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 | ||
b6459415 | 18 | #include <linux/bpf.h> |
1da177e4 LT |
19 | #include <linux/errno.h> |
20 | #include <linux/types.h> | |
21 | #include <linux/net.h> | |
22 | #include <linux/route.h> | |
23 | #include <linux/netdevice.h> | |
24 | #include <linux/in6.h> | |
25 | #include <linux/init.h> | |
c71099ac | 26 | #include <linux/list.h> |
5a0e3ad6 | 27 | #include <linux/slab.h> |
1da177e4 | 28 | |
cc5f0eb2 | 29 | #include <net/ip.h> |
1da177e4 LT |
30 | #include <net/ipv6.h> |
31 | #include <net/ndisc.h> | |
32 | #include <net/addrconf.h> | |
19e42e45 | 33 | #include <net/lwtunnel.h> |
df77fe4d | 34 | #include <net/fib_notifier.h> |
1da177e4 | 35 | |
ed13923f | 36 | #include <net/ip_fib.h> |
1da177e4 LT |
37 | #include <net/ip6_fib.h> |
38 | #include <net/ip6_route.h> | |
39 | ||
437de07c | 40 | static struct kmem_cache *fib6_node_kmem __read_mostly; |
1da177e4 | 41 | |
94b2cfe0 HFS |
42 | struct fib6_cleaner { |
43 | struct fib6_walker w; | |
ec7d43c2 | 44 | struct net *net; |
8d1c802b | 45 | int (*func)(struct fib6_info *, void *arg); |
327571cb | 46 | int sernum; |
1da177e4 | 47 | void *arg; |
7c6bb7d2 | 48 | bool skip_notify; |
1da177e4 LT |
49 | }; |
50 | ||
1da177e4 LT |
51 | #ifdef CONFIG_IPV6_SUBTREES |
52 | #define FWS_INIT FWS_S | |
1da177e4 LT |
53 | #else |
54 | #define FWS_INIT FWS_L | |
1da177e4 LT |
55 | #endif |
56 | ||
8d1c802b | 57 | static struct fib6_info *fib6_find_prefix(struct net *net, |
66f5d6ce WW |
58 | struct fib6_table *table, |
59 | struct fib6_node *fn); | |
60 | static struct fib6_node *fib6_repair_tree(struct net *net, | |
61 | struct fib6_table *table, | |
62 | struct fib6_node *fn); | |
9a03cd8f | 63 | static int fib6_walk(struct net *net, struct fib6_walker *w); |
94b2cfe0 | 64 | static int fib6_walk_continue(struct fib6_walker *w); |
1da177e4 LT |
65 | |
66 | /* | |
67 | * A routing update causes an increase of the serial number on the | |
68 | * affected subtree. This allows for cached routes to be asynchronously | |
69 | * tested when modifications are made to the destination cache as a | |
70 | * result of redirects, path MTU changes, etc. | |
71 | */ | |
72 | ||
86cb30ec | 73 | static void fib6_gc_timer_cb(struct timer_list *t); |
5b7c931d | 74 | |
9a03cd8f MK |
75 | #define FOR_WALKERS(net, w) \ |
76 | list_for_each_entry(w, &(net)->ipv6.fib6_walkers, lh) | |
1da177e4 | 77 | |
9a03cd8f | 78 | static void fib6_walker_link(struct net *net, struct fib6_walker *w) |
90d41122 | 79 | { |
9a03cd8f MK |
80 | write_lock_bh(&net->ipv6.fib6_walker_lock); |
81 | list_add(&w->lh, &net->ipv6.fib6_walkers); | |
82 | write_unlock_bh(&net->ipv6.fib6_walker_lock); | |
90d41122 AB |
83 | } |
84 | ||
9a03cd8f | 85 | static void fib6_walker_unlink(struct net *net, struct fib6_walker *w) |
90d41122 | 86 | { |
9a03cd8f | 87 | write_lock_bh(&net->ipv6.fib6_walker_lock); |
bbef49da | 88 | list_del(&w->lh); |
9a03cd8f | 89 | write_unlock_bh(&net->ipv6.fib6_walker_lock); |
90d41122 | 90 | } |
94b2cfe0 | 91 | |
812918c4 | 92 | static int fib6_new_sernum(struct net *net) |
1da177e4 | 93 | { |
30189806 | 94 | int new, old = atomic_read(&net->ipv6.fib6_sernum); |
42b18706 HFS |
95 | |
96 | do { | |
42b18706 | 97 | new = old < INT_MAX ? old + 1 : 1; |
30189806 ED |
98 | } while (!atomic_try_cmpxchg(&net->ipv6.fib6_sernum, &old, new)); |
99 | ||
42b18706 | 100 | return new; |
1da177e4 LT |
101 | } |
102 | ||
327571cb HFS |
103 | enum { |
104 | FIB6_NO_SERNUM_CHANGE = 0, | |
105 | }; | |
106 | ||
93c2fb25 | 107 | void fib6_update_sernum(struct net *net, struct fib6_info *f6i) |
180ca444 | 108 | { |
180ca444 WW |
109 | struct fib6_node *fn; |
110 | ||
93c2fb25 DA |
111 | fn = rcu_dereference_protected(f6i->fib6_node, |
112 | lockdep_is_held(&f6i->fib6_table->tb6_lock)); | |
180ca444 | 113 | if (fn) |
aafc2e32 | 114 | WRITE_ONCE(fn->fn_sernum, fib6_new_sernum(net)); |
180ca444 WW |
115 | } |
116 | ||
1da177e4 LT |
117 | /* |
118 | * Auxiliary address test functions for the radix tree. | |
119 | * | |
1ab1457c | 120 | * These assume a 32bit processor (although it will work on |
1da177e4 LT |
121 | * 64bit processors) |
122 | */ | |
123 | ||
124 | /* | |
125 | * test bit | |
126 | */ | |
02cdce53 YH |
127 | #if defined(__LITTLE_ENDIAN) |
128 | # define BITOP_BE32_SWIZZLE (0x1F & ~7) | |
129 | #else | |
130 | # define BITOP_BE32_SWIZZLE 0 | |
131 | #endif | |
1da177e4 | 132 | |
94b2cfe0 | 133 | static __be32 addr_bit_set(const void *token, int fn_bit) |
1da177e4 | 134 | { |
b71d1d42 | 135 | const __be32 *addr = token; |
02cdce53 YH |
136 | /* |
137 | * Here, | |
8db46f1d | 138 | * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f) |
02cdce53 YH |
139 | * is optimized version of |
140 | * htonl(1 << ((~fn_bit)&0x1F)) | |
141 | * See include/asm-generic/bitops/le.h. | |
142 | */ | |
0eae88f3 ED |
143 | return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) & |
144 | addr[fn_bit >> 5]; | |
1da177e4 LT |
145 | } |
146 | ||
1cf844c7 | 147 | struct fib6_info *fib6_info_alloc(gfp_t gfp_flags, bool with_fib6_nh) |
a64efe14 | 148 | { |
8d1c802b | 149 | struct fib6_info *f6i; |
1cf844c7 | 150 | size_t sz = sizeof(*f6i); |
a64efe14 | 151 | |
1cf844c7 DA |
152 | if (with_fib6_nh) |
153 | sz += sizeof(struct fib6_nh); | |
154 | ||
155 | f6i = kzalloc(sz, gfp_flags); | |
a64efe14 DA |
156 | if (!f6i) |
157 | return NULL; | |
158 | ||
f88d8ea6 | 159 | /* fib6_siblings is a union with nh_list, so this initializes both */ |
93c2fb25 | 160 | INIT_LIST_HEAD(&f6i->fib6_siblings); |
f05713e0 | 161 | refcount_set(&f6i->fib6_ref, 1); |
a64efe14 | 162 | |
5eb902b8 KFL |
163 | INIT_HLIST_NODE(&f6i->gc_link); |
164 | ||
a64efe14 DA |
165 | return f6i; |
166 | } | |
167 | ||
9b0a8da8 | 168 | void fib6_info_destroy_rcu(struct rcu_head *head) |
a64efe14 | 169 | { |
9b0a8da8 | 170 | struct fib6_info *f6i = container_of(head, struct fib6_info, rcu); |
a64efe14 | 171 | |
93c2fb25 | 172 | WARN_ON(f6i->fib6_node); |
a64efe14 | 173 | |
f88d8ea6 DA |
174 | if (f6i->nh) |
175 | nexthop_put(f6i->nh); | |
176 | else | |
177 | fib6_nh_release(f6i->fib6_nh); | |
178 | ||
cc5f0eb2 | 179 | ip_fib_metrics_put(f6i->fib6_metrics); |
a64efe14 DA |
180 | kfree(f6i); |
181 | } | |
9b0a8da8 | 182 | EXPORT_SYMBOL_GPL(fib6_info_destroy_rcu); |
a64efe14 | 183 | |
81eb8447 | 184 | static struct fib6_node *node_alloc(struct net *net) |
1da177e4 LT |
185 | { |
186 | struct fib6_node *fn; | |
187 | ||
c3762229 | 188 | fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC); |
81eb8447 WW |
189 | if (fn) |
190 | net->ipv6.rt6_stats->fib_nodes++; | |
1da177e4 LT |
191 | |
192 | return fn; | |
193 | } | |
194 | ||
81eb8447 | 195 | static void node_free_immediate(struct net *net, struct fib6_node *fn) |
c5cff856 WW |
196 | { |
197 | kmem_cache_free(fib6_node_kmem, fn); | |
81eb8447 | 198 | net->ipv6.rt6_stats->fib_nodes--; |
c5cff856 WW |
199 | } |
200 | ||
81eb8447 | 201 | static void node_free(struct net *net, struct fib6_node *fn) |
c5cff856 | 202 | { |
85e48bcf | 203 | kfree_rcu(fn, rcu); |
81eb8447 | 204 | net->ipv6.rt6_stats->fib_nodes--; |
c5cff856 WW |
205 | } |
206 | ||
ba1cc08d SD |
207 | static void fib6_free_table(struct fib6_table *table) |
208 | { | |
209 | inetpeer_invalidate_tree(&table->tb6_peers); | |
210 | kfree(table); | |
211 | } | |
212 | ||
58f09b78 | 213 | static void fib6_link_table(struct net *net, struct fib6_table *tb) |
1b43af54 PM |
214 | { |
215 | unsigned int h; | |
216 | ||
375216ad TG |
217 | /* |
218 | * Initialize table lock at a single place to give lockdep a key, | |
219 | * tables aren't visible prior to being linked to the list. | |
220 | */ | |
66f5d6ce | 221 | spin_lock_init(&tb->tb6_lock); |
a33bc5c1 | 222 | h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1); |
1b43af54 PM |
223 | |
224 | /* | |
225 | * No protection necessary, this is the only list mutatation | |
226 | * operation, tables never disappear once they exist. | |
227 | */ | |
58f09b78 | 228 | hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]); |
1b43af54 | 229 | } |
c71099ac | 230 | |
1b43af54 | 231 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 232 | |
8ed67789 | 233 | static struct fib6_table *fib6_alloc_table(struct net *net, u32 id) |
c71099ac TG |
234 | { |
235 | struct fib6_table *table; | |
236 | ||
237 | table = kzalloc(sizeof(*table), GFP_ATOMIC); | |
507c9b1e | 238 | if (table) { |
c71099ac | 239 | table->tb6_id = id; |
66f5d6ce | 240 | rcu_assign_pointer(table->tb6_root.leaf, |
421842ed | 241 | net->ipv6.fib6_null_entry); |
c71099ac | 242 | table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; |
8e773277 | 243 | inet_peer_base_init(&table->tb6_peers); |
5eb902b8 | 244 | INIT_HLIST_HEAD(&table->tb6_gc_hlist); |
c71099ac TG |
245 | } |
246 | ||
247 | return table; | |
248 | } | |
249 | ||
58f09b78 | 250 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac | 251 | { |
834d9784 | 252 | struct fib6_table *tb, *new_tb; |
c71099ac TG |
253 | |
254 | if (id == 0) | |
255 | id = RT6_TABLE_MAIN; | |
834d9784 | 256 | |
58f09b78 | 257 | tb = fib6_get_table(net, id); |
c71099ac TG |
258 | if (tb) |
259 | return tb; | |
260 | ||
834d9784 KI |
261 | new_tb = fib6_alloc_table(net, id); |
262 | if (!new_tb) | |
263 | return NULL; | |
264 | ||
265 | spin_lock_bh(&net->ipv6.fib_table_hash_lock); | |
266 | ||
267 | tb = fib6_get_table(net, id); | |
268 | if (unlikely(tb)) { | |
269 | spin_unlock_bh(&net->ipv6.fib_table_hash_lock); | |
270 | kfree(new_tb); | |
271 | return tb; | |
272 | } | |
c71099ac | 273 | |
834d9784 KI |
274 | fib6_link_table(net, new_tb); |
275 | ||
276 | spin_unlock_bh(&net->ipv6.fib_table_hash_lock); | |
277 | ||
278 | return new_tb; | |
c71099ac | 279 | } |
b3b4663c | 280 | EXPORT_SYMBOL_GPL(fib6_new_table); |
c71099ac | 281 | |
58f09b78 | 282 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac | 283 | { |
58f09b78 | 284 | struct hlist_head *head; |
f1a8d107 | 285 | struct fib6_table *tb; |
c71099ac | 286 | |
f1a8d107 | 287 | if (!id) |
c71099ac | 288 | id = RT6_TABLE_MAIN; |
f1a8d107 KI |
289 | |
290 | head = &net->ipv6.fib_table_hash[id & (FIB6_TABLE_HASHSZ - 1)]; | |
291 | ||
292 | /* See comment in fib6_link_table(). RCU is not required, | |
293 | * but rcu_dereference_raw() is used to avoid data-race. | |
294 | */ | |
295 | hlist_for_each_entry_rcu(tb, head, tb6_hlist, true) | |
296 | if (tb->tb6_id == id) | |
c71099ac | 297 | return tb; |
c71099ac TG |
298 | |
299 | return NULL; | |
300 | } | |
c4850687 | 301 | EXPORT_SYMBOL_GPL(fib6_get_table); |
c71099ac | 302 | |
2c8c1e72 | 303 | static void __net_init fib6_tables_init(struct net *net) |
c71099ac | 304 | { |
58f09b78 DL |
305 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
306 | fib6_link_table(net, net->ipv6.fib6_local_tbl); | |
c71099ac | 307 | } |
c71099ac TG |
308 | #else |
309 | ||
58f09b78 | 310 | struct fib6_table *fib6_new_table(struct net *net, u32 id) |
c71099ac | 311 | { |
58f09b78 | 312 | return fib6_get_table(net, id); |
c71099ac TG |
313 | } |
314 | ||
58f09b78 | 315 | struct fib6_table *fib6_get_table(struct net *net, u32 id) |
c71099ac | 316 | { |
58f09b78 | 317 | return net->ipv6.fib6_main_tbl; |
c71099ac TG |
318 | } |
319 | ||
4c9483b2 | 320 | struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6, |
b75cc8f9 | 321 | const struct sk_buff *skb, |
58f09b78 | 322 | int flags, pol_lookup_t lookup) |
c71099ac | 323 | { |
ab997ad4 | 324 | struct rt6_info *rt; |
325 | ||
55cced4f BV |
326 | rt = pol_lookup_func(lookup, |
327 | net, net->ipv6.fib6_main_tbl, fl6, skb, flags); | |
07f61557 | 328 | if (rt->dst.error == -EAGAIN) { |
d64a1f57 | 329 | ip6_rt_put_flags(rt, flags); |
ab997ad4 | 330 | rt = net->ipv6.ip6_null_entry; |
7b09c2d0 | 331 | if (!(flags & RT6_LOOKUP_F_DST_NOREF)) |
d64a1f57 | 332 | dst_hold(&rt->dst); |
ab997ad4 | 333 | } |
334 | ||
335 | return &rt->dst; | |
c71099ac TG |
336 | } |
337 | ||
138118ec | 338 | /* called with rcu lock held; no reference taken on fib6_info */ |
effda4dd DA |
339 | int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6, |
340 | struct fib6_result *res, int flags) | |
138118ec | 341 | { |
effda4dd DA |
342 | return fib6_table_lookup(net, net->ipv6.fib6_main_tbl, oif, fl6, |
343 | res, flags); | |
138118ec DA |
344 | } |
345 | ||
2c8c1e72 | 346 | static void __net_init fib6_tables_init(struct net *net) |
c71099ac | 347 | { |
58f09b78 | 348 | fib6_link_table(net, net->ipv6.fib6_main_tbl); |
c71099ac TG |
349 | } |
350 | ||
351 | #endif | |
352 | ||
e60ea454 | 353 | unsigned int fib6_tables_seq_read(const struct net *net) |
e1ee0a5b IS |
354 | { |
355 | unsigned int h, fib_seq = 0; | |
356 | ||
357 | rcu_read_lock(); | |
358 | for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { | |
e60ea454 ED |
359 | const struct hlist_head *head = &net->ipv6.fib_table_hash[h]; |
360 | const struct fib6_table *tb; | |
e1ee0a5b | 361 | |
66f5d6ce | 362 | hlist_for_each_entry_rcu(tb, head, tb6_hlist) |
e60ea454 | 363 | fib_seq += READ_ONCE(tb->fib_seq); |
e1ee0a5b IS |
364 | } |
365 | rcu_read_unlock(); | |
366 | ||
367 | return fib_seq; | |
368 | } | |
369 | ||
