Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 | 2 | /* |
1ab1457c | 3 | * Linux INET6 implementation |
1da177e4 LT |
4 | * Forwarding Information Database |
5 | * | |
6 | * Authors: | |
1ab1457c | 7 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 | 8 | * |
8db46f1d WY |
9 | * Changes: |
10 | * Yuji SEKIYA @USAGI: Support default route on router node; | |
11 | * remove ip6_null_entry from the top of | |
12 | * routing table. | |
13 | * Ville Nuorvala: Fixed routing subtrees. | |
1da177e4 | 14 | */ |
f3213831 JP |
15 | |
16 | #define pr_fmt(fmt) "IPv6: " fmt | |
17 | ||
1da177e4 LT |
18 | #include <linux/errno.h> |
19 | #include <linux/types.h> | |
20 | #include <linux/net.h> | |
21 | #include <linux/route.h> | |
22 | #include <linux/netdevice.h> | |
23 | #include <linux/in6.h> | |
24 | #include <linux/init.h> | |
c71099ac | 25 | #include <linux/list.h> |
5a0e3ad6 | 26 | #include <linux/slab.h> |
1da177e4 | 27 | |
cc5f0eb2 | 28 | #include <net/ip.h> |
1da177e4 LT |
29 | #include <net/ipv6.h> |
30 | #include <net/ndisc.h> | |
31 | #include <net/addrconf.h> | |
19e42e45 | 32 | #include <net/lwtunnel.h> |
df77fe4d | 33 | #include <net/fib_notifier.h> |
1da177e4 LT |
34 | |
35 | #include <net/ip6_fib.h> | |
36 | #include <net/ip6_route.h> | |
37 | ||
437de07c | 38 | static struct kmem_cache *fib6_node_kmem __read_mostly; |
1da177e4 | 39 | |
94b2cfe0 HFS |
40 | struct fib6_cleaner { |
41 | struct fib6_walker w; | |
ec7d43c2 | 42 | struct net *net; |
8d1c802b | 43 | int (*func)(struct fib6_info *, void *arg); |
327571cb | 44 | int sernum; |
1da177e4 | 45 | void *arg; |
7c6bb7d2 | 46 | bool skip_notify; |
1da177e4 LT |
47 | }; |
48 | ||
1da177e4 LT |
49 | #ifdef CONFIG_IPV6_SUBTREES |
50 | #define FWS_INIT FWS_S | |
1da177e4 LT |
51 | #else |
52 | #define FWS_INIT FWS_L | |
1da177e4 LT |
53 | #endif |
54 | ||
8d1c802b | 55 | static struct fib6_info *fib6_find_prefix(struct net *net, |
66f5d6ce WW |
56 | struct fib6_table *table, |
57 | struct fib6_node *fn); | |
58 | static struct fib6_node *fib6_repair_tree(struct net *net, | |
59 | struct fib6_table *table, | |
60 | struct fib6_node *fn); | |
9a03cd8f | 61 | static int fib6_walk(struct net *net, struct fib6_walker *w); |
94b2cfe0 | 62 | static int fib6_walk_continue(struct fib6_walker *w); |
1da177e4 LT |
63 | |
64 | /* | |
65 | * A routing update causes an increase of the serial number on the | |
66 | * affected subtree. This allows for cached routes to be asynchronously | |
67 | * tested when modifications are made to the destination cache as a | |
68 | * result of redirects, path MTU changes, etc. | |
69 | */ | |
70 | ||
86cb30ec | 71 | static void fib6_gc_timer_cb(struct timer_list *t); |
5b7c931d | 72 | |
9a03cd8f MK |
73 | #define FOR_WALKERS(net, w) \ |
74 | list_for_each_entry(w, &(net)->ipv6.fib6_walkers, lh) | |
1da177e4 | 75 | |
9a03cd8f | 76 | static void fib6_walker_link(struct net *net, struct fib6_walker *w) |
90d41122 | 77 | { |
9a03cd8f MK |
78 | write_lock_bh(&net->ipv6.fib6_walker_lock); |
79 | list_add(&w->lh, &net->ipv6.fib6_walkers); | |
80 | write_unlock_bh(&net->ipv6.fib6_walker_lock); | |
90d41122 AB |
81 | } |
82 | ||
9a03cd8f | 83 | static void fib6_walker_unlink(struct net *net, struct fib6_walker *w) |
90d41122 | 84 | { |
9a03cd8f | 85 | write_lock_bh(&net->ipv6.fib6_walker_lock); |
bbef49da | 86 | list_del(&w->lh); |
9a03cd8f | 87 | write_unlock_bh(&net->ipv6.fib6_walker_lock); |
90d41122 | 88 | } |
94b2cfe0 | 89 | |
812918c4 | 90 | static int fib6_new_sernum(struct net *net) |
1da177e4 | 91 | { |
42b18706 HFS |
92 | int new, old; |
93 | ||
94 | do { | |
812918c4 | 95 | old = atomic_read(&net->ipv6.fib6_sernum); |
42b18706 | 96 | new = old < INT_MAX ? old + 1 : 1; |
812918c4 HFS |
97 | } while (atomic_cmpxchg(&net->ipv6.fib6_sernum, |
98 | old, new) != old); | |
42b18706 | 99 | return new; |
1da177e4 LT |
100 | } |
101 | ||
327571cb HFS |
102 | enum { |
103 | FIB6_NO_SERNUM_CHANGE = 0, | |
104 | }; | |
105 | ||
93c2fb25 | 106 | void fib6_update_sernum(struct net *net, struct fib6_info *f6i) |
180ca444 | 107 | { |
180ca444 WW |
108 | struct fib6_node *fn; |
109 | ||
93c2fb25 DA |
110 | fn = rcu_dereference_protected(f6i->fib6_node, |
111 | lockdep_is_held(&f6i->fib6_table->tb6_lock)); | |
180ca444 WW |
112 | if (fn) |
113 | fn->fn_sernum = fib6_new_sernum(net); | |
180ca444 WW |
114 | } |
115 | ||
1da177e4 LT |
116 | /* |
117 | * Auxiliary address test functions for the radix tree. | |
118 | * | |
1ab1457c | 119 | * These assume a 32bit processor (although it will work on |
1da177e4 LT |
120 | * 64bit processors) |
121 | */ | |
122 | ||
123 | /* | |
124 | * test bit | |
125 | */ | |
02cdce53 YH |
126 | #if defined(__LITTLE_ENDIAN) |
127 | # define BITOP_BE32_SWIZZLE (0x1F & ~7) | |
128 | #else | |
129 | # define BITOP_BE32_SWIZZLE 0 | |
130 | #endif | |
1da177e4 | 131 | |
94b2cfe0 | 132 | static __be32 addr_bit_set(const void *token, int fn_bit) |
1da177e4 | 133 | { |
b71d1d42 | 134 | const __be32 *addr = token; |
02cdce53 YH |
135 | /* |
136 | * Here, | |
8db46f1d | 137 | * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f) |
02cdce53 YH |
138 | * is optimized version of |
139 | * htonl(1 << ((~fn_bit)&0x1F)) | |
140 | * See include/asm-generic/bitops/le.h. | |
141 | */ | |
0eae88f3 ED |
142 | return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) & |
143 | addr[fn_bit >> 5]; | |
1da177e4 LT |
144 | } |
145 | ||
1cf844c7 | 146 | struct fib6_info *fib6_info_alloc(gfp_t gfp_flags, bool with_fib6_nh) |
a64efe14 | 147 | { |
8d1c802b | 148 | struct fib6_info *f6i; |
1cf844c7 | 149 | size_t sz = sizeof(*f6i); |
a64efe14 | 150 | |
1cf844c7 DA |
151 | if (with_fib6_nh) |
152 | sz += sizeof(struct fib6_nh); | |
153 | ||
154 | f6i = kzalloc(sz, gfp_flags); | |
a64efe14 DA |
155 | if (!f6i) |
156 | return NULL; | |
157 | ||
f88d8ea6 | 158 | /* fib6_siblings is a union with nh_list, so this initializes both */ |
93c2fb25 | 159 | INIT_LIST_HEAD(&f6i->fib6_siblings); |
f05713e0 | 160 | refcount_set(&f6i->fib6_ref, 1); |
a64efe14 DA |
161 | |
162 | return f6i; | |
163 | } | |
164 | ||
9b0a8da8 | 165 | void fib6_info_destroy_rcu(struct rcu_head *head) |
a64efe14 | 166 | { |
9b0a8da8 | 167 | struct fib6_info *f6i = container_of(head, struct fib6_info, rcu); |
a64efe14 | 168 | |
93c2fb25 | 169 | WARN_ON(f6i->fib6_node); |
a64efe14 | 170 | |
f88d8ea6 DA |
171 | if (f6i->nh) |
172 | nexthop_put(f6i->nh); | |
173 | else | |
174 | fib6_nh_release(f6i->fib6_nh); | |
175 | ||
cc5f0eb2 | 176 | ip_fib_metrics_put(f6i->fib6_metrics); |
a64efe14 DA |
177 | kfree(f6i); |
178 | } | |
9b0a8da8 | 179 | EXPORT_SYMBOL_GPL(fib6_info_destroy_rcu); |
a64efe14 | 180 | |
81eb8447 | 181 | static struct fib6_node *node_alloc(struct net *net) |
1da177e4 LT |
182 | { |
183 | struct fib6_node *fn; | |
184 | ||
c3762229 | 185 | fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC); |
81eb8447 WW |
186 | if (fn) |
187 | net->ipv6.rt6_stats->fib_nodes++; | |
1da177e4 LT |
188 | |
189 | return fn; | |
190 | } | |
191 | ||
81eb8447 | 192 | static void node_free_immediate(struct net *net, struct fib6_node *fn) |
c5cff856 WW |
193 | { |
194 | kmem_cache_free(fib6_node_kmem, fn); | |
81eb8447 | 195 | net->ipv6.rt6_stats->fib_nodes--; |
c5cff856 WW |
196 | } |
197 | ||
198 | static void node_free_rcu(struct rcu_head *head) | |
1da177e4 | 199 | { |
c5cff856 WW |
200 | struct fib6_node *fn = container_of(head, struct fib6_node, rcu); |
201 | ||
1da177e4 LT |
202 | kmem_cache_free(fib6_node_kmem, fn); |
203 | } | |
204 | ||
81eb8447 | 205 | static void node_free(struct net *net, struct fib6_node *fn) |
c5cff856 WW |
206 | { |
207 | call_rcu(&fn->rcu, node_free_rcu); | |
81eb8447 | 208 | net->ipv6.rt6_stats->fib_nodes--; |
c5cff856 WW |
209 | } |
210 | ||
ba1cc08d SD |
211 | static void fib6_free_table(struct fib6_table *table) |
212 | { | |
213 | inetpeer_invalidate_tree(&table->tb6_peers); | |
214 | kfree(table); | |
215 | } | |
216 | ||
58f09b78 | 217 | static void fib6_link_table(struct net *net, struct fib6_table *tb) |
1b43af54 PM |
218 | { |
219 | unsigned int h; | |
220 | ||
375216ad TG |
221 | /* |
222 | * Initialize table lock at a single place to give lockdep a key, | |
223 | * tables aren't visible prior to being linked to the list. | |
224 | */ | |
66f5d6ce | 225 | spin_lock_init(&tb->tb6_lock); |
a33bc5c1 | 226 | h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1); |
1b43af54 PM |
227 | |
228 | /* | |
229 | * No protection necessary, this is the only list mutatation | |
230 | * operation, tables never disappear once they exist. | |
231 | */ | |
58f09b78 | 232 | hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]); |
1b43af54 | 233 | } |
c71099ac | 234 | |
1b43af54 | 235 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 236 | |
8ed67789 | 237 | static struct fib6_table *fib6_alloc_table(struct net *net, u32 id) |
c71099ac TG |
238 | { |
239 | struct fib6_table *table; | |
240 | ||
241 | table = kzalloc(sizeof(*table), GFP_ATOMIC); | |
507c9b1e | 242 | if (table) { |
c71099ac | 243 | table->tb6_id = id; |
66f5d6ce | 244 | rcu_assign_pointer(table->tb6_root.leaf, |
421842ed | 245 | net->ipv6.fib6_null_entry); |
c71099ac | 246 | table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; |
8e773277 | 247 | inet_peer_base_init(&table->tb6_peers); |
c71099ac TG |
248 | } |
249 | ||
250 | return table; | |
251 | } | |
252 | ||
58f09b78 | 253 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac TG |
254 | { |
255 | struct fib6_table *tb; | |
256 | ||
257 | if (id == 0) | |
258 | id = RT6_TABLE_MAIN; | |
58f09b78 | 259 | tb = fib6_get_table(net, id); |
c71099ac TG |
260 | if (tb) |
261 | return tb; | |
262 | ||
8ed67789 | 263 | tb = fib6_alloc_table(net, id); |
507c9b1e | 264 | if (tb) |
58f09b78 | 265 | fib6_link_table(net, tb); |
c71099ac TG |
266 | |
267 | return tb; | |
268 | } | |
b3b4663c | 269 | EXPORT_SYMBOL_GPL(fib6_new_table); |
c71099ac | 270 | |
58f09b78 | 271 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac TG |
272 | { |
273 | struct fib6_table *tb; | |
58f09b78 | 274 | struct hlist_head *head; |
c71099ac TG |
275 | unsigned int h; |
276 | ||
277 | if (id == 0) | |
278 | id = RT6_TABLE_MAIN; | |
a33bc5c1 | 279 | h = id & (FIB6_TABLE_HASHSZ - 1); |
c71099ac | 280 | rcu_read_lock(); |
58f09b78 | 281 | head = &net->ipv6.fib_table_hash[h]; |
b67bfe0d | 282 | hlist_for_each_entry_rcu(tb, head, tb6_hlist) { |
c71099ac TG |
283 | if (tb->tb6_id == id) { |
284 | rcu_read_unlock(); | |
285 | return tb; | |
286 | } | |
287 | } | |
288 | rcu_read_unlock(); | |
289 | ||
290 | return NULL; | |
291 | } | |
c4850687 | 292 | EXPORT_SYMBOL_GPL(fib6_get_table); |
c71099ac | 293 | |
2c8c1e72 | 294 | static void __net_init fib6_tables_init(struct net *net) |
c71099ac | 295 | { |
58f09b78 DL |
296 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
297 | fib6_link_table(net, net->ipv6.fib6_local_tbl); | |
c71099ac | 298 | } |
c71099ac TG |
299 | #else |
300 | ||
58f09b78 | 301 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac | 302 | { |
58f09b78 | 303 | return fib6_get_table(net, id); |
c71099ac TG |
304 | } |
305 | ||
58f09b78 | 306 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac | 307 | { |
58f09b78 | 308 | return net->ipv6.fib6_main_tbl; |
c71099ac TG |
309 | } |
310 | ||
4c9483b2 | 311 | struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6, |
b75cc8f9 | 312 | const struct sk_buff *skb, |
58f09b78 | 313 | int flags, pol_lookup_t lookup) |
c71099ac | 314 | { |
ab997ad4 | 315 | struct rt6_info *rt; |
316 | ||
b75cc8f9 | 317 | rt = lookup(net, net->ipv6.fib6_main_tbl, fl6, skb, flags); |
07f61557 | 318 | if (rt->dst.error == -EAGAIN) { |
d64a1f57 | 319 | ip6_rt_put_flags(rt, flags); |
ab997ad4 | 320 | rt = net->ipv6.ip6_null_entry; |
7b09c2d0 | 321 | if (!(flags & RT6_LOOKUP_F_DST_NOREF)) |
