1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Linux INET6 implementation
4 * Forwarding Information Database
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
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(new: &w->lh, head: &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(entry: &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(v: &net->ipv6.fib6_sernum);
95
96 do {
97 new = old < INT_MAX ? old + 1 : 1;
98 } while (!atomic_try_cmpxchg(v: &net->ipv6.fib6_sernum, old: &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(list: &f6i->fib6_siblings);
161 refcount_set(r: &f6i->fib6_ref, n: 1);
162
163 INIT_HLIST_NODE(h: &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(nh: f6i->nh);
176 else
177 fib6_nh_release(fib6_nh: f6i->fib6_nh);
178
179 ip_fib_metrics_put(fib_metrics: f6i->fib6_metrics);
180 kfree(objp: 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(s: fib6_node_kmem, objp: 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(objp: 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(n: &tb->tb6_hlist, h: &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(sizeof(*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, table: 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(dst: &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, table: net->ipv6.fib6_main_tbl, oif, fl6,
343 res, strict: flags);
344}
345
346static void __net_init fib6_tables_init(struct net *net)
347{
348 fib6_link_table(net, tb: net->ipv6.fib6_main_tbl);
349}
350
351#endif
352
353unsigned int fib6_tables_seq_read(const struct net *net)
354{
355 unsigned int h, fib_seq = 0;
356
357 rcu_read_lock();
358 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
359 const struct hlist_head *head = &net->ipv6.fib_table_hash[h];
360 const struct fib6_table *tb;
361
362 hlist_for_each_entry_rcu(tb, head, tb6_hlist)
363 fib_seq += READ_ONCE(tb->fib_seq);
364 }
365 rcu_read_unlock();
366
367 return fib_seq;
368}
369
370static int call_fib6_entry_notifier(struct notifier_block *nb,
371 enum fib_event_type event_type,
372 struct fib6_info *rt,
373 struct netlink_ext_ack *extack)
374{
375 struct fib6_entry_notifier_info info = {
376 .info.extack = extack,
377 .rt = rt,
378 };
379
380 return call_fib6_notifier(nb, event_type, info: &info.info);
381}
382
383static 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.info);
396}
397
398int 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)
402{
403 struct fib6_entry_notifier_info info = {
404 .info.extack = extack,
405 .rt = rt,
406 };
407
408 WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1);
409 return call_fib6_notifiers(net, event_type, info: &info.info);
410}
411
412int 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,
422 };
423
424 WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1);
425 return call_fib6_notifiers(net, event_type, info: &info.info);
426}
427
428int 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
435 WRITE_ONCE(rt->fib6_table->fib_seq, rt->fib6_table->fib_seq + 1);
436 return call_fib6_notifiers(net, event_type: FIB_EVENT_ENTRY_REPLACE, info: &info.info);
437}
438
439struct fib6_dump_arg {
440 struct net *net;
441 struct notifier_block *nb;
442 struct netlink_ext_ack *extack;
443};
444
445static int fib6_rt_dump(struct fib6_info *rt, struct fib6_dump_arg *arg)
446{
447 enum fib_event_type fib_event = FIB_EVENT_ENTRY_REPLACE;
448 unsigned int nsiblings;
449 int err;
450
451 if (!rt || rt == arg->net->ipv6.fib6_null_entry)
452 return 0;
453
454 nsiblings = READ_ONCE(rt->fib6_nsiblings);
455 if (nsiblings)
456 err = call_fib6_multipath_entry_notifier(nb: arg->nb, event_type: fib_event,
457 rt,
458 nsiblings,
459 extack: arg->extack);
460 else
461 err = call_fib6_entry_notifier(nb: arg->nb, event_type: fib_event, rt,
462 extack: arg->extack);
463
464 return err;
465}
466
467static int fib6_node_dump(struct fib6_walker *w)
468{
469 int err;
470
471 err = fib6_rt_dump(rt: w->leaf, arg: w->args);
472 w->leaf = NULL;
473 return err;
474}
475
476static int fib6_table_dump(struct net *net, struct fib6_table *tb,
477 struct fib6_walker *w)
478{
479 int err;
480
481 w->root = &tb->tb6_root;
482 spin_lock_bh(lock: &tb->tb6_lock);
483 err = fib6_walk(net, w);
484 spin_unlock_bh(lock: &tb->tb6_lock);
485 return err;
486}
487
488/* Called with rcu_read_lock() */
489int fib6_tables_dump(struct net *net, struct notifier_block *nb,
490 struct netlink_ext_ack *extack)
491{
492 struct fib6_dump_arg arg;
493 struct fib6_walker *w;
494 unsigned int h;
495 int err = 0;
496
497 w = kzalloc(sizeof(*w), GFP_ATOMIC);
498 if (!w)
499 return -ENOMEM;
500
501 w->func = fib6_node_dump;
502 arg.net = net;
503 arg.nb = nb;
504 arg.extack = extack;
505 w->args = &arg;
506
507 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
508 struct hlist_head *head = &net->ipv6.fib_table_hash[h];
509 struct fib6_table *tb;
510
511 hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
512 err = fib6_table_dump(net, tb, w);
513 if (err)
514 goto out;
515 }
516 }
517
518out:
519 kfree(objp: w);
520
521 /* The tree traversal function should never return a positive value. */
522 return err > 0 ? -EINVAL : err;
523}
524
525static int fib6_dump_node(struct fib6_walker *w)
526{
527 int res;
528 struct fib6_info *rt;
529
530 for_each_fib6_walker_rt(w) {
531 res = rt6_dump_route(f6i: rt, p_arg: w->args, skip: w->skip_in_node);
532 if (res >= 0) {
533 /* Frame is full, suspend walking */
534 w->leaf = rt;
535
536 /* We'll restart from this node, so if some routes were
537 * already dumped, skip them next time.
538 */
539 w->skip_in_node += res;
540
541 return 1;
542 }
543 w->skip_in_node = 0;
544
545 /* Multipath routes are dumped in one route with the
546 * RTA_MULTIPATH attribute. Jump 'rt' to point to the
547 * last sibling of this route (no need to dump the
548 * sibling routes again)
549 */
550 if (rt->fib6_nsiblings)
551 rt = list_last_entry(&rt->fib6_siblings,
552 struct fib6_info,
553 fib6_siblings);
554 }
555 w->leaf = NULL;
556 return 0;
557}
558
559static void fib6_dump_end(struct netlink_callback *cb)
560{
561 struct net *net = sock_net(sk: cb->skb->sk);
562 struct fib6_walker *w = (void *)cb->args[2];
563
564 if (w) {
565 if (cb->args[4]) {
566 cb->args[4] = 0;
567 fib6_walker_unlink(net, w);
568 }
569 cb->args[2] = 0;
570 kfree(objp: w);
571 }
572 cb->done = (void *)cb->args[3];
573 cb->args[1] = 3;
574}
575
576static int fib6_dump_done(struct netlink_callback *cb)
577{
578 fib6_dump_end(cb);
579 return cb->done ? cb->done(cb) : 0;
580}
581
582static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
583 struct netlink_callback *cb)
584{
585 struct net *net = sock_net(sk: skb->sk);
586 struct fib6_walker *w;
587 int res;
588
589 w = (void *)cb->args[2];
590 w->root = &table->tb6_root;
591
592 if (cb->args[4] == 0) {
593 w->count = 0;
594 w->skip = 0;
595 w->skip_in_node = 0;
596
597 spin_lock_bh(lock: &table->tb6_lock);
598 res = fib6_walk(net, w);
599 spin_unlock_bh(lock: &table->tb6_lock);
600 if (res > 0) {
601 cb->args[4] = 1;
602 cb->args[5] = READ_ONCE(w->root->fn_sernum);
603 }
604 } else {
605 int sernum = READ_ONCE(w->root->fn_sernum);
606 if (cb->args[5] != sernum) {
607 /* Begin at the root if the tree changed */
608 cb->args[5] = sernum;
609 w->state = FWS_INIT;
610 w->node = w->root;
611 w->skip = w->count;
612 w->skip_in_node = 0;
613 } else
614 w->skip = 0;
615
616 spin_lock_bh(lock: &table->tb6_lock);
617 res = fib6_walk_continue(w);
618 spin_unlock_bh(lock: &table->tb6_lock);
619 if (res <= 0) {
620 fib6_walker_unlink(net, w);
621 cb->args[4] = 0;
622 }
623 }
624
625 return res;
626}
627
628static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
629{
630 struct rt6_rtnl_dump_arg arg = {
631 .filter.dump_exceptions = true,
632 .filter.dump_routes = true,
633 .filter.rtnl_held = false,
634 };
635 const struct nlmsghdr *nlh = cb->nlh;
636 struct net *net = sock_net(sk: skb->sk);
637 unsigned int e = 0, s_e;
638 struct hlist_head *head;
639 struct fib6_walker *w;
640 struct fib6_table *tb;
641 unsigned int h, s_h;
642 int err = 0;
643
644 rcu_read_lock();
645 if (cb->strict_check) {
646 err = ip_valid_fib_dump_req(net, nlh, filter: &arg.filter, cb);
647 if (err < 0)
648 goto unlock;
649 } else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) {
650 struct rtmsg *rtm = nlmsg_data(nlh);
651
652 if (rtm->rtm_flags & RTM_F_PREFIX)
653 arg.filter.flags = RTM_F_PREFIX;
654 }
655
656 w = (void *)cb->args[2];
657 if (!w) {
658 /* New dump:
659 *
660 * 1. allocate and initialize walker.
661 */
662 w = kzalloc(sizeof(*w), GFP_ATOMIC);
663 if (!w) {
664 err = -ENOMEM;
665 goto unlock;
666 }
667 w->func = fib6_dump_node;
668 cb->args[2] = (long)w;
669
670 /* 2. hook callback destructor.
