xref: /openbmc/linux/net/ipv6/ip6_fib.c (revision b2765275)
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/errno.h>
19 #include <linux/types.h>
20 #include <linux/net.h>
21 #include <linux/route.h>
22 #include <linux/netdevice.h>
23 #include <linux/in6.h>
24 #include <linux/init.h>
25 #include <linux/list.h>
26 #include <linux/slab.h>
27 
28 #include <net/ip.h>
29 #include <net/ipv6.h>
30 #include <net/ndisc.h>
31 #include <net/addrconf.h>
32 #include <net/lwtunnel.h>
33 #include <net/fib_notifier.h>
34 
35 #include <net/ip6_fib.h>
36 #include <net/ip6_route.h>
37 
38 static struct kmem_cache *fib6_node_kmem __read_mostly;
39 
40 struct fib6_cleaner {
41 	struct fib6_walker w;
42 	struct net *net;
43 	int (*func)(struct fib6_info *, void *arg);
44 	int sernum;
45 	void *arg;
46 	bool skip_notify;
47 };
48 
49 #ifdef CONFIG_IPV6_SUBTREES
50 #define FWS_INIT FWS_S
51 #else
52 #define FWS_INIT FWS_L
53 #endif
54 
55 static struct fib6_info *fib6_find_prefix(struct net *net,
56 					 struct fib6_table *table,
57 					 struct fib6_node *fn);
58 static struct fib6_node *fib6_repair_tree(struct net *net,
59 					  struct fib6_table *table,
60 					  struct fib6_node *fn);
61 static int fib6_walk(struct net *net, struct fib6_walker *w);
62 static int fib6_walk_continue(struct fib6_walker *w);
63 
64 /*
65  *	A routing update causes an increase of the serial number on the
66  *	affected subtree. This allows for cached routes to be asynchronously
67  *	tested when modifications are made to the destination cache as a
68  *	result of redirects, path MTU changes, etc.
69  */
70 
71 static void fib6_gc_timer_cb(struct timer_list *t);
72 
73 #define FOR_WALKERS(net, w) \
74 	list_for_each_entry(w, &(net)->ipv6.fib6_walkers, lh)
75 
76 static void fib6_walker_link(struct net *net, struct fib6_walker *w)
77 {
78 	write_lock_bh(&net->ipv6.fib6_walker_lock);
79 	list_add(&w->lh, &net->ipv6.fib6_walkers);
80 	write_unlock_bh(&net->ipv6.fib6_walker_lock);
81 }
82 
83 static void fib6_walker_unlink(struct net *net, struct fib6_walker *w)
84 {
85 	write_lock_bh(&net->ipv6.fib6_walker_lock);
86 	list_del(&w->lh);
87 	write_unlock_bh(&net->ipv6.fib6_walker_lock);
88 }
89 
90 static int fib6_new_sernum(struct net *net)
91 {
92 	int new, old;
93 
94 	do {
95 		old = atomic_read(&net->ipv6.fib6_sernum);
96 		new = old < INT_MAX ? old + 1 : 1;
97 	} while (atomic_cmpxchg(&net->ipv6.fib6_sernum,
98 				old, new) != old);
99 	return new;
100 }
101 
102 enum {
103 	FIB6_NO_SERNUM_CHANGE = 0,
104 };
105 
106 void fib6_update_sernum(struct net *net, struct fib6_info *f6i)
107 {
108 	struct fib6_node *fn;
109 
110 	fn = rcu_dereference_protected(f6i->fib6_node,
111 			lockdep_is_held(&f6i->fib6_table->tb6_lock));
112 	if (fn)
113 		fn->fn_sernum = fib6_new_sernum(net);
114 }
115 
116 /*
117  *	Auxiliary address test functions for the radix tree.
118  *
119  *	These assume a 32bit processor (although it will work on
120  *	64bit processors)
121  */
122 
123 /*
124  *	test bit
125  */
126 #if defined(__LITTLE_ENDIAN)
127 # define BITOP_BE32_SWIZZLE	(0x1F & ~7)
128 #else
129 # define BITOP_BE32_SWIZZLE	0
130 #endif
131 
132 static __be32 addr_bit_set(const void *token, int fn_bit)
133 {
134 	const __be32 *addr = token;
135 	/*
136 	 * Here,
137 	 *	1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
138 	 * is optimized version of
139 	 *	htonl(1 << ((~fn_bit)&0x1F))
140 	 * See include/asm-generic/bitops/le.h.
141 	 */
142 	return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) &
143 	       addr[fn_bit >> 5];
144 }
145 
146 struct fib6_info *fib6_info_alloc(gfp_t gfp_flags, bool with_fib6_nh)
147 {
148 	struct fib6_info *f6i;
149 	size_t sz = sizeof(*f6i);
150 
151 	if (with_fib6_nh)
152 		sz += sizeof(struct fib6_nh);
153 
154 	f6i = kzalloc(sz, gfp_flags);
155 	if (!f6i)
156 		return NULL;
157 
158 	/* fib6_siblings is a union with nh_list, so this initializes both */
159 	INIT_LIST_HEAD(&f6i->fib6_siblings);
160 	refcount_set(&f6i->fib6_ref, 1);
161 
162 	return f6i;
163 }
164 
165 void fib6_info_destroy_rcu(struct rcu_head *head)
166 {
167 	struct fib6_info *f6i = container_of(head, struct fib6_info, rcu);
168 
169 	WARN_ON(f6i->fib6_node);
170 
171 	if (f6i->nh)
172 		nexthop_put(f6i->nh);
173 	else
174 		fib6_nh_release(f6i->fib6_nh);
175 
176 	ip_fib_metrics_put(f6i->fib6_metrics);
177 	kfree(f6i);
178 }
179 EXPORT_SYMBOL_GPL(fib6_info_destroy_rcu);
180 
181 static struct fib6_node *node_alloc(struct net *net)
182 {
183 	struct fib6_node *fn;
184 
185 	fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC);
186 	if (fn)
187 		net->ipv6.rt6_stats->fib_nodes++;
188 
189 	return fn;
190 }
191 
192 static void node_free_immediate(struct net *net, struct fib6_node *fn)
193 {
194 	kmem_cache_free(fib6_node_kmem, fn);
195 	net->ipv6.rt6_stats->fib_nodes--;
196 }
197 
198 static void node_free_rcu(struct rcu_head *head)
199 {
200 	struct fib6_node *fn = container_of(head, struct fib6_node, rcu);
201 
202 	kmem_cache_free(fib6_node_kmem, fn);
203 }
204 
205 static void node_free(struct net *net, struct fib6_node *fn)
206 {
207 	call_rcu(&fn->rcu, node_free_rcu);
208 	net->ipv6.rt6_stats->fib_nodes--;
209 }
210 
211 static void fib6_free_table(struct fib6_table *table)
212 {
213 	inetpeer_invalidate_tree(&table->tb6_peers);
214 	kfree(table);
215 }
216 
217 static void fib6_link_table(struct net *net, struct fib6_table *tb)
218 {
219 	unsigned int h;
220 
221 	/*
222 	 * Initialize table lock at a single place to give lockdep a key,
223 	 * tables aren't visible prior to being linked to the list.
224 	 */
225 	spin_lock_init(&tb->tb6_lock);
226 	h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1);
227 
228 	/*
229 	 * No protection necessary, this is the only list mutatation
230 	 * operation, tables never disappear once they exist.
231 	 */
232 	hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]);
233 }
234 
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
236 
237 static struct fib6_table *fib6_alloc_table(struct net *net, u32 id)
238 {
239 	struct fib6_table *table;
240 
241 	table = kzalloc(sizeof(*table), GFP_ATOMIC);
242 	if (table) {
243 		table->tb6_id = id;
244 		rcu_assign_pointer(table->tb6_root.leaf,
245 				   net->ipv6.fib6_null_entry);
246 		table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
247 		inet_peer_base_init(&table->tb6_peers);
248 	}
249 
250 	return table;
251 }
252 
253 struct fib6_table *fib6_new_table(struct net *net, u32 id)
254 {
255 	struct fib6_table *tb;
256 
257 	if (id == 0)
258 		id = RT6_TABLE_MAIN;
259 	tb = fib6_get_table(net, id);
260 	if (tb)
261 		return tb;
262 
263 	tb = fib6_alloc_table(net, id);
264 	if (tb)
265 		fib6_link_table(net, tb);
266 
267 	return tb;
268 }
269 EXPORT_SYMBOL_GPL(fib6_new_table);
270 
271 struct fib6_table *fib6_get_table(struct net *net, u32 id)
272 {
273 	struct fib6_table *tb;
274 	struct hlist_head *head;
275 	unsigned int h;
276 
277 	if (id == 0)
278 		id = RT6_TABLE_MAIN;
279 	h = id & (FIB6_TABLE_HASHSZ - 1);
280 	rcu_read_lock();
281 	head = &net->ipv6.fib_table_hash[h];
282 	hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
283 		if (tb->tb6_id == id) {
284 			rcu_read_unlock();
285 			return tb;
286 		}
287 	}
288 	rcu_read_unlock();
289 
290 	return NULL;
291 }
292 EXPORT_SYMBOL_GPL(fib6_get_table);
293 
294 static void __net_init fib6_tables_init(struct net *net)
295 {
296 	fib6_link_table(net, net->ipv6.fib6_main_tbl);
297 	fib6_link_table(net, net->ipv6.fib6_local_tbl);
298 }
299 #else
300 
301 struct fib6_table *fib6_new_table(struct net *net, u32 id)
302 {
303 	return fib6_get_table(net, id);
304 }
305 
306 struct fib6_table *fib6_get_table(struct net *net, u32 id)
307 {
308 	  return net->ipv6.fib6_main_tbl;
309 }
310 
311 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
312 				   const struct sk_buff *skb,
313 				   int flags, pol_lookup_t lookup)
314 {
315 	struct rt6_info *rt;
316 
317 	rt = lookup(net, net->ipv6.fib6_main_tbl, fl6, skb, flags);
318 	if (rt->dst.error == -EAGAIN) {
319 		ip6_rt_put_flags(rt, flags);
320 		rt = net->ipv6.ip6_null_entry;
321 		if (!(flags & RT6_LOOKUP_F_DST_NOREF))
322 			dst_hold(&rt->dst);
323 	}
324 
325 	return &rt->dst;
326 }
327 
328 /* called with rcu lock held; no reference taken on fib6_info */
329 int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
330 		struct fib6_result *res, int flags)
331 {
332 	return fib6_table_lookup(net, net->ipv6.fib6_main_tbl, oif, fl6,
333 				 res, flags);
334 }
335 
336 static void __net_init fib6_tables_init(struct net *net)
337 {
338 	fib6_link_table(net, net->ipv6.fib6_main_tbl);
339 }
340 
341 #endif
342 
343 unsigned int fib6_tables_seq_read(struct net *net)
344 {
345 	unsigned int h, fib_seq = 0;
346 
347 	rcu_read_lock();
348 	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
349 		struct hlist_head *head = &net->ipv6.fib_table_hash[h];
350 		struct fib6_table *tb;
351 
352 		hlist_for_each_entry_rcu(tb, head, tb6_hlist)
353 			fib_seq += tb->fib_seq;
354 	}
355 	rcu_read_unlock();
356 
357 	return fib_seq;
358 }
359 
360 static int call_fib6_entry_notifier(struct notifier_block *nb,
361 				    enum fib_event_type event_type,
362 				    struct fib6_info *rt,