7c550daf | 370 | static int call_fib6_entry_notifier(struct notifier_block *nb, |
e1ee0a5b | 371 | enum fib_event_type event_type, |
b7a59557 JP |
372 | struct fib6_info *rt, |
373 | struct netlink_ext_ack *extack) | |
e1ee0a5b IS |
374 | { |
375 | struct fib6_entry_notifier_info info = { | |
b7a59557 | 376 | .info.extack = extack, |
e1ee0a5b IS |
377 | .rt = rt, |
378 | }; | |
379 | ||
7c550daf | 380 | return call_fib6_notifier(nb, event_type, &info.info); |
e1ee0a5b IS |
381 | } |
382 | ||
9c6ecd3c IS |
383 | static int call_fib6_multipath_entry_notifier(struct notifier_block *nb, |
384 | enum fib_event_type event_type, | |
385 | struct fib6_info *rt, | |
386 | unsigned int nsiblings, | |
387 | struct netlink_ext_ack *extack) | |
388 | { | |
389 | struct fib6_entry_notifier_info info = { | |
390 | .info.extack = extack, | |
391 | .rt = rt, | |
392 | .nsiblings = nsiblings, | |
393 | }; | |
394 | ||
395 | return call_fib6_notifier(nb, event_type, &info.info); | |
396 | } | |
397 | ||
19a3b7ee DA |
398 | int call_fib6_entry_notifiers(struct net *net, |
399 | enum fib_event_type event_type, | |
400 | struct fib6_info *rt, | |
401 | struct netlink_ext_ack *extack) | |
df77fe4d IS |
402 | { |
403 | struct fib6_entry_notifier_info info = { | |
6c31e5a9 | 404 | .info.extack = extack, |
df77fe4d IS |
405 | .rt = rt, |
406 | }; | |
407 | ||
e60ea454 | 408 | WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1); |
df77fe4d IS |
409 | return call_fib6_notifiers(net, event_type, &info.info); |
410 | } | |
411 | ||
d4b96c7b IS |
412 | int call_fib6_multipath_entry_notifiers(struct net *net, |
413 | enum fib_event_type event_type, | |
414 | struct fib6_info *rt, | |
415 | unsigned int nsiblings, | |
416 | struct netlink_ext_ack *extack) | |
417 | { | |
418 | struct fib6_entry_notifier_info info = { | |
419 | .info.extack = extack, | |
420 | .rt = rt, | |
421 | .nsiblings = nsiblings, | |
d4b96c7b IS |
422 | }; |
423 | ||
e60ea454 | 424 | WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1); |
d4b96c7b IS |
425 | return call_fib6_notifiers(net, event_type, &info.info); |
426 | } | |
427 | ||
d2f0c9b1 IS |
428 | int call_fib6_entry_notifiers_replace(struct net *net, struct fib6_info *rt) |
429 | { | |
430 | struct fib6_entry_notifier_info info = { | |
431 | .rt = rt, | |
432 | .nsiblings = rt->fib6_nsiblings, | |
433 | }; | |
434 | ||
e60ea454 | 435 | WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1); |
caafb250 | 436 | return call_fib6_notifiers(net, FIB_EVENT_ENTRY_REPLACE, &info.info); |
d2f0c9b1 IS |
437 | } |
438 | ||
e1ee0a5b IS |
439 | struct fib6_dump_arg { |
440 | struct net *net; | |
441 | struct notifier_block *nb; | |
b7a59557 | 442 | struct netlink_ext_ack *extack; |
e1ee0a5b IS |
443 | }; |
444 | ||
55c894f7 | 445 | static int fib6_rt_dump(struct fib6_info *rt, struct fib6_dump_arg *arg) |
e1ee0a5b | 446 | { |
caafb250 | 447 | enum fib_event_type fib_event = FIB_EVENT_ENTRY_REPLACE; |
9c6ecd3c IS |
448 | int err; |
449 | ||
450 | if (!rt || rt == arg->net->ipv6.fib6_null_entry) | |
451 | return 0; | |
452 | ||
453 | if (rt->fib6_nsiblings) | |
454 | err = call_fib6_multipath_entry_notifier(arg->nb, fib_event, | |
455 | rt, | |
456 | rt->fib6_nsiblings, | |
457 | arg->extack); | |
458 | else | |
459 | err = call_fib6_entry_notifier(arg->nb, fib_event, rt, | |
460 | arg->extack); | |
461 | ||
462 | return err; | |
463 | } | |
464 | ||
e1ee0a5b IS |
465 | static int fib6_node_dump(struct fib6_walker *w) |
466 | { | |
caafb250 | 467 | int err; |
9c6ecd3c | 468 | |
caafb250 | 469 | err = fib6_rt_dump(w->leaf, w->args); |
e1ee0a5b | 470 | w->leaf = NULL; |
55c894f7 | 471 | return err; |
e1ee0a5b IS |
472 | } |
473 | ||
55c894f7 JP |
474 | static int fib6_table_dump(struct net *net, struct fib6_table *tb, |
475 | struct fib6_walker *w) | |
e1ee0a5b | 476 | { |
55c894f7 JP |
477 | int err; |
478 | ||
e1ee0a5b | 479 | w->root = &tb->tb6_root; |
66f5d6ce | 480 | spin_lock_bh(&tb->tb6_lock); |
55c894f7 | 481 | err = fib6_walk(net, w); |
66f5d6ce | 482 | spin_unlock_bh(&tb->tb6_lock); |
55c894f7 | 483 | return err; |
e1ee0a5b IS |
484 | } |
485 | ||
486 | /* Called with rcu_read_lock() */ | |
b7a59557 JP |
487 | int fib6_tables_dump(struct net *net, struct notifier_block *nb, |
488 | struct netlink_ext_ack *extack) | |
e1ee0a5b IS |
489 | { |
490 | struct fib6_dump_arg arg; | |
491 | struct fib6_walker *w; | |
492 | unsigned int h; | |
55c894f7 | 493 | int err = 0; |
e1ee0a5b IS |
494 | |
495 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
496 | if (!w) | |
497 | return -ENOMEM; | |
498 | ||
499 | w->func = fib6_node_dump; | |
500 | arg.net = net; | |
501 | arg.nb = nb; | |
b7a59557 | 502 | arg.extack = extack; |
e1ee0a5b IS |
503 | w->args = &arg; |
504 | ||
505 | for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { | |
506 | struct hlist_head *head = &net->ipv6.fib_table_hash[h]; | |
507 | struct fib6_table *tb; | |
508 | ||
55c894f7 JP |
509 | hlist_for_each_entry_rcu(tb, head, tb6_hlist) { |
510 | err = fib6_table_dump(net, tb, w); | |
6f199552 | 511 | if (err) |
55c894f7 JP |
512 | goto out; |
513 | } | |
e1ee0a5b IS |
514 | } |
515 | ||
55c894f7 | 516 | out: |
e1ee0a5b IS |
517 | kfree(w); |
518 | ||
6f199552 VB |
519 | /* The tree traversal function should never return a positive value. */ |
520 | return err > 0 ? -EINVAL : err; | |
e1ee0a5b IS |
521 | } |
522 | ||
94b2cfe0 | 523 | static int fib6_dump_node(struct fib6_walker *w) |
1b43af54 PM |
524 | { |
525 | int res; | |
8d1c802b | 526 | struct fib6_info *rt; |
1b43af54 | 527 | |
66f5d6ce | 528 | for_each_fib6_walker_rt(w) { |
1e47b483 | 529 | res = rt6_dump_route(rt, w->args, w->skip_in_node); |
bf9a8a06 | 530 | if (res >= 0) { |
1b43af54 PM |
531 | /* Frame is full, suspend walking */ |
532 | w->leaf = rt; | |
1e47b483 SB |
533 | |
534 | /* We'll restart from this node, so if some routes were | |
535 | * already dumped, skip them next time. | |
536 | */ | |
537 | w->skip_in_node += res; | |
538 | ||
1b43af54 PM |
539 | return 1; |
540 | } | |
1e47b483 | 541 | w->skip_in_node = 0; |
beb1afac DA |
542 | |
543 | /* Multipath routes are dumped in one route with the | |
544 | * RTA_MULTIPATH attribute. Jump 'rt' to point to the | |
545 | * last sibling of this route (no need to dump the | |
546 | * sibling routes again) | |
547 | */ | |
93c2fb25 DA |
548 | if (rt->fib6_nsiblings) |
549 | rt = list_last_entry(&rt->fib6_siblings, | |
8d1c802b | 550 | struct fib6_info, |
93c2fb25 | 551 | fib6_siblings); |
1b43af54 PM |
552 | } |
553 | w->leaf = NULL; | |
554 | return 0; | |
555 | } | |
556 | ||
557 | static void fib6_dump_end(struct netlink_callback *cb) | |
558 | { | |
9a03cd8f | 559 | struct net *net = sock_net(cb->skb->sk); |
94b2cfe0 | 560 | struct fib6_walker *w = (void *)cb->args[2]; |
1b43af54 PM |
561 | |
562 | if (w) { | |
7891cc81 HX |
563 | if (cb->args[4]) { |
564 | cb->args[4] = 0; | |
9a03cd8f | 565 | fib6_walker_unlink(net, w); |
7891cc81 | 566 | } |
1b43af54 PM |
567 | cb->args[2] = 0; |
568 | kfree(w); | |
569 | } | |
437de07c | 570 | cb->done = (void *)cb->args[3]; |
1b43af54 PM |
571 | cb->args[1] = 3; |
572 | } | |
573 | ||
574 | static int fib6_dump_done(struct netlink_callback *cb) | |
575 | { | |
576 | fib6_dump_end(cb); | |
577 | return cb->done ? cb->done(cb) : 0; | |
578 | } | |
579 | ||
580 | static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb, | |
581 | struct netlink_callback *cb) | |
582 | { | |
9a03cd8f | 583 | struct net *net = sock_net(skb->sk); |
94b2cfe0 | 584 | struct fib6_walker *w; |
1b43af54 PM |
585 | int res; |
586 | ||
587 | w = (void *)cb->args[2]; | |
588 | w->root = &table->tb6_root; | |
589 | ||
590 | if (cb->args[4] == 0) { | |
2bec5a36 PM |
591 | w->count = 0; |
592 | w->skip = 0; | |
1e47b483 | 593 | w->skip_in_node = 0; |
2bec5a36 | 594 | |
66f5d6ce | 595 | spin_lock_bh(&table->tb6_lock); |
9a03cd8f | 596 | res = fib6_walk(net, w); |
66f5d6ce | 597 | spin_unlock_bh(&table->tb6_lock); |
2bec5a36 | 598 | if (res > 0) { |
1b43af54 | 599 | cb->args[4] = 1; |
aafc2e32 | 600 | cb->args[5] = READ_ONCE(w->root->fn_sernum); |
2bec5a36 | 601 | } |
1b43af54 | 602 | } else { |
aafc2e32 ED |
603 | int sernum = READ_ONCE(w->root->fn_sernum); |
604 | if (cb->args[5] != sernum) { | |
2bec5a36 | 605 | /* Begin at the root if the tree changed */ |
aafc2e32 | 606 | cb->args[5] = sernum; |
2bec5a36 PM |
607 | w->state = FWS_INIT; |
608 | w->node = w->root; | |
609 | w->skip = w->count; | |
1e47b483 | 610 | w->skip_in_node = 0; |
2bec5a36 PM |
611 | } else |
612 | w->skip = 0; | |
613 | ||
66f5d6ce | 614 | spin_lock_bh(&table->tb6_lock); |
1b43af54 | 615 | res = fib6_walk_continue(w); |
66f5d6ce | 616 | spin_unlock_bh(&table->tb6_lock); |
7891cc81 | 617 | if (res <= 0) { |
9a03cd8f | 618 | fib6_walker_unlink(net, w); |
7891cc81 | 619 | cb->args[4] = 0; |
1b43af54 | 620 | } |
1b43af54 | 621 | } |
7891cc81 | 622 | |
1b43af54 PM |
623 | return res; |
624 | } | |
625 | ||
c127ea2c | 626 | static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) |
1b43af54 | 627 | { |
22e36ea9 ED |
628 | struct rt6_rtnl_dump_arg arg = { |
629 | .filter.dump_exceptions = true, | |
630 | .filter.dump_routes = true, | |
5fc68320 | 631 | .filter.rtnl_held = false, |
22e36ea9 | 632 | }; |
e8ba330a | 633 | const struct nlmsghdr *nlh = cb->nlh; |
3b1e0a65 | 634 | struct net *net = sock_net(skb->sk); |
1b43af54 | 635 | unsigned int e = 0, s_e; |
5fc68320 | 636 | struct hlist_head *head; |
94b2cfe0 | 637 | struct fib6_walker *w; |
1b43af54 | 638 | struct fib6_table *tb; |
5fc68320 ED |
639 | unsigned int h, s_h; |
640 | int err = 0; | |
1b43af54 | 641 | |
5fc68320 | 642 | rcu_read_lock(); |
e8ba330a | 643 | if (cb->strict_check) { |
effe6792 | 644 | err = ip_valid_fib_dump_req(net, nlh, &arg.filter, cb); |
e8ba330a | 645 | if (err < 0) |
5fc68320 | 646 | goto unlock; |
13e38901 DA |
647 | } else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) { |
648 | struct rtmsg *rtm = nlmsg_data(nlh); | |
e8ba330a | 649 | |
ef11209d SB |
650 | if (rtm->rtm_flags & RTM_F_PREFIX) |
651 | arg.filter.flags = RTM_F_PREFIX; | |
13e38901 | 652 | } |
1b43af54 PM |
653 | |
654 | w = (void *)cb->args[2]; | |
507c9b1e | 655 | if (!w) { |
1b43af54 PM |
656 | /* New dump: |
657 | * | |
d21d4060 | 658 | * 1. allocate and initialize walker. |
1b43af54 PM |
659 | */ |
660 | w = kzalloc(sizeof(*w), GFP_ATOMIC); | |
5fc68320 ED |
661 | if (!w) { |
662 | err = -ENOMEM; | |
663 | goto unlock; | |
664 | } | |
1b43af54 PM |
665 | w->func = fib6_dump_node; |
666 | cb->args[2] = (long)w; | |
d21d4060 KI |
667 | |
668 | /* 2. hook callback destructor. | |
669 | */ | |
670 | cb->args[3] = (long)cb->done; | |
671 | cb->done = fib6_dump_done; | |
672 | ||
1b43af54 PM |
673 | } |
674 | ||
675 | arg.skb = skb; | |
676 | arg.cb = cb; | |
191cd582 | 677 | arg.net = net; |
1b43af54 PM |
678 | w->args = &arg; |
679 | ||
13e38901 DA |
680 | if (arg.filter.table_id) { |
681 | tb = fib6_get_table(net, arg.filter.table_id); | |
682 | if (!tb) { | |
41b4bd98 | 683 | if (rtnl_msg_family(cb->nlh) != PF_INET6) |
5fc68320 | 684 | goto unlock; |
ae677bbb | 685 | |
13e38901 | 686 | NL_SET_ERR_MSG_MOD(cb->extack, "FIB table does not exist"); |
5fc68320 ED |
687 | err = -ENOENT; |
688 | goto unlock; | |
13e38901 DA |
689 | } |
690 | ||
73155879 | 691 | if (!cb->args[0]) { |
5fc68320 ED |
692 | err = fib6_dump_table(tb, skb, cb); |
693 | if (!err) | |
73155879 DA |
694 | cb->args[0] = 1; |
695 | } | |
5fc68320 | 696 | goto unlock; |
13e38901 DA |
697 | } |
698 | ||
699 | s_h = cb->args[0]; | |
700 | s_e = cb->args[1]; | |
701 | ||
a33bc5c1 | 702 | for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) { |
1b43af54 | 703 | e = 0; |
58f09b78 | 704 | head = &net->ipv6.fib_table_hash[h]; |
b67bfe0d | 705 | hlist_for_each_entry_rcu(tb, head, tb6_hlist) { |
1b43af54 PM |
706 | if (e < s_e) |
707 | goto next; | |
5fc68320 ED |
708 | err = fib6_dump_table(tb, skb, cb); |
709 | if (err != 0) | |
710 | goto out; | |
1b43af54 PM |
711 | next: |
712 | e++; | |
713 | } | |
714 | } | |
5fc68320 | 715 | out: |
1b43af54 PM |
716 | cb->args[1] = e; |
717 | cb->args[0] = h; | |
5fc68320 ED |
718 | |
719 | unlock: | |
720 | rcu_read_unlock(); | |
721 | if (err <= 0) | |
1b43af54 | 722 | fib6_dump_end(cb); |
5fc68320 | 723 | return err; |
1b43af54 | 724 | } |
1da177e4 | 725 | |
8d1c802b | 726 | void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val) |
d4ead6b3 DA |
727 | { |
728 | if (!f6i) | |
729 | return; | |
730 | ||
731 | if (f6i->fib6_metrics == &dst_default_metrics) { | |
732 | struct dst_metrics *p = kzalloc(sizeof(*p), GFP_ATOMIC); | |
733 | ||
734 | if (!p) | |
735 | return; | |
736 | ||
737 | refcount_set(&p->refcnt, 1); | |
738 | f6i->fib6_metrics = p; | |
739 | } | |
740 | ||
741 | f6i->fib6_metrics->metrics[metric - 1] = val; | |
742 | } | |
743 | ||
1da177e4 LT |
744 | /* |
745 | * Routing Table | |
746 | * | |
747 | * return the appropriate node for a routing tree "add" operation | |
748 | * by either creating and inserting or by returning an existing | |
749 | * node. | |
750 | */ | |
751 | ||
81eb8447 WW |
752 | static struct fib6_node *fib6_add_1(struct net *net, |