d64a1f57 | 322 | dst_hold(&rt->dst); |
ab997ad4 | 323 | } |
324 | ||
325 | return &rt->dst; | |
c71099ac TG |
326 | } |
327 | ||
138118ec | 328 | /* called with rcu lock held; no reference taken on fib6_info */ |
effda4dd DA |
329 | int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6, |
330 | struct fib6_result *res, int flags) | |
138118ec | 331 | { |
effda4dd DA |
332 | return fib6_table_lookup(net, net->ipv6.fib6_main_tbl, oif, fl6, |
333 | res, flags); | |
138118ec DA |
334 | } |
335 | ||
2c8c1e72 | 336 | static void __net_init fib6_tables_init(struct net *net) |
c71099ac | 337 | { |
58f09b78 | 338 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
c71099ac TG |
339 | } |
340 | ||
341 | #endif | |
342 | ||
e1ee0a5b IS |
343 | unsigned int fib6_tables_seq_read(struct net *net) |
344 | { | |
345 | unsigned int h, fib_seq = 0; | |
346 | ||
347 | rcu_read_lock(); | |
348 | for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { | |
349 | struct hlist_head *head = &net->ipv6.fib_table_hash[h]; | |
350 | struct fib6_table *tb; | |
351 | ||
66f5d6ce | 352 | hlist_for_each_entry_rcu(tb, head, tb6_hlist) |
e1ee0a5b | 353 | fib_seq += tb->fib_seq; |
e1ee0a5b IS |
354 | } |
355 | rcu_read_unlock(); | |
356 | ||
357 | return fib_seq; | |
358 | } | |
359 | ||
360 | static int call_fib6_entry_notifier(struct notifier_block *nb, struct net *net, | |
361 | enum fib_event_type event_type, | |
8d1c802b | 362 | struct fib6_info *rt) |
e1ee0a5b IS |
363 | { |
364 | struct fib6_entry_notifier_info info = { | |
365 | .rt = rt, | |
366 | }; | |
367 | ||
368 | return call_fib6_notifier(nb, net, event_type, &info.info); | |
369 | } | |
370 | ||
19a3b7ee DA |
371 | int call_fib6_entry_notifiers(struct net *net, |
372 | enum fib_event_type event_type, | |
373 | struct fib6_info *rt, | |
374 | struct netlink_ext_ack *extack) | |
df77fe4d IS |
375 | { |
376 | struct fib6_entry_notifier_info info = { | |
6c31e5a9 | 377 | .info.extack = extack, |
df77fe4d IS |
378 | .rt = rt, |
379 | }; | |
380 | ||
93c2fb25 | 381 | rt->fib6_table->fib_seq++; |
df77fe4d IS |
382 | return call_fib6_notifiers(net, event_type, &info.info); |
383 | } | |
384 | ||
d4b96c7b IS |
385 | int call_fib6_multipath_entry_notifiers(struct net *net, |
386 | enum fib_event_type event_type, | |
387 | struct fib6_info *rt, | |
388 | unsigned int nsiblings, | |
389 | struct netlink_ext_ack *extack) | |
390 | { | |
391 | struct fib6_entry_notifier_info info = { | |
392 | .info.extack = extack, | |
393 | .rt = rt, | |
394 | .nsiblings = nsiblings, | |
d4b96c7b IS |
395 | }; |
396 | ||
397 | rt->fib6_table->fib_seq++; | |
398 | return call_fib6_notifiers(net, event_type, &info.info); | |
399 | } | |
400 | ||
e1ee0a5b IS |
401 | struct fib6_dump_arg { |
402 | struct net *net; | |
403 | struct notifier_block *nb; | |
404 | }; | |
405 | ||
8d1c802b | 406 | static void fib6_rt_dump(struct fib6_info *rt, struct fib6_dump_arg *arg) |
e1ee0a5b | 407 | { |
421842ed | 408 | if (rt == arg->net->ipv6.fib6_null_entry) |
e1ee0a5b IS |
409 | return; |
410 | call_fib6_entry_notifier(arg->nb, arg->net, FIB_EVENT_ENTRY_ADD, rt); | |
411 | } | |
412 | ||
413 | static int fib6_node_dump(struct fib6_walker *w) | |
414 | { | |
8d1c802b | 415 | struct fib6_info *rt; |
e1ee0a5b | 416 | |
66f5d6ce | 417 | for_each_fib6_walker_rt(w) |
e1ee0a5b IS |
418 | fib6_rt_dump(rt, w->args); |
419 | w->leaf = NULL; | |
420 | return 0; | |
421 | } | |
422 | ||
423 | static void fib6_table_dump(struct net *net, struct fib6_table *tb, | |
424 | struct fib6_walker *w) | |
425 | { | |
426 | w->root = &tb->tb6_root; | |
66f5d6ce | 427 | spin_lock_bh(&tb->tb6_lock); |
e1ee0a5b | 428 | fib6_walk(net, w); |
66f5d6ce | 429 | spin_unlock_bh(&tb->tb6_lock); |
e1ee0a5b IS |
430 | } |
431 | ||
432 | /* Called with rcu_read_lock() */ | |
433 | int fib6_tables_dump(struct net *net, struct notifier_block *nb) | |
434 | { | |
435 | struct fib6_dump_arg arg; | |
436 | struct fib6_walker *w; | |
437 | unsigned int h; | |
438 | ||
439 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
440 | if (!w) | |
441 | return -ENOMEM; | |
442 | ||
443 | w->func = fib6_node_dump; | |
444 | arg.net = net; | |
445 | arg.nb = nb; | |
446 | w->args = &arg; | |
447 | ||
448 | for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { | |
449 | struct hlist_head *head = &net->ipv6.fib_table_hash[h]; | |
450 | struct fib6_table *tb; | |
451 | ||
452 | hlist_for_each_entry_rcu(tb, head, tb6_hlist) | |
453 | fib6_table_dump(net, tb, w); | |
454 | } | |
455 | ||
456 | kfree(w); | |
457 | ||
458 | return 0; | |
459 | } | |
460 | ||
94b2cfe0 | 461 | static int fib6_dump_node(struct fib6_walker *w) |
1b43af54 PM |
462 | { |
463 | int res; | |
8d1c802b | 464 | struct fib6_info *rt; |
1b43af54 | 465 | |
66f5d6ce | 466 | for_each_fib6_walker_rt(w) { |
1e47b483 | 467 | res = rt6_dump_route(rt, w->args, w->skip_in_node); |
bf9a8a06 | 468 | if (res >= 0) { |
1b43af54 PM |
469 | /* Frame is full, suspend walking */ |
470 | w->leaf = rt; | |
1e47b483 SB |
471 | |
472 | /* We'll restart from this node, so if some routes were | |
473 | * already dumped, skip them next time. | |
474 | */ | |
475 | w->skip_in_node += res; | |
476 | ||
1b43af54 PM |
477 | return 1; |
478 | } | |
1e47b483 | 479 | w->skip_in_node = 0; |
beb1afac DA |
480 | |
481 | /* Multipath routes are dumped in one route with the | |
482 | * RTA_MULTIPATH attribute. Jump 'rt' to point to the | |
483 | * last sibling of this route (no need to dump the | |
484 | * sibling routes again) | |
485 | */ | |
93c2fb25 DA |
486 | if (rt->fib6_nsiblings) |
487 | rt = list_last_entry(&rt->fib6_siblings, | |
8d1c802b | 488 | struct fib6_info, |
93c2fb25 | 489 | fib6_siblings); |
1b43af54 PM |
490 | } |
491 | w->leaf = NULL; | |
492 | return 0; | |
493 | } | |
494 | ||
495 | static void fib6_dump_end(struct netlink_callback *cb) | |
496 | { | |
9a03cd8f | 497 | struct net *net = sock_net(cb->skb->sk); |
94b2cfe0 | 498 | struct fib6_walker *w = (void *)cb->args[2]; |
1b43af54 PM |
499 | |
500 | if (w) { | |
7891cc81 HX |
501 | if (cb->args[4]) { |
502 | cb->args[4] = 0; | |
9a03cd8f | 503 | fib6_walker_unlink(net, w); |
7891cc81 | 504 | } |
1b43af54 PM |
505 | cb->args[2] = 0; |
506 | kfree(w); | |
507 | } | |
437de07c | 508 | cb->done = (void *)cb->args[3]; |
1b43af54 PM |
509 | cb->args[1] = 3; |
510 | } | |
511 | ||
512 | static int fib6_dump_done(struct netlink_callback *cb) | |
513 | { | |
514 | fib6_dump_end(cb); | |
515 | return cb->done ? cb->done(cb) : 0; | |
516 | } | |
517 | ||
518 | static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb, | |
519 | struct netlink_callback *cb) | |
520 | { | |
9a03cd8f | 521 | struct net *net = sock_net(skb->sk); |
94b2cfe0 | 522 | struct fib6_walker *w; |
1b43af54 PM |
523 | int res; |
524 | ||
525 | w = (void *)cb->args[2]; | |
526 | w->root = &table->tb6_root; | |
527 | ||
528 | if (cb->args[4] == 0) { | |
2bec5a36 PM |
529 | w->count = 0; |
530 | w->skip = 0; | |
1e47b483 | 531 | w->skip_in_node = 0; |
2bec5a36 | 532 | |
66f5d6ce | 533 | spin_lock_bh(&table->tb6_lock); |
9a03cd8f | 534 | res = fib6_walk(net, w); |
66f5d6ce | 535 | spin_unlock_bh(&table->tb6_lock); |
2bec5a36 | 536 | if (res > 0) { |
1b43af54 | 537 | cb->args[4] = 1; |
2bec5a36 PM |
538 | cb->args[5] = w->root->fn_sernum; |
539 | } | |
1b43af54 | 540 | } else { |
2bec5a36 PM |
541 | if (cb->args[5] != w->root->fn_sernum) { |
542 | /* Begin at the root if the tree changed */ | |
543 | cb->args[5] = w->root->fn_sernum; | |
544 | w->state = FWS_INIT; | |
545 | w->node = w->root; | |
546 | w->skip = w->count; | |
1e47b483 | 547 | w->skip_in_node = 0; |
2bec5a36 PM |
548 | } else |
549 | w->skip = 0; | |
550 | ||
66f5d6ce | 551 | spin_lock_bh(&table->tb6_lock); |
1b43af54 | 552 | res = fib6_walk_continue(w); |
66f5d6ce | 553 | spin_unlock_bh(&table->tb6_lock); |
7891cc81 | 554 | if (res <= 0) { |
9a03cd8f | 555 | fib6_walker_unlink(net, w); |
7891cc81 | 556 | cb->args[4] = 0; |
1b43af54 | 557 | } |
1b43af54 | 558 | } |
7891cc81 | 559 | |
1b43af54 PM |
560 | return res; |
561 | } | |
562 | ||
c127ea2c | 563 | static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) |
1b43af54 | 564 | { |
564c91f7 SB |
565 | struct rt6_rtnl_dump_arg arg = { .filter.dump_exceptions = true, |
566 | .filter.dump_routes = true }; | |
e8ba330a | 567 | const struct nlmsghdr *nlh = cb->nlh; |
3b1e0a65 | 568 | struct net *net = sock_net(skb->sk); |
1b43af54 PM |
569 | unsigned int h, s_h; |
570 | unsigned int e = 0, s_e; | |
94b2cfe0 | 571 | struct fib6_walker *w; |
1b43af54 | 572 | struct fib6_table *tb; |
58f09b78 | 573 | struct hlist_head *head; |
1b43af54 PM |
574 | int res = 0; |
575 | ||
e8ba330a | 576 | if (cb->strict_check) { |
4724676d | 577 | int err; |
e8ba330a | 578 | |
effe6792 | 579 | err = ip_valid_fib_dump_req(net, nlh, &arg.filter, cb); |
e8ba330a DA |
580 | if (err < 0) |
581 | return err; | |
13e38901 DA |
582 | } else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) { |
583 | struct rtmsg *rtm = nlmsg_data(nlh); | |
e8ba330a | 584 | |
ef11209d SB |
585 | if (rtm->rtm_flags & RTM_F_PREFIX) |
586 | arg.filter.flags = RTM_F_PREFIX; | |
13e38901 | 587 | } |
1b43af54 PM |
588 | |
589 | w = (void *)cb->args[2]; | |
507c9b1e | 590 | if (!w) { |
1b43af54 PM |
591 | /* New dump: |
592 | * | |
593 | * 1. hook callback destructor. | |
594 | */ | |
595 | cb->args[3] = (long)cb->done; | |
596 | cb->done = fib6_dump_done; | |
597 | ||
598 | /* | |
599 | * 2. allocate and initialize walker. | |
600 | */ | |
601 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
507c9b1e | 602 | if (!w) |
1b43af54 PM |
603 | return -ENOMEM; |
604 | w->func = fib6_dump_node; | |
605 | cb->args[2] = (long)w; | |
606 | } | |
607 | ||
608 | arg.skb = skb; | |
609 | arg.cb = cb; | |
191cd582 | 610 | arg.net = net; |
1b43af54 PM |
611 | w->args = &arg; |
612 | ||
13e38901 DA |
613 | if (arg.filter.table_id) { |
614 | tb = fib6_get_table(net, arg.filter.table_id); | |
615 | if (!tb) { | |
ae677bbb | 616 | if (arg.filter.dump_all_families) |
e22d0bfa | 617 | goto out; |
ae677bbb | 618 | |
13e38901 DA |
619 | NL_SET_ERR_MSG_MOD(cb->extack, "FIB table does not exist"); |
620 | return -ENOENT; | |
621 | } | |
622 | ||
73155879 DA |
623 | if (!cb->args[0]) { |
624 | res = fib6_dump_table(tb, skb, cb); | |
625 | if (!res) | |
626 | cb->args[0] = 1; | |
627 | } | |
13e38901 DA |
628 | goto out; |
629 | } | |
630 | ||
631 | s_h = cb->args[0]; | |
632 | s_e = cb->args[1]; | |
633 | ||
e67f88dd | 634 | rcu_read_lock(); |
a33bc5c1 | 635 | for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) { |
1b43af54 | 636 | e = 0; |
58f09b78 | 637 | head = &net->ipv6.fib_table_hash[h]; |
b67bfe0d | 638 | hlist_for_each_entry_rcu(tb, head, tb6_hlist) { |
1b43af54 PM |
639 | if (e < s_e) |
640 | goto next; | |
641 | res = fib6_dump_table(tb, skb, cb); | |
642 | if (res != 0) | |
13e38901 | 643 | goto out_unlock; |
1b43af54 PM |
644 | next: |
645 | e++; | |
646 | } | |
647 | } | |
13e38901 | 648 | out_unlock: |
e67f88dd | 649 | rcu_read_unlock(); |
1b43af54 PM |
650 | cb->args[1] = e; |
651 | cb->args[0] = h; | |
13e38901 | 652 | out: |
1b43af54 PM |
653 | res = res < 0 ? res : skb->len; |
654 | if (res <= 0) | |
655 | fib6_dump_end(cb); | |
656 | return res; | |
657 | } | |
1da177e4 | 658 | |
8d1c802b | 659 | void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val) |
d4ead6b3 DA |
660 | { |
661 | if (!f6i) | |
662 | return; | |
663 | ||
664 | if (f6i->fib6_metrics == &dst_default_metrics) { | |
665 | struct dst_metrics *p = kzalloc(sizeof(*p), GFP_ATOMIC); | |
666 | ||
667 | if (!p) | |
668 | return; | |
669 | ||
670 | refcount_set(&p->refcnt, 1); | |
671 | f6i->fib6_metrics = p; | |
672 | } | |
673 | ||
674 | f6i->fib6_metrics->metrics[metric - 1] = val; | |
675 | } | |
