xref: /linux/net/ipv6/route.c (revision d4b996f9ef1fe83d9ce9ad5c1ca0bd8231638ce5)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Linux INET6 implementation
4  *	FIB front-end.
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  */
9 
10 /*	Changes:
11  *
12  *	YOSHIFUJI Hideaki @USAGI
13  *		reworked default router selection.
14  *		- respect outgoing interface
15  *		- select from (probably) reachable routers (i.e.
16  *		routers in REACHABLE, STALE, DELAY or PROBE states).
17  *		- always select the same router if it is (probably)
18  *		reachable.  otherwise, round-robin the list.
19  *	Ville Nuorvala
20  *		Fixed routing subtrees.
21  */
22 
23 #define pr_fmt(fmt) "IPv6: " fmt
24 
25 #include <linux/capability.h>
26 #include <linux/errno.h>
27 #include <linux/export.h>
28 #include <linux/types.h>
29 #include <linux/times.h>
30 #include <linux/socket.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/route.h>
34 #include <linux/netdevice.h>
35 #include <linux/in6.h>
36 #include <linux/mroute6.h>
37 #include <linux/init.h>
38 #include <linux/if_arp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/nsproxy.h>
42 #include <linux/slab.h>
43 #include <linux/jhash.h>
44 #include <net/net_namespace.h>
45 #include <net/snmp.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
51 #include <net/tcp.h>
52 #include <linux/rtnetlink.h>
53 #include <net/dst.h>
54 #include <net/dst_metadata.h>
55 #include <net/xfrm.h>
56 #include <net/netevent.h>
57 #include <net/netlink.h>
58 #include <net/rtnh.h>
59 #include <net/lwtunnel.h>
60 #include <net/ip_tunnels.h>
61 #include <net/l3mdev.h>
62 #include <net/ip.h>
63 #include <linux/uaccess.h>
64 #include <linux/btf_ids.h>
65 
66 #ifdef CONFIG_SYSCTL
67 #include <linux/sysctl.h>
68 #endif
69 
70 static int ip6_rt_type_to_error(u8 fib6_type);
71 
72 #define CREATE_TRACE_POINTS
73 #include <trace/events/fib6.h>
74 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
75 #undef CREATE_TRACE_POINTS
76 
77 enum rt6_nud_state {
78 	RT6_NUD_FAIL_HARD = -3,
79 	RT6_NUD_FAIL_PROBE = -2,
80 	RT6_NUD_FAIL_DO_RR = -1,
81 	RT6_NUD_SUCCEED = 1
82 };
83 
84 INDIRECT_CALLABLE_SCOPE
85 struct dst_entry	*ip6_dst_check(struct dst_entry *dst, u32 cookie);
86 static unsigned int	 ip6_default_advmss(const struct dst_entry *dst);
87 INDIRECT_CALLABLE_SCOPE
88 unsigned int		ip6_mtu(const struct dst_entry *dst);
89 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
90 static void		ip6_dst_destroy(struct dst_entry *);
91 static void		ip6_dst_ifdown(struct dst_entry *,
92 				       struct net_device *dev, int how);
93 static int		 ip6_dst_gc(struct dst_ops *ops);
94 
95 static int		ip6_pkt_discard(struct sk_buff *skb);
96 static int		ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
97 static int		ip6_pkt_prohibit(struct sk_buff *skb);
98 static int		ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
99 static void		ip6_link_failure(struct sk_buff *skb);
100 static void		ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
101 					   struct sk_buff *skb, u32 mtu,
102 					   bool confirm_neigh);
103 static void		rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
104 					struct sk_buff *skb);
105 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
106 			   int strict);
107 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
108 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