753 | struct fib6_table *table, | |
66f5d6ce WW |
754 | struct fib6_node *root, |
755 | struct in6_addr *addr, int plen, | |
756 | int offset, int allow_create, | |
757 | int replace_required, | |
758 | struct netlink_ext_ack *extack) | |
1da177e4 LT |
759 | { |
760 | struct fib6_node *fn, *in, *ln; | |
761 | struct fib6_node *pn = NULL; | |
762 | struct rt6key *key; | |
763 | int bit; | |
1ab1457c | 764 | __be32 dir = 0; |
1da177e4 | 765 | |
1da177e4 LT |
766 | /* insert node in tree */ |
767 | ||
768 | fn = root; | |
769 | ||
770 | do { | |
8d1c802b | 771 | struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, |
66f5d6ce WW |
772 | lockdep_is_held(&table->tb6_lock)); |
773 | key = (struct rt6key *)((u8 *)leaf + offset); | |
1da177e4 LT |
774 | |
775 | /* | |
776 | * Prefix match | |
777 | */ | |
778 | if (plen < fn->fn_bit || | |
4a287eba | 779 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) { |
14df015b MV |
780 | if (!allow_create) { |
781 | if (replace_required) { | |
d5d531cb DA |
782 | NL_SET_ERR_MSG(extack, |
783 | "Can not replace route - no match found"); | |
f3213831 | 784 | pr_warn("Can't replace route, no match found\n"); |
14df015b MV |
785 | return ERR_PTR(-ENOENT); |
786 | } | |
f3213831 | 787 | pr_warn("NLM_F_CREATE should be set when creating new route\n"); |
14df015b | 788 | } |
1da177e4 | 789 | goto insert_above; |
4a287eba | 790 | } |
1ab1457c | 791 | |
1da177e4 LT |
792 | /* |
793 | * Exact match ? | |
794 | */ | |
1ab1457c | 795 | |
1da177e4 LT |
796 | if (plen == fn->fn_bit) { |
797 | /* clean up an intermediate node */ | |
507c9b1e | 798 | if (!(fn->fn_flags & RTN_RTINFO)) { |
66f5d6ce | 799 | RCU_INIT_POINTER(fn->leaf, NULL); |
93531c67 | 800 | fib6_info_release(leaf); |
4512c43e WW |
801 | /* remove null_entry in the root node */ |
802 | } else if (fn->fn_flags & RTN_TL_ROOT && | |
803 | rcu_access_pointer(fn->leaf) == | |
421842ed | 804 | net->ipv6.fib6_null_entry) { |
4512c43e | 805 | RCU_INIT_POINTER(fn->leaf, NULL); |
1da177e4 | 806 | } |
1ab1457c | 807 | |
1da177e4 LT |
808 | return fn; |
809 | } | |
810 | ||
811 | /* | |
812 | * We have more bits to go | |
813 | */ | |
1ab1457c | 814 | |
1da177e4 | 815 | /* Try to walk down on tree. */ |
1da177e4 LT |
816 | dir = addr_bit_set(addr, fn->fn_bit); |
817 | pn = fn; | |
66f5d6ce WW |
818 | fn = dir ? |
819 | rcu_dereference_protected(fn->right, | |
820 | lockdep_is_held(&table->tb6_lock)) : | |
821 | rcu_dereference_protected(fn->left, | |
822 | lockdep_is_held(&table->tb6_lock)); | |
1da177e4 LT |
823 | } while (fn); |
824 | ||
14df015b | 825 | if (!allow_create) { |
4a287eba MV |
826 | /* We should not create new node because |
827 | * NLM_F_REPLACE was specified without NLM_F_CREATE | |
828 | * I assume it is safe to require NLM_F_CREATE when | |
829 | * REPLACE flag is used! Later we may want to remove the | |
830 | * check for replace_required, because according | |
831 | * to netlink specification, NLM_F_CREATE | |
832 | * MUST be specified if new route is created. | |
833 | * That would keep IPv6 consistent with IPv4 | |
834 | */ | |
14df015b | 835 | if (replace_required) { |
d5d531cb DA |
836 | NL_SET_ERR_MSG(extack, |
837 | "Can not replace route - no match found"); | |
f3213831 | 838 | pr_warn("Can't replace route, no match found\n"); |
14df015b MV |
839 | return ERR_PTR(-ENOENT); |
840 | } | |
f3213831 | 841 | pr_warn("NLM_F_CREATE should be set when creating new route\n"); |
4a287eba | 842 | } |
1da177e4 LT |
843 | /* |
844 | * We walked to the bottom of tree. | |
845 | * Create new leaf node without children. | |
846 | */ | |
847 | ||
81eb8447 | 848 | ln = node_alloc(net); |
1da177e4 | 849 | |
507c9b1e | 850 | if (!ln) |
188c517a | 851 | return ERR_PTR(-ENOMEM); |
1da177e4 | 852 | ln->fn_bit = plen; |
66f5d6ce | 853 | RCU_INIT_POINTER(ln->parent, pn); |
1da177e4 LT |
854 | |
855 | if (dir) | |
66f5d6ce | 856 | rcu_assign_pointer(pn->right, ln); |
1da177e4 | 857 | else |
66f5d6ce | 858 | rcu_assign_pointer(pn->left, ln); |
1da177e4 LT |
859 | |
860 | return ln; | |
861 | ||
862 | ||
863 | insert_above: | |
864 | /* | |
1ab1457c | 865 | * split since we don't have a common prefix anymore or |
1da177e4 LT |
866 | * we have a less significant route. |
867 | * we've to insert an intermediate node on the list | |
868 | * this new node will point to the one we need to create | |
869 | * and the current | |
870 | */ | |
871 | ||
66f5d6ce WW |
872 | pn = rcu_dereference_protected(fn->parent, |
873 | lockdep_is_held(&table->tb6_lock)); | |
1da177e4 LT |
874 | |
875 | /* find 1st bit in difference between the 2 addrs. | |
876 | ||
971f359d | 877 | See comment in __ipv6_addr_diff: bit may be an invalid value, |
1da177e4 LT |
878 | but if it is >= plen, the value is ignored in any case. |
879 | */ | |
1ab1457c | 880 | |
9225b230 | 881 | bit = __ipv6_addr_diff(addr, &key->addr, sizeof(*addr)); |
1da177e4 | 882 | |
1ab1457c YH |
883 | /* |
884 | * (intermediate)[in] | |
1da177e4 LT |
885 | * / \ |
886 | * (new leaf node)[ln] (old node)[fn] | |
887 | */ | |
888 | if (plen > bit) { | |
81eb8447 WW |
889 | in = node_alloc(net); |
890 | ln = node_alloc(net); | |
1ab1457c | 891 | |
507c9b1e | 892 | if (!in || !ln) { |
1da177e4 | 893 | if (in) |
81eb8447 | 894 | node_free_immediate(net, in); |
1da177e4 | 895 | if (ln) |
81eb8447 | 896 | node_free_immediate(net, ln); |
188c517a | 897 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
898 | } |
899 | ||
1ab1457c YH |
900 | /* |
901 | * new intermediate node. | |
1da177e4 LT |
902 | * RTN_RTINFO will |
903 | * be off since that an address that chooses one of | |
904 | * the branches would not match less specific routes | |
905 | * in the other branch | |
906 | */ | |
907 | ||
908 | in->fn_bit = bit; | |
909 | ||
66f5d6ce | 910 | RCU_INIT_POINTER(in->parent, pn); |
1da177e4 | 911 | in->leaf = fn->leaf; |
5ea71528 ED |
912 | fib6_info_hold(rcu_dereference_protected(in->leaf, |
913 | lockdep_is_held(&table->tb6_lock))); | |
1da177e4 | 914 | |
1da177e4 LT |
915 | /* update parent pointer */ |
916 | if (dir) | |
66f5d6ce | 917 | rcu_assign_pointer(pn->right, in); |
1da177e4 | 918 | else |
66f5d6ce | 919 | rcu_assign_pointer(pn->left, in); |
1da177e4 LT |
920 | |
921 | ln->fn_bit = plen; | |
922 | ||
66f5d6ce WW |
923 | RCU_INIT_POINTER(ln->parent, in); |
924 | rcu_assign_pointer(fn->parent, in); | |
1da177e4 | 925 | |
1da177e4 | 926 | if (addr_bit_set(addr, bit)) { |
66f5d6ce WW |
927 | rcu_assign_pointer(in->right, ln); |
928 | rcu_assign_pointer(in->left, fn); | |
1da177e4 | 929 | } else { |
66f5d6ce WW |
930 | rcu_assign_pointer(in->left, ln); |
931 | rcu_assign_pointer(in->right, fn); | |
1da177e4 LT |
932 | } |
933 | } else { /* plen <= bit */ | |
934 | ||
1ab1457c | 935 | /* |
1da177e4 LT |
936 | * (new leaf node)[ln] |
937 | * / \ | |
938 | * (old node)[fn] NULL | |
939 | */ | |
940 | ||
81eb8447 | 941 | ln = node_alloc(net); |
1da177e4 | 942 | |
507c9b1e | 943 | if (!ln) |
188c517a | 944 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
945 | |
946 | ln->fn_bit = plen; | |
947 | ||
66f5d6ce | 948 | RCU_INIT_POINTER(ln->parent, pn); |
1da177e4 LT |
949 | |
950 | if (addr_bit_set(&key->addr, plen)) | |
66f5d6ce | 951 | RCU_INIT_POINTER(ln->right, fn); |
1da177e4 | 952 | else |
66f5d6ce WW |
953 | RCU_INIT_POINTER(ln->left, fn); |
954 | ||
955 | rcu_assign_pointer(fn->parent, ln); | |
1da177e4 | 956 | |
66f5d6ce WW |
957 | if (dir) |
958 | rcu_assign_pointer(pn->right, ln); | |
959 | else | |
960 | rcu_assign_pointer(pn->left, ln); | |
1da177e4 LT |
961 | } |
962 | return ln; | |
963 | } | |
964 | ||
7d88d8b5 DA |
965 | static void __fib6_drop_pcpu_from(struct fib6_nh *fib6_nh, |
966 | const struct fib6_info *match, | |
967 | const struct fib6_table *table) | |
e715b6d3 | 968 | { |
5bcaa41b | 969 | int cpu; |
e715b6d3 | 970 | |
f40b6ae2 DA |
971 | if (!fib6_nh->rt6i_pcpu) |
972 | return; | |
973 | ||
b01e1c03 | 974 | rcu_read_lock(); |
5bcaa41b DA |
975 | /* release the reference to this fib entry from |
976 | * all of its cached pcpu routes | |
977 | */ | |
978 | for_each_possible_cpu(cpu) { | |
979 | struct rt6_info **ppcpu_rt; | |
980 | struct rt6_info *pcpu_rt; | |
e715b6d3 | 981 | |
f40b6ae2 | 982 | ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu); |
b01e1c03 ED |
983 | |
984 | /* Paired with xchg() in rt6_get_pcpu_route() */ | |
985 | pcpu_rt = READ_ONCE(*ppcpu_rt); | |
7d88d8b5 DA |
986 | |
987 | /* only dropping the 'from' reference if the cached route | |
988 | * is using 'match'. The cached pcpu_rt->from only changes | |
989 | * from a fib6_info to NULL (ip6_dst_destroy); it can never | |
990 | * change from one fib6_info reference to another | |
991 | */ | |
992 | if (pcpu_rt && rcu_access_pointer(pcpu_rt->from) == match) { | |
a68886a6 | 993 | struct fib6_info *from; |
e715b6d3 | 994 | |
b4cb4a13 | 995 | from = unrcu_pointer(xchg(&pcpu_rt->from, NULL)); |
a68886a6 | 996 | fib6_info_release(from); |
5bcaa41b | 997 | } |
e5fd387a | 998 | } |
b01e1c03 | 999 | rcu_read_unlock(); |
e5fd387a MK |
1000 | } |
1001 | ||
2ab75bfb DA |
1002 | struct fib6_nh_pcpu_arg { |
1003 | struct fib6_info *from; | |
1004 | const struct fib6_table *table; | |
1005 | }; | |
1006 | ||
1007 | static int fib6_nh_drop_pcpu_from(struct fib6_nh *nh, void *_arg) | |
1008 | { | |
1009 | struct fib6_nh_pcpu_arg *arg = _arg; | |
1010 | ||
1011 | __fib6_drop_pcpu_from(nh, arg->from, arg->table); | |
1012 | return 0; | |
1013 | } | |
1014 | ||
7d88d8b5 DA |
1015 | static void fib6_drop_pcpu_from(struct fib6_info *f6i, |
1016 | const struct fib6_table *table) | |
1017 | { | |
7d88d8b5 DA |
1018 | /* Make sure rt6_make_pcpu_route() wont add other percpu routes |
1019 | * while we are cleaning them here. | |
1020 | */ | |
1021 | f6i->fib6_destroying = 1; | |
1022 | mb(); /* paired with the cmpxchg() in rt6_make_pcpu_route() */ | |
1023 | ||
2ab75bfb DA |
1024 | if (f6i->nh) { |
1025 | struct fib6_nh_pcpu_arg arg = { | |
1026 | .from = f6i, | |
1027 | .table = table | |
1028 | }; | |
1029 | ||
8e5f1bb8 KI |
1030 | rcu_read_lock(); |
1031 | nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_drop_pcpu_from, &arg); | |
1032 | rcu_read_unlock(); | |
2ab75bfb DA |
1033 | } else { |
1034 | struct fib6_nh *fib6_nh; | |
1035 | ||
1036 | fib6_nh = f6i->fib6_nh; | |
1037 | __fib6_drop_pcpu_from(fib6_nh, f6i, table); | |
1038 | } | |
7d88d8b5 DA |
1039 | } |
1040 | ||
8d1c802b | 1041 | static void fib6_purge_rt(struct fib6_info *rt, struct fib6_node *fn, |
6e9e16e6 HFS |
1042 | struct net *net) |
1043 | { | |
93c2fb25 | 1044 | struct fib6_table *table = rt->fib6_table; |
66f5d6ce | 1045 | |
d8f5c296 ST |
1046 | /* Flush all cached dst in exception table */ |
1047 | rt6_flush_exceptions(rt); | |
f40b6ae2 | 1048 | fib6_drop_pcpu_from(rt, table); |
61fb0d01 | 1049 | |
081efd18 KI |
1050 | if (rt->nh) { |
1051 | spin_lock(&rt->nh->lock); | |
1052 | ||
1053 | if (!list_empty(&rt->nh_list)) | |
1054 | list_del_init(&rt->nh_list); | |
1055 | ||
1056 | spin_unlock(&rt->nh->lock); | |
1057 | } | |
f88d8ea6 | 1058 | |
f05713e0 | 1059 | if (refcount_read(&rt->fib6_ref) != 1) { |
6e9e16e6 HFS |
1060 | /* This route is used as dummy address holder in some split |
1061 | * nodes. It is not leaked, but it still holds other resources, | |
1062 | * which must be released in time. So, scan ascendant nodes | |
1063 | * and replace dummy references to this route with references | |
1064 | * to still alive ones. | |
1065 | */ | |
1066 | while (fn) { | |
8d1c802b | 1067 | struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, |
66f5d6ce | 1068 | lockdep_is_held(&table->tb6_lock)); |
8d1c802b | 1069 | struct fib6_info *new_leaf; |
66f5d6ce WW |
1070 | if (!(fn->fn_flags & RTN_RTINFO) && leaf == rt) { |
1071 | new_leaf = fib6_find_prefix(net, table, fn); | |
5ea71528 | 1072 | fib6_info_hold(new_leaf); |
93531c67 | 1073 | |
66f5d6ce | 1074 | rcu_assign_pointer(fn->leaf, new_leaf); |
93531c67 | 1075 | fib6_info_release(rt); |
6e9e16e6 | 1076 | } |
66f5d6ce WW |
1077 | fn = rcu_dereference_protected(fn->parent, |
1078 | lockdep_is_held(&table->tb6_lock)); | |
6e9e16e6 | 1079 | } |
6e9e16e6 | 1080 | } |
5eb902b8 KFL |
1081 | |
1082 | fib6_clean_expires(rt); | |
1083 | fib6_remove_gc_list(rt); | |
6e9e16e6 HFS |
1084 | } |
1085 | ||
1da177e4 LT |
1086 | /* |
1087 | * Insert routing information in a node. | |
1088 | */ | |
1089 | ||
8d1c802b | 1090 | static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt, |
accb46b5 KI |
1091 | struct nl_info *info, struct netlink_ext_ack *extack, |
1092 | struct list_head *purge_list) | |
1da177e4 | 1093 | { |
8d1c802b | 1094 | struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, |
93c2fb25 | 1095 | lockdep_is_held(&rt->fib6_table->tb6_lock)); |
33bd5ac5 | 1096 | struct fib6_info *iter = NULL; |
8d1c802b | 1097 | struct fib6_info __rcu **ins; |
33bd5ac5 | 1098 | struct fib6_info __rcu **fallback_ins = NULL; |
507c9b1e DM |
1099 | int replace = (info->nlh && |
1100 | (info->nlh->nlmsg_flags & NLM_F_REPLACE)); | |
1101 | int add = (!info->nlh || | |
1102 | (info->nlh->nlmsg_flags & NLM_F_CREATE)); | |
4a287eba | 1103 | int found = 0; |