671 */
672 cb->args[3] = (long)cb->done;
673 cb->done = fib6_dump_done;
674
675 }
676
677 arg.skb = skb;
678 arg.cb = cb;
679 arg.net = net;
680 w->args = &arg;
681
682 if (arg.filter.table_id) {
683 tb = fib6_get_table(net, id: arg.filter.table_id);
684 if (!tb) {
685 if (rtnl_msg_family(nlh: cb->nlh) != PF_INET6)
686 goto unlock;
687
688 NL_SET_ERR_MSG_MOD(cb->extack, "FIB table does not exist");
689 err = -ENOENT;
690 goto unlock;
691 }
692
693 if (!cb->args[0]) {
694 err = fib6_dump_table(table: tb, skb, cb);
695 if (!err)
696 cb->args[0] = 1;
697 }
698 goto unlock;
699 }
700
701 s_h = cb->args[0];
702 s_e = cb->args[1];
703
704 for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) {
705 e = 0;
706 head = &net->ipv6.fib_table_hash[h];
707 hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
708 if (e < s_e)
709 goto next;
710 err = fib6_dump_table(table: tb, skb, cb);
711 if (err != 0)
712 goto out;
713next:
714 e++;
715 }
716 }
717out:
718 cb->args[1] = e;
719 cb->args[0] = h;
720
721unlock:
722 rcu_read_unlock();
723 if (err <= 0)
724 fib6_dump_end(cb);
725 return err;
726}
727
728void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val)
729{
730 if (!f6i)
731 return;
732
733 if (f6i->fib6_metrics == &dst_default_metrics) {
734 struct dst_metrics *p = kzalloc(sizeof(*p), GFP_ATOMIC);
735
736 if (!p)
737 return;
738
739 refcount_set(r: &p->refcnt, n: 1);
740 f6i->fib6_metrics = p;
741 }
742
743 f6i->fib6_metrics->metrics[metric - 1] = val;
744}
745
746/*
747 * Routing Table
748 *
749 * return the appropriate node for a routing tree "add" operation
750 * by either creating and inserting or by returning an existing
751 * node.
752 */
753
754static struct fib6_node *fib6_add_1(struct net *net,
755 struct fib6_table *table,
756 struct fib6_node *root,
757 struct in6_addr *addr, int plen,
758 int offset, int allow_create,
759 int replace_required,
760 struct netlink_ext_ack *extack)
761{
762 struct fib6_node *fn, *in, *ln;
763 struct fib6_node *pn = NULL;
764 struct rt6key *key;
765 int bit;
766 __be32 dir = 0;
767
768 /* insert node in tree */
769
770 fn = root;
771
772 do {
773 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
774 lockdep_is_held(&table->tb6_lock));
775 key = (struct rt6key *)((u8 *)leaf + offset);
776
777 /*
778 * Prefix match
779 */
780 if (plen < fn->fn_bit ||
781 !ipv6_prefix_equal(addr1: &key->addr, addr2: addr, prefixlen: fn->fn_bit)) {
782 if (!allow_create) {
783 if (replace_required) {
784 NL_SET_ERR_MSG(extack,
785 "Can not replace route - no match found");
786 pr_warn("Can't replace route, no match found\n");
787 return ERR_PTR(error: -ENOENT);
788 }
789 pr_warn("NLM_F_CREATE should be set when creating new route\n");
790 }
791 goto insert_above;
792 }
793
794 /*
795 * Exact match ?
796 */
797
798 if (plen == fn->fn_bit) {
799 /* clean up an intermediate node */
800 if (!(fn->fn_flags & RTN_RTINFO)) {
801 RCU_INIT_POINTER(fn->leaf, NULL);
802 fib6_info_release(f6i: leaf);
803 /* remove null_entry in the root node */
804 } else if (fn->fn_flags & RTN_TL_ROOT &&
805 rcu_access_pointer(fn->leaf) ==
806 net->ipv6.fib6_null_entry) {
807 RCU_INIT_POINTER(fn->leaf, NULL);
808 }
809
810 return fn;
811 }
812
813 /*
814 * We have more bits to go
815 */
816
817 /* Try to walk down on tree. */
818 dir = addr_bit_set(token: addr, fn_bit: fn->fn_bit);
819 pn = fn;
820 fn = dir ?
821 rcu_dereference_protected(fn->right,
822 lockdep_is_held(&table->tb6_lock)) :
823 rcu_dereference_protected(fn->left,
824 lockdep_is_held(&table->tb6_lock));
825 } while (fn);
826
827 if (!allow_create) {
828 /* We should not create new node because
829 * NLM_F_REPLACE was specified without NLM_F_CREATE
830 * I assume it is safe to require NLM_F_CREATE when
831 * REPLACE flag is used! Later we may want to remove the
832 * check for replace_required, because according
833 * to netlink specification, NLM_F_CREATE
834 * MUST be specified if new route is created.
835 * That would keep IPv6 consistent with IPv4
836 */
837 if (replace_required) {
838 NL_SET_ERR_MSG(extack,
839 "Can not replace route - no match found");
840 pr_warn("Can't replace route, no match found\n");
841 return ERR_PTR(error: -ENOENT);
842 }
843 pr_warn("NLM_F_CREATE should be set when creating new route\n");
844 }
845 /*
846 * We walked to the bottom of tree.
847 * Create new leaf node without children.
848 */
849
850 ln = node_alloc(net);
851
852 if (!ln)
853 return ERR_PTR(error: -ENOMEM);
854 ln->fn_bit = plen;
855 RCU_INIT_POINTER(ln->parent, pn);
856
857 if (dir)
858 rcu_assign_pointer(pn->right, ln);
859 else
860 rcu_assign_pointer(pn->left, ln);
861
862 return ln;
863
864
865insert_above:
866 /*
867 * split since we don't have a common prefix anymore or
868 * we have a less significant route.
869 * we've to insert an intermediate node on the list
870 * this new node will point to the one we need to create
871 * and the current
872 */
873
874 pn = rcu_dereference_protected(fn->parent,
875 lockdep_is_held(&table->tb6_lock));
876
877 /* find 1st bit in difference between the 2 addrs.
878
879 See comment in __ipv6_addr_diff: bit may be an invalid value,
880 but if it is >= plen, the value is ignored in any case.
881 */
882
883 bit = __ipv6_addr_diff(token1: addr, token2: &key->addr, addrlen: sizeof(*addr));
884
885 /*
886 * (intermediate)[in]
887 * / \
888 * (new leaf node)[ln] (old node)[fn]
889 */
890 if (plen > bit) {
891 in = node_alloc(net);
892 ln = node_alloc(net);
893
894 if (!in || !ln) {
895 if (in)
896 node_free_immediate(net, fn: in);
897 if (ln)
898 node_free_immediate(net, fn: ln);
899 return ERR_PTR(error: -ENOMEM);
900 }
901
902 /*
903 * new intermediate node.
904 * RTN_RTINFO will
905 * be off since that an address that chooses one of
906 * the branches would not match less specific routes
907 * in the other branch
908 */
909
910 in->fn_bit = bit;
911
912 RCU_INIT_POINTER(in->parent, pn);
913 in->leaf = fn->leaf;
914 fib6_info_hold(rcu_dereference_protected(in->leaf,
915 lockdep_is_held(&table->tb6_lock)));
916
917 /* update parent pointer */
918 if (dir)
919 rcu_assign_pointer(pn->right, in);
920 else
921 rcu_assign_pointer(pn->left, in);
922
923 ln->fn_bit = plen;
924
925 RCU_INIT_POINTER(ln->parent, in);
926 rcu_assign_pointer(fn->parent, in);
927
928 if (addr_bit_set(token: addr, fn_bit: bit)) {
929 rcu_assign_pointer(in->right, ln);
930 rcu_assign_pointer(in->left, fn);
931 } else {
932 rcu_assign_pointer(in->left, ln);
933 rcu_assign_pointer(in->right, fn);
934 }
935 } else { /* plen <= bit */
936
937 /*
938 * (new leaf node)[ln]
939 * / \
940 * (old node)[fn] NULL
941 */
942
943 ln = node_alloc(net);
944
945 if (!ln)
946 return ERR_PTR(error: -ENOMEM);
947
948 ln->fn_bit = plen;
949
950 RCU_INIT_POINTER(ln->parent, pn);
951
952 if (addr_bit_set(token: &key->addr, fn_bit: plen))
953 RCU_INIT_POINTER(ln->right, fn);
954 else
955 RCU_INIT_POINTER(ln->left, fn);
956
957 rcu_assign_pointer(fn->parent, ln);
958
959 if (dir)
960 rcu_assign_pointer(pn->right, ln);
961 else
962 rcu_assign_pointer(pn->left, ln);
963 }
964 return ln;
965}
966
967static void __fib6_drop_pcpu_from(struct fib6_nh *fib6_nh,
968 const struct fib6_info *match)
969{
970 int cpu;
971
972 if (!fib6_nh->rt6i_pcpu)
973 return;
974
975 rcu_read_lock();
976 /* release the reference to this fib entry from
977 * all of its cached pcpu routes
978 */
979 for_each_possible_cpu(cpu) {
980 struct rt6_info **ppcpu_rt;
981 struct rt6_info *pcpu_rt;
982
983 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
984
985 /* Paired with xchg() in rt6_get_pcpu_route() */
986 pcpu_rt = READ_ONCE(*ppcpu_rt);
987
988 /* only dropping the 'from' reference if the cached route
989 * is using 'match'. The cached pcpu_rt->from only changes
990 * from a fib6_info to NULL (ip6_dst_destroy); it can never
991 * change from one fib6_info reference to another
992 */
993 if (pcpu_rt && rcu_access_pointer(pcpu_rt->from) == match) {
994 struct fib6_info *from;
995
996 from = unrcu_pointer(xchg(&pcpu_rt->from, NULL));
997 fib6_info_release(f6i: from);
998 }
999 }
1000 rcu_read_unlock();
1001}
1002
1003static int fib6_nh_drop_pcpu_from(struct fib6_nh *nh, void *_arg)
1004{
1005 struct fib6_info *arg = _arg;
1006
1007 __fib6_drop_pcpu_from(fib6_nh: nh, match: arg);
1008 return 0;
1009}
1010
1011static void fib6_drop_pcpu_from(struct fib6_info *f6i)
1012{
1013 /* Make sure rt6_make_pcpu_route() wont add other percpu routes
1014 * while we are cleaning them here.
1015 */
1016 f6i->fib6_destroying = 1;
1017 mb(); /* paired with the cmpxchg() in rt6_make_pcpu_route() */
1018
1019 if (f6i->nh) {
1020 rcu_read_lock();
1021 nexthop_for_each_fib6_nh(nh: f6i->nh, cb: fib6_nh_drop_pcpu_from, arg: f6i);
1022 rcu_read_unlock();
1023 } else {
1024 struct fib6_nh *fib6_nh;
1025
1026 fib6_nh = f6i->fib6_nh;
1027 __fib6_drop_pcpu_from(fib6_nh, match: f6i);
1028 }
1029}
1030
1031static void fib6_purge_rt(struct fib6_info *rt, struct fib6_node *fn,
1032 struct net *net)
1033{
1034 struct fib6_table *table = rt->fib6_table;
1035
1036 /* Flush all cached dst in exception table */
1037 rt6_flush_exceptions(f6i: rt);
1038 fib6_drop_pcpu_from(f6i: rt);
1039
1040 if (rt->nh) {
1041 spin_lock(lock: &rt->nh->lock);
1042
1043 if (!list_empty(head: &rt->nh_list))
1044 list_del_init(entry: &rt->nh_list);
1045
1046 spin_unlock(lock: &rt->nh->lock);
1047 }
1048
1049 if (refcount_read(r: &rt->fib6_ref) != 1) {
1050 /* This route is used as dummy address holder in some split
1051 * nodes. It is not leaked, but it still holds other resources,
1052 * which must be released in time. So, scan ascendant nodes
1053 * and replace dummy references to this route with references
1054 * to still alive ones.