676 | ||
1da177e4 LT |
677 | /* |
678 | * Routing Table | |
679 | * | |
680 | * return the appropriate node for a routing tree "add" operation | |
681 | * by either creating and inserting or by returning an existing | |
682 | * node. | |
683 | */ | |
684 | ||
81eb8447 WW |
685 | static struct fib6_node *fib6_add_1(struct net *net, |
686 | struct fib6_table *table, | |
66f5d6ce WW |
687 | struct fib6_node *root, |
688 | struct in6_addr *addr, int plen, | |
689 | int offset, int allow_create, | |
690 | int replace_required, | |
691 | struct netlink_ext_ack *extack) | |
1da177e4 LT |
692 | { |
693 | struct fib6_node *fn, *in, *ln; | |
694 | struct fib6_node *pn = NULL; | |
695 | struct rt6key *key; | |
696 | int bit; | |
1ab1457c | 697 | __be32 dir = 0; |
1da177e4 LT |
698 | |
699 | RT6_TRACE("fib6_add_1\n"); | |
700 | ||
701 | /* insert node in tree */ | |
702 | ||
703 | fn = root; | |
704 | ||
705 | do { | |
8d1c802b | 706 | struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, |
66f5d6ce WW |
707 | lockdep_is_held(&table->tb6_lock)); |
708 | key = (struct rt6key *)((u8 *)leaf + offset); | |
1da177e4 LT |
709 | |
710 | /* | |
711 | * Prefix match | |
712 | */ | |
713 | if (plen < fn->fn_bit || | |
4a287eba | 714 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) { |
14df015b MV |
715 | if (!allow_create) { |
716 | if (replace_required) { | |
d5d531cb DA |
717 | NL_SET_ERR_MSG(extack, |
718 | "Can not replace route - no match found"); | |
f3213831 | 719 | pr_warn("Can't replace route, no match found\n"); |
14df015b MV |
720 | return ERR_PTR(-ENOENT); |
721 | } | |
f3213831 | 722 | pr_warn("NLM_F_CREATE should be set when creating new route\n"); |
14df015b | 723 | } |
1da177e4 | 724 | goto insert_above; |
4a287eba | 725 | } |
1ab1457c | 726 | |
1da177e4 LT |
727 | /* |
728 | * Exact match ? | |
729 | */ | |
1ab1457c | 730 | |
1da177e4 LT |
731 | if (plen == fn->fn_bit) { |
732 | /* clean up an intermediate node */ | |
507c9b1e | 733 | if (!(fn->fn_flags & RTN_RTINFO)) { |
66f5d6ce | 734 | RCU_INIT_POINTER(fn->leaf, NULL); |
93531c67 | 735 | fib6_info_release(leaf); |
4512c43e WW |
736 | /* remove null_entry in the root node */ |
737 | } else if (fn->fn_flags & RTN_TL_ROOT && | |
738 | rcu_access_pointer(fn->leaf) == | |
421842ed | 739 | net->ipv6.fib6_null_entry) { |
4512c43e | 740 | RCU_INIT_POINTER(fn->leaf, NULL); |
1da177e4 | 741 | } |
1ab1457c | 742 | |
1da177e4 LT |
743 | return fn; |
744 | } | |
745 | ||
746 | /* | |
747 | * We have more bits to go | |
748 | */ | |
1ab1457c | 749 | |
1da177e4 | 750 | /* Try to walk down on tree. */ |
1da177e4 LT |
751 | dir = addr_bit_set(addr, fn->fn_bit); |
752 | pn = fn; | |
66f5d6ce WW |
753 | fn = dir ? |
754 | rcu_dereference_protected(fn->right, | |
755 | lockdep_is_held(&table->tb6_lock)) : | |
756 | rcu_dereference_protected(fn->left, | |
757 | lockdep_is_held(&table->tb6_lock)); | |
1da177e4 LT |
758 | } while (fn); |
759 | ||
14df015b | 760 | if (!allow_create) { |
4a287eba MV |
761 | /* We should not create new node because |
762 | * NLM_F_REPLACE was specified without NLM_F_CREATE | |
763 | * I assume it is safe to require NLM_F_CREATE when | |
764 | * REPLACE flag is used! Later we may want to remove the | |
765 | * check for replace_required, because according | |
766 | * to netlink specification, NLM_F_CREATE | |
767 | * MUST be specified if new route is created. | |
768 | * That would keep IPv6 consistent with IPv4 | |
769 | */ | |
14df015b | 770 | if (replace_required) { |
d5d531cb DA |
771 | NL_SET_ERR_MSG(extack, |
772 | "Can not replace route - no match found"); | |
f3213831 | 773 | pr_warn("Can't replace route, no match found\n"); |
14df015b MV |
774 | return ERR_PTR(-ENOENT); |
775 | } | |
f3213831 | 776 | pr_warn("NLM_F_CREATE should be set when creating new route\n"); |
4a287eba | 777 | } |
1da177e4 LT |
778 | /* |
779 | * We walked to the bottom of tree. | |
780 | * Create new leaf node without children. | |
781 | */ | |
782 | ||
81eb8447 | 783 | ln = node_alloc(net); |
1da177e4 | 784 | |
507c9b1e | 785 | if (!ln) |
188c517a | 786 | return ERR_PTR(-ENOMEM); |
1da177e4 | 787 | ln->fn_bit = plen; |
66f5d6ce | 788 | RCU_INIT_POINTER(ln->parent, pn); |
1da177e4 LT |
789 | |
790 | if (dir) | |
66f5d6ce | 791 | rcu_assign_pointer(pn->right, ln); |
1da177e4 | 792 | else |
66f5d6ce | 793 | rcu_assign_pointer(pn->left, ln); |
1da177e4 LT |
794 | |
795 | return ln; | |
796 | ||
797 | ||
798 | insert_above: | |
799 | /* | |
1ab1457c | 800 | * split since we don't have a common prefix anymore or |
1da177e4 LT |
801 | * we have a less significant route. |
802 | * we've to insert an intermediate node on the list | |
803 | * this new node will point to the one we need to create | |
804 | * and the current | |
805 | */ | |
806 | ||
66f5d6ce WW |
807 | pn = rcu_dereference_protected(fn->parent, |
808 | lockdep_is_held(&table->tb6_lock)); | |
1da177e4 LT |
809 | |
810 | /* find 1st bit in difference between the 2 addrs. | |
811 | ||
971f359d | 812 | See comment in __ipv6_addr_diff: bit may be an invalid value, |
1da177e4 LT |
813 | but if it is >= plen, the value is ignored in any case. |
814 | */ | |
1ab1457c | 815 | |
9225b230 | 816 | bit = __ipv6_addr_diff(addr, &key->addr, sizeof(*addr)); |
1da177e4 | 817 | |
1ab1457c YH |
818 | /* |
819 | * (intermediate)[in] | |
1da177e4 LT |
820 | * / \ |
821 | * (new leaf node)[ln] (old node)[fn] | |
822 | */ | |
823 | if (plen > bit) { | |
81eb8447 WW |
824 | in = node_alloc(net); |
825 | ln = node_alloc(net); | |
1ab1457c | 826 | |
507c9b1e | 827 | if (!in || !ln) { |
1da177e4 | 828 | if (in) |
81eb8447 | 829 | node_free_immediate(net, in); |
1da177e4 | 830 | if (ln) |
81eb8447 | 831 | node_free_immediate(net, ln); |
188c517a | 832 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
833 | } |
834 | ||
1ab1457c YH |
835 | /* |
836 | * new intermediate node. | |
1da177e4 LT |
837 | * RTN_RTINFO will |
838 | * be off since that an address that chooses one of | |
839 | * the branches would not match less specific routes | |
840 | * in the other branch | |
841 | */ | |
842 | ||
843 | in->fn_bit = bit; | |
844 | ||
66f5d6ce | 845 | RCU_INIT_POINTER(in->parent, pn); |
1da177e4 | 846 | in->leaf = fn->leaf; |
5ea71528 ED |
847 | fib6_info_hold(rcu_dereference_protected(in->leaf, |
848 | lockdep_is_held(&table->tb6_lock))); | |
1da177e4 | 849 | |
1da177e4 LT |
850 | /* update parent pointer */ |
851 | if (dir) | |
66f5d6ce | 852 | rcu_assign_pointer(pn->right, in); |
1da177e4 | 853 | else |
66f5d6ce | 854 | rcu_assign_pointer(pn->left, in); |
1da177e4 LT |
855 | |
856 | ln->fn_bit = plen; | |
857 | ||
66f5d6ce WW |
858 | RCU_INIT_POINTER(ln->parent, in); |
859 | rcu_assign_pointer(fn->parent, in); | |
1da177e4 | 860 | |
1da177e4 | 861 | if (addr_bit_set(addr, bit)) { |
66f5d6ce WW |
862 | rcu_assign_pointer(in->right, ln); |
863 | rcu_assign_pointer(in->left, fn); | |
1da177e4 | 864 | } else { |
66f5d6ce WW |
865 | rcu_assign_pointer(in->left, ln); |
866 | rcu_assign_pointer(in->right, fn); | |
1da177e4 LT |
867 | } |
868 | } else { /* plen <= bit */ | |
869 | ||
1ab1457c | 870 | /* |
1da177e4 LT |
871 | * (new leaf node)[ln] |
872 | * / \ | |
873 | * (old node)[fn] NULL | |
874 | */ | |
875 | ||
81eb8447 | 876 | ln = node_alloc(net); |
1da177e4 | 877 | |
507c9b1e | 878 | if (!ln) |
188c517a | 879 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
880 | |
881 | ln->fn_bit = plen; | |
882 | ||
66f5d6ce | 883 | RCU_INIT_POINTER(ln->parent, pn); |
1da177e4 LT |
884 | |
885 | if (addr_bit_set(&key->addr, plen)) | |
66f5d6ce | 886 | RCU_INIT_POINTER(ln->right, fn); |
1da177e4 | 887 | else |
66f5d6ce WW |
888 | RCU_INIT_POINTER(ln->left, fn); |
889 | ||
890 | rcu_assign_pointer(fn->parent, ln); | |
1da177e4 | 891 | |
66f5d6ce WW |
892 | if (dir) |
893 | rcu_assign_pointer(pn->right, ln); | |
894 | else | |
895 | rcu_assign_pointer(pn->left, ln); | |
1da177e4 LT |
896 | } |
897 | return ln; | |
898 | } | |
899 | ||
7d88d8b5 DA |
900 | static void __fib6_drop_pcpu_from(struct fib6_nh *fib6_nh, |
901 | const struct fib6_info *match, | |
902 | const struct fib6_table *table) | |
e715b6d3 | 903 | { |
5bcaa41b | 904 | int cpu; |
e715b6d3 | 905 | |
f40b6ae2 DA |
906 | if (!fib6_nh->rt6i_pcpu) |
907 | return; | |
908 | ||
5bcaa41b DA |
909 | /* release the reference to this fib entry from |
910 | * all of its cached pcpu routes | |
911 | */ | |
912 | for_each_possible_cpu(cpu) { | |
913 | struct rt6_info **ppcpu_rt; | |
914 | struct rt6_info *pcpu_rt; | |
e715b6d3 | 915 | |
f40b6ae2 | 916 | ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu); |
5bcaa41b | 917 | pcpu_rt = *ppcpu_rt; |
7d88d8b5 DA |
918 | |
919 | /* only dropping the 'from' reference if the cached route | |
920 | * is using 'match'. The cached pcpu_rt->from only changes | |
921 | * from a fib6_info to NULL (ip6_dst_destroy); it can never | |
922 | * change from one fib6_info reference to another | |
923 | */ | |
924 | if (pcpu_rt && rcu_access_pointer(pcpu_rt->from) == match) { | |
a68886a6 | 925 | struct fib6_info *from; |
e715b6d3 | 926 | |
0e233874 | 927 | from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL); |
a68886a6 | 928 | fib6_info_release(from); |
5bcaa41b | 929 | } |
e5fd387a | 930 | } |
e5fd387a MK |
931 | } |
932 | ||
2ab75bfb DA |
933 | struct fib6_nh_pcpu_arg { |
934 | struct fib6_info *from; | |
935 | const struct fib6_table *table; | |
936 | }; | |
937 | ||
938 | static int fib6_nh_drop_pcpu_from(struct fib6_nh *nh, void *_arg) | |
939 | { | |
940 | struct fib6_nh_pcpu_arg *arg = _arg; | |
941 | ||
942 | __fib6_drop_pcpu_from(nh, arg->from, arg->table); | |
943 | return 0; | |
944 | } | |
945 | ||
7d88d8b5 DA |
946 | static void fib6_drop_pcpu_from(struct fib6_info *f6i, |
947 | const struct fib6_table *table) | |
948 | { | |
7d88d8b5 DA |
949 | /* Make sure rt6_make_pcpu_route() wont add other percpu routes |
950 | * while we are cleaning them here. | |
951 | */ | |
952 | f6i->fib6_destroying = 1; | |
953 | mb(); /* paired with the cmpxchg() in rt6_make_pcpu_route() */ | |
954 | ||
2ab75bfb DA |
955 | if (f6i->nh) { |
956 | struct fib6_nh_pcpu_arg arg = { | |
957 | .from = f6i, | |
958 | .table = table | |
959 | }; | |
960 | ||
961 | nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_drop_pcpu_from, | |
962 | &arg); | |
963 | } else { | |
964 | struct fib6_nh *fib6_nh; | |
965 | ||
966 | fib6_nh = f6i->fib6_nh; | |
967 | __fib6_drop_pcpu_from(fib6_nh, f6i, table); | |
968 | } | |
7d88d8b5 DA |
969 | } |
970 | ||
8d1c802b | 971 | static void fib6_purge_rt(struct fib6_info *rt, struct fib6_node *fn, |
6e9e16e6 HFS |
972 | struct net *net) |
973 | { | |
93c2fb25 | 974 | struct fib6_table *table = rt->fib6_table; |
66f5d6ce | 975 | |
f40b6ae2 | 976 | fib6_drop_pcpu_from(rt, table); |
61fb0d01 | 977 | |
f88d8ea6 DA |
978 | if (rt->nh && !list_empty(&rt->nh_list)) |
979 | list_del_init(&rt->nh_list); | |
980 | ||
f05713e0 | 981 | if (refcount_read(&rt->fib6_ref) != 1) { |
6e9e16e6 HFS |
982 | /* This route is used as dummy address holder in some split |
983 | * nodes. It is not leaked, but it still holds other resources, | |
984 | * which must be released in time. So, scan ascendant nodes | |
985 | * and replace dummy references to this route with references | |
986 | * to still alive ones. | |
987 | */ | |
988 | while (fn) { | |
8d1c802b | 989 | struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, |
66f5d6ce | 990 | lockdep_is_held(&table->tb6_lock)); |
8d1c802b | 991 | struct fib6_info *new_leaf; |
66f5d6ce WW |
992 | if (!(fn->fn_flags & RTN_RTINFO) && leaf == rt) { |
993 | new_leaf = fib6_find_prefix(net, table, fn); | |
5ea71528 | 994 | fib6_info_hold(new_leaf); |
93531c67 | 995 | |
66f5d6ce | 996 | rcu_assign_pointer(fn->leaf, new_leaf); |
93531c67 | 997 | fib6_info_release(rt); |
