]>
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 | 7 | * |
1da177e4 LT |
8 | * This program is free software; you can redistribute it and/or |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
12 | */ | |
13 | ||
14 | /* | |
15 | * Changes: | |
16 | * Yuji SEKIYA @USAGI: Support default route on router node; | |
17 | * remove ip6_null_entry from the top of | |
18 | * routing table. | |
c0bece9f | 19 | * Ville Nuorvala: Fixed routing subtrees. |
1da177e4 | 20 | */ |
1da177e4 LT |
21 | #include <linux/errno.h> |
22 | #include <linux/types.h> | |
23 | #include <linux/net.h> | |
24 | #include <linux/route.h> | |
25 | #include <linux/netdevice.h> | |
26 | #include <linux/in6.h> | |
27 | #include <linux/init.h> | |
c71099ac | 28 | #include <linux/list.h> |
1da177e4 LT |
29 | |
30 | #ifdef CONFIG_PROC_FS | |
31 | #include <linux/proc_fs.h> | |
32 | #endif | |
33 | ||
34 | #include <net/ipv6.h> | |
35 | #include <net/ndisc.h> | |
36 | #include <net/addrconf.h> | |
37 | ||
38 | #include <net/ip6_fib.h> | |
39 | #include <net/ip6_route.h> | |
40 | ||
41 | #define RT6_DEBUG 2 | |
42 | ||
43 | #if RT6_DEBUG >= 3 | |
44 | #define RT6_TRACE(x...) printk(KERN_DEBUG x) | |
45 | #else | |
46 | #define RT6_TRACE(x...) do { ; } while (0) | |
47 | #endif | |
48 | ||
e18b890b | 49 | static struct kmem_cache * fib6_node_kmem __read_mostly; |
1da177e4 LT |
50 | |
51 | enum fib_walk_state_t | |
52 | { | |
53 | #ifdef CONFIG_IPV6_SUBTREES | |
54 | FWS_S, | |
55 | #endif | |
56 | FWS_L, | |
57 | FWS_R, | |
58 | FWS_C, | |
59 | FWS_U | |
60 | }; | |
61 | ||
62 | struct fib6_cleaner_t | |
63 | { | |
64 | struct fib6_walker_t w; | |
ec7d43c2 | 65 | struct net *net; |
1da177e4 LT |
66 | int (*func)(struct rt6_info *, void *arg); |
67 | void *arg; | |
68 | }; | |
69 | ||
90d41122 | 70 | static DEFINE_RWLOCK(fib6_walker_lock); |
1da177e4 LT |
71 | |
72 | #ifdef CONFIG_IPV6_SUBTREES | |
73 | #define FWS_INIT FWS_S | |
1da177e4 LT |
74 | #else |
75 | #define FWS_INIT FWS_L | |
1da177e4 LT |
76 | #endif |
77 | ||
ec7d43c2 BT |
78 | static void fib6_prune_clones(struct net *net, struct fib6_node *fn, |
79 | struct rt6_info *rt); | |
8ed67789 DL |
80 | static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn); |
81 | static struct fib6_node *fib6_repair_tree(struct net *net, struct fib6_node *fn); | |
90d41122 AB |
82 | static int fib6_walk(struct fib6_walker_t *w); |
83 | static int fib6_walk_continue(struct fib6_walker_t *w); | |
1da177e4 LT |
84 | |
85 | /* | |
86 | * A routing update causes an increase of the serial number on the | |
87 | * affected subtree. This allows for cached routes to be asynchronously | |
88 | * tested when modifications are made to the destination cache as a | |
89 | * result of redirects, path MTU changes, etc. | |
90 | */ | |
91 | ||
92 | static __u32 rt_sernum; | |
93 | ||
5b7c931d DL |
94 | static void fib6_gc_timer_cb(unsigned long arg); |
95 | ||
90d41122 | 96 | static struct fib6_walker_t fib6_walker_list = { |
1da177e4 | 97 | .prev = &fib6_walker_list, |
1ab1457c | 98 | .next = &fib6_walker_list, |
1da177e4 LT |
99 | }; |
100 | ||
101 | #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next) | |
102 | ||
90d41122 AB |
103 | static inline void fib6_walker_link(struct fib6_walker_t *w) |
104 | { | |
105 | write_lock_bh(&fib6_walker_lock); | |
106 | w->next = fib6_walker_list.next; | |
107 | w->prev = &fib6_walker_list; | |
108 | w->next->prev = w; | |
109 | w->prev->next = w; | |
110 | write_unlock_bh(&fib6_walker_lock); | |
111 | } | |
112 | ||
113 | static inline void fib6_walker_unlink(struct fib6_walker_t *w) | |
114 | { | |
115 | write_lock_bh(&fib6_walker_lock); | |
116 | w->next->prev = w->prev; | |
117 | w->prev->next = w->next; | |
118 | w->prev = w->next = w; | |
119 | write_unlock_bh(&fib6_walker_lock); | |
120 | } | |
1da177e4 LT |
121 | static __inline__ u32 fib6_new_sernum(void) |
122 | { | |
123 | u32 n = ++rt_sernum; | |
124 | if ((__s32)n <= 0) | |
125 | rt_sernum = n = 1; | |
126 | return n; | |
127 | } | |
128 | ||
129 | /* | |
130 | * Auxiliary address test functions for the radix tree. | |
131 | * | |
1ab1457c | 132 | * These assume a 32bit processor (although it will work on |
1da177e4 LT |
133 | * 64bit processors) |
134 | */ | |
135 | ||
136 | /* | |
137 | * test bit | |
138 | */ | |
139 | ||
e69a4adc | 140 | static __inline__ __be32 addr_bit_set(void *token, int fn_bit) |
1da177e4 | 141 | { |
e69a4adc | 142 | __be32 *addr = token; |
1da177e4 LT |
143 | |
144 | return htonl(1 << ((~fn_bit)&0x1F)) & addr[fn_bit>>5]; | |
145 | } | |
146 | ||
1da177e4 LT |
147 | static __inline__ struct fib6_node * node_alloc(void) |
148 | { | |
149 | struct fib6_node *fn; | |
150 | ||
c3762229 | 151 | fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC); |
1da177e4 LT |
152 | |
153 | return fn; | |
154 | } | |
155 | ||
156 | static __inline__ void node_free(struct fib6_node * fn) | |
157 | { | |
158 | kmem_cache_free(fib6_node_kmem, fn); | |
159 | } | |
160 | ||
161 | static __inline__ void rt6_release(struct rt6_info *rt) | |
162 | { | |
163 | if (atomic_dec_and_test(&rt->rt6i_ref)) | |
164 | dst_free(&rt->u.dst); | |
165 | } | |
166 | ||
c71099ac | 167 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
1b43af54 PM |
168 | #define FIB_TABLE_HASHSZ 256 |
169 | #else | |
170 | #define FIB_TABLE_HASHSZ 1 | |
171 | #endif | |
1b43af54 | 172 | |
58f09b78 | 173 | static void fib6_link_table(struct net *net, struct fib6_table *tb) |
1b43af54 PM |
174 | { |
175 | unsigned int h; | |
176 | ||
375216ad TG |
177 | /* |
178 | * Initialize table lock at a single place to give lockdep a key, | |
179 | * tables aren't visible prior to being linked to the list. | |
180 | */ | |
181 | rwlock_init(&tb->tb6_lock); | |
182 | ||
1b43af54 PM |
183 | h = tb->tb6_id & (FIB_TABLE_HASHSZ - 1); |
184 | ||
185 | /* | |
186 | * No protection necessary, this is the only list mutatation | |
187 | * operation, tables never disappear once they exist. | |
188 | */ | |
58f09b78 | 189 | hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]); |
1b43af54 | 190 | } |
c71099ac | 191 | |
1b43af54 | 192 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 193 | |