109 			 struct fib6_info *rt, struct dst_entry *dst,
110 			 struct in6_addr *dest, struct in6_addr *src,
111 			 int iif, int type, u32 portid, u32 seq,
112 			 unsigned int flags);
113 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
114 					   const struct in6_addr *daddr,
115 					   const struct in6_addr *saddr);
116 
117 #ifdef CONFIG_IPV6_ROUTE_INFO
118 static struct fib6_info *rt6_add_route_info(struct net *net,
119 					   const struct in6_addr *prefix, int prefixlen,
120 					   const struct in6_addr *gwaddr,
121 					   struct net_device *dev,
122 					   unsigned int pref);
123 static struct fib6_info *rt6_get_route_info(struct net *net,
124 					   const struct in6_addr *prefix, int prefixlen,
125 					   const struct in6_addr *gwaddr,
126 					   struct net_device *dev);
127 #endif
128 
129 struct uncached_list {
130 	spinlock_t		lock;
131 	struct list_head	head;
132 };
133 
134 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
135 
136 void rt6_uncached_list_add(struct rt6_info *rt)
137 {
138 	struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
139 
140 	rt->rt6i_uncached_list = ul;
141 
142 	spin_lock_bh(&ul->lock);
143 	list_add_tail(&rt->rt6i_uncached, &ul->head);
144 	spin_unlock_bh(&ul->lock);
145 }
146 
147 void rt6_uncached_list_del(struct rt6_info *rt)
148 {
149 	if (!list_empty(&rt->rt6i_uncached)) {
150 		struct uncached_list *ul = rt->rt6i_uncached_list;
151 		struct net *net = dev_net(rt->dst.dev);
152 
153 		spin_lock_bh(&ul->lock);
154 		list_del(&rt->rt6i_uncached);
155 		atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
156 		spin_unlock_bh(&ul->lock);
157 	}
158 }
159 
160 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
161 {
162 	struct net_device *loopback_dev = net->loopback_dev;
163 	int cpu;
164 
165 	if (dev == loopback_dev)
166 		return;
167 
168 	for_each_possible_cpu(cpu) {
169 		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
170 		struct rt6_info *rt;
171 
172 		spin_lock_bh(&ul->lock);
173 		list_for_each_entry(rt, &ul->head, rt6i_uncached) {
174 			struct inet6_dev *rt_idev = rt->rt6i_idev;
175 			struct net_device *rt_dev = rt->dst.dev;
176 
177 			if (rt_idev->dev == dev) {
178 				rt->rt6i_idev = in6_dev_get(loopback_dev);
179 				in6_dev_put(rt_idev);
180 			}
181 
182 			if (rt_dev == dev) {
183 				rt->dst.dev = blackhole_netdev;
184 				dev_hold(rt->dst.dev);
185 				dev_put(rt_dev);
186 			}
187 		}
188 		spin_unlock_bh(&ul->lock);
189 	}
190 }
191 
192 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
193 					     struct sk_buff *skb,
194 					     const void *daddr)
195 {
196 	if (!ipv6_addr_any(p))
197 		return (const void *) p;
198 	else if (skb)
199 		return &ipv6_hdr(skb)->daddr;
200 	return daddr;
201 }
202 
203 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
204 				   struct net_device *dev,
205 				   struct sk_buff *skb,
206 				   const void *daddr)
207 {
208 	struct neighbour *n;
209 
210 	daddr = choose_neigh_daddr(gw, skb, daddr);
211 	n = __ipv6_neigh_lookup(dev, daddr);
212 	if (n)
213 		return n;
214 
215 	n = neigh_create(&nd_tbl, daddr, dev);
216 	return IS_ERR(n) ? NULL : n;
217 }
218 
219 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
220 					      struct sk_buff *skb,
221 					      const void *daddr)
222 {
223 	const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
224 
225 	return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