6e9e16e6 | 998 | } |
66f5d6ce WW |
999 | fn = rcu_dereference_protected(fn->parent, |
1000 | lockdep_is_held(&table->tb6_lock)); | |
6e9e16e6 | 1001 | } |
6e9e16e6 HFS |
1002 | } |
1003 | } | |
1004 | ||
1da177e4 LT |
1005 | /* |
1006 | * Insert routing information in a node. | |
1007 | */ | |
1008 | ||
8d1c802b | 1009 | static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt, |
d4ead6b3 | 1010 | struct nl_info *info, |
6c31e5a9 | 1011 | struct netlink_ext_ack *extack) |
1da177e4 | 1012 | { |
8d1c802b | 1013 | struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, |
93c2fb25 | 1014 | lockdep_is_held(&rt->fib6_table->tb6_lock)); |
33bd5ac5 | 1015 | struct fib6_info *iter = NULL; |
8d1c802b | 1016 | struct fib6_info __rcu **ins; |
33bd5ac5 | 1017 | struct fib6_info __rcu **fallback_ins = NULL; |
507c9b1e DM |
1018 | int replace = (info->nlh && |
1019 | (info->nlh->nlmsg_flags & NLM_F_REPLACE)); | |
1020 | int add = (!info->nlh || | |
1021 | (info->nlh->nlmsg_flags & NLM_F_CREATE)); | |
4a287eba | 1022 | int found = 0; |
33bd5ac5 | 1023 | bool rt_can_ecmp = rt6_qualify_for_ecmp(rt); |
73483c12 | 1024 | u16 nlflags = NLM_F_EXCL; |
e5fd387a | 1025 | int err; |
1da177e4 | 1026 | |
33bd5ac5 | 1027 | if (info->nlh && (info->nlh->nlmsg_flags & NLM_F_APPEND)) |
1f5e29ce DA |
1028 | nlflags |= NLM_F_APPEND; |
1029 | ||
1da177e4 LT |
1030 | ins = &fn->leaf; |
1031 | ||
66f5d6ce | 1032 | for (iter = leaf; iter; |
8fb11a9a | 1033 | iter = rcu_dereference_protected(iter->fib6_next, |
93c2fb25 | 1034 | lockdep_is_held(&rt->fib6_table->tb6_lock))) { |
1da177e4 LT |
1035 | /* |
1036 | * Search for duplicates | |
1037 | */ | |
1038 | ||
93c2fb25 | 1039 | if (iter->fib6_metric == rt->fib6_metric) { |
1da177e4 LT |
1040 | /* |
1041 | * Same priority level | |
1042 | */ | |
507c9b1e DM |
1043 | if (info->nlh && |
1044 | (info->nlh->nlmsg_flags & NLM_F_EXCL)) | |
4a287eba | 1045 | return -EEXIST; |
73483c12 GN |
1046 | |
1047 | nlflags &= ~NLM_F_EXCL; | |
4a287eba | 1048 | if (replace) { |
33bd5ac5 DA |
1049 | if (rt_can_ecmp == rt6_qualify_for_ecmp(iter)) { |
1050 | found++; | |
1051 | break; | |
1052 | } | |
1053 | if (rt_can_ecmp) | |
1054 | fallback_ins = fallback_ins ?: ins; | |
1055 | goto next_iter; | |
4a287eba | 1056 | } |
1da177e4 | 1057 | |
f06b7549 | 1058 | if (rt6_duplicate_nexthop(iter, rt)) { |
93c2fb25 DA |
1059 | if (rt->fib6_nsiblings) |
1060 | rt->fib6_nsiblings = 0; | |
1061 | if (!(iter->fib6_flags & RTF_EXPIRES)) | |
1da177e4 | 1062 | return -EEXIST; |
93c2fb25 | 1063 | if (!(rt->fib6_flags & RTF_EXPIRES)) |
14895687 | 1064 | fib6_clean_expires(iter); |
1716a961 | 1065 | else |
14895687 | 1066 | fib6_set_expires(iter, rt->expires); |
15a81b41 DA |
1067 | |
1068 | if (rt->fib6_pmtu) | |
1069 | fib6_metric_set(iter, RTAX_MTU, | |
1070 | rt->fib6_pmtu); | |
1da177e4 LT |
1071 | return -EEXIST; |
1072 | } | |
33bd5ac5 DA |
1073 | /* If we have the same destination and the same metric, |
1074 | * but not the same gateway, then the route we try to | |
1075 | * add is sibling to this route, increment our counter | |
1076 | * of siblings, and later we will add our route to the | |
1077 | * list. | |
1078 | * Only static routes (which don't have flag | |
1079 | * RTF_EXPIRES) are used for ECMPv6. | |
1080 | * | |
1081 | * To avoid long list, we only had siblings if the | |
1082 | * route have a gateway. | |
1083 | */ | |
1084 | if (rt_can_ecmp && | |
1085 | rt6_qualify_for_ecmp(iter)) | |
1086 | rt->fib6_nsiblings++; | |
1da177e4 LT |
1087 | } |
1088 | ||
93c2fb25 | 1089 | if (iter->fib6_metric > rt->fib6_metric) |
1da177e4 LT |
1090 | break; |
1091 | ||
33bd5ac5 | 1092 | next_iter: |
8fb11a9a | 1093 | ins = &iter->fib6_next; |
27596472 MK |
1094 | } |
1095 | ||
33bd5ac5 DA |
1096 | if (fallback_ins && !found) { |
1097 | /* No ECMP-able route found, replace first non-ECMP one */ | |
1098 | ins = fallback_ins; | |
1099 | iter = rcu_dereference_protected(*ins, | |
1100 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
1101 | found++; | |
1102 | } | |
1103 | ||
f11e6659 DM |
1104 | /* Reset round-robin state, if necessary */ |
1105 | if (ins == &fn->leaf) | |
1106 | fn->rr_ptr = NULL; | |
1107 | ||
51ebd318 | 1108 | /* Link this route to others same route. */ |
33bd5ac5 DA |
1109 | if (rt->fib6_nsiblings) { |
1110 | unsigned int fib6_nsiblings; | |
8d1c802b | 1111 | struct fib6_info *sibling, *temp_sibling; |
51ebd318 | 1112 | |
33bd5ac5 DA |
1113 | /* Find the first route that have the same metric */ |
1114 | sibling = leaf; | |
1115 | while (sibling) { | |
1116 | if (sibling->fib6_metric == rt->fib6_metric && | |
1117 | rt6_qualify_for_ecmp(sibling)) { | |
1118 | list_add_tail(&rt->fib6_siblings, | |
1119 | &sibling->fib6_siblings); | |
1120 | break; | |
1121 | } | |
1122 | sibling = rcu_dereference_protected(sibling->fib6_next, | |
1123 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
51ebd318 ND |
1124 | } |
1125 | /* For each sibling in the list, increment the counter of | |
1126 | * siblings. BUG() if counters does not match, list of siblings | |
1127 | * is broken! | |
1128 | */ | |
33bd5ac5 | 1129 | fib6_nsiblings = 0; |
51ebd318 | 1130 | list_for_each_entry_safe(sibling, temp_sibling, |
33bd5ac5 | 1131 | &rt->fib6_siblings, fib6_siblings) { |
93c2fb25 | 1132 | sibling->fib6_nsiblings++; |
33bd5ac5 DA |
1133 | BUG_ON(sibling->fib6_nsiblings != rt->fib6_nsiblings); |
1134 | fib6_nsiblings++; | |
51ebd318 | 1135 | } |
33bd5ac5 DA |
1136 | BUG_ON(fib6_nsiblings != rt->fib6_nsiblings); |
1137 | rt6_multipath_rebalance(temp_sibling); | |
51ebd318 ND |
1138 | } |
1139 | ||
1da177e4 LT |
1140 | /* |
1141 | * insert node | |
1142 | */ | |
4a287eba MV |
1143 | if (!replace) { |
1144 | if (!add) | |
f3213831 | 1145 | pr_warn("NLM_F_CREATE should be set when creating new route\n"); |
4a287eba MV |
1146 | |
1147 | add: | |
73483c12 | 1148 | nlflags |= NLM_F_CREATE; |
e715b6d3 | 1149 | |
d5382fef IS |
1150 | if (!info->skip_notify_kernel) { |
1151 | err = call_fib6_entry_notifiers(info->nl_net, | |
1152 | FIB_EVENT_ENTRY_ADD, | |
1153 | rt, extack); | |
54851aa9 IS |
1154 | if (err) { |
1155 | struct fib6_info *sibling, *next_sibling; | |
1156 | ||
1157 | /* If the route has siblings, then it first | |
1158 | * needs to be unlinked from them. | |
1159 | */ | |
1160 | if (!rt->fib6_nsiblings) | |
1161 | return err; | |
1162 | ||
1163 | list_for_each_entry_safe(sibling, next_sibling, | |
1164 | &rt->fib6_siblings, | |
1165 | fib6_siblings) | |
1166 | sibling->fib6_nsiblings--; | |
1167 | rt->fib6_nsiblings = 0; | |
1168 | list_del_init(&rt->fib6_siblings); | |
1169 | rt6_multipath_rebalance(next_sibling); | |
d5382fef | 1170 | return err; |
54851aa9 | 1171 | } |
d5382fef | 1172 | } |
2233000c | 1173 | |
8fb11a9a | 1174 | rcu_assign_pointer(rt->fib6_next, iter); |
5ea71528 | 1175 | fib6_info_hold(rt); |
93c2fb25 | 1176 | rcu_assign_pointer(rt->fib6_node, fn); |
66f5d6ce | 1177 | rcu_assign_pointer(*ins, rt); |
3b1137fe DA |
1178 | if (!info->skip_notify) |
1179 | inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags); | |
4a287eba MV |
1180 | info->nl_net->ipv6.rt6_stats->fib_rt_entries++; |
1181 | ||
507c9b1e | 1182 | if (!(fn->fn_flags & RTN_RTINFO)) { |
4a287eba MV |
1183 | info->nl_net->ipv6.rt6_stats->fib_route_nodes++; |
1184 | fn->fn_flags |= RTN_RTINFO; | |
1185 | } | |
1da177e4 | 1186 | |
4a287eba | 1187 | } else { |
33bd5ac5 | 1188 | int nsiblings; |
27596472 | 1189 | |
4a287eba MV |
1190 | if (!found) { |
1191 | if (add) | |
1192 | goto add; | |
f3213831 | 1193 | pr_warn("NLM_F_REPLACE set, but no existing node found!\n"); |
4a287eba MV |
1194 | return -ENOENT; |
1195 | } | |
e715b6d3 | 1196 | |
d5382fef IS |
1197 | if (!info->skip_notify_kernel) { |
1198 | err = call_fib6_entry_notifiers(info->nl_net, | |
1199 | FIB_EVENT_ENTRY_REPLACE, | |
1200 | rt, extack); | |
1201 | if (err) | |
1202 | return err; | |
1203 | } | |
2233000c | 1204 | |
5ea71528 | 1205 | fib6_info_hold(rt); |
93c2fb25 | 1206 | rcu_assign_pointer(rt->fib6_node, fn); |
33bd5ac5 | 1207 | rt->fib6_next = iter->fib6_next; |
66f5d6ce | 1208 | rcu_assign_pointer(*ins, rt); |
3b1137fe DA |
1209 | if (!info->skip_notify) |
1210 | inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE); | |
507c9b1e | 1211 | if (!(fn->fn_flags & RTN_RTINFO)) { |
4a287eba MV |
1212 | info->nl_net->ipv6.rt6_stats->fib_route_nodes++; |
1213 | fn->fn_flags |= RTN_RTINFO; | |
1214 | } | |
33bd5ac5 DA |
1215 | nsiblings = iter->fib6_nsiblings; |
1216 | iter->fib6_node = NULL; | |
1217 | fib6_purge_rt(iter, fn, info->nl_net); | |
1218 | if (rcu_access_pointer(fn->rr_ptr) == iter) | |
1219 | fn->rr_ptr = NULL; | |
1220 | fib6_info_release(iter); | |
27596472 | 1221 | |
33bd5ac5 | 1222 | if (nsiblings) { |
27596472 | 1223 | /* Replacing an ECMP route, remove all siblings */ |
33bd5ac5 DA |
1224 | ins = &rt->fib6_next; |
1225 | iter = rcu_dereference_protected(*ins, | |
1226 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
1227 | while (iter) { | |
1228 | if (iter->fib6_metric > rt->fib6_metric) | |
1229 | break; | |
1230 | if (rt6_qualify_for_ecmp(iter)) { | |
1231 | *ins = iter->fib6_next; | |
1232 | iter->fib6_node = NULL; | |
1233 | fib6_purge_rt(iter, fn, info->nl_net); | |
1234 | if (rcu_access_pointer(fn->rr_ptr) == iter) | |
1235 | fn->rr_ptr = NULL; | |
1236 | fib6_info_release(iter); | |
1237 | nsiblings--; | |
1238 | info->nl_net->ipv6.rt6_stats->fib_rt_entries--; | |
1239 | } else { | |
1240 | ins = &iter->fib6_next; | |
1241 | } | |
1242 | iter = rcu_dereference_protected(*ins, | |
1243 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
27596472 | 1244 | } |
33bd5ac5 | 1245 | WARN_ON(nsiblings != 0); |
27596472 | 1246 | } |
1da177e4 LT |
1247 | } |
1248 | ||
1249 | return 0; | |
1250 | } | |
1251 | ||
8d1c802b | 1252 | static void fib6_start_gc(struct net *net, struct fib6_info *rt) |
1da177e4 | 1253 | { |
417f28bb | 1254 | if (!timer_pending(&net->ipv6.ip6_fib_timer) && |
93c2fb25 | 1255 | (rt->fib6_flags & RTF_EXPIRES)) |
417f28bb | 1256 | mod_timer(&net->ipv6.ip6_fib_timer, |
847499ce | 1257 | jiffies + net->ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 LT |
1258 | } |
1259 | ||
63152fc0 | 1260 | void fib6_force_start_gc(struct net *net) |
1da177e4 | 1261 | { |
417f28bb SH |
1262 | if (!timer_pending(&net->ipv6.ip6_fib_timer)) |
1263 | mod_timer(&net->ipv6.ip6_fib_timer, | |
847499ce | 1264 | jiffies + net->ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 LT |
1265 | } |
1266 | ||
8d1c802b | 1267 | static void __fib6_update_sernum_upto_root(struct fib6_info *rt, |
4a8e56ee | 1268 | int sernum) |
bbd63f06 | 1269 | { |
93c2fb25 DA |
1270 | struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node, |
1271 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
bbd63f06 WW |
1272 | |
1273 | /* paired with smp_rmb() in rt6_get_cookie_safe() */ | |
1274 | smp_wmb(); | |
1275 | while (fn) { | |
1276 | fn->fn_sernum = sernum; | |
66f5d6ce | 1277 | fn = rcu_dereference_protected(fn->parent, |
93c2fb25 | 1278 | lockdep_is_held(&rt->fib6_table->tb6_lock)); |
bbd63f06 WW |
1279 | } |
1280 | } | |
1281 | ||
8d1c802b | 1282 | void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt) |
4a8e56ee IS |
1283 | { |
1284 | __fib6_update_sernum_upto_root(rt, fib6_new_sernum(net)); | |
1285 | } | |
1286 | ||
cdaa16a4 DA |
1287 | /* allow ipv4 to update sernum via ipv6_stub */ |
1288 | void fib6_update_sernum_stub(struct net *net, struct fib6_info *f6i) | |
1289 | { | |
1290 | spin_lock_bh(&f6i->fib6_table->tb6_lock); | |
1291 | fib6_update_sernum_upto_root(net, f6i); | |
1292 | spin_unlock_bh(&f6i->fib6_table->tb6_lock); | |
1293 | } | |
1294 | ||
1da177e4 LT |
1295 | /* |
1296 | * Add routing information to the routing tree. | |
1297 | * <destination addr>/<source addr> | |
1298 | * with source addr info in sub-trees | |