8ed67789 | 194 | static struct fib6_table *fib6_alloc_table(struct net *net, u32 id) |
c71099ac TG |
195 | { |
196 | struct fib6_table *table; | |
197 | ||
198 | table = kzalloc(sizeof(*table), GFP_ATOMIC); | |
199 | if (table != NULL) { | |
200 | table->tb6_id = id; | |
8ed67789 | 201 | table->tb6_root.leaf = net->ipv6.ip6_null_entry; |
c71099ac TG |
202 | table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; |
203 | } | |
204 | ||
205 | return table; | |
206 | } | |
207 | ||
58f09b78 | 208 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac TG |
209 | { |
210 | struct fib6_table *tb; | |
211 | ||
212 | if (id == 0) | |
213 | id = RT6_TABLE_MAIN; | |
58f09b78 | 214 | tb = fib6_get_table(net, id); |
c71099ac TG |
215 | if (tb) |
216 | return tb; | |
217 | ||
8ed67789 | 218 | tb = fib6_alloc_table(net, id); |
c71099ac | 219 | if (tb != NULL) |
58f09b78 | 220 | fib6_link_table(net, tb); |
c71099ac TG |
221 | |
222 | return tb; | |
223 | } | |
224 | ||
58f09b78 | 225 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac TG |
226 | { |
227 | struct fib6_table *tb; | |
58f09b78 | 228 | struct hlist_head *head; |
c71099ac TG |
229 | struct hlist_node *node; |
230 | unsigned int h; | |
231 | ||
232 | if (id == 0) | |
233 | id = RT6_TABLE_MAIN; | |
234 | h = id & (FIB_TABLE_HASHSZ - 1); | |
235 | rcu_read_lock(); | |
58f09b78 DL |
236 | head = &net->ipv6.fib_table_hash[h]; |
237 | hlist_for_each_entry_rcu(tb, node, head, tb6_hlist) { | |
c71099ac TG |
238 | if (tb->tb6_id == id) { |
239 | rcu_read_unlock(); | |
240 | return tb; | |
241 | } | |
242 | } | |
243 | rcu_read_unlock(); | |
244 | ||
245 | return NULL; | |
246 | } | |
247 | ||
58f09b78 | 248 | static void fib6_tables_init(struct net *net) |
c71099ac | 249 | { |
58f09b78 DL |
250 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
251 | fib6_link_table(net, net->ipv6.fib6_local_tbl); | |
c71099ac | 252 | } |
c71099ac TG |
253 | #else |
254 | ||
58f09b78 | 255 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac | 256 | { |
58f09b78 | 257 | return fib6_get_table(net, id); |
c71099ac TG |
258 | } |
259 | ||
58f09b78 | 260 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac | 261 | { |
58f09b78 | 262 | return net->ipv6.fib6_main_tbl; |
c71099ac TG |
263 | } |
264 | ||
58f09b78 DL |
265 | struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi *fl, |
266 | int flags, pol_lookup_t lookup) | |
c71099ac | 267 | { |
8ed67789 | 268 | return (struct dst_entry *) lookup(net, net->ipv6.fib6_main_tbl, fl, flags); |
c71099ac TG |
269 | } |
270 | ||
58f09b78 | 271 | static void fib6_tables_init(struct net *net) |
c71099ac | 272 | { |
58f09b78 | 273 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
c71099ac TG |
274 | } |
275 | ||
276 | #endif | |
277 | ||
1b43af54 PM |
278 | static int fib6_dump_node(struct fib6_walker_t *w) |
279 | { | |
280 | int res; | |
281 | struct rt6_info *rt; | |
282 | ||
7cc48263 | 283 | for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) { |
1b43af54 PM |
284 | res = rt6_dump_route(rt, w->args); |
285 | if (res < 0) { | |
286 | /* Frame is full, suspend walking */ | |
287 | w->leaf = rt; | |
288 | return 1; | |
289 | } | |
290 | BUG_TRAP(res!=0); | |
291 | } | |
292 | w->leaf = NULL; | |
293 | return 0; | |
294 | } | |
295 | ||
296 | static void fib6_dump_end(struct netlink_callback *cb) | |
297 | { | |
298 | struct fib6_walker_t *w = (void*)cb->args[2]; | |
299 | ||
300 | if (w) { | |
301 | cb->args[2] = 0; | |
302 | kfree(w); | |
303 | } | |
304 | cb->done = (void*)cb->args[3]; | |
305 | cb->args[1] = 3; | |
306 | } | |
307 | ||
308 | static int fib6_dump_done(struct netlink_callback *cb) | |
309 | { | |
310 | fib6_dump_end(cb); | |
311 | return cb->done ? cb->done(cb) : 0; | |
312 | } | |
313 | ||
314 | static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb, | |
315 | struct netlink_callback *cb) | |
316 | { | |
317 | struct fib6_walker_t *w; | |
318 | int res; | |
319 | ||
320 | w = (void *)cb->args[2]; | |
321 | w->root = &table->tb6_root; | |
322 | ||
323 | if (cb->args[4] == 0) { | |
324 | read_lock_bh(&table->tb6_lock); | |
325 | res = fib6_walk(w); | |
326 | read_unlock_bh(&table->tb6_lock); | |
327 | if (res > 0) | |
328 | cb->args[4] = 1; | |
329 | } else { | |
330 | read_lock_bh(&table->tb6_lock); | |
331 | res = fib6_walk_continue(w); | |
332 | read_unlock_bh(&table->tb6_lock); | |
333 | if (res != 0) { | |
334 | if (res < 0) | |
335 | fib6_walker_unlink(w); | |
336 | goto end; | |
337 | } | |
338 | fib6_walker_unlink(w); | |
339 | cb->args[4] = 0; | |
340 | } | |
341 | end: | |
342 | return res; | |
343 | } | |
344 | ||
c127ea2c | 345 | static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) |
1b43af54 | 346 | { |
3b1e0a65 | 347 | struct net *net = sock_net(skb->sk); |
1b43af54 PM |
348 | unsigned int h, s_h; |
349 | unsigned int e = 0, s_e; | |
350 | struct rt6_rtnl_dump_arg arg; | |
351 | struct fib6_walker_t *w; | |
352 | struct fib6_table *tb; | |
353 | struct hlist_node *node; | |
58f09b78 | 354 | struct hlist_head *head; |
1b43af54 PM |
355 | int res = 0; |
356 | ||
357 | s_h = cb->args[0]; | |
358 | s_e = cb->args[1]; | |
359 | ||
360 | w = (void *)cb->args[2]; | |
361 | if (w == NULL) { | |
362 | /* New dump: | |
363 | * | |
364 | * 1. hook callback destructor. | |
365 | */ | |
366 | cb->args[3] = (long)cb->done; | |
367 | cb->done = fib6_dump_done; | |
368 | ||
369 | /* | |
370 | * 2. allocate and initialize walker. | |
371 | */ | |
372 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
373 | if (w == NULL) | |
374 | return -ENOMEM; | |
375 | w->func = fib6_dump_node; | |
376 | cb->args[2] = (long)w; | |
377 | } | |
378 | ||
379 | arg.skb = skb; | |
380 | arg.cb = cb; | |
381 | w->args = &arg; | |
382 | ||
383 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { | |
384 | e = 0; | |
58f09b78 DL |
385 | head = &net->ipv6.fib_table_hash[h]; |
386 | hlist_for_each_entry(tb, node, head, tb6_hlist) { | |
1b43af54 PM |
387 | if (e < s_e) |
388 | goto next; | |
389 | res = fib6_dump_table(tb, skb, cb); | |
390 | if (res != 0) | |
391 | goto out; | |
392 | next: | |
393 | e++; | |
394 | } | |
395 | } | |
396 | out: | |
397 | cb->args[1] = e; | |
398 | cb->args[0] = h; | |
399 | ||
400 | res = res < 0 ? res : skb->len; | |