1055 */
1056 while (fn) {
1057 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
1058 lockdep_is_held(&table->tb6_lock));
1059 struct fib6_info *new_leaf;
1060 if (!(fn->fn_flags & RTN_RTINFO) && leaf == rt) {
1061 new_leaf = fib6_find_prefix(net, table, fn);
1062 fib6_info_hold(f6i: new_leaf);
1063
1064 rcu_assign_pointer(fn->leaf, new_leaf);
1065 fib6_info_release(f6i: rt);
1066 }
1067 fn = rcu_dereference_protected(fn->parent,
1068 lockdep_is_held(&table->tb6_lock));
1069 }
1070 }
1071
1072 fib6_clean_expires(f6i: rt);
1073 fib6_remove_gc_list(f6i: rt);
1074}
1075
1076/*
1077 * Insert routing information in a node.
1078 */
1079
1080static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
1081 struct nl_info *info, struct netlink_ext_ack *extack,
1082 struct list_head *purge_list)
1083{
1084 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
1085 lockdep_is_held(&rt->fib6_table->tb6_lock));
1086 struct fib6_info *iter = NULL;
1087 struct fib6_info __rcu **ins;
1088 struct fib6_info __rcu **fallback_ins = NULL;
1089 int replace = (info->nlh &&
1090 (info->nlh->nlmsg_flags & NLM_F_REPLACE));
1091 int add = (!info->nlh ||
1092 (info->nlh->nlmsg_flags & NLM_F_CREATE));
1093 int found = 0;
1094 bool rt_can_ecmp = rt6_qualify_for_ecmp(f6i: rt);
1095 bool notify_sibling_rt = false;
1096 u16 nlflags = NLM_F_EXCL;
1097 int err;
1098
1099 if (info->nlh && (info->nlh->nlmsg_flags & NLM_F_APPEND))
1100 nlflags |= NLM_F_APPEND;
1101
1102 ins = &fn->leaf;
1103
1104 for (iter = leaf; iter;
1105 iter = rcu_dereference_protected(iter->fib6_next,
1106 lockdep_is_held(&rt->fib6_table->tb6_lock))) {
1107 /*
1108 * Search for duplicates
1109 */
1110
1111 if (iter->fib6_metric == rt->fib6_metric) {
1112 /*
1113 * Same priority level
1114 */
1115 if (info->nlh &&
1116 (info->nlh->nlmsg_flags & NLM_F_EXCL))
1117 return -EEXIST;
1118
1119 nlflags &= ~NLM_F_EXCL;
1120 if (replace) {
1121 if (rt_can_ecmp == rt6_qualify_for_ecmp(f6i: iter)) {
1122 found++;
1123 break;
1124 }
1125 fallback_ins = fallback_ins ?: ins;
1126 goto next_iter;
1127 }
1128
1129 if (rt6_duplicate_nexthop(a: iter, b: rt)) {
1130 if (rt->fib6_nsiblings)
1131 WRITE_ONCE(rt->fib6_nsiblings, 0);
1132 if (!(iter->fib6_flags & RTF_EXPIRES))
1133 return -EEXIST;
1134 if (!(rt->fib6_flags & RTF_EXPIRES)) {
1135 fib6_clean_expires(f6i: iter);
1136 fib6_remove_gc_list(f6i: iter);
1137 } else {
1138 fib6_set_expires(f6i: iter, expires: rt->expires);
1139 fib6_add_gc_list(f6i: iter);
1140 }
1141
1142 if (rt->fib6_pmtu)
1143 fib6_metric_set(f6i: iter, RTAX_MTU,
1144 val: rt->fib6_pmtu);
1145 return -EEXIST;
1146 }
1147 /* If we have the same destination and the same metric,
1148 * but not the same gateway, then the route we try to
1149 * add is sibling to this route, increment our counter
1150 * of siblings, and later we will add our route to the
1151 * list.
1152 * Only static routes (which don't have flag
1153 * RTF_EXPIRES) are used for ECMPv6.
1154 *
1155 * To avoid long list, we only had siblings if the
1156 * route have a gateway.
1157 */
1158 if (rt_can_ecmp &&
1159 rt6_qualify_for_ecmp(f6i: iter))
1160 WRITE_ONCE(rt->fib6_nsiblings,
1161 rt->fib6_nsiblings + 1);
1162 }
1163
1164 if (iter->fib6_metric > rt->fib6_metric)
1165 break;
1166
1167next_iter:
1168 ins = &iter->fib6_next;
1169 }
1170
1171 if (fallback_ins && !found) {
1172 /* No matching route with same ecmp-able-ness found, replace
1173 * first matching route
1174 */
1175 ins = fallback_ins;
1176 iter = rcu_dereference_protected(*ins,
1177 lockdep_is_held(&rt->fib6_table->tb6_lock));
1178 found++;
1179 }
1180
1181 /* Reset round-robin state, if necessary */
1182 if (ins == &fn->leaf)
1183 fn->rr_ptr = NULL;
1184
1185 /* Link this route to others same route. */
1186 if (rt->fib6_nsiblings) {
1187 unsigned int fib6_nsiblings;
1188 struct fib6_info *sibling, *temp_sibling;
1189
1190 /* Find the first route that have the same metric */
1191 sibling = leaf;
1192 notify_sibling_rt = true;
1193 while (sibling) {
1194 if (sibling->fib6_metric == rt->fib6_metric &&
1195 rt6_qualify_for_ecmp(f6i: sibling)) {
1196 list_add_tail_rcu(new: &rt->fib6_siblings,
1197 head: &sibling->fib6_siblings);
1198 break;
1199 }
1200 sibling = rcu_dereference_protected(sibling->fib6_next,
1201 lockdep_is_held(&rt->fib6_table->tb6_lock));
1202 notify_sibling_rt = false;
1203 }
1204 /* For each sibling in the list, increment the counter of
1205 * siblings. BUG() if counters does not match, list of siblings
1206 * is broken!
1207 */
1208 fib6_nsiblings = 0;
1209 list_for_each_entry_safe(sibling, temp_sibling,
1210 &rt->fib6_siblings, fib6_siblings) {
1211 WRITE_ONCE(sibling->fib6_nsiblings,
1212 sibling->fib6_nsiblings + 1);
1213 BUG_ON(sibling->fib6_nsiblings != rt->fib6_nsiblings);
1214 fib6_nsiblings++;
1215 }
1216 BUG_ON(fib6_nsiblings != rt->fib6_nsiblings);
1217 rcu_read_lock();
1218 rt6_multipath_rebalance(f6i: temp_sibling);
1219 rcu_read_unlock();
1220 }
1221
1222 /*
1223 * insert node
1224 */
1225 if (!replace) {
1226 if (!add)
1227 pr_warn("NLM_F_CREATE should be set when creating new route\n");
1228
1229add:
1230 nlflags |= NLM_F_CREATE;
1231
1232 /* The route should only be notified if it is the first
1233 * route in the node or if it is added as a sibling
1234 * route to the first route in the node.
1235 */
1236 if (!info->skip_notify_kernel &&
1237 (notify_sibling_rt || ins == &fn->leaf)) {
1238 enum fib_event_type fib_event;
1239
1240 if (notify_sibling_rt)
1241 fib_event = FIB_EVENT_ENTRY_APPEND;
1242 else
1243 fib_event = FIB_EVENT_ENTRY_REPLACE;
1244 err = call_fib6_entry_notifiers(net: info->nl_net,
1245 event_type: fib_event, rt,
1246 extack);
1247 if (err) {
1248 struct fib6_info *sibling, *next_sibling;
1249
1250 /* If the route has siblings, then it first
1251 * needs to be unlinked from them.