226 				dst->dev, skb, daddr);
227 }
228 
229 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
230 {
231 	struct net_device *dev = dst->dev;
232 	struct rt6_info *rt = (struct rt6_info *)dst;
233 
234 	daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
235 	if (!daddr)
236 		return;
237 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
238 		return;
239 	if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
240 		return;
241 	__ipv6_confirm_neigh(dev, daddr);
242 }
243 
244 static struct dst_ops ip6_dst_ops_template = {
245 	.family			=	AF_INET6,
246 	.gc			=	ip6_dst_gc,
247 	.gc_thresh		=	1024,
248 	.check			=	ip6_dst_check,
249 	.default_advmss		=	ip6_default_advmss,
250 	.mtu			=	ip6_mtu,
251 	.cow_metrics		=	dst_cow_metrics_generic,
252 	.destroy		=	ip6_dst_destroy,
253 	.ifdown			=	ip6_dst_ifdown,
254 	.negative_advice	=	ip6_negative_advice,
255 	.link_failure		=	ip6_link_failure,
256 	.update_pmtu		=	ip6_rt_update_pmtu,
257 	.redirect		=	rt6_do_redirect,
258 	.local_out		=	__ip6_local_out,
259 	.neigh_lookup		=	ip6_dst_neigh_lookup,
260 	.confirm_neigh		=	ip6_confirm_neigh,
261 };
262 
263 static struct dst_ops ip6_dst_blackhole_ops = {
264 	.family			= AF_INET6,
265 	.default_advmss		= ip6_default_advmss,
266 	.neigh_lookup		= ip6_dst_neigh_lookup,
267 	.check			= ip6_dst_check,
268 	.destroy		= ip6_dst_destroy,
269 	.cow_metrics		= dst_cow_metrics_generic,
270 	.update_pmtu		= dst_blackhole_update_pmtu,
271 	.redirect		= dst_blackhole_redirect,
272 	.mtu			= dst_blackhole_mtu,
273 };
274 
275 static const u32 ip6_template_metrics[RTAX_MAX] = {
276 	[RTAX_HOPLIMIT - 1] = 0,
277 };
278 
279 static const struct fib6_info fib6_null_entry_template = {
280 	.fib6_flags	= (RTF_REJECT | RTF_NONEXTHOP),
281 	.fib6_protocol  = RTPROT_KERNEL,
282 	.fib6_metric	= ~(u32)0,
283 	.fib6_ref	= REFCOUNT_INIT(1),
284 	.fib6_type	= RTN_UNREACHABLE,
285 	.fib6_metrics	= (struct dst_metrics *)&dst_default_metrics,
286 };
287 
288 static const struct rt6_info ip6_null_entry_template = {
289 	.dst = {
290 		.__refcnt	= ATOMIC_INIT(1),
291 		.__use		= 1,
292 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
293 		.error		= -ENETUNREACH,
294 		.input		= ip6_pkt_discard,
295 		.output		= ip6_pkt_discard_out,
296 	},
297 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
298 };
299 
300 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
301 
302 static const struct rt6_info ip6_prohibit_entry_template = {
303 	.dst = {
304 		.__refcnt	= ATOMIC_INIT(1),
305 		.__use		= 1,
306 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
307 		.error		= -EACCES,
308 		.input		= ip6_pkt_prohibit,
309 		.output		= ip6_pkt_prohibit_out,
310 	},
311 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
312 };
313 
314 static const struct rt6_info ip6_blk_hole_entry_template = {
315 	.dst = {
316 		.__refcnt	= ATOMIC_INIT(1),
317 		.__use		= 1,
318 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
319 		.error		= -EINVAL,
320 		.input		= dst_discard,
321 		.output		= dst_discard_out,
322 	},
323 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
324 };
325 
326 #endif
327 
328 static void rt6_info_init(struct rt6_info *rt)
329 {
330 	struct dst_entry *dst = &rt->dst;
331 
332 	memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
333 	INIT_LIST_HEAD(&rt->rt6i_uncached);
334 }
335 
336 /* allocate dst with ip6_dst_ops */
