projects/linux·blob master · refs

2840 lines · 66.0 KiB

1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Linux INET6 implementation
4 * Forwarding Information Database
5 *
6 * Authors:
7 * Pedro Roque <[email protected]>
8 *
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.
14 */
15
16#define pr_fmt(fmt) "IPv6: " fmt
17
18#include <linux/bpf.h>
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>
26#include <linux/list.h>
27#include <linux/slab.h>
28
29#include <net/ip.h>
30#include <net/ipv6.h>
31#include <net/ndisc.h>
32#include <net/addrconf.h>
33#include <net/lwtunnel.h>
34#include <net/fib_notifier.h>
35
36#include <net/ip_fib.h>
37#include <net/ip6_fib.h>
38#include <net/ip6_route.h>
39
40static struct kmem_cache *fib6_node_kmem __read_mostly;
41
42struct fib6_cleaner {
43 struct fib6_walker w;
44 struct net *net;
45 int (*func)(struct fib6_info *, void *arg);
46 int sernum;
47 void *arg;
48 bool skip_notify;
49};
50
51#ifdef CONFIG_IPV6_SUBTREES
52#define FWS_INIT FWS_S
53#else
54#define FWS_INIT FWS_L
55#endif
56
57static struct fib6_info *fib6_find_prefix(struct net *net,
58 struct fib6_table *table,
59 struct fib6_node *fn);
60static struct fib6_node *fib6_repair_tree(struct net *net,
61 struct fib6_table *table,
62 struct fib6_node *fn);
63static int fib6_walk(struct net *net, struct fib6_walker *w);
64static int fib6_walk_continue(struct fib6_walker *w);
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
73static void fib6_gc_timer_cb(struct timer_list *t);
74
75#define FOR_WALKERS(net, w) \
76 list_for_each_entry(w, &(net)->ipv6.fib6_walkers, lh)
77
78static void fib6_walker_link(struct net *net, struct fib6_walker *w)
79{
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);
83}
84
85static void fib6_walker_unlink(struct net *net, struct fib6_walker *w)
86{
87 write_lock_bh(&net->ipv6.fib6_walker_lock);
88 list_del_init(&w->lh);
89 write_unlock_bh(&net->ipv6.fib6_walker_lock);
90}
91
92static int fib6_new_sernum(struct net *net)
93{
94 int new, old = atomic_read(&net->ipv6.fib6_sernum);
95
96 do {
97 new = old < INT_MAX ? old + 1 : 1;
98 } while (!atomic_try_cmpxchg(&net->ipv6.fib6_sernum, &old, new));
99
100 return new;
101}
102
103enum {
104 FIB6_NO_SERNUM_CHANGE = 0,
105};
106
107void fib6_update_sernum(struct net *net, struct fib6_info *f6i)
108{
109 struct fib6_node *fn;
110
111 fn = rcu_dereference_protected(f6i->fib6_node,
112 lockdep_is_held(&f6i->fib6_table->tb6_lock));
113 if (fn)
114 WRITE_ONCE(fn->fn_sernum, fib6_new_sernum(net));
115}
116
117/*
118 * Auxiliary address test functions for the radix tree.
119 *
120 * These assume a 32bit processor (although it will work on
121 * 64bit processors)
122 */
123
124/*
125 * test bit
126 */
127#if defined(__LITTLE_ENDIAN)
128# define BITOP_BE32_SWIZZLE (0x1F & ~7)
129#else
130# define BITOP_BE32_SWIZZLE 0
131#endif
132
133static __be32 addr_bit_set(const void *token, int fn_bit)
134{
135 const __be32 *addr = token;
136 /*
137 * Here,
138 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
139 * is optimized version of
140 * htonl(1 << ((~fn_bit)&0x1F))
141 * See include/asm-generic/bitops/le.h.
142 */
143 return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) &
144 addr[fn_bit >> 5];
145}
146
147struct fib6_info *fib6_info_alloc(gfp_t gfp_flags, bool with_fib6_nh)
148{
149 struct fib6_info *f6i;
150 size_t sz = sizeof(*f6i);
151
152 if (with_fib6_nh)
153 sz += sizeof(struct fib6_nh);
154
155 f6i = kzalloc(sz, gfp_flags);
156 if (!f6i)
157 return NULL;
158
159 /* fib6_siblings is a union with nh_list, so this initializes both */
160 INIT_LIST_HEAD(&f6i->fib6_siblings);
161 refcount_set(&f6i->fib6_ref, 1);
162
163 INIT_HLIST_NODE(&f6i->gc_link);
164
165 return f6i;
166}
167
168void fib6_info_destroy_rcu(struct rcu_head *head)
169{
170 struct fib6_info *f6i = container_of(head, struct fib6_info, rcu);
171
172 WARN_ON(f6i->fib6_node);
173
174 if (f6i->nh)
175 nexthop_put(f6i->nh);
176 else
177 fib6_nh_release(f6i->fib6_nh);
178
179 ip_fib_metrics_put(f6i->fib6_metrics);
180 kfree(f6i);
181}
182EXPORT_SYMBOL_GPL(fib6_info_destroy_rcu);
183
184static struct fib6_node *node_alloc(struct net *net)
185{
186 struct fib6_node *fn;
187
188 fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
189 if (fn)
190 net->ipv6.rt6_stats->fib_nodes++;
191
192 return fn;
193}
194
195static void node_free_immediate(struct net *net, struct fib6_node *fn)
196{
197 kmem_cache_free(fib6_node_kmem, fn);
198 net->ipv6.rt6_stats->fib_nodes--;
199}
200
201static void node_free(struct net *net, struct fib6_node *fn)
202{
203 kfree_rcu(fn, rcu);
204 net->ipv6.rt6_stats->fib_nodes--;
205}
206
207static void fib6_free_table(struct fib6_table *table)
208{
209 inetpeer_invalidate_tree(&table->tb6_peers);
210 kfree(table);
211}
212
213static void fib6_link_table(struct net *net, struct fib6_table *tb)
214{
215 unsigned int h;
216
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 */
221 spin_lock_init(&tb->tb6_lock);
222 h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1);
223
224 /*
225 * No protection necessary, this is the only list mutatation
226 * operation, tables never disappear once they exist.
227 */
228 hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]);
229}
230
231#ifdef CONFIG_IPV6_MULTIPLE_TABLES
232
233static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
234{
235 struct fib6_table *table;
236
237 table = kzalloc_obj(*table, GFP_ATOMIC);
238 if (table) {
239 table->tb6_id = id;
240 rcu_assign_pointer(table->tb6_root.leaf,
241 net->ipv6.fib6_null_entry);
242 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
243 inet_peer_base_init(&table->tb6_peers);
244 INIT_HLIST_HEAD(&table->tb6_gc_hlist);
245 }
246
247 return table;
248}
249
250struct fib6_table *fib6_new_table(struct net *net, u32 id)
251{
252 struct fib6_table *tb, *new_tb;
253
254 if (id == 0)
255 id = RT6_TABLE_MAIN;
256
257 tb = fib6_get_table(net, id);
258 if (tb)
259 return tb;
260
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 }
273
274 fib6_link_table(net, new_tb);
275
276 spin_unlock_bh(&net->ipv6.fib_table_hash_lock);
277
278 return new_tb;
279}
280EXPORT_SYMBOL_GPL(fib6_new_table);
281
282struct fib6_table *fib6_get_table(struct net *net, u32 id)
283{
284 struct hlist_head *head;
285 struct fib6_table *tb;
286
287 if (!id)
288 id = RT6_TABLE_MAIN;
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)
297 return tb;
298
299 return NULL;
300}
301EXPORT_SYMBOL_GPL(fib6_get_table);
302
303static void __net_init fib6_tables_init(struct net *net)
304{
305 fib6_link_table(net, net->ipv6.fib6_main_tbl);
306 fib6_link_table(net, net->ipv6.fib6_local_tbl);
307}
308#else
309
310struct fib6_table *fib6_new_table(struct net *net, u32 id)
311{
312 return fib6_get_table(net, id);
313}
314
315struct fib6_table *fib6_get_table(struct net *net, u32 id)
316{
317 return net->ipv6.fib6_main_tbl;
318}
319
320struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
321 const struct sk_buff *skb,
322 int flags, pol_lookup_t lookup)
323{
324 struct rt6_info *rt;
325
326 rt = pol_lookup_func(lookup,
327 net, net->ipv6.fib6_main_tbl, fl6, skb, flags);
328 if (rt->dst.error == -EAGAIN) {
329 ip6_rt_put_flags(rt, flags);
330 rt = net->ipv6.ip6_null_entry;
331 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
332 dst_hold(&rt->dst);
333 }
334
335 return &rt->dst;
336}
337
338/* called with rcu lock held; no reference taken on fib6_info */
339int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
340 struct fib6_result *res, int flags)
341{
342 return fib6_table_lookup(net, net->ipv6.fib6_main_tbl, oif, fl6,
343 res, flags);
344}
345#if IS_MODULE(CONFIG_NFT_FIB_IPV6)
346EXPORT_SYMBOL_GPL(fib6_lookup);
347#endif
348
349static void __net_init fib6_tables_init(struct net *net)
350{
351 fib6_link_table(net, net->ipv6.fib6_main_tbl);
352}
353
354#endif
355
356unsigned int fib6_tables_seq_read(const struct net *net)
357{
358 unsigned int h, fib_seq = 0;
359
360 rcu_read_lock();
361 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
362 const struct hlist_head *head = &net->ipv6.fib_table_hash[h];
363 const struct fib6_table *tb;
364
365 hlist_for_each_entry_rcu(tb, head, tb6_hlist)
366 fib_seq += READ_ONCE(tb->fib_seq);
367 }
368 rcu_read_unlock();
369
370 return fib_seq;
371}
372
373static int call_fib6_entry_notifier(struct notifier_block *nb,
374 enum fib_event_type event_type,
375 struct fib6_info *rt,
376 struct netlink_ext_ack *extack)
377{
378 struct fib6_entry_notifier_info info = {
379 .info.extack = extack,
380 .rt = rt,
381 };
382
383 return call_fib6_notifier(nb, event_type, &info.info);
384}
385
386static int call_fib6_multipath_entry_notifier(struct notifier_block *nb,
387 enum fib_event_type event_type,
388 struct fib6_info *rt,
389 unsigned int nsiblings,
390 struct netlink_ext_ack *extack)
391{
392 struct fib6_entry_notifier_info info = {
393 .info.extack = extack,
394 .rt = rt,
395 .nsiblings = nsiblings,
396 };
397
398 return call_fib6_notifier(nb, event_type, &info.info);
399}
400
401int call_fib6_entry_notifiers(struct net *net,
402 enum fib_event_type event_type,
403 struct fib6_info *rt,
404 struct netlink_ext_ack *extack)
405{
406 struct fib6_entry_notifier_info info = {
407 .info.extack = extack,
408 .rt = rt,
409 };
410