363 				    struct netlink_ext_ack *extack)
364 {
365 	struct fib6_entry_notifier_info info = {
366 		.info.extack = extack,
367 		.rt = rt,
368 	};
369 
370 	return call_fib6_notifier(nb, event_type, &info.info);
371 }
372 
373 static int call_fib6_multipath_entry_notifier(struct notifier_block *nb,
374 					      enum fib_event_type event_type,
375 					      struct fib6_info *rt,
376 					      unsigned int nsiblings,
377 					      struct netlink_ext_ack *extack)
378 {
379 	struct fib6_entry_notifier_info info = {
380 		.info.extack = extack,
381 		.rt = rt,
382 		.nsiblings = nsiblings,
383 	};
384 
385 	return call_fib6_notifier(nb, event_type, &info.info);
386 }
387 
388 int call_fib6_entry_notifiers(struct net *net,
389 			      enum fib_event_type event_type,
390 			      struct fib6_info *rt,
391 			      struct netlink_ext_ack *extack)
392 {
393 	struct fib6_entry_notifier_info info = {
394 		.info.extack = extack,
395 		.rt = rt,
396 	};
397 
398 	rt->fib6_table->fib_seq++;
399 	return call_fib6_notifiers(net, event_type, &info.info);
400 }
401 
402 int call_fib6_multipath_entry_notifiers(struct net *net,
403 					enum fib_event_type event_type,
404 					struct fib6_info *rt,
405 					unsigned int nsiblings,
406 					struct netlink_ext_ack *extack)
407 {
408 	struct fib6_entry_notifier_info info = {
409 		.info.extack = extack,
410 		.rt = rt,
411 		.nsiblings = nsiblings,
412 	};
413 
414 	rt->fib6_table->fib_seq++;
415 	return call_fib6_notifiers(net, event_type, &info.info);
416 }
417 
418 int call_fib6_entry_notifiers_replace(struct net *net, struct fib6_info *rt)
419 {
420 	struct fib6_entry_notifier_info info = {
421 		.rt = rt,
422 		.nsiblings = rt->fib6_nsiblings,
423 	};
424 
425 	rt->fib6_table->fib_seq++;
426 	return call_fib6_notifiers(net, FIB_EVENT_ENTRY_REPLACE, &info.info);
427 }
428 
429 struct fib6_dump_arg {
430 	struct net *net;
431 	struct notifier_block *nb;
432 	struct netlink_ext_ack *extack;
433 };
434 
435 static int fib6_rt_dump(struct fib6_info *rt, struct fib6_dump_arg *arg)
436 {
437 	enum fib_event_type fib_event = FIB_EVENT_ENTRY_REPLACE;
438 	int err;
439 
440 	if (!rt || rt == arg->net->ipv6.fib6_null_entry)
441 		return 0;
442 
443 	if (rt->fib6_nsiblings)
444 		err = call_fib6_multipath_entry_notifier(arg->nb, fib_event,
445 							 rt,
446 							 rt->fib6_nsiblings,
447 							 arg->extack);
448 	else
449 		err = call_fib6_entry_notifier(arg->nb, fib_event, rt,
450 					       arg->extack);
451 
452 	return err;
453 }
454 
455 static int fib6_node_dump(struct fib6_walker *w)
456 {
457 	int err;
458 
459 	err = fib6_rt_dump(w->leaf, w->args);
460 	w->leaf = NULL;
461 	return err;
462 }
463 
464 static int fib6_table_dump(struct net *net, struct fib6_table *tb,
465 			   struct fib6_walker *w)
466 {
467 	int err;
468 
469 	w->root = &tb->tb6_root;
470 	spin_lock_bh(&tb->tb6_lock);
471 	err = fib6_walk(net, w);
472 	spin_unlock_bh(&tb->tb6_lock);
473 	return err;
474 }
475 
476 /* Called with rcu_read_lock() */
477 int fib6_tables_dump(struct net *net, struct notifier_block *nb,
478 		     struct netlink_ext_ack *extack)
479 {
480 	struct fib6_dump_arg arg;
481 	struct fib6_walker *w;
482 	unsigned int h;
483 	int err = 0;
484 
485 	w = kzalloc(sizeof(*w), GFP_ATOMIC);
486 	if (!w)
487 		return -ENOMEM;
488 
489 	w->func = fib6_node_dump;
490 	arg.net = net;
491 	arg.nb = nb;
492 	arg.extack = extack;
493 	w->args = &arg;
494 
495 	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
496 		struct hlist_head *head = &net->ipv6.fib_table_hash[h];
497 		struct fib6_table *tb;
498 
499 		hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
500 			err = fib6_table_dump(net, tb, w);
501 			if (err < 0)
502 				goto out;
503 		}
504 	}
505 
506 out:
507 	kfree(w);
508 
509 	return err;
510 }
511 
512 static int fib6_dump_node(struct fib6_walker *w)
513 {
514 	int res;
515 	struct fib6_info *rt;
516 
517 	for_each_fib6_walker_rt(w) {
518 		res = rt6_dump_route(rt, w->args, w->skip_in_node);
519 		if (res >= 0) {
520 			/* Frame is full, suspend walking */
521 			w->leaf = rt;
522 
523 			/* We'll restart from this node, so if some routes were
524 			 * already dumped, skip them next time.
525 			 */
526 			w->skip_in_node += res;
527 
528 			return 1;
529 		}
530 		w->skip_in_node = 0;
531 
532 		/* Multipath routes are dumped in one route with the
533 		 * RTA_MULTIPATH attribute. Jump 'rt' to point to the
534 		 * last sibling of this route (no need to dump the
535 		 * sibling routes again)
536 		 */
537 		if (rt->fib6_nsiblings)
538 			rt = list_last_entry(&rt->fib6_siblings,
539 					     struct fib6_info,
540 					     fib6_siblings);
541 	}
542 	w->leaf = NULL;
543 	return 0;
544 }
545 
546 static void fib6_dump_end(struct netlink_callback *cb)
547 {
548 	struct net *net = sock_net(cb->skb->sk);
549 	struct fib6_walker *w = (void *)cb->args[2];
550 
551 	if (w) {
552 		if (cb->args[4]) {
553 			cb->args[4] = 0;
554 			fib6_walker_unlink(net, w);
555 		}
556 		cb->args[2] = 0;
557 		kfree(w);
558 	}
559 	cb->done = (void *)cb->args[3];
560 	cb->args[1] = 3;
561 }
562 
563 static int fib6_dump_done(struct netlink_callback *cb)
564 {
565 	fib6_dump_end(cb);
566 	return cb->done ? cb->done(cb) : 0;
567 }
568 
569 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb,
570 			   struct netlink_callback *cb)
571 {
572 	struct net *net = sock_net(skb->sk);
573 	struct fib6_walker *w;
574 	int res;
575 
576 	w = (void *)cb->args[2];
577 	w->root = &table->tb6_root;
578 
579 	if (cb->args[4] == 0) {
580 		w->count = 0;
581 		w->skip = 0;
582 		w->skip_in_node = 0;
583 
584 		spin_lock_bh(&table->tb6_lock);
585 		res = fib6_walk(net, w);
586 		spin_unlock_bh(&table->tb6_lock);
587 		if (res > 0) {
588 			cb->args[4] = 1;
589 			cb->args[5] = w->root->fn_sernum;
590 		}
591 	} else {
592 		if (cb->args[5] != w->root->fn_sernum) {
593 			/* Begin at the root if the tree changed */
594 			cb->args[5] = w->root->fn_sernum;
595 			w->state = FWS_INIT;
596 			w->node = w->root;
597 			w->skip = w->count;
598 			w->skip_in_node = 0;
599 		} else
600 			w->skip = 0;
601 
602 		spin_lock_bh(&table->tb6_lock);
603 		res = fib6_walk_continue(w);
604 		spin_unlock_bh(&table->tb6_lock);
605 		if (res <= 0) {
606 			fib6_walker_unlink(net, w);
607 			cb->args[4] = 0;
608 		}
609 	}
610 
611 	return res;
612 }
613 
614 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
615 {
616 	struct rt6_rtnl_dump_arg arg = { .filter.dump_exceptions = true,
617 					 .filter.dump_routes = true };
618 	const struct nlmsghdr *nlh = cb->nlh;
619 	struct net *net = sock_net(skb->sk);
620 	unsigned int h, s_h;
621 	unsigned int e = 0, s_e;
622 	struct fib6_walker *w;
623 	struct fib6_table *tb;
624 	struct hlist_head *head;
625 	int res = 0;
626 
627 	if (cb->strict_check) {
628 		int err;
629 
630 		err = ip_valid_fib_dump_req(net, nlh, &arg.filter, cb);
631 		if (err < 0)
632 			return err;
633 	} else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) {
634 		struct rtmsg *rtm = nlmsg_data(nlh);
635 
636 		if (rtm->rtm_flags & RTM_F_PREFIX)
637 			arg.filter.flags = RTM_F_PREFIX;
638 	}
639 
640 	w = (void *)cb->args[2];
641 	if (!w) {
642 		/* New dump:
643 		 *
644 		 * 1. hook callback destructor.
645 		 */
646 		cb->args[3] = (long)cb->done;
647 		cb->done = fib6_dump_done;
648 
649 		/*
650 		 * 2. allocate and initialize walker.
651 		 */
652 		w = kzalloc(sizeof(*w), GFP_ATOMIC);
653 		if (!w)
654 			return -ENOMEM;
655 		w->func = fib6_dump_node;
656 		cb->args[2] = (long)w;
657 	}
658 
659 	arg.skb = skb;
660 	arg.cb = cb;
661 	arg.net = net;
662 	w->args = &arg;
663 
664 	if (arg.filter.table_id) {
665 		tb = fib6_get_table(net, arg.filter.table_id);
666 		if (!tb) {
667 			if (arg.filter.dump_all_families)
668 				goto out;
669 
670 			NL_SET_ERR_MSG_MOD(cb->extack, "FIB table does not exist");
671 			return -ENOENT;
672 		}
673 
674 		if (!cb->args[0]) {
675 			res = fib6_dump_table(tb, skb, cb);
676 			if (!res)
677 				cb->args[0] = 1;
678 		}
679 		goto out;
680 	}
681 
682 	s_h = cb->args[0];
683 	s_e = cb->args[1];
684 
685 	rcu_read_lock();
686 	for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) {
687 		e = 0;
688 		head = &net->ipv6.fib_table_hash[h];
689 		hlist_for_each_entry_rcu(tb, head, tb6_hlist) {
690 			if (e < s_e)
691 				goto next;
692 			res = fib6_dump_table(tb, skb, cb);
693 			if (res != 0)
694 				goto out_unlock;
695 next:
696 			e++;
697 		}
698 	}
699 out_unlock:
700 	rcu_read_unlock();
701 	cb->args[1] = e;
702 	cb->args[0] = h;
703 out:
704 	res = res < 0 ? res : skb->len;
705 	if (res <= 0)
706 		fib6_dump_end(cb);
707 	return res;
708 }
709 
710 void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val)
711 {
712 	if (!f6i)
713 		return;
714 
715 	if (f6i->fib6_metrics == &dst_default_metrics) {
716 		struct dst_metrics *p = kzalloc(sizeof(*p), GFP_ATOMIC);
717 
718 		if (!p)
719 			return;
720 
721 		refcount_set(&p->refcnt, 1);
722 		f6i->fib6_metrics = p;
723 	}
724 
725 	f6i->fib6_metrics->metrics[metric - 1] = val;
726 }
727 
728 /*
729  *	Routing Table
730  *
731  *	return the appropriate node for a routing tree "add" operation
732  *	by either creating and inserting or by returning an existing
733  *	node.