33bd5ac5 | 1104 | bool rt_can_ecmp = rt6_qualify_for_ecmp(rt); |
c10c4279 | 1105 | bool notify_sibling_rt = false; |
73483c12 | 1106 | u16 nlflags = NLM_F_EXCL; |
e5fd387a | 1107 | int err; |
1da177e4 | 1108 | |
33bd5ac5 | 1109 | if (info->nlh && (info->nlh->nlmsg_flags & NLM_F_APPEND)) |
1f5e29ce DA |
1110 | nlflags |= NLM_F_APPEND; |
1111 | ||
1da177e4 LT |
1112 | ins = &fn->leaf; |
1113 | ||
66f5d6ce | 1114 | for (iter = leaf; iter; |
8fb11a9a | 1115 | iter = rcu_dereference_protected(iter->fib6_next, |
93c2fb25 | 1116 | lockdep_is_held(&rt->fib6_table->tb6_lock))) { |
1da177e4 LT |
1117 | /* |
1118 | * Search for duplicates | |
1119 | */ | |
1120 | ||
93c2fb25 | 1121 | if (iter->fib6_metric == rt->fib6_metric) { |
1da177e4 LT |
1122 | /* |
1123 | * Same priority level | |
1124 | */ | |
507c9b1e DM |
1125 | if (info->nlh && |
1126 | (info->nlh->nlmsg_flags & NLM_F_EXCL)) | |
4a287eba | 1127 | return -EEXIST; |
73483c12 GN |
1128 | |
1129 | nlflags &= ~NLM_F_EXCL; | |
4a287eba | 1130 | if (replace) { |
33bd5ac5 DA |
1131 | if (rt_can_ecmp == rt6_qualify_for_ecmp(iter)) { |
1132 | found++; | |
1133 | break; | |
1134 | } | |
e404b8c7 | 1135 | fallback_ins = fallback_ins ?: ins; |
33bd5ac5 | 1136 | goto next_iter; |
4a287eba | 1137 | } |
1da177e4 | 1138 | |
f06b7549 | 1139 | if (rt6_duplicate_nexthop(iter, rt)) { |
93c2fb25 DA |
1140 | if (rt->fib6_nsiblings) |
1141 | rt->fib6_nsiblings = 0; | |
1142 | if (!(iter->fib6_flags & RTF_EXPIRES)) | |
1da177e4 | 1143 | return -EEXIST; |
5eb902b8 | 1144 | if (!(rt->fib6_flags & RTF_EXPIRES)) { |
dade3f6a | 1145 | fib6_clean_expires(iter); |
5eb902b8 KFL |
1146 | fib6_remove_gc_list(iter); |
1147 | } else { | |
dade3f6a | 1148 | fib6_set_expires(iter, rt->expires); |
5eb902b8 KFL |
1149 | fib6_add_gc_list(iter); |
1150 | } | |
15a81b41 DA |
1151 | |
1152 | if (rt->fib6_pmtu) | |
1153 | fib6_metric_set(iter, RTAX_MTU, | |
1154 | rt->fib6_pmtu); | |
1da177e4 LT |
1155 | return -EEXIST; |
1156 | } | |
33bd5ac5 DA |
1157 | /* If we have the same destination and the same metric, |
1158 | * but not the same gateway, then the route we try to | |
1159 | * add is sibling to this route, increment our counter | |
1160 | * of siblings, and later we will add our route to the | |
1161 | * list. | |
1162 | * Only static routes (which don't have flag | |
1163 | * RTF_EXPIRES) are used for ECMPv6. | |
1164 | * | |
1165 | * To avoid long list, we only had siblings if the | |
1166 | * route have a gateway. | |
1167 | */ | |
1168 | if (rt_can_ecmp && | |
1169 | rt6_qualify_for_ecmp(iter)) | |
1170 | rt->fib6_nsiblings++; | |
1da177e4 LT |
1171 | } |
1172 | ||
93c2fb25 | 1173 | if (iter->fib6_metric > rt->fib6_metric) |
1da177e4 LT |
1174 | break; |
1175 | ||
33bd5ac5 | 1176 | next_iter: |
8fb11a9a | 1177 | ins = &iter->fib6_next; |
27596472 MK |
1178 | } |
1179 | ||
33bd5ac5 | 1180 | if (fallback_ins && !found) { |
e404b8c7 BP |
1181 | /* No matching route with same ecmp-able-ness found, replace |
1182 | * first matching route | |
1183 | */ | |
33bd5ac5 DA |
1184 | ins = fallback_ins; |
1185 | iter = rcu_dereference_protected(*ins, | |
1186 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
1187 | found++; | |
1188 | } | |
1189 | ||
f11e6659 DM |
1190 | /* Reset round-robin state, if necessary */ |
1191 | if (ins == &fn->leaf) | |
1192 | fn->rr_ptr = NULL; | |
1193 | ||
51ebd318 | 1194 | /* Link this route to others same route. */ |
33bd5ac5 DA |
1195 | if (rt->fib6_nsiblings) { |
1196 | unsigned int fib6_nsiblings; | |
8d1c802b | 1197 | struct fib6_info *sibling, *temp_sibling; |
51ebd318 | 1198 | |
33bd5ac5 DA |
1199 | /* Find the first route that have the same metric */ |
1200 | sibling = leaf; | |
c10c4279 | 1201 | notify_sibling_rt = true; |
33bd5ac5 DA |
1202 | while (sibling) { |
1203 | if (sibling->fib6_metric == rt->fib6_metric && | |
1204 | rt6_qualify_for_ecmp(sibling)) { | |
d9ccb18f OEH |
1205 | list_add_tail_rcu(&rt->fib6_siblings, |
1206 | &sibling->fib6_siblings); | |
33bd5ac5 DA |
1207 | break; |
1208 | } | |
1209 | sibling = rcu_dereference_protected(sibling->fib6_next, | |
1210 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
c10c4279 | 1211 | notify_sibling_rt = false; |
51ebd318 ND |
1212 | } |
1213 | /* For each sibling in the list, increment the counter of | |
1214 | * siblings. BUG() if counters does not match, list of siblings | |
1215 | * is broken! | |
1216 | */ | |
33bd5ac5 | 1217 | fib6_nsiblings = 0; |
51ebd318 | 1218 | list_for_each_entry_safe(sibling, temp_sibling, |
33bd5ac5 | 1219 | &rt->fib6_siblings, fib6_siblings) { |
93c2fb25 | 1220 | sibling->fib6_nsiblings++; |
33bd5ac5 DA |
1221 | BUG_ON(sibling->fib6_nsiblings != rt->fib6_nsiblings); |
1222 | fib6_nsiblings++; | |
51ebd318 | 1223 | } |
33bd5ac5 | 1224 | BUG_ON(fib6_nsiblings != rt->fib6_nsiblings); |
8e5f1bb8 | 1225 | rcu_read_lock(); |
33bd5ac5 | 1226 | rt6_multipath_rebalance(temp_sibling); |
8e5f1bb8 | 1227 | rcu_read_unlock(); |
51ebd318 ND |
1228 | } |
1229 | ||
1da177e4 LT |
1230 | /* |
1231 | * insert node | |
1232 | */ | |
4a287eba MV |
1233 | if (!replace) { |
1234 | if (!add) | |
f3213831 | 1235 | pr_warn("NLM_F_CREATE should be set when creating new route\n"); |
4a287eba MV |
1236 | |
1237 | add: | |
73483c12 | 1238 | nlflags |= NLM_F_CREATE; |
e715b6d3 | 1239 | |
caafb250 IS |
1240 | /* The route should only be notified if it is the first |
1241 | * route in the node or if it is added as a sibling | |
1242 | * route to the first route in the node. | |
1243 | */ | |
1244 | if (!info->skip_notify_kernel && | |
1245 | (notify_sibling_rt || ins == &fn->leaf)) { | |
c10c4279 IS |
1246 | enum fib_event_type fib_event; |
1247 | ||
1248 | if (notify_sibling_rt) | |
1249 | fib_event = FIB_EVENT_ENTRY_APPEND; | |
1250 | else | |
caafb250 | 1251 | fib_event = FIB_EVENT_ENTRY_REPLACE; |
d5382fef | 1252 | err = call_fib6_entry_notifiers(info->nl_net, |
caafb250 IS |
1253 | fib_event, rt, |
1254 | extack); | |
54851aa9 IS |
1255 | if (err) { |
1256 | struct fib6_info *sibling, *next_sibling; | |
1257 | ||
1258 | /* If the route has siblings, then it first | |
1259 | * needs to be unlinked from them. | |
1260 | */ | |
1261 | if (!rt->fib6_nsiblings) | |
1262 | return err; | |
1263 | ||
1264 | list_for_each_entry_safe(sibling, next_sibling, | |
1265 | &rt->fib6_siblings, | |
1266 | fib6_siblings) | |
1267 | sibling->fib6_nsiblings--; | |
1268 | rt->fib6_nsiblings = 0; | |
d9ccb18f | 1269 | list_del_rcu(&rt->fib6_siblings); |
8e5f1bb8 | 1270 | rcu_read_lock(); |
54851aa9 | 1271 | rt6_multipath_rebalance(next_sibling); |
8e5f1bb8 | 1272 | rcu_read_unlock(); |
d5382fef | 1273 | return err; |
54851aa9 | 1274 | } |
d5382fef | 1275 | } |
2233000c | 1276 | |
8fb11a9a | 1277 | rcu_assign_pointer(rt->fib6_next, iter); |
5ea71528 | 1278 | fib6_info_hold(rt); |
93c2fb25 | 1279 | rcu_assign_pointer(rt->fib6_node, fn); |
66f5d6ce | 1280 | rcu_assign_pointer(*ins, rt); |
3b1137fe DA |
1281 | if (!info->skip_notify) |
1282 | inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags); | |
4a287eba MV |
1283 | info->nl_net->ipv6.rt6_stats->fib_rt_entries++; |
1284 | ||
507c9b1e | 1285 | if (!(fn->fn_flags & RTN_RTINFO)) { |
4a287eba MV |
1286 | info->nl_net->ipv6.rt6_stats->fib_route_nodes++; |
1287 | fn->fn_flags |= RTN_RTINFO; | |
1288 | } | |
1da177e4 | 1289 | |
4a287eba | 1290 | } else { |
33bd5ac5 | 1291 | int nsiblings; |
27596472 | 1292 | |
4a287eba MV |
1293 | if (!found) { |
1294 | if (add) | |
1295 | goto add; | |
f3213831 | 1296 | pr_warn("NLM_F_REPLACE set, but no existing node found!\n"); |
4a287eba MV |
1297 | return -ENOENT; |
1298 | } | |
e715b6d3 | 1299 | |
caafb250 | 1300 | if (!info->skip_notify_kernel && ins == &fn->leaf) { |
d5382fef IS |
1301 | err = call_fib6_entry_notifiers(info->nl_net, |
1302 | FIB_EVENT_ENTRY_REPLACE, | |
1303 | rt, extack); | |
1304 | if (err) | |
1305 | return err; | |
1306 | } | |
2233000c | 1307 | |
5ea71528 | 1308 | fib6_info_hold(rt); |
93c2fb25 | 1309 | rcu_assign_pointer(rt->fib6_node, fn); |
33bd5ac5 | 1310 | rt->fib6_next = iter->fib6_next; |
66f5d6ce | 1311 | rcu_assign_pointer(*ins, rt); |
3b1137fe DA |
1312 | if (!info->skip_notify) |
1313 | inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE); | |
507c9b1e | 1314 | if (!(fn->fn_flags & RTN_RTINFO)) { |
4a287eba MV |
1315 | info->nl_net->ipv6.rt6_stats->fib_route_nodes++; |
1316 | fn->fn_flags |= RTN_RTINFO; | |
1317 | } | |
33bd5ac5 DA |
1318 | nsiblings = iter->fib6_nsiblings; |
1319 | iter->fib6_node = NULL; | |
accb46b5 | 1320 | list_add(&iter->purge_link, purge_list); |
33bd5ac5 DA |
1321 | if (rcu_access_pointer(fn->rr_ptr) == iter) |
1322 | fn->rr_ptr = NULL; | |
27596472 | 1323 | |
33bd5ac5 | 1324 | if (nsiblings) { |
27596472 | 1325 | /* Replacing an ECMP route, remove all siblings */ |
33bd5ac5 DA |
1326 | ins = &rt->fib6_next; |
1327 | iter = rcu_dereference_protected(*ins, | |
1328 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
1329 | while (iter) { | |
1330 | if (iter->fib6_metric > rt->fib6_metric) | |
1331 | break; | |
1332 | if (rt6_qualify_for_ecmp(iter)) { | |
1333 | *ins = iter->fib6_next; | |
1334 | iter->fib6_node = NULL; | |
accb46b5 | 1335 | list_add(&iter->purge_link, purge_list); |
33bd5ac5 DA |
1336 | if (rcu_access_pointer(fn->rr_ptr) == iter) |
1337 | fn->rr_ptr = NULL; | |
33bd5ac5 DA |
1338 | nsiblings--; |
1339 | info->nl_net->ipv6.rt6_stats->fib_rt_entries--; | |
1340 | } else { | |
1341 | ins = &iter->fib6_next; | |
1342 | } | |
1343 | iter = rcu_dereference_protected(*ins, | |
1344 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
27596472 | 1345 | } |
33bd5ac5 | 1346 | WARN_ON(nsiblings != 0); |
27596472 | 1347 | } |
1da177e4 LT |
1348 | } |
1349 | ||
1350 | return 0; | |
1351 | } | |
1352 | ||
081efd18 KI |
1353 | static int fib6_add_rt2node_nh(struct fib6_node *fn, struct fib6_info *rt, |
1354 | struct nl_info *info, struct netlink_ext_ack *extack, | |
1355 | struct list_head *purge_list) | |
1356 | { | |
1357 | int err; | |
1358 | ||
1359 | spin_lock(&rt->nh->lock); | |
1360 | ||
1361 | if (rt->nh->dead) { | |
1362 | NL_SET_ERR_MSG(extack, "Nexthop has been deleted"); | |
1363 | err = -EINVAL; | |
1364 | } else { | |
1365 | err = fib6_add_rt2node(fn, rt, info, extack, purge_list); | |
1366 | if (!err) | |
1367 | list_add(&rt->nh_list, &rt->nh->f6i_list); | |
1368 | } | |
1369 | ||
1370 | spin_unlock(&rt->nh->lock); | |
1371 | ||
1372 | return err; | |
1373 | } | |
1374 | ||
8d1c802b | 1375 | static void fib6_start_gc(struct net *net, struct fib6_info *rt) |
1da177e4 | 1376 | { |
417f28bb | 1377 | if (!timer_pending(&net->ipv6.ip6_fib_timer) && |
93c2fb25 | 1378 | (rt->fib6_flags & RTF_EXPIRES)) |
417f28bb | 1379 | mod_timer(&net->ipv6.ip6_fib_timer, |
847499ce | 1380 | jiffies + net->ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 LT |
1381 | } |
1382 | ||
63152fc0 | 1383 | void fib6_force_start_gc(struct net *net) |
1da177e4 | 1384 | { |
417f28bb SH |
1385 | if (!timer_pending(&net->ipv6.ip6_fib_timer)) |
1386 | mod_timer(&net->ipv6.ip6_fib_timer, | |
847499ce | 1387 | jiffies + net->ipv6.sysctl.ip6_rt_gc_interval); |
1da177e4 LT |
1388 | } |
1389 | ||
8d1c802b | 1390 | static void __fib6_update_sernum_upto_root(struct fib6_info *rt, |
4a8e56ee | 1391 | int sernum) |
bbd63f06 | 1392 | { |
93c2fb25 DA |
1393 | struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node, |
1394 | lockdep_is_held(&rt->fib6_table->tb6_lock)); | |
bbd63f06 | 1395 | |
446e7f21 | 1396 | /* paired with smp_rmb() in fib6_get_cookie_safe() */ |
bbd63f06 WW |
1397 | smp_wmb(); |
1398 | while (fn) { | |
aafc2e32 | 1399 | WRITE_ONCE(fn->fn_sernum, sernum); |
66f5d6ce | 1400 | fn = rcu_dereference_protected(fn->parent, |
93c2fb25 | 1401 | lockdep_is_held(&rt->fib6_table->tb6_lock)); |
bbd63f06 WW |
1402 | } |
1403 | } | |
1404 | ||
8d1c802b | 1405 | void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt) |
4a8e56ee IS |
1406 | { |
1407 | __fib6_update_sernum_upto_root(rt, fib6_new_sernum(net)); | |
1408 | } | |
1409 | ||
cdaa16a4 DA |
1410 | /* allow ipv4 to update sernum via ipv6_stub */ |
1411 | void fib6_update_sernum_stub(struct net *net, struct fib6_info *f6i) | |
1412 | { | |
1413 | spin_lock_bh(&f6i->fib6_table->tb6_lock); | |
1414 | fib6_update_sernum_upto_root(net, f6i); | |
1415 | spin_unlock_bh(&f6i->fib6_table->tb6_lock); | |
1416 | } | |
1417 | ||
1da177e4 LT |
1418 | /* |
1419 | * Add routing information to the routing tree. | |
1420 | * <destination addr>/<source addr> | |
1421 | * with source addr info in sub-trees | |
66f5d6ce | 1422 | * Need to own table->tb6_lock |
1da177e4 LT |
1423 | */ |
1424 | ||
8d1c802b | 1425 | int fib6_add(struct fib6_node *root, struct fib6_info *rt, |
d4ead6b3 | 1426 | struct nl_info *info, struct netlink_ext_ack *extack) |
1da177e4 | 1427 | { |
93c2fb25 | 1428 | struct fib6_table *table = rt->fib6_table; |
accb46b5 | 1429 | LIST_HEAD(purge_list); |
74043489 AB |
1430 | struct fib6_node *fn; |
1431 | #ifdef CONFIG_IPV6_SUBTREES | |
1432 | struct fib6_node *pn = NULL; | |
1433 | #endif | |
1da177e4 | 1434 | int err = -ENOMEM; |
4a287eba MV |
1435 | int allow_create = 1; |