411 WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1);
412 return call_fib6_notifiers(net, event_type, &info.info);
413}
414
415int call_fib6_multipath_entry_notifiers(struct net *net,
416 enum fib_event_type event_type,
417 struct fib6_info *rt,
418 unsigned int nsiblings,
419 struct netlink_ext_ack *extack)
420{
421 struct fib6_entry_notifier_info info = {
422 .info.extack = extack,
423 .rt = rt,
424 .nsiblings = nsiblings,
425 };
426
427 WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1);
428 return call_fib6_notifiers(net, event_type, &info.info);
429}
430
431int call_fib6_entry_notifiers_replace(struct net *net, struct fib6_info *rt)
432{
433 struct fib6_entry_notifier_info info = {
434 .rt = rt,
435 .nsiblings = rt->fib6_nsiblings,
436 };
437
438 WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1);
439 return call_fib6_notifiers(net, FIB_EVENT_ENTRY_REPLACE, &info.info);
440}
441
442struct fib6_dump_arg {
443 struct net *net;
444 struct notifier_block *nb;
445 struct netlink_ext_ack *extack;
446};
447
448static int fib6_rt_dump(struct fib6_info *rt, struct fib6_dump_arg *arg)
449{
450 enum fib_event_type fib_event = FIB_EVENT_ENTRY_REPLACE;
451 unsigned int nsiblings;
452 int err;
453
454 if (!rt || rt == arg->net->ipv6.fib6_null_entry)
455 return 0;
456
457 nsiblings = READ_ONCE(rt->fib6_nsiblings);
458 if (nsiblings)
459 err = call_fib6_multipath_entry_notifier(arg->nb, fib_event,
460 rt,
461 nsiblings,
462 arg->extack);
463 else
464 err = call_fib6_entry_notifier(arg->nb, fib_event, rt,
465 arg->extack);
466
467 return err;
468}
469
470static int fib6_node_dump(struct fib6_walker *w)
471{
472 int err;
473
474 err = fib6_rt_dump(w->leaf, w->args);
475 w->leaf = NULL;
476 return err;
477}
478
479static int fib6_table_dump(struct net *net, struct fib6_table *tb,
480 struct fib6_walker *w)
481{
482 int err;
483
484 w->root = &tb->tb6_root;
485 spin_lock_bh(&tb->tb6_lock);
486 err = fib6_walk(net, w);
487 spin_unlock_bh(&tb->tb6_lock);
488 return err;
489}
490
491/* Called with rcu_read_lock() */
492int fib6_tables_dump(struct net *net, struct notifier_block *nb,
493 struct netlink_ext_ack *extack)
494{
495 struct fib6_dump_arg arg;
496 struct fib6_walker *w;
497 unsigned int h;
498 int err = 0;
499
500 w = kzalloc_obj(*w, GFP_ATOMIC);
501 if (!w)
502 return -ENOMEM;
503
504 w->func = fib6_node_dump;
505 arg.net = net;
506 arg.nb = nb;
507 arg.extack = extack;
508 w->args = &arg;
509
510 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
511 struct hlist_head *head = &net->ipv6.fib_table_hash[h];
512 struct fib6_table *tb;
513
514 hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
515 err = fib6_table_dump(net, tb, w);
516 if (err)
517 goto out;
518 }
519 }
520
521out:
522 kfree(w);
523
524 /* The tree traversal function should never return a positive value. */
525 return err > 0 ? -EINVAL : err;
526}
527
528static int fib6_dump_node(struct fib6_walker *w)
529{
530 int res;
531 struct fib6_info *rt;
532
533 for_each_fib6_walker_rt(w) {
534 res = rt6_dump_route(rt, w->args, w->skip_in_node);
535 if (res >= 0) {
536 /* Frame is full, suspend walking */
537 w->leaf = rt;
538
539 /* We'll restart from this node, so if some routes were
540 * already dumped, skip them next time.
541 */
542 w->skip_in_node += res;
543
544 return 1;
545 }
546 w->skip_in_node = 0;
547
548 /* Multipath routes are dumped in one route with the
549 * RTA_MULTIPATH attribute. Jump 'rt' to point to the
550 * last sibling of this route (no need to dump the
551 * sibling routes again)
552 */
553 if (rt->fib6_nsiblings)
554 rt = list_last_entry(&rt->fib6_siblings,
555 struct fib6_info,
556 fib6_siblings);
557 }
558 w->leaf = NULL;
559 return 0;
560}
561
562static void fib6_dump_end(struct netlink_callback *cb)
563{
564 struct net *net = sock_net(cb->skb->sk);
565 struct fib6_walker *w = (void *)cb->args[2];
566
567 if (w) {
568 if (cb->args[4]) {
569 cb->args[4] = 0;
570 fib6_walker_unlink(net, w);
571 }
572 cb->args[2] = 0;
573 kfree(w);
574 }
575 cb->done = (void *)cb->args[3];
576 cb->args[1] = 3;
577}
578
579static int fib6_dump_done(struct netlink_callback *cb)
580{
581 fib6_dump_end(cb);
582 return cb->done ? cb->done(cb) : 0;
583}
584
585static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
586 struct netlink_callback *cb)
587{
588 struct net *net = sock_net(skb->sk);
589 struct fib6_walker *w;
590 int res;
591
592 w = (void *)cb->args[2];
593 w->root = &table->tb6_root;
594
595 if (cb->args[4] == 0) {
596 w->count = 0;
597 w->skip = 0;
598 w->skip_in_node = 0;
599
600 spin_lock_bh(&table->tb6_lock);
601 res = fib6_walk(net, w);
602 spin_unlock_bh(&table->tb6_lock);
603 if (res > 0) {
604 cb->args[4] = 1;
605 cb->args[5] = READ_ONCE(w->root->fn_sernum);
606 }
607 } else {
608 int sernum = READ_ONCE(w->root->fn_sernum);
609 if (cb->args[5] != sernum) {
610 /* Begin at the root if the tree changed */
611 cb->args[5] = sernum;
612 w->state = FWS_INIT;
613 w->node = w->root;
614 w->skip = w->count;
615 w->skip_in_node = 0;
616 } else
617 w->skip = 0;
618
619 spin_lock_bh(&table->tb6_lock);
620 res = fib6_walk_continue(w);
621 spin_unlock_bh(&table->tb6_lock);
622 if (res <= 0) {
623 fib6_walker_unlink(net, w);
624 cb->args[4] = 0;
625 }
626 }
627
628 return res;
629}
630
631static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
632{
633 struct rt6_rtnl_dump_arg arg = {
634 .filter.dump_exceptions = true,
635 .filter.dump_routes = true,
636 };
637 const struct nlmsghdr *nlh = cb->nlh;
638 struct net *net = sock_net(skb->sk);
639 struct hlist_head *head;
640 struct fib6_walker *w;
641 struct fib6_table *tb;
642 unsigned int h, s_h;
643 int err = 0;
644 u32 s_id;
645
646 rcu_read_lock();
647 if (cb->strict_check) {
648 err = ip_valid_fib_dump_req(net, nlh, &arg.filter, cb);
649 if (err < 0)
650 goto unlock;
651 } else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) {
652 struct rtmsg *rtm = nlmsg_data(nlh);
653
654 if (rtm->rtm_flags & RTM_F_PREFIX)
655 arg.filter.flags = RTM_F_PREFIX;
656 }
657
658 w = (void *)cb->args[2];
659 if (!w) {
660 /* New dump:
661 *
662 * 1. allocate and initialize walker.
663 */
664 w = kzalloc_obj(*w, GFP_ATOMIC);
665 if (!w) {
666 err = -ENOMEM;
667 goto unlock;
668 }
669 w->func = fib6_dump_node;
670 cb->args[2] = (long)w;
671
672 /* 2. hook callback destructor.
673 */
674 cb->args[3] = (long)cb->done;
675 cb->done = fib6_dump_done;
676
677 }
678
679 arg.skb = skb;
680 arg.cb = cb;
681 arg.net = net;
682 w->args = &arg;
683
684 if (arg.filter.table_id) {
685 tb = fib6_get_table(net, arg.filter.table_id);
686 if (!tb) {
687 if (rtnl_msg_family(cb->nlh) != PF_INET6)
688 goto unlock;
689
690 NL_SET_ERR_MSG_MOD(cb->extack, "FIB table does not exist");
691 err = -ENOENT;
692 goto unlock;
693 }
694
695 if (!cb->args[0]) {
696 err = fib6_dump_table(tb, skb, cb);
697 if (!err)
698 cb->args[0] = 1;
699 }
700 goto unlock;
701 }
702
703 s_h = cb->args[0];
704 s_id = cb->args[1];
705
706 for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_id = 0) {
707 head = &net->ipv6.fib_table_hash[h];
708 hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
709 if (s_id && tb->tb6_id != s_id)
710 continue;
711
712 s_id = 0;
713 cb->args[1] = tb->tb6_id;
714 err = fib6_dump_table(tb, skb, cb);
715 if (err != 0)
716 goto out;
717 }
718 }
719out:
720 cb->args[0] = h;
721
722unlock:
723 rcu_read_unlock();
724 if (err <= 0)
725 fib6_dump_end(cb);
726 return err;
727}
728
729void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val)
730{
731 struct dst_metrics *m;
732
733 if (!f6i)
734 return;
735
736 if (READ_ONCE(f6i->fib6_metrics) == &dst_default_metrics) {
737 struct dst_metrics *dflt = (struct dst_metrics *)&dst_default_metrics;
738 struct dst_metrics *p = kzalloc_obj(*p, GFP_ATOMIC);
739
740 if (!p)
741 return;
742
743 p->metrics[metric - 1] = val;
744 refcount_set(&p->refcnt, 1);
745 if (cmpxchg(&f6i->fib6_metrics, dflt, p) != dflt)
746 kfree(p);
747 else
748 return;
749 }
750
751 m = READ_ONCE(f6i->fib6_metrics);
752 WRITE_ONCE(m->metrics[metric - 1], val);
753}
754
755/*
756 * Routing Table
757 *
758 * return the appropriate node for a routing tree "add" operation
759 * by either creating and inserting or by returning an existing
760 * node.
761 */
762
763static struct fib6_node *fib6_add_1(struct net *net,
764 struct fib6_table *table,
765 struct fib6_node *root,
766 struct in6_addr *addr, int plen,
767 int offset, int allow_create,
768 int replace_required,
769 struct netlink_ext_ack *extack)
770{
771 struct fib6_node *fn, *in, *ln;
772 struct fib6_node *pn = NULL;
773 struct rt6key *key;
774 int bit;
775 __be32 dir = 0;
776
777 /* insert node in tree */
778
779 fn = root;
780
781 do {
782 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
783 lockdep_is_held(&table->tb6_lock));
784 key = (struct rt6key *)((u8 *)leaf + offset);
785
786 /*
787 * Prefix match
788 */
789 if (plen < fn->fn_bit ||
790 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) {
791 if (!allow_create) {
792 if (replace_required) {
793 NL_SET_ERR_MSG(extack,
794 "Can not replace route - no match found");
795 pr_warn("Can't replace route, no match found\n");
796 return ERR_PTR(-ENOENT);
797 }
798 pr_warn("NLM_F_CREATE should be set when creating new route\n");
799 }
800 goto insert_above;
801 }
802
803 /*
804 * Exact match ?