1252 */
1253 if (!rt->fib6_nsiblings)
1254 return err;
1255
1256 list_for_each_entry_safe(sibling, next_sibling,
1257 &rt->fib6_siblings,
1258 fib6_siblings)
1259 WRITE_ONCE(sibling->fib6_nsiblings,
1260 sibling->fib6_nsiblings - 1);
1261 WRITE_ONCE(rt->fib6_nsiblings, 0);
1262 list_del_rcu(entry: &rt->fib6_siblings);
1263 rcu_read_lock();
1264 rt6_multipath_rebalance(f6i: next_sibling);
1265 rcu_read_unlock();
1266 return err;
1267 }
1268 }
1269
1270 rcu_assign_pointer(rt->fib6_next, iter);
1271 fib6_info_hold(f6i: rt);
1272 rcu_assign_pointer(rt->fib6_node, fn);
1273 rcu_assign_pointer(*ins, rt);
1274 if (!info->skip_notify)
1275 inet6_rt_notify(RTM_NEWROUTE, rt, info, flags: nlflags);
1276 info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
1277
1278 if (!(fn->fn_flags & RTN_RTINFO)) {
1279 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
1280 fn->fn_flags |= RTN_RTINFO;
1281 }
1282
1283 } else {
1284 int nsiblings;
1285
1286 if (!found) {
1287 if (add)
1288 goto add;
1289 pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
1290 return -ENOENT;
1291 }
1292
1293 if (!info->skip_notify_kernel && ins == &fn->leaf) {
1294 err = call_fib6_entry_notifiers(net: info->nl_net,
1295 event_type: FIB_EVENT_ENTRY_REPLACE,
1296 rt, extack);
1297 if (err)
1298 return err;
1299 }
1300
1301 fib6_info_hold(f6i: rt);
1302 rcu_assign_pointer(rt->fib6_node, fn);
1303 rt->fib6_next = iter->fib6_next;
1304 rcu_assign_pointer(*ins, rt);
1305 if (!info->skip_notify)
1306 inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE);
1307 if (!(fn->fn_flags & RTN_RTINFO)) {
1308 info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
1309 fn->fn_flags |= RTN_RTINFO;
1310 }
1311 nsiblings = iter->fib6_nsiblings;
1312 iter->fib6_node = NULL;
1313 list_add(new: &iter->purge_link, head: purge_list);
1314 if (rcu_access_pointer(fn->rr_ptr) == iter)
1315 fn->rr_ptr = NULL;
1316
1317 if (nsiblings) {
1318 /* Replacing an ECMP route, remove all siblings */
1319 ins = &rt->fib6_next;
1320 iter = rcu_dereference_protected(*ins,
1321 lockdep_is_held(&rt->fib6_table->tb6_lock));
1322 while (iter) {
1323 if (iter->fib6_metric > rt->fib6_metric)
1324 break;
1325 if (rt6_qualify_for_ecmp(f6i: iter)) {
1326 *ins = iter->fib6_next;
1327 iter->fib6_node = NULL;
1328 list_add(new: &iter->purge_link, head: purge_list);
1329 if (rcu_access_pointer(fn->rr_ptr) == iter)
1330 fn->rr_ptr = NULL;
1331 nsiblings--;
1332 info->nl_net->ipv6.rt6_stats->fib_rt_entries--;
1333 } else {
1334 ins = &iter->fib6_next;
1335 }
1336 iter = rcu_dereference_protected(*ins,
1337 lockdep_is_held(&rt->fib6_table->tb6_lock));
1338 }
1339 WARN_ON(nsiblings != 0);
1340 }
1341 }
1342
1343 return 0;
1344}
1345
1346static int fib6_add_rt2node_nh(struct fib6_node *fn, struct fib6_info *rt,
1347 struct nl_info *info, struct netlink_ext_ack *extack,
1348 struct list_head *purge_list)
1349{
1350 int err;
1351
1352 spin_lock(lock: &rt->nh->lock);
1353
1354 if (rt->nh->dead) {
1355 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
1356 err = -EINVAL;
1357 } else {
1358 err = fib6_add_rt2node(fn, rt, info, extack, purge_list);
1359 if (!err)
1360 list_add(new: &rt->nh_list, head: &rt->nh->f6i_list);
1361 }
1362
1363 spin_unlock(lock: &rt->nh->lock);
1364
1365 return err;
1366}
1367
1368static void fib6_start_gc(struct net *net, struct fib6_info *rt)
1369{
1370 if (!timer_pending(timer: &net->ipv6.ip6_fib_timer) &&
1371 (rt->fib6_flags & RTF_EXPIRES))
1372 mod_timer(timer: &net->ipv6.ip6_fib_timer,
1373 expires: jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
1374}
1375
1376void fib6_force_start_gc(struct net *net)
1377{
1378 if (!timer_pending(timer: &net->ipv6.ip6_fib_timer))
1379 mod_timer(timer: &net->ipv6.ip6_fib_timer,
1380 expires: jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
1381}
1382
1383static void __fib6_update_sernum_upto_root(struct fib6_info *rt,
1384 int sernum)
1385{
1386 struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node,
1387 lockdep_is_held(&rt->fib6_table->tb6_lock));
1388
1389 /* paired with smp_rmb() in fib6_get_cookie_safe() */
1390 smp_wmb();
1391 while (fn) {
1392 WRITE_ONCE(fn->fn_sernum, sernum);
1393 fn = rcu_dereference_protected(fn->parent,
1394 lockdep_is_held(&rt->fib6_table->tb6_lock));
1395 }
1396}
1397
1398void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt)
1399{
1400 __fib6_update_sernum_upto_root(rt, sernum: fib6_new_sernum(net));
1401}
1402
1403/* allow ipv4 to update sernum via ipv6_stub */
1404void fib6_update_sernum_stub(struct net *net, struct fib6_info *f6i)
1405{
1406 spin_lock_bh(lock: &f6i->fib6_table->tb6_lock);
1407 fib6_update_sernum_upto_root(net, rt: f6i);
1408 spin_unlock_bh(lock: &f6i->fib6_table->tb6_lock);
1409}
1410
1411/*
1412 * Add routing information to the routing tree.
1413 * <destination addr>/<source addr>
1414 * with source addr info in sub-trees
1415 * Need to own table->tb6_lock
1416 */
1417
1418int fib6_add(struct fib6_node *root, struct fib6_info *rt,
1419 struct nl_info *info, struct netlink_ext_ack *extack)
1420{
1421 struct fib6_table *table = rt->fib6_table;
1422 LIST_HEAD(purge_list);
1423 struct fib6_node *fn;
1424#ifdef CONFIG_IPV6_SUBTREES
1425 struct fib6_node *pn = NULL;
1426#endif
1427 int err = -ENOMEM;
1428 int allow_create = 1;
1429 int replace_required = 0;
1430
1431 if (info->nlh) {
1432 if (!(info->nlh->nlmsg_flags & NLM_F_CREATE))
1433 allow_create = 0;
1434 if (info->nlh->nlmsg_flags & NLM_F_REPLACE)
1435 replace_required = 1;
1436 }
1437 if (!allow_create && !replace_required)
1438 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
1439
1440 fn = fib6_add_1(net: info->nl_net, table, root,
1441 addr: &rt->fib6_dst.addr, plen: rt->fib6_dst.plen,
1442 offsetof(struct fib6_info, fib6_dst), allow_create,
1443 replace_required, extack);
1444 if (IS_ERR(ptr: fn)) {
1445 err = PTR_ERR(ptr: fn);
1446 fn = NULL;
1447 goto out;
1448 }
1449
1450#ifdef CONFIG_IPV6_SUBTREES
1451 pn = fn;
1452
1453 if (rt->fib6_src.plen) {
1454 struct fib6_node *sn;
1455
1456 if (!rcu_access_pointer(fn->subtree)) {
1457 struct fib6_node *sfn;
1458
1459 /*
1460 * Create subtree.
1461 *
1462 * fn[main tree]
1463 * |
1464 * sfn[subtree root]
1465 * \
1466 * sn[new leaf node]
1467 */
1468
1469 /* Create subtree root node */
1470 sfn = node_alloc(info->nl_net);
1471 if (!sfn)
1472 goto failure;
1473
1474 fib6_info_hold(info->nl_net->ipv6.fib6_null_entry);
1475 rcu_assign_pointer(sfn->leaf,
1476 info->nl_net->ipv6.fib6_null_entry);
1477 sfn->fn_flags = RTN_ROOT;
1478
1479 /* Now add the first leaf node to new subtree */
1480
1481 sn = fib6_add_1(info->nl_net, table, sfn,
1482 &rt->fib6_src.addr, rt->fib6_src.plen,
1483 offsetof(struct fib6_info, fib6_src),
1484 allow_create, replace_required, extack);
1485
1486 if (IS_ERR(sn)) {
1487 /* If it is failed, discard just allocated
1488 root, and then (in failure) stale node
1489 in main tree.
1490 */
1491 node_free_immediate(info->nl_net, sfn);
1492 err = PTR_ERR(sn);
1493 goto failure;
1494 }
1495
1496 /* Now link new subtree to main tree */
1497 rcu_assign_pointer(sfn->parent, fn);
1498 rcu_assign_pointer(fn->subtree, sfn);
1499 } else {
1500 sn = fib6_add_1(info->nl_net, table, FIB6_SUBTREE(fn),
1501 &rt->fib6_src.addr, rt->fib6_src.plen,
1502 offsetof(struct fib6_info, fib6_src),
1503 allow_create, replace_required, extack);
1504
1505 if (IS_ERR(sn)) {
1506 err = PTR_ERR(sn);
1507 goto failure;
1508 }
1509 }
1510
1511 if (!rcu_access_pointer(fn->leaf)) {
1512 if (fn->fn_flags & RTN_TL_ROOT) {
1513 /* put back null_entry for root node */
1514 rcu_assign_pointer(fn->leaf,
1515 info->nl_net->ipv6.fib6_null_entry);
1516 } else {
1517 fib6_info_hold(rt);
1518 rcu_assign_pointer(fn->leaf, rt);
1519 }
1520 }
1521 fn = sn;
1522 }
1523#endif
1524
1525 if (rt->nh)
1526 err = fib6_add_rt2node_nh(fn, rt, info, extack, purge_list: &purge_list);
1527 else
1528 err = fib6_add_rt2node(fn, rt, info, extack, purge_list: &purge_list);
1529 if (!err) {
1530 struct fib6_info *iter, *next;
1531
1532 list_for_each_entry_safe(iter, next, &purge_list, purge_link) {
1533 list_del(entry: &iter->purge_link);
1534 fib6_purge_rt(rt: iter, fn, net: info->nl_net);
1535 fib6_info_release(f6i: iter);
1536 }
1537
1538 __fib6_update_sernum_upto_root(rt, sernum: fib6_new_sernum(net: info->nl_net));
1539
1540 if (rt->fib6_flags & RTF_EXPIRES)
1541 fib6_add_gc_list(f6i: rt);
1542
1543 fib6_start_gc(net: info->nl_net, rt);
1544 }
1545
1546out:
1547 if (err) {
1548#ifdef CONFIG_IPV6_SUBTREES
1549 /*
1550 * If fib6_add_1 has cleared the old leaf pointer in the
1551 * super-tree leaf node we have to find a new one for it.
1552 */
1553 if (pn != fn) {
1554 struct fib6_info *pn_leaf =
1555 rcu_dereference_protected(pn->leaf,
1556 lockdep_is_held(&table->tb6_lock));
1557 if (pn_leaf == rt) {
1558 pn_leaf = NULL;
1559 RCU_INIT_POINTER(pn->leaf, NULL);
1560 fib6_info_release(rt);
1561 }
1562 if (!pn_leaf && !(pn->fn_flags & RTN_RTINFO)) {
1563 pn_leaf = fib6_find_prefix(info->nl_net, table,
1564 pn);
1565 if (!pn_leaf)
1566 pn_leaf =
1567 info->nl_net->ipv6.fib6_null_entry;
1568 fib6_info_hold(pn_leaf);
1569 rcu_assign_pointer(pn->leaf, pn_leaf);
1570 }
1571 }
1572#endif
1573 goto failure;
1574 } else if (fib6_requires_src(rt)) {
1575 fib6_routes_require_src_inc(net: info->nl_net);
1576 }
1577 return err;
1578
1579failure:
1580 /* fn->leaf could be NULL and fib6_repair_tree() needs to be called if:
1581 * 1. fn is an intermediate node and we failed to add the new
1582 * route to it in both subtree creation failure and fib6_add_rt2node()
1583 * failure case.
1584 * 2. fn is the root node in the table and we fail to add the first
1585 * default route to it.