337 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
338 			       int flags)
339 {
340 	struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
341 					1, DST_OBSOLETE_FORCE_CHK, flags);
342 
343 	if (rt) {
344 		rt6_info_init(rt);
345 		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
346 	}
347 
348 	return rt;
349 }
350 EXPORT_SYMBOL(ip6_dst_alloc);
351 
352 static void ip6_dst_destroy(struct dst_entry *dst)
353 {
354 	struct rt6_info *rt = (struct rt6_info *)dst;
355 	struct fib6_info *from;
356 	struct inet6_dev *idev;
357 
358 	ip_dst_metrics_put(dst);
359 	rt6_uncached_list_del(rt);
360 
361 	idev = rt->rt6i_idev;
362 	if (idev) {
363 		rt->rt6i_idev = NULL;
364 		in6_dev_put(idev);
365 	}
366 
367 	from = xchg((__force struct fib6_info **)&rt->from, NULL);
368 	fib6_info_release(from);
369 }
370 
371 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
372 			   int how)
373 {
374 	struct rt6_info *rt = (struct rt6_info *)dst;
375 	struct inet6_dev *idev = rt->rt6i_idev;
376 	struct net_device *loopback_dev =
377 		dev_net(dev)->loopback_dev;
378 
379 	if (idev && idev->dev != loopback_dev) {
380 		struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
381 		if (loopback_idev) {
382 			rt->rt6i_idev = loopback_idev;
383 			in6_dev_put(idev);
384 		}
385 	}
386 }
387 
388 static bool __rt6_check_expired(const struct rt6_info *rt)
389 {
390 	if (rt->rt6i_flags & RTF_EXPIRES)
391 		return time_after(jiffies, rt->dst.expires);
392 	else
393 		return false;
394 }
395 
396 static bool rt6_check_expired(const struct rt6_info *rt)
397 {
398 	struct fib6_info *from;
399 
400 	from = rcu_dereference(rt->from);
401 
402 	if (rt->rt6i_flags & RTF_EXPIRES) {
403 		if (time_after(jiffies, rt->dst.expires))
404 			return true;
405 	} else if (from) {
406 		return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
407 			fib6_check_expired(from);
408 	}
409 	return false;
410 }
411 
412 void fib6_select_path(const struct net *net, struct fib6_result *res,
413 		      struct flowi6 *fl6, int oif, bool have_oif_match,
414 		      const struct sk_buff *skb, int strict)
415 {
416 	struct fib6_info *sibling, *next_sibling;
417 	struct fib6_info *match = res->f6i;
418 
419 	if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
420 		goto out;
421 
422 	if (match->nh && have_oif_match && res->nh)
423 		return;
424 
425 	/* We might have already computed the hash for ICMPv6 errors. In such
426 	 * case it will always be non-zero. Otherwise now is the time to do it.
427 	 */
428 	if (!fl6->mp_hash &&
429 	    (!match->nh || nexthop_is_multipath(match->nh)))
430 		fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
431 
432 	if (unlikely(match->nh)) {
433 		nexthop_path_fib6_result(res, fl6->mp_hash);
434 		return;
435 	}
436 
437 	if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
438 		goto out;
439 
440 	list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
441 				 fib6_siblings) {
442 		const struct fib6_nh *nh = sibling->fib6_nh;
443 		int nh_upper_bound;
444 
445 		nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
446 		if (fl6->mp_hash > nh_upper_bound)
447 			continue;
448 		if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
449 			break;
450 		match = sibling;
451 		break;
452 	}
453 
454 out:
455 	res->f6i = match;
456 	res->nh = match->fib6_nh;
457 }
458 
459 /*
460  *	Route lookup. rcu_read_lock() should be held.