734  */
735 
736 static struct fib6_node *fib6_add_1(struct net *net,
737 				    struct fib6_table *table,
738 				    struct fib6_node *root,
739 				    struct in6_addr *addr, int plen,
740 				    int offset, int allow_create,
741 				    int replace_required,
742 				    struct netlink_ext_ack *extack)
743 {
744 	struct fib6_node *fn, *in, *ln;
745 	struct fib6_node *pn = NULL;
746 	struct rt6key *key;
747 	int	bit;
748 	__be32	dir = 0;
749 
750 	RT6_TRACE("fib6_add_1\n");
751 
752 	/* insert node in tree */
753 
754 	fn = root;
755 
756 	do {
757 		struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
758 					    lockdep_is_held(&table->tb6_lock));
759 		key = (struct rt6key *)((u8 *)leaf + offset);
760 
761 		/*
762 		 *	Prefix match
763 		 */
764 		if (plen < fn->fn_bit ||
765 		    !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) {
766 			if (!allow_create) {
767 				if (replace_required) {
768 					NL_SET_ERR_MSG(extack,
769 						       "Can not replace route - no match found");
770 					pr_warn("Can't replace route, no match found\n");
771 					return ERR_PTR(-ENOENT);
772 				}
773 				pr_warn("NLM_F_CREATE should be set when creating new route\n");
774 			}
775 			goto insert_above;
776 		}
777 
778 		/*
779 		 *	Exact match ?
780 		 */
781 
782 		if (plen == fn->fn_bit) {
783 			/* clean up an intermediate node */
784 			if (!(fn->fn_flags & RTN_RTINFO)) {
785 				RCU_INIT_POINTER(fn->leaf, NULL);
786 				fib6_info_release(leaf);
787 			/* remove null_entry in the root node */
788 			} else if (fn->fn_flags & RTN_TL_ROOT &&
789 				   rcu_access_pointer(fn->leaf) ==
790 				   net->ipv6.fib6_null_entry) {
791 				RCU_INIT_POINTER(fn->leaf, NULL);
792 			}
793 
794 			return fn;
795 		}
796 
797 		/*
798 		 *	We have more bits to go
799 		 */
800 
801 		/* Try to walk down on tree. */
802 		dir = addr_bit_set(addr, fn->fn_bit);
803 		pn = fn;
804 		fn = dir ?
805 		     rcu_dereference_protected(fn->right,
806 					lockdep_is_held(&table->tb6_lock)) :
807 		     rcu_dereference_protected(fn->left,
808 					lockdep_is_held(&table->tb6_lock));
809 	} while (fn);
810 
811 	if (!allow_create) {
812 		/* We should not create new node because
813 		 * NLM_F_REPLACE was specified without NLM_F_CREATE
814 		 * I assume it is safe to require NLM_F_CREATE when
815 		 * REPLACE flag is used! Later we may want to remove the
816 		 * check for replace_required, because according
817 		 * to netlink specification, NLM_F_CREATE
818 		 * MUST be specified if new route is created.
819 		 * That would keep IPv6 consistent with IPv4
820 		 */
821 		if (replace_required) {
822 			NL_SET_ERR_MSG(extack,
823 				       "Can not replace route - no match found");
824 			pr_warn("Can't replace route, no match found\n");
825 			return ERR_PTR(-ENOENT);
826 		}
827 		pr_warn("NLM_F_CREATE should be set when creating new route\n");
828 	}
829 	/*
830 	 *	We walked to the bottom of tree.
831 	 *	Create new leaf node without children.
832 	 */
833 
834 	ln = node_alloc(net);
835 
836 	if (!ln)
837 		return ERR_PTR(-ENOMEM);
838 	ln->fn_bit = plen;
839 	RCU_INIT_POINTER(ln->parent, pn);
840 
841 	if (dir)
842 		rcu_assign_pointer(pn->right, ln);
843 	else
844 		rcu_assign_pointer(pn->left, ln);
845 
846 	return ln;
847 
848 
849 insert_above:
850 	/*
851 	 * split since we don't have a common prefix anymore or
852 	 * we have a less significant route.
853 	 * we've to insert an intermediate node on the list
854 	 * this new node will point to the one we need to create
855 	 * and the current
856 	 */
857 
858 	pn = rcu_dereference_protected(fn->parent,
859 				       lockdep_is_held(&table->tb6_lock));
860 
861 	/* find 1st bit in difference between the 2 addrs.
862 
863 	   See comment in __ipv6_addr_diff: bit may be an invalid value,
864 	   but if it is >= plen, the value is ignored in any case.
865 	 */
866 
867 	bit = __ipv6_addr_diff(addr, &key->addr, sizeof(*addr));
868 
869 	/*
870 	 *		(intermediate)[in]
871 	 *	          /	   \
872 	 *	(new leaf node)[ln] (old node)[fn]
873 	 */
874 	if (plen > bit) {
875 		in = node_alloc(net);
876 		ln = node_alloc(net);
877 
878 		if (!in || !ln) {
879 			if (in)
880 				node_free_immediate(net, in);
881 			if (ln)
882 				node_free_immediate(net, ln);
883 			return ERR_PTR(-ENOMEM);
884 		}
885 
886 		/*
887 		 * new intermediate node.
888 		 * RTN_RTINFO will
889 		 * be off since that an address that chooses one of
890 		 * the branches would not match less specific routes
891 		 * in the other branch
892 		 */
893 
894 		in->fn_bit = bit;
895 
896 		RCU_INIT_POINTER(in->parent, pn);
897 		in->leaf = fn->leaf;
898 		fib6_info_hold(rcu_dereference_protected(in->leaf,
899 				lockdep_is_held(&table->tb6_lock)));
900 
901 		/* update parent pointer */
902 		if (dir)
903 			rcu_assign_pointer(pn->right, in);
904 		else
905 			rcu_assign_pointer(pn->left, in);
906 
907 		ln->fn_bit = plen;
908 
909 		RCU_INIT_POINTER(ln->parent, in);
910 		rcu_assign_pointer(fn->parent, in);
911 
912 		if (addr_bit_set(addr, bit)) {
913 			rcu_assign_pointer(in->right, ln);
914 			rcu_assign_pointer(in->left, fn);
915 		} else {
916 			rcu_assign_pointer(in->left, ln);
917 			rcu_assign_pointer(in->right, fn);
918 		}
919 	} else { /* plen <= bit */
920 
921 		/*
922 		 *		(new leaf node)[ln]
923 		 *	          /	   \
924 		 *	     (old node)[fn] NULL
925 		 */
926 
927 		ln = node_alloc(net);
928 
929 		if (!ln)
930 			return ERR_PTR(-ENOMEM);
931 
932 		ln->fn_bit = plen;
933 
934 		RCU_INIT_POINTER(ln->parent, pn);
935 
936 		if (addr_bit_set(&key->addr, plen))
937 			RCU_INIT_POINTER(ln->right, fn);
938 		else
939 			RCU_INIT_POINTER(ln->left, fn);
940 
941 		rcu_assign_pointer(fn->parent, ln);
942 
943 		if (dir)
944 			rcu_assign_pointer(pn->right, ln);
945 		else
946 			rcu_assign_pointer(pn->left, ln);
947 	}
948 	return ln;
949 }
950 
951 static void __fib6_drop_pcpu_from(struct fib6_nh *fib6_nh,
952 				  const struct fib6_info *match,
953 				  const struct fib6_table *table)
954 {
955 	int cpu;
956 
957 	if (!fib6_nh->rt6i_pcpu)
958 		return;
959 
960 	/* release the reference to this fib entry from
961 	 * all of its cached pcpu routes
962 	 */
963 	for_each_possible_cpu(cpu) {
964 		struct rt6_info **ppcpu_rt;
965 		struct rt6_info *pcpu_rt;
966 
967 		ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
968 		pcpu_rt = *ppcpu_rt;
969 
970 		/* only dropping the 'from' reference if the cached route
971 		 * is using 'match'. The cached pcpu_rt->from only changes
972 		 * from a fib6_info to NULL (ip6_dst_destroy); it can never
973 		 * change from one fib6_info reference to another
974 		 */
975 		if (pcpu_rt && rcu_access_pointer(pcpu_rt->from) == match) {
976 			struct fib6_info *from;
977 
978 			from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
979 			fib6_info_release(from);
980 		}
981 	}
982 }
983 
984 struct fib6_nh_pcpu_arg {
985 	struct fib6_info	*from;
986 	const struct fib6_table *table;
987 };
988 
989 static int fib6_nh_drop_pcpu_from(struct fib6_nh *nh, void *_arg)
990 {
991 	struct fib6_nh_pcpu_arg *arg = _arg;
992 
993 	__fib6_drop_pcpu_from(nh, arg->from, arg->table);
994 	return 0;
995 }
996 
997 static void fib6_drop_pcpu_from(struct fib6_info *f6i,
998 				const struct fib6_table *table)
999 {
1000 	/* Make sure rt6_make_pcpu_route() wont add other percpu routes
1001 	 * while we are cleaning them here.
1002 	 */
1003 	f6i->fib6_destroying = 1;
1004 	mb(); /* paired with the cmpxchg() in rt6_make_pcpu_route() */
1005 
1006 	if (f6i->nh) {
1007 		struct fib6_nh_pcpu_arg arg = {
1008 			.from = f6i,
1009 			.table = table
1010 		};
1011 
1012 		nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_drop_pcpu_from,
1013 					 &arg);
1014 	} else {
1015 		struct fib6_nh *fib6_nh;
1016 
1017 		fib6_nh = f6i->fib6_nh;
1018 		__fib6_drop_pcpu_from(fib6_nh, f6i, table);
1019 	}
1020 }
1021 
1022 static void fib6_purge_rt(struct fib6_info *rt, struct fib6_node *fn,
1023 			  struct net *net)
1024 {
1025 	struct fib6_table *table = rt->fib6_table;
1026 
1027 	fib6_drop_pcpu_from(rt, table);
1028 
1029 	if (rt->nh && !list_empty(&rt->nh_list))
1030 		list_del_init(&rt->nh_list);
1031 
1032 	if (refcount_read(&rt->fib6_ref) != 1) {
1033 		/* This route is used as dummy address holder in some split
1034 		 * nodes. It is not leaked, but it still holds other resources,
1035 		 * which must be released in time. So, scan ascendant nodes
1036 		 * and replace dummy references to this route with references
1037 		 * to still alive ones.
1038 		 */
1039 		while (fn) {
1040 			struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
1041 					    lockdep_is_held(&table->tb6_lock));
1042 			struct fib6_info *new_leaf;
1043 			if (!(fn->fn_flags & RTN_RTINFO) && leaf == rt) {
1044 				new_leaf = fib6_find_prefix(net, table, fn);
1045 				fib6_info_hold(new_leaf);
1046 
1047 				rcu_assign_pointer(fn->leaf, new_leaf);
1048 				fib6_info_release(rt);
1049 			}
1050 			fn = rcu_dereference_protected(fn->parent,
1051 				    lockdep_is_held(&table->tb6_lock));
1052 		}
1053 	}
1054 }
1055 
1056 /*
1057  *	Insert routing information in a node.