1436 | int replace_required = 0; | |
507c9b1e DM |
1437 | |
1438 | if (info->nlh) { | |
1439 | if (!(info->nlh->nlmsg_flags & NLM_F_CREATE)) | |
4a287eba | 1440 | allow_create = 0; |
507c9b1e | 1441 | if (info->nlh->nlmsg_flags & NLM_F_REPLACE) |
4a287eba MV |
1442 | replace_required = 1; |
1443 | } | |
1444 | if (!allow_create && !replace_required) | |
f3213831 | 1445 | pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n"); |
1da177e4 | 1446 | |
81eb8447 | 1447 | fn = fib6_add_1(info->nl_net, table, root, |
93c2fb25 DA |
1448 | &rt->fib6_dst.addr, rt->fib6_dst.plen, |
1449 | offsetof(struct fib6_info, fib6_dst), allow_create, | |
bbd63f06 | 1450 | replace_required, extack); |
4a287eba MV |
1451 | if (IS_ERR(fn)) { |
1452 | err = PTR_ERR(fn); | |
ae7b4e1f | 1453 | fn = NULL; |
1da177e4 | 1454 | goto out; |
188c517a | 1455 | } |
1da177e4 | 1456 | |
74043489 | 1457 | #ifdef CONFIG_IPV6_SUBTREES |
66729e18 YH |
1458 | pn = fn; |
1459 | ||
93c2fb25 | 1460 | if (rt->fib6_src.plen) { |
1da177e4 LT |
1461 | struct fib6_node *sn; |
1462 | ||
66f5d6ce | 1463 | if (!rcu_access_pointer(fn->subtree)) { |
1da177e4 LT |
1464 | struct fib6_node *sfn; |
1465 | ||
1466 | /* | |
1467 | * Create subtree. | |
1468 | * | |
1469 | * fn[main tree] | |
1470 | * | | |
1471 | * sfn[subtree root] | |
1472 | * \ | |
1473 | * sn[new leaf node] | |
1474 | */ | |
1475 | ||
1476 | /* Create subtree root node */ | |
81eb8447 | 1477 | sfn = node_alloc(info->nl_net); |
507c9b1e | 1478 | if (!sfn) |
348a4002 | 1479 | goto failure; |
1da177e4 | 1480 | |
5ea71528 | 1481 | fib6_info_hold(info->nl_net->ipv6.fib6_null_entry); |
66f5d6ce | 1482 | rcu_assign_pointer(sfn->leaf, |
421842ed | 1483 | info->nl_net->ipv6.fib6_null_entry); |
1da177e4 | 1484 | sfn->fn_flags = RTN_ROOT; |
1da177e4 LT |
1485 | |
1486 | /* Now add the first leaf node to new subtree */ | |
1487 | ||
81eb8447 | 1488 | sn = fib6_add_1(info->nl_net, table, sfn, |
93c2fb25 DA |
1489 | &rt->fib6_src.addr, rt->fib6_src.plen, |
1490 | offsetof(struct fib6_info, fib6_src), | |
bbd63f06 | 1491 | allow_create, replace_required, extack); |
1da177e4 | 1492 | |
f950c0ec | 1493 | if (IS_ERR(sn)) { |
1da177e4 | 1494 | /* If it is failed, discard just allocated |
348a4002 | 1495 | root, and then (in failure) stale node |
1da177e4 LT |
1496 | in main tree. |
1497 | */ | |
81eb8447 | 1498 | node_free_immediate(info->nl_net, sfn); |
188c517a | 1499 | err = PTR_ERR(sn); |
348a4002 | 1500 | goto failure; |
1da177e4 LT |
1501 | } |
1502 | ||
1503 | /* Now link new subtree to main tree */ | |
66f5d6ce WW |
1504 | rcu_assign_pointer(sfn->parent, fn); |
1505 | rcu_assign_pointer(fn->subtree, sfn); | |
1da177e4 | 1506 | } else { |
81eb8447 | 1507 | sn = fib6_add_1(info->nl_net, table, FIB6_SUBTREE(fn), |
93c2fb25 DA |
1508 | &rt->fib6_src.addr, rt->fib6_src.plen, |
1509 | offsetof(struct fib6_info, fib6_src), | |
bbd63f06 | 1510 | allow_create, replace_required, extack); |
1da177e4 | 1511 | |
4a287eba MV |
1512 | if (IS_ERR(sn)) { |
1513 | err = PTR_ERR(sn); | |
348a4002 | 1514 | goto failure; |
188c517a | 1515 | } |
1da177e4 LT |
1516 | } |
1517 | ||
66f5d6ce | 1518 | if (!rcu_access_pointer(fn->leaf)) { |
591ff9ea WW |
1519 | if (fn->fn_flags & RTN_TL_ROOT) { |
1520 | /* put back null_entry for root node */ | |
1521 | rcu_assign_pointer(fn->leaf, | |
421842ed | 1522 | info->nl_net->ipv6.fib6_null_entry); |
591ff9ea | 1523 | } else { |
5ea71528 | 1524 | fib6_info_hold(rt); |
591ff9ea WW |
1525 | rcu_assign_pointer(fn->leaf, rt); |
1526 | } | |
66729e18 | 1527 | } |
1da177e4 LT |
1528 | fn = sn; |
1529 | } | |
1530 | #endif | |
1531 | ||
081efd18 KI |
1532 | if (rt->nh) |
1533 | err = fib6_add_rt2node_nh(fn, rt, info, extack, &purge_list); | |
1534 | else | |
1535 | err = fib6_add_rt2node(fn, rt, info, extack, &purge_list); | |
bbd63f06 | 1536 | if (!err) { |
accb46b5 KI |
1537 | struct fib6_info *iter, *next; |
1538 | ||
1539 | list_for_each_entry_safe(iter, next, &purge_list, purge_link) { | |
1540 | list_del(&iter->purge_link); | |
1541 | fib6_purge_rt(iter, fn, info->nl_net); | |
1542 | fib6_info_release(iter); | |
1543 | } | |
1544 | ||
e87b5052 | 1545 | __fib6_update_sernum_upto_root(rt, fib6_new_sernum(info->nl_net)); |
5eb902b8 KFL |
1546 | |
1547 | if (rt->fib6_flags & RTF_EXPIRES) | |
1548 | fib6_add_gc_list(rt); | |
1549 | ||
63152fc0 | 1550 | fib6_start_gc(info->nl_net, rt); |
bbd63f06 | 1551 | } |
1da177e4 LT |
1552 | |
1553 | out: | |
66729e18 YH |
1554 | if (err) { |
1555 | #ifdef CONFIG_IPV6_SUBTREES | |
1556 | /* | |
1557 | * If fib6_add_1 has cleared the old leaf pointer in the | |
1558 | * super-tree leaf node we have to find a new one for it. | |
1559 | */ | |
7bbfe00e | 1560 | if (pn != fn) { |
8d1c802b | 1561 | struct fib6_info *pn_leaf = |
7bbfe00e WW |
1562 | rcu_dereference_protected(pn->leaf, |
1563 | lockdep_is_held(&table->tb6_lock)); | |
1564 | if (pn_leaf == rt) { | |
1565 | pn_leaf = NULL; | |
1566 | RCU_INIT_POINTER(pn->leaf, NULL); | |
93531c67 | 1567 | fib6_info_release(rt); |
66729e18 | 1568 | } |
7bbfe00e WW |
1569 | if (!pn_leaf && !(pn->fn_flags & RTN_RTINFO)) { |
1570 | pn_leaf = fib6_find_prefix(info->nl_net, table, | |
1571 | pn); | |
75475bb5 | 1572 | if (!pn_leaf) |
7bbfe00e | 1573 | pn_leaf = |
421842ed | 1574 | info->nl_net->ipv6.fib6_null_entry; |
93531c67 | 1575 | fib6_info_hold(pn_leaf); |
7bbfe00e WW |
1576 | rcu_assign_pointer(pn->leaf, pn_leaf); |
1577 | } | |
66729e18 YH |
1578 | } |
1579 | #endif | |
348a4002 | 1580 | goto failure; |
b9b33e7c PA |
1581 | } else if (fib6_requires_src(rt)) { |
1582 | fib6_routes_require_src_inc(info->nl_net); | |
66729e18 | 1583 | } |
1da177e4 LT |
1584 | return err; |
1585 | ||
348a4002 | 1586 | failure: |
4512c43e WW |
1587 | /* fn->leaf could be NULL and fib6_repair_tree() needs to be called if: |
1588 | * 1. fn is an intermediate node and we failed to add the new | |
1589 | * route to it in both subtree creation failure and fib6_add_rt2node() | |
1590 | * failure case. | |
1591 | * 2. fn is the root node in the table and we fail to add the first | |
1592 | * default route to it. | |
1da177e4 | 1593 | */ |
4512c43e WW |
1594 | if (fn && |
1595 | (!(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)) || | |
1596 | (fn->fn_flags & RTN_TL_ROOT && | |
1597 | !rcu_access_pointer(fn->leaf)))) | |
66f5d6ce | 1598 | fib6_repair_tree(info->nl_net, table, fn); |
1da177e4 | 1599 | return err; |
1da177e4 LT |
1600 | } |
1601 | ||
1602 | /* | |
1603 | * Routing tree lookup | |
1604 | * | |
1605 | */ | |
1606 | ||
1607 | struct lookup_args { | |
8d1c802b | 1608 | int offset; /* key offset on fib6_info */ |
b71d1d42 | 1609 | const struct in6_addr *addr; /* search key */ |
1da177e4 LT |
1610 | }; |
1611 | ||
6454743b DA |
1612 | static struct fib6_node *fib6_node_lookup_1(struct fib6_node *root, |
1613 | struct lookup_args *args) | |
1da177e4 LT |
1614 | { |
1615 | struct fib6_node *fn; | |
e69a4adc | 1616 | __be32 dir; |
1da177e4 | 1617 | |
825e288e YH |
1618 | if (unlikely(args->offset == 0)) |
1619 | return NULL; | |
1620 | ||
1da177e4 LT |
1621 | /* |
1622 | * Descend on a tree | |
1623 | */ | |
1624 | ||
1625 | fn = root; | |
1626 | ||
1627 | for (;;) { | |
1628 | struct fib6_node *next; | |
1629 | ||
1630 | dir = addr_bit_set(args->addr, fn->fn_bit); | |
1631 | ||
66f5d6ce WW |
1632 | next = dir ? rcu_dereference(fn->right) : |
1633 | rcu_dereference(fn->left); | |
1da177e4 LT |
1634 | |
1635 | if (next) { | |
1636 | fn = next; | |
1637 | continue; | |
1638 | } | |
1da177e4 LT |
1639 | break; |
1640 | } | |
1641 | ||
507c9b1e | 1642 | while (fn) { |
66f5d6ce WW |
1643 | struct fib6_node *subtree = FIB6_SUBTREE(fn); |
1644 | ||
1645 | if (subtree || fn->fn_flags & RTN_RTINFO) { | |
8d1c802b | 1646 | struct fib6_info *leaf = rcu_dereference(fn->leaf); |
1da177e4 LT |
1647 | struct rt6key *key; |
1648 | ||
8d1040e8 WW |
1649 | if (!leaf) |
1650 | goto backtrack; | |
1651 | ||
1652 | key = (struct rt6key *) ((u8 *)leaf + args->offset); | |
1da177e4 | 1653 | |
3fc5e044 YH |
1654 | if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) { |
1655 | #ifdef CONFIG_IPV6_SUBTREES | |
66f5d6ce | 1656 | if (subtree) { |
3e3be275 | 1657 | struct fib6_node *sfn; |
6454743b DA |
1658 | sfn = fib6_node_lookup_1(subtree, |
1659 | args + 1); | |
3e3be275 HFS |
1660 | if (!sfn) |
1661 | goto backtrack; | |
1662 | fn = sfn; | |
1663 | } | |
3fc5e044 | 1664 | #endif |
3e3be275 | 1665 | if (fn->fn_flags & RTN_RTINFO) |
3fc5e044 YH |
1666 | return fn; |
1667 | } | |
1da177e4 | 1668 | } |
3e3be275 | 1669 | backtrack: |
3fc5e044 YH |
1670 | if (fn->fn_flags & RTN_ROOT) |
1671 | break; | |
1672 | ||
66f5d6ce | 1673 | fn = rcu_dereference(fn->parent); |
1da177e4 LT |
1674 | } |
1675 | ||
1676 | return NULL; | |
1677 | } | |
1678 | ||
66f5d6ce WW |
1679 | /* called with rcu_read_lock() held |
1680 | */ | |
6454743b DA |
1681 | struct fib6_node *fib6_node_lookup(struct fib6_node *root, |
1682 | const struct in6_addr *daddr, | |
1683 | const struct in6_addr *saddr) | |
1da177e4 | 1684 | { |
1da177e4 | 1685 | struct fib6_node *fn; |
825e288e YH |
1686 | struct lookup_args args[] = { |
1687 | { | |
93c2fb25 | 1688 | .offset = offsetof(struct fib6_info, fib6_dst), |
825e288e YH |
1689 | .addr = daddr, |
1690 | }, | |
1da177e4 | 1691 | #ifdef CONFIG_IPV6_SUBTREES |
825e288e | 1692 | { |
93c2fb25 | 1693 | .offset = offsetof(struct fib6_info, fib6_src), |
825e288e YH |
1694 | .addr = saddr, |
1695 | }, | |
1da177e4 | 1696 | #endif |
825e288e YH |
1697 | { |
1698 | .offset = 0, /* sentinel */ | |
1699 | } | |
1700 | }; | |
1da177e4 | 1701 | |
6454743b | 1702 | fn = fib6_node_lookup_1(root, daddr ? args : args + 1); |
507c9b1e | 1703 | if (!fn || fn->fn_flags & RTN_TL_ROOT) |
1da177e4 LT |
1704 | fn = root; |
1705 | ||
1706 | return fn; | |
1707 | } | |
1708 | ||
1709 | /* | |
1710 | * Get node with specified destination prefix (and source prefix, | |
1711 | * if subtrees are used) | |
38fbeeee WW |
1712 | * exact_match == true means we try to find fn with exact match of |
1713 | * the passed in prefix addr | |
1714 | * exact_match == false means we try to find fn with longest prefix | |
1715 | * match of the passed in prefix addr. This is useful for finding fn | |
1716 | * for cached route as it will be stored in the exception table under | |
1717 | * the node with longest prefix length. | |
1da177e4 LT |
1718 | */ |
1719 | ||
1720 | ||
437de07c WY |
1721 | static struct fib6_node *fib6_locate_1(struct fib6_node *root, |
1722 | const struct in6_addr *addr, | |
38fbeeee WW |
1723 | int plen, int offset, |
1724 | bool exact_match) | |
1da177e4 | 1725 | { |
38fbeeee | 1726 | struct fib6_node *fn, *prev = NULL; |
1da177e4 LT |
1727 | |
1728 | for (fn = root; fn ; ) { | |
8d1c802b | 1729 | struct fib6_info *leaf = rcu_dereference(fn->leaf); |
8d1040e8 WW |
1730 | struct rt6key *key; |
1731 | ||
1732 | /* This node is being deleted */ | |
1733 | if (!leaf) { | |
1734 | if (plen <= fn->fn_bit) | |
1735 | goto out; | |
1736 | else | |
1737 | goto next; | |
1738 | } | |
1739 | ||
1740 | key = (struct rt6key *)((u8 *)leaf + offset); | |
1da177e4 LT |
1741 | |
1742 | /* | |
1743 | * Prefix match | |
1744 | */ | |
1745 | if (plen < fn->fn_bit || | |
1746 | !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) | |
38fbeeee | 1747 | goto out; |
1da177e4 LT |
1748 | |
1749 | if (plen == fn->fn_bit) | |
1750 | return fn; | |
1751 | ||
40cb35d5 SB |
1752 | if (fn->fn_flags & RTN_RTINFO) |
1753 | prev = fn; | |
38fbeeee | 1754 | |
8d1040e8 | 1755 | next: |
1da177e4 LT |
1756 | /* |
1757 | * We have more bits to go | |
1758 | */ | |
1759 | if (addr_bit_set(addr, fn->fn_bit)) | |
66f5d6ce | 1760 | fn = rcu_dereference(fn->right); |
1da177e4 | 1761 | else |
66f5d6ce | 1762 | fn = rcu_dereference(fn->left); |
1da177e4 | 1763 | } |
38fbeeee WW |
1764 | out: |
1765 | if (exact_match) | |
1766 | return NULL; | |
1767 | else | |
1768 | return prev; | |
1da177e4 LT |
1769 | } |
1770 | ||
437de07c WY |
1771 | struct fib6_node *fib6_locate(struct fib6_node *root, |
1772 | const struct in6_addr *daddr, int dst_len, | |
38fbeeee WW |
1773 | const struct in6_addr *saddr, int src_len, |
1774 | bool exact_match) | |
1da177e4 LT |
1775 | { |
1776 | struct fib6_node *fn; | |
1777 | ||
1778 | fn = fib6_locate_1(root, daddr, dst_len, | |
93c2fb25 | 1779 | offsetof(struct fib6_info, fib6_dst), |
38fbeeee | 1780 | exact_match); |
1da177e4 LT |
1781 | |
1782 | #ifdef CONFIG_IPV6_SUBTREES | |
1783 | if (src_len) { | |
547b792c | 1784 | WARN_ON(saddr == NULL); |
0e80193b WW |
1785 | if (fn) { |
1786 | struct fib6_node *subtree = FIB6_SUBTREE(fn); | |
1787 | ||
1788 | if (subtree) { | |
1789 | fn = fib6_locate_1(subtree, saddr, src_len, | |
93c2fb25 | 1790 | offsetof(struct fib6_info, fib6_src), |
38fbeeee | 1791 | exact_match); |
0e80193b WW |
1792 | } |
1793 | } | |
1da177e4 LT |
1794 | } |
1795 | #endif | |
1796 | ||
507c9b1e | 1797 | if (fn && fn->fn_flags & RTN_RTINFO) |
1da177e4 LT |
1798 | return fn; |
1799 | ||
1800 | return NULL; | |
1801 | } | |
1802 | ||
1803 | ||
1804 | /* | |
1805 | * Deletion | |
1806 | * | |
1807 | */ | |
1808 | ||