1586 */
1587 if (fn &&
1588 (!(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)) ||
1589 (fn->fn_flags & RTN_TL_ROOT &&
1590 !rcu_access_pointer(fn->leaf))))
1591 fib6_repair_tree(net: info->nl_net, table, fn);
1592 return err;
1593}
1594
1595/*
1596 * Routing tree lookup
1597 *
1598 */
1599
1600struct lookup_args {
1601 int offset; /* key offset on fib6_info */
1602 const struct in6_addr *addr; /* search key */
1603};
1604
1605static struct fib6_node *fib6_node_lookup_1(struct fib6_node *root,
1606 struct lookup_args *args)
1607{
1608 struct fib6_node *fn;
1609 __be32 dir;
1610
1611 if (unlikely(args->offset == 0))
1612 return NULL;
1613
1614 /*
1615 * Descend on a tree
1616 */
1617
1618 fn = root;
1619
1620 for (;;) {
1621 struct fib6_node *next;
1622
1623 dir = addr_bit_set(token: args->addr, fn_bit: fn->fn_bit);
1624
1625 next = dir ? rcu_dereference(fn->right) :
1626 rcu_dereference(fn->left);
1627
1628 if (next) {
1629 fn = next;
1630 continue;
1631 }
1632 break;
1633 }
1634
1635 while (fn) {
1636 struct fib6_node *subtree = FIB6_SUBTREE(fn);
1637
1638 if (subtree || fn->fn_flags & RTN_RTINFO) {
1639 struct fib6_info *leaf = rcu_dereference(fn->leaf);
1640 struct rt6key *key;
1641
1642 if (!leaf)
1643 goto backtrack;
1644
1645 key = (struct rt6key *) ((u8 *)leaf + args->offset);
1646
1647 if (ipv6_prefix_equal(addr1: &key->addr, addr2: args->addr, prefixlen: key->plen)) {
1648#ifdef CONFIG_IPV6_SUBTREES
1649 if (subtree) {
1650 struct fib6_node *sfn;
1651 sfn = fib6_node_lookup_1(subtree,
1652 args + 1);
1653 if (!sfn)
1654 goto backtrack;
1655 fn = sfn;
1656 }
1657#endif
1658 if (fn->fn_flags & RTN_RTINFO)
1659 return fn;
1660 }
1661 }
1662backtrack:
1663 if (fn->fn_flags & RTN_ROOT)
1664 break;
1665
1666 fn = rcu_dereference(fn->parent);
1667 }
1668
1669 return NULL;
1670}
1671
1672/* called with rcu_read_lock() held
1673 */
1674struct fib6_node *fib6_node_lookup(struct fib6_node *root,
1675 const struct in6_addr *daddr,
1676 const struct in6_addr *saddr)
1677{
1678 struct fib6_node *fn;
1679 struct lookup_args args[] = {
1680 {
1681 .offset = offsetof(struct fib6_info, fib6_dst),
1682 .addr = daddr,
1683 },
1684#ifdef CONFIG_IPV6_SUBTREES
1685 {
1686 .offset = offsetof(struct fib6_info, fib6_src),
1687 .addr = saddr,
1688 },
1689#endif
1690 {
1691 .offset = 0, /* sentinel */
1692 }
1693 };
1694
1695 fn = fib6_node_lookup_1(root, args: daddr ? args : args + 1);
1696 if (!fn || fn->fn_flags & RTN_TL_ROOT)
1697 fn = root;
1698
1699 return fn;
1700}
1701
1702/*
1703 * Get node with specified destination prefix (and source prefix,
1704 * if subtrees are used)
1705 * exact_match == true means we try to find fn with exact match of
1706 * the passed in prefix addr
1707 * exact_match == false means we try to find fn with longest prefix
1708 * match of the passed in prefix addr. This is useful for finding fn
1709 * for cached route as it will be stored in the exception table under
1710 * the node with longest prefix length.
1711 */
1712
1713
1714static struct fib6_node *fib6_locate_1(struct fib6_node *root,
1715 const struct in6_addr *addr,
1716 int plen, int offset,
1717 bool exact_match)
1718{
1719 struct fib6_node *fn, *prev = NULL;
1720
1721 for (fn = root; fn ; ) {
1722 struct fib6_info *leaf = rcu_dereference(fn->leaf);
1723 struct rt6key *key;
1724
1725 /* This node is being deleted */
1726 if (!leaf) {
1727 if (plen <= fn->fn_bit)
1728 goto out;
1729 else
1730 goto next;
1731 }
1732
1733 key = (struct rt6key *)((u8 *)leaf + offset);
1734
1735 /*
1736 * Prefix match
1737 */
1738 if (plen < fn->fn_bit ||
1739 !ipv6_prefix_equal(addr1: &key->addr, addr2: addr, prefixlen: fn->fn_bit))
1740 goto out;
1741
1742 if (plen == fn->fn_bit)
1743 return fn;
1744
1745 if (fn->fn_flags & RTN_RTINFO)
1746 prev = fn;
1747
1748next:
1749 /*
1750 * We have more bits to go
1751 */
1752 if (addr_bit_set(token: addr, fn_bit: fn->fn_bit))
1753 fn = rcu_dereference(fn->right);
1754 else
1755 fn = rcu_dereference(fn->left);
1756 }
1757out:
1758 if (exact_match)
1759 return NULL;
1760 else
1761 return prev;
1762}
1763
1764struct fib6_node *fib6_locate(struct fib6_node *root,
1765 const struct in6_addr *daddr, int dst_len,
1766 const struct in6_addr *saddr, int src_len,
1767 bool exact_match)
1768{
1769 struct fib6_node *fn;
1770
1771 fn = fib6_locate_1(root, addr: daddr, plen: dst_len,
1772 offsetof(struct fib6_info, fib6_dst),
1773 exact_match);
1774
1775#ifdef CONFIG_IPV6_SUBTREES
1776 if (src_len) {
1777 WARN_ON(saddr == NULL);
1778 if (fn) {
1779 struct fib6_node *subtree = FIB6_SUBTREE(fn);
1780
1781 if (subtree) {
1782 fn = fib6_locate_1(subtree, saddr, src_len,
1783 offsetof(struct fib6_info, fib6_src),
1784 exact_match);
1785 }
1786 }
1787 }
1788#endif
1789
1790 if (fn && fn->fn_flags & RTN_RTINFO)
1791 return fn;
1792
1793 return NULL;
1794}
1795
1796
1797/*
1798 * Deletion
1799 *
1800 */
1801
1802static struct fib6_info *fib6_find_prefix(struct net *net,
1803 struct fib6_table *table,
1804 struct fib6_node *fn)
1805{
1806 struct fib6_node *child_left, *child_right;
1807
1808 if (fn->fn_flags & RTN_ROOT)
1809 return net->ipv6.fib6_null_entry;
1810
1811 while (fn) {
1812 child_left = rcu_dereference_protected(fn->left,
1813 lockdep_is_held(&table->tb6_lock));
1814 child_right = rcu_dereference_protected(fn->right,
1815 lockdep_is_held(&table->tb6_lock));
1816 if (child_left)
1817 return rcu_dereference_protected(child_left->leaf,
1818 lockdep_is_held(&table->tb6_lock));
1819 if (child_right)
1820 return rcu_dereference_protected(child_right->leaf,
1821 lockdep_is_held(&table->tb6_lock));
1822
1823 fn = FIB6_SUBTREE(fn);
1824 }
1825 return NULL;
1826}
1827
1828/*
1829 * Called to trim the tree of intermediate nodes when possible. "fn"
1830 * is the node we want to try and remove.
1831 * Need to own table->tb6_lock
1832 */
1833
1834static struct fib6_node *fib6_repair_tree(struct net *net,
1835 struct fib6_table *table,
1836 struct fib6_node *fn)
1837{
1838 int children;
1839 int nstate;
1840 struct fib6_node *child;
1841 struct fib6_walker *w;
1842 int iter = 0;
1843
1844 /* Set fn->leaf to null_entry for root node. */
1845 if (fn->fn_flags & RTN_TL_ROOT) {
1846 rcu_assign_pointer(fn->leaf, net->ipv6.fib6_null_entry);
1847 return fn;
1848 }
1849
1850 for (;;) {
1851 struct fib6_node *fn_r = rcu_dereference_protected(fn->right,
1852 lockdep_is_held(&table->tb6_lock));
1853 struct fib6_node *fn_l = rcu_dereference_protected(fn->left,
1854 lockdep_is_held(&table->tb6_lock));
1855 struct fib6_node *pn = rcu_dereference_protected(fn->parent,
1856 lockdep_is_held(&table->tb6_lock));
1857 struct fib6_node *pn_r = rcu_dereference_protected(pn->right,
1858 lockdep_is_held(&table->tb6_lock));
1859 struct fib6_node *pn_l = rcu_dereference_protected(pn->left,
1860 lockdep_is_held(&table->tb6_lock));
1861 struct fib6_info *fn_leaf = rcu_dereference_protected(fn->leaf,
1862 lockdep_is_held(&table->tb6_lock));
1863 struct fib6_info *pn_leaf = rcu_dereference_protected(pn->leaf,
1864 lockdep_is_held(&table->tb6_lock));
1865 struct fib6_info *new_fn_leaf;
1866
1867 pr_debug("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1868 iter++;
1869
1870 WARN_ON(fn->fn_flags & RTN_RTINFO);
1871 WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1872 WARN_ON(fn_leaf);
1873
1874 children = 0;
1875 child = NULL;
1876 if (fn_r) {
1877 child = fn_r;
1878 children |= 1;
1879 }
1880 if (fn_l) {
1881 child = fn_l;
1882 children |= 2;
1883 }
1884
1885 if (children == 3 || FIB6_SUBTREE(fn)
1886#ifdef CONFIG_IPV6_SUBTREES
1887 /* Subtree root (i.e. fn) may have one child */
1888 || (children && fn->fn_flags & RTN_ROOT)
1889#endif
1890 ) {
1891 new_fn_leaf = fib6_find_prefix(net, table, fn);
1892#if RT6_DEBUG >= 2
1893 if (!new_fn_leaf) {
1894 WARN_ON(!new_fn_leaf);
1895 new_fn_leaf = net->ipv6.fib6_null_entry;
1896 }
1897#endif
1898 fib6_info_hold(f6i: new_fn_leaf);
1899 rcu_assign_pointer(fn->leaf, new_fn_leaf);
1900 return pn;
1901 }
1902
1903#ifdef CONFIG_IPV6_SUBTREES
1904 if (FIB6_SUBTREE(pn) == fn) {
1905 WARN_ON(!(fn->fn_flags & RTN_ROOT));
1906 RCU_INIT_POINTER(pn->subtree, NULL);
1907 nstate = FWS_L;
1908 } else {
1909 WARN_ON(fn->fn_flags & RTN_ROOT);
1910#endif
1911 if (pn_r == fn)
1912 rcu_assign_pointer(pn->right, child);
1913 else if (pn_l == fn)
1914 rcu_assign_pointer(pn->left, child);
1915#if RT6_DEBUG >= 2
1916 else
1917 WARN_ON(1);
1918#endif
1919 if (child)
1920 rcu_assign_pointer(child->parent, pn);
1921 nstate = FWS_R;
1922#ifdef CONFIG_IPV6_SUBTREES
1923 }
1924#endif
1925
1926 read_lock(&net->ipv6.fib6_walker_lock);
1927 FOR_WALKERS(net, w) {
1928 if (!child) {
1929 if (w->node == fn) {
1930 pr_debug("W %p adjusted by delnode 1, s=%d/%d\n",
1931 w, w->state, nstate);
1932 w->node = pn;
1933 w->state = nstate;
1934 }
1935 } else {
1936 if (w->node == fn) {
1937 w->node = child;
1938 if (children&2) {
1939 pr_debug("W %p adjusted by delnode 2, s=%d\n",
1940 w, w->state);
1941 w->state = w->state >= FWS_R ? FWS_U : FWS_INIT;
1942 } else {
1943 pr_debug("W %p adjusted by delnode 2, s=%d\n",
1944 w, w->state);
1945 w->state = w->state >= FWS_C ? FWS_U : FWS_INIT;
1946 }
1947 }
1948 }
1949 }
1950 read_unlock(&net->ipv6.fib6_walker_lock);
1951
1952 node_free(net, fn);
1953 if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn))
1954 return pn;
1955
1956 RCU_INIT_POINTER(pn->leaf, NULL);
1957 fib6_info_release(f6i: pn_leaf);
1958 fn = pn;
1959 }
1960}
1961
1962static void fib6_del_route(struct fib6_table *table, struct fib6_node *fn,
1963 struct fib6_info __rcu **rtp, struct nl_info *info)
1964{
1965 struct fib6_info *leaf, *replace_rt = NULL;
1966 struct fib6_walker *w;
1967 struct fib6_info *rt = rcu_dereference_protected(*rtp,
1968 lockdep_is_held(&table->tb6_lock));
1969 struct net *net = info->nl_net;
1970 bool notify_del = false;
1971
1972 /* If the deleted route is the first in the node and it is not part of
1973 * a multipath route, then we need to replace it with the next route
1974 * in the node, if exists.