401 | if (res <= 0) | |
402 | fib6_dump_end(cb); | |
403 | return res; | |
404 | } | |
1da177e4 LT |
405 | |
406 | /* | |
407 | * Routing Table | |
408 | * | |
409 | * return the appropriate node for a routing tree "add" operation | |
410 | * by either creating and inserting or by returning an existing | |
411 | * node. | |
412 | */ | |
413 | ||
414 | static struct fib6_node * fib6_add_1(struct fib6_node *root, void *addr, | |
415 | int addrlen, int plen, | |
416 | int offset) | |
417 | { | |
418 | struct fib6_node *fn, *in, *ln; | |
419 | struct fib6_node *pn = NULL; | |
420 | struct rt6key *key; | |
421 | int bit; | |
1ab1457c | 422 | __be32 dir = 0; |
1da177e4 LT |
423 | __u32 sernum = fib6_new_sernum(); |
424 | ||
425 | RT6_TRACE("fib6_add_1\n"); | |
426 | ||
427 | /* insert node in tree */ | |
428 | ||
429 | fn = root; | |
430 | ||
431 | do { | |
432 | key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
433 | ||
434 | /* | |
435 | * Prefix match | |
436 | */ | |
437 | if (plen < fn->fn_bit || | |
438 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
439 | goto insert_above; | |
1ab1457c | 440 | |
1da177e4 LT |
441 | /* |
442 | * Exact match ? | |
443 | */ | |
1ab1457c | 444 | |
1da177e4 LT |
445 | if (plen == fn->fn_bit) { |
446 | /* clean up an intermediate node */ | |
447 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
448 | rt6_release(fn->leaf); | |
449 | fn->leaf = NULL; | |
450 | } | |
1ab1457c | 451 | |
1da177e4 | 452 | fn->fn_sernum = sernum; |
1ab1457c | 453 | |
1da177e4 LT |
454 | return fn; |
455 | } | |
456 | ||
457 | /* | |
458 | * We have more bits to go | |
459 | */ | |
1ab1457c | 460 | |
1da177e4 LT |
461 | /* Try to walk down on tree. */ |
462 | fn->fn_sernum = sernum; | |
463 | dir = addr_bit_set(addr, fn->fn_bit); | |
464 | pn = fn; | |
465 | fn = dir ? fn->right: fn->left; | |
466 | } while (fn); | |
467 | ||
468 | /* | |
469 | * We walked to the bottom of tree. | |
470 | * Create new leaf node without children. | |
471 | */ | |
472 | ||
473 | ln = node_alloc(); | |
474 | ||
475 | if (ln == NULL) | |
476 | return NULL; | |
477 | ln->fn_bit = plen; | |
1ab1457c | 478 | |
1da177e4 LT |
479 | ln->parent = pn; |
480 | ln->fn_sernum = sernum; | |
481 | ||
482 | if (dir) | |
483 | pn->right = ln; | |
484 | else | |
485 | pn->left = ln; | |
486 | ||
487 | return ln; | |
488 | ||
489 | ||
490 | insert_above: | |
491 | /* | |
1ab1457c | 492 | * split since we don't have a common prefix anymore or |
1da177e4 LT |
493 | * we have a less significant route. |
494 | * we've to insert an intermediate node on the list | |
495 | * this new node will point to the one we need to create | |
496 | * and the current | |
497 | */ | |
498 | ||
499 | pn = fn->parent; | |
500 | ||
501 | /* find 1st bit in difference between the 2 addrs. | |
502 | ||
971f359d | 503 | See comment in __ipv6_addr_diff: bit may be an invalid value, |
1da177e4 LT |
504 | but if it is >= plen, the value is ignored in any case. |
505 | */ | |
1ab1457c | 506 | |
971f359d | 507 | bit = __ipv6_addr_diff(addr, &key->addr, addrlen); |
1da177e4 | 508 | |
1ab1457c YH |
509 | /* |
510 | * (intermediate)[in] | |
1da177e4 LT |
511 | * / \ |
512 | * (new leaf node)[ln] (old node)[fn] | |
513 | */ | |
514 | if (plen > bit) { | |
515 | in = node_alloc(); | |
516 | ln = node_alloc(); | |
1ab1457c | 517 | |
1da177e4 LT |
518 | if (in == NULL || ln == NULL) { |
519 | if (in) | |
520 | node_free(in); | |
521 | if (ln) | |
522 | node_free(ln); | |
523 | return NULL; | |
524 | } | |
525 | ||
1ab1457c YH |
526 | /* |
527 | * new intermediate node. | |
1da177e4 LT |
528 | * RTN_RTINFO will |
529 | * be off since that an address that chooses one of | |
530 | * the branches would not match less specific routes | |
531 | * in the other branch | |
532 | */ | |
533 | ||
534 | in->fn_bit = bit; | |
535 | ||
536 | in->parent = pn; | |
537 | in->leaf = fn->leaf; | |
538 | atomic_inc(&in->leaf->rt6i_ref); | |
539 | ||
540 | in->fn_sernum = sernum; | |
541 | ||
542 | /* update parent pointer */ | |
543 | if (dir) | |
544 | pn->right = in; | |
545 | else | |
546 | pn->left = in; | |
547 | ||
548 | ln->fn_bit = plen; | |
549 | ||
550 | ln->parent = in; | |
551 | fn->parent = in; | |
552 | ||
553 | ln->fn_sernum = sernum; | |
554 | ||
555 | if (addr_bit_set(addr, bit)) { | |
556 | in->right = ln; | |
557 | in->left = fn; | |
558 | } else { | |
559 | in->left = ln; | |
560 | in->right = fn; | |
561 | } | |
562 | } else { /* plen <= bit */ | |
563 | ||
1ab1457c | 564 | /* |
1da177e4 LT |
565 | * (new leaf node)[ln] |
566 | * / \ | |
567 | * (old node)[fn] NULL | |
568 | */ | |
569 | ||
570 | ln = node_alloc(); | |
571 | ||
572 | if (ln == NULL) | |
573 | return NULL; | |
574 | ||
575 | ln->fn_bit = plen; | |
576 | ||
577 | ln->parent = pn; | |
578 | ||
579 | ln->fn_sernum = sernum; | |
1ab1457c | 580 | |
1da177e4 LT |
581 | if (dir) |
582 | pn->right = ln; | |
583 | else | |
584 | pn->left = ln; | |
585 | ||
586 | if (addr_bit_set(&key->addr, plen)) | |
587 | ln->right = fn; | |
588 | else | |
589 | ln->left = fn; | |
590 | ||
591 | fn->parent = ln; | |
592 | } | |
593 | return ln; | |
594 | } | |
595 | ||
596 | /* | |
597 | * Insert routing information in a node. | |
598 | */ | |
599 | ||
600 | static int fib6_add_rt2node(struct fib6_node *fn, struct rt6_info *rt, | |
86872cb5 | 601 | struct nl_info *info) |
1da177e4 LT |
602 | { |
603 | struct rt6_info *iter = NULL; | |
604 | struct rt6_info **ins; | |
605 | ||
606 | ins = &fn->leaf; | |
607 | ||
7cc48263 | 608 | for (iter = fn->leaf; iter; iter=iter->u.dst.rt6_next) { |
1da177e4 LT |
609 | /* |
610 | * Search for duplicates | |
611 | */ | |
612 | ||
613 | if (iter->rt6i_metric == rt->rt6i_metric) { | |
614 | /* | |
615 | * Same priority level | |
616 | */ | |
617 | ||
618 | if (iter->rt6i_dev == rt->rt6i_dev && | |
619 | iter->rt6i_idev == rt->rt6i_idev && | |
620 | ipv6_addr_equal(&iter->rt6i_gateway, | |
621 | &rt->rt6i_gateway)) { | |
622 | if (!(iter->rt6i_flags&RTF_EXPIRES)) | |
623 | return -EEXIST; | |
624 | iter->rt6i_expires = rt->rt6i_expires; | |
625 | if (!(rt->rt6i_flags&RTF_EXPIRES)) { | |
626 | iter->rt6i_flags &= ~RTF_EXPIRES; | |
627 | iter->rt6i_expires = 0; | |