8d1c802b | 1809 | static struct fib6_info *fib6_find_prefix(struct net *net, |
66f5d6ce WW |
1810 | struct fib6_table *table, |
1811 | struct fib6_node *fn) | |
1da177e4 | 1812 | { |
66f5d6ce WW |
1813 | struct fib6_node *child_left, *child_right; |
1814 | ||
507c9b1e | 1815 | if (fn->fn_flags & RTN_ROOT) |
421842ed | 1816 | return net->ipv6.fib6_null_entry; |
1da177e4 | 1817 | |
507c9b1e | 1818 | while (fn) { |
66f5d6ce WW |
1819 | child_left = rcu_dereference_protected(fn->left, |
1820 | lockdep_is_held(&table->tb6_lock)); | |
1821 | child_right = rcu_dereference_protected(fn->right, | |
1822 | lockdep_is_held(&table->tb6_lock)); | |
1823 | if (child_left) | |
1824 | return rcu_dereference_protected(child_left->leaf, | |
1825 | lockdep_is_held(&table->tb6_lock)); | |
1826 | if (child_right) | |
1827 | return rcu_dereference_protected(child_right->leaf, | |
1828 | lockdep_is_held(&table->tb6_lock)); | |
1da177e4 | 1829 | |
7fc33165 | 1830 | fn = FIB6_SUBTREE(fn); |
1da177e4 LT |
1831 | } |
1832 | return NULL; | |
1833 | } | |
1834 | ||
1835 | /* | |
1836 | * Called to trim the tree of intermediate nodes when possible. "fn" | |
1837 | * is the node we want to try and remove. | |
66f5d6ce | 1838 | * Need to own table->tb6_lock |
1da177e4 LT |
1839 | */ |
1840 | ||
8ed67789 | 1841 | static struct fib6_node *fib6_repair_tree(struct net *net, |
66f5d6ce WW |
1842 | struct fib6_table *table, |
1843 | struct fib6_node *fn) | |
1da177e4 LT |
1844 | { |
1845 | int children; | |
1846 | int nstate; | |
66f5d6ce | 1847 | struct fib6_node *child; |
94b2cfe0 | 1848 | struct fib6_walker *w; |
1da177e4 LT |
1849 | int iter = 0; |
1850 | ||
4512c43e WW |
1851 | /* Set fn->leaf to null_entry for root node. */ |
1852 | if (fn->fn_flags & RTN_TL_ROOT) { | |
421842ed | 1853 | rcu_assign_pointer(fn->leaf, net->ipv6.fib6_null_entry); |
4512c43e WW |
1854 | return fn; |
1855 | } | |
1856 | ||
1da177e4 | 1857 | for (;;) { |
66f5d6ce WW |
1858 | struct fib6_node *fn_r = rcu_dereference_protected(fn->right, |
1859 | lockdep_is_held(&table->tb6_lock)); | |
1860 | struct fib6_node *fn_l = rcu_dereference_protected(fn->left, | |
1861 | lockdep_is_held(&table->tb6_lock)); | |
1862 | struct fib6_node *pn = rcu_dereference_protected(fn->parent, | |
1863 | lockdep_is_held(&table->tb6_lock)); | |
1864 | struct fib6_node *pn_r = rcu_dereference_protected(pn->right, | |
1865 | lockdep_is_held(&table->tb6_lock)); | |
1866 | struct fib6_node *pn_l = rcu_dereference_protected(pn->left, | |
1867 | lockdep_is_held(&table->tb6_lock)); | |
8d1c802b | 1868 | struct fib6_info *fn_leaf = rcu_dereference_protected(fn->leaf, |
66f5d6ce | 1869 | lockdep_is_held(&table->tb6_lock)); |
8d1c802b | 1870 | struct fib6_info *pn_leaf = rcu_dereference_protected(pn->leaf, |
66f5d6ce | 1871 | lockdep_is_held(&table->tb6_lock)); |
8d1c802b | 1872 | struct fib6_info *new_fn_leaf; |
66f5d6ce | 1873 | |
20df28fb | 1874 | pr_debug("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter); |
1da177e4 LT |
1875 | iter++; |
1876 | ||
547b792c IJ |
1877 | WARN_ON(fn->fn_flags & RTN_RTINFO); |
1878 | WARN_ON(fn->fn_flags & RTN_TL_ROOT); | |
66f5d6ce | 1879 | WARN_ON(fn_leaf); |
1da177e4 LT |
1880 | |
1881 | children = 0; | |
1882 | child = NULL; | |
dee84779 JP |
1883 | if (fn_r) { |
1884 | child = fn_r; | |
1885 | children |= 1; | |
1886 | } | |
1887 | if (fn_l) { | |
1888 | child = fn_l; | |
1889 | children |= 2; | |
1890 | } | |
1da177e4 | 1891 | |
7fc33165 | 1892 | if (children == 3 || FIB6_SUBTREE(fn) |
1da177e4 LT |
1893 | #ifdef CONFIG_IPV6_SUBTREES |
1894 | /* Subtree root (i.e. fn) may have one child */ | |
507c9b1e | 1895 | || (children && fn->fn_flags & RTN_ROOT) |
1da177e4 LT |
1896 | #endif |
1897 | ) { | |
66f5d6ce | 1898 | new_fn_leaf = fib6_find_prefix(net, table, fn); |
1da177e4 | 1899 | #if RT6_DEBUG >= 2 |
66f5d6ce WW |
1900 | if (!new_fn_leaf) { |
1901 | WARN_ON(!new_fn_leaf); | |
421842ed | 1902 | new_fn_leaf = net->ipv6.fib6_null_entry; |
1da177e4 LT |
1903 | } |
1904 | #endif | |
93531c67 | 1905 | fib6_info_hold(new_fn_leaf); |
66f5d6ce WW |
1906 | rcu_assign_pointer(fn->leaf, new_fn_leaf); |
1907 | return pn; | |
1da177e4 LT |
1908 | } |
1909 | ||
1da177e4 | 1910 | #ifdef CONFIG_IPV6_SUBTREES |
7fc33165 | 1911 | if (FIB6_SUBTREE(pn) == fn) { |
547b792c | 1912 | WARN_ON(!(fn->fn_flags & RTN_ROOT)); |
66f5d6ce | 1913 | RCU_INIT_POINTER(pn->subtree, NULL); |
1da177e4 LT |
1914 | nstate = FWS_L; |
1915 | } else { | |
547b792c | 1916 | WARN_ON(fn->fn_flags & RTN_ROOT); |
1da177e4 | 1917 | #endif |
66f5d6ce WW |
1918 | if (pn_r == fn) |
1919 | rcu_assign_pointer(pn->right, child); | |
1920 | else if (pn_l == fn) | |
1921 | rcu_assign_pointer(pn->left, child); | |
1da177e4 | 1922 | #if RT6_DEBUG >= 2 |
547b792c IJ |
1923 | else |
1924 | WARN_ON(1); | |
1da177e4 LT |
1925 | #endif |
1926 | if (child) | |
66f5d6ce | 1927 | rcu_assign_pointer(child->parent, pn); |
1da177e4 LT |
1928 | nstate = FWS_R; |
1929 | #ifdef CONFIG_IPV6_SUBTREES | |
1930 | } | |
1931 | #endif | |
1932 | ||
9a03cd8f MK |
1933 | read_lock(&net->ipv6.fib6_walker_lock); |
1934 | FOR_WALKERS(net, w) { | |
507c9b1e | 1935 | if (!child) { |
2b760fcf | 1936 | if (w->node == fn) { |
20df28fb BL |
1937 | pr_debug("W %p adjusted by delnode 1, s=%d/%d\n", |
1938 | w, w->state, nstate); | |
1da177e4 LT |
1939 | w->node = pn; |
1940 | w->state = nstate; | |
1941 | } | |
1942 | } else { | |
1da177e4 LT |
1943 | if (w->node == fn) { |
1944 | w->node = child; | |
1945 | if (children&2) { | |
20df28fb BL |
1946 | pr_debug("W %p adjusted by delnode 2, s=%d\n", |
1947 | w, w->state); | |
8db46f1d | 1948 | w->state = w->state >= FWS_R ? FWS_U : FWS_INIT; |
1da177e4 | 1949 | } else { |
20df28fb BL |
1950 | pr_debug("W %p adjusted by delnode 2, s=%d\n", |
1951 | w, w->state); | |
8db46f1d | 1952 | w->state = w->state >= FWS_C ? FWS_U : FWS_INIT; |
1da177e4 LT |
1953 | } |
1954 | } | |
1955 | } | |
1956 | } | |
9a03cd8f | 1957 | read_unlock(&net->ipv6.fib6_walker_lock); |
1da177e4 | 1958 | |
81eb8447 | 1959 | node_free(net, fn); |
507c9b1e | 1960 | if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn)) |
1da177e4 LT |
1961 | return pn; |
1962 | ||
66f5d6ce | 1963 | RCU_INIT_POINTER(pn->leaf, NULL); |
93531c67 | 1964 | fib6_info_release(pn_leaf); |
1da177e4 LT |
1965 | fn = pn; |
1966 | } | |
1967 | } | |
1968 | ||
66f5d6ce | 1969 | static void fib6_del_route(struct fib6_table *table, struct fib6_node *fn, |
8d1c802b | 1970 | struct fib6_info __rcu **rtp, struct nl_info *info) |
1da177e4 | 1971 | { |
d2f0c9b1 | 1972 | struct fib6_info *leaf, *replace_rt = NULL; |
94b2cfe0 | 1973 | struct fib6_walker *w; |
8d1c802b | 1974 | struct fib6_info *rt = rcu_dereference_protected(*rtp, |
66f5d6ce | 1975 | lockdep_is_held(&table->tb6_lock)); |
c572872f | 1976 | struct net *net = info->nl_net; |
d2f0c9b1 | 1977 | bool notify_del = false; |
1da177e4 | 1978 | |
d2f0c9b1 IS |
1979 | /* If the deleted route is the first in the node and it is not part of |
1980 | * a multipath route, then we need to replace it with the next route | |
1981 | * in the node, if exists. | |
1982 | */ | |
1983 | leaf = rcu_dereference_protected(fn->leaf, | |
1984 | lockdep_is_held(&table->tb6_lock)); | |
1985 | if (leaf == rt && !rt->fib6_nsiblings) { | |
1986 | if (rcu_access_pointer(rt->fib6_next)) | |
1987 | replace_rt = rcu_dereference_protected(rt->fib6_next, | |
1988 | lockdep_is_held(&table->tb6_lock)); | |
1989 | else | |
1990 | notify_del = true; | |
1991 | } | |
1992 | ||
1da177e4 | 1993 | /* Unlink it */ |
8fb11a9a | 1994 | *rtp = rt->fib6_next; |
93c2fb25 | 1995 | rt->fib6_node = NULL; |
c572872f BT |
1996 | net->ipv6.rt6_stats->fib_rt_entries--; |
1997 | net->ipv6.rt6_stats->fib_discarded_routes++; | |
1da177e4 | 1998 | |
f11e6659 | 1999 | /* Reset round-robin state, if necessary */ |
66f5d6ce | 2000 | if (rcu_access_pointer(fn->rr_ptr) == rt) |
f11e6659 DM |
2001 | fn->rr_ptr = NULL; |
2002 | ||
51ebd318 | 2003 | /* Remove this entry from other siblings */ |
93c2fb25 | 2004 | if (rt->fib6_nsiblings) { |
8d1c802b | 2005 | struct fib6_info *sibling, *next_sibling; |
51ebd318 | 2006 | |
d2f0c9b1 IS |
2007 | /* The route is deleted from a multipath route. If this |
2008 | * multipath route is the first route in the node, then we need | |
2009 | * to emit a delete notification. Otherwise, we need to skip | |
2010 | * the notification. | |
2011 | */ | |
2012 | if (rt->fib6_metric == leaf->fib6_metric && | |
2013 | rt6_qualify_for_ecmp(leaf)) | |
2014 | notify_del = true; | |
51ebd318 | 2015 | list_for_each_entry_safe(sibling, next_sibling, |
93c2fb25 DA |
2016 | &rt->fib6_siblings, fib6_siblings) |
2017 | sibling->fib6_nsiblings--; | |
2018 | rt->fib6_nsiblings = 0; | |
d9ccb18f | 2019 | list_del_rcu(&rt->fib6_siblings); |
d7dedee1 | 2020 | rt6_multipath_rebalance(next_sibling); |
51ebd318 ND |
2021 | } |
2022 | ||
1da177e4 | 2023 | /* Adjust walkers */ |
9a03cd8f MK |
2024 | read_lock(&net->ipv6.fib6_walker_lock); |
2025 | FOR_WALKERS(net, w) { | |
1da177e4 | 2026 | if (w->state == FWS_C && w->leaf == rt) { |
20df28fb | 2027 | pr_debug("walker %p adjusted by delroute\n", w); |
8fb11a9a | 2028 | w->leaf = rcu_dereference_protected(rt->fib6_next, |
66f5d6ce | 2029 | lockdep_is_held(&table->tb6_lock)); |
507c9b1e | 2030 | if (!w->leaf) |
1da177e4 LT |
2031 | w->state = FWS_U; |
2032 | } | |
2033 | } | |
9a03cd8f | 2034 | read_unlock(&net->ipv6.fib6_walker_lock); |
1da177e4 | 2035 | |
4512c43e WW |
2036 | /* If it was last route, call fib6_repair_tree() to: |
2037 | * 1. For root node, put back null_entry as how the table was created. | |
2038 | * 2. For other nodes, expunge its radix tree node. | |
2039 | */ | |
66f5d6ce | 2040 | if (!rcu_access_pointer(fn->leaf)) { |
4512c43e WW |
2041 | if (!(fn->fn_flags & RTN_TL_ROOT)) { |
2042 | fn->fn_flags &= ~RTN_RTINFO; | |
2043 | net->ipv6.rt6_stats->fib_route_nodes--; | |
2044 | } | |
66f5d6ce | 2045 | fn = fib6_repair_tree(net, table, fn); |
1da177e4 LT |
2046 | } |
2047 | ||
6e9e16e6 | 2048 | fib6_purge_rt(rt, fn, net); |
1da177e4 | 2049 | |
d2f0c9b1 IS |
2050 | if (!info->skip_notify_kernel) { |
2051 | if (notify_del) | |
caafb250 | 2052 | call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, |
d2f0c9b1 IS |
2053 | rt, NULL); |
2054 | else if (replace_rt) | |
2055 | call_fib6_entry_notifiers_replace(net, replace_rt); | |
d2f0c9b1 | 2056 | } |
16a16cd3 DA |
2057 | if (!info->skip_notify) |
2058 | inet6_rt_notify(RTM_DELROUTE, rt, info, 0); | |
d5382fef | 2059 | |
93531c67 | 2060 | fib6_info_release(rt); |
1da177e4 LT |
2061 | } |
2062 | ||
66f5d6ce | 2063 | /* Need to own table->tb6_lock */ |
8d1c802b | 2064 | int fib6_del(struct fib6_info *rt, struct nl_info *info) |
1da177e4 | 2065 | { |
8ed67789 | 2066 | struct net *net = info->nl_net; |
8d1c802b DA |
2067 | struct fib6_info __rcu **rtp; |
2068 | struct fib6_info __rcu **rtp_next; | |
843d926b ED |
2069 | struct fib6_table *table; |
2070 | struct fib6_node *fn; | |
2071 | ||
2072 | if (rt == net->ipv6.fib6_null_entry) | |
2073 | return -ENOENT; | |
1da177e4 | 2074 | |
843d926b ED |
2075 | table = rt->fib6_table; |
2076 | fn = rcu_dereference_protected(rt->fib6_node, | |
2077 | lockdep_is_held(&table->tb6_lock)); | |
2078 | if (!fn) | |
1da177e4 LT |
2079 | return -ENOENT; |
2080 | ||
547b792c | 2081 | WARN_ON(!(fn->fn_flags & RTN_RTINFO)); |
1da177e4 | 2082 | |
1da177e4 LT |
2083 | /* |
2084 | * Walk the leaf entries looking for ourself | |
2085 | */ | |
2086 | ||
66f5d6ce | 2087 | for (rtp = &fn->leaf; *rtp; rtp = rtp_next) { |
8d1c802b | 2088 | struct fib6_info *cur = rcu_dereference_protected(*rtp, |
66f5d6ce WW |
2089 | lockdep_is_held(&table->tb6_lock)); |
2090 | if (rt == cur) { | |
b9b33e7c PA |
2091 | if (fib6_requires_src(cur)) |
2092 | fib6_routes_require_src_dec(info->nl_net); | |
66f5d6ce | 2093 | fib6_del_route(table, fn, rtp, info); |
1da177e4 LT |
2094 | return 0; |
2095 | } | |
8fb11a9a | 2096 | rtp_next = &cur->fib6_next; |
1da177e4 LT |
2097 | } |
2098 | return -ENOENT; | |
2099 | } | |
2100 | ||
2101 | /* | |
2102 | * Tree traversal function. | |
2103 | * | |
2104 | * Certainly, it is not interrupt safe. | |
2105 | * However, it is internally reenterable wrt itself and fib6_add/fib6_del. | |
2106 | * It means, that we can modify tree during walking | |
2107 | * and use this function for garbage collection, clone pruning, | |
1ab1457c | 2108 | * cleaning tree when a device goes down etc. etc. |
1da177e4 LT |
2109 | * |
2110 | * It guarantees that every node will be traversed, | |
2111 | * and that it will be traversed only once. | |
2112 | * | |
2113 | * Callback function w->func may return: | |
2114 | * 0 -> continue walking. | |
2115 | * positive value -> walking is suspended (used by tree dumps, | |
2116 | * and probably by gc, if it will be split to several slices) | |
2117 | * negative value -> terminate walking. | |
2118 | * | |
2119 | * The function itself returns: | |
2120 | * 0 -> walk is complete. | |
2121 | * >0 -> walk is incomplete (i.e. suspended) | |
2122 | * <0 -> walk is terminated by an error. | |
66f5d6ce WW |
2123 | * |
2124 | * This function is called with tb6_lock held. | |