461  */
462 
463 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
464 			       const struct in6_addr *saddr, int oif, int flags)
465 {
466 	const struct net_device *dev;
467 
468 	if (nh->fib_nh_flags & RTNH_F_DEAD)
469 		return false;
470 
471 	dev = nh->fib_nh_dev;
472 	if (oif) {
473 		if (dev->ifindex == oif)
474 			return true;
475 	} else {
476 		if (ipv6_chk_addr(net, saddr, dev,
477 				  flags & RT6_LOOKUP_F_IFACE))
478 			return true;
479 	}
480 
481 	return false;
482 }
483 
484 struct fib6_nh_dm_arg {
485 	struct net		*net;
486 	const struct in6_addr	*saddr;
487 	int			oif;
488 	int			flags;
489 	struct fib6_nh		*nh;
490 };
491 
492 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
493 {
494 	struct fib6_nh_dm_arg *arg = _arg;
495 
496 	arg->nh = nh;
497 	return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
498 				  arg->flags);
499 }
500 
501 /* returns fib6_nh from nexthop or NULL */
502 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
503 					struct fib6_result *res,
504 					const struct in6_addr *saddr,
505 					int oif, int flags)
506 {
507 	struct fib6_nh_dm_arg arg = {
508 		.net   = net,
509 		.saddr = saddr,
510 		.oif   = oif,
511 		.flags = flags,
512 	};
513 
514 	if (nexthop_is_blackhole(nh))
515 		return NULL;
516 
517 	if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
518 		return arg.nh;
519 
520 	return NULL;
521 }
522 
523 static void rt6_device_match(struct net *net, struct fib6_result *res,
524 			     const struct in6_addr *saddr, int oif, int flags)
525 {
526 	struct fib6_info *f6i = res->f6i;
527 	struct fib6_info *spf6i;
528 	struct fib6_nh *nh;
529 
530 	if (!oif && ipv6_addr_any(saddr)) {
531 		if (unlikely(f6i->nh)) {
532 			nh = nexthop_fib6_nh(f6i->nh);
533 			if (nexthop_is_blackhole(f6i->nh))
534 				goto out_blackhole;
535 		} else {
536 			nh = f6i->fib6_nh;
537 		}
538 		if (!(nh->fib_nh_flags & RTNH_F_DEAD))
539 			goto out;
540 	}
541 
542 	for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
543 		bool matched = false;
544 
545 		if (unlikely(spf6i->nh)) {
546 			nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
547 					      oif, flags);
548 			if (nh)
549 				matched = true;
550 		} else {
551 			nh = spf6i->fib6_nh;
552 			if (__rt6_device_match(net, nh, saddr, oif, flags))
553 				matched = true;
554 		}
555 		if (matched) {
556 			res->f6i = spf6i;
557 			goto out;
558 		}
559 	}
560 
561 	if (oif && flags & RT6_LOOKUP_F_IFACE) {
562 		res->f6i = net->ipv6.fib6_null_entry;
563 		nh = res->f6i->fib6_nh;
564 		goto out;
565 	}
566 
567 	if (unlikely(f6i->nh)) {
568 		nh = nexthop_fib6_nh(f6i->nh);
569 		if (nexthop_is_blackhole(f6i->nh))
570 			goto out_blackhole;
571 	} else {
572 		nh = f6i->fib6_nh;
573 	}
574 
575 	if (nh->fib_nh_flags & RTNH_F_DEAD) {
576 		res->f6i = net->ipv6.fib6_null_entry;
577 		nh = res->f6i->fib6_nh;
578 	}
579 out:
580 	res->nh = nh;
581 	res->fib6_type = res->f6i->fib6_type;
582 	res->fib6_flags = res->f6i->fib6_flags;
583 	return;
584 
585 out_blackhole:
586 	res->fib6_flags |= RTF_REJECT;
587 	res->fib6_type = RTN_BLACKHOLE;
588 	res->nh = nh;
589 }
590 
591 #ifdef CONFIG_IPV6_ROUTER_PREF
592 struct __rt6_probe_work {
593 	struct work_struct work;
594 	struct in6_addr target;
595 	struct net_device *dev;
596 };
597 
598 static void rt6_probe_deferred(struct work_struct *w)
599 {
600 	struct in6_addr mcaddr;
601 	struct __rt6_probe_work *work =
602 		container_of(w, struct __rt6_probe_work, work);
603 
604 	addrconf_addr_solict_mult(&work->target, &mcaddr);
605 	ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
606 	dev_put(work->dev);
607 	kfree(work);
608 }
609 
610 static void rt6_probe(struct fib6_nh *fib6_nh)
611 {
612 	struct __rt6_probe_work *work = NULL;
613 	const struct in6_addr *nh_gw;
614 	unsigned long last_probe;
615 	struct neighbour *neigh;
616 	struct net_device *dev;
617 	struct inet6_dev *idev;
618 
619 	/*
620 	 * Okay, this does not seem to be appropriate
621 	 * for now, however, we need to check if it
622 	 * is really so; aka Router Reachability Probing.
623 	 *
624 	 * Router Reachability Probe MUST be rate-limited
625 	 * to no more than one per minute.
