignoring pure split nodes) will be scanned. | |
1317 | */ | |
1318 | ||
8ce11e6a AB |
1319 | static void fib6_clean_tree(struct fib6_node *root, |
1320 | int (*func)(struct rt6_info *, void *arg), | |
1321 | int prune, void *arg) | |
1da177e4 LT |
1322 | { |
1323 | struct fib6_cleaner_t c; | |
1324 | ||
1325 | c.w.root = root; | |
1326 | c.w.func = fib6_clean_node; | |
1327 | c.w.prune = prune; | |
1328 | c.func = func; | |
1329 | c.arg = arg; | |
1330 | ||
1331 | fib6_walk(&c.w); | |
1332 | } | |
1333 | ||
c71099ac TG |
1334 | void fib6_clean_all(int (*func)(struct rt6_info *, void *arg), |
1335 | int prune, void *arg) | |
1336 | { | |
c71099ac | 1337 | struct fib6_table *table; |
1b43af54 PM |
1338 | struct hlist_node *node; |
1339 | unsigned int h; | |
c71099ac | 1340 | |
1b43af54 PM |
1341 | rcu_read_lock(); |
1342 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { | |
1343 | hlist_for_each_entry_rcu(table, node, &fib_table_hash[h], | |
1344 | tb6_hlist) { | |
c71099ac TG |
1345 | write_lock_bh(&table->tb6_lock); |
1346 | fib6_clean_tree(&table->tb6_root, func, prune, arg); | |
1347 | write_unlock_bh(&table->tb6_lock); | |
1348 | } | |
1349 | } | |
1b43af54 | 1350 | rcu_read_unlock(); |
c71099ac TG |
1351 | } |
1352 | ||
1da177e4 LT |
1353 | static int fib6_prune_clone(struct rt6_info *rt, void *arg) |
1354 | { | |
1355 | if (rt->rt6i_flags & RTF_CACHE) { | |
1356 | RT6_TRACE("pruning clone %p\n", rt); | |
1357 | return -1; | |
1358 | } | |
1359 | ||
1360 | return 0; | |
1361 | } | |
1362 | ||
1363 | static void fib6_prune_clones(struct fib6_node *fn, struct rt6_info *rt) | |
1364 | { | |
1365 | fib6_clean_tree(fn, fib6_prune_clone, 1, rt); | |
1366 | } | |
1367 | ||
1368 | /* | |
1369 | * Garbage collection | |
1370 | */ | |
1371 | ||
1372 | static struct fib6_gc_args | |
1373 | { | |
1374 | int timeout; | |
1375 | int more; | |
1376 | } gc_args; | |
1377 | ||
1378 | static int fib6_age(struct rt6_info *rt, void *arg) | |
1379 | { | |
1380 | unsigned long now = jiffies; | |
1381 | ||
1382 | /* | |
1383 | * check addrconf expiration here. | |
1384 | * Routes are expired even if they are in use. | |
1385 | * | |
1386 | * Also age clones. Note, that clones are aged out | |
1387 | * only if they are not in use now. | |
1388 | */ | |
1389 | ||
1390 | if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) { | |
1391 | if (time_after(now, rt->rt6i_expires)) { | |
1392 | RT6_TRACE("expiring %p\n", rt); | |
1da177e4 LT |
1393 | return -1; |
1394 | } | |
1395 | gc_args.more++; | |
1396 | } else if (rt->rt6i_flags & RTF_CACHE) { | |
1397 | if (atomic_read(&rt->u.dst.__refcnt) == 0 && | |
1398 | time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) { | |
1399 | RT6_TRACE("aging clone %p\n", rt); | |
1400 | return -1; | |
1401 | } else if ((rt->rt6i_flags & RTF_GATEWAY) && | |
1402 | (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) { | |
1403 | RT6_TRACE("purging route %p via non-router but gateway\n", | |
1404 | rt); | |
1405 | return -1; | |
1406 | } | |
1407 | gc_args.more++; | |
1408 | } | |
1409 | ||
1410 | return 0; | |
1411 | } | |
1412 | ||
1413 | static DEFINE_SPINLOCK(fib6_gc_lock); | |
1414 | ||
1415 | void fib6_run_gc(unsigned long dummy) | |
1416 | { | |
1417 | if (dummy != ~0UL) { | |
1418 | spin_lock_bh(&fib6_gc_lock); | |
1419 | gc_args.timeout = dummy ? (int)dummy : ip6_rt_gc_interval; | |
1420 | } else { | |
1421 | local_bh_disable(); | |
1422 | if (!spin_trylock(&fib6_gc_lock)) { | |
1423 | mod_timer(&ip6_fib_timer, jiffies + HZ); | |
1424 | local_bh_enable(); | |
1425 | return; | |
1426 | } | |
1427 | gc_args.timeout = ip6_rt_gc_interval; | |
1428 | } | |
1429 | gc_args.more = 0; | |
1430 | ||
1da177e4 | 1431 | ndisc_dst_gc(&gc_args.more); |
c71099ac | 1432 | fib6_clean_all(fib6_age, 0, NULL); |
1da177e4 LT |
1433 | |
1434 | if (gc_args.more) | |
1435 | mod_timer(&ip6_fib_timer, jiffies + ip6_rt_gc_interval); | |
1436 | else { | |
1437 | del_timer(&ip6_fib_timer); | |
1438 | ip6_fib_timer.expires = 0; | |
1439 | } | |
1440 | spin_unlock_bh(&fib6_gc_lock); | |
1441 | } | |
1442 | ||
1443 | void __init fib6_init(void) | |
1444 | { | |
1445 | fib6_node_kmem = kmem_cache_create("fib6_nodes", | |
1446 | sizeof(struct fib6_node), | |
1447 | 0, SLAB_HWCACHE_ALIGN, | |
1448 | NULL, NULL); | |
1449 | if (!fib6_node_kmem) | |
1450 | panic("cannot create fib6_nodes cache"); | |
c71099ac TG |
1451 | |
1452 | fib6_tables_init(); | |
1da177e4 LT |
1453 | } |
1454 | ||
1455 | void fib6_gc_cleanup(void) | |
1456 | { | |
1457 | del_timer(&ip6_fib_timer); | |
1458 | kmem_cache_destroy(fib6_node_kmem); | |
1459 | } |