628 | } | |
629 | return -EEXIST; | |
630 | } | |
631 | } | |
632 | ||
633 | if (iter->rt6i_metric > rt->rt6i_metric) | |
634 | break; | |
635 | ||
7cc48263 | 636 | ins = &iter->u.dst.rt6_next; |
1da177e4 LT |
637 | } |
638 | ||
f11e6659 DM |
639 | /* Reset round-robin state, if necessary */ |
640 | if (ins == &fn->leaf) | |
641 | fn->rr_ptr = NULL; | |
642 | ||
1da177e4 LT |
643 | /* |
644 | * insert node | |
645 | */ | |
646 | ||
7cc48263 | 647 | rt->u.dst.rt6_next = iter; |
1da177e4 LT |
648 | *ins = rt; |
649 | rt->rt6i_node = fn; | |
650 | atomic_inc(&rt->rt6i_ref); | |
86872cb5 | 651 | inet6_rt_notify(RTM_NEWROUTE, rt, info); |
c572872f | 652 | info->nl_net->ipv6.rt6_stats->fib_rt_entries++; |
1da177e4 LT |
653 | |
654 | if ((fn->fn_flags & RTN_RTINFO) == 0) { | |
c572872f | 655 | info->nl_net->ipv6.rt6_stats->fib_route_nodes++; |
1da177e4 LT |
656 | fn->fn_flags |= RTN_RTINFO; |
657 | } | |
658 | ||
659 | return 0; | |
660 | } | |
661 | ||
63152fc0 | 662 | static __inline__ void fib6_start_gc(struct net *net, struct rt6_info *rt) |
1da177e4 | 663 | { |
417f28bb | 664 | if (!timer_pending(&net->ipv6.ip6_fib_timer) && |
1da177e4 | 665 | (rt->rt6i_flags & (RTF_EXPIRES|RTF_CACHE))) |
417f28bb | 666 | mod_timer(&net->ipv6.ip6_fib_timer, |
847499ce | 667 | jiffies + net->ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 LT |
668 | } |
669 | ||
63152fc0 | 670 | void fib6_force_start_gc(struct net *net) |
1da177e4 | 671 | { |
417f28bb SH |
672 | if (!timer_pending(&net->ipv6.ip6_fib_timer)) |
673 | mod_timer(&net->ipv6.ip6_fib_timer, | |
847499ce | 674 | jiffies + net->ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 LT |
675 | } |
676 | ||
677 | /* | |
678 | * Add routing information to the routing tree. | |
679 | * <destination addr>/<source addr> | |
680 | * with source addr info in sub-trees | |
681 | */ | |
682 | ||
86872cb5 | 683 | int fib6_add(struct fib6_node *root, struct rt6_info *rt, struct nl_info *info) |
1da177e4 | 684 | { |
66729e18 | 685 | struct fib6_node *fn, *pn = NULL; |
1da177e4 LT |
686 | int err = -ENOMEM; |
687 | ||
688 | fn = fib6_add_1(root, &rt->rt6i_dst.addr, sizeof(struct in6_addr), | |
689 | rt->rt6i_dst.plen, offsetof(struct rt6_info, rt6i_dst)); | |
690 | ||
691 | if (fn == NULL) | |
692 | goto out; | |
693 | ||
66729e18 YH |
694 | pn = fn; |
695 | ||
1da177e4 LT |
696 | #ifdef CONFIG_IPV6_SUBTREES |
697 | if (rt->rt6i_src.plen) { | |
698 | struct fib6_node *sn; | |
699 | ||
700 | if (fn->subtree == NULL) { | |
701 | struct fib6_node *sfn; | |
702 | ||
703 | /* | |
704 | * Create subtree. | |
705 | * | |
706 | * fn[main tree] | |
707 | * | | |
708 | * sfn[subtree root] | |
709 | * \ | |
710 | * sn[new leaf node] | |
711 | */ | |
712 | ||
713 | /* Create subtree root node */ | |
714 | sfn = node_alloc(); | |
715 | if (sfn == NULL) | |
716 | goto st_failure; | |
717 | ||
8ed67789 DL |
718 | sfn->leaf = info->nl_net->ipv6.ip6_null_entry; |
719 | atomic_inc(&info->nl_net->ipv6.ip6_null_entry->rt6i_ref); | |
1da177e4 LT |
720 | sfn->fn_flags = RTN_ROOT; |
721 | sfn->fn_sernum = fib6_new_sernum(); | |
722 | ||
723 | /* Now add the first leaf node to new subtree */ | |
724 | ||
725 | sn = fib6_add_1(sfn, &rt->rt6i_src.addr, | |
726 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
727 | offsetof(struct rt6_info, rt6i_src)); | |
728 | ||
729 | if (sn == NULL) { | |
730 | /* If it is failed, discard just allocated | |
731 | root, and then (in st_failure) stale node | |
732 | in main tree. | |
733 | */ | |
734 | node_free(sfn); | |
735 | goto st_failure; | |
736 | } | |
737 | ||
738 | /* Now link new subtree to main tree */ | |
739 | sfn->parent = fn; | |
740 | fn->subtree = sfn; | |
1da177e4 LT |
741 | } else { |
742 | sn = fib6_add_1(fn->subtree, &rt->rt6i_src.addr, | |
743 | sizeof(struct in6_addr), rt->rt6i_src.plen, | |
744 | offsetof(struct rt6_info, rt6i_src)); | |
745 | ||
746 | if (sn == NULL) | |
747 | goto st_failure; | |
748 | } | |
749 | ||
66729e18 YH |
750 | if (fn->leaf == NULL) { |
751 | fn->leaf = rt; | |
752 | atomic_inc(&rt->rt6i_ref); | |
753 | } | |
1da177e4 LT |
754 | fn = sn; |
755 | } | |
756 | #endif | |
757 | ||
86872cb5 | 758 | err = fib6_add_rt2node(fn, rt, info); |
1da177e4 LT |
759 | |
760 | if (err == 0) { | |
63152fc0 | 761 | fib6_start_gc(info->nl_net, rt); |
1da177e4 | 762 | if (!(rt->rt6i_flags&RTF_CACHE)) |
ec7d43c2 | 763 | fib6_prune_clones(info->nl_net, pn, rt); |
1da177e4 LT |
764 | } |
765 | ||
766 | out: | |
66729e18 YH |
767 | if (err) { |
768 | #ifdef CONFIG_IPV6_SUBTREES | |
769 | /* | |
770 | * If fib6_add_1 has cleared the old leaf pointer in the | |
771 | * super-tree leaf node we have to find a new one for it. | |
772 | */ | |
3c051235 DM |
773 | if (pn != fn && pn->leaf == rt) { |
774 | pn->leaf = NULL; | |
775 | atomic_dec(&rt->rt6i_ref); | |
776 | } | |
66729e18 | 777 | if (pn != fn && !pn->leaf && !(pn->fn_flags & RTN_RTINFO)) { |
8ed67789 | 778 | pn->leaf = fib6_find_prefix(info->nl_net, pn); |
66729e18 YH |
779 | #if RT6_DEBUG >= 2 |
780 | if (!pn->leaf) { | |
781 | BUG_TRAP(pn->leaf != NULL); | |
8ed67789 | 782 | pn->leaf = info->nl_net->ipv6.ip6_null_entry; |
66729e18 YH |
783 | } |
784 | #endif | |
785 | atomic_inc(&pn->leaf->rt6i_ref); | |
786 | } | |
787 | #endif | |
1da177e4 | 788 | dst_free(&rt->u.dst); |
66729e18 | 789 | } |
1da177e4 LT |
790 | return err; |
791 | ||
792 | #ifdef CONFIG_IPV6_SUBTREES | |
793 | /* Subtree creation failed, probably main tree node | |
794 | is orphan. If it is, shoot it. | |
795 | */ | |
796 | st_failure: | |
797 | if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT))) | |
8ed67789 | 798 | fib6_repair_tree(info->nl_net, fn); |
1da177e4 LT |
799 | dst_free(&rt->u.dst); |
800 | return err; | |
801 | #endif | |
802 | } | |
803 | ||
804 | /* | |
805 | * Routing tree lookup | |
806 | * | |
807 | */ | |
808 | ||
809 | struct lookup_args { | |
810 | int offset; /* key offset on rt6_info */ | |
811 | struct in6_addr *addr; /* search key */ | |
812 | }; | |
813 | ||
814 | static struct fib6_node * fib6_lookup_1(struct fib6_node *root, | |
815 | struct lookup_args *args) | |
816 | { | |
817 | struct fib6_node *fn; | |
e69a4adc | 818 | __be32 dir; |
1da177e4 | 819 | |