626 	 */
627 	if (!fib6_nh->fib_nh_gw_family)
628 		return;
629 
630 	nh_gw = &fib6_nh->fib_nh_gw6;
631 	dev = fib6_nh->fib_nh_dev;
632 	rcu_read_lock_bh();
633 	last_probe = READ_ONCE(fib6_nh->last_probe);
634 	idev = __in6_dev_get(dev);
635 	neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
636 	if (neigh) {
637 		if (neigh->nud_state & NUD_VALID)
638 			goto out;
639 
640 		write_lock(&neigh->lock);
641 		if (!(neigh->nud_state & NUD_VALID) &&
642 		    time_after(jiffies,
643 			       neigh->updated + idev->cnf.rtr_probe_interval)) {
644 			work = kmalloc(sizeof(*work), GFP_ATOMIC);
645 			if (work)
646 				__neigh_set_probe_once(neigh);
647 		}
648 		write_unlock(&neigh->lock);
649 	} else if (time_after(jiffies, last_probe +
650 				       idev->cnf.rtr_probe_interval)) {
651 		work = kmalloc(sizeof(*work), GFP_ATOMIC);
652 	}
653 
654 	if (!work || cmpxchg(&fib6_nh->last_probe,
655 			     last_probe, jiffies) != last_probe) {
656 		kfree(work);
657 	} else {
658 		INIT_WORK(&work->work, rt6_probe_deferred);
659 		work->target = *nh_gw;
660 		dev_hold(dev);
661 		work->dev = dev;
662 		schedule_work(&work->work);
663 	}
664 
665 out:
666 	rcu_read_unlock_bh();
667 }
668 #else
669 static inline void rt6_probe(struct fib6_nh *fib6_nh)
670 {
671 }
672 #endif
673 
674 /*
675  * Default Router Selection (RFC 2461 6.3.6)
676  */
677 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
678 {
679 	enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
680 	struct neighbour *neigh;
681 
682 	rcu_read_lock_bh();
683 	neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
684 					  &fib6_nh->fib_nh_gw6);
685 	if (neigh) {
686 		read_lock(&neigh->lock);
687 		if (neigh->nud_state & NUD_VALID)
688 			ret = RT6_NUD_SUCCEED;
689 #ifdef CONFIG_IPV6_ROUTER_PREF
690 		else if (!(neigh->nud_state & NUD_FAILED))
691 			ret = RT6_NUD_SUCCEED;
692 		else
693 			ret = RT6_NUD_FAIL_PROBE;
694 #endif
695 		read_unlock(&neigh->lock);
696 	} else {
697 		ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
698 		      RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
699 	}
700 	rcu_read_unlock_bh();
701 
702 	return ret;
703 }
704 
705 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
706 			   int strict)
707 {
708 	int m = 0;
709 
710 	if (!oif || nh->fib_nh_dev->ifindex == oif)
711 		m = 2;
712 
713 	if (!m && (strict & RT6_LOOKUP_F_IFACE))
714 		return RT6_NUD_FAIL_HARD;
715 #ifdef CONFIG_IPV6_ROUTER_PREF
716 	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
717 #endif
718 	if ((strict & RT6_LOOKUP_F_REACHABLE) &&
719 	    !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
720 		int n = rt6_check_neigh(nh);
721 		if (n < 0)
722 			return n;
723 	}
724 	return m;
725 }
726 
727 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
728 		       int oif, int strict, int *mpri, bool *do_rr)
729 {
730 	bool match_do_rr = false;
731 	bool rc = false;
732 	int m;
733 
734 	if (nh->fib_nh_flags & RTNH_F_DEAD)
735 		goto out;
736 
737 	if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
738 	    nh->fib_nh_flags & RTNH_F_LINKDOWN &&
739 	    !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
740 		goto out;
741 
742 	m = rt6_score_route(nh, fib6_flags, oif, strict);
743 	if (m == RT6_NUD_FAIL_DO_RR) {
744 		match_do_rr = true;
745 		m = 0; /* lowest valid score */
746 	} else if (m == RT6_NUD_FAIL_HARD) {
747 		goto out;
748 	}
749 
750 	if (strict & RT6_LOOKUP_F_REACHABLE)