1975 */
1976 leaf = rcu_dereference_protected(fn->leaf,
1977 lockdep_is_held(&table->tb6_lock));
1978 if (leaf == rt && !rt->fib6_nsiblings) {
1979 if (rcu_access_pointer(rt->fib6_next))
1980 replace_rt = rcu_dereference_protected(rt->fib6_next,
1981 lockdep_is_held(&table->tb6_lock));
1982 else
1983 notify_del = true;
1984 }
1985
1986 /* Unlink it */
1987 *rtp = rt->fib6_next;
1988 rt->fib6_node = NULL;
1989 net->ipv6.rt6_stats->fib_rt_entries--;
1990 net->ipv6.rt6_stats->fib_discarded_routes++;
1991
1992 /* Reset round-robin state, if necessary */
1993 if (rcu_access_pointer(fn->rr_ptr) == rt)
1994 fn->rr_ptr = NULL;
1995
1996 /* Remove this entry from other siblings */
1997 if (rt->fib6_nsiblings) {
1998 struct fib6_info *sibling, *next_sibling;
1999
2000 /* The route is deleted from a multipath route. If this
2001 * multipath route is the first route in the node, then we need
2002 * to emit a delete notification. Otherwise, we need to skip
2003 * the notification.
2004 */
2005 if (rt->fib6_metric == leaf->fib6_metric &&
2006 rt6_qualify_for_ecmp(f6i: leaf))
2007 notify_del = true;
2008 list_for_each_entry_safe(sibling, next_sibling,
2009 &rt->fib6_siblings, fib6_siblings)
2010 WRITE_ONCE(sibling->fib6_nsiblings,
2011 sibling->fib6_nsiblings - 1);
2012 WRITE_ONCE(rt->fib6_nsiblings, 0);
2013 list_del_rcu(entry: &rt->fib6_siblings);
2014 rt6_multipath_rebalance(f6i: next_sibling);
2015 }
2016
2017 /* Adjust walkers */
2018 read_lock(&net->ipv6.fib6_walker_lock);
2019 FOR_WALKERS(net, w) {
2020 if (w->state == FWS_C && w->leaf == rt) {
2021 pr_debug("walker %p adjusted by delroute\n", w);
2022 w->leaf = rcu_dereference_protected(rt->fib6_next,
2023 lockdep_is_held(&table->tb6_lock));
2024 if (!w->leaf)
2025 w->state = FWS_U;
2026 }
2027 }
2028 read_unlock(&net->ipv6.fib6_walker_lock);
2029
2030 /* If it was last route, call fib6_repair_tree() to:
2031 * 1. For root node, put back null_entry as how the table was created.
2032 * 2. For other nodes, expunge its radix tree node.
2033 */
2034 if (!rcu_access_pointer(fn->leaf)) {
2035 if (!(fn->fn_flags & RTN_TL_ROOT)) {
2036 fn->fn_flags &= ~RTN_RTINFO;
2037 net->ipv6.rt6_stats->fib_route_nodes--;
2038 }
2039 fn = fib6_repair_tree(net, table, fn);
2040 }
2041
2042 fib6_purge_rt(rt, fn, net);
2043
2044 if (!info->skip_notify_kernel) {
2045 if (notify_del)
2046 call_fib6_entry_notifiers(net, event_type: FIB_EVENT_ENTRY_DEL,
2047 rt, NULL);
2048 else if (replace_rt)
2049 call_fib6_entry_notifiers_replace(net, rt: replace_rt);
2050 }
2051 if (!info->skip_notify)
2052 inet6_rt_notify(RTM_DELROUTE, rt, info, flags: 0);
2053
2054 fib6_info_release(f6i: rt);
2055}
2056
2057/* Need to own table->tb6_lock */
2058int fib6_del(struct fib6_info *rt, struct nl_info *info)
2059{
2060 struct net *net = info->nl_net;
2061 struct fib6_info __rcu **rtp;
2062 struct fib6_info __rcu **rtp_next;
2063 struct fib6_table *table;
2064 struct fib6_node *fn;
2065
2066 if (rt == net->ipv6.fib6_null_entry)
2067 return -ENOENT;
2068
2069 table = rt->fib6_table;
2070 fn = rcu_dereference_protected(rt->fib6_node,
2071 lockdep_is_held(&table->tb6_lock));
2072 if (!fn)
2073 return -ENOENT;
2074
2075 WARN_ON(!(fn->fn_flags & RTN_RTINFO));
2076
2077 /*
2078 * Walk the leaf entries looking for ourself
2079 */
2080
2081 for (rtp = &fn->leaf; *rtp; rtp = rtp_next) {
2082 struct fib6_info *cur = rcu_dereference_protected(*rtp,
2083 lockdep_is_held(&table->tb6_lock));
2084 if (rt == cur) {
2085 if (fib6_requires_src(rt: cur))
2086 fib6_routes_require_src_dec(net: info->nl_net);
2087 fib6_del_route(table, fn, rtp, info);
2088 return 0;
2089 }
2090 rtp_next = &cur->fib6_next;
2091 }
2092 return -ENOENT;
2093}
2094
2095/*
2096 * Tree traversal function.
2097 *
2098 * Certainly, it is not interrupt safe.
2099 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
2100 * It means, that we can modify tree during walking
2101 * and use this function for garbage collection, clone pruning,
2102 * cleaning tree when a device goes down etc. etc.
2103 *
2104 * It guarantees that every node will be traversed,
2105 * and that it will be traversed only once.
2106 *
2107 * Callback function w->func may return:
2108 * 0 -> continue walking.
2109 * positive value -> walking is suspended (used by tree dumps,
2110 * and probably by gc, if it will be split to several slices)
2111 * negative value -> terminate walking.
2112 *
2113 * The function itself returns:
2114 * 0 -> walk is complete.
2115 * >0 -> walk is incomplete (i.e. suspended)
2116 * <0 -> walk is terminated by an error.
2117 *
2118 * This function is called with tb6_lock held.
2119 */
2120
2121static int fib6_walk_continue(struct fib6_walker *w)
2122{
2123 struct fib6_node *fn, *pn, *left, *right;
2124
2125 /* w->root should always be table->tb6_root */
2126 WARN_ON_ONCE(!(w->root->fn_flags & RTN_TL_ROOT));
2127
2128 for (;;) {
2129 fn = w->node;
2130 if (!fn)
2131 return 0;
2132
2133 switch (w->state) {
2134#ifdef CONFIG_IPV6_SUBTREES
2135 case FWS_S:
2136 if (FIB6_SUBTREE(fn)) {
2137 w->node = FIB6_SUBTREE(fn);
2138 continue;
2139 }
2140 w->state = FWS_L;
2141 fallthrough;
2142#endif
2143 case FWS_L:
2144 left = rcu_dereference_protected(fn->left, 1);
2145 if (left) {
2146 w->node = left;
2147 w->state = FWS_INIT;
2148 continue;
2149 }
2150 w->state = FWS_R;
2151 fallthrough;
2152 case FWS_R:
2153 right = rcu_dereference_protected(fn->right, 1);
2154 if (right) {
2155 w->node = right;
2156 w->state = FWS_INIT;
2157 continue;
2158 }
2159 w->state = FWS_C;
2160 w->leaf = rcu_dereference_protected(fn->leaf, 1);
2161 fallthrough;
2162 case FWS_C:
2163 if (w->leaf && fn->fn_flags & RTN_RTINFO) {
2164 int err;
2165
2166 if (w->skip) {
2167 w->skip--;
2168 goto skip;
2169 }
2170
2171 err = w->func(w);
2172 if (err)
2173 return err;
2174
2175 w->count++;
2176 continue;
2177 }
2178skip:
2179 w->state = FWS_U;
2180 fallthrough;
2181 case FWS_U:
2182 if (fn == w->root)
2183 return 0;
2184 pn = rcu_dereference_protected(fn->parent, 1);
2185 left = rcu_dereference_protected(pn->left, 1);
2186 right = rcu_dereference_protected(pn->right, 1);
2187 w->node = pn;
2188#ifdef CONFIG_IPV6_SUBTREES
2189 if (FIB6_SUBTREE(pn) == fn) {
2190 WARN_ON(!(fn->fn_flags & RTN_ROOT));
2191 w->state = FWS_L;
2192 continue;
2193 }
2194#endif
2195 if (left == fn) {
2196 w->state = FWS_R;
2197 continue;
2198 }
2199 if (right == fn) {
2200 w->state = FWS_C;
2201 w->leaf = rcu_dereference_protected(w->node->leaf, 1);
2202 continue;
2203 }
2204#if RT6_DEBUG >= 2
2205 WARN_ON(1);
2206#endif
2207 }
2208 }
2209}
2210
2211static int fib6_walk(struct net *net, struct fib6_walker *w)
2212{
2213 int res;
2214
2215 w->state = FWS_INIT;
2216 w->node = w->root;
2217
2218 fib6_walker_link(net, w);
2219 res = fib6_walk_continue(w);
2220 if (res <= 0)
2221 fib6_walker_unlink(net, w);
2222 return res;
2223}
2224
2225static int fib6_clean_node(struct fib6_walker *w)
2226{
2227 int res;
2228 struct fib6_info *rt;
2229 struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w);
2230 struct nl_info info = {
2231 .nl_net = c->net,
2232 .skip_notify = c->skip_notify,
2233 };
2234
2235 if (c->sernum != FIB6_NO_SERNUM_CHANGE &&
2236 READ_ONCE(w->node->fn_sernum) != c->sernum)
2237 WRITE_ONCE(w->node->fn_sernum, c->sernum);
2238
2239 if (!c->func) {
2240 WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE);
2241 w->leaf = NULL;
2242 return 0;
2243 }
2244
2245 for_each_fib6_walker_rt(w) {
2246 res = c->func(rt, c->arg);
2247 if (res == -1) {
2248 w->leaf = rt;
2249 res = fib6_del(rt, info: &info);
2250 if (res) {
2251#if RT6_DEBUG >= 2
2252 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
2253 __func__, rt,
2254 rcu_access_pointer(rt->fib6_node),
2255 res);
2256#endif
2257 continue;
2258 }
2259 return 0;
2260 } else if (res == -2) {
2261 if (WARN_ON(!rt->fib6_nsiblings))
2262 continue;
2263 rt = list_last_entry(&rt->fib6_siblings,
2264 struct fib6_info, fib6_siblings);
2265 continue;
2266 }
2267 WARN_ON(res != 0);
2268 }
2269 w->leaf = rt;
2270 return 0;
2271}
2272
2273/*
2274 * Convenient frontend to tree walker.