805 */
806
807 if (plen == fn->fn_bit) {
808 /* clean up an intermediate node */
809 if (!(fn->fn_flags & RTN_RTINFO)) {
810 RCU_INIT_POINTER(fn->leaf, NULL);
811 fib6_info_release(leaf);
812 /* remove null_entry in the root node */
813 } else if (fn->fn_flags & RTN_TL_ROOT &&
814 rcu_access_pointer(fn->leaf) ==
815 net->ipv6.fib6_null_entry) {
816 RCU_INIT_POINTER(fn->leaf, NULL);
817 }
818
819 return fn;
820 }
821
822 /*
823 * We have more bits to go
824 */
825
826 /* Try to walk down on tree. */
827 dir = addr_bit_set(addr, fn->fn_bit);
828 pn = fn;
829 fn = dir ?
830 rcu_dereference_protected(fn->right,
831 lockdep_is_held(&table->tb6_lock)) :
832 rcu_dereference_protected(fn->left,
833 lockdep_is_held(&table->tb6_lock));
834 } while (fn);
835
836 if (!allow_create) {
837 /* We should not create new node because
838 * NLM_F_REPLACE was specified without NLM_F_CREATE
839 * I assume it is safe to require NLM_F_CREATE when
840 * REPLACE flag is used! Later we may want to remove the
841 * check for replace_required, because according
842 * to netlink specification, NLM_F_CREATE
843 * MUST be specified if new route is created.
844 * That would keep IPv6 consistent with IPv4
845 */
846 if (replace_required) {
847 NL_SET_ERR_MSG(extack,
848 "Can not replace route - no match found");
849 pr_warn("Can't replace route, no match found\n");
850 return ERR_PTR(-ENOENT);
851 }
852 pr_warn("NLM_F_CREATE should be set when creating new route\n");
853 }
854 /*
855 * We walked to the bottom of tree.
856 * Create new leaf node without children.
857 */
858
859 ln = node_alloc(net);
860
861 if (!ln)
862 return ERR_PTR(-ENOMEM);
863 ln->fn_bit = plen;
864 RCU_INIT_POINTER(ln->parent, pn);
865
866 if (dir)
867 rcu_assign_pointer(pn->right, ln);
868 else
869 rcu_assign_pointer(pn->left, ln);
870
871 return ln;
872
873
874insert_above:
875 /*
876 * split since we don't have a common prefix anymore or
877 * we have a less significant route.
878 * we've to insert an intermediate node on the list
879 * this new node will point to the one we need to create
880 * and the current
881 */
882
883 pn = rcu_dereference_protected(fn->parent,
884 lockdep_is_held(&table->tb6_lock));
885
886 /* find 1st bit in difference between the 2 addrs.
887
888 See comment in __ipv6_addr_diff: bit may be an invalid value,
889 but if it is >= plen, the value is ignored in any case.
890 */
891
892 bit = __ipv6_addr_diff(addr, &key->addr, sizeof(*addr));
893
894 /*
895 * (intermediate)[in]
896 * / \
897 * (new leaf node)[ln] (old node)[fn]
898 */
899 if (plen > bit) {
900 in = node_alloc(net);
901 ln = node_alloc(net);
902
903 if (!in || !ln) {
904 if (in)
905 node_free_immediate(net, in);
906 if (ln)
907 node_free_immediate(net, ln);
908 return ERR_PTR(-ENOMEM);
909 }
910
911 /*
912 * new intermediate node.
913 * RTN_RTINFO will
914 * be off since that an address that chooses one of
915 * the branches would not match less specific routes
916 * in the other branch
917 */
918
919 in->fn_bit = bit;
920
921 RCU_INIT_POINTER(in->parent, pn);
922 in->leaf = fn->leaf;
923 fib6_info_hold(rcu_dereference_protected(in->leaf,
924 lockdep_is_held(&table->tb6_lock)));
925
926 /* update parent pointer */
927 if (dir)
928 rcu_assign_pointer(pn->right, in);
929 else
930 rcu_assign_pointer(pn->left, in);
931
932 ln->fn_bit = plen;
933
934 RCU_INIT_POINTER(ln->parent, in);
935 rcu_assign_pointer(fn->parent, in);
936
937 if (addr_bit_set(addr, bit)) {
938 rcu_assign_pointer(in->right, ln);
939 rcu_assign_pointer(in->left, fn);
940 } else {
941 rcu_assign_pointer(in->left, ln);
942 rcu_assign_pointer(in->right, fn);
943 }
944 } else { /* plen <= bit */
945
946 /*
947 * (new leaf node)[ln]
948 * / \
949 * (old node)[fn] NULL
950 */
951
952 ln = node_alloc(net);
953
954 if (!ln)
955 return ERR_PTR(-ENOMEM);
956
957 ln->fn_bit = plen;
958
959 RCU_INIT_POINTER(ln->parent, pn);
960
961 if (addr_bit_set(&key->addr, plen))
962 RCU_INIT_POINTER(ln->right, fn);
963 else
964 RCU_INIT_POINTER(ln->left, fn);
965
966 rcu_assign_pointer(fn->parent, ln);
967
968 if (dir)
969 rcu_assign_pointer(pn->right, ln);
970 else
971 rcu_assign_pointer(pn->left, ln);
972 }
973 return ln;
974}
975
976static void __fib6_drop_pcpu_from(struct fib6_nh *fib6_nh,
977 const struct fib6_info *match)
978{
979 int cpu;
980
981 if (!fib6_nh->rt6i_pcpu)
982 return;
983
984 rcu_read_lock();
985 /* release the reference to this fib entry from
986 * all of its cached pcpu routes
987 */
988 for_each_possible_cpu(cpu) {
989 struct rt6_info **ppcpu_rt;
990 struct rt6_info *pcpu_rt;
991
992 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
993
994 /* Paired with xchg() in rt6_get_pcpu_route() */
995 pcpu_rt = READ_ONCE(*ppcpu_rt);
996
997 /* only dropping the 'from' reference if the cached route
998 * is using 'match'. The cached pcpu_rt->from only changes
999 * from a fib6_info to NULL (ip6_dst_destroy); it can never
1000 * change from one fib6_info reference to another
1001 */
1002 if (pcpu_rt && rcu_access_pointer(pcpu_rt->from) == match) {
1003 struct fib6_info *from;
1004
1005 from = unrcu_pointer(xchg(&pcpu_rt->from, NULL));
1006 fib6_info_release(from);
1007 }
1008 }
1009 rcu_read_unlock();
1010}
1011
1012static int fib6_nh_drop_pcpu_from(struct fib6_nh *nh, void *_arg)
1013{
1014 struct fib6_info *arg = _arg;
1015
1016 __fib6_drop_pcpu_from(nh, arg);
1017 return 0;
1018}
1019
1020static void fib6_drop_pcpu_from(struct fib6_info *f6i)
1021{
1022 /* Make sure rt6_make_pcpu_route() wont add other percpu routes
1023 * while we are cleaning them here.
1024 */
1025 f6i->fib6_destroying = 1;
1026 mb(); /* paired with the cmpxchg() in rt6_make_pcpu_route() */
1027
1028 if (f6i->nh) {
1029 rcu_read_lock();
1030 nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_drop_pcpu_from, f6i);
1031 rcu_read_unlock();
1032 } else {
1033 struct fib6_nh *fib6_nh;
1034
1035 fib6_nh = f6i->fib6_nh;
1036 __fib6_drop_pcpu_from(fib6_nh, f6i);
1037 }
1038}
1039
1040static void fib6_purge_rt(struct fib6_info *rt, struct fib6_node *fn,
1041 struct net *net)
1042{
1043 struct fib6_table *table = rt->fib6_table;
1044
1045 /* Flush all cached dst in exception table */
1046 fib6_drop_pcpu_from(rt);
1047 rt6_flush_exceptions(rt);
1048
1049 if (rt->nh) {
1050 spin_lock(&rt->nh->lock);
1051
1052 if (!list_empty(&rt->nh_list))
1053 list_del_init(&rt->nh_list);
1054
1055 spin_unlock(&rt->nh->lock);
1056 }
1057
1058 if (refcount_read(&rt->fib6_ref) != 1) {
1059 /* This route is used as dummy address holder in some split
1060 * nodes. It is not leaked, but it still holds other resources,
1061 * which must be released in time. So, scan ascendant nodes
1062 * and replace dummy references to this route with references
1063 * to still alive ones.
1064 */
1065 while (fn) {
1066 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
1067 lockdep_is_held(&table->tb6_lock));
1068 struct fib6_info *new_leaf;
1069 if (!(fn->fn_flags & RTN_RTINFO) && leaf == rt) {
1070 new_leaf = fib6_find_prefix(net, table, fn);
1071 fib6_info_hold(new_leaf);
1072
1073 rcu_assign_pointer(fn->leaf, new_leaf);
1074 fib6_info_release(rt);
1075 }
1076 fn = rcu_dereference_protected(fn->parent,
1077 lockdep_is_held(&table->tb6_lock));
1078 }
1079 }
1080
1081 fib6_clean_expires(rt);
1082 fib6_remove_gc_list(rt);
1083}
1084
1085/*
1086 * Insert routing information in a node.
1087 */
1088
1089static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
1090 struct nl_info *info, struct netlink_ext_ack *extack,
1091 struct list_head *purge_list)
1092{
1093 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
1094 lockdep_is_held(&rt->fib6_table->tb6_lock));
1095 struct fib6_info *iter = NULL;
1096 struct fib6_info __rcu **ins;
1097 struct fib6_info __rcu **fallback_ins = NULL;
1098 int replace = (info->nlh &&
1099 (info->nlh->nlmsg_flags & NLM_F_REPLACE));
1100 int add = (!info->nlh ||
1101 (info->nlh->nlmsg_flags & NLM_F_CREATE));
1102 int found = 0;
1103 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
1104 bool notify_sibling_rt = false;
1105 u16 nlflags = NLM_F_EXCL;
1106 int err;
1107
1108 if (info->nlh && (info->nlh->nlmsg_flags & NLM_F_APPEND))
1109 nlflags |= NLM_F_APPEND;
1110
1111 ins = &fn->leaf;
1112
1113 for (iter = leaf; iter;
1114 iter = rcu_dereference_protected(iter->fib6_next,
1115 lockdep_is_held(&rt->fib6_table->tb6_lock))) {
1116 /*
1117 * Search for duplicates
1118 */
1119
1120 if (iter->fib6_metric == rt->fib6_metric) {
1121 /*
1122 * Same priority level
1123 */
1124 if (info->nlh &&
1125 (info->nlh->nlmsg_flags & NLM_F_EXCL))
1126 return -EEXIST;
1127
1128 nlflags &= ~NLM_F_EXCL;
1129 if (replace) {
1130 if (rt_can_ecmp == rt6_qualify_for_ecmp(iter)) {
1131 found++;
1132 break;
1133 }
1134 fallback_ins = fallback_ins ?: ins;
1135 goto next_iter;
1136 }
1137
1138 if (rt6_duplicate_nexthop(iter, rt)) {
1139 if (rt->fib6_nsiblings)
1140 WRITE_ONCE(rt->fib6_nsiblings, 0);
1141 if (!(iter->fib6_flags & RTF_EXPIRES))
1142 return -EEXIST;
1143 if (!(rt->fib6_flags & RTF_EXPIRES)) {
1144 fib6_clean_expires(iter);
1145 fib6_may_remove_gc_list(info->nl_net, iter);
1146 } else {
1147 fib6_set_expires(iter, rt->expires);
1148 fib6_add_gc_list(iter);
1149 }
1150 if (!(rt->fib6_flags & (RTF_ADDRCONF | RTF_PREFIX_RT)) &&
1151 (iter->nh || !iter->fib6_nh->fib_nh_gw_family)) {
1152 iter->fib6_flags &= ~RTF_ADDRCONF;
1153 iter->fib6_flags &= ~RTF_PREFIX_RT;
1154 }
1155
1156 if (rt->fib6_pmtu)
1157 fib6_metric_set(iter, RTAX_MTU,
1158 rt->fib6_pmtu);
1159 return -EEXIST;
1160 }
1161 /* If we have the same destination and the same metric,
1162 * but not the same gateway, then the route we try to
1163 * add is sibling to this route, increment our counter
1164 * of siblings, and later we will add our route to the
1165 * list.