825e288e YH |
820 | if (unlikely(args->offset == 0)) |
821 | return NULL; | |
822 | ||
1da177e4 LT |
823 | /* |
824 | * Descend on a tree | |
825 | */ | |
826 | ||
827 | fn = root; | |
828 | ||
829 | for (;;) { | |
830 | struct fib6_node *next; | |
831 | ||
832 | dir = addr_bit_set(args->addr, fn->fn_bit); | |
833 | ||
834 | next = dir ? fn->right : fn->left; | |
835 | ||
836 | if (next) { | |
837 | fn = next; | |
838 | continue; | |
839 | } | |
840 | ||
841 | break; | |
842 | } | |
843 | ||
3fc5e044 | 844 | while(fn) { |
7fc33165 | 845 | if (FIB6_SUBTREE(fn) || fn->fn_flags & RTN_RTINFO) { |
1da177e4 LT |
846 | struct rt6key *key; |
847 | ||
848 | key = (struct rt6key *) ((u8 *) fn->leaf + | |
849 | args->offset); | |
850 | ||
3fc5e044 YH |
851 | if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) { |
852 | #ifdef CONFIG_IPV6_SUBTREES | |
853 | if (fn->subtree) | |
854 | fn = fib6_lookup_1(fn->subtree, args + 1); | |
855 | #endif | |
856 | if (!fn || fn->fn_flags & RTN_RTINFO) | |
857 | return fn; | |
858 | } | |
1da177e4 LT |
859 | } |
860 | ||
3fc5e044 YH |
861 | if (fn->fn_flags & RTN_ROOT) |
862 | break; | |
863 | ||
1da177e4 LT |
864 | fn = fn->parent; |
865 | } | |
866 | ||
867 | return NULL; | |
868 | } | |
869 | ||
870 | struct fib6_node * fib6_lookup(struct fib6_node *root, struct in6_addr *daddr, | |
871 | struct in6_addr *saddr) | |
872 | { | |
1da177e4 | 873 | struct fib6_node *fn; |
825e288e YH |
874 | struct lookup_args args[] = { |
875 | { | |
876 | .offset = offsetof(struct rt6_info, rt6i_dst), | |
877 | .addr = daddr, | |
878 | }, | |
1da177e4 | 879 | #ifdef CONFIG_IPV6_SUBTREES |
825e288e YH |
880 | { |
881 | .offset = offsetof(struct rt6_info, rt6i_src), | |
882 | .addr = saddr, | |
883 | }, | |
1da177e4 | 884 | #endif |
825e288e YH |
885 | { |
886 | .offset = 0, /* sentinel */ | |
887 | } | |
888 | }; | |
1da177e4 | 889 | |
fefc2a6c | 890 | fn = fib6_lookup_1(root, daddr ? args : args + 1); |
1da177e4 LT |
891 | |
892 | if (fn == NULL || fn->fn_flags & RTN_TL_ROOT) | |
893 | fn = root; | |
894 | ||
895 | return fn; | |
896 | } | |
897 | ||
898 | /* | |
899 | * Get node with specified destination prefix (and source prefix, | |
900 | * if subtrees are used) | |
901 | */ | |
902 | ||
903 | ||
904 | static struct fib6_node * fib6_locate_1(struct fib6_node *root, | |
905 | struct in6_addr *addr, | |
906 | int plen, int offset) | |
907 | { | |
908 | struct fib6_node *fn; | |
909 | ||
910 | for (fn = root; fn ; ) { | |
911 | struct rt6key *key = (struct rt6key *)((u8 *)fn->leaf + offset); | |
912 | ||
913 | /* | |
914 | * Prefix match | |
915 | */ | |
916 | if (plen < fn->fn_bit || | |
917 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
918 | return NULL; | |
919 | ||
920 | if (plen == fn->fn_bit) | |
921 | return fn; | |
922 | ||
923 | /* | |
924 | * We have more bits to go | |
925 | */ | |
926 | if (addr_bit_set(addr, fn->fn_bit)) | |
927 | fn = fn->right; | |
928 | else | |
929 | fn = fn->left; | |
930 | } | |
931 | return NULL; | |
932 | } | |
933 | ||
934 | struct fib6_node * fib6_locate(struct fib6_node *root, | |
935 | struct in6_addr *daddr, int dst_len, | |
936 | struct in6_addr *saddr, int src_len) | |
937 | { | |
938 | struct fib6_node *fn; | |
939 | ||
940 | fn = fib6_locate_1(root, daddr, dst_len, | |
941 | offsetof(struct rt6_info, rt6i_dst)); | |
942 | ||
943 | #ifdef CONFIG_IPV6_SUBTREES | |
944 | if (src_len) { | |
945 | BUG_TRAP(saddr!=NULL); | |
3fc5e044 YH |
946 | if (fn && fn->subtree) |
947 | fn = fib6_locate_1(fn->subtree, saddr, src_len, | |
1da177e4 LT |
948 | offsetof(struct rt6_info, rt6i_src)); |
949 | } | |
950 | #endif | |
951 | ||
952 | if (fn && fn->fn_flags&RTN_RTINFO) | |
953 | return fn; | |
954 | ||
955 | return NULL; | |
956 | } | |
957 | ||
958 | ||
959 | /* | |
960 | * Deletion | |
961 | * | |
962 | */ | |
963 | ||
8ed67789 | 964 | static struct rt6_info *fib6_find_prefix(struct net *net, struct fib6_node *fn) |
1da177e4 LT |
965 | { |
966 | if (fn->fn_flags&RTN_ROOT) | |
8ed67789 | 967 | return net->ipv6.ip6_null_entry; |
1da177e4 LT |
968 | |
969 | while(fn) { | |
970 | if(fn->left) | |
971 | return fn->left->leaf; | |
972 | ||
973 | if(fn->right) | |
974 | return fn->right->leaf; | |
975 | ||
7fc33165 | 976 | fn = FIB6_SUBTREE(fn); |
1da177e4 LT |
977 | } |
978 | return NULL; | |
979 | } | |
980 | ||
981 | /* | |
982 | * Called to trim the tree of intermediate nodes when possible. "fn" | |
983 | * is the node we want to try and remove. | |
984 | */ | |
985 | ||
8ed67789 DL |
986 | static struct fib6_node *fib6_repair_tree(struct net *net, |
987 | struct fib6_node *fn) | |
1da177e4 LT |
988 | { |
989 | int children; | |
990 | int nstate; | |
991 | struct fib6_node *child, *pn; | |
992 | struct fib6_walker_t *w; | |
993 | int iter = 0; | |
994 | ||
995 | for (;;) { | |
996 | RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter); | |
997 | iter++; | |
998 | ||
999 | BUG_TRAP(!(fn->fn_flags&RTN_RTINFO)); | |
1000 | BUG_TRAP(!(fn->fn_flags&RTN_TL_ROOT)); | |
1001 | BUG_TRAP(fn->leaf==NULL); | |
1002 | ||
1003 | children = 0; | |
1004 | child = NULL; | |
1005 | if (fn->right) child = fn->right, children |= 1; | |
1006 | if (fn->left) child = fn->left, children |= 2; | |
1007 | ||
7fc33165 | 1008 | if (children == 3 || FIB6_SUBTREE(fn) |
1da177e4 LT |
1009 | #ifdef CONFIG_IPV6_SUBTREES |
1010 | /* Subtree root (i.e. fn) may have one child */ | |
1011 | || (children && fn->fn_flags&RTN_ROOT) | |
1012 | #endif | |
1013 | ) { | |
8ed67789 | 1014 | fn->leaf = fib6_find_prefix(net, fn); |
1da177e4 LT |
1015 | #if RT6_DEBUG >= 2 |
1016 | if (fn->leaf==NULL) { | |
1017 | BUG_TRAP(fn->leaf); | |
8ed67789 | 1018 | fn->leaf = net->ipv6.ip6_null_entry; |
1da177e4 LT |
1019 | } |
1020 | #endif | |
1021 | atomic_inc(&fn->leaf->rt6i_ref); | |
1022 | return fn->parent; | |
1023 | } | |
1024 | ||
1025 | pn = fn->parent; | |
1026 | #ifdef CONFIG_IPV6_SUBTREES | |
7fc33165 | 1027 | if (FIB6_SUBTREE(pn) == fn) { |
1da177e4 | 1028 | BUG_TRAP(fn->fn_flags&RTN_ROOT); |
7fc33165 | 1029 | FIB6_SUBTREE(pn) = NULL; |
1da177e4 LT |
1030 | nstate = FWS_L; |
1031 | } else { | |
1032 | BUG_TRAP(!(fn->fn_flags&RTN_ROOT)); | |
1033 | #endif | |