1058  */
1059 
1060 static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
1061 			    struct nl_info *info,
1062 			    struct netlink_ext_ack *extack)
1063 {
1064 	struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
1065 				    lockdep_is_held(&rt->fib6_table->tb6_lock));
1066 	struct fib6_info *iter = NULL;
1067 	struct fib6_info __rcu **ins;
1068 	struct fib6_info __rcu **fallback_ins = NULL;
1069 	int replace = (info->nlh &&
1070 		       (info->nlh->nlmsg_flags & NLM_F_REPLACE));
1071 	int add = (!info->nlh ||
1072 		   (info->nlh->nlmsg_flags & NLM_F_CREATE));
1073 	int found = 0;
1074 	bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
1075 	bool notify_sibling_rt = false;
1076 	u16 nlflags = NLM_F_EXCL;
1077 	int err;
1078 
1079 	if (info->nlh && (info->nlh->nlmsg_flags & NLM_F_APPEND))
1080 		nlflags |= NLM_F_APPEND;
1081 
1082 	ins = &fn->leaf;
1083 
1084 	for (iter = leaf; iter;
1085 	     iter = rcu_dereference_protected(iter->fib6_next,
1086 				lockdep_is_held(&rt->fib6_table->tb6_lock))) {
1087 		/*
1088 		 *	Search for duplicates
1089 		 */
1090 
1091 		if (iter->fib6_metric == rt->fib6_metric) {
1092 			/*
1093 			 *	Same priority level
1094 			 */
1095 			if (info->nlh &&
1096 			    (info->nlh->nlmsg_flags & NLM_F_EXCL))
1097 				return -EEXIST;
1098 
1099 			nlflags &= ~NLM_F_EXCL;
1100 			if (replace) {
1101 				if (rt_can_ecmp == rt6_qualify_for_ecmp(iter)) {
1102 					found++;
1103 					break;
1104 				}
1105 				fallback_ins = fallback_ins ?: ins;
1106 				goto next_iter;
1107 			}
1108 
1109 			if (rt6_duplicate_nexthop(iter, rt)) {
1110 				if (rt->fib6_nsiblings)
1111 					rt->fib6_nsiblings = 0;
1112 				if (!(iter->fib6_flags & RTF_EXPIRES))
1113 					return -EEXIST;
1114 				if (!(rt->fib6_flags & RTF_EXPIRES))
1115 					fib6_clean_expires(iter);
1116 				else
1117 					fib6_set_expires(iter, rt->expires);
1118 
1119 				if (rt->fib6_pmtu)
1120 					fib6_metric_set(iter, RTAX_MTU,
1121 							rt->fib6_pmtu);
1122 				return -EEXIST;
1123 			}
1124 			/* If we have the same destination and the same metric,
1125 			 * but not the same gateway, then the route we try to
1126 			 * add is sibling to this route, increment our counter
1127 			 * of siblings, and later we will add our route to the
1128 			 * list.
1129 			 * Only static routes (which don't have flag
1130 			 * RTF_EXPIRES) are used for ECMPv6.
1131 			 *
1132 			 * To avoid long list, we only had siblings if the
1133 			 * route have a gateway.
1134 			 */
1135 			if (rt_can_ecmp &&
1136 			    rt6_qualify_for_ecmp(iter))
1137 				rt->fib6_nsiblings++;
1138 		}
1139 
1140 		if (iter->fib6_metric > rt->fib6_metric)
1141 			break;
1142 
1143 next_iter:
1144 		ins = &iter->fib6_next;
1145 	}
1146 
1147 	if (fallback_ins && !found) {
1148 		/* No matching route with same ecmp-able-ness found, replace
1149 		 * first matching route
1150 		 */
1151 		ins = fallback_ins;
1152 		iter = rcu_dereference_protected(*ins,
1153 				    lockdep_is_held(&rt->fib6_table->tb6_lock));
1154 		found++;
1155 	}
1156 
1157 	/* Reset round-robin state, if necessary */
1158 	if (ins == &fn->leaf)
1159 		fn->rr_ptr = NULL;
1160 
1161 	/* Link this route to others same route. */
1162 	if (rt->fib6_nsiblings) {
1163 		unsigned int fib6_nsiblings;
1164 		struct fib6_info *sibling, *temp_sibling;
1165 
1166 		/* Find the first route that have the same metric */
1167 		sibling = leaf;
1168 		notify_sibling_rt = true;
1169 		while (sibling) {
1170 			if (sibling->fib6_metric == rt->fib6_metric &&
1171 			    rt6_qualify_for_ecmp(sibling)) {
1172 				list_add_tail(&rt->fib6_siblings,
1173 					      &sibling->fib6_siblings);
1174 				break;
1175 			}
1176 			sibling = rcu_dereference_protected(sibling->fib6_next,
1177 				    lockdep_is_held(&rt->fib6_table->tb6_lock));
1178 			notify_sibling_rt = false;
1179 		}
1180 		/* For each sibling in the list, increment the counter of
1181 		 * siblings. BUG() if counters does not match, list of siblings
1182 		 * is broken!
1183 		 */
1184 		fib6_nsiblings = 0;
1185 		list_for_each_entry_safe(sibling, temp_sibling,
1186 					 &rt->fib6_siblings, fib6_siblings) {
1187 			sibling->fib6_nsiblings++;
1188 			BUG_ON(sibling->fib6_nsiblings != rt->fib6_nsiblings);
1189 			fib6_nsiblings++;
1190 		}
1191 		BUG_ON(fib6_nsiblings != rt->fib6_nsiblings);
1192 		rt6_multipath_rebalance(temp_sibling);
1193 	}
1194 
1195 	/*
1196 	 *	insert node
1197 	 */
1198 	if (!replace) {
1199 		if (!add)
1200 			pr_warn("NLM_F_CREATE should be set when creating new route\n");
1201 
1202 add:
1203 		nlflags |= NLM_F_CREATE;
1204 
1205 		/* The route should only be notified if it is the first
1206 		 * route in the node or if it is added as a sibling
1207 		 * route to the first route in the node.
1208 		 */
1209 		if (!info->skip_notify_kernel &&
1210 		    (notify_sibling_rt || ins == &fn->leaf)) {
1211 			enum fib_event_type fib_event;
1212 
1213 			if (notify_sibling_rt)
1214 				fib_event = FIB_EVENT_ENTRY_APPEND;
1215 			else
1216 				fib_event = FIB_EVENT_ENTRY_REPLACE;
1217 			err = call_fib6_entry_notifiers(info->nl_net,
1218 							fib_event, rt,
1219 							extack);
1220 			if (err) {
1221 				struct fib6_info *sibling, *next_sibling;
1222 
1223 				/* If the route has siblings, then it first
1224 				 * needs to be unlinked from them.
1225 				 */
1226 				if (!rt->fib6_nsiblings)
1227 					return err;
1228 
1229 				list_for_each_entry_safe(sibling, next_sibling,
1230 							 &rt->fib6_siblings,
1231 							 fib6_siblings)
1232 					sibling->fib6_nsiblings--;
1233 				rt->fib6_nsiblings = 0;
1234 				list_del_init(&rt->fib6_siblings);
1235 				rt6_multipath_rebalance(next_sibling);
1236 				return err;
1237 			}
1238 		}
1239 
1240 		rcu_assign_pointer(rt->fib6_next, iter);
1241 		fib6_info_hold(rt);
1242 		rcu_assign_pointer(rt->fib6_node, fn);
1243 		rcu_assign_pointer(*ins, rt);
1244 		if (!info->skip_notify)
1245 			inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
1246 		info->nl_net->ipv6.rt6_stats->fib_rt_entries++;
1247 
1248 		if (!(fn->fn_flags & RTN_RTINFO)) {
1249 			info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
1250 			fn->fn_flags |= RTN_RTINFO;
1251 		}
1252 
1253 	} else {
1254 		int nsiblings;
1255 
1256 		if (!found) {
1257 			if (add)
1258 				goto add;
1259 			pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
1260 			return -ENOENT;
1261 		}
1262 
1263 		if (!info->skip_notify_kernel && ins == &fn->leaf) {
1264 			err = call_fib6_entry_notifiers(info->nl_net,
1265 							FIB_EVENT_ENTRY_REPLACE,
1266 							rt, extack);
1267 			if (err)
1268 				return err;
1269 		}
1270 
1271 		fib6_info_hold(rt);
1272 		rcu_assign_pointer(rt->fib6_node, fn);
1273 		rt->fib6_next = iter->fib6_next;
1274 		rcu_assign_pointer(*ins, rt);
1275 		if (!info->skip_notify)
1276 			inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE);
1277 		if (!(fn->fn_flags & RTN_RTINFO)) {
1278 			info->nl_net->ipv6.rt6_stats->fib_route_nodes++;
1279 			fn->fn_flags |= RTN_RTINFO;
1280 		}
1281 		nsiblings = iter->fib6_nsiblings;
1282 		iter->fib6_node = NULL;
1283 		fib6_purge_rt(iter, fn, info->nl_net);
1284 		if (rcu_access_pointer(fn->rr_ptr) == iter)
1285 			fn->rr_ptr = NULL;
1286 		fib6_info_release(iter);
1287 
1288 		if (nsiblings) {
1289 			/* Replacing an ECMP route, remove all siblings */
1290 			ins = &rt->fib6_next;
1291 			iter = rcu_dereference_protected(*ins,
1292 				    lockdep_is_held(&rt->fib6_table->tb6_lock));
1293 			while (iter) {
1294 				if (iter->fib6_metric > rt->fib6_metric)
1295 					break;
1296 				if (rt6_qualify_for_ecmp(iter)) {
1297 					*ins = iter->fib6_next;
1298 					iter->fib6_node = NULL;
1299 					fib6_purge_rt(iter, fn, info->nl_net);
1300 					if (rcu_access_pointer(fn->rr_ptr) == iter)
1301 						fn->rr_ptr = NULL;
1302 					fib6_info_release(iter);
1303 					nsiblings--;
1304 					info->nl_net->ipv6.rt6_stats->fib_rt_entries--;
1305 				} else {
1306 					ins = &iter->fib6_next;
1307 				}
1308 				iter = rcu_dereference_protected(*ins,
1309 					lockdep_is_held(&rt->fib6_table->tb6_lock));
1310 			}
1311 			WARN_ON(nsiblings != 0);
1312 		}
1313 	}
1314 
1315 	return 0;
1316 }
1317 
1318 static void fib6_start_gc(struct net *net, struct fib6_info *rt)
1319 {
1320 	if (!timer_pending(&net->ipv6.ip6_fib_timer) &&
1321 	    (rt->fib6_flags & RTF_EXPIRES))
1322 		mod_timer(&net->ipv6.ip6_fib_timer,
1323 			  jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
1324 }
1325 
1326 void fib6_force_start_gc(struct net *net)
1327 {
1328 	if (!timer_pending(&net->ipv6.ip6_fib_timer))
1329 		mod_timer(&net->ipv6.ip6_fib_timer,
1330 			  jiffies + net->ipv6.sysctl.ip6_rt_gc_interval);
1331 }
1332 
1333 static void __fib6_update_sernum_upto_root(struct fib6_info *rt,
1334 					   int sernum)
1335 {
1336 	struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node,
1337 				lockdep_is_held(&rt->fib6_table->tb6_lock));
1338 
1339 	/* paired with smp_rmb() in rt6_get_cookie_safe() */
1340 	smp_wmb();
1341 	while (fn) {
1342 		fn->fn_sernum = sernum;
1343 		fn = rcu_dereference_protected(fn->parent,
1344 				lockdep_is_held(&rt->fib6_table->tb6_lock));
1345 	}
1346 }
1347 
1348 void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt)
1349 {
1350 	__fib6_update_sernum_upto_root(rt, fib6_new_sernum(net));
1351 }
1352 
1353 /* allow ipv4 to update sernum via ipv6_stub */
1354 void fib6_update_sernum_stub(struct net *net, struct fib6_info *f6i)
1355 {
1356 	spin_lock_bh(&f6i->fib6_table->tb6_lock);
1357 	fib6_update_sernum_upto_root(net, f6i);
1358 	spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1359 }
1360 
1361 /*
1362  *	Add routing information to the routing tree.