751 		rt6_probe(nh);
752 
753 	/* note that m can be RT6_NUD_FAIL_PROBE at this point */
754 	if (m > *mpri) {
755 		*do_rr = match_do_rr;
756 		*mpri = m;
757 		rc = true;
758 	}
759 out:
760 	return rc;
761 }
762 
763 struct fib6_nh_frl_arg {
764 	u32		flags;
765 	int		oif;
766 	int		strict;
767 	int		*mpri;
768 	bool		*do_rr;
769 	struct fib6_nh	*nh;
770 };
771 
772 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
773 {
774 	struct fib6_nh_frl_arg *arg = _arg;
775 
776 	arg->nh = nh;
777 	return find_match(nh, arg->flags, arg->oif, arg->strict,
778 			  arg->mpri, arg->do_rr);
779 }
780 
781 static void __find_rr_leaf(struct fib6_info *f6i_start,
782 			   struct fib6_info *nomatch, u32 metric,
783 			   struct fib6_result *res, struct fib6_info **cont,
784 			   int oif, int strict, bool *do_rr, int *mpri)
785 {
786 	struct fib6_info *f6i;
787 
788 	for (f6i = f6i_start;
789 	     f6i && f6i != nomatch;
790 	     f6i = rcu_dereference(f6i->fib6_next)) {
791 		bool matched = false;
792 		struct fib6_nh *nh;
793 
794 		if (cont && f6i->fib6_metric != metric) {
795 			*cont = f6i;
796 			return;
797 		}
798 
799 		if (fib6_check_expired(f6i))
800 			continue;
801 
802 		if (unlikely(f6i->nh)) {
803 			struct fib6_nh_frl_arg arg = {
804 				.flags  = f6i->fib6_flags,
805 				.oif    = oif,
806 				.strict = strict,
807 				.mpri   = mpri,
808 				.do_rr  = do_rr
809 			};
810 
811 			if (nexthop_is_blackhole(f6i->nh)) {
812 				res->fib6_flags = RTF_REJECT;
813 				res->fib6_type = RTN_BLACKHOLE;
814 				res->f6i = f6i;
815 				res->nh = nexthop_fib6_nh(f6i->nh);
816 				return;
817 			}
818 			if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
819 						     &arg)) {
820 				matched = true;
821 				nh = arg.nh;
822 			}
823 		} else {
824 			nh = f6i->fib6_nh;
825 			if (find_match(nh, f6i->fib6_flags, oif, strict,
826 				       mpri, do_rr))
827 				matched = true;
828 		}
829 		if (matched) {
830 			res->f6i = f6i;
831 			res->nh = nh;
832 			res->fib6_flags = f6i->fib6_flags;
833 			res->fib6_type = f6i->fib6_type;
834 		}
835 	}
836 }
837 
838 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
839 			 struct fib6_info *rr_head, int oif, int strict,
840 			 bool *do_rr, struct fib6_result *res)
841 {
842 	u32 metric = rr_head->fib6_metric;
843 	struct fib6_info *cont = NULL;
844 	int mpri = -1;
845 
846 	__find_rr_leaf(rr_head, NULL, metric, res, &cont,
847 		       oif, strict, do_rr, &mpri);
848 
849 	__find_rr_leaf(leaf, rr_head, metric, res, &cont,
850 		       oif, strict, do_rr, &mpri);
851 
852 	if (res->f6i || !cont)
853 		return;
854 
855 	__find_rr_leaf(cont, NULL, metric, res, NULL,
856 		       oif, strict, do_rr, &mpri);
857 }
858 
859 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
860 		       struct fib6_result *res, int strict)
861 {
862 	struct fib6_info *leaf = rcu_dereference(fn->leaf);
863 	struct fib6_info *rt0;
864 	bool do_rr = false;
865 	int key_plen;
866 
867 	/* make sure this function or its helpers sets f6i */
868 	res->f6i = NULL;
869 
870 	if (!leaf || leaf == net->ipv6.fib6_null_entry)
871 		goto out;
872 
873 	rt0 = rcu_dereference(fn->rr_ptr);
874 	if (!rt0)
875 		rt0 = leaf;
876 
877 	/* Double check to make sure fn is not an intermediate node
878 	 * and fn->leaf does not points to its child's leaf
879 	 * (This might happen if all routes under fn are deleted from
880 	 * the tree and fib6_repair_tree() is called on the node.)