1034 | if (pn->right == fn) pn->right = child; | |
1035 | else if (pn->left == fn) pn->left = child; | |
1036 | #if RT6_DEBUG >= 2 | |
1037 | else BUG_TRAP(0); | |
1038 | #endif | |
1039 | if (child) | |
1040 | child->parent = pn; | |
1041 | nstate = FWS_R; | |
1042 | #ifdef CONFIG_IPV6_SUBTREES | |
1043 | } | |
1044 | #endif | |
1045 | ||
1046 | read_lock(&fib6_walker_lock); | |
1047 | FOR_WALKERS(w) { | |
1048 | if (child == NULL) { | |
1049 | if (w->root == fn) { | |
1050 | w->root = w->node = NULL; | |
1051 | RT6_TRACE("W %p adjusted by delroot 1\n", w); | |
1052 | } else if (w->node == fn) { | |
1053 | RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate); | |
1054 | w->node = pn; | |
1055 | w->state = nstate; | |
1056 | } | |
1057 | } else { | |
1058 | if (w->root == fn) { | |
1059 | w->root = child; | |
1060 | RT6_TRACE("W %p adjusted by delroot 2\n", w); | |
1061 | } | |
1062 | if (w->node == fn) { | |
1063 | w->node = child; | |
1064 | if (children&2) { | |
1065 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1066 | w->state = w->state>=FWS_R ? FWS_U : FWS_INIT; | |
1067 | } else { | |
1068 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
1069 | w->state = w->state>=FWS_C ? FWS_U : FWS_INIT; | |
1070 | } | |
1071 | } | |
1072 | } | |
1073 | } | |
1074 | read_unlock(&fib6_walker_lock); | |
1075 | ||
1076 | node_free(fn); | |
7fc33165 | 1077 | if (pn->fn_flags&RTN_RTINFO || FIB6_SUBTREE(pn)) |
1da177e4 LT |
1078 | return pn; |
1079 | ||
1080 | rt6_release(pn->leaf); | |
1081 | pn->leaf = NULL; | |
1082 | fn = pn; | |
1083 | } | |
1084 | } | |
1085 | ||
1086 | static void fib6_del_route(struct fib6_node *fn, struct rt6_info **rtp, | |
86872cb5 | 1087 | struct nl_info *info) |
1da177e4 LT |
1088 | { |
1089 | struct fib6_walker_t *w; | |
1090 | struct rt6_info *rt = *rtp; | |
c572872f | 1091 | struct net *net = info->nl_net; |
1da177e4 LT |
1092 | |
1093 | RT6_TRACE("fib6_del_route\n"); | |
1094 | ||
1095 | /* Unlink it */ | |
7cc48263 | 1096 | *rtp = rt->u.dst.rt6_next; |
1da177e4 | 1097 | rt->rt6i_node = NULL; |
c572872f BT |
1098 | net->ipv6.rt6_stats->fib_rt_entries--; |
1099 | net->ipv6.rt6_stats->fib_discarded_routes++; | |
1da177e4 | 1100 | |
f11e6659 DM |
1101 | /* Reset round-robin state, if necessary */ |
1102 | if (fn->rr_ptr == rt) | |
1103 | fn->rr_ptr = NULL; | |
1104 | ||
1da177e4 LT |
1105 | /* Adjust walkers */ |
1106 | read_lock(&fib6_walker_lock); | |
1107 | FOR_WALKERS(w) { | |
1108 | if (w->state == FWS_C && w->leaf == rt) { | |
1109 | RT6_TRACE("walker %p adjusted by delroute\n", w); | |
7cc48263 | 1110 | w->leaf = rt->u.dst.rt6_next; |
1da177e4 LT |
1111 | if (w->leaf == NULL) |
1112 | w->state = FWS_U; | |
1113 | } | |
1114 | } | |
1115 | read_unlock(&fib6_walker_lock); | |
1116 | ||
7cc48263 | 1117 | rt->u.dst.rt6_next = NULL; |
1da177e4 | 1118 | |
1da177e4 LT |
1119 | /* If it was last route, expunge its radix tree node */ |
1120 | if (fn->leaf == NULL) { | |
1121 | fn->fn_flags &= ~RTN_RTINFO; | |
c572872f | 1122 | net->ipv6.rt6_stats->fib_route_nodes--; |
8ed67789 | 1123 | fn = fib6_repair_tree(net, fn); |
1da177e4 LT |
1124 | } |
1125 | ||
1126 | if (atomic_read(&rt->rt6i_ref) != 1) { | |
1127 | /* This route is used as dummy address holder in some split | |
1128 | * nodes. It is not leaked, but it still holds other resources, | |
1129 | * which must be released in time. So, scan ascendant nodes | |
1130 | * and replace dummy references to this route with references | |
1131 | * to still alive ones. | |
1132 | */ | |
1133 | while (fn) { | |
1134 | if (!(fn->fn_flags&RTN_RTINFO) && fn->leaf == rt) { | |
8ed67789 | 1135 | fn->leaf = fib6_find_prefix(net, fn); |
1da177e4 LT |
1136 | atomic_inc(&fn->leaf->rt6i_ref); |
1137 | rt6_release(rt); | |
1138 | } | |
1139 | fn = fn->parent; | |
1140 | } | |
1141 | /* No more references are possible at this point. */ | |
2df96af0 | 1142 | BUG_ON(atomic_read(&rt->rt6i_ref) != 1); |
1da177e4 LT |
1143 | } |
1144 | ||
86872cb5 | 1145 | inet6_rt_notify(RTM_DELROUTE, rt, info); |
1da177e4 LT |
1146 | rt6_release(rt); |
1147 | } | |
1148 | ||
86872cb5 | 1149 | int fib6_del(struct rt6_info *rt, struct nl_info *info) |
1da177e4 | 1150 | { |
8ed67789 | 1151 | struct net *net = info->nl_net; |
1da177e4 LT |
1152 | struct fib6_node *fn = rt->rt6i_node; |
1153 | struct rt6_info **rtp; | |
1154 | ||
1155 | #if RT6_DEBUG >= 2 | |
1156 | if (rt->u.dst.obsolete>0) { | |
1157 | BUG_TRAP(fn==NULL); | |
1158 | return -ENOENT; | |
1159 | } | |
1160 | #endif | |
8ed67789 | 1161 | if (fn == NULL || rt == net->ipv6.ip6_null_entry) |
1da177e4 LT |
1162 | return -ENOENT; |
1163 | ||
1164 | BUG_TRAP(fn->fn_flags&RTN_RTINFO); | |
1165 | ||
150730d5 YH |
1166 | if (!(rt->rt6i_flags&RTF_CACHE)) { |
1167 | struct fib6_node *pn = fn; | |
1168 | #ifdef CONFIG_IPV6_SUBTREES | |
1169 | /* clones of this route might be in another subtree */ | |
1170 | if (rt->rt6i_src.plen) { | |
1171 | while (!(pn->fn_flags&RTN_ROOT)) | |
1172 | pn = pn->parent; | |
1173 | pn = pn->parent; | |
1174 | } | |
1175 | #endif | |
ec7d43c2 | 1176 | fib6_prune_clones(info->nl_net, pn, rt); |
150730d5 | 1177 | } |
1da177e4 LT |
1178 | |
1179 | /* | |
1180 | * Walk the leaf entries looking for ourself | |
1181 | */ | |
1182 | ||
7cc48263 | 1183 | for (rtp = &fn->leaf; *rtp; rtp = &(*rtp)->u.dst.rt6_next) { |
1da177e4 | 1184 | if (*rtp == rt) { |
86872cb5 | 1185 | fib6_del_route(fn, rtp, info); |
1da177e4 LT |
1186 | return 0; |
1187 | } | |
1188 | } | |
1189 | return -ENOENT; | |
1190 | } | |
1191 | ||
1192 | /* | |
1193 | * Tree traversal function. | |
1194 | * | |
1195 | * Certainly, it is not interrupt safe. | |
1196 | * However, it is internally reenterable wrt itself and fib6_add/fib6_del. | |
1197 | * It means, that we can modify tree during walking | |
1198 | * and use this function for garbage collection, clone pruning, | |
1ab1457c | 1199 | * cleaning tree when a device goes down etc. etc. |
1da177e4 LT |
1200 | * |
1201 | * It guarantees that every node will be traversed, | |
1202 | * and that it will be traversed only once. | |
1203 | * | |
1204 | * Callback function w->func may return: | |
1205 | * 0 -> continue walking. | |
1206 | * positive value -> walking is suspended (used by tree dumps, | |