1da177e4 LT |
2125 | */ |
2126 | ||
94b2cfe0 | 2127 | static int fib6_walk_continue(struct fib6_walker *w) |
1da177e4 | 2128 | { |
66f5d6ce | 2129 | struct fib6_node *fn, *pn, *left, *right; |
1da177e4 | 2130 | |
2b760fcf WW |
2131 | /* w->root should always be table->tb6_root */ |
2132 | WARN_ON_ONCE(!(w->root->fn_flags & RTN_TL_ROOT)); | |
2133 | ||
1da177e4 LT |
2134 | for (;;) { |
2135 | fn = w->node; | |
507c9b1e | 2136 | if (!fn) |
1da177e4 LT |
2137 | return 0; |
2138 | ||
1da177e4 LT |
2139 | switch (w->state) { |
2140 | #ifdef CONFIG_IPV6_SUBTREES | |
2141 | case FWS_S: | |
7fc33165 YH |
2142 | if (FIB6_SUBTREE(fn)) { |
2143 | w->node = FIB6_SUBTREE(fn); | |
1da177e4 LT |
2144 | continue; |
2145 | } | |
2146 | w->state = FWS_L; | |
a8eceea8 | 2147 | fallthrough; |
1ab1457c | 2148 | #endif |
1da177e4 | 2149 | case FWS_L: |
66f5d6ce WW |
2150 | left = rcu_dereference_protected(fn->left, 1); |
2151 | if (left) { | |
2152 | w->node = left; | |
1da177e4 LT |
2153 | w->state = FWS_INIT; |
2154 | continue; | |
2155 | } | |
2156 | w->state = FWS_R; | |
a8eceea8 | 2157 | fallthrough; |
1da177e4 | 2158 | case FWS_R: |
66f5d6ce WW |
2159 | right = rcu_dereference_protected(fn->right, 1); |
2160 | if (right) { | |
2161 | w->node = right; | |
1da177e4 LT |
2162 | w->state = FWS_INIT; |
2163 | continue; | |
2164 | } | |
2165 | w->state = FWS_C; | |
66f5d6ce | 2166 | w->leaf = rcu_dereference_protected(fn->leaf, 1); |
a8eceea8 | 2167 | fallthrough; |
1da177e4 | 2168 | case FWS_C: |
507c9b1e | 2169 | if (w->leaf && fn->fn_flags & RTN_RTINFO) { |
2bec5a36 PM |
2170 | int err; |
2171 | ||
fa809e2f ED |
2172 | if (w->skip) { |
2173 | w->skip--; | |
1c265854 | 2174 | goto skip; |
2bec5a36 PM |
2175 | } |
2176 | ||
2177 | err = w->func(w); | |
1da177e4 LT |
2178 | if (err) |
2179 | return err; | |
2bec5a36 PM |
2180 | |
2181 | w->count++; | |
1da177e4 LT |
2182 | continue; |
2183 | } | |
1c265854 | 2184 | skip: |
1da177e4 | 2185 | w->state = FWS_U; |
a8eceea8 | 2186 | fallthrough; |
1da177e4 LT |
2187 | case FWS_U: |
2188 | if (fn == w->root) | |
2189 | return 0; | |
66f5d6ce WW |
2190 | pn = rcu_dereference_protected(fn->parent, 1); |
2191 | left = rcu_dereference_protected(pn->left, 1); | |
2192 | right = rcu_dereference_protected(pn->right, 1); | |
1da177e4 LT |
2193 | w->node = pn; |
2194 | #ifdef CONFIG_IPV6_SUBTREES | |
7fc33165 | 2195 | if (FIB6_SUBTREE(pn) == fn) { |
547b792c | 2196 | WARN_ON(!(fn->fn_flags & RTN_ROOT)); |
1da177e4 LT |
2197 | w->state = FWS_L; |
2198 | continue; | |
2199 | } | |
2200 | #endif | |
66f5d6ce | 2201 | if (left == fn) { |
1da177e4 LT |
2202 | w->state = FWS_R; |
2203 | continue; | |
2204 | } | |
66f5d6ce | 2205 | if (right == fn) { |
1da177e4 | 2206 | w->state = FWS_C; |
66f5d6ce | 2207 | w->leaf = rcu_dereference_protected(w->node->leaf, 1); |
1da177e4 LT |
2208 | continue; |
2209 | } | |
2210 | #if RT6_DEBUG >= 2 | |
547b792c | 2211 | WARN_ON(1); |
1da177e4 LT |
2212 | #endif |
2213 | } | |
2214 | } | |
2215 | } | |
2216 | ||
9a03cd8f | 2217 | static int fib6_walk(struct net *net, struct fib6_walker *w) |
1da177e4 LT |
2218 | { |
2219 | int res; | |
2220 | ||
2221 | w->state = FWS_INIT; | |
2222 | w->node = w->root; | |
2223 | ||
9a03cd8f | 2224 | fib6_walker_link(net, w); |
1da177e4 LT |
2225 | res = fib6_walk_continue(w); |
2226 | if (res <= 0) | |
9a03cd8f | 2227 | fib6_walker_unlink(net, w); |
1da177e4 LT |
2228 | return res; |
2229 | } | |
2230 | ||
94b2cfe0 | 2231 | static int fib6_clean_node(struct fib6_walker *w) |
1da177e4 LT |
2232 | { |
2233 | int res; | |
8d1c802b | 2234 | struct fib6_info *rt; |
94b2cfe0 | 2235 | struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w); |
ec7d43c2 BT |
2236 | struct nl_info info = { |
2237 | .nl_net = c->net, | |
7c6bb7d2 | 2238 | .skip_notify = c->skip_notify, |
ec7d43c2 | 2239 | }; |
1da177e4 | 2240 | |
327571cb | 2241 | if (c->sernum != FIB6_NO_SERNUM_CHANGE && |
aafc2e32 ED |
2242 | READ_ONCE(w->node->fn_sernum) != c->sernum) |
2243 | WRITE_ONCE(w->node->fn_sernum, c->sernum); | |
327571cb HFS |
2244 | |
2245 | if (!c->func) { | |
2246 | WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE); | |
2247 | w->leaf = NULL; | |
2248 | return 0; | |
2249 | } | |
2250 | ||
66f5d6ce | 2251 | for_each_fib6_walker_rt(w) { |
1da177e4 | 2252 | res = c->func(rt, c->arg); |
b5cb5a75 | 2253 | if (res == -1) { |
1da177e4 | 2254 | w->leaf = rt; |
528c4ceb | 2255 | res = fib6_del(rt, &info); |
1da177e4 LT |
2256 | if (res) { |
2257 | #if RT6_DEBUG >= 2 | |
91df42be | 2258 | pr_debug("%s: del failed: rt=%p@%p err=%d\n", |
4e587ea7 | 2259 | __func__, rt, |
93c2fb25 | 2260 | rcu_access_pointer(rt->fib6_node), |
4e587ea7 | 2261 | res); |
1da177e4 LT |
2262 | #endif |
2263 | continue; | |
2264 | } | |
2265 | return 0; | |
b5cb5a75 | 2266 | } else if (res == -2) { |
93c2fb25 | 2267 | if (WARN_ON(!rt->fib6_nsiblings)) |
b5cb5a75 | 2268 | continue; |
93c2fb25 DA |
2269 | rt = list_last_entry(&rt->fib6_siblings, |
2270 | struct fib6_info, fib6_siblings); | |
b5cb5a75 | 2271 | continue; |
1da177e4 | 2272 | } |
547b792c | 2273 | WARN_ON(res != 0); |
1da177e4 LT |
2274 | } |
2275 | w->leaf = rt; | |
2276 | return 0; | |
2277 | } | |
2278 | ||
2279 | /* | |
2280 | * Convenient frontend to tree walker. | |
1ab1457c | 2281 | * |
1da177e4 | 2282 | * func is called on each route. |
b5cb5a75 IS |
2283 | * It may return -2 -> skip multipath route. |
2284 | * -1 -> delete this route. | |
1da177e4 | 2285 | * 0 -> continue walking |
1da177e4 LT |
2286 | */ |
2287 | ||
ec7d43c2 | 2288 | static void fib6_clean_tree(struct net *net, struct fib6_node *root, |
8d1c802b | 2289 | int (*func)(struct fib6_info *, void *arg), |
7c6bb7d2 | 2290 | int sernum, void *arg, bool skip_notify) |
1da177e4 | 2291 | { |
94b2cfe0 | 2292 | struct fib6_cleaner c; |
1da177e4 LT |
2293 | |
2294 | c.w.root = root; | |
2295 | c.w.func = fib6_clean_node; | |
2bec5a36 PM |
2296 | c.w.count = 0; |
2297 | c.w.skip = 0; | |
1e47b483 | 2298 | c.w.skip_in_node = 0; |
1da177e4 | 2299 | c.func = func; |
327571cb | 2300 | c.sernum = sernum; |
1da177e4 | 2301 | c.arg = arg; |
ec7d43c2 | 2302 | c.net = net; |
7c6bb7d2 | 2303 | c.skip_notify = skip_notify; |
1da177e4 | 2304 | |
9a03cd8f | 2305 | fib6_walk(net, &c.w); |
1da177e4 LT |
2306 | } |
2307 | ||
327571cb | 2308 | static void __fib6_clean_all(struct net *net, |
8d1c802b | 2309 | int (*func)(struct fib6_info *, void *), |
7c6bb7d2 | 2310 | int sernum, void *arg, bool skip_notify) |
c71099ac | 2311 | { |
c71099ac | 2312 | struct fib6_table *table; |
58f09b78 | 2313 | struct hlist_head *head; |
1b43af54 | 2314 | unsigned int h; |
c71099ac | 2315 | |
1b43af54 | 2316 | rcu_read_lock(); |
a33bc5c1 | 2317 | for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { |
f3db4851 | 2318 | head = &net->ipv6.fib_table_hash[h]; |
b67bfe0d | 2319 | hlist_for_each_entry_rcu(table, head, tb6_hlist) { |
66f5d6ce | 2320 | spin_lock_bh(&table->tb6_lock); |
ec7d43c2 | 2321 | fib6_clean_tree(net, &table->tb6_root, |
7c6bb7d2 | 2322 | func, sernum, arg, skip_notify); |
66f5d6ce | 2323 | spin_unlock_bh(&table->tb6_lock); |
c71099ac TG |
2324 | } |
2325 | } | |
1b43af54 | 2326 | rcu_read_unlock(); |
c71099ac TG |
2327 | } |
2328 | ||
8d1c802b | 2329 | void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *), |
327571cb HFS |
2330 | void *arg) |
2331 | { | |
7c6bb7d2 DA |
2332 | __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, false); |
2333 | } | |
2334 | ||
2335 | void fib6_clean_all_skip_notify(struct net *net, | |
2336 | int (*func)(struct fib6_info *, void *), | |
2337 | void *arg) | |
2338 | { | |
2339 | __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, true); | |
327571cb HFS |
2340 | } |
2341 | ||
705f1c86 HFS |
2342 | static void fib6_flush_trees(struct net *net) |
2343 | { | |
812918c4 | 2344 | int new_sernum = fib6_new_sernum(net); |
705f1c86 | 2345 | |
7c6bb7d2 | 2346 | __fib6_clean_all(net, NULL, new_sernum, NULL, false); |
705f1c86 HFS |
2347 | } |
2348 | ||
1da177e4 LT |
2349 | /* |
2350 | * Garbage collection | |
2351 | */ | |
2352 | ||
5eb902b8 | 2353 | static int fib6_age(struct fib6_info *rt, struct fib6_gc_args *gc_args) |
1da177e4 LT |
2354 | { |
2355 | unsigned long now = jiffies; | |
2356 | ||
2357 | /* | |
2358 | * check addrconf expiration here. | |
2359 | * Routes are expired even if they are in use. | |
1da177e4 LT |
2360 | */ |
2361 | ||
dade3f6a | 2362 | if (rt->fib6_flags & RTF_EXPIRES && rt->expires) { |
14895687 | 2363 | if (time_after(now, rt->expires)) { |
20df28fb | 2364 | pr_debug("expiring %p\n", rt); |
1da177e4 LT |
2365 | return -1; |
2366 | } | |
3570df91 | 2367 | gc_args->more++; |
1da177e4 LT |
2368 | } |
2369 | ||
c757faa8 WW |
2370 | /* Also age clones in the exception table. |
2371 | * Note, that clones are aged out | |
2372 | * only if they are not in use now. | |
2373 | */ | |
2374 | rt6_age_exceptions(rt, gc_args, now); | |
2375 | ||
1da177e4 LT |
2376 | return 0; |
2377 | } | |
2378 | ||
5eb902b8 KFL |
2379 | static void fib6_gc_table(struct net *net, |
2380 | struct fib6_table *tb6, | |
2381 | struct fib6_gc_args *gc_args) | |
2382 | { | |
2383 | struct fib6_info *rt; | |
2384 | struct hlist_node *n; | |
2385 | struct nl_info info = { | |
2386 | .nl_net = net, | |
2387 | .skip_notify = false, | |
2388 | }; | |
2389 | ||
2390 | hlist_for_each_entry_safe(rt, n, &tb6->tb6_gc_hlist, gc_link) | |
2391 | if (fib6_age(rt, gc_args) == -1) | |
2392 | fib6_del(rt, &info); | |
2393 | } | |
2394 | ||
2395 | static void fib6_gc_all(struct net *net, struct fib6_gc_args *gc_args) | |
2396 | { | |
2397 | struct fib6_table *table; | |
2398 | struct hlist_head *head; | |
2399 | unsigned int h; | |
2400 | ||
2401 | rcu_read_lock(); | |
2402 | for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { | |
2403 | head = &net->ipv6.fib_table_hash[h]; | |
2404 | hlist_for_each_entry_rcu(table, head, tb6_hlist) { | |
2405 | spin_lock_bh(&table->tb6_lock); | |
2406 | ||
2407 | fib6_gc_table(net, table, gc_args); | |
2408 | ||
2409 | spin_unlock_bh(&table->tb6_lock); | |
2410 | } | |
2411 | } | |
2412 | rcu_read_unlock(); | |
2413 | } | |
2414 | ||
2ac3ac8f | 2415 | void fib6_run_gc(unsigned long expires, struct net *net, bool force) |
1da177e4 | 2416 | { |
3570df91 | 2417 | struct fib6_gc_args gc_args; |
49a18d86 MK |
2418 | unsigned long now; |
2419 | ||
2ac3ac8f | 2420 | if (force) { |
3dc94f93 MK |
2421 | spin_lock_bh(&net->ipv6.fib6_gc_lock); |
2422 | } else if (!spin_trylock_bh(&net->ipv6.fib6_gc_lock)) { | |
2ac3ac8f MK |
2423 | mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ); |
2424 | return; | |
1da177e4 | 2425 | } |
2ac3ac8f MK |
2426 | gc_args.timeout = expires ? (int)expires : |
2427 | net->ipv6.sysctl.ip6_rt_gc_interval; | |
db916649 | 2428 | gc_args.more = 0; |
f3db4851 | 2429 | |
5eb902b8 | 2430 | fib6_gc_all(net, &gc_args); |
49a18d86 MK |
2431 | now = jiffies; |
2432 | net->ipv6.ip6_rt_last_gc = now; | |
1da177e4 LT |
2433 | |
2434 | if (gc_args.more) | |
c8a45222 | 2435 | mod_timer(&net->ipv6.ip6_fib_timer, |
49a18d86 | 2436 | round_jiffies(now |
c8a45222 | 2437 | + net->ipv6.sysctl.ip6_rt_gc_interval)); |
417f28bb | 2438 | else |
8fa7292f | 2439 | timer_delete(&net->ipv6.ip6_fib_timer); |
3dc94f93 | 2440 | spin_unlock_bh(&net->ipv6.fib6_gc_lock); |
1da177e4 LT |
2441 | } |
2442 | ||
86cb30ec | 2443 | static void fib6_gc_timer_cb(struct timer_list *t) |
5b7c931d | 2444 | { |
41cb0855 | 2445 | struct net *arg = timer_container_of(arg, t, ipv6.ip6_fib_timer); |
86cb30ec KC |
2446 | |
2447 | fib6_run_gc(0, arg, true); | |
5b7c931d DL |
2448 | } |
2449 | ||
2c8c1e72 | 2450 | static int __net_init fib6_net_init(struct net *net) |
1da177e4 | 2451 | { |
10da66f7 | 2452 | size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ; |
16ab6d7d IS |
2453 | int err; |
2454 | ||
2455 | err = fib6_notifier_init(net); | |
2456 | if (err) | |
2457 | return err; | |
10da66f7 | 2458 | |
ed13923f IS |
2459 | /* Default to 3-tuple */ |
2460 | net->ipv6.sysctl.multipath_hash_fields = | |
2461 | FIB_MULTIPATH_HASH_FIELD_DEFAULT_MASK; | |
2462 | ||
3dc94f93 | 2463 | spin_lock_init(&net->ipv6.fib6_gc_lock); |
9a03cd8f MK |
2464 | rwlock_init(&net->ipv6.fib6_walker_lock); |
2465 | INIT_LIST_HEAD(&net->ipv6.fib6_walkers); | |
86cb30ec | 2466 | timer_setup(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, 0); |
63152fc0 | 2467 | |
c572872f BT |
2468 | net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL); |
2469 | if (!net->ipv6.rt6_stats) | |
67db5ca7 | 2470 | goto out_notifier; |
c572872f | 2471 | |
10da66f7 ED |
2472 | /* Avoid false sharing : Use at least a full cache line */ |
2473 | size = max_t(size_t, size, L1_CACHE_BYTES); | |
2474 | ||
2475 | net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL); | |
58f09b78 | 2476 | if (!net->ipv6.fib_table_hash) |
c572872f | 2477 | goto out_rt6_stats; |
e0b85590 | 2478 | |
834d9784 KI |
2479 | spin_lock_init(&net->ipv6.fib_table_hash_lock); |
2480 | ||
58f09b78 DL |
2481 | net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl), |
2482 | GFP_KERNEL); | |
2483 | if (!net->ipv6.fib6_main_tbl) | |
e0b85590 DL |