1363  *	<destination addr>/<source addr>
1364  *	with source addr info in sub-trees
1365  *	Need to own table->tb6_lock
1366  */
1367 
1368 int fib6_add(struct fib6_node *root, struct fib6_info *rt,
1369 	     struct nl_info *info, struct netlink_ext_ack *extack)
1370 {
1371 	struct fib6_table *table = rt->fib6_table;
1372 	struct fib6_node *fn, *pn = NULL;
1373 	int err = -ENOMEM;
1374 	int allow_create = 1;
1375 	int replace_required = 0;
1376 	int sernum = fib6_new_sernum(info->nl_net);
1377 
1378 	if (info->nlh) {
1379 		if (!(info->nlh->nlmsg_flags & NLM_F_CREATE))
1380 			allow_create = 0;
1381 		if (info->nlh->nlmsg_flags & NLM_F_REPLACE)
1382 			replace_required = 1;
1383 	}
1384 	if (!allow_create && !replace_required)
1385 		pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
1386 
1387 	fn = fib6_add_1(info->nl_net, table, root,
1388 			&rt->fib6_dst.addr, rt->fib6_dst.plen,
1389 			offsetof(struct fib6_info, fib6_dst), allow_create,
1390 			replace_required, extack);
1391 	if (IS_ERR(fn)) {
1392 		err = PTR_ERR(fn);
1393 		fn = NULL;
1394 		goto out;
1395 	}
1396 
1397 	pn = fn;
1398 
1399 #ifdef CONFIG_IPV6_SUBTREES
1400 	if (rt->fib6_src.plen) {
1401 		struct fib6_node *sn;
1402 
1403 		if (!rcu_access_pointer(fn->subtree)) {
1404 			struct fib6_node *sfn;
1405 
1406 			/*
1407 			 * Create subtree.
1408 			 *
1409 			 *		fn[main tree]
1410 			 *		|
1411 			 *		sfn[subtree root]
1412 			 *		   \
1413 			 *		    sn[new leaf node]
1414 			 */
1415 
1416 			/* Create subtree root node */
1417 			sfn = node_alloc(info->nl_net);
1418 			if (!sfn)
1419 				goto failure;
1420 
1421 			fib6_info_hold(info->nl_net->ipv6.fib6_null_entry);
1422 			rcu_assign_pointer(sfn->leaf,
1423 					   info->nl_net->ipv6.fib6_null_entry);
1424 			sfn->fn_flags = RTN_ROOT;
1425 
1426 			/* Now add the first leaf node to new subtree */
1427 
1428 			sn = fib6_add_1(info->nl_net, table, sfn,
1429 					&rt->fib6_src.addr, rt->fib6_src.plen,
1430 					offsetof(struct fib6_info, fib6_src),
1431 					allow_create, replace_required, extack);
1432 
1433 			if (IS_ERR(sn)) {
1434 				/* If it is failed, discard just allocated
1435 				   root, and then (in failure) stale node
1436 				   in main tree.
1437 				 */
1438 				node_free_immediate(info->nl_net, sfn);
1439 				err = PTR_ERR(sn);
1440 				goto failure;
1441 			}
1442 
1443 			/* Now link new subtree to main tree */
1444 			rcu_assign_pointer(sfn->parent, fn);
1445 			rcu_assign_pointer(fn->subtree, sfn);
1446 		} else {
1447 			sn = fib6_add_1(info->nl_net, table, FIB6_SUBTREE(fn),
1448 					&rt->fib6_src.addr, rt->fib6_src.plen,
1449 					offsetof(struct fib6_info, fib6_src),
1450 					allow_create, replace_required, extack);
1451 
1452 			if (IS_ERR(sn)) {
1453 				err = PTR_ERR(sn);
1454 				goto failure;
1455 			}
1456 		}
1457 
1458 		if (!rcu_access_pointer(fn->leaf)) {
1459 			if (fn->fn_flags & RTN_TL_ROOT) {
1460 				/* put back null_entry for root node */
1461 				rcu_assign_pointer(fn->leaf,
1462 					    info->nl_net->ipv6.fib6_null_entry);
1463 			} else {
1464 				fib6_info_hold(rt);
1465 				rcu_assign_pointer(fn->leaf, rt);
1466 			}
1467 		}
1468 		fn = sn;
1469 	}
1470 #endif
1471 
1472 	err = fib6_add_rt2node(fn, rt, info, extack);
1473 	if (!err) {
1474 		if (rt->nh)
1475 			list_add(&rt->nh_list, &rt->nh->f6i_list);
1476 		__fib6_update_sernum_upto_root(rt, sernum);
1477 		fib6_start_gc(info->nl_net, rt);
1478 	}
1479 
1480 out:
1481 	if (err) {
1482 #ifdef CONFIG_IPV6_SUBTREES
1483 		/*
1484 		 * If fib6_add_1 has cleared the old leaf pointer in the
1485 		 * super-tree leaf node we have to find a new one for it.
1486 		 */
1487 		if (pn != fn) {
1488 			struct fib6_info *pn_leaf =
1489 				rcu_dereference_protected(pn->leaf,
1490 				    lockdep_is_held(&table->tb6_lock));
1491 			if (pn_leaf == rt) {
1492 				pn_leaf = NULL;
1493 				RCU_INIT_POINTER(pn->leaf, NULL);
1494 				fib6_info_release(rt);
1495 			}
1496 			if (!pn_leaf && !(pn->fn_flags & RTN_RTINFO)) {
1497 				pn_leaf = fib6_find_prefix(info->nl_net, table,
1498 							   pn);
1499 #if RT6_DEBUG >= 2
1500 				if (!pn_leaf) {
1501 					WARN_ON(!pn_leaf);
1502 					pn_leaf =
1503 					    info->nl_net->ipv6.fib6_null_entry;
1504 				}
1505 #endif
1506 				fib6_info_hold(pn_leaf);
1507 				rcu_assign_pointer(pn->leaf, pn_leaf);
1508 			}
1509 		}
1510 #endif
1511 		goto failure;
1512 	} else if (fib6_requires_src(rt)) {
1513 		fib6_routes_require_src_inc(info->nl_net);
1514 	}
1515 	return err;
1516 
1517 failure:
1518 	/* fn->leaf could be NULL and fib6_repair_tree() needs to be called if:
1519 	 * 1. fn is an intermediate node and we failed to add the new
1520 	 * route to it in both subtree creation failure and fib6_add_rt2node()
1521 	 * failure case.
1522 	 * 2. fn is the root node in the table and we fail to add the first
1523 	 * default route to it.
1524 	 */
1525 	if (fn &&
1526 	    (!(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)) ||
1527 	     (fn->fn_flags & RTN_TL_ROOT &&
1528 	      !rcu_access_pointer(fn->leaf))))
1529 		fib6_repair_tree(info->nl_net, table, fn);
1530 	return err;
1531 }
1532 
1533 /*
1534  *	Routing tree lookup
1535  *
1536  */
1537 
1538 struct lookup_args {
1539 	int			offset;		/* key offset on fib6_info */
1540 	const struct in6_addr	*addr;		/* search key			*/
1541 };
1542 
1543 static struct fib6_node *fib6_node_lookup_1(struct fib6_node *root,
1544 					    struct lookup_args *args)
1545 {
1546 	struct fib6_node *fn;
1547 	__be32 dir;
1548 
1549 	if (unlikely(args->offset == 0))
1550 		return NULL;
1551 
1552 	/*
1553 	 *	Descend on a tree
1554 	 */
1555 
1556 	fn = root;
1557 
1558 	for (;;) {
1559 		struct fib6_node *next;
1560 
1561 		dir = addr_bit_set(args->addr, fn->fn_bit);
1562 
1563 		next = dir ? rcu_dereference(fn->right) :
1564 			     rcu_dereference(fn->left);
1565 
1566 		if (next) {
1567 			fn = next;
1568 			continue;
1569 		}
1570 		break;
1571 	}
1572 
1573 	while (fn) {
1574 		struct fib6_node *subtree = FIB6_SUBTREE(fn);
1575 
1576 		if (subtree || fn->fn_flags & RTN_RTINFO) {
1577 			struct fib6_info *leaf = rcu_dereference(fn->leaf);
1578 			struct rt6key *key;
1579 
1580 			if (!leaf)
1581 				goto backtrack;
1582 
1583 			key = (struct rt6key *) ((u8 *)leaf + args->offset);
1584 
1585 			if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) {
1586 #ifdef CONFIG_IPV6_SUBTREES
1587 				if (subtree) {
1588 					struct fib6_node *sfn;
1589 					sfn = fib6_node_lookup_1(subtree,
1590 								 args + 1);
1591 					if (!sfn)
1592 						goto backtrack;
1593 					fn = sfn;
1594 				}
1595 #endif
1596 				if (fn->fn_flags & RTN_RTINFO)
1597 					return fn;
1598 			}
1599 		}
1600 backtrack:
1601 		if (fn->fn_flags & RTN_ROOT)
1602 			break;
1603 
1604 		fn = rcu_dereference(fn->parent);
1605 	}
1606 
1607 	return NULL;
1608 }
1609 
1610 /* called with rcu_read_lock() held
1611  */
1612 struct fib6_node *fib6_node_lookup(struct fib6_node *root,
1613 				   const struct in6_addr *daddr,
1614 				   const struct in6_addr *saddr)
1615 {
1616 	struct fib6_node *fn;
1617 	struct lookup_args args[] = {
1618 		{
1619 			.offset = offsetof(struct fib6_info, fib6_dst),
1620 			.addr = daddr,
1621 		},
1622 #ifdef CONFIG_IPV6_SUBTREES
1623 		{
1624 			.offset = offsetof(struct fib6_info, fib6_src),
1625 			.addr = saddr,
1626 		},
1627 #endif
1628 		{
1629 			.offset = 0,	/* sentinel */
1630 		}
1631 	};
1632 
1633 	fn = fib6_node_lookup_1(root, daddr ? args : args + 1);
1634 	if (!fn || fn->fn_flags & RTN_TL_ROOT)
1635 		fn = root;
1636 
1637 	return fn;
1638 }
1639 
1640 /*
1641  *	Get node with specified destination prefix (and source prefix,
1642  *	if subtrees are used)
1643  *	exact_match == true means we try to find fn with exact match of
1644  *	the passed in prefix addr
1645  *	exact_match == false means we try to find fn with longest prefix
1646  *	match of the passed in prefix addr. This is useful for finding fn
1647  *	for cached route as it will be stored in the exception table under
1648  *	the node with longest prefix length.