2275 *
2276 * func is called on each route.
2277 * It may return -2 -> skip multipath route.
2278 * -1 -> delete this route.
2279 * 0 -> continue walking
2280 */
2281
2282static void fib6_clean_tree(struct net *net, struct fib6_node *root,
2283 int (*func)(struct fib6_info *, void *arg),
2284 int sernum, void *arg, bool skip_notify)
2285{
2286 struct fib6_cleaner c;
2287
2288 c.w.root = root;
2289 c.w.func = fib6_clean_node;
2290 c.w.count = 0;
2291 c.w.skip = 0;
2292 c.w.skip_in_node = 0;
2293 c.func = func;
2294 c.sernum = sernum;
2295 c.arg = arg;
2296 c.net = net;
2297 c.skip_notify = skip_notify;
2298
2299 fib6_walk(net, w: &c.w);
2300}
2301
2302static void __fib6_clean_all(struct net *net,
2303 int (*func)(struct fib6_info *, void *),
2304 int sernum, void *arg, bool skip_notify)
2305{
2306 struct fib6_table *table;
2307 struct hlist_head *head;
2308 unsigned int h;
2309
2310 rcu_read_lock();
2311 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2312 head = &net->ipv6.fib_table_hash[h];
2313 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2314 spin_lock_bh(lock: &table->tb6_lock);
2315 fib6_clean_tree(net, root: &table->tb6_root,
2316 func, sernum, arg, skip_notify);
2317 spin_unlock_bh(lock: &table->tb6_lock);
2318 }
2319 }
2320 rcu_read_unlock();
2321}
2322
2323void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *),
2324 void *arg)
2325{
2326 __fib6_clean_all(net, func, sernum: FIB6_NO_SERNUM_CHANGE, arg, skip_notify: false);
2327}
2328
2329void fib6_clean_all_skip_notify(struct net *net,
2330 int (*func)(struct fib6_info *, void *),
2331 void *arg)
2332{
2333 __fib6_clean_all(net, func, sernum: FIB6_NO_SERNUM_CHANGE, arg, skip_notify: true);
2334}
2335
2336static void fib6_flush_trees(struct net *net)
2337{
2338 int new_sernum = fib6_new_sernum(net);
2339
2340 __fib6_clean_all(net, NULL, sernum: new_sernum, NULL, skip_notify: false);
2341}
2342
2343/*
2344 * Garbage collection
2345 */
2346
2347static int fib6_age(struct fib6_info *rt, struct fib6_gc_args *gc_args)
2348{
2349 unsigned long now = jiffies;
2350
2351 /*
2352 * check addrconf expiration here.
2353 * Routes are expired even if they are in use.
2354 */
2355
2356 if (rt->fib6_flags & RTF_EXPIRES && rt->expires) {
2357 if (time_after(now, rt->expires)) {
2358 pr_debug("expiring %p\n", rt);
2359 return -1;
2360 }
2361 gc_args->more++;
2362 }
2363
2364 /* Also age clones in the exception table.
2365 * Note, that clones are aged out
2366 * only if they are not in use now.
2367 */
2368 rt6_age_exceptions(f6i: rt, gc_args, now);
2369
2370 return 0;
2371}
2372
2373static void fib6_gc_table(struct net *net,
2374 struct fib6_table *tb6,
2375 struct fib6_gc_args *gc_args)
2376{
2377 struct fib6_info *rt;
2378 struct hlist_node *n;
2379 struct nl_info info = {
2380 .nl_net = net,
2381 .skip_notify = false,
2382 };
2383
2384 hlist_for_each_entry_safe(rt, n, &tb6->tb6_gc_hlist, gc_link)
2385 if (fib6_age(rt, gc_args) == -1)
2386 fib6_del(rt, info: &info);
2387}
2388
2389static void fib6_gc_all(struct net *net, struct fib6_gc_args *gc_args)
2390{
2391 struct fib6_table *table;
2392 struct hlist_head *head;
2393 unsigned int h;
2394
2395 rcu_read_lock();
2396 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2397 head = &net->ipv6.fib_table_hash[h];
2398 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2399 spin_lock_bh(lock: &table->tb6_lock);
2400
2401 fib6_gc_table(net, tb6: table, gc_args);
2402
2403 spin_unlock_bh(lock: &table->tb6_lock);
2404 }
2405 }
2406 rcu_read_unlock();
2407}
2408
2409void fib6_run_gc(unsigned long expires, struct net *net, bool force)
2410{
2411 struct fib6_gc_args gc_args;
2412 unsigned long now;
2413
2414 if (force) {
2415 spin_lock_bh(lock: &net->ipv6.fib6_gc_lock);
2416 } else if (!spin_trylock_bh(lock: &net->ipv6.fib6_gc_lock)) {
2417 mod_timer(timer: &net->ipv6.ip6_fib_timer, expires: jiffies + HZ);
2418 return;
2419 }
2420 gc_args.timeout = expires ? (int)expires :
2421 net->ipv6.sysctl.ip6_rt_gc_interval;
2422 gc_args.more = 0;
2423
2424 fib6_gc_all(net, gc_args: &gc_args);
2425 now = jiffies;
2426 net->ipv6.ip6_rt_last_gc = now;
2427
2428 if (gc_args.more)
2429 mod_timer(timer: &net->ipv6.ip6_fib_timer,
2430 expires: round_jiffies(j: now
2431 + net->ipv6.sysctl.ip6_rt_gc_interval));
2432 else
2433 timer_delete(timer: &net->ipv6.ip6_fib_timer);
2434 spin_unlock_bh(lock: &net->ipv6.fib6_gc_lock);
2435}
2436
2437static void fib6_gc_timer_cb(struct timer_list *t)
2438{
2439 struct net *arg = timer_container_of(arg, t, ipv6.ip6_fib_timer);
2440
2441 fib6_run_gc(expires: 0, net: arg, force: true);
2442}
2443
2444static int __net_init fib6_net_init(struct net *net)
2445{
2446 size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ;
2447 int err;
2448
2449 err = fib6_notifier_init(net);
2450 if (err)
2451 return err;
2452
2453 /* Default to 3-tuple */
2454 net->ipv6.sysctl.multipath_hash_fields =
2455 FIB_MULTIPATH_HASH_FIELD_DEFAULT_MASK;
2456
2457 spin_lock_init(&net->ipv6.fib6_gc_lock);
2458 rwlock_init(&net->ipv6.fib6_walker_lock);
2459 INIT_LIST_HEAD(list: &net->ipv6.fib6_walkers);
2460 timer_setup(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, 0);
2461
2462 net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL);
2463 if (!net->ipv6.rt6_stats)
2464 goto out_notifier;
2465
2466 /* Avoid false sharing : Use at least a full cache line */
2467 size = max_t(size_t, size, L1_CACHE_BYTES);
2468
2469 net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL);
2470 if (!net->ipv6.fib_table_hash)
2471 goto out_rt6_stats;
2472
2473 spin_lock_init(&net->ipv6.fib_table_hash_lock);
2474
2475 net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
2476 GFP_KERNEL);
2477 if (!net->ipv6.fib6_main_tbl)
2478 goto out_fib_table_hash;
2479
2480 net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
2481 rcu_assign_pointer(net->ipv6.fib6_main_tbl->tb6_root.leaf,
2482 net->ipv6.fib6_null_entry);
2483 net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
2484 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
2485 inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers);
2486 INIT_HLIST_HEAD(&net->ipv6.fib6_main_tbl->tb6_gc_hlist);
2487
2488#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2489 net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
2490 GFP_KERNEL);
2491 if (!net->ipv6.fib6_local_tbl)
2492 goto out_fib6_main_tbl;
2493 net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
2494 rcu_assign_pointer(net->ipv6.fib6_local_tbl->tb6_root.leaf,
2495 net->ipv6.fib6_null_entry);
2496 net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
2497 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
2498 inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers);
2499 INIT_HLIST_HEAD(&net->ipv6.fib6_local_tbl->tb6_gc_hlist);
2500#endif
2501 fib6_tables_init(net);
2502
2503 return 0;
2504
2505#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2506out_fib6_main_tbl:
2507 kfree(net->ipv6.fib6_main_tbl);
2508#endif
2509out_fib_table_hash:
2510 kfree(objp: net->ipv6.fib_table_hash);
2511out_rt6_stats:
2512 kfree(objp: net->ipv6.rt6_stats);
2513out_notifier:
2514 fib6_notifier_exit(net);
2515 return -ENOMEM;
2516}
2517
2518static void fib6_net_exit(struct net *net)
2519{
2520 unsigned int i;
2521
2522 timer_delete_sync(timer: &net->ipv6.ip6_fib_timer);
2523
2524 for (i = 0; i < FIB6_TABLE_HASHSZ; i++) {
2525 struct hlist_head *head = &net->ipv6.fib_table_hash[i];
2526 struct hlist_node *tmp;
2527 struct fib6_table *tb;
2528
2529 hlist_for_each_entry_safe(tb, tmp, head, tb6_hlist) {
2530 hlist_del(n: &tb->tb6_hlist);
2531 fib6_free_table(table: tb);
2532 }
2533 }
2534
2535 kfree(objp: net->ipv6.fib_table_hash);
2536 kfree(objp: net->ipv6.rt6_stats);