1207 | * and probably by gc, if it will be split to several slices) | |
1208 | * negative value -> terminate walking. | |
1209 | * | |
1210 | * The function itself returns: | |
1211 | * 0 -> walk is complete. | |
1212 | * >0 -> walk is incomplete (i.e. suspended) | |
1213 | * <0 -> walk is terminated by an error. | |
1214 | */ | |
1215 | ||
90d41122 | 1216 | static int fib6_walk_continue(struct fib6_walker_t *w) |
1da177e4 LT |
1217 | { |
1218 | struct fib6_node *fn, *pn; | |
1219 | ||
1220 | for (;;) { | |
1221 | fn = w->node; | |
1222 | if (fn == NULL) | |
1223 | return 0; | |
1224 | ||
1225 | if (w->prune && fn != w->root && | |
1226 | fn->fn_flags&RTN_RTINFO && w->state < FWS_C) { | |
1227 | w->state = FWS_C; | |
1228 | w->leaf = fn->leaf; | |
1229 | } | |
1230 | switch (w->state) { | |
1231 | #ifdef CONFIG_IPV6_SUBTREES | |
1232 | case FWS_S: | |
7fc33165 YH |
1233 | if (FIB6_SUBTREE(fn)) { |
1234 | w->node = FIB6_SUBTREE(fn); | |
1da177e4 LT |
1235 | continue; |
1236 | } | |
1237 | w->state = FWS_L; | |
1ab1457c | 1238 | #endif |
1da177e4 LT |
1239 | case FWS_L: |
1240 | if (fn->left) { | |
1241 | w->node = fn->left; | |
1242 | w->state = FWS_INIT; | |
1243 | continue; | |
1244 | } | |
1245 | w->state = FWS_R; | |
1246 | case FWS_R: | |
1247 | if (fn->right) { | |
1248 | w->node = fn->right; | |
1249 | w->state = FWS_INIT; | |
1250 | continue; | |
1251 | } | |
1252 | w->state = FWS_C; | |
1253 | w->leaf = fn->leaf; | |
1254 | case FWS_C: | |
1255 | if (w->leaf && fn->fn_flags&RTN_RTINFO) { | |
1256 | int err = w->func(w); | |
1257 | if (err) | |
1258 | return err; | |
1259 | continue; | |
1260 | } | |
1261 | w->state = FWS_U; | |
1262 | case FWS_U: | |
1263 | if (fn == w->root) | |
1264 | return 0; | |
1265 | pn = fn->parent; | |
1266 | w->node = pn; | |
1267 | #ifdef CONFIG_IPV6_SUBTREES | |
7fc33165 | 1268 | if (FIB6_SUBTREE(pn) == fn) { |
1da177e4 LT |
1269 | BUG_TRAP(fn->fn_flags&RTN_ROOT); |
1270 | w->state = FWS_L; | |
1271 | continue; | |
1272 | } | |
1273 | #endif | |
1274 | if (pn->left == fn) { | |
1275 | w->state = FWS_R; | |
1276 | continue; | |
1277 | } | |
1278 | if (pn->right == fn) { | |
1279 | w->state = FWS_C; | |
1280 | w->leaf = w->node->leaf; | |
1281 | continue; | |
1282 | } | |
1283 | #if RT6_DEBUG >= 2 | |
1284 | BUG_TRAP(0); | |
1285 | #endif | |
1286 | } | |
1287 | } | |
1288 | } | |
1289 | ||
90d41122 | 1290 | static int fib6_walk(struct fib6_walker_t *w) |
1da177e4 LT |
1291 | { |
1292 | int res; | |
1293 | ||
1294 | w->state = FWS_INIT; | |
1295 | w->node = w->root; | |
1296 | ||
1297 | fib6_walker_link(w); | |
1298 | res = fib6_walk_continue(w); | |
1299 | if (res <= 0) | |
1300 | fib6_walker_unlink(w); | |
1301 | return res; | |
1302 | } | |
1303 | ||
1304 | static int fib6_clean_node(struct fib6_walker_t *w) | |
1305 | { | |
1306 | int res; | |
1307 | struct rt6_info *rt; | |
0a8891a0 | 1308 | struct fib6_cleaner_t *c = container_of(w, struct fib6_cleaner_t, w); |
ec7d43c2 BT |
1309 | struct nl_info info = { |
1310 | .nl_net = c->net, | |
1311 | }; | |
1da177e4 | 1312 | |
7cc48263 | 1313 | for (rt = w->leaf; rt; rt = rt->u.dst.rt6_next) { |
1da177e4 LT |
1314 | res = c->func(rt, c->arg); |
1315 | if (res < 0) { | |
1316 | w->leaf = rt; | |
528c4ceb | 1317 | res = fib6_del(rt, &info); |
1da177e4 LT |
1318 | if (res) { |
1319 | #if RT6_DEBUG >= 2 | |
1320 | printk(KERN_DEBUG "fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt, rt->rt6i_node, res); | |
1321 | #endif | |
1322 | continue; | |
1323 | } | |
1324 | return 0; | |
1325 | } | |
1326 | BUG_TRAP(res==0); | |
1327 | } | |
1328 | w->leaf = rt; | |
1329 | return 0; | |
1330 | } | |
1331 | ||
1332 | /* | |
1333 | * Convenient frontend to tree walker. | |
1ab1457c | 1334 | * |
1da177e4 LT |
1335 | * func is called on each route. |
1336 | * It may return -1 -> delete this route. | |
1337 | * 0 -> continue walking | |
1338 | * | |
1339 | * prune==1 -> only immediate children of node (certainly, | |
1340 | * ignoring pure split nodes) will be scanned. | |
1341 | */ | |
1342 | ||
ec7d43c2 | 1343 | static void fib6_clean_tree(struct net *net, struct fib6_node *root, |
8ce11e6a AB |
1344 | int (*func)(struct rt6_info *, void *arg), |
1345 | int prune, void *arg) | |
1da177e4 LT |
1346 | { |
1347 | struct fib6_cleaner_t c; | |
1348 | ||
1349 | c.w.root = root; | |
1350 | c.w.func = fib6_clean_node; | |
1351 | c.w.prune = prune; | |
1352 | c.func = func; | |
1353 | c.arg = arg; | |
ec7d43c2 | 1354 | c.net = net; |
1da177e4 LT |
1355 | |
1356 | fib6_walk(&c.w); | |
1357 | } | |
1358 | ||
f3db4851 | 1359 | void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg), |
c71099ac TG |
1360 | int prune, void *arg) |
1361 | { | |
c71099ac | 1362 | struct fib6_table *table; |
1b43af54 | 1363 | struct hlist_node *node; |
58f09b78 | 1364 | struct hlist_head *head; |
1b43af54 | 1365 | unsigned int h; |
c71099ac | 1366 | |
1b43af54 PM |
1367 | rcu_read_lock(); |
1368 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { | |
f3db4851 | 1369 | head = &net->ipv6.fib_table_hash[h]; |
58f09b78 | 1370 | hlist_for_each_entry_rcu(table, node, head, tb6_hlist) { |
c71099ac | 1371 | write_lock_bh(&table->tb6_lock); |
ec7d43c2 BT |
1372 | fib6_clean_tree(net, &table->tb6_root, |
1373 | func, prune, arg); | |
c71099ac TG |
1374 | write_unlock_bh(&table->tb6_lock); |
1375 | } | |
1376 | } | |
1b43af54 | 1377 | rcu_read_unlock(); |
c71099ac TG |
1378 | } |
1379 | ||
1da177e4 LT |
1380 | static int fib6_prune_clone(struct rt6_info *rt, void *arg) |
1381 | { | |
1382 | if (rt->rt6i_flags & RTF_CACHE) { | |
1383 | RT6_TRACE("pruning clone %p\n", rt); | |
1384 | return -1; | |
1385 | } | |
1386 | ||
1387 | return 0; | |
1388 | } | |
1389 | ||
ec7d43c2 BT |
1390 | static void fib6_prune_clones(struct net *net, struct fib6_node *fn, |
1391 | struct rt6_info *rt) | |
1da177e4 | 1392 | { |
ec7d43c2 | 1393 | fib6_clean_tree(net, fn, fib6_prune_clone, 1, rt); |
1da177e4 LT |
1394 | } |
1395 | ||
1396 | /* | |
1397 | * Garbage collection | |
1398 | */ | |
1399 | ||
1400 | static struct fib6_gc_args | |
1401 | { | |
1402 | int timeout; | |
1403 | int more; | |
1404 | } gc_args; | |
1405 | ||
1406 | static int fib6_age(struct rt6_info *rt, void *arg) | |