1166 * Only static routes (which don't have flag
1167 * RTF_EXPIRES) are used for ECMPv6.
1168 *
1169 * To avoid long list, we only had siblings if the
1170 * route have a gateway.
1171 */
1172 if (rt_can_ecmp &&
1173 rt6_qualify_for_ecmp(iter))
1174 WRITE_ONCE(rt->fib6_nsiblings,
1175 rt->fib6_nsiblings + 1);
1176 }
1177
1178 if (iter->fib6_metric > rt->fib6_metric)
1179 break;
1180
1181next_iter:
1182 ins = &iter->fib6_next;
1183 }
1184
1185 if (fallback_ins && !found) {
1186 /* No matching route with same ecmp-able-ness found, replace
1187 * first matching route
1188 */
1189 ins = fallback_ins;
1190 iter = rcu_dereference_protected(*ins,
1191 lockdep_is_held(&rt->fib6_table->tb6_lock));
1192 found++;
1193 }
1194
1195 /* Reset round-robin state, if necessary */
1196 if (ins == &fn->leaf)
1197 fn->rr_ptr = NULL;
1198
1199 /* Link this route to others same route. */
1200 if (rt->fib6_nsiblings) {
1201 unsigned int fib6_nsiblings;
1202 struct fib6_info *sibling, *temp_sibling;
1203
1204 /* Find the first route that have the same metric */
1205 sibling = leaf;
1206 notify_sibling_rt = true;
1207 while (sibling) {
1208 if (sibling->fib6_metric == rt->fib6_metric &&
1209 rt6_qualify_for_ecmp(sibling)) {
1210 list_add_tail_rcu(&rt->fib6_siblings,
1211 &sibling->fib6_siblings);
1212 break;
1213 }
1214 sibling = rcu_dereference_protected(sibling->fib6_next,
1215 lockdep_is_held(&rt->fib6_table->tb6_lock));
1216 notify_sibling_rt = false;
1217 }
1218 /* For each sibling in the list, increment the counter of
1219 * siblings. BUG() if counters does not match, list of siblings
1220 * is broken!
1221 */
1222 fib6_nsiblings = 0;
1223 list_for_each_entry_safe(sibling, temp_sibling,
1224 &rt->fib6_siblings, fib6_siblings) {
1225 WRITE_ONCE(sibling->fib6_nsiblings,
1226 sibling->fib6_nsiblings + 1);
1227 BUG_ON(sibling->fib6_nsiblings != rt->fib6_nsiblings);
1228 fib6_nsiblings++;
1229 }
1230 BUG_ON(fib6_nsiblings != rt->fib6_nsiblings);
1231 rcu_read_lock();
1232 rt6_multipath_rebalance(temp_sibling);
1233 rcu_read_unlock();
1234 }
1235
1236 /*
1237 * insert node
1238 */
1239 if (!replace) {
1240 if (!add)
1241 pr_warn("NLM_F_CREATE should be set when creating new route\n");
1242
1243add:
1244 nlflags |= NLM_F_CREATE;
1245
1246 /* The route should only be notified if it is the first
1247 * route in the node or if it is added as a sibling
1248 * route to the first route in the node.
1249 */
1250 if (!info->skip_notify_kernel &&
1251 (notify_sibling_rt || ins == &fn->leaf)) {
1252 enum fib_event_type fib_event;
1253
1254 if (notify_sibling_rt)
1255 fib_event = FIB_EVENT_ENTRY_APPEND;
1256 else
1257 fib_event = FIB_EVENT_ENTRY_REPLACE;
1258 err = call_fib6_entry_notifiers(info->nl_net,
1259 fib_event, rt,
1260 extack);
1261 if (err) {
1262 struct fib6_info *sibling, *next_sibling;
1263
1264 /* If the route has siblings, then it first
1265 * needs to be unlinked from them.
1266 */
1267 if (!rt->fib6_nsiblings)
1268 return err;
1269
1270 list_for_each_entry_safe(sibling, next_sibling,
1271 &rt->fib6_siblings,
1272 fib6_siblings)
1273 WRITE_ONCE(sibling->fib6_nsiblings,
1274 sibling->fib6_nsiblings - 1);
1275 WRITE_ONCE(rt->fib6_nsiblings, 0);
1276 list_del_rcu(&rt->fib6_siblings);
1277 rcu_read_lock();
1278 rt6_multipath_rebalance(next_sibling);
1279 rcu_read_unlock();
1280 return err;
1281 }
1282 }
1283
1284 rcu_assign_pointer(rt->fib6_next, iter);
1285 fib6_info_hold(rt);
1286 rcu_assign_pointer(rt->fib6_node, fn);
1287 rcu_assign_pointer(*ins, rt);
1288 if (!info->skip_notify)
1289 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
1290 info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
1291
1292 if (!(fn->fn_flags & RTN_RTINFO)) {
1293 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
1294 fn->fn_flags |= RTN_RTINFO;
1295 }
1296
1297 } else {
1298 int nsiblings;
1299
1300 if (!found) {
1301 if (add)
1302 goto add;
1303 pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
1304 return -ENOENT;
1305 }
1306
1307 if (!info->skip_notify_kernel && ins == &fn->leaf) {
1308 err = call_fib6_entry_notifiers(info->nl_net,
1309 FIB_EVENT_ENTRY_REPLACE,
1310 rt, extack);
1311 if (err)
1312 return err;
1313 }
1314
1315 fib6_info_hold(rt);
1316 rcu_assign_pointer(rt->fib6_node, fn);
1317 rt->fib6_next = iter->fib6_next;
1318 rcu_assign_pointer(*ins, rt);
1319 if (!info->skip_notify)
1320 inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE);
1321 if (!(fn->fn_flags & RTN_RTINFO)) {
1322 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
1323 fn->fn_flags |= RTN_RTINFO;
1324 }
1325 nsiblings = iter->fib6_nsiblings;
1326 iter->fib6_node = NULL;
1327 list_add(&iter->purge_link, purge_list);
1328 if (rcu_access_pointer(fn->rr_ptr) == iter)
1329 fn->rr_ptr = NULL;
1330
1331 if (nsiblings) {
1332 /* Replacing an ECMP route, remove all siblings */
1333 ins = &rt->fib6_next;
1334 iter = rcu_dereference_protected(*ins,
1335 lockdep_is_held(&rt->fib6_table->tb6_lock));
1336 while (iter) {
1337 if (iter->fib6_metric > rt->fib6_metric)
1338 break;
1339 if (rt6_qualify_for_ecmp(iter)) {
1340 *ins = iter->fib6_next;
1341 iter->fib6_node = NULL;
1342 list_add(&iter->purge_link, purge_list);
1343 if (rcu_access_pointer(fn->rr_ptr) == iter)
1344 fn->rr_ptr = NULL;
1345 nsiblings--;
1346 info->nl_net->ipv6.rt6_stats->fib_rt_entries--;
1347 } else {
1348 ins = &iter->fib6_next;
1349 }
1350 iter = rcu_dereference_protected(*ins,
1351 lockdep_is_held(&rt->fib6_table->tb6_lock));
1352 }
1353 WARN_ON(nsiblings != 0);
1354 }
1355 }
1356
1357 return 0;
1358}
1359
1360static int fib6_add_rt2node_nh(struct fib6_node *fn, struct fib6_info *rt,
1361 struct nl_info *info, struct netlink_ext_ack *extack,
1362 struct list_head *purge_list)
1363{
1364 int err;
1365
1366 spin_lock(&rt->nh->lock);
1367
1368 if (rt->nh->dead) {
1369 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
1370 err = -EINVAL;
1371 } else {
1372 err = fib6_add_rt2node(fn, rt, info, extack, purge_list);
1373 if (!err)
1374 list_add(&rt->nh_list, &rt->nh->f6i_list);
1375 }
1376
1377 spin_unlock(&rt->nh->lock);
1378
1379 return err;
1380}
1381
1382static void fib6_start_gc(struct net *net, struct fib6_info *rt)
1383{
1384 if (!timer_pending(&net->ipv6.ip6_fib_timer) &&
1385 (rt->fib6_flags & RTF_EXPIRES))
1386 mod_timer(&net->ipv6.ip6_fib_timer,
1387 jiffies + READ_ONCE(net->ipv6.sysctl.ip6_rt_gc_interval));
1388}
1389
1390void fib6_force_start_gc(struct net *net)
1391{
1392 if (!timer_pending(&net->ipv6.ip6_fib_timer))
1393 mod_timer(&net->ipv6.ip6_fib_timer,
1394 jiffies + READ_ONCE(net->ipv6.sysctl.ip6_rt_gc_interval));
1395}
1396
1397static void __fib6_update_sernum_upto_root(struct fib6_info *rt,
1398 int sernum)
1399{
1400 struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node,
1401 lockdep_is_held(&rt->fib6_table->tb6_lock));
1402
1403 /* paired with smp_rmb() in fib6_get_cookie_safe() */
1404 smp_wmb();
1405 while (fn) {
1406 WRITE_ONCE(fn->fn_sernum, sernum);
1407 fn = rcu_dereference_protected(fn->parent,
1408 lockdep_is_held(&rt->fib6_table->tb6_lock));
1409 }
1410}
1411
1412void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt)
1413{
1414 __fib6_update_sernum_upto_root(rt, fib6_new_sernum(net));
1415}
1416
1417/*
1418 * Add routing information to the routing tree.
1419 * <destination addr>/<source addr>
1420 * with source addr info in sub-trees
1421 * Need to own table->tb6_lock
1422 */
1423
1424int fib6_add(struct fib6_node *root, struct fib6_info *rt,
1425 struct nl_info *info, struct netlink_ext_ack *extack)
1426{
1427 struct fib6_table *table = rt->fib6_table;
1428 LIST_HEAD(purge_list);
1429 struct fib6_node *fn;
1430#ifdef CONFIG_IPV6_SUBTREES
1431 struct fib6_node *pn = NULL;
1432#endif
1433 int err = -ENOMEM;
1434 int allow_create = 1;
1435 int replace_required = 0;
1436
1437 if (info->nlh) {
1438 if (!(info->nlh->nlmsg_flags & NLM_F_CREATE))
1439 allow_create = 0;
1440 if (info->nlh->nlmsg_flags & NLM_F_REPLACE)
1441 replace_required = 1;
1442 }
1443 if (!allow_create && !replace_required)
1444 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
1445
1446 fn = fib6_add_1(info->nl_net, table, root,
1447 &rt->fib6_dst.addr, rt->fib6_dst.plen,
1448 offsetof(struct fib6_info, fib6_dst), allow_create,
1449 replace_required, extack);
1450 if (IS_ERR(fn)) {
1451 err = PTR_ERR(fn);
1452 fn = NULL;
1453 goto out;
1454 }
1455
1456#ifdef CONFIG_IPV6_SUBTREES
1457 pn = fn;
1458
1459 if (rt->fib6_src.plen) {
1460 struct fib6_node *sn;
1461
1462 if (!rcu_access_pointer(fn->subtree)) {
1463 struct fib6_node *sfn;
1464
1465 /*
1466 * Create subtree.