2537 fib6_notifier_exit(net);
2538}
2539
2540static struct pernet_operations fib6_net_ops = {
2541 .init = fib6_net_init,
2542 .exit = fib6_net_exit,
2543};
2544
2545static const struct rtnl_msg_handler fib6_rtnl_msg_handlers[] __initconst_or_module = {
2546 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETROUTE,
2547 .dumpit = inet6_dump_fib,
2548 .flags = RTNL_FLAG_DUMP_UNLOCKED | RTNL_FLAG_DUMP_SPLIT_NLM_DONE},
2549};
2550
2551int __init fib6_init(void)
2552{
2553 int ret = -ENOMEM;
2554
2555 fib6_node_kmem = KMEM_CACHE(fib6_node,
2556 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT);
2557 if (!fib6_node_kmem)
2558 goto out;
2559
2560 ret = register_pernet_subsys(&fib6_net_ops);
2561 if (ret)
2562 goto out_kmem_cache_create;
2563
2564 ret = rtnl_register_many(fib6_rtnl_msg_handlers);
2565 if (ret)
2566 goto out_unregister_subsys;
2567
2568 __fib6_flush_trees = fib6_flush_trees;
2569out:
2570 return ret;
2571
2572out_unregister_subsys:
2573 unregister_pernet_subsys(&fib6_net_ops);
2574out_kmem_cache_create:
2575 kmem_cache_destroy(s: fib6_node_kmem);
2576 goto out;
2577}
2578
2579void fib6_gc_cleanup(void)
2580{
2581 unregister_pernet_subsys(&fib6_net_ops);
2582 kmem_cache_destroy(s: fib6_node_kmem);
2583}
2584
2585#ifdef CONFIG_PROC_FS
2586static int ipv6_route_native_seq_show(struct seq_file *seq, void *v)
2587{
2588 struct fib6_info *rt = v;
2589 struct ipv6_route_iter *iter = seq->private;
2590 struct fib6_nh *fib6_nh = rt->fib6_nh;
2591 unsigned int flags = rt->fib6_flags;
2592 const struct net_device *dev;
2593
2594 if (rt->nh)
2595 fib6_nh = nexthop_fib6_nh(nh: rt->nh);
2596
2597 seq_printf(m: seq, fmt: "%pi6 %02x ", &rt->fib6_dst.addr, rt->fib6_dst.plen);
2598
2599#ifdef CONFIG_IPV6_SUBTREES
2600 seq_printf(seq, "%pi6 %02x ", &rt->fib6_src.addr, rt->fib6_src.plen);
2601#else
2602 seq_puts(m: seq, s: "00000000000000000000000000000000 00 ");
2603#endif
2604 if (fib6_nh->fib_nh_gw_family) {
2605 flags |= RTF_GATEWAY;
2606 seq_printf(m: seq, fmt: "%pi6", &fib6_nh->fib_nh_gw6);
2607 } else {
2608 seq_puts(m: seq, s: "00000000000000000000000000000000");
2609 }
2610
2611 dev = fib6_nh->fib_nh_dev;
2612 seq_printf(m: seq, fmt: " %08x %08x %08x %08x %8s\n",
2613 rt->fib6_metric, refcount_read(r: &rt->fib6_ref), 0,
2614 flags, dev ? dev->name : "");
2615 iter->w.leaf = NULL;
2616 return 0;
2617}
2618
2619static int ipv6_route_yield(struct fib6_walker *w)
2620{
2621 struct ipv6_route_iter *iter = w->args;
2622
2623 if (!iter->skip)
2624 return 1;
2625
2626 do {
2627 iter->w.leaf = rcu_dereference_protected(
2628 iter->w.leaf->fib6_next,
2629 lockdep_is_held(&iter->tbl->tb6_lock));
2630 iter->skip--;
2631 if (!iter->skip && iter->w.leaf)
2632 return 1;
2633 } while (iter->w.leaf);
2634
2635 return 0;
2636}
2637
2638static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter,
2639 struct net *net)
2640{
2641 memset(s: &iter->w, c: 0, n: sizeof(iter->w));
2642 iter->w.func = ipv6_route_yield;
2643 iter->w.root = &iter->tbl->tb6_root;
2644 iter->w.state = FWS_INIT;
2645 iter->w.node = iter->w.root;
2646 iter->w.args = iter;
2647 iter->sernum = READ_ONCE(iter->w.root->fn_sernum);
2648 INIT_LIST_HEAD(list: &iter->w.lh);
2649 fib6_walker_link(net, w: &iter->w);
2650}
2651
2652static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl,
2653 struct net *net)
2654{
2655 unsigned int h;
2656 struct hlist_node *node;
2657
2658 if (tbl) {
2659 h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1;
2660 node = rcu_dereference(hlist_next_rcu(&tbl->tb6_hlist));
2661 } else {
2662 h = 0;
2663 node = NULL;
2664 }
2665
2666 while (!node && h < FIB6_TABLE_HASHSZ) {
2667 node = rcu_dereference(
2668 hlist_first_rcu(&net->ipv6.fib_table_hash[h++]));
2669 }
2670 return hlist_entry_safe(node, struct fib6_table, tb6_hlist);
2671}
2672
2673static void ipv6_route_check_sernum(struct ipv6_route_iter *iter)
2674{
2675 int sernum = READ_ONCE(iter->w.root->fn_sernum);
2676
2677 if (iter->sernum != sernum) {
2678 iter->sernum = sernum;
2679 iter->w.state = FWS_INIT;
2680 iter->w.node = iter->w.root;
2681 WARN_ON(iter->w.skip);
2682 iter->w.skip = iter->w.count;
2683 }
2684}
2685
2686static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2687{
2688 int r;
2689 struct fib6_info *n;
2690 struct net *net = seq_file_net(seq);
2691 struct ipv6_route_iter *iter = seq->private;
2692
2693 ++(*pos);
2694 if (!v)
2695 goto iter_table;
2696
2697 n = rcu_dereference(((struct fib6_info *)v)->fib6_next);
2698 if (n)
2699 return n;
2700
2701iter_table:
2702 ipv6_route_check_sernum(iter);
2703 spin_lock_bh(lock: &iter->tbl->tb6_lock);
2704 r = fib6_walk_continue(w: &iter->w);
2705 spin_unlock_bh(lock: &iter->tbl->tb6_lock);
2706 if (r > 0) {
2707 return iter->w.leaf;
2708 } else if (r < 0) {
2709 fib6_walker_unlink(net, w: &iter->w);
2710 return NULL;
2711 }
2712 fib6_walker_unlink(net, w: &iter->w);
2713
2714 iter->tbl = ipv6_route_seq_next_table(tbl: iter->tbl, net);
2715 if (!iter->tbl)
2716 return NULL;
2717
2718 ipv6_route_seq_setup_walk(iter, net);
2719 goto iter_table;
2720}
2721
2722static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos)
2723 __acquires(RCU)
2724{
2725 struct net *net = seq_file_net(seq);
2726 struct ipv6_route_iter *iter = seq->private;
2727
2728 rcu_read_lock();
2729 iter->tbl = ipv6_route_seq_next_table(NULL, net);
2730 iter->skip = *pos;
2731
2732 if (iter->tbl) {
2733 loff_t p = 0;
2734
2735 ipv6_route_seq_setup_walk(iter, net);
2736 return ipv6_route_seq_next(seq, NULL, pos: &p);
2737 } else {
2738 return NULL;
2739 }
2740}
2741
2742static bool ipv6_route_iter_active(struct ipv6_route_iter *iter)
2743{
2744 struct fib6_walker *w = &iter->w;
2745 return w->node && !(w->state == FWS_U && w->node == w->root);
2746}
2747
2748static void ipv6_route_native_seq_stop(struct seq_file *seq, void *v)
2749 __releases(RCU)
2750{
2751 struct net *net = seq_file_net(seq);
2752 struct ipv6_route_iter *iter = seq->private;
2753
2754 if (ipv6_route_iter_active(iter))
2755 fib6_walker_unlink(net, w: &iter->w);
2756
2757 rcu_read_unlock();
2758}
2759
2760#if IS_BUILTIN(CONFIG_IPV6) && defined(CONFIG_BPF_SYSCALL)
2761static int ipv6_route_prog_seq_show(struct bpf_prog *prog,
2762 struct bpf_iter_meta *meta,
2763 void *v)
2764{
2765 struct bpf_iter__ipv6_route ctx;
2766
2767 ctx.meta = meta;
2768 ctx.rt = v;
2769 return bpf_iter_run_prog(prog, &ctx);
2770}
2771
2772static int ipv6_route_seq_show(struct seq_file *seq, void *v)
2773{
2774 struct ipv6_route_iter *iter = seq->private;
2775 struct bpf_iter_meta meta;
2776 struct bpf_prog *prog;
2777 int ret;
2778
2779 meta.seq = seq;
2780 prog = bpf_iter_get_info(&meta, false);
2781 if (!prog)
2782 return ipv6_route_native_seq_show(seq, v);
2783
2784 ret = ipv6_route_prog_seq_show(prog, &meta, v);
2785 iter->w.leaf = NULL;
2786
2787 return ret;
2788}
2789
2790static void ipv6_route_seq_stop(struct seq_file *seq, void *v)
2791{
2792 struct bpf_iter_meta meta;
2793 struct bpf_prog *prog;
2794
2795 if (!v) {
2796 meta.seq = seq;
2797 prog = bpf_iter_get_info(&meta, true);
2798 if (prog)
2799 (void)ipv6_route_prog_seq_show(prog, &meta, v);
2800 }
2801
2802 ipv6_route_native_seq_stop(seq, v);
2803}
2804#else
2805static int ipv6_route_seq_show(struct seq_file *seq, void *v)
2806{
2807 return ipv6_route_native_seq_show(seq, v);
2808}
2809
2810static void ipv6_route_seq_stop(struct seq_file *seq, void *v)
2811{
2812 ipv6_route_native_seq_stop(seq, v);
2813}
2814#endif
2815
2816const struct seq_operations ipv6_route_seq_ops = {
2817 .start = ipv6_route_seq_start,
2818 .next = ipv6_route_seq_next,
2819 .stop = ipv6_route_seq_stop,
2820 .show = ipv6_route_seq_show
2821};
2822#endif /* CONFIG_PROC_FS */
2823