66f5d6ce | 1299 | * Need to own table->tb6_lock |
1da177e4 LT |
1300 | */ |
1301 | ||
8d1c802b | 1302 | int fib6_add(struct fib6_node *root, struct fib6_info *rt, |
d4ead6b3 | 1303 | struct nl_info *info, struct netlink_ext_ack *extack) |
1da177e4 | 1304 | { |
93c2fb25 | 1305 | struct fib6_table *table = rt->fib6_table; |
66729e18 | 1306 | struct fib6_node *fn, *pn = NULL; |
1da177e4 | 1307 | int err = -ENOMEM; |
4a287eba MV |
1308 | int allow_create = 1; |
1309 | int replace_required = 0; | |
812918c4 | 1310 | int sernum = fib6_new_sernum(info->nl_net); |
507c9b1e DM |
1311 | |
1312 | if (info->nlh) { | |
1313 | if (!(info->nlh->nlmsg_flags & NLM_F_CREATE)) | |
4a287eba | 1314 | allow_create = 0; |
507c9b1e | 1315 | if (info->nlh->nlmsg_flags & NLM_F_REPLACE) |
4a287eba MV |
1316 | replace_required = 1; |
1317 | } | |
1318 | if (!allow_create && !replace_required) | |
f3213831 | 1319 | pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n"); |
1da177e4 | 1320 | |
81eb8447 | 1321 | fn = fib6_add_1(info->nl_net, table, root, |
93c2fb25 DA |
1322 | &rt->fib6_dst.addr, rt->fib6_dst.plen, |
1323 | offsetof(struct fib6_info, fib6_dst), allow_create, | |
bbd63f06 | 1324 | replace_required, extack); |
4a287eba MV |
1325 | if (IS_ERR(fn)) { |
1326 | err = PTR_ERR(fn); | |
ae7b4e1f | 1327 | fn = NULL; |
1da177e4 | 1328 | goto out; |
188c517a | 1329 | } |
1da177e4 | 1330 | |
66729e18 YH |
1331 | pn = fn; |
1332 | ||
1da177e4 | 1333 | #ifdef CONFIG_IPV6_SUBTREES |
93c2fb25 | 1334 | if (rt->fib6_src.plen) { |
1da177e4 LT |
1335 | struct fib6_node *sn; |
1336 | ||
66f5d6ce | 1337 | if (!rcu_access_pointer(fn->subtree)) { |
1da177e4 LT |
1338 | struct fib6_node *sfn; |
1339 | ||
1340 | /* | |
1341 | * Create subtree. | |
1342 | * | |
1343 | * fn[main tree] | |
1344 | * | | |
1345 | * sfn[subtree root] | |
1346 | * \ | |
1347 | * sn[new leaf node] | |
1348 | */ | |
1349 | ||
1350 | /* Create subtree root node */ | |
81eb8447 | 1351 | sfn = node_alloc(info->nl_net); |
507c9b1e | 1352 | if (!sfn) |
348a4002 | 1353 | goto failure; |
1da177e4 | 1354 | |
5ea71528 | 1355 | fib6_info_hold(info->nl_net->ipv6.fib6_null_entry); |
66f5d6ce | 1356 | rcu_assign_pointer(sfn->leaf, |
421842ed | 1357 | info->nl_net->ipv6.fib6_null_entry); |
1da177e4 | 1358 | sfn->fn_flags = RTN_ROOT; |
1da177e4 LT |
1359 | |
1360 | /* Now add the first leaf node to new subtree */ | |
1361 | ||
81eb8447 | 1362 | sn = fib6_add_1(info->nl_net, table, sfn, |
93c2fb25 DA |
1363 | &rt->fib6_src.addr, rt->fib6_src.plen, |
1364 | offsetof(struct fib6_info, fib6_src), | |
bbd63f06 | 1365 | allow_create, replace_required, extack); |
1da177e4 | 1366 | |
f950c0ec | 1367 | if (IS_ERR(sn)) { |
1da177e4 | 1368 | /* If it is failed, discard just allocated |
348a4002 | 1369 | root, and then (in failure) stale node |
1da177e4 LT |
1370 | in main tree. |
1371 | */ | |
81eb8447 | 1372 | node_free_immediate(info->nl_net, sfn); |
188c517a | 1373 | err = PTR_ERR(sn); |
348a4002 | 1374 | goto failure; |
1da177e4 LT |
1375 | } |
1376 | ||
1377 | /* Now link new subtree to main tree */ | |
66f5d6ce WW |
1378 | rcu_assign_pointer(sfn->parent, fn); |
1379 | rcu_assign_pointer(fn->subtree, sfn); | |
1da177e4 | 1380 | } else { |
81eb8447 | 1381 | sn = fib6_add_1(info->nl_net, table, FIB6_SUBTREE(fn), |
93c2fb25 DA |
1382 | &rt->fib6_src.addr, rt->fib6_src.plen, |
1383 | offsetof(struct fib6_info, fib6_src), | |
bbd63f06 | 1384 | allow_create, replace_required, extack); |
1da177e4 | 1385 | |
4a287eba MV |
1386 | if (IS_ERR(sn)) { |
1387 | err = PTR_ERR(sn); | |
348a4002 | 1388 | goto failure; |
188c517a | 1389 | } |
1da177e4 LT |
1390 | } |
1391 | ||
66f5d6ce | 1392 | if (!rcu_access_pointer(fn->leaf)) { |
591ff9ea WW |
1393 | if (fn->fn_flags & RTN_TL_ROOT) { |
1394 | /* put back null_entry for root node */ | |
1395 | rcu_assign_pointer(fn->leaf, | |
421842ed | 1396 | info->nl_net->ipv6.fib6_null_entry); |
591ff9ea | 1397 | } else { |
5ea71528 | 1398 | fib6_info_hold(rt); |
591ff9ea WW |
1399 | rcu_assign_pointer(fn->leaf, rt); |
1400 | } | |
66729e18 | 1401 | } |
1da177e4 LT |
1402 | fn = sn; |
1403 | } | |
1404 | #endif | |
1405 | ||
d4ead6b3 | 1406 | err = fib6_add_rt2node(fn, rt, info, extack); |
bbd63f06 | 1407 | if (!err) { |
f88d8ea6 DA |
1408 | if (rt->nh) |
1409 | list_add(&rt->nh_list, &rt->nh->f6i_list); | |
4a8e56ee | 1410 | __fib6_update_sernum_upto_root(rt, sernum); |
63152fc0 | 1411 | fib6_start_gc(info->nl_net, rt); |
bbd63f06 | 1412 | } |
1da177e4 LT |
1413 | |
1414 | out: | |
66729e18 YH |
1415 | if (err) { |
1416 | #ifdef CONFIG_IPV6_SUBTREES | |
1417 | /* | |
1418 | * If fib6_add_1 has cleared the old leaf pointer in the | |
1419 | * super-tree leaf node we have to find a new one for it. | |
1420 | */ | |
7bbfe00e | 1421 | if (pn != fn) { |
8d1c802b | 1422 | struct fib6_info *pn_leaf = |
7bbfe00e WW |
1423 | rcu_dereference_protected(pn->leaf, |
1424 | lockdep_is_held(&table->tb6_lock)); | |
1425 | if (pn_leaf == rt) { | |
1426 | pn_leaf = NULL; | |
1427 | RCU_INIT_POINTER(pn->leaf, NULL); | |
93531c67 | 1428 | fib6_info_release(rt); |
66729e18 | 1429 | } |
7bbfe00e WW |
1430 | if (!pn_leaf && !(pn->fn_flags & RTN_RTINFO)) { |
1431 | pn_leaf = fib6_find_prefix(info->nl_net, table, | |
1432 | pn); | |
1433 | #if RT6_DEBUG >= 2 | |
1434 | if (!pn_leaf) { | |
1435 | WARN_ON(!pn_leaf); | |
1436 | pn_leaf = | |
421842ed | 1437 | info->nl_net->ipv6.fib6_null_entry; |
7bbfe00e | 1438 | } |
66729e18 | 1439 | #endif |
93531c67 | 1440 | fib6_info_hold(pn_leaf); |
7bbfe00e WW |
1441 | rcu_assign_pointer(pn->leaf, pn_leaf); |
1442 | } | |
66729e18 YH |
1443 | } |
1444 | #endif | |
348a4002 | 1445 | goto failure; |
66729e18 | 1446 | } |
1da177e4 LT |
1447 | return err; |
1448 | ||
348a4002 | 1449 | failure: |
4512c43e WW |
1450 | /* fn->leaf could be NULL and fib6_repair_tree() needs to be called if: |
1451 | * 1. fn is an intermediate node and we failed to add the new | |
1452 | * route to it in both subtree creation failure and fib6_add_rt2node() | |
1453 | * failure case. | |
1454 | * 2. fn is the root node in the table and we fail to add the first | |
1455 | * default route to it. | |
1da177e4 | 1456 | */ |
4512c43e WW |
1457 | if (fn && |
1458 | (!(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)) || | |
1459 | (fn->fn_flags & RTN_TL_ROOT && | |
1460 | !rcu_access_pointer(fn->leaf)))) | |
66f5d6ce | 1461 | fib6_repair_tree(info->nl_net, table, fn); |
1da177e4 | 1462 | return err; |
1da177e4 LT |
1463 | } |
1464 | ||
1465 | /* | |
1466 | * Routing tree lookup | |
1467 | * | |
1468 | */ | |
1469 | ||
1470 | struct lookup_args { | |
8d1c802b | 1471 | int offset; /* key offset on fib6_info */ |
b71d1d42 | 1472 | const struct in6_addr *addr; /* search key */ |
1da177e4 LT |
1473 | }; |
1474 | ||
6454743b DA |
1475 | static struct fib6_node *fib6_node_lookup_1(struct fib6_node *root, |
1476 | struct lookup_args *args) | |
1da177e4 LT |
1477 | { |
1478 | struct fib6_node *fn; | |
e69a4adc | 1479 | __be32 dir; |
1da177e4 | 1480 | |
825e288e YH |
1481 | if (unlikely(args->offset == 0)) |
1482 | return NULL; | |
1483 | ||
1da177e4 LT |
1484 | /* |
1485 | * Descend on a tree | |
1486 | */ | |
1487 | ||
1488 | fn = root; | |
1489 | ||
1490 | for (;;) { | |
1491 | struct fib6_node *next; | |
1492 | ||
1493 | dir = addr_bit_set(args->addr, fn->fn_bit); | |
1494 | ||
66f5d6ce WW |
1495 | next = dir ? rcu_dereference(fn->right) : |
1496 | rcu_dereference(fn->left); | |
1da177e4 LT |
1497 | |
1498 | if (next) { | |
1499 | fn = next; | |
1500 | continue; | |
1501 | } | |
1da177e4 LT |
1502 | break; |
1503 | } | |
1504 | ||
507c9b1e | 1505 | while (fn) { |
66f5d6ce WW |
1506 | struct fib6_node *subtree = FIB6_SUBTREE(fn); |
1507 | ||
1508 | if (subtree || fn->fn_flags & RTN_RTINFO) { | |
8d1c802b | 1509 | struct fib6_info *leaf = rcu_dereference(fn->leaf); |
1da177e4 LT |
1510 | struct rt6key *key; |
1511 | ||
8d1040e8 WW |
1512 | if (!leaf) |
1513 | goto backtrack; | |
1514 | ||
1515 | key = (struct rt6key *) ((u8 *)leaf + args->offset); | |
1da177e4 | 1516 | |
3fc5e044 YH |
1517 | if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) { |
1518 | #ifdef CONFIG_IPV6_SUBTREES | |
66f5d6ce | 1519 | if (subtree) { |
3e3be275 | 1520 | struct fib6_node *sfn; |
6454743b DA |
1521 | sfn = fib6_node_lookup_1(subtree, |
1522 | args + 1); | |
3e3be275 HFS |
1523 | if (!sfn) |
1524 | goto backtrack; | |
1525 | fn = sfn; | |
1526 | } | |
3fc5e044 | 1527 | #endif |
3e3be275 | 1528 | if (fn->fn_flags & RTN_RTINFO) |
3fc5e044 YH |
1529 | return fn; |
1530 | } | |
1da177e4 | 1531 | } |
3e3be275 | 1532 | backtrack: |
3fc5e044 YH |
1533 | if (fn->fn_flags & RTN_ROOT) |
1534 | break; | |
1535 | ||
66f5d6ce | 1536 | fn = rcu_dereference(fn->parent); |
1da177e4 LT |
1537 | } |
1538 | ||
1539 | return NULL; | |
1540 | } | |
1541 | ||
66f5d6ce WW |
1542 | /* called with rcu_read_lock() held |
1543 | */ | |
6454743b DA |
1544 | struct fib6_node *fib6_node_lookup(struct fib6_node *root, |
1545 | const struct in6_addr *daddr, | |
1546 | const struct in6_addr *saddr) | |
1da177e4 | 1547 | { |
1da177e4 | 1548 | struct fib6_node *fn; |
825e288e YH |
1549 | struct lookup_args args[] = { |
1550 | { | |
93c2fb25 | 1551 | .offset = offsetof(struct fib6_info, fib6_dst), |
825e288e YH |
1552 | .addr = daddr, |
1553 | }, | |
1da177e4 | 1554 | #ifdef CONFIG_IPV6_SUBTREES |
825e288e | 1555 | { |
93c2fb25 | 1556 | .offset = offsetof(struct fib6_info, fib6_src), |
825e288e YH |
1557 | .addr = saddr, |
1558 | }, | |
1da177e4 | 1559 | #endif |
825e288e YH |
1560 | { |
1561 | .offset = 0, /* sentinel */ | |
1562 | } | |
1563 | }; | |
1da177e4 | 1564 | |
6454743b | 1565 | fn = fib6_node_lookup_1(root, daddr ? args : args + 1); |
507c9b1e | 1566 | if (!fn || fn->fn_flags & RTN_TL_ROOT) |
1da177e4 LT |
1567 | fn = root; |
1568 | ||
1569 | return fn; | |
1570 | } | |
1571 | ||
1572 | /* | |
1573 | * Get node with specified destination prefix (and source prefix, | |
1574 | * if subtrees are used) | |
38fbeeee WW |
1575 | * exact_match == true means we try to find fn with exact match of |
1576 | * the passed in prefix addr | |
1577 | * exact_match == false means we try to find fn with longest prefix | |
1578 | * match of the passed in prefix addr. This is useful for finding fn | |
1579 | * for cached route as it will be stored in the exception table under | |
1580 | * the node with longest prefix length. | |
1da177e4 LT |
1581 | */ |
1582 | ||
1583 | ||
437de07c WY |
1584 | static struct fib6_node *fib6_locate_1(struct fib6_node *root, |
1585 | const struct in6_addr *addr, | |
38fbeeee WW |
1586 | int plen, int offset, |
1587 | bool exact_match) | |
1da177e4 | 1588 | { |
38fbeeee | 1589 | struct fib6_node *fn, *prev = NULL; |
1da177e4 LT |
1590 | |
1591 | for (fn = root; fn ; ) { | |
8d1c802b | 1592 | struct fib6_info *leaf = rcu_dereference(fn->leaf); |
8d1040e8 WW |
1593 | struct rt6key *key; |
1594 | ||
1595 | /* This node is being deleted */ | |
1596 | if (!leaf) { | |
1597 | if (plen <= fn->fn_bit) | |
1598 | goto out; | |
1599 | else | |
1600 | goto next; | |
1601 | } | |
1602 | ||
1603 | key = (struct rt6key *)((u8 *)leaf + offset); | |
1da177e4 LT |
1604 | |
1605 | /* | |
1606 | * Prefix match | |
1607 | */ | |
1608 | if (plen < fn->fn_bit || | |
1609 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
38fbeeee | 1610 | goto out; |
1da177e4 LT |
1611 | |
1612 | if (plen == fn->fn_bit) | |
1613 | return fn; | |
1614 | ||
40cb35d5 SB |
1615 | if (fn->fn_flags & RTN_RTINFO) |
1616 | prev = fn; | |
38fbeeee | 1617 | |
8d1040e8 | 1618 | next: |
1da177e4 LT |
1619 | /* |
1620 | * We have more bits to go | |
1621 | */ | |
1622 | if (addr_bit_set(addr, fn->fn_bit)) | |
66f5d6ce | 1623 | fn = rcu_dereference(fn->right); |
1da177e4 | 1624 | else |