1467 *
1468 * fn[main tree]
1469 * |
1470 * sfn[subtree root]
1471 * \
1472 * sn[new leaf node]
1473 */
1474
1475 /* Create subtree root node */
1476 sfn = node_alloc(info->nl_net);
1477 if (!sfn)
1478 goto failure;
1479
1480 fib6_info_hold(info->nl_net->ipv6.fib6_null_entry);
1481 rcu_assign_pointer(sfn->leaf,
1482 info->nl_net->ipv6.fib6_null_entry);
1483 sfn->fn_flags = RTN_ROOT;
1484
1485 /* Now add the first leaf node to new subtree */
1486
1487 sn = fib6_add_1(info->nl_net, table, sfn,
1488 &rt->fib6_src.addr, rt->fib6_src.plen,
1489 offsetof(struct fib6_info, fib6_src),
1490 allow_create, replace_required, extack);
1491
1492 if (IS_ERR(sn)) {
1493 /* If it is failed, discard just allocated
1494 root, and then (in failure) stale node
1495 in main tree.
1496 */
1497 fib6_info_release(info->nl_net->ipv6.fib6_null_entry);
1498 node_free_immediate(info->nl_net, sfn);
1499 err = PTR_ERR(sn);
1500 goto failure;
1501 }
1502
1503 /* Now link new subtree to main tree */
1504 rcu_assign_pointer(sfn->parent, fn);
1505 rcu_assign_pointer(fn->subtree, sfn);
1506 } else {
1507 sn = fib6_add_1(info->nl_net, table, FIB6_SUBTREE(fn),
1508 &rt->fib6_src.addr, rt->fib6_src.plen,
1509 offsetof(struct fib6_info, fib6_src),
1510 allow_create, replace_required, extack);
1511
1512 if (IS_ERR(sn)) {
1513 err = PTR_ERR(sn);
1514 goto failure;
1515 }
1516 }
1517
1518 if (!rcu_access_pointer(fn->leaf)) {
1519 if (fn->fn_flags & RTN_TL_ROOT) {
1520 /* put back null_entry for root node */
1521 rcu_assign_pointer(fn->leaf,
1522 info->nl_net->ipv6.fib6_null_entry);
1523 } else {
1524 fib6_info_hold(rt);
1525 rcu_assign_pointer(fn->leaf, rt);
1526 }
1527 }
1528 fn = sn;
1529 }
1530#endif
1531
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);
1536 if (!err) {
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
1545 __fib6_update_sernum_upto_root(rt, fib6_new_sernum(info->nl_net));
1546
1547 if (rt->fib6_flags & RTF_EXPIRES)
1548 fib6_add_gc_list(rt);
1549
1550 fib6_start_gc(info->nl_net, rt);
1551 }
1552
1553out:
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 */
1560 if (pn != fn) {
1561 struct fib6_info *pn_leaf =
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);
1567 fib6_info_release(rt);
1568 }
1569 if (!pn_leaf && !(pn->fn_flags & RTN_RTINFO)) {
1570 pn_leaf = fib6_find_prefix(info->nl_net, table,
1571 pn);
1572 if (!pn_leaf)
1573 pn_leaf =
1574 info->nl_net->ipv6.fib6_null_entry;
1575 fib6_info_hold(pn_leaf);
1576 rcu_assign_pointer(pn->leaf, pn_leaf);
1577 }
1578 }
1579#endif
1580 goto failure;
1581 } else if (fib6_requires_src(rt)) {
1582 fib6_routes_require_src_inc(info->nl_net);
1583 }
1584 return err;
1585
1586failure:
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.
1593 */
1594 if (fn &&
1595 (!(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)) ||
1596 (fn->fn_flags & RTN_TL_ROOT &&
1597 !rcu_access_pointer(fn->leaf))))
1598 fib6_repair_tree(info->nl_net, table, fn);
1599 return err;
1600}
1601
1602/*
1603 * Routing tree lookup
1604 *
1605 */
1606
1607struct lookup_args {
1608 int offset; /* key offset on fib6_info */
1609 const struct in6_addr *addr; /* search key */
1610};
1611
1612static struct fib6_node *fib6_node_lookup_1(struct fib6_node *root,
1613 struct lookup_args *args)
1614{
1615 struct fib6_node *fn;
1616 __be32 dir;
1617
1618 if (unlikely(args->offset == 0))
1619 return NULL;
1620
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
1632 next = dir ? rcu_dereference(fn->right) :
1633 rcu_dereference(fn->left);
1634
1635 if (next) {
1636 fn = next;
1637 continue;
1638 }
1639 break;
1640 }
1641
1642 while (fn) {
1643 struct fib6_node *subtree = FIB6_SUBTREE(fn);
1644
1645 if (subtree || fn->fn_flags & RTN_RTINFO) {
1646 struct fib6_info *leaf = rcu_dereference(fn->leaf);
1647 struct rt6key *key;
1648
1649 if (!leaf)
1650 goto backtrack;
1651
1652 key = (struct rt6key *) ((u8 *)leaf + args->offset);
1653
1654 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
1655#ifdef CONFIG_IPV6_SUBTREES
1656 if (subtree) {
1657 struct fib6_node *sfn;
1658 sfn = fib6_node_lookup_1(subtree,
1659 args + 1);
1660 if (!sfn)
1661 goto backtrack;
1662 fn = sfn;
1663 }
1664#endif
1665 if (fn->fn_flags & RTN_RTINFO)
1666 return fn;
1667 }
1668 }
1669backtrack:
1670 if (fn->fn_flags & RTN_ROOT)
1671 break;
1672
1673 fn = rcu_dereference(fn->parent);
1674 }
1675
1676 return NULL;
1677}
1678
1679/* called with rcu_read_lock() held
1680 */
1681struct fib6_node *fib6_node_lookup(struct fib6_node *root,
1682 const struct in6_addr *daddr,
1683 const struct in6_addr *saddr)
1684{
1685 struct fib6_node *fn;
1686 struct lookup_args args[] = {
1687 {
1688 .offset = offsetof(struct fib6_info, fib6_dst),
1689 .addr = daddr,
1690 },
1691#ifdef CONFIG_IPV6_SUBTREES
1692 {
1693 .offset = offsetof(struct fib6_info, fib6_src),
1694 .addr = saddr,
1695 },
1696#endif
1697 {
1698 .offset = 0, /* sentinel */
1699 }
1700 };
1701
1702 fn = fib6_node_lookup_1(root, daddr ? args : args + 1);
1703 if (!fn || fn->fn_flags & RTN_TL_ROOT)
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)
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.
1718 */
1719
1720
1721static struct fib6_node *fib6_locate_1(struct fib6_node *root,
1722 const struct in6_addr *addr,
1723 int plen, int offset,
1724 bool exact_match)
1725{
1726 struct fib6_node *fn, *prev = NULL;
1727
1728 for (fn = root; fn ; ) {
1729 struct fib6_info *leaf = rcu_dereference(fn->leaf);
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);
1741
1742 /*
1743 * Prefix match
1744 */
1745 if (plen < fn->fn_bit ||
1746 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
1747 goto out;
1748
1749 if (plen == fn->fn_bit)
1750 return fn;
1751
1752 if (fn->fn_flags & RTN_RTINFO)
1753 prev = fn;
1754
1755next:
1756 /*
1757 * We have more bits to go
1758 */
1759 if (addr_bit_set(addr, fn->fn_bit))
1760 fn = rcu_dereference(fn->right);
1761 else
1762 fn = rcu_dereference(fn->left);
1763 }
1764out:
1765 if (exact_match)
1766 return NULL;
1767 else
1768 return prev;
1769}
1770
1771struct fib6_node *fib6_locate(struct fib6_node *root,
1772 const struct in6_addr *daddr, int dst_len,
1773 const struct in6_addr *saddr, int src_len,
1774 bool exact_match)
1775{
1776 struct fib6_node *fn;
1777
1778 fn = fib6_locate_1(root, daddr, dst_len,
1779 offsetof(struct fib6_info, fib6_dst),
1780 exact_match);
1781
1782#ifdef CONFIG_IPV6_SUBTREES
1783 if (src_len) {
1784 WARN_ON(saddr == NULL);
1785 if (fn) {
1786 struct fib6_node *subtree = FIB6_SUBTREE(fn);
1787
1788 if (subtree) {
1789 fn = fib6_locate_1(subtree, saddr, src_len,
1790 offsetof(struct fib6_info, fib6_src),
1791 exact_match);
1792 }
1793 }
1794 }
1795#endif
1796
1797 if (fn && fn->fn_flags & RTN_RTINFO)
1798 return fn;
1799
1800 return NULL;
1801}
1802
1803
1804/*
1805 * Deletion
1806 *
1807 */
1808
1809static struct fib6_info *fib6_find_prefix(struct net *net,
1810 struct fib6_table *table,
1811 struct fib6_node *fn)
1812{
1813 struct fib6_node *child_left, *child_right;
1814
1815 if (fn->fn_flags & RTN_ROOT)
1816 return net->ipv6.fib6_null_entry;
1817
1818 while (fn) {
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));
1829
1830 fn = FIB6_SUBTREE(fn);
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.