1649  */
1650 
1651 
1652 static struct fib6_node *fib6_locate_1(struct fib6_node *root,
1653 				       const struct in6_addr *addr,
1654 				       int plen, int offset,
1655 				       bool exact_match)
1656 {
1657 	struct fib6_node *fn, *prev = NULL;
1658 
1659 	for (fn = root; fn ; ) {
1660 		struct fib6_info *leaf = rcu_dereference(fn->leaf);
1661 		struct rt6key *key;
1662 
1663 		/* This node is being deleted */
1664 		if (!leaf) {
1665 			if (plen <= fn->fn_bit)
1666 				goto out;
1667 			else
1668 				goto next;
1669 		}
1670 
1671 		key = (struct rt6key *)((u8 *)leaf + offset);
1672 
1673 		/*
1674 		 *	Prefix match
1675 		 */
1676 		if (plen < fn->fn_bit ||
1677 		    !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit))
1678 			goto out;
1679 
1680 		if (plen == fn->fn_bit)
1681 			return fn;
1682 
1683 		if (fn->fn_flags & RTN_RTINFO)
1684 			prev = fn;
1685 
1686 next:
1687 		/*
1688 		 *	We have more bits to go
1689 		 */
1690 		if (addr_bit_set(addr, fn->fn_bit))
1691 			fn = rcu_dereference(fn->right);
1692 		else
1693 			fn = rcu_dereference(fn->left);
1694 	}
1695 out:
1696 	if (exact_match)
1697 		return NULL;
1698 	else
1699 		return prev;
1700 }
1701 
1702 struct fib6_node *fib6_locate(struct fib6_node *root,
1703 			      const struct in6_addr *daddr, int dst_len,
1704 			      const struct in6_addr *saddr, int src_len,
1705 			      bool exact_match)
1706 {
1707 	struct fib6_node *fn;
1708 
1709 	fn = fib6_locate_1(root, daddr, dst_len,
1710 			   offsetof(struct fib6_info, fib6_dst),
1711 			   exact_match);
1712 
1713 #ifdef CONFIG_IPV6_SUBTREES
1714 	if (src_len) {
1715 		WARN_ON(saddr == NULL);
1716 		if (fn) {
1717 			struct fib6_node *subtree = FIB6_SUBTREE(fn);
1718 
1719 			if (subtree) {
1720 				fn = fib6_locate_1(subtree, saddr, src_len,
1721 					   offsetof(struct fib6_info, fib6_src),
1722 					   exact_match);
1723 			}
1724 		}
1725 	}
1726 #endif
1727 
1728 	if (fn && fn->fn_flags & RTN_RTINFO)
1729 		return fn;
1730 
1731 	return NULL;
1732 }
1733 
1734 
1735 /*
1736  *	Deletion
1737  *
1738  */
1739 
1740 static struct fib6_info *fib6_find_prefix(struct net *net,
1741 					 struct fib6_table *table,
1742 					 struct fib6_node *fn)
1743 {
1744 	struct fib6_node *child_left, *child_right;
1745 
1746 	if (fn->fn_flags & RTN_ROOT)
1747 		return net->ipv6.fib6_null_entry;
1748 
1749 	while (fn) {
1750 		child_left = rcu_dereference_protected(fn->left,
1751 				    lockdep_is_held(&table->tb6_lock));
1752 		child_right = rcu_dereference_protected(fn->right,
1753 				    lockdep_is_held(&table->tb6_lock));
1754 		if (child_left)
1755 			return rcu_dereference_protected(child_left->leaf,
1756 					lockdep_is_held(&table->tb6_lock));
1757 		if (child_right)
1758 			return rcu_dereference_protected(child_right->leaf,
1759 					lockdep_is_held(&table->tb6_lock));
1760 
1761 		fn = FIB6_SUBTREE(fn);
1762 	}
1763 	return NULL;
1764 }
1765 
1766 /*
1767  *	Called to trim the tree of intermediate nodes when possible. "fn"
1768  *	is the node we want to try and remove.
1769  *	Need to own table->tb6_lock
1770  */
1771 
1772 static struct fib6_node *fib6_repair_tree(struct net *net,
1773 					  struct fib6_table *table,
1774 					  struct fib6_node *fn)
1775 {
1776 	int children;
1777 	int nstate;
1778 	struct fib6_node *child;
1779 	struct fib6_walker *w;
1780 	int iter = 0;
1781 
1782 	/* Set fn->leaf to null_entry for root node. */
1783 	if (fn->fn_flags & RTN_TL_ROOT) {
1784 		rcu_assign_pointer(fn->leaf, net->ipv6.fib6_null_entry);
1785 		return fn;
1786 	}
1787 
1788 	for (;;) {
1789 		struct fib6_node *fn_r = rcu_dereference_protected(fn->right,
1790 					    lockdep_is_held(&table->tb6_lock));
1791 		struct fib6_node *fn_l = rcu_dereference_protected(fn->left,
1792 					    lockdep_is_held(&table->tb6_lock));
1793 		struct fib6_node *pn = rcu_dereference_protected(fn->parent,
1794 					    lockdep_is_held(&table->tb6_lock));
1795 		struct fib6_node *pn_r = rcu_dereference_protected(pn->right,
1796 					    lockdep_is_held(&table->tb6_lock));
1797 		struct fib6_node *pn_l = rcu_dereference_protected(pn->left,
1798 					    lockdep_is_held(&table->tb6_lock));
1799 		struct fib6_info *fn_leaf = rcu_dereference_protected(fn->leaf,
1800 					    lockdep_is_held(&table->tb6_lock));
1801 		struct fib6_info *pn_leaf = rcu_dereference_protected(pn->leaf,
1802 					    lockdep_is_held(&table->tb6_lock));
1803 		struct fib6_info *new_fn_leaf;
1804 
1805 		RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter);
1806 		iter++;
1807 
1808 		WARN_ON(fn->fn_flags & RTN_RTINFO);
1809 		WARN_ON(fn->fn_flags & RTN_TL_ROOT);
1810 		WARN_ON(fn_leaf);
1811 
1812 		children = 0;
1813 		child = NULL;
1814 		if (fn_r)
1815 			child = fn_r, children |= 1;
1816 		if (fn_l)
1817 			child = fn_l, children |= 2;
1818 
1819 		if (children == 3 || FIB6_SUBTREE(fn)
1820 #ifdef CONFIG_IPV6_SUBTREES
1821 		    /* Subtree root (i.e. fn) may have one child */
1822 		    || (children && fn->fn_flags & RTN_ROOT)
1823 #endif
1824 		    ) {
1825 			new_fn_leaf = fib6_find_prefix(net, table, fn);
1826 #if RT6_DEBUG >= 2
1827 			if (!new_fn_leaf) {
1828 				WARN_ON(!new_fn_leaf);
1829 				new_fn_leaf = net->ipv6.fib6_null_entry;
1830 			}
1831 #endif
1832 			fib6_info_hold(new_fn_leaf);
1833 			rcu_assign_pointer(fn->leaf, new_fn_leaf);
1834 			return pn;
1835 		}
1836 
1837 #ifdef CONFIG_IPV6_SUBTREES
1838 		if (FIB6_SUBTREE(pn) == fn) {
1839 			WARN_ON(!(fn->fn_flags & RTN_ROOT));
1840 			RCU_INIT_POINTER(pn->subtree, NULL);
1841 			nstate = FWS_L;
1842 		} else {
1843 			WARN_ON(fn->fn_flags & RTN_ROOT);
1844 #endif
1845 			if (pn_r == fn)
1846 				rcu_assign_pointer(pn->right, child);
1847 			else if (pn_l == fn)
1848 				rcu_assign_pointer(pn->left, child);
1849 #if RT6_DEBUG >= 2
1850 			else
1851 				WARN_ON(1);
1852 #endif
1853 			if (child)
1854 				rcu_assign_pointer(child->parent, pn);
1855 			nstate = FWS_R;
1856 #ifdef CONFIG_IPV6_SUBTREES
1857 		}
1858 #endif
1859 
1860 		read_lock(&net->ipv6.fib6_walker_lock);
1861 		FOR_WALKERS(net, w) {
1862 			if (!child) {
1863 				if (w->node == fn) {
1864 					RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate);
1865 					w->node = pn;
1866 					w->state = nstate;
1867 				}
1868 			} else {
1869 				if (w->node == fn) {
1870 					w->node = child;
1871 					if (children&2) {
1872 						RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1873 						w->state = w->state >= FWS_R ? FWS_U : FWS_INIT;
1874 					} else {
1875 						RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state);
1876 						w->state = w->state >= FWS_C ? FWS_U : FWS_INIT;
1877 					}
1878 				}
1879 			}
1880 		}
1881 		read_unlock(&net->ipv6.fib6_walker_lock);
1882 
1883 		node_free(net, fn);
1884 		if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn))
1885 			return pn;
1886 
1887 		RCU_INIT_POINTER(pn->leaf, NULL);
1888 		fib6_info_release(pn_leaf);
1889 		fn = pn;
1890 	}
1891 }
1892 
1893 static void fib6_del_route(struct fib6_table *table, struct fib6_node *fn,
1894 			   struct fib6_info __rcu **rtp, struct nl_info *info)
1895 {
1896 	struct fib6_info *leaf, *replace_rt = NULL;
1897 	struct fib6_walker *w;
1898 	struct fib6_info *rt = rcu_dereference_protected(*rtp,
1899 				    lockdep_is_held(&table->tb6_lock));
1900 	struct net *net = info->nl_net;
1901 	bool notify_del = false;
1902 
1903 	RT6_TRACE("fib6_del_route\n");
1904 
1905 	/* If the deleted route is the first in the node and it is not part of
1906 	 * a multipath route, then we need to replace it with the next route
1907 	 * in the node, if exists.
1908 	 */
1909 	leaf = rcu_dereference_protected(fn->leaf,
1910 					 lockdep_is_held(&table->tb6_lock));
1911 	if (leaf == rt && !rt->fib6_nsiblings) {
1912 		if (rcu_access_pointer(rt->fib6_next))
1913 			replace_rt = rcu_dereference_protected(rt->fib6_next,
1914 					    lockdep_is_held(&table->tb6_lock));
1915 		else
1916 			notify_del = true;
1917 	}
1918 
1919 	/* Unlink it */
1920 	*rtp = rt->fib6_next;
1921 	rt->fib6_node = NULL;
1922 	net->ipv6.rt6_stats->fib_rt_entries--;
1923 	net->ipv6.rt6_stats->fib_discarded_routes++;
1924 
1925 	/* Flush all cached dst in exception table */
1926 	rt6_flush_exceptions(rt);
1927 
1928 	/* Reset round-robin state, if necessary */
1929 	if (rcu_access_pointer(fn->rr_ptr) == rt)
1930 		fn->rr_ptr = NULL;
1931 
1932 	/* Remove this entry from other siblings */
1933 	if (rt->fib6_nsiblings) {
1934 		struct fib6_info *sibling, *next_sibling;
1935 
1936 		/* The route is deleted from a multipath route. If this
1937 		 * multipath route is the first route in the node, then we need
1938 		 * to emit a delete notification. Otherwise, we need to skip
1939 		 * the notification.
1940 		 */
1941 		if (rt->fib6_metric == leaf->fib6_metric &&
1942 		    rt6_qualify_for_ecmp(leaf))
1943 			notify_del = true;
1944 		list_for_each_entry_safe(sibling, next_sibling,
1945 					 &rt->fib6_siblings, fib6_siblings)
1946 			sibling->fib6_nsiblings--;
1947 		rt->fib6_nsiblings = 0;
1948 		list_del_init(&rt->fib6_siblings);
1949 		rt6_multipath_rebalance(next_sibling);
1950 	}
1951 
1952 	/* Adjust walkers */
1953 	read_lock(&net->ipv6.fib6_walker_lock);
1954 	FOR_WALKERS(net, w) {
1955 		if (w->state == FWS_C && w->leaf == rt) {
1956 			RT6_TRACE("walker %p adjusted by delroute\n", w);
1957 			w->leaf = rcu_dereference_protected(rt->fib6_next,
1958 					    lockdep_is_held(&table->tb6_lock));
1959 			if (!w->leaf)
1960 				w->state = FWS_U;
1961 		}
1962 	}
1963 	read_unlock(&net->ipv6.fib6_walker_lock);
1964 
1965 	/* If it was last route, call fib6_repair_tree() to:
1966 	 * 1. For root node, put back null_entry as how the table was created.
1967 	 * 2. For other nodes, expunge its radix tree node.