66f5d6ce | 1625 | fn = rcu_dereference(fn->left); |
1da177e4 | 1626 | } |
38fbeeee WW |
1627 | out: |
1628 | if (exact_match) | |
1629 | return NULL; | |
1630 | else | |
1631 | return prev; | |
1da177e4 LT |
1632 | } |
1633 | ||
437de07c WY |
1634 | struct fib6_node *fib6_locate(struct fib6_node *root, |
1635 | const struct in6_addr *daddr, int dst_len, | |
38fbeeee WW |
1636 | const struct in6_addr *saddr, int src_len, |
1637 | bool exact_match) | |
1da177e4 LT |
1638 | { |
1639 | struct fib6_node *fn; | |
1640 | ||
1641 | fn = fib6_locate_1(root, daddr, dst_len, | |
93c2fb25 | 1642 | offsetof(struct fib6_info, fib6_dst), |
38fbeeee | 1643 | exact_match); |
1da177e4 LT |
1644 | |
1645 | #ifdef CONFIG_IPV6_SUBTREES | |
1646 | if (src_len) { | |
547b792c | 1647 | WARN_ON(saddr == NULL); |
0e80193b WW |
1648 | if (fn) { |
1649 | struct fib6_node *subtree = FIB6_SUBTREE(fn); | |
1650 | ||
1651 | if (subtree) { | |
1652 | fn = fib6_locate_1(subtree, saddr, src_len, | |
93c2fb25 | 1653 | offsetof(struct fib6_info, fib6_src), |
38fbeeee | 1654 | exact_match); |
0e80193b WW |
1655 | } |
1656 | } | |
1da177e4 LT |
1657 | } |
1658 | #endif | |
1659 | ||
507c9b1e | 1660 | if (fn && fn->fn_flags & RTN_RTINFO) |
1da177e4 LT |
1661 | return fn; |
1662 | ||
1663 | return NULL; | |
1664 | } | |
1665 | ||
1666 | ||
1667 | /* | |
1668 | * Deletion | |
1669 | * | |
1670 | */ | |
1671 | ||
8d1c802b | 1672 | static struct fib6_info *fib6_find_prefix(struct net *net, |
66f5d6ce WW |
1673 | struct fib6_table *table, |
1674 | struct fib6_node *fn) | |
1da177e4 | 1675 | { |
66f5d6ce WW |
1676 | struct fib6_node *child_left, *child_right; |
1677 | ||
507c9b1e | 1678 | if (fn->fn_flags & RTN_ROOT) |
421842ed | 1679 | return net->ipv6.fib6_null_entry; |
1da177e4 | 1680 | |
507c9b1e | 1681 | while (fn) { |
66f5d6ce WW |
1682 | child_left = rcu_dereference_protected(fn->left, |
1683 | lockdep_is_held(&table->tb6_lock)); | |
1684 | child_right = rcu_dereference_protected(fn->right, | |
1685 | lockdep_is_held(&table->tb6_lock)); | |
1686 | if (child_left) | |
1687 | return rcu_dereference_protected(child_left->leaf, | |
1688 | lockdep_is_held(&table->tb6_lock)); | |
1689 | if (child_right) | |
1690 | return rcu_dereference_protected(child_right->leaf, | |
1691 | lockdep_is_held(&table->tb6_lock)); | |
1da177e4 | 1692 | |
7fc33165 | 1693 | fn = FIB6_SUBTREE(fn); |
1da177e4 LT |
1694 | } |
1695 | return NULL; | |
1696 | } | |
1697 | ||
1698 | /* | |
1699 | * Called to trim the tree of intermediate nodes when possible. "fn" | |
1700 | * is the node we want to try and remove. | |
66f5d6ce | 1701 | * Need to own table->tb6_lock |
1da177e4 LT |
1702 | */ |
1703 | ||
8ed67789 | 1704 | static struct fib6_node *fib6_repair_tree(struct net *net, |
66f5d6ce WW |
1705 | struct fib6_table *table, |
1706 | struct fib6_node *fn) | |
1da177e4 LT |
1707 | { |
1708 | int children; | |
1709 | int nstate; | |
66f5d6ce | 1710 | struct fib6_node *child; |
94b2cfe0 | 1711 | struct fib6_walker *w; |
1da177e4 LT |
1712 | int iter = 0; |
1713 | ||
4512c43e WW |
1714 | /* Set fn->leaf to null_entry for root node. */ |
1715 | if (fn->fn_flags & RTN_TL_ROOT) { | |
421842ed | 1716 | rcu_assign_pointer(fn->leaf, net->ipv6.fib6_null_entry); |
4512c43e WW |
1717 | return fn; |
1718 | } | |
1719 | ||
1da177e4 | 1720 | for (;;) { |
66f5d6ce WW |
1721 | struct fib6_node *fn_r = rcu_dereference_protected(fn->right, |
1722 | lockdep_is_held(&table->tb6_lock)); | |
1723 | struct fib6_node *fn_l = rcu_dereference_protected(fn->left, | |
1724 | lockdep_is_held(&table->tb6_lock)); | |
1725 | struct fib6_node *pn = rcu_dereference_protected(fn->parent, | |
1726 | lockdep_is_held(&table->tb6_lock)); | |
1727 | struct fib6_node *pn_r = rcu_dereference_protected(pn->right, | |
1728 | lockdep_is_held(&table->tb6_lock)); | |
1729 | struct fib6_node *pn_l = rcu_dereference_protected(pn->left, | |
1730 | lockdep_is_held(&table->tb6_lock)); | |
8d1c802b | 1731 | struct fib6_info *fn_leaf = rcu_dereference_protected(fn->leaf, |
66f5d6ce | 1732 | lockdep_is_held(&table->tb6_lock)); |
8d1c802b | 1733 | struct fib6_info *pn_leaf = rcu_dereference_protected(pn->leaf, |
66f5d6ce | 1734 | lockdep_is_held(&table->tb6_lock)); |
8d1c802b | 1735 | struct fib6_info *new_fn_leaf; |
66f5d6ce | 1736 | |
1da177e4 LT |
1737 | RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter); |
1738 | iter++; | |
1739 | ||
547b792c IJ |
1740 | WARN_ON(fn->fn_flags & RTN_RTINFO); |
1741 | WARN_ON(fn->fn_flags & RTN_TL_ROOT); | |
66f5d6ce | 1742 | WARN_ON(fn_leaf); |
1da177e4 LT |
1743 | |
1744 | children = 0; | |
1745 | child = NULL; | |
66f5d6ce WW |
1746 | if (fn_r) |
1747 | child = fn_r, children |= 1; | |
1748 | if (fn_l) | |
1749 | child = fn_l, children |= 2; | |
1da177e4 | 1750 | |
7fc33165 | 1751 | if (children == 3 || FIB6_SUBTREE(fn) |
1da177e4 LT |
1752 | #ifdef CONFIG_IPV6_SUBTREES |
1753 | /* Subtree root (i.e. fn) may have one child */ | |
507c9b1e | 1754 | || (children && fn->fn_flags & RTN_ROOT) |
1da177e4 LT |
1755 | #endif |
1756 | ) { | |
66f5d6ce | 1757 | new_fn_leaf = fib6_find_prefix(net, table, fn); |
1da177e4 | 1758 | #if RT6_DEBUG >= 2 |
66f5d6ce WW |
1759 | if (!new_fn_leaf) { |
1760 | WARN_ON(!new_fn_leaf); | |
421842ed | 1761 | new_fn_leaf = net->ipv6.fib6_null_entry; |
1da177e4 LT |
1762 | } |
1763 | #endif | |
93531c67 | 1764 | fib6_info_hold(new_fn_leaf); |
66f5d6ce WW |
1765 | rcu_assign_pointer(fn->leaf, new_fn_leaf); |
1766 | return pn; | |
1da177e4 LT |
1767 | } |
1768 | ||
1da177e4 | 1769 | #ifdef CONFIG_IPV6_SUBTREES |
7fc33165 | 1770 | if (FIB6_SUBTREE(pn) == fn) { |
547b792c | 1771 | WARN_ON(!(fn->fn_flags & RTN_ROOT)); |
66f5d6ce | 1772 | RCU_INIT_POINTER(pn->subtree, NULL); |
1da177e4 LT |
1773 | nstate = FWS_L; |
1774 | } else { | |
547b792c | 1775 | WARN_ON(fn->fn_flags & RTN_ROOT); |
1da177e4 | 1776 | #endif |
66f5d6ce WW |
1777 | if (pn_r == fn) |
1778 | rcu_assign_pointer(pn->right, child); | |
1779 | else if (pn_l == fn) | |
1780 | rcu_assign_pointer(pn->left, child); | |
1da177e4 | 1781 | #if RT6_DEBUG >= 2 |
547b792c IJ |
1782 | else |
1783 | WARN_ON(1); | |
1da177e4 LT |
1784 | #endif |
1785 | if (child) | |
66f5d6ce | 1786 | rcu_assign_pointer(child->parent, pn); |
1da177e4 LT |
1787 | nstate = FWS_R; |
1788 | #ifdef CONFIG_IPV6_SUBTREES | |
1789 | } | |
1790 | #endif | |
1791 | ||
9a03cd8f MK |
1792 | read_lock(&net->ipv6.fib6_walker_lock); |
1793 | FOR_WALKERS(net, w) { | |
507c9b1e | 1794 | if (!child) { |
2b760fcf | 1795 | if (w->node == fn) { |
1da177e4 LT |
1796 | RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate); |
1797 | w->node = pn; | |
1798 | w->state = nstate; | |
1799 | } | |
1800 | } else { | |
1da177e4 LT |
1801 | if (w->node == fn) { |
1802 | w->node = child; | |
1803 | if (children&2) { | |
1804 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
8db46f1d | 1805 | w->state = w->state >= FWS_R ? FWS_U : FWS_INIT; |
1da177e4 LT |
1806 | } else { |
1807 | RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); | |
8db46f1d | 1808 | w->state = w->state >= FWS_C ? FWS_U : FWS_INIT; |
1da177e4 LT |
1809 | } |
1810 | } | |
1811 | } | |
1812 | } | |
9a03cd8f | 1813 | read_unlock(&net->ipv6.fib6_walker_lock); |
1da177e4 | 1814 | |
81eb8447 | 1815 | node_free(net, fn); |
507c9b1e | 1816 | if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn)) |
1da177e4 LT |
1817 | return pn; |
1818 | ||
66f5d6ce | 1819 | RCU_INIT_POINTER(pn->leaf, NULL); |
93531c67 | 1820 | fib6_info_release(pn_leaf); |
1da177e4 LT |
1821 | fn = pn; |
1822 | } | |
1823 | } | |
1824 | ||
66f5d6ce | 1825 | static void fib6_del_route(struct fib6_table *table, struct fib6_node *fn, |
8d1c802b | 1826 | struct fib6_info __rcu **rtp, struct nl_info *info) |
1da177e4 | 1827 | { |
94b2cfe0 | 1828 | struct fib6_walker *w; |
8d1c802b | 1829 | struct fib6_info *rt = rcu_dereference_protected(*rtp, |
66f5d6ce | 1830 | lockdep_is_held(&table->tb6_lock)); |
c572872f | 1831 | struct net *net = info->nl_net; |
1da177e4 LT |
1832 | |
1833 | RT6_TRACE("fib6_del_route\n"); | |
1834 | ||
1835 | /* Unlink it */ | |
8fb11a9a | 1836 | *rtp = rt->fib6_next; |
93c2fb25 | 1837 | rt->fib6_node = NULL; |
c572872f BT |
1838 | net->ipv6.rt6_stats->fib_rt_entries--; |
1839 | net->ipv6.rt6_stats->fib_discarded_routes++; | |
1da177e4 | 1840 | |
2b760fcf WW |
1841 | /* Flush all cached dst in exception table */ |
1842 | rt6_flush_exceptions(rt); | |
1843 | ||
f11e6659 | 1844 | /* Reset round-robin state, if necessary */ |
66f5d6ce | 1845 | if (rcu_access_pointer(fn->rr_ptr) == rt) |
f11e6659 DM |
1846 | fn->rr_ptr = NULL; |
1847 | ||
51ebd318 | 1848 | /* Remove this entry from other siblings */ |
93c2fb25 | 1849 | if (rt->fib6_nsiblings) { |
8d1c802b | 1850 | struct fib6_info *sibling, *next_sibling; |
51ebd318 ND |
1851 | |
1852 | list_for_each_entry_safe(sibling, next_sibling, | |
93c2fb25 DA |
1853 | &rt->fib6_siblings, fib6_siblings) |
1854 | sibling->fib6_nsiblings--; | |
1855 | rt->fib6_nsiblings = 0; | |
1856 | list_del_init(&rt->fib6_siblings); | |
d7dedee1 | 1857 | rt6_multipath_rebalance(next_sibling); |
51ebd318 ND |
1858 | } |
1859 | ||
1da177e4 | 1860 | /* Adjust walkers */ |
9a03cd8f MK |
1861 | read_lock(&net->ipv6.fib6_walker_lock); |
1862 | FOR_WALKERS(net, w) { | |
1da177e4 LT |
1863 | if (w->state == FWS_C && w->leaf == rt) { |
1864 | RT6_TRACE("walker %p adjusted by delroute\n", w); | |
8fb11a9a | 1865 | w->leaf = rcu_dereference_protected(rt->fib6_next, |
66f5d6ce | 1866 | lockdep_is_held(&table->tb6_lock)); |
507c9b1e | 1867 | if (!w->leaf) |
1da177e4 LT |
1868 | w->state = FWS_U; |
1869 | } | |
1870 | } | |
9a03cd8f | 1871 | read_unlock(&net->ipv6.fib6_walker_lock); |
1da177e4 | 1872 | |
4512c43e WW |
1873 | /* If it was last route, call fib6_repair_tree() to: |
1874 | * 1. For root node, put back null_entry as how the table was created. | |
1875 | * 2. For other nodes, expunge its radix tree node. | |
1876 | */ | |
66f5d6ce | 1877 | if (!rcu_access_pointer(fn->leaf)) { |
4512c43e WW |
1878 | if (!(fn->fn_flags & RTN_TL_ROOT)) { |
1879 | fn->fn_flags &= ~RTN_RTINFO; | |
1880 | net->ipv6.rt6_stats->fib_route_nodes--; | |
1881 | } | |
66f5d6ce | 1882 | fn = fib6_repair_tree(net, table, fn); |
1da177e4 LT |
1883 | } |
1884 | ||
6e9e16e6 | 1885 | fib6_purge_rt(rt, fn, net); |
1da177e4 | 1886 | |
d5382fef IS |
1887 | if (!info->skip_notify_kernel) |
1888 | call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, rt, NULL); | |
16a16cd3 DA |
1889 | if (!info->skip_notify) |
1890 | inet6_rt_notify(RTM_DELROUTE, rt, info, 0); | |
d5382fef | 1891 | |
93531c67 | 1892 | fib6_info_release(rt); |
1da177e4 LT |
1893 | } |
1894 | ||
66f5d6ce | 1895 | /* Need to own table->tb6_lock */ |
8d1c802b | 1896 | int fib6_del(struct fib6_info *rt, struct nl_info *info) |
1da177e4 | 1897 | { |
93c2fb25 DA |
1898 | struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node, |
1899 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
1900 | struct fib6_table *table = rt->fib6_table; | |
8ed67789 | 1901 | struct net *net = info->nl_net; |
8d1c802b DA |
1902 | struct fib6_info __rcu **rtp; |
1903 | struct fib6_info __rcu **rtp_next; | |
1da177e4 | 1904 | |
421842ed | 1905 | if (!fn || rt == net->ipv6.fib6_null_entry) |
1da177e4 LT |
1906 | return -ENOENT; |
1907 | ||
547b792c | 1908 | WARN_ON(!(fn->fn_flags & RTN_RTINFO)); |
1da177e4 | 1909 | |
1da177e4 LT |
1910 | /* |
1911 | * Walk the leaf entries looking for ourself | |
1912 | */ | |
1913 | ||
66f5d6ce | 1914 | for (rtp = &fn->leaf; *rtp; rtp = rtp_next) { |
8d1c802b | 1915 | struct fib6_info *cur = rcu_dereference_protected(*rtp, |
66f5d6ce WW |
1916 | lockdep_is_held(&table->tb6_lock)); |
1917 | if (rt == cur) { | |
1918 | fib6_del_route(table, fn, rtp, info); | |
1da177e4 LT |
1919 | return 0; |
1920 | } | |