1407 | { | |
1408 | unsigned long now = jiffies; | |
1409 | ||
1410 | /* | |
1411 | * check addrconf expiration here. | |
1412 | * Routes are expired even if they are in use. | |
1413 | * | |
1414 | * Also age clones. Note, that clones are aged out | |
1415 | * only if they are not in use now. | |
1416 | */ | |
1417 | ||
1418 | if (rt->rt6i_flags&RTF_EXPIRES && rt->rt6i_expires) { | |
1419 | if (time_after(now, rt->rt6i_expires)) { | |
1420 | RT6_TRACE("expiring %p\n", rt); | |
1da177e4 LT |
1421 | return -1; |
1422 | } | |
1423 | gc_args.more++; | |
1424 | } else if (rt->rt6i_flags & RTF_CACHE) { | |
1425 | if (atomic_read(&rt->u.dst.__refcnt) == 0 && | |
1426 | time_after_eq(now, rt->u.dst.lastuse + gc_args.timeout)) { | |
1427 | RT6_TRACE("aging clone %p\n", rt); | |
1428 | return -1; | |
1429 | } else if ((rt->rt6i_flags & RTF_GATEWAY) && | |
1430 | (!(rt->rt6i_nexthop->flags & NTF_ROUTER))) { | |
1431 | RT6_TRACE("purging route %p via non-router but gateway\n", | |
1432 | rt); | |
1433 | return -1; | |
1434 | } | |
1435 | gc_args.more++; | |
1436 | } | |
1437 | ||
1438 | return 0; | |
1439 | } | |
1440 | ||
1441 | static DEFINE_SPINLOCK(fib6_gc_lock); | |
1442 | ||
5b7c931d | 1443 | void fib6_run_gc(unsigned long expires, struct net *net) |
1da177e4 | 1444 | { |
5b7c931d | 1445 | if (expires != ~0UL) { |
1da177e4 | 1446 | spin_lock_bh(&fib6_gc_lock); |
5b7c931d DL |
1447 | gc_args.timeout = expires ? (int)expires : |
1448 | net->ipv6.sysctl.ip6_rt_gc_interval; | |
1da177e4 | 1449 | } else { |
a76d7345 | 1450 | if (!spin_trylock_bh(&fib6_gc_lock)) { |
417f28bb | 1451 | mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ); |
1da177e4 LT |
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) | |
c8a45222 SH |
1463 | mod_timer(&net->ipv6.ip6_fib_timer, |
1464 | round_jiffies(jiffies | |
1465 | + net->ipv6.sysctl.ip6_rt_gc_interval)); | |
417f28bb SH |
1466 | else |
1467 | del_timer(&net->ipv6.ip6_fib_timer); | |
1da177e4 LT |
1468 | spin_unlock_bh(&fib6_gc_lock); |
1469 | } | |
1470 | ||
5b7c931d DL |
1471 | static void fib6_gc_timer_cb(unsigned long arg) |
1472 | { | |
1473 | fib6_run_gc(0, (struct net *)arg); | |
1474 | } | |
1475 | ||
58f09b78 | 1476 | static int fib6_net_init(struct net *net) |
1da177e4 | 1477 | { |
417f28bb | 1478 | setup_timer(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, (unsigned long)net); |
63152fc0 | 1479 | |
c572872f BT |
1480 | net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL); |
1481 | if (!net->ipv6.rt6_stats) | |
1482 | goto out_timer; | |
1483 | ||
58f09b78 DL |
1484 | net->ipv6.fib_table_hash = |
1485 | kzalloc(sizeof(*net->ipv6.fib_table_hash)*FIB_TABLE_HASHSZ, | |
1486 | GFP_KERNEL); | |
1487 | if (!net->ipv6.fib_table_hash) | |
c572872f | 1488 | goto out_rt6_stats; |
e0b85590 | 1489 | |
58f09b78 DL |
1490 | net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl), |
1491 | GFP_KERNEL); | |
1492 | if (!net->ipv6.fib6_main_tbl) | |
e0b85590 DL |
1493 | goto out_fib_table_hash; |
1494 | ||
58f09b78 | 1495 | net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN; |
8ed67789 | 1496 | net->ipv6.fib6_main_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry; |
58f09b78 DL |
1497 | net->ipv6.fib6_main_tbl->tb6_root.fn_flags = |
1498 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
e0b85590 DL |
1499 | |
1500 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES | |
58f09b78 DL |
1501 | net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl), |
1502 | GFP_KERNEL); | |
1503 | if (!net->ipv6.fib6_local_tbl) | |
e0b85590 | 1504 | goto out_fib6_main_tbl; |
58f09b78 | 1505 | net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL; |
8ed67789 | 1506 | net->ipv6.fib6_local_tbl->tb6_root.leaf = net->ipv6.ip6_null_entry; |
58f09b78 DL |
1507 | net->ipv6.fib6_local_tbl->tb6_root.fn_flags = |
1508 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
e0b85590 | 1509 | #endif |
58f09b78 | 1510 | fib6_tables_init(net); |
f845ab6b | 1511 | |
417f28bb | 1512 | return 0; |
d63bddbe | 1513 | |
e0b85590 | 1514 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 1515 | out_fib6_main_tbl: |
58f09b78 | 1516 | kfree(net->ipv6.fib6_main_tbl); |
e0b85590 | 1517 | #endif |
e0b85590 | 1518 | out_fib_table_hash: |
58f09b78 | 1519 | kfree(net->ipv6.fib_table_hash); |
c572872f BT |
1520 | out_rt6_stats: |
1521 | kfree(net->ipv6.rt6_stats); | |
63152fc0 | 1522 | out_timer: |
417f28bb | 1523 | return -ENOMEM; |
58f09b78 DL |
1524 | } |
1525 | ||
1526 | static void fib6_net_exit(struct net *net) | |
1527 | { | |
8ed67789 | 1528 | rt6_ifdown(net, NULL); |
417f28bb SH |
1529 | del_timer_sync(&net->ipv6.ip6_fib_timer); |
1530 | ||
58f09b78 DL |
1531 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
1532 | kfree(net->ipv6.fib6_local_tbl); | |
1533 | #endif | |
1534 | kfree(net->ipv6.fib6_main_tbl); | |
1535 | kfree(net->ipv6.fib_table_hash); | |
c572872f | 1536 | kfree(net->ipv6.rt6_stats); |
58f09b78 DL |
1537 | } |
1538 | ||
1539 | static struct pernet_operations fib6_net_ops = { | |
1540 | .init = fib6_net_init, | |
1541 | .exit = fib6_net_exit, | |
1542 | }; | |
1543 | ||
1544 | int __init fib6_init(void) | |
1545 | { | |
1546 | int ret = -ENOMEM; | |
63152fc0 | 1547 | |
58f09b78 DL |
1548 | fib6_node_kmem = kmem_cache_create("fib6_nodes", |
1549 | sizeof(struct fib6_node), | |
1550 | 0, SLAB_HWCACHE_ALIGN, | |
1551 | NULL); | |
1552 | if (!fib6_node_kmem) | |
1553 | goto out; | |
1554 | ||
1555 | ret = register_pernet_subsys(&fib6_net_ops); | |
1556 | if (ret) | |
c572872f | 1557 | goto out_kmem_cache_create; |
58f09b78 DL |
1558 | |
1559 | ret = __rtnl_register(PF_INET6, RTM_GETROUTE, NULL, inet6_dump_fib); | |
1560 | if (ret) | |
1561 | goto out_unregister_subsys; | |
1562 | out: | |
1563 | return ret; | |
1564 | ||
1565 | out_unregister_subsys: | |
1566 | unregister_pernet_subsys(&fib6_net_ops); | |
d63bddbe DL |
1567 | out_kmem_cache_create: |
1568 | kmem_cache_destroy(fib6_node_kmem); | |
1569 | goto out; | |
1da177e4 LT |
1570 | } |
1571 | ||
1572 | void fib6_gc_cleanup(void) | |
1573 | { | |
58f09b78 | 1574 | unregister_pernet_subsys(&fib6_net_ops); |
1da177e4 LT |
1575 | kmem_cache_destroy(fib6_node_kmem); |
1576 | } |