1968 	 */
1969 	if (!rcu_access_pointer(fn->leaf)) {
1970 		if (!(fn->fn_flags & RTN_TL_ROOT)) {
1971 			fn->fn_flags &= ~RTN_RTINFO;
1972 			net->ipv6.rt6_stats->fib_route_nodes--;
1973 		}
1974 		fn = fib6_repair_tree(net, table, fn);
1975 	}
1976 
1977 	fib6_purge_rt(rt, fn, net);
1978 
1979 	if (!info->skip_notify_kernel) {
1980 		if (notify_del)
1981 			call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_DEL,
1982 						  rt, NULL);
1983 		else if (replace_rt)
1984 			call_fib6_entry_notifiers_replace(net, replace_rt);
1985 	}
1986 	if (!info->skip_notify)
1987 		inet6_rt_notify(RTM_DELROUTE, rt, info, 0);
1988 
1989 	fib6_info_release(rt);
1990 }
1991 
1992 /* Need to own table->tb6_lock */
1993 int fib6_del(struct fib6_info *rt, struct nl_info *info)
1994 {
1995 	struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node,
1996 				    lockdep_is_held(&rt->fib6_table->tb6_lock));
1997 	struct fib6_table *table = rt->fib6_table;
1998 	struct net *net = info->nl_net;
1999 	struct fib6_info __rcu **rtp;
2000 	struct fib6_info __rcu **rtp_next;
2001 
2002 	if (!fn || rt == net->ipv6.fib6_null_entry)
2003 		return -ENOENT;
2004 
2005 	WARN_ON(!(fn->fn_flags & RTN_RTINFO));
2006 
2007 	/*
2008 	 *	Walk the leaf entries looking for ourself
2009 	 */
2010 
2011 	for (rtp = &fn->leaf; *rtp; rtp = rtp_next) {
2012 		struct fib6_info *cur = rcu_dereference_protected(*rtp,
2013 					lockdep_is_held(&table->tb6_lock));
2014 		if (rt == cur) {
2015 			if (fib6_requires_src(cur))
2016 				fib6_routes_require_src_dec(info->nl_net);
2017 			fib6_del_route(table, fn, rtp, info);
2018 			return 0;
2019 		}
2020 		rtp_next = &cur->fib6_next;
2021 	}
2022 	return -ENOENT;
2023 }
2024 
2025 /*
2026  *	Tree traversal function.
2027  *
2028  *	Certainly, it is not interrupt safe.
2029  *	However, it is internally reenterable wrt itself and fib6_add/fib6_del.
2030  *	It means, that we can modify tree during walking
2031  *	and use this function for garbage collection, clone pruning,
2032  *	cleaning tree when a device goes down etc. etc.
2033  *
2034  *	It guarantees that every node will be traversed,
2035  *	and that it will be traversed only once.
2036  *
2037  *	Callback function w->func may return:
2038  *	0 -> continue walking.
2039  *	positive value -> walking is suspended (used by tree dumps,
2040  *	and probably by gc, if it will be split to several slices)
2041  *	negative value -> terminate walking.
2042  *
2043  *	The function itself returns:
2044  *	0   -> walk is complete.
2045  *	>0  -> walk is incomplete (i.e. suspended)
2046  *	<0  -> walk is terminated by an error.
2047  *
2048  *	This function is called with tb6_lock held.
2049  */
2050 
2051 static int fib6_walk_continue(struct fib6_walker *w)
2052 {
2053 	struct fib6_node *fn, *pn, *left, *right;
2054 
2055 	/* w->root should always be table->tb6_root */
2056 	WARN_ON_ONCE(!(w->root->fn_flags & RTN_TL_ROOT));
2057 
2058 	for (;;) {
2059 		fn = w->node;
2060 		if (!fn)
2061 			return 0;
2062 
2063 		switch (w->state) {
2064 #ifdef CONFIG_IPV6_SUBTREES
2065 		case FWS_S:
2066 			if (FIB6_SUBTREE(fn)) {
2067 				w->node = FIB6_SUBTREE(fn);
2068 				continue;
2069 			}
2070 			w->state = FWS_L;
2071 #endif
2072 			/* fall through */
2073 		case FWS_L:
2074 			left = rcu_dereference_protected(fn->left, 1);
2075 			if (left) {
2076 				w->node = left;
2077 				w->state = FWS_INIT;
2078 				continue;
2079 			}
2080 			w->state = FWS_R;
2081 			/* fall through */
2082 		case FWS_R:
2083 			right = rcu_dereference_protected(fn->right, 1);
2084 			if (right) {
2085 				w->node = right;
2086 				w->state = FWS_INIT;
2087 				continue;
2088 			}
2089 			w->state = FWS_C;
2090 			w->leaf = rcu_dereference_protected(fn->leaf, 1);
2091 			/* fall through */
2092 		case FWS_C:
2093 			if (w->leaf && fn->fn_flags & RTN_RTINFO) {
2094 				int err;
2095 
2096 				if (w->skip) {
2097 					w->skip--;
2098 					goto skip;
2099 				}
2100 
2101 				err = w->func(w);
2102 				if (err)
2103 					return err;
2104 
2105 				w->count++;
2106 				continue;
2107 			}
2108 skip:
2109 			w->state = FWS_U;
2110 			/* fall through */
2111 		case FWS_U:
2112 			if (fn == w->root)
2113 				return 0;
2114 			pn = rcu_dereference_protected(fn->parent, 1);
2115 			left = rcu_dereference_protected(pn->left, 1);
2116 			right = rcu_dereference_protected(pn->right, 1);
2117 			w->node = pn;
2118 #ifdef CONFIG_IPV6_SUBTREES
2119 			if (FIB6_SUBTREE(pn) == fn) {
2120 				WARN_ON(!(fn->fn_flags & RTN_ROOT));
2121 				w->state = FWS_L;
2122 				continue;
2123 			}
2124 #endif
2125 			if (left == fn) {
2126 				w->state = FWS_R;
2127 				continue;
2128 			}
2129 			if (right == fn) {
2130 				w->state = FWS_C;
2131 				w->leaf = rcu_dereference_protected(w->node->leaf, 1);
2132 				continue;
2133 			}
2134 #if RT6_DEBUG >= 2
2135 			WARN_ON(1);
2136 #endif
2137 		}
2138 	}
2139 }
2140 
2141 static int fib6_walk(struct net *net, struct fib6_walker *w)
2142 {
2143 	int res;
2144 
2145 	w->state = FWS_INIT;
2146 	w->node = w->root;
2147 
2148 	fib6_walker_link(net, w);
2149 	res = fib6_walk_continue(w);
2150 	if (res <= 0)
2151 		fib6_walker_unlink(net, w);
2152 	return res;
2153 }
2154 
2155 static int fib6_clean_node(struct fib6_walker *w)
2156 {
2157 	int res;
2158 	struct fib6_info *rt;
2159 	struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w);
2160 	struct nl_info info = {
2161 		.nl_net = c->net,
2162 		.skip_notify = c->skip_notify,
2163 	};
2164 
2165 	if (c->sernum != FIB6_NO_SERNUM_CHANGE &&
2166 	    w->node->fn_sernum != c->sernum)
2167 		w->node->fn_sernum = c->sernum;
2168 
2169 	if (!c->func) {
2170 		WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE);
2171 		w->leaf = NULL;
2172 		return 0;
2173 	}
2174 
2175 	for_each_fib6_walker_rt(w) {
2176 		res = c->func(rt, c->arg);
2177 		if (res == -1) {
2178 			w->leaf = rt;
2179 			res = fib6_del(rt, &info);
2180 			if (res) {
2181 #if RT6_DEBUG >= 2
2182 				pr_debug("%s: del failed: rt=%p@%p err=%d\n",
2183 					 __func__, rt,
2184 					 rcu_access_pointer(rt->fib6_node),
2185 					 res);
2186 #endif
2187 				continue;
2188 			}
2189 			return 0;
2190 		} else if (res == -2) {
2191 			if (WARN_ON(!rt->fib6_nsiblings))
2192 				continue;
2193 			rt = list_last_entry(&rt->fib6_siblings,
2194 					     struct fib6_info, fib6_siblings);
2195 			continue;
2196 		}
2197 		WARN_ON(res != 0);
2198 	}
2199 	w->leaf = rt;
2200 	return 0;
2201 }
2202 
2203 /*
2204  *	Convenient frontend to tree walker.
2205  *
2206  *	func is called on each route.
2207  *		It may return -2 -> skip multipath route.
2208  *			      -1 -> delete this route.
2209  *		              0  -> continue walking
2210  */
2211 
2212 static void fib6_clean_tree(struct net *net, struct fib6_node *root,
2213 			    int (*func)(struct fib6_info *, void *arg),
2214 			    int sernum, void *arg, bool skip_notify)
2215 {
2216 	struct fib6_cleaner c;
2217 
2218 	c.w.root = root;
2219 	c.w.func = fib6_clean_node;
2220 	c.w.count = 0;
2221 	c.w.skip = 0;
2222 	c.w.skip_in_node = 0;
2223 	c.func = func;
2224 	c.sernum = sernum;
2225 	c.arg = arg;
2226 	c.net = net;
2227 	c.skip_notify = skip_notify;
2228 
2229 	fib6_walk(net, &c.w);
2230 }
2231 
2232 static void __fib6_clean_all(struct net *net,
2233 			     int (*func)(struct fib6_info *, void *),
2234 			     int sernum, void *arg, bool skip_notify)
2235 {
2236 	struct fib6_table *table;
2237 	struct hlist_head *head;
2238 	unsigned int h;
2239 
2240 	rcu_read_lock();
2241 	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2242 		head = &net->ipv6.fib_table_hash[h];
2243 		hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2244 			spin_lock_bh(&table->tb6_lock);
2245 			fib6_clean_tree(net, &table->tb6_root,
2246 					func, sernum, arg, skip_notify);
2247 			spin_unlock_bh(&table->tb6_lock);
2248 		}
2249 	}
2250 	rcu_read_unlock();
2251 }
2252 
2253 void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *),
2254 		    void *arg)
2255 {
2256 	__fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, false);
2257 }
2258 
2259 void fib6_clean_all_skip_notify(struct net *net,
2260 				int (*func)(struct fib6_info *, void *),
2261 				void *arg)
2262 {
2263 	__fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, true);
2264 }
2265 
2266 static void fib6_flush_trees(struct net *net)
2267 {
2268 	int new_sernum = fib6_new_sernum(net);
2269 
2270 	__fib6_clean_all(net, NULL, new_sernum, NULL, false);
2271 }
2272 
2273 /*
2274  *	Garbage collection
2275  */
2276 
2277 static int fib6_age(struct fib6_info *rt, void *arg)
2278 {
2279 	struct fib6_gc_args *gc_args = arg;
2280 	unsigned long now = jiffies;
2281 
2282 	/*
2283 	 *	check addrconf expiration here.
2284 	 *	Routes are expired even if they are in use.
2285 	 */
2286 
2287 	if (rt->fib6_flags & RTF_EXPIRES && rt->expires) {
2288 		if (time_after(now, rt->expires)) {
2289 			RT6_TRACE("expiring %p\n", rt);
2290 			return -1;
2291 		}
2292 		gc_args->more++;
2293 	}
2294 
2295 	/*	Also age clones in the exception table.
2296 	 *	Note, that clones are aged out
2297 	 *	only if they are not in use now.