8fb11a9a | 1921 | rtp_next = &cur->fib6_next; |
1da177e4 LT |
1922 | } |
1923 | return -ENOENT; | |
1924 | } | |
1925 | ||
1926 | /* | |
1927 | * Tree traversal function. | |
1928 | * | |
1929 | * Certainly, it is not interrupt safe. | |
1930 | * However, it is internally reenterable wrt itself and fib6_add/fib6_del. | |
1931 | * It means, that we can modify tree during walking | |
1932 | * and use this function for garbage collection, clone pruning, | |
1ab1457c | 1933 | * cleaning tree when a device goes down etc. etc. |
1da177e4 LT |
1934 | * |
1935 | * It guarantees that every node will be traversed, | |
1936 | * and that it will be traversed only once. | |
1937 | * | |
1938 | * Callback function w->func may return: | |
1939 | * 0 -> continue walking. | |
1940 | * positive value -> walking is suspended (used by tree dumps, | |
1941 | * and probably by gc, if it will be split to several slices) | |
1942 | * negative value -> terminate walking. | |
1943 | * | |
1944 | * The function itself returns: | |
1945 | * 0 -> walk is complete. | |
1946 | * >0 -> walk is incomplete (i.e. suspended) | |
1947 | * <0 -> walk is terminated by an error. | |
66f5d6ce WW |
1948 | * |
1949 | * This function is called with tb6_lock held. | |
1da177e4 LT |
1950 | */ |
1951 | ||
94b2cfe0 | 1952 | static int fib6_walk_continue(struct fib6_walker *w) |
1da177e4 | 1953 | { |
66f5d6ce | 1954 | struct fib6_node *fn, *pn, *left, *right; |
1da177e4 | 1955 | |
2b760fcf WW |
1956 | /* w->root should always be table->tb6_root */ |
1957 | WARN_ON_ONCE(!(w->root->fn_flags & RTN_TL_ROOT)); | |
1958 | ||
1da177e4 LT |
1959 | for (;;) { |
1960 | fn = w->node; | |
507c9b1e | 1961 | if (!fn) |
1da177e4 LT |
1962 | return 0; |
1963 | ||
1da177e4 LT |
1964 | switch (w->state) { |
1965 | #ifdef CONFIG_IPV6_SUBTREES | |
1966 | case FWS_S: | |
7fc33165 YH |
1967 | if (FIB6_SUBTREE(fn)) { |
1968 | w->node = FIB6_SUBTREE(fn); | |
1da177e4 LT |
1969 | continue; |
1970 | } | |
1971 | w->state = FWS_L; | |
1ab1457c | 1972 | #endif |
275757e6 | 1973 | /* fall through */ |
1da177e4 | 1974 | case FWS_L: |
66f5d6ce WW |
1975 | left = rcu_dereference_protected(fn->left, 1); |
1976 | if (left) { | |
1977 | w->node = left; | |
1da177e4 LT |
1978 | w->state = FWS_INIT; |
1979 | continue; | |
1980 | } | |
1981 | w->state = FWS_R; | |
275757e6 | 1982 | /* fall through */ |
1da177e4 | 1983 | case FWS_R: |
66f5d6ce WW |
1984 | right = rcu_dereference_protected(fn->right, 1); |
1985 | if (right) { | |
1986 | w->node = right; | |
1da177e4 LT |
1987 | w->state = FWS_INIT; |
1988 | continue; | |
1989 | } | |
1990 | w->state = FWS_C; | |
66f5d6ce | 1991 | w->leaf = rcu_dereference_protected(fn->leaf, 1); |
275757e6 | 1992 | /* fall through */ |
1da177e4 | 1993 | case FWS_C: |
507c9b1e | 1994 | if (w->leaf && fn->fn_flags & RTN_RTINFO) { |
2bec5a36 PM |
1995 | int err; |
1996 | ||
fa809e2f ED |
1997 | if (w->skip) { |
1998 | w->skip--; | |
1c265854 | 1999 | goto skip; |
2bec5a36 PM |
2000 | } |
2001 | ||
2002 | err = w->func(w); | |
1da177e4 LT |
2003 | if (err) |
2004 | return err; | |
2bec5a36 PM |
2005 | |
2006 | w->count++; | |
1da177e4 LT |
2007 | continue; |
2008 | } | |
1c265854 | 2009 | skip: |
1da177e4 | 2010 | w->state = FWS_U; |
275757e6 | 2011 | /* fall through */ |
1da177e4 LT |
2012 | case FWS_U: |
2013 | if (fn == w->root) | |
2014 | return 0; | |
66f5d6ce WW |
2015 | pn = rcu_dereference_protected(fn->parent, 1); |
2016 | left = rcu_dereference_protected(pn->left, 1); | |
2017 | right = rcu_dereference_protected(pn->right, 1); | |
1da177e4 LT |
2018 | w->node = pn; |
2019 | #ifdef CONFIG_IPV6_SUBTREES | |
7fc33165 | 2020 | if (FIB6_SUBTREE(pn) == fn) { |
547b792c | 2021 | WARN_ON(!(fn->fn_flags & RTN_ROOT)); |
1da177e4 LT |
2022 | w->state = FWS_L; |
2023 | continue; | |
2024 | } | |
2025 | #endif | |
66f5d6ce | 2026 | if (left == fn) { |
1da177e4 LT |
2027 | w->state = FWS_R; |
2028 | continue; | |
2029 | } | |
66f5d6ce | 2030 | if (right == fn) { |
1da177e4 | 2031 | w->state = FWS_C; |
66f5d6ce | 2032 | w->leaf = rcu_dereference_protected(w->node->leaf, 1); |
1da177e4 LT |
2033 | continue; |
2034 | } | |
2035 | #if RT6_DEBUG >= 2 | |
547b792c | 2036 | WARN_ON(1); |
1da177e4 LT |
2037 | #endif |
2038 | } | |
2039 | } | |
2040 | } | |
2041 | ||
9a03cd8f | 2042 | static int fib6_walk(struct net *net, struct fib6_walker *w) |
1da177e4 LT |
2043 | { |
2044 | int res; | |
2045 | ||
2046 | w->state = FWS_INIT; | |
2047 | w->node = w->root; | |
2048 | ||
9a03cd8f | 2049 | fib6_walker_link(net, w); |
1da177e4 LT |
2050 | res = fib6_walk_continue(w); |
2051 | if (res <= 0) | |
9a03cd8f | 2052 | fib6_walker_unlink(net, w); |
1da177e4 LT |
2053 | return res; |
2054 | } | |
2055 | ||
94b2cfe0 | 2056 | static int fib6_clean_node(struct fib6_walker *w) |
1da177e4 LT |
2057 | { |
2058 | int res; | |
8d1c802b | 2059 | struct fib6_info *rt; |
94b2cfe0 | 2060 | struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w); |
ec7d43c2 BT |
2061 | struct nl_info info = { |
2062 | .nl_net = c->net, | |
7c6bb7d2 | 2063 | .skip_notify = c->skip_notify, |
ec7d43c2 | 2064 | }; |
1da177e4 | 2065 | |
327571cb HFS |
2066 | if (c->sernum != FIB6_NO_SERNUM_CHANGE && |
2067 | w->node->fn_sernum != c->sernum) | |
2068 | w->node->fn_sernum = c->sernum; | |
2069 | ||
2070 | if (!c->func) { | |
2071 | WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE); | |
2072 | w->leaf = NULL; | |
2073 | return 0; | |
2074 | } | |
2075 | ||
66f5d6ce | 2076 | for_each_fib6_walker_rt(w) { |
1da177e4 | 2077 | res = c->func(rt, c->arg); |
b5cb5a75 | 2078 | if (res == -1) { |
1da177e4 | 2079 | w->leaf = rt; |
528c4ceb | 2080 | res = fib6_del(rt, &info); |
1da177e4 LT |
2081 | if (res) { |
2082 | #if RT6_DEBUG >= 2 | |
91df42be | 2083 | pr_debug("%s: del failed: rt=%p@%p err=%d\n", |
4e587ea7 | 2084 | __func__, rt, |
93c2fb25 | 2085 | rcu_access_pointer(rt->fib6_node), |
4e587ea7 | 2086 | res); |
1da177e4 LT |
2087 | #endif |
2088 | continue; | |
2089 | } | |
2090 | return 0; | |
b5cb5a75 | 2091 | } else if (res == -2) { |
93c2fb25 | 2092 | if (WARN_ON(!rt->fib6_nsiblings)) |
b5cb5a75 | 2093 | continue; |
93c2fb25 DA |
2094 | rt = list_last_entry(&rt->fib6_siblings, |
2095 | struct fib6_info, fib6_siblings); | |
b5cb5a75 | 2096 | continue; |
1da177e4 | 2097 | } |
547b792c | 2098 | WARN_ON(res != 0); |
1da177e4 LT |
2099 | } |
2100 | w->leaf = rt; | |
2101 | return 0; | |
2102 | } | |
2103 | ||
2104 | /* | |
2105 | * Convenient frontend to tree walker. | |
1ab1457c | 2106 | * |
1da177e4 | 2107 | * func is called on each route. |
b5cb5a75 IS |
2108 | * It may return -2 -> skip multipath route. |
2109 | * -1 -> delete this route. | |
1da177e4 | 2110 | * 0 -> continue walking |
1da177e4 LT |
2111 | */ |
2112 | ||
ec7d43c2 | 2113 | static void fib6_clean_tree(struct net *net, struct fib6_node *root, |
8d1c802b | 2114 | int (*func)(struct fib6_info *, void *arg), |
7c6bb7d2 | 2115 | int sernum, void *arg, bool skip_notify) |
1da177e4 | 2116 | { |
94b2cfe0 | 2117 | struct fib6_cleaner c; |
1da177e4 LT |
2118 | |
2119 | c.w.root = root; | |
2120 | c.w.func = fib6_clean_node; | |
2bec5a36 PM |
2121 | c.w.count = 0; |
2122 | c.w.skip = 0; | |
1e47b483 | 2123 | c.w.skip_in_node = 0; |
1da177e4 | 2124 | c.func = func; |
327571cb | 2125 | c.sernum = sernum; |
1da177e4 | 2126 | c.arg = arg; |
ec7d43c2 | 2127 | c.net = net; |
7c6bb7d2 | 2128 | c.skip_notify = skip_notify; |
1da177e4 | 2129 | |
9a03cd8f | 2130 | fib6_walk(net, &c.w); |
1da177e4 LT |
2131 | } |
2132 | ||
327571cb | 2133 | static void __fib6_clean_all(struct net *net, |
8d1c802b | 2134 | int (*func)(struct fib6_info *, void *), |
7c6bb7d2 | 2135 | int sernum, void *arg, bool skip_notify) |
c71099ac | 2136 | { |
c71099ac | 2137 | struct fib6_table *table; |
58f09b78 | 2138 | struct hlist_head *head; |
1b43af54 | 2139 | unsigned int h; |
c71099ac | 2140 | |
1b43af54 | 2141 | rcu_read_lock(); |
a33bc5c1 | 2142 | for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { |
f3db4851 | 2143 | head = &net->ipv6.fib_table_hash[h]; |
b67bfe0d | 2144 | hlist_for_each_entry_rcu(table, head, tb6_hlist) { |
66f5d6ce | 2145 | spin_lock_bh(&table->tb6_lock); |
ec7d43c2 | 2146 | fib6_clean_tree(net, &table->tb6_root, |
7c6bb7d2 | 2147 | func, sernum, arg, skip_notify); |
66f5d6ce | 2148 | spin_unlock_bh(&table->tb6_lock); |
c71099ac TG |
2149 | } |
2150 | } | |
1b43af54 | 2151 | rcu_read_unlock(); |
c71099ac TG |
2152 | } |
2153 | ||
8d1c802b | 2154 | void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *), |
327571cb HFS |
2155 | void *arg) |
2156 | { | |
7c6bb7d2 DA |
2157 | __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, false); |
2158 | } | |
2159 | ||
2160 | void fib6_clean_all_skip_notify(struct net *net, | |
2161 | int (*func)(struct fib6_info *, void *), | |
2162 | void *arg) | |
2163 | { | |
2164 | __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, true); | |
327571cb HFS |
2165 | } |
2166 | ||
705f1c86 HFS |
2167 | static void fib6_flush_trees(struct net *net) |
2168 | { | |
812918c4 | 2169 | int new_sernum = fib6_new_sernum(net); |
705f1c86 | 2170 | |
7c6bb7d2 | 2171 | __fib6_clean_all(net, NULL, new_sernum, NULL, false); |
705f1c86 HFS |
2172 | } |
2173 | ||
1da177e4 LT |
2174 | /* |
2175 | * Garbage collection | |
2176 | */ | |
2177 | ||
8d1c802b | 2178 | static int fib6_age(struct fib6_info *rt, void *arg) |
1da177e4 | 2179 | { |
3570df91 | 2180 | struct fib6_gc_args *gc_args = arg; |
1da177e4 LT |
2181 | unsigned long now = jiffies; |
2182 | ||
2183 | /* | |
2184 | * check addrconf expiration here. | |
2185 | * Routes are expired even if they are in use. | |
1da177e4 LT |
2186 | */ |
2187 | ||
93c2fb25 | 2188 | if (rt->fib6_flags & RTF_EXPIRES && rt->expires) { |
14895687 | 2189 | if (time_after(now, rt->expires)) { |
1da177e4 | 2190 | RT6_TRACE("expiring %p\n", rt); |
1da177e4 LT |
2191 | return -1; |
2192 | } | |
3570df91 | 2193 | gc_args->more++; |
1da177e4 LT |
2194 | } |
2195 | ||
c757faa8 WW |
2196 | /* Also age clones in the exception table. |
2197 | * Note, that clones are aged out | |
2198 | * only if they are not in use now. | |
2199 | */ | |
2200 | rt6_age_exceptions(rt, gc_args, now); | |
2201 | ||
1da177e4 LT |
2202 | return 0; |
2203 | } | |
2204 | ||
2ac3ac8f | 2205 | void fib6_run_gc(unsigned long expires, struct net *net, bool force) |
1da177e4 | 2206 | { |
3570df91 | 2207 | struct fib6_gc_args gc_args; |
49a18d86 MK |
2208 | unsigned long now; |
2209 | ||
2ac3ac8f | 2210 | if (force) { |
3dc94f93 MK |
2211 | spin_lock_bh(&net->ipv6.fib6_gc_lock); |
2212 | } else if (!spin_trylock_bh(&net->ipv6.fib6_gc_lock)) { | |
2ac3ac8f MK |
2213 | mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ); |
2214 | return; | |
1da177e4 | 2215 | } |
2ac3ac8f MK |
2216 | gc_args.timeout = expires ? (int)expires : |
2217 | net->ipv6.sysctl.ip6_rt_gc_interval; | |
db916649 | 2218 | gc_args.more = 0; |
f3db4851 | 2219 | |
3570df91 | 2220 | fib6_clean_all(net, fib6_age, &gc_args); |
49a18d86 MK |
2221 | now = jiffies; |
2222 | net->ipv6.ip6_rt_last_gc = now; | |
1da177e4 LT |
2223 | |
2224 | if (gc_args.more) | |
c8a45222 | 2225 | mod_timer(&net->ipv6.ip6_fib_timer, |
49a18d86 | 2226 | round_jiffies(now |
c8a45222 | 2227 | + net->ipv6.sysctl.ip6_rt_gc_interval)); |
417f28bb SH |
2228 | else |
2229 | del_timer(&net->ipv6.ip6_fib_timer); | |
3dc94f93 | 2230 | spin_unlock_bh(&net->ipv6.fib6_gc_lock); |
1da177e4 LT |
2231 | } |
2232 | ||
86cb30ec | 2233 | static void fib6_gc_timer_cb(struct timer_list *t) |
5b7c931d | 2234 | { |
86cb30ec KC |
2235 | struct net *arg = from_timer(arg, t, ipv6.ip6_fib_timer); |
2236 | ||
2237 | fib6_run_gc(0, arg, true); | |
5b7c931d DL |
2238 | } |
2239 | ||
2c8c1e72 | 2240 | static int __net_init fib6_net_init(struct net *net) |
1da177e4 | 2241 | { |