1838 * Need to own table->tb6_lock
1839 */
1840
1841static struct fib6_node *fib6_repair_tree(struct net *net,
1842 struct fib6_table *table,
1843 struct fib6_node *fn)
1844{
1845 int children;
1846 int nstate;
1847 struct fib6_node *child;
1848 struct fib6_walker *w;
1849 int iter = 0;
1850
1851 /* Set fn->leaf to null_entry for root node. */
1852 if (fn->fn_flags & RTN_TL_ROOT) {
1853 rcu_assign_pointer(fn->leaf, net->ipv6.fib6_null_entry);
1854 return fn;
1855 }
1856
1857 for (;;) {
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));
1868 struct fib6_info *fn_leaf = rcu_dereference_protected(fn->leaf,
1869 lockdep_is_held(&table->tb6_lock));
1870 struct fib6_info *pn_leaf = rcu_dereference_protected(pn->leaf,
1871 lockdep_is_held(&table->tb6_lock));
1872 struct fib6_info *new_fn_leaf;
1873
1874 pr_debug("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1875 iter++;
1876
1877 WARN_ON(fn->fn_flags & RTN_RTINFO);
1878 WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1879 WARN_ON(fn_leaf);
1880
1881 children = 0;
1882 child = NULL;
1883 if (fn_r) {
1884 child = fn_r;
1885 children |= 1;
1886 }
1887 if (fn_l) {
1888 child = fn_l;
1889 children |= 2;
1890 }
1891
1892 if (children == 3 || FIB6_SUBTREE(fn)
1893#ifdef CONFIG_IPV6_SUBTREES
1894 /* Subtree root (i.e. fn) may have one child */
1895 || (children && fn->fn_flags & RTN_ROOT)
1896#endif
1897 ) {
1898 new_fn_leaf = fib6_find_prefix(net, table, fn);
1899#if RT6_DEBUG >= 2
1900 if (!new_fn_leaf) {
1901 WARN_ON(!new_fn_leaf);
1902 new_fn_leaf = net->ipv6.fib6_null_entry;
1903 }
1904#endif
1905 fib6_info_hold(new_fn_leaf);
1906 rcu_assign_pointer(fn->leaf, new_fn_leaf);
1907 return pn;
1908 }
1909
1910#ifdef CONFIG_IPV6_SUBTREES
1911 if (FIB6_SUBTREE(pn) == fn) {
1912 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1913 RCU_INIT_POINTER(pn->subtree, NULL);
1914 nstate = FWS_L;
1915 } else {
1916 WARN_ON(fn->fn_flags & RTN_ROOT);
1917#endif
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);
1922#if RT6_DEBUG >= 2
1923 else
1924 WARN_ON(1);
1925#endif
1926 if (child)
1927 rcu_assign_pointer(child->parent, pn);
1928 nstate = FWS_R;
1929#ifdef CONFIG_IPV6_SUBTREES
1930 }
1931#endif
1932
1933 read_lock(&net->ipv6.fib6_walker_lock);
1934 FOR_WALKERS(net, w) {
1935 if (!child) {
1936 if (w->node == fn) {
1937 pr_debug("W %p adjusted by delnode 1, s=%d/%d\n",
1938 w, w->state, nstate);
1939 w->node = pn;
1940 w->state = nstate;
1941 }
1942 } else {
1943 if (w->node == fn) {
1944 w->node = child;
1945 if (children&2) {
1946 pr_debug("W %p adjusted by delnode 2, s=%d\n",
1947 w, w->state);
1948 w->state = w->state >= FWS_R ? FWS_U : FWS_INIT;
1949 } else {
1950 pr_debug("W %p adjusted by delnode 2, s=%d\n",
1951 w, w->state);
1952 w->state = w->state >= FWS_C ? FWS_U : FWS_INIT;
1953 }
1954 }
1955 }
1956 }
1957 read_unlock(&net->ipv6.fib6_walker_lock);
1958
1959 node_free(net, fn);
1960 if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn))
1961 return pn;
1962
1963 RCU_INIT_POINTER(pn->leaf, NULL);
1964 fib6_info_release(pn_leaf);
1965 fn = pn;
1966 }
1967}
1968
1969static void fib6_del_route(struct fib6_table *table, struct fib6_node *fn,
1970 struct fib6_info __rcu **rtp, struct nl_info *info,
1971 enum rt_del_reason del_reason)
1972{
1973 struct fib6_info *leaf, *replace_rt = NULL;
1974 struct fib6_walker *w;
1975 struct fib6_info *rt = rcu_dereference_protected(*rtp,
1976 lockdep_is_held(&table->tb6_lock));
1977 struct net *net = info->nl_net;
1978 bool notify_del = false;
1979
1980 /* If the deleted route is the first in the node and it is not part of
1981 * a multipath route, then we need to replace it with the next route
1982 * in the node, if exists.
1983 */
1984 leaf = rcu_dereference_protected(fn->leaf,
1985 lockdep_is_held(&table->tb6_lock));
1986 if (leaf == rt && !rt->fib6_nsiblings) {
1987 if (rcu_access_pointer(rt->fib6_next))
1988 replace_rt = rcu_dereference_protected(rt->fib6_next,
1989 lockdep_is_held(&table->tb6_lock));
1990 else
1991 notify_del = true;
1992 }
1993
1994 /* Unlink it */
1995 *rtp = rt->fib6_next;
1996 rt->fib6_node = NULL;
1997 net->ipv6.rt6_stats->fib_rt_entries--;
1998 net->ipv6.rt6_stats->fib_discarded_routes++;
1999
2000 /* Reset round-robin state, if necessary */
2001 if (rcu_access_pointer(fn->rr_ptr) == rt)
2002 fn->rr_ptr = NULL;
2003
2004 /* Remove this entry from other siblings */
2005 if (rt->fib6_nsiblings) {
2006 struct fib6_info *sibling, *next_sibling;
2007
2008 /* The route is deleted from a multipath route. If this
2009 * multipath route is the first route in the node, then we need
2010 * to emit a delete notification. Otherwise, we need to skip
2011 * the notification.
2012 */
2013 if (rt->fib6_metric == leaf->fib6_metric &&
2014 rt6_qualify_for_ecmp(leaf))
2015 notify_del = true;
2016 list_for_each_entry_safe(sibling, next_sibling,
2017 &rt->fib6_siblings, fib6_siblings)
2018 WRITE_ONCE(sibling->fib6_nsiblings,
2019 sibling->fib6_nsiblings - 1);
2020 WRITE_ONCE(rt->fib6_nsiblings, 0);
2021 list_del_rcu(&rt->fib6_siblings);
2022 rt6_multipath_rebalance(next_sibling);
2023 }
2024
2025 /* Adjust walkers */
2026 read_lock(&net->ipv6.fib6_walker_lock);
2027 FOR_WALKERS(net, w) {
2028 if (w->state == FWS_C && w->leaf == rt) {
2029 pr_debug("walker %p adjusted by delroute\n", w);
2030 w->leaf = rcu_dereference_protected(rt->fib6_next,
2031 lockdep_is_held(&table->tb6_lock));
2032 if (!w->leaf)
2033 w->state = FWS_U;
2034 }
2035 }
2036 read_unlock(&net->ipv6.fib6_walker_lock);
2037
2038 /* If it was last route, call fib6_repair_tree() to:
2039 * 1. For root node, put back null_entry as how the table was created.
2040 * 2. For other nodes, expunge its radix tree node.
2041 */
2042 if (!rcu_access_pointer(fn->leaf)) {
2043 if (!(fn->fn_flags & RTN_TL_ROOT)) {
2044 fn->fn_flags &= ~RTN_RTINFO;
2045 net->ipv6.rt6_stats->fib_route_nodes--;
2046 }
2047 fn = fib6_repair_tree(net, table, fn);
2048 }
2049
2050 fib6_purge_rt(rt, fn, net);
2051
2052 if (!info->skip_notify_kernel) {
2053 if (notify_del)
2054 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_DEL,
2055 rt, NULL);
2056 else if (replace_rt)
2057 call_fib6_entry_notifiers_replace(net, replace_rt);
2058 }
2059 if (!info->skip_notify)
2060 inet6_rt_del_notify(rt, info, del_reason);
2061
2062 fib6_info_release(rt);
2063}
2064
2065/* Need to own table->tb6_lock */
2066int fib6_del(struct fib6_info *rt, struct nl_info *info,
2067 enum rt_del_reason del_reason)
2068{
2069 struct net *net = info->nl_net;
2070 struct fib6_info __rcu **rtp;
2071 struct fib6_info __rcu **rtp_next;
2072 struct fib6_table *table;
2073 struct fib6_node *fn;
2074
2075 if (rt == net->ipv6.fib6_null_entry)
2076 return -ENOENT;
2077
2078 table = rt->fib6_table;
2079 fn = rcu_dereference_protected(rt->fib6_node,
2080 lockdep_is_held(&table->tb6_lock));
2081 if (!fn)
2082 return -ENOENT;
2083
2084 WARN_ON(!(fn->fn_flags & RTN_RTINFO));
2085
2086 /*
2087 * Walk the leaf entries looking for ourself
2088 */
2089
2090 for (rtp = &fn->leaf; *rtp; rtp = rtp_next) {
2091 struct fib6_info *cur = rcu_dereference_protected(*rtp,
2092 lockdep_is_held(&table->tb6_lock));
2093 if (rt == cur) {
2094 if (fib6_requires_src(cur))
2095 fib6_routes_require_src_dec(info->nl_net);
2096 fib6_del_route(table, fn, rtp, info, del_reason);
2097 return 0;
2098 }
2099 rtp_next = &cur->fib6_next;
2100 }
2101 return -ENOENT;
2102}
2103
2104/*
2105 * Tree traversal function.
2106 *
2107 * Certainly, it is not interrupt safe.
2108 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
2109 * It means, that we can modify tree during walking
2110 * and use this function for garbage collection, clone pruning,
2111 * cleaning tree when a device goes down etc. etc.
2112 *
2113 * It guarantees that every node will be traversed,
2114 * and that it will be traversed only once.
2115 *
2116 * Callback function w->func may return:
2117 * 0 -> continue walking.
2118 * positive value -> walking is suspended (used by tree dumps,
2119 * and probably by gc, if it will be split to several slices)
2120 * negative value -> terminate walking.
2121 *
2122 * The function itself returns:
2123 * 0 -> walk is complete.
2124 * >0 -> walk is incomplete (i.e. suspended)
2125 * <0 -> walk is terminated by an error.
2126 *
2127 * This function is called with tb6_lock held.