2298 	 */
2299 	rt6_age_exceptions(rt, gc_args, now);
2300 
2301 	return 0;
2302 }
2303 
2304 void fib6_run_gc(unsigned long expires, struct net *net, bool force)
2305 {
2306 	struct fib6_gc_args gc_args;
2307 	unsigned long now;
2308 
2309 	if (force) {
2310 		spin_lock_bh(&net->ipv6.fib6_gc_lock);
2311 	} else if (!spin_trylock_bh(&net->ipv6.fib6_gc_lock)) {
2312 		mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ);
2313 		return;
2314 	}
2315 	gc_args.timeout = expires ? (int)expires :
2316 			  net->ipv6.sysctl.ip6_rt_gc_interval;
2317 	gc_args.more = 0;
2318 
2319 	fib6_clean_all(net, fib6_age, &gc_args);
2320 	now = jiffies;
2321 	net->ipv6.ip6_rt_last_gc = now;
2322 
2323 	if (gc_args.more)
2324 		mod_timer(&net->ipv6.ip6_fib_timer,
2325 			  round_jiffies(now
2326 					+ net->ipv6.sysctl.ip6_rt_gc_interval));
2327 	else
2328 		del_timer(&net->ipv6.ip6_fib_timer);
2329 	spin_unlock_bh(&net->ipv6.fib6_gc_lock);
2330 }
2331 
2332 static void fib6_gc_timer_cb(struct timer_list *t)
2333 {
2334 	struct net *arg = from_timer(arg, t, ipv6.ip6_fib_timer);
2335 
2336 	fib6_run_gc(0, arg, true);
2337 }
2338 
2339 static int __net_init fib6_net_init(struct net *net)
2340 {
2341 	size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ;
2342 	int err;
2343 
2344 	err = fib6_notifier_init(net);
2345 	if (err)
2346 		return err;
2347 
2348 	spin_lock_init(&net->ipv6.fib6_gc_lock);
2349 	rwlock_init(&net->ipv6.fib6_walker_lock);
2350 	INIT_LIST_HEAD(&net->ipv6.fib6_walkers);
2351 	timer_setup(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, 0);
2352 
2353 	net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL);
2354 	if (!net->ipv6.rt6_stats)
2355 		goto out_timer;
2356 
2357 	/* Avoid false sharing : Use at least a full cache line */
2358 	size = max_t(size_t, size, L1_CACHE_BYTES);
2359 
2360 	net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL);
2361 	if (!net->ipv6.fib_table_hash)
2362 		goto out_rt6_stats;
2363 
2364 	net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl),
2365 					  GFP_KERNEL);
2366 	if (!net->ipv6.fib6_main_tbl)
2367 		goto out_fib_table_hash;
2368 
2369 	net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN;
2370 	rcu_assign_pointer(net->ipv6.fib6_main_tbl->tb6_root.leaf,
2371 			   net->ipv6.fib6_null_entry);
2372 	net->ipv6.fib6_main_tbl->tb6_root.fn_flags =
2373 		RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
2374 	inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers);
2375 
2376 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2377 	net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl),
2378 					   GFP_KERNEL);
2379 	if (!net->ipv6.fib6_local_tbl)
2380 		goto out_fib6_main_tbl;
2381 	net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL;
2382 	rcu_assign_pointer(net->ipv6.fib6_local_tbl->tb6_root.leaf,
2383 			   net->ipv6.fib6_null_entry);
2384 	net->ipv6.fib6_local_tbl->tb6_root.fn_flags =
2385 		RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO;
2386 	inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers);
2387 #endif
2388 	fib6_tables_init(net);
2389 
2390 	return 0;
2391 
2392 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2393 out_fib6_main_tbl:
2394 	kfree(net->ipv6.fib6_main_tbl);
2395 #endif
2396 out_fib_table_hash:
2397 	kfree(net->ipv6.fib_table_hash);
2398 out_rt6_stats:
2399 	kfree(net->ipv6.rt6_stats);
2400 out_timer:
2401 	fib6_notifier_exit(net);
2402 	return -ENOMEM;
2403 }
2404 
2405 static void fib6_net_exit(struct net *net)
2406 {
2407 	unsigned int i;
2408 
2409 	del_timer_sync(&net->ipv6.ip6_fib_timer);
2410 
2411 	for (i = 0; i < FIB6_TABLE_HASHSZ; i++) {
2412 		struct hlist_head *head = &net->ipv6.fib_table_hash[i];
2413 		struct hlist_node *tmp;
2414 		struct fib6_table *tb;
2415 
2416 		hlist_for_each_entry_safe(tb, tmp, head, tb6_hlist) {
2417 			hlist_del(&tb->tb6_hlist);
2418 			fib6_free_table(tb);
2419 		}
2420 	}
2421 
2422 	kfree(net->ipv6.fib_table_hash);
2423 	kfree(net->ipv6.rt6_stats);
2424 	fib6_notifier_exit(net);
2425 }
2426 
2427 static struct pernet_operations fib6_net_ops = {
2428 	.init = fib6_net_init,
2429 	.exit = fib6_net_exit,
2430 };
2431 
2432 int __init fib6_init(void)
2433 {
2434 	int ret = -ENOMEM;
2435 
2436 	fib6_node_kmem = kmem_cache_create("fib6_nodes",
2437 					   sizeof(struct fib6_node),
2438 					   0, SLAB_HWCACHE_ALIGN,
2439 					   NULL);
2440 	if (!fib6_node_kmem)
2441 		goto out;
2442 
2443 	ret = register_pernet_subsys(&fib6_net_ops);
2444 	if (ret)
2445 		goto out_kmem_cache_create;
2446 
2447 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE, NULL,
2448 				   inet6_dump_fib, 0);
2449 	if (ret)
2450 		goto out_unregister_subsys;
2451 
2452 	__fib6_flush_trees = fib6_flush_trees;
2453 out:
2454 	return ret;
2455 
2456 out_unregister_subsys:
2457 	unregister_pernet_subsys(&fib6_net_ops);
2458 out_kmem_cache_create:
2459 	kmem_cache_destroy(fib6_node_kmem);
2460 	goto out;
2461 }
2462 
2463 void fib6_gc_cleanup(void)
2464 {
2465 	unregister_pernet_subsys(&fib6_net_ops);
2466 	kmem_cache_destroy(fib6_node_kmem);
2467 }
2468 
2469 #ifdef CONFIG_PROC_FS
2470 static int ipv6_route_seq_show(struct seq_file *seq, void *v)
2471 {
2472 	struct fib6_info *rt = v;
2473 	struct ipv6_route_iter *iter = seq->private;
2474 	struct fib6_nh *fib6_nh = rt->fib6_nh;
2475 	unsigned int flags = rt->fib6_flags;
2476 	const struct net_device *dev;
2477 
2478 	if (rt->nh)
2479 		fib6_nh = nexthop_fib6_nh(rt->nh);
2480 
2481 	seq_printf(seq, "%pi6 %02x ", &rt->fib6_dst.addr, rt->fib6_dst.plen);
2482 
2483 #ifdef CONFIG_IPV6_SUBTREES
2484 	seq_printf(seq, "%pi6 %02x ", &rt->fib6_src.addr, rt->fib6_src.plen);
2485 #else
2486 	seq_puts(seq, "00000000000000000000000000000000 00 ");
2487 #endif
2488 	if (fib6_nh->fib_nh_gw_family) {
2489 		flags |= RTF_GATEWAY;
2490 		seq_printf(seq, "%pi6", &fib6_nh->fib_nh_gw6);
2491 	} else {
2492 		seq_puts(seq, "00000000000000000000000000000000");
2493 	}
2494 
2495 	dev = fib6_nh->fib_nh_dev;
2496 	seq_printf(seq, " %08x %08x %08x %08x %8s\n",
2497 		   rt->fib6_metric, refcount_read(&rt->fib6_ref), 0,
2498 		   flags, dev ? dev->name : "");
2499 	iter->w.leaf = NULL;
2500 	return 0;
2501 }
2502 
2503 static int ipv6_route_yield(struct fib6_walker *w)
2504 {
2505 	struct ipv6_route_iter *iter = w->args;
2506 
2507 	if (!iter->skip)
2508 		return 1;
2509 
2510 	do {
2511 		iter->w.leaf = rcu_dereference_protected(
2512 				iter->w.leaf->fib6_next,
2513 				lockdep_is_held(&iter->tbl->tb6_lock));
2514 		iter->skip--;
2515 		if (!iter->skip && iter->w.leaf)
2516 			return 1;
2517 	} while (iter->w.leaf);
2518 
2519 	return 0;
2520 }
2521 
2522 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter,
2523 				      struct net *net)
2524 {
2525 	memset(&iter->w, 0, sizeof(iter->w));
2526 	iter->w.func = ipv6_route_yield;
2527 	iter->w.root = &iter->tbl->tb6_root;
2528 	iter->w.state = FWS_INIT;
2529 	iter->w.node = iter->w.root;
2530 	iter->w.args = iter;
2531 	iter->sernum = iter->w.root->fn_sernum;
2532 	INIT_LIST_HEAD(&iter->w.lh);
2533 	fib6_walker_link(net, &iter->w);
2534 }
2535 
2536 static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl,
2537 						    struct net *net)
2538 {
2539 	unsigned int h;
2540 	struct hlist_node *node;
2541 
2542 	if (tbl) {
2543 		h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1;
2544 		node = rcu_dereference_bh(hlist_next_rcu(&tbl->tb6_hlist));
2545 	} else {
2546 		h = 0;
2547 		node = NULL;
2548 	}
2549 
2550 	while (!node && h < FIB6_TABLE_HASHSZ) {
2551 		node = rcu_dereference_bh(
2552 			hlist_first_rcu(&net->ipv6.fib_table_hash[h++]));
2553 	}
2554 	return hlist_entry_safe(node, struct fib6_table, tb6_hlist);
2555 }
2556 
2557 static void ipv6_route_check_sernum(struct ipv6_route_iter *iter)
2558 {
2559 	if (iter->sernum != iter->w.root->fn_sernum) {
2560 		iter->sernum = iter->w.root->fn_sernum;
2561 		iter->w.state = FWS_INIT;
2562 		iter->w.node = iter->w.root;
2563 		WARN_ON(iter->w.skip);
2564 		iter->w.skip = iter->w.count;
2565 	}
2566 }
2567 
2568 static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2569 {
2570 	int r;
2571 	struct fib6_info *n;
2572 	struct net *net = seq_file_net(seq);
2573 	struct ipv6_route_iter *iter = seq->private;
2574 
2575 	++(*pos);
2576 	if (!v)
2577 		goto iter_table;
2578 
2579 	n = rcu_dereference_bh(((struct fib6_info *)v)->fib6_next);
2580 	if (n)
2581 		return n;
2582 
2583 iter_table:
2584 	ipv6_route_check_sernum(iter);
2585 	spin_lock_bh(&iter->tbl->tb6_lock);
2586 	r = fib6_walk_continue(&iter->w);
2587 	spin_unlock_bh(&iter->tbl->tb6_lock);
2588 	if (r > 0) {
2589 		return iter->w.leaf;
2590 	} else if (r < 0) {
2591 		fib6_walker_unlink(net, &iter->w);
2592 		return NULL;
2593 	}
2594 	fib6_walker_unlink(net, &iter->w);
2595 
2596 	iter->tbl = ipv6_route_seq_next_table(iter->tbl, net);
2597 	if (!iter->tbl)
2598 		return NULL;
2599 
2600 	ipv6_route_seq_setup_walk(iter, net);
2601 	goto iter_table;
2602 }
2603 
2604 static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos)
2605 	__acquires(RCU_BH)
2606 {
2607 	struct net *net = seq_file_net(seq);
2608 	struct ipv6_route_iter *iter = seq->private;
2609 
2610 	rcu_read_lock_bh();
2611 	iter->tbl = ipv6_route_seq_next_table(NULL, net);
2612 	iter->skip = *pos;
2613 
2614 	if (iter->tbl) {
2615 		ipv6_route_seq_setup_walk(iter, net);
2616 		return ipv6_route_seq_next(seq, NULL, pos);
2617 	} else {
2618 		return NULL;
2619 	}
2620 }
2621 
2622 static bool ipv6_route_iter_active(struct ipv6_route_iter *iter)
2623 {
2624 	struct fib6_walker *w = &iter->w;
2625 	return w->node && !(w->state == FWS_U && w->node == w->root);
2626 }
2627 
2628 static void ipv6_route_seq_stop(struct seq_file *seq, void *v)
2629 	__releases(RCU_BH)
2630 {
2631 	struct net *net = seq_file_net(seq);
2632 	struct ipv6_route_iter *iter = seq->private;
2633 
2634 	if (ipv6_route_iter_active(iter))
2635 		fib6_walker_unlink(net, &iter->w);
2636 
2637 	rcu_read_unlock_bh();
2638 }
2639 
2640 const struct seq_operations ipv6_route_seq_ops = {
2641 	.start	= ipv6_route_seq_start,
2642 	.next	= ipv6_route_seq_next,
2643 	.stop	= ipv6_route_seq_stop,
2644 	.show	= ipv6_route_seq_show
2645 };
2646 #endif /* CONFIG_PROC_FS */
2647