2484 | goto out_fib_table_hash; |
2485 | ||
58f09b78 | 2486 | net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN; |
66f5d6ce | 2487 | rcu_assign_pointer(net->ipv6.fib6_main_tbl->tb6_root.leaf, |
421842ed | 2488 | net->ipv6.fib6_null_entry); |
58f09b78 DL |
2489 | net->ipv6.fib6_main_tbl->tb6_root.fn_flags = |
2490 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
8e773277 | 2491 | inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers); |
5eb902b8 | 2492 | INIT_HLIST_HEAD(&net->ipv6.fib6_main_tbl->tb6_gc_hlist); |
e0b85590 DL |
2493 | |
2494 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES | |
58f09b78 DL |
2495 | net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl), |
2496 | GFP_KERNEL); | |
2497 | if (!net->ipv6.fib6_local_tbl) | |
e0b85590 | 2498 | goto out_fib6_main_tbl; |
58f09b78 | 2499 | net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL; |
66f5d6ce | 2500 | rcu_assign_pointer(net->ipv6.fib6_local_tbl->tb6_root.leaf, |
421842ed | 2501 | net->ipv6.fib6_null_entry); |
58f09b78 DL |
2502 | net->ipv6.fib6_local_tbl->tb6_root.fn_flags = |
2503 | RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; | |
8e773277 | 2504 | inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers); |
5eb902b8 | 2505 | INIT_HLIST_HEAD(&net->ipv6.fib6_local_tbl->tb6_gc_hlist); |
e0b85590 | 2506 | #endif |
58f09b78 | 2507 | fib6_tables_init(net); |
f845ab6b | 2508 | |
417f28bb | 2509 | return 0; |
d63bddbe | 2510 | |
e0b85590 | 2511 | #ifdef CONFIG_IPV6_MULTIPLE_TABLES |
e0b85590 | 2512 | out_fib6_main_tbl: |
58f09b78 | 2513 | kfree(net->ipv6.fib6_main_tbl); |
e0b85590 | 2514 | #endif |
e0b85590 | 2515 | out_fib_table_hash: |
58f09b78 | 2516 | kfree(net->ipv6.fib_table_hash); |
c572872f BT |
2517 | out_rt6_stats: |
2518 | kfree(net->ipv6.rt6_stats); | |
67db5ca7 | 2519 | out_notifier: |
16ab6d7d | 2520 | fib6_notifier_exit(net); |
417f28bb | 2521 | return -ENOMEM; |
8db46f1d | 2522 | } |
58f09b78 DL |
2523 | |
2524 | static void fib6_net_exit(struct net *net) | |
2525 | { | |
ba1cc08d SD |
2526 | unsigned int i; |
2527 | ||
8fa7292f | 2528 | timer_delete_sync(&net->ipv6.ip6_fib_timer); |
417f28bb | 2529 | |
32a805ba | 2530 | for (i = 0; i < FIB6_TABLE_HASHSZ; i++) { |
ba1cc08d SD |
2531 | struct hlist_head *head = &net->ipv6.fib_table_hash[i]; |
2532 | struct hlist_node *tmp; | |
2533 | struct fib6_table *tb; | |
2534 | ||
2535 | hlist_for_each_entry_safe(tb, tmp, head, tb6_hlist) { | |
2536 | hlist_del(&tb->tb6_hlist); | |
2537 | fib6_free_table(tb); | |
2538 | } | |
2539 | } | |
2540 | ||
58f09b78 | 2541 | kfree(net->ipv6.fib_table_hash); |
c572872f | 2542 | kfree(net->ipv6.rt6_stats); |
16ab6d7d | 2543 | fib6_notifier_exit(net); |
58f09b78 DL |
2544 | } |
2545 | ||
2546 | static struct pernet_operations fib6_net_ops = { | |
2547 | .init = fib6_net_init, | |
2548 | .exit = fib6_net_exit, | |
2549 | }; | |
2550 | ||
a37b0e4e KI |
2551 | static const struct rtnl_msg_handler fib6_rtnl_msg_handlers[] __initconst_or_module = { |
2552 | {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETROUTE, | |
2553 | .dumpit = inet6_dump_fib, | |
2554 | .flags = RTNL_FLAG_DUMP_UNLOCKED | RTNL_FLAG_DUMP_SPLIT_NLM_DONE}, | |
2555 | }; | |
2556 | ||
58f09b78 DL |
2557 | int __init fib6_init(void) |
2558 | { | |
2559 | int ret = -ENOMEM; | |
63152fc0 | 2560 | |
072f88ca KC |
2561 | fib6_node_kmem = KMEM_CACHE(fib6_node, |
2562 | SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT); | |
58f09b78 DL |
2563 | if (!fib6_node_kmem) |
2564 | goto out; | |
2565 | ||
2566 | ret = register_pernet_subsys(&fib6_net_ops); | |
2567 | if (ret) | |
c572872f | 2568 | goto out_kmem_cache_create; |
e8803b6c | 2569 | |
a37b0e4e | 2570 | ret = rtnl_register_many(fib6_rtnl_msg_handlers); |
e8803b6c DM |
2571 | if (ret) |
2572 | goto out_unregister_subsys; | |
705f1c86 HFS |
2573 | |
2574 | __fib6_flush_trees = fib6_flush_trees; | |
58f09b78 DL |
2575 | out: |
2576 | return ret; | |
2577 | ||
e8803b6c DM |
2578 | out_unregister_subsys: |
2579 | unregister_pernet_subsys(&fib6_net_ops); | |
d63bddbe DL |
2580 | out_kmem_cache_create: |
2581 | kmem_cache_destroy(fib6_node_kmem); | |
2582 | goto out; | |
1da177e4 LT |
2583 | } |
2584 | ||
2585 | void fib6_gc_cleanup(void) | |
2586 | { | |
58f09b78 | 2587 | unregister_pernet_subsys(&fib6_net_ops); |
1da177e4 LT |
2588 | kmem_cache_destroy(fib6_node_kmem); |
2589 | } | |
8d2ca1d7 HFS |
2590 | |
2591 | #ifdef CONFIG_PROC_FS | |
138d0be3 | 2592 | static int ipv6_route_native_seq_show(struct seq_file *seq, void *v) |
8d2ca1d7 | 2593 | { |
8d1c802b | 2594 | struct fib6_info *rt = v; |
8d2ca1d7 | 2595 | struct ipv6_route_iter *iter = seq->private; |
f88d8ea6 | 2596 | struct fib6_nh *fib6_nh = rt->fib6_nh; |
2b2450ca | 2597 | unsigned int flags = rt->fib6_flags; |
5e670d84 | 2598 | const struct net_device *dev; |
8d2ca1d7 | 2599 | |
f88d8ea6 | 2600 | if (rt->nh) |
ef1148d4 | 2601 | fib6_nh = nexthop_fib6_nh(rt->nh); |
f88d8ea6 | 2602 | |
93c2fb25 | 2603 | seq_printf(seq, "%pi6 %02x ", &rt->fib6_dst.addr, rt->fib6_dst.plen); |
8d2ca1d7 HFS |
2604 | |
2605 | #ifdef CONFIG_IPV6_SUBTREES | |
93c2fb25 | 2606 | seq_printf(seq, "%pi6 %02x ", &rt->fib6_src.addr, rt->fib6_src.plen); |
8d2ca1d7 HFS |
2607 | #else |
2608 | seq_puts(seq, "00000000000000000000000000000000 00 "); | |
2609 | #endif | |
f88d8ea6 | 2610 | if (fib6_nh->fib_nh_gw_family) { |
2b2450ca | 2611 | flags |= RTF_GATEWAY; |
f88d8ea6 | 2612 | seq_printf(seq, "%pi6", &fib6_nh->fib_nh_gw6); |
2b2450ca | 2613 | } else { |
8d2ca1d7 | 2614 | seq_puts(seq, "00000000000000000000000000000000"); |
2b2450ca | 2615 | } |
8d2ca1d7 | 2616 | |
f88d8ea6 | 2617 | dev = fib6_nh->fib_nh_dev; |
8d2ca1d7 | 2618 | seq_printf(seq, " %08x %08x %08x %08x %8s\n", |
f05713e0 | 2619 | rt->fib6_metric, refcount_read(&rt->fib6_ref), 0, |
2b2450ca | 2620 | flags, dev ? dev->name : ""); |
8d2ca1d7 HFS |
2621 | iter->w.leaf = NULL; |
2622 | return 0; | |
2623 | } | |
2624 | ||
94b2cfe0 | 2625 | static int ipv6_route_yield(struct fib6_walker *w) |
8d2ca1d7 HFS |
2626 | { |
2627 | struct ipv6_route_iter *iter = w->args; | |
2628 | ||
2629 | if (!iter->skip) | |
2630 | return 1; | |
2631 | ||
2632 | do { | |
66f5d6ce | 2633 | iter->w.leaf = rcu_dereference_protected( |
8fb11a9a | 2634 | iter->w.leaf->fib6_next, |
66f5d6ce | 2635 | lockdep_is_held(&iter->tbl->tb6_lock)); |
8d2ca1d7 HFS |
2636 | iter->skip--; |
2637 | if (!iter->skip && iter->w.leaf) | |
2638 | return 1; | |
2639 | } while (iter->w.leaf); | |
2640 | ||
2641 | return 0; | |
2642 | } | |
2643 | ||
9a03cd8f MK |
2644 | static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter, |
2645 | struct net *net) | |
8d2ca1d7 HFS |
2646 | { |
2647 | memset(&iter->w, 0, sizeof(iter->w)); | |
2648 | iter->w.func = ipv6_route_yield; | |
2649 | iter->w.root = &iter->tbl->tb6_root; | |
2650 | iter->w.state = FWS_INIT; | |
2651 | iter->w.node = iter->w.root; | |
2652 | iter->w.args = iter; | |
aafc2e32 | 2653 | iter->sernum = READ_ONCE(iter->w.root->fn_sernum); |
8d2ca1d7 | 2654 | INIT_LIST_HEAD(&iter->w.lh); |
9a03cd8f | 2655 | fib6_walker_link(net, &iter->w); |
8d2ca1d7 HFS |
2656 | } |
2657 | ||
2658 | static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl, | |
2659 | struct net *net) | |
2660 | { | |
2661 | unsigned int h; | |
2662 | struct hlist_node *node; | |
2663 | ||
2664 | if (tbl) { | |
2665 | h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1; | |
ef1148d4 | 2666 | node = rcu_dereference(hlist_next_rcu(&tbl->tb6_hlist)); |
8d2ca1d7 HFS |
2667 | } else { |
2668 | h = 0; | |
2669 | node = NULL; | |
2670 | } | |
2671 | ||
2672 | while (!node && h < FIB6_TABLE_HASHSZ) { | |
ef1148d4 | 2673 | node = rcu_dereference( |
8d2ca1d7 HFS |
2674 | hlist_first_rcu(&net->ipv6.fib_table_hash[h++])); |
2675 | } | |
2676 | return hlist_entry_safe(node, struct fib6_table, tb6_hlist); | |
2677 | } | |
2678 | ||
0a67d3ef HFS |
2679 | static void ipv6_route_check_sernum(struct ipv6_route_iter *iter) |
2680 | { | |
aafc2e32 ED |
2681 | int sernum = READ_ONCE(iter->w.root->fn_sernum); |
2682 | ||
2683 | if (iter->sernum != sernum) { | |
2684 | iter->sernum = sernum; | |
0a67d3ef HFS |
2685 | iter->w.state = FWS_INIT; |
2686 | iter->w.node = iter->w.root; | |
2687 | WARN_ON(iter->w.skip); | |
2688 | iter->w.skip = iter->w.count; | |
2689 | } | |
2690 | } | |
2691 | ||
8d2ca1d7 HFS |
2692 | static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos) |
2693 | { | |
2694 | int r; | |
8d1c802b | 2695 | struct fib6_info *n; |
8d2ca1d7 HFS |
2696 | struct net *net = seq_file_net(seq); |
2697 | struct ipv6_route_iter *iter = seq->private; | |
2698 | ||
4fc427e0 | 2699 | ++(*pos); |
8d2ca1d7 HFS |
2700 | if (!v) |
2701 | goto iter_table; | |
2702 | ||
ef1148d4 | 2703 | n = rcu_dereference(((struct fib6_info *)v)->fib6_next); |
4fc427e0 | 2704 | if (n) |
8d2ca1d7 | 2705 | return n; |
8d2ca1d7 HFS |
2706 | |
2707 | iter_table: | |
0a67d3ef | 2708 | ipv6_route_check_sernum(iter); |
66f5d6ce | 2709 | spin_lock_bh(&iter->tbl->tb6_lock); |
8d2ca1d7 | 2710 | r = fib6_walk_continue(&iter->w); |
66f5d6ce | 2711 | spin_unlock_bh(&iter->tbl->tb6_lock); |
8d2ca1d7 | 2712 | if (r > 0) { |
8d2ca1d7 HFS |
2713 | return iter->w.leaf; |
2714 | } else if (r < 0) { | |
9a03cd8f | 2715 | fib6_walker_unlink(net, &iter->w); |
8d2ca1d7 HFS |
2716 | return NULL; |
2717 | } | |
9a03cd8f | 2718 | fib6_walker_unlink(net, &iter->w); |
8d2ca1d7 HFS |
2719 | |
2720 | iter->tbl = ipv6_route_seq_next_table(iter->tbl, net); | |
2721 | if (!iter->tbl) | |
2722 | return NULL; | |
2723 | ||
9a03cd8f | 2724 | ipv6_route_seq_setup_walk(iter, net); |
8d2ca1d7 HFS |
2725 | goto iter_table; |
2726 | } | |
2727 | ||
2728 | static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos) | |
ef1148d4 | 2729 | __acquires(RCU) |
8d2ca1d7 HFS |
2730 | { |
2731 | struct net *net = seq_file_net(seq); | |
2732 | struct ipv6_route_iter *iter = seq->private; | |
2733 | ||
ef1148d4 | 2734 | rcu_read_lock(); |
8d2ca1d7 HFS |
2735 | iter->tbl = ipv6_route_seq_next_table(NULL, net); |
2736 | iter->skip = *pos; | |
2737 | ||
2738 | if (iter->tbl) { | |
6617dfd4 YS |
2739 | loff_t p = 0; |
2740 | ||
9a03cd8f | 2741 | ipv6_route_seq_setup_walk(iter, net); |
6617dfd4 | 2742 | return ipv6_route_seq_next(seq, NULL, &p); |
8d2ca1d7 HFS |
2743 | } else { |
2744 | return NULL; | |
2745 | } | |
2746 | } | |
2747 | ||
2748 | static bool ipv6_route_iter_active(struct ipv6_route_iter *iter) | |
2749 | { | |
94b2cfe0 | 2750 | struct fib6_walker *w = &iter->w; |
8d2ca1d7 HFS |
2751 | return w->node && !(w->state == FWS_U && w->node == w->root); |
2752 | } | |
2753 | ||
138d0be3 | 2754 | static void ipv6_route_native_seq_stop(struct seq_file *seq, void *v) |
ef1148d4 | 2755 | __releases(RCU) |
8d2ca1d7 | 2756 | { |
9a03cd8f | 2757 | struct net *net = seq_file_net(seq); |
8d2ca1d7 HFS |
2758 | struct ipv6_route_iter *iter = seq->private; |
2759 | ||
2760 | if (ipv6_route_iter_active(iter)) | |
9a03cd8f | 2761 | fib6_walker_unlink(net, &iter->w); |
8d2ca1d7 | 2762 | |
ef1148d4 | 2763 | rcu_read_unlock(); |
8d2ca1d7 HFS |
2764 | } |
2765 | ||
138d0be3 | 2766 | #if IS_BUILTIN(CONFIG_IPV6) && defined(CONFIG_BPF_SYSCALL) |
138d0be3 YS |
2767 | static int ipv6_route_prog_seq_show(struct bpf_prog *prog, |
2768 | struct bpf_iter_meta *meta, | |
2769 | void *v) | |
2770 | { | |
2771 | struct bpf_iter__ipv6_route ctx; | |
2772 | ||
2773 | ctx.meta = meta; | |
2774 | ctx.rt = v; | |
2775 | return bpf_iter_run_prog(prog, &ctx); | |
2776 | } | |
2777 | ||
2778 | static int ipv6_route_seq_show(struct seq_file *seq, void *v) | |
2779 | { | |
2780 | struct ipv6_route_iter *iter = seq->private; | |
2781 | struct bpf_iter_meta meta; | |
2782 | struct bpf_prog *prog; | |
2783 | int ret; | |
2784 | ||
2785 | meta.seq = seq; | |
2786 | prog = bpf_iter_get_info(&meta, false); | |
2787 | if (!prog) | |
2788 | return ipv6_route_native_seq_show(seq, v); | |
2789 | ||
2790 | ret = ipv6_route_prog_seq_show(prog, &meta, v); | |
2791 | iter->w.leaf = NULL; | |
2792 | ||
2793 | return ret; | |
2794 | } | |
2795 | ||
2796 | static void ipv6_route_seq_stop(struct seq_file *seq, void *v) | |
2797 | { | |
2798 | struct bpf_iter_meta meta; | |
2799 | struct bpf_prog *prog; | |
2800 | ||
2801 | if (!v) { | |
2802 | meta.seq = seq; | |
2803 | prog = bpf_iter_get_info(&meta, true); | |
2804 | if (prog) | |
2805 | (void)ipv6_route_prog_seq_show(prog, &meta, v); | |
2806 | } | |
2807 | ||
2808 | ipv6_route_native_seq_stop(seq, v); | |
2809 | } | |
2810 | #else | |
2811 | static int ipv6_route_seq_show(struct seq_file *seq, void *v) | |
2812 | { | |
2813 | return ipv6_route_native_seq_show(seq, v); | |
2814 | } | |
2815 | ||
2816 | static void ipv6_route_seq_stop(struct seq_file *seq, void *v) | |
2817 | { | |
2818 | ipv6_route_native_seq_stop(seq, v); | |
2819 | } | |
2820 | #endif | |
2821 | ||
c3506372 | 2822 | const struct seq_operations ipv6_route_seq_ops = { |
8d2ca1d7 HFS |
2823 | .start = ipv6_route_seq_start, |
2824 | .next = ipv6_route_seq_next, | |
2825 | .stop = ipv6_route_seq_stop, | |
2826 | .show = ipv6_route_seq_show | |
2827 | }; | |
8d2ca1d7 | 2828 | #endif /* CONFIG_PROC_FS */ |