2128 */
2129
2130static int fib6_walk_continue(struct fib6_walker *w)
2131{
2132 struct fib6_node *fn, *pn, *left, *right;
2133
2134 /* w->root should always be table->tb6_root */
2135 WARN_ON_ONCE(!(w->root->fn_flags & RTN_TL_ROOT));
2136
2137 for (;;) {
2138 fn = w->node;
2139 if (!fn)
2140 return 0;
2141
2142 switch (w->state) {
2143#ifdef CONFIG_IPV6_SUBTREES
2144 case FWS_S:
2145 if (FIB6_SUBTREE(fn)) {
2146 w->node = FIB6_SUBTREE(fn);
2147 continue;
2148 }
2149 w->state = FWS_L;
2150 fallthrough;
2151#endif
2152 case FWS_L:
2153 left = rcu_dereference_protected(fn->left, 1);
2154 if (left) {
2155 w->node = left;
2156 w->state = FWS_INIT;
2157 continue;
2158 }
2159 w->state = FWS_R;
2160 fallthrough;
2161 case FWS_R:
2162 right = rcu_dereference_protected(fn->right, 1);
2163 if (right) {
2164 w->node = right;
2165 w->state = FWS_INIT;
2166 continue;
2167 }
2168 w->state = FWS_C;
2169 w->leaf = rcu_dereference_protected(fn->leaf, 1);
2170 fallthrough;
2171 case FWS_C:
2172 if (w->leaf && fn->fn_flags & RTN_RTINFO) {
2173 int err;
2174
2175 if (w->skip) {
2176 w->skip--;
2177 goto skip;
2178 }
2179
2180 err = w->func(w);
2181 if (err)
2182 return err;
2183
2184 w->count++;
2185 continue;
2186 }
2187skip:
2188 w->state = FWS_U;
2189 fallthrough;
2190 case FWS_U:
2191 if (fn == w->root)
2192 return 0;
2193 pn = rcu_dereference_protected(fn->parent, 1);
2194 left = rcu_dereference_protected(pn->left, 1);
2195 right = rcu_dereference_protected(pn->right, 1);
2196 w->node = pn;
2197#ifdef CONFIG_IPV6_SUBTREES
2198 if (FIB6_SUBTREE(pn) == fn) {
2199 WARN_ON(!(fn->fn_flags & RTN_ROOT));
2200 w->state = FWS_L;
2201 continue;
2202 }
2203#endif
2204 if (left == fn) {
2205 w->state = FWS_R;
2206 continue;
2207 }
2208 if (right == fn) {
2209 w->state = FWS_C;
2210 w->leaf = rcu_dereference_protected(w->node->leaf, 1);
2211 continue;
2212 }
2213#if RT6_DEBUG >= 2
2214 WARN_ON(1);
2215#endif
2216 }
2217 }
2218}
2219
2220static int fib6_walk(struct net *net, struct fib6_walker *w)
2221{
2222 int res;
2223
2224 w->state = FWS_INIT;
2225 w->node = w->root;
2226
2227 fib6_walker_link(net, w);
2228 res = fib6_walk_continue(w);
2229 if (res <= 0)
2230 fib6_walker_unlink(net, w);
2231 return res;
2232}
2233
2234static int fib6_clean_node(struct fib6_walker *w)
2235{
2236 int res;
2237 struct fib6_info *rt;
2238 struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w);
2239 struct nl_info info = {
2240 .nl_net = c->net,
2241 .skip_notify = c->skip_notify,
2242 };
2243
2244 if (c->sernum != FIB6_NO_SERNUM_CHANGE &&
2245 READ_ONCE(w->node->fn_sernum) != c->sernum)
2246 WRITE_ONCE(w->node->fn_sernum, c->sernum);
2247
2248 if (!c->func) {
2249 WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE);
2250 w->leaf = NULL;
2251 return 0;
2252 }
2253
2254 for_each_fib6_walker_rt(w) {
2255 res = c->func(rt, c->arg);
2256 if (res == -1) {
2257 w->leaf = rt;
2258 res = fib6_del(rt, &info, RT_DEL_REASON_UNSPEC);
2259 if (res) {
2260#if RT6_DEBUG >= 2
2261 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
2262 __func__, rt,
2263 rcu_access_pointer(rt->fib6_node),
2264 res);
2265#endif
2266 continue;
2267 }
2268 return 0;
2269 } else if (res == -2) {
2270 if (WARN_ON(!rt->fib6_nsiblings))
2271 continue;
2272 rt = list_last_entry(&rt->fib6_siblings,
2273 struct fib6_info, fib6_siblings);
2274 continue;
2275 }
2276 WARN_ON(res != 0);
2277 }
2278 w->leaf = rt;
2279 return 0;
2280}
2281
2282/*
2283 * Convenient frontend to tree walker.
2284 *
2285 * func is called on each route.
2286 * It may return -2 -> skip multipath route.
2287 * -1 -> delete this route.
2288 * 0 -> continue walking
2289 */
2290
2291static void fib6_clean_tree(struct net *net, struct fib6_node *root,
2292 int (*func)(struct fib6_info *, void *arg),
2293 int sernum, void *arg, bool skip_notify)
2294{
2295 struct fib6_cleaner c;
2296
2297 c.w.root = root;
2298 c.w.func = fib6_clean_node;
2299 c.w.count = 0;
2300 c.w.skip = 0;
2301 c.w.skip_in_node = 0;
2302 c.func = func;
2303 c.sernum = sernum;
2304 c.arg = arg;
2305 c.net = net;
2306 c.skip_notify = skip_notify;
2307
2308 fib6_walk(net, &c.w);
2309}
2310
2311static void __fib6_clean_all(struct net *net,
2312 int (*func)(struct fib6_info *, void *),
2313 int sernum, void *arg, bool skip_notify)
2314{
2315 struct fib6_table *table;
2316 struct hlist_head *head;
2317 unsigned int h;
2318
2319 rcu_read_lock();
2320 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2321 head = &net->ipv6.fib_table_hash[h];
2322 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2323 spin_lock_bh(&table->tb6_lock);
2324 fib6_clean_tree(net, &table->tb6_root,
2325 func, sernum, arg, skip_notify);
2326 spin_unlock_bh(&table->tb6_lock);
2327 }
2328 }
2329 rcu_read_unlock();
2330}
2331
2332void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *),
2333 void *arg)
2334{
2335 __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, false);
2336}
2337
2338void fib6_clean_all_skip_notify(struct net *net,
2339 int (*func)(struct fib6_info *, void *),
2340 void *arg)
2341{
2342 __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, true);
2343}
2344
2345static void fib6_flush_trees(struct net *net)
2346{
2347 int new_sernum = fib6_new_sernum(net);
2348
2349 __fib6_clean_all(net, NULL, new_sernum, NULL, false);
2350}
2351
2352/*
2353 * Garbage collection
2354 */
2355void fib6_age_exceptions(struct fib6_info *rt, struct fib6_gc_args *gc_args,
2356 unsigned long now)
2357{
2358 bool may_expire = rt->fib6_flags & RTF_EXPIRES && rt->expires;
2359 int old_more = gc_args->more;
2360
2361 rt6_age_exceptions(rt, gc_args, now);
2362
2363 if (!may_expire && old_more == gc_args->more)
2364 fib6_remove_gc_list(rt);
2365}
2366
2367static int fib6_age(struct fib6_info *rt, struct fib6_gc_args *gc_args)
2368{
2369 unsigned long now = jiffies;
2370
2371 /*
2372 * check addrconf expiration here.
2373 * Routes are expired even if they are in use.
2374 */
2375
2376 if (rt->fib6_flags & RTF_EXPIRES && rt->expires) {
2377 if (time_after(now, rt->expires)) {
2378 pr_debug("expiring %p\n", rt);
2379 return -1;
2380 }
2381 gc_args->more++;
2382 }
2383
2384 /* Also age clones in the exception table.
2385 * Note, that clones are aged out
2386 * only if they are not in use now.
2387 */
2388 fib6_age_exceptions(rt, gc_args, now);
2389
2390 return 0;
2391}
2392
2393static void fib6_gc_table(struct net *net,
2394 struct fib6_table *tb6,
2395 struct fib6_gc_args *gc_args)
2396{
2397 struct fib6_info *rt;
2398 struct hlist_node *n;
2399 struct nl_info info = {
2400 .nl_net = net,
2401 .skip_notify = false,
2402 };
2403
2404 hlist_for_each_entry_safe(rt, n, &tb6->tb6_gc_hlist, gc_link)
2405 if (fib6_age(rt, gc_args) == -1)
2406 fib6_del(rt, &info, RT_DEL_REASON_EXPIRED);
2407}
2408
2409static void fib6_gc_all(struct net *net, struct fib6_gc_args *gc_args)
2410{
2411 struct fib6_table *table;
2412 struct hlist_head *head;
2413 unsigned int h;
2414
2415 rcu_read_lock();
2416 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2417 head = &net->ipv6.fib_table_hash[h];
2418 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2419 spin_lock_bh(&table->tb6_lock);
2420
2421 fib6_gc_table(net, table, gc_args);
2422
2423 spin_unlock_bh(&table->tb6_lock);
2424 }
2425 }
2426 rcu_read_unlock();
2427}
2428
2429void fib6_run_gc(unsigned long expires, struct net *net, bool force)
2430{
2431 struct fib6_gc_args gc_args;
2432 int ip6_rt_gc_interval;
2433 unsigned long now;
2434
2435 if (force) {
2436 spin_lock_bh(&net->ipv6.fib6_gc_lock);
2437 } else if (!spin_trylock_bh(&net->ipv6.fib6_gc_lock)) {
2438 mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
2439 return;
2440 }
2441 ip6_rt_gc_interval = READ_ONCE(net->ipv6.sysctl.ip6_rt_gc_interval);
2442 gc_args.timeout = expires ? (int)expires : ip6_rt_gc_interval;
2443 gc_args.more = 0;
2444
2445 fib6_gc_all(net, &gc_args);
2446 now = jiffies;
2447 net->ipv6.ip6_rt_last_gc = now;
2448
2449 if (gc_args.more)
2450 mod_timer(&net->ipv6.ip6_fib_timer,
2451 round_jiffies(now + ip6_rt_gc_interval));
2452 else
2453 timer_delete(&net->ipv6.ip6_fib_timer);
2454 spin_unlock_bh(&net->ipv6.fib6_gc_lock);
2455}
2456
2457static void fib6_gc_timer_cb(struct timer_list *t)
2458{
2459 struct net *arg = timer_container_of(arg, t, ipv6.ip6_fib_timer);
2460
2461 fib6_run_gc(0, arg, true);
2462}
2463
2464static int __net_init fib6_net_init(struct net *net)
2465{
2466 size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ;
2467 int err;
2468
2469 err = fib6_notifier_init(net);
2470 if (err)
2471 return err;
2472
2473 /* Default to 3-tuple */
2474 net->ipv6.sysctl.multipath_hash_fields =
2475 FIB_MULTIPATH_HASH_FIELD_DEFAULT_MASK;
2476
2477 spin_lock_init(&net->ipv6.fib6_gc_lock);
2478 rwlock_init(&net->ipv6.fib6_walker_lock);
2479 INIT_LIST_HEAD(&net->ipv6.fib6_walkers);
2480 timer_setup(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, 0);
2481
2482 net->ipv6.rt6_stats = kzalloc_obj(*net->ipv6.rt6_stats);
2483 if (!net->ipv6.rt6_stats)
2484 goto out_notifier;
2485
2486 /* Avoid false sharing : Use at least a full cache line */
2487 size = max_t(size_t, size, L1_CACHE_BYTES);
2488
2489 net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL);
2490 if (!net->ipv6.fib_table_hash)
2491 goto out_rt6_stats;
2492
2493 spin_lock_init(&net->ipv6.fib_table_hash_lock);
2494
2495 net->ipv6.fib6_main_tbl = kzalloc_obj(*net->ipv6.fib6_main_tbl);
2496 if (!net->ipv6.fib6_main_tbl)
2497 goto out_fib_table_hash;