10da66f7 | 2242 | size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ; |
16ab6d7d IS |
2243 | int err; |
2244 | ||
2245 | err = fib6_notifier_init(net); | |
2246 | if (err) | |
2247 | return err; | |
10da66f7 | 2248 | |
3dc94f93 | 2249 | spin_lock_init(&net->ipv6.fib6_gc_lock); |
9a03cd8f MK |
2250 | rwlock_init(&net->ipv6.fib6_walker_lock); |
2251 | INIT_LIST_HEAD(&net->ipv6.fib6_walkers); | |
86cb30ec | 2252 | timer_setup(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, 0); |
63152fc0 | 2253 | |
c572872f BT |
2254 | net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL); |
2255 | if (!net->ipv6.rt6_stats) | |
2256 | goto out_timer; | |
2257 | ||
10da66f7 ED |
2258 | /* Avoid false sharing : Use at least a full cache line */ |
2259 | size = max_t(size_t, size, L1_CACHE_BYTES); | |
2260 | ||
2261 | net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL); | |
58f09b78 | 2262 | if (!net->ipv6.fib_table_hash) |
c572872f | 2263 | goto out_rt6_stats; |
e0b85590 | 2264 | |
58f09b78 DL |
2265 | net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl), |
2266 | GFP_KERNEL); | |
2267 | if (!net->ipv6.fib6_main_tbl) | |
e0b85590 DL |
2268 | goto out_fib_table_hash; |
2269 | ||
58f09b78 | 2270 | net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN; |
66f5d6ce | 2271 | rcu_assign_pointer(net->ipv6.fib6_main_tbl->tb6_root.leaf, |
421842ed | 2272 | net->ipv6.fib6_null_entry); |
58f09b78 DL |
2273 | net->ipv6.fib6_main_tbl->tb6_root.fn_flags = |
2274 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
8e773277 | 2275 | inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers); |
e0b85590 DL |
2276 | |
2277 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES | |
58f09b78 DL |
2278 | net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl), |
2279 | GFP_KERNEL); | |
2280 | if (!net->ipv6.fib6_local_tbl) | |
e0b85590 | 2281 | goto out_fib6_main_tbl; |
58f09b78 | 2282 | net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL; |
66f5d6ce | 2283 | rcu_assign_pointer(net->ipv6.fib6_local_tbl->tb6_root.leaf, |
421842ed | 2284 | net->ipv6.fib6_null_entry); |
58f09b78 DL |
2285 | net->ipv6.fib6_local_tbl->tb6_root.fn_flags = |
2286 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
8e773277 | 2287 | inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers); |
e0b85590 | 2288 | #endif |
58f09b78 | 2289 | fib6_tables_init(net); |
f845ab6b | 2290 | |
417f28bb | 2291 | return 0; |
d63bddbe | 2292 | |
e0b85590 | 2293 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 2294 | out_fib6_main_tbl: |
58f09b78 | 2295 | kfree(net->ipv6.fib6_main_tbl); |
e0b85590 | 2296 | #endif |
e0b85590 | 2297 | out_fib_table_hash: |
58f09b78 | 2298 | kfree(net->ipv6.fib_table_hash); |
c572872f BT |
2299 | out_rt6_stats: |
2300 | kfree(net->ipv6.rt6_stats); | |
63152fc0 | 2301 | out_timer: |
16ab6d7d | 2302 | fib6_notifier_exit(net); |
417f28bb | 2303 | return -ENOMEM; |
8db46f1d | 2304 | } |
58f09b78 DL |
2305 | |
2306 | static void fib6_net_exit(struct net *net) | |
2307 | { | |
ba1cc08d SD |
2308 | unsigned int i; |
2309 | ||
417f28bb SH |
2310 | del_timer_sync(&net->ipv6.ip6_fib_timer); |
2311 | ||
32a805ba | 2312 | for (i = 0; i < FIB6_TABLE_HASHSZ; i++) { |
ba1cc08d SD |
2313 | struct hlist_head *head = &net->ipv6.fib_table_hash[i]; |
2314 | struct hlist_node *tmp; | |
2315 | struct fib6_table *tb; | |
2316 | ||
2317 | hlist_for_each_entry_safe(tb, tmp, head, tb6_hlist) { | |
2318 | hlist_del(&tb->tb6_hlist); | |
2319 | fib6_free_table(tb); | |
2320 | } | |
2321 | } | |
2322 | ||
58f09b78 | 2323 | kfree(net->ipv6.fib_table_hash); |
c572872f | 2324 | kfree(net->ipv6.rt6_stats); |
16ab6d7d | 2325 | fib6_notifier_exit(net); |
58f09b78 DL |
2326 | } |
2327 | ||
2328 | static struct pernet_operations fib6_net_ops = { | |
2329 | .init = fib6_net_init, | |
2330 | .exit = fib6_net_exit, | |
2331 | }; | |
2332 | ||
2333 | int __init fib6_init(void) | |
2334 | { | |
2335 | int ret = -ENOMEM; | |
63152fc0 | 2336 | |
58f09b78 DL |
2337 | fib6_node_kmem = kmem_cache_create("fib6_nodes", |
2338 | sizeof(struct fib6_node), | |
2339 | 0, SLAB_HWCACHE_ALIGN, | |
2340 | NULL); | |
2341 | if (!fib6_node_kmem) | |
2342 | goto out; | |
2343 | ||
2344 | ret = register_pernet_subsys(&fib6_net_ops); | |
2345 | if (ret) | |
c572872f | 2346 | goto out_kmem_cache_create; |
e8803b6c | 2347 | |
16feebcf FW |
2348 | ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE, NULL, |
2349 | inet6_dump_fib, 0); | |
e8803b6c DM |
2350 | if (ret) |
2351 | goto out_unregister_subsys; | |
705f1c86 HFS |
2352 | |
2353 | __fib6_flush_trees = fib6_flush_trees; | |
58f09b78 DL |
2354 | out: |
2355 | return ret; | |
2356 | ||
e8803b6c DM |
2357 | out_unregister_subsys: |
2358 | unregister_pernet_subsys(&fib6_net_ops); | |
d63bddbe DL |
2359 | out_kmem_cache_create: |
2360 | kmem_cache_destroy(fib6_node_kmem); | |
2361 | goto out; | |
1da177e4 LT |
2362 | } |
2363 | ||
2364 | void fib6_gc_cleanup(void) | |
2365 | { | |
58f09b78 | 2366 | unregister_pernet_subsys(&fib6_net_ops); |
1da177e4 LT |
2367 | kmem_cache_destroy(fib6_node_kmem); |
2368 | } | |
8d2ca1d7 HFS |
2369 | |
2370 | #ifdef CONFIG_PROC_FS | |
8d2ca1d7 HFS |
2371 | static int ipv6_route_seq_show(struct seq_file *seq, void *v) |
2372 | { | |
8d1c802b | 2373 | struct fib6_info *rt = v; |
8d2ca1d7 | 2374 | struct ipv6_route_iter *iter = seq->private; |
f88d8ea6 | 2375 | struct fib6_nh *fib6_nh = rt->fib6_nh; |
2b2450ca | 2376 | unsigned int flags = rt->fib6_flags; |
5e670d84 | 2377 | const struct net_device *dev; |
8d2ca1d7 | 2378 | |
f88d8ea6 DA |
2379 | if (rt->nh) |
2380 | fib6_nh = nexthop_fib6_nh(rt->nh); | |
2381 | ||
93c2fb25 | 2382 | seq_printf(seq, "%pi6 %02x ", &rt->fib6_dst.addr, rt->fib6_dst.plen); |
8d2ca1d7 HFS |
2383 | |
2384 | #ifdef CONFIG_IPV6_SUBTREES | |
93c2fb25 | 2385 | seq_printf(seq, "%pi6 %02x ", &rt->fib6_src.addr, rt->fib6_src.plen); |
8d2ca1d7 HFS |
2386 | #else |
2387 | seq_puts(seq, "00000000000000000000000000000000 00 "); | |
2388 | #endif | |
f88d8ea6 | 2389 | if (fib6_nh->fib_nh_gw_family) { |
2b2450ca | 2390 | flags |= RTF_GATEWAY; |
f88d8ea6 | 2391 | seq_printf(seq, "%pi6", &fib6_nh->fib_nh_gw6); |
2b2450ca | 2392 | } else { |
8d2ca1d7 | 2393 | seq_puts(seq, "00000000000000000000000000000000"); |
2b2450ca | 2394 | } |
8d2ca1d7 | 2395 | |
f88d8ea6 | 2396 | dev = fib6_nh->fib_nh_dev; |
8d2ca1d7 | 2397 | seq_printf(seq, " %08x %08x %08x %08x %8s\n", |
f05713e0 | 2398 | rt->fib6_metric, refcount_read(&rt->fib6_ref), 0, |
2b2450ca | 2399 | flags, dev ? dev->name : ""); |
8d2ca1d7 HFS |
2400 | iter->w.leaf = NULL; |
2401 | return 0; | |
2402 | } | |
2403 | ||
94b2cfe0 | 2404 | static int ipv6_route_yield(struct fib6_walker *w) |
8d2ca1d7 HFS |
2405 | { |
2406 | struct ipv6_route_iter *iter = w->args; | |
2407 | ||
2408 | if (!iter->skip) | |
2409 | return 1; | |
2410 | ||
2411 | do { | |
66f5d6ce | 2412 | iter->w.leaf = rcu_dereference_protected( |
8fb11a9a | 2413 | iter->w.leaf->fib6_next, |
66f5d6ce | 2414 | lockdep_is_held(&iter->tbl->tb6_lock)); |
8d2ca1d7 HFS |
2415 | iter->skip--; |
2416 | if (!iter->skip && iter->w.leaf) | |
2417 | return 1; | |
2418 | } while (iter->w.leaf); | |
2419 | ||
2420 | return 0; | |
2421 | } | |
2422 | ||
9a03cd8f MK |
2423 | static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter, |
2424 | struct net *net) | |
8d2ca1d7 HFS |
2425 | { |
2426 | memset(&iter->w, 0, sizeof(iter->w)); | |
2427 | iter->w.func = ipv6_route_yield; | |
2428 | iter->w.root = &iter->tbl->tb6_root; | |
2429 | iter->w.state = FWS_INIT; | |
2430 | iter->w.node = iter->w.root; | |
2431 | iter->w.args = iter; | |
0a67d3ef | 2432 | iter->sernum = iter->w.root->fn_sernum; |
8d2ca1d7 | 2433 | INIT_LIST_HEAD(&iter->w.lh); |
9a03cd8f | 2434 | fib6_walker_link(net, &iter->w); |
8d2ca1d7 HFS |
2435 | } |
2436 | ||
2437 | static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl, | |
2438 | struct net *net) | |
2439 | { | |
2440 | unsigned int h; | |
2441 | struct hlist_node *node; | |
2442 | ||
2443 | if (tbl) { | |
2444 | h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1; | |
2445 | node = rcu_dereference_bh(hlist_next_rcu(&tbl->tb6_hlist)); | |
2446 | } else { | |
2447 | h = 0; | |
2448 | node = NULL; | |
2449 | } | |
2450 | ||
2451 | while (!node && h < FIB6_TABLE_HASHSZ) { | |
2452 | node = rcu_dereference_bh( | |
2453 | hlist_first_rcu(&net->ipv6.fib_table_hash[h++])); | |
2454 | } | |
2455 | return hlist_entry_safe(node, struct fib6_table, tb6_hlist); | |
2456 | } | |
2457 | ||
0a67d3ef HFS |
2458 | static void ipv6_route_check_sernum(struct ipv6_route_iter *iter) |
2459 | { | |
2460 | if (iter->sernum != iter->w.root->fn_sernum) { | |
2461 | iter->sernum = iter->w.root->fn_sernum; | |
2462 | iter->w.state = FWS_INIT; | |
2463 | iter->w.node = iter->w.root; | |
2464 | WARN_ON(iter->w.skip); | |
2465 | iter->w.skip = iter->w.count; | |
2466 | } | |
2467 | } | |
2468 | ||
8d2ca1d7 HFS |
2469 | static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos) |
2470 | { | |
2471 | int r; | |
8d1c802b | 2472 | struct fib6_info *n; |
8d2ca1d7 HFS |
2473 | struct net *net = seq_file_net(seq); |
2474 | struct ipv6_route_iter *iter = seq->private; | |
2475 | ||
2476 | if (!v) | |
2477 | goto iter_table; | |
2478 | ||
8fb11a9a | 2479 | n = rcu_dereference_bh(((struct fib6_info *)v)->fib6_next); |
8d2ca1d7 HFS |
2480 | if (n) { |
2481 | ++*pos; | |
2482 | return n; | |
2483 | } | |
2484 | ||
2485 | iter_table: | |
0a67d3ef | 2486 | ipv6_route_check_sernum(iter); |
66f5d6ce | 2487 | spin_lock_bh(&iter->tbl->tb6_lock); |
8d2ca1d7 | 2488 | r = fib6_walk_continue(&iter->w); |
66f5d6ce | 2489 | spin_unlock_bh(&iter->tbl->tb6_lock); |
8d2ca1d7 HFS |
2490 | if (r > 0) { |
2491 | if (v) | |
2492 | ++*pos; | |
2493 | return iter->w.leaf; | |
2494 | } else if (r < 0) { | |
9a03cd8f | 2495 | fib6_walker_unlink(net, &iter->w); |
8d2ca1d7 HFS |
2496 | return NULL; |
2497 | } | |
9a03cd8f | 2498 | fib6_walker_unlink(net, &iter->w); |
8d2ca1d7 HFS |
2499 | |
2500 | iter->tbl = ipv6_route_seq_next_table(iter->tbl, net); | |
2501 | if (!iter->tbl) | |
2502 | return NULL; | |
2503 | ||
9a03cd8f | 2504 | ipv6_route_seq_setup_walk(iter, net); |
8d2ca1d7 HFS |
2505 | goto iter_table; |
2506 | } | |
2507 | ||
2508 | static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos) | |
2509 | __acquires(RCU_BH) | |
2510 | { | |
2511 | struct net *net = seq_file_net(seq); | |
2512 | struct ipv6_route_iter *iter = seq->private; | |
2513 | ||
2514 | rcu_read_lock_bh(); | |
2515 | iter->tbl = ipv6_route_seq_next_table(NULL, net); | |
2516 | iter->skip = *pos; | |
2517 | ||
2518 | if (iter->tbl) { | |
9a03cd8f | 2519 | ipv6_route_seq_setup_walk(iter, net); |
8d2ca1d7 HFS |
2520 | return ipv6_route_seq_next(seq, NULL, pos); |
2521 | } else { | |
2522 | return NULL; | |
2523 | } | |
2524 | } | |
2525 | ||
2526 | static bool ipv6_route_iter_active(struct ipv6_route_iter *iter) | |
2527 | { | |
94b2cfe0 | 2528 | struct fib6_walker *w = &iter->w; |
8d2ca1d7 HFS |
2529 | return w->node && !(w->state == FWS_U && w->node == w->root); |
2530 | } | |
2531 | ||
2532 | static void ipv6_route_seq_stop(struct seq_file *seq, void *v) | |
2533 | __releases(RCU_BH) | |
2534 | { | |
9a03cd8f | 2535 | struct net *net = seq_file_net(seq); |
8d2ca1d7 HFS |
2536 | struct ipv6_route_iter *iter = seq->private; |
2537 | ||
2538 | if (ipv6_route_iter_active(iter)) | |
9a03cd8f | 2539 | fib6_walker_unlink(net, &iter->w); |
8d2ca1d7 HFS |
2540 | |
2541 | rcu_read_unlock_bh(); | |
2542 | } | |
2543 | ||
c3506372 | 2544 | const struct seq_operations ipv6_route_seq_ops = { |
8d2ca1d7 HFS |
2545 | .start = ipv6_route_seq_start, |
2546 | .next = ipv6_route_seq_next, | |
2547 | .stop = ipv6_route_seq_stop, | |
2548 | .show = ipv6_route_seq_show | |
2549 | }; | |
8d2ca1d7 | 2550 | #endif /* CONFIG_PROC_FS */ |