2498
2499 net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
2500 rcu_assign_pointer(net->ipv6.fib6_main_tbl->tb6_root.leaf,
2501 net->ipv6.fib6_null_entry);
2502 net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
2503 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
2504 inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers);
2505 INIT_HLIST_HEAD(&net->ipv6.fib6_main_tbl->tb6_gc_hlist);
2506
2507#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2508 net->ipv6.fib6_local_tbl = kzalloc_obj(*net->ipv6.fib6_local_tbl);
2509 if (!net->ipv6.fib6_local_tbl)
2510 goto out_fib6_main_tbl;
2511 net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
2512 rcu_assign_pointer(net->ipv6.fib6_local_tbl->tb6_root.leaf,
2513 net->ipv6.fib6_null_entry);
2514 net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
2515 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
2516 inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers);
2517 INIT_HLIST_HEAD(&net->ipv6.fib6_local_tbl->tb6_gc_hlist);
2518#endif
2519 fib6_tables_init(net);
2520
2521 return 0;
2522
2523#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2524out_fib6_main_tbl:
2525 kfree(net->ipv6.fib6_main_tbl);
2526#endif
2527out_fib_table_hash:
2528 kfree(net->ipv6.fib_table_hash);
2529out_rt6_stats:
2530 kfree(net->ipv6.rt6_stats);
2531out_notifier:
2532 fib6_notifier_exit(net);
2533 return -ENOMEM;
2534}
2535
2536static void fib6_net_exit(struct net *net)
2537{
2538 unsigned int i;
2539
2540 timer_delete_sync(&net->ipv6.ip6_fib_timer);
2541
2542 for (i = 0; i < FIB6_TABLE_HASHSZ; i++) {
2543 struct hlist_head *head = &net->ipv6.fib_table_hash[i];
2544 struct hlist_node *tmp;
2545 struct fib6_table *tb;
2546
2547 hlist_for_each_entry_safe(tb, tmp, head, tb6_hlist) {
2548 hlist_del(&tb->tb6_hlist);
2549 fib6_free_table(tb);
2550 }
2551 }
2552
2553 kfree(net->ipv6.fib_table_hash);
2554 kfree(net->ipv6.rt6_stats);
2555 fib6_notifier_exit(net);
2556}
2557
2558static struct pernet_operations fib6_net_ops = {
2559 .init = fib6_net_init,
2560 .exit = fib6_net_exit,
2561};
2562
2563static const struct rtnl_msg_handler fib6_rtnl_msg_handlers[] __initconst_or_module = {
2564 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETROUTE,
2565 .dumpit = inet6_dump_fib,
2566 .flags = RTNL_FLAG_DUMP_UNLOCKED | RTNL_FLAG_DUMP_SPLIT_NLM_DONE},
2567};
2568
2569int __init fib6_init(void)
2570{
2571 int ret = -ENOMEM;
2572
2573 fib6_node_kmem = KMEM_CACHE(fib6_node,
2574 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT);
2575 if (!fib6_node_kmem)
2576 goto out;
2577
2578 ret = register_pernet_subsys(&fib6_net_ops);
2579 if (ret)
2580 goto out_kmem_cache_create;
2581
2582 ret = rtnl_register_many(fib6_rtnl_msg_handlers);
2583 if (ret)
2584 goto out_unregister_subsys;
2585
2586 __fib6_flush_trees = fib6_flush_trees;
2587out:
2588 return ret;
2589
2590out_unregister_subsys:
2591 unregister_pernet_subsys(&fib6_net_ops);
2592out_kmem_cache_create:
2593 kmem_cache_destroy(fib6_node_kmem);
2594 goto out;
2595}
2596
2597void fib6_gc_cleanup(void)
2598{
2599 unregister_pernet_subsys(&fib6_net_ops);
2600 kmem_cache_destroy(fib6_node_kmem);
2601}
2602
2603#ifdef CONFIG_PROC_FS
2604static int ipv6_route_native_seq_show(struct seq_file *seq, void *v)
2605{
2606 struct fib6_info *rt = v;
2607 struct ipv6_route_iter *iter = seq->private;
2608 struct fib6_nh *fib6_nh = rt->fib6_nh;
2609 unsigned int flags = rt->fib6_flags;
2610 const struct net_device *dev;
2611
2612 if (rt->nh)
2613 fib6_nh = nexthop_fib6_nh(rt->nh);
2614
2615 seq_printf(seq, "%pi6 %02x ", &rt->fib6_dst.addr, rt->fib6_dst.plen);
2616
2617#ifdef CONFIG_IPV6_SUBTREES
2618 seq_printf(seq, "%pi6 %02x ", &rt->fib6_src.addr, rt->fib6_src.plen);
2619#else
2620 seq_puts(seq, "00000000000000000000000000000000 00 ");
2621#endif
2622 if (fib6_nh->fib_nh_gw_family) {
2623 flags |= RTF_GATEWAY;
2624 seq_printf(seq, "%pi6", &fib6_nh->fib_nh_gw6);
2625 } else {
2626 seq_puts(seq, "00000000000000000000000000000000");
2627 }
2628
2629 dev = fib6_nh->fib_nh_dev;
2630 seq_printf(seq, " %08x %08x %08x %08x %8s\n",
2631 rt->fib6_metric, refcount_read(&rt->fib6_ref), 0,
2632 flags, dev ? dev->name : "");
2633 iter->w.leaf = NULL;
2634 return 0;
2635}
2636
2637static int ipv6_route_yield(struct fib6_walker *w)
2638{
2639 struct ipv6_route_iter *iter = w->args;
2640
2641 if (!iter->skip)
2642 return 1;
2643
2644 do {
2645 iter->w.leaf = rcu_dereference_protected(
2646 iter->w.leaf->fib6_next,
2647 lockdep_is_held(&iter->tbl->tb6_lock));
2648 iter->skip--;
2649 if (!iter->skip && iter->w.leaf)
2650 return 1;
2651 } while (iter->w.leaf);
2652
2653 return 0;
2654}
2655
2656static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter,
2657 struct net *net)
2658{
2659 memset(&iter->w, 0, sizeof(iter->w));
2660 iter->w.func = ipv6_route_yield;
2661 iter->w.root = &iter->tbl->tb6_root;
2662 iter->w.state = FWS_INIT;
2663 iter->w.node = iter->w.root;
2664 iter->w.args = iter;
2665 iter->sernum = READ_ONCE(iter->w.root->fn_sernum);
2666 INIT_LIST_HEAD(&iter->w.lh);
2667 fib6_walker_link(net, &iter->w);
2668}
2669
2670static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl,
2671 struct net *net)
2672{
2673 unsigned int h;
2674 struct hlist_node *node;
2675
2676 if (tbl) {
2677 h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1;
2678 node = rcu_dereference(hlist_next_rcu(&tbl->tb6_hlist));
2679 } else {
2680 h = 0;
2681 node = NULL;
2682 }
2683
2684 while (!node && h < FIB6_TABLE_HASHSZ) {
2685 node = rcu_dereference(
2686 hlist_first_rcu(&net->ipv6.fib_table_hash[h++]));
2687 }
2688 return hlist_entry_safe(node, struct fib6_table, tb6_hlist);
2689}
2690
2691static void ipv6_route_check_sernum(struct ipv6_route_iter *iter)
2692{
2693 int sernum = READ_ONCE(iter->w.root->fn_sernum);
2694
2695 if (iter->sernum != sernum) {
2696 iter->sernum = sernum;
2697 iter->w.state = FWS_INIT;
2698 iter->w.node = iter->w.root;
2699 WARN_ON(iter->w.skip);
2700 iter->w.skip = iter->w.count;
2701 }
2702}
2703
2704static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2705{
2706 int r;
2707 struct fib6_info *n;
2708 struct net *net = seq_file_net(seq);
2709 struct ipv6_route_iter *iter = seq->private;
2710
2711 ++(*pos);
2712 if (!v)
2713 goto iter_table;
2714
2715 n = rcu_dereference(((struct fib6_info *)v)->fib6_next);
2716 if (n)
2717 return n;
2718
2719iter_table:
2720 ipv6_route_check_sernum(iter);
2721 spin_lock_bh(&iter->tbl->tb6_lock);
2722 r = fib6_walk_continue(&iter->w);
2723 spin_unlock_bh(&iter->tbl->tb6_lock);
2724 if (r > 0) {
2725 return iter->w.leaf;
2726 } else if (r < 0) {
2727 fib6_walker_unlink(net, &iter->w);
2728 return NULL;
2729 }
2730 fib6_walker_unlink(net, &iter->w);
2731
2732 iter->tbl = ipv6_route_seq_next_table(iter->tbl, net);
2733 if (!iter->tbl)
2734 return NULL;
2735
2736 ipv6_route_seq_setup_walk(iter, net);
2737 goto iter_table;
2738}
2739
2740static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos)
2741 __acquires(RCU)
2742{
2743 struct net *net = seq_file_net(seq);
2744 struct ipv6_route_iter *iter = seq->private;
2745
2746 rcu_read_lock();
2747 iter->tbl = ipv6_route_seq_next_table(NULL, net);
2748 iter->skip = *pos;
2749
2750 if (iter->tbl) {
2751 loff_t p = 0;
2752
2753 ipv6_route_seq_setup_walk(iter, net);
2754 return ipv6_route_seq_next(seq, NULL, &p);
2755 } else {
2756 return NULL;
2757 }
2758}
2759
2760static bool ipv6_route_iter_active(struct ipv6_route_iter *iter)
2761{
2762 struct fib6_walker *w = &iter->w;
2763 return w->node && !list_empty(&w->lh);
2764}
2765
2766static void ipv6_route_native_seq_stop(struct seq_file *seq, void *v)
2767 __releases(RCU)
2768{
2769 struct net *net = seq_file_net(seq);
2770 struct ipv6_route_iter *iter = seq->private;
2771
2772 if (ipv6_route_iter_active(iter))
2773 fib6_walker_unlink(net, &iter->w);
2774
2775 rcu_read_unlock();
2776}
2777
2778#if defined(CONFIG_BPF_SYSCALL)
2779static int ipv6_route_prog_seq_show(struct bpf_prog *prog,
2780 struct bpf_iter_meta *meta,
2781 void *v)
2782{
2783 struct bpf_iter__ipv6_route ctx;
2784
2785 ctx.meta = meta;
2786 ctx.rt = v;
2787 return bpf_iter_run_prog(prog, &ctx);
2788}
2789
2790static int ipv6_route_seq_show(struct seq_file *seq, void *v)
2791{
2792 struct ipv6_route_iter *iter = seq->private;
2793 struct bpf_iter_meta meta;
2794 struct bpf_prog *prog;
2795 int ret;
2796
2797 meta.seq = seq;
2798 prog = bpf_iter_get_info(&meta, false);
2799 if (!prog)
2800 return ipv6_route_native_seq_show(seq, v);
2801
2802 ret = ipv6_route_prog_seq_show(prog, &meta, v);
2803 iter->w.leaf = NULL;
2804
2805 return ret;
2806}
2807
2808static void ipv6_route_seq_stop(struct seq_file *seq, void *v)
2809{
2810 struct bpf_iter_meta meta;
2811 struct bpf_prog *prog;
2812
2813 if (!v) {
2814 meta.seq = seq;
2815 prog = bpf_iter_get_info(&meta, true);
2816 if (prog)
2817 (void)ipv6_route_prog_seq_show(prog, &meta, v);
2818 }
2819
2820 ipv6_route_native_seq_stop(seq, v);
2821}
2822#else
2823static int ipv6_route_seq_show(struct seq_file *seq, void *v)
2824{
2825 return ipv6_route_native_seq_show(seq, v);
2826}
2827
2828static void ipv6_route_seq_stop(struct seq_file *seq, void *v)
2829{
2830 ipv6_route_native_seq_stop(seq, v);
2831}
2832#endif
2833
2834const struct seq_operations ipv6_route_seq_ops = {
2835 .start = ipv6_route_seq_start,
2836 .next = ipv6_route_seq_next,
2837 .stop = ipv6_route_seq_stop,
2838 .show = ipv6_route_seq_show
2839};
2840#endif /* CONFIG_PROC_FS */