881 	 */
882 	key_plen = rt0->fib6_dst.plen;
883 #ifdef CONFIG_IPV6_SUBTREES
884 	if (rt0->fib6_src.plen)
885 		key_plen = rt0->fib6_src.plen;
886 #endif
887 	if (fn->fn_bit != key_plen)
888 		goto out;
889 
890 	find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
891 	if (do_rr) {
892 		struct fib6_info *next = rcu_dereference(rt0->fib6_next);
893 
894 		/* no entries matched; do round-robin */
895 		if (!next || next->fib6_metric != rt0->fib6_metric)
896 			next = leaf;
897 
898 		if (next != rt0) {
899 			spin_lock_bh(&leaf->fib6_table->tb6_lock);
900 			/* make sure next is not being deleted from the tree */
901 			if (next->fib6_node)
902 				rcu_assign_pointer(fn->rr_ptr, next);
903 			spin_unlock_bh(&leaf->fib6_table->tb6_lock);
904 		}
905 	}
906 
907 out:
908 	if (!res->f6i) {
909 		res->f6i = net->ipv6.fib6_null_entry;
910 		res->nh = res->f6i->fib6_nh;
911 		res->fib6_flags = res->f6i->fib6_flags;
912 		res->fib6_type = res->f6i->fib6_type;
913 	}
914 }
915 
916 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
917 {
918 	return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
919 	       res->nh->fib_nh_gw_family;
920 }
921 
922 #ifdef CONFIG_IPV6_ROUTE_INFO
923 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
924 		  const struct in6_addr *gwaddr)
925 {
926 	struct net *net = dev_net(dev);
927 	struct route_info *rinfo = (struct route_info *) opt;
928 	struct in6_addr prefix_buf, *prefix;
929 	unsigned int pref;
930 	unsigned long lifetime;
931 	struct fib6_info *rt;
932 
933 	if (len < sizeof(struct route_info)) {
934 		return -EINVAL;
935 	}
936 
937 	/* Sanity check for prefix_len and length */
938 	if (rinfo->length > 3) {
939 		return -EINVAL;
940 	} else if (rinfo->prefix_len > 128) {
941 		return -EINVAL;
942 	} else if (rinfo->prefix_len > 64) {
943 		if (rinfo->length < 2) {
944 			return -EINVAL;
945 		}
946 	} else if (rinfo->prefix_len > 0) {
947 		if (rinfo->length < 1) {
948 			return -EINVAL;
949 		}
950 	}
951 
952 	pref = rinfo->route_pref;
953 	if (pref == ICMPV6_ROUTER_PREF_INVALID)
954 		return -EINVAL;
955 
956 	lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
957 
958 	if (rinfo->length == 3)
959 		prefix = (struct in6_addr *)rinfo->prefix;
960 	else {
961 		/* this function is safe */
962 		ipv6_addr_prefix(&prefix_buf,
963 				 (struct in6_addr *)rinfo->prefix,
964 				 rinfo->prefix_len);
965 		prefix = &prefix_buf;
966 	}
967 
968 	if (rinfo->prefix_len == 0)
969 		rt = rt6_get_dflt_router(net, gwaddr, dev);
970 	else
971 		rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
972 					gwaddr, dev);
973 
974 	if (rt && !lifetime) {
975 		ip6_del_rt(net, rt, false);
976 		rt = NULL;
977 	}
978 
979 	if (!rt && lifetime)
980 		rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
981 					dev, pref);
982 	else if (rt)
983 		rt->fib6_flags = RTF_ROUTEINFO |
984 				 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
985 
986 	if (rt) {
987 		if (!addrconf_finite_timeout(lifetime))
988 			fib6_clean_expires(rt);
989 		else
990 			fib6_set_expires(rt, jiffies + HZ * lifetime);
991 
992 		fib6_info_release(rt);
993 	}
994 	return 0;
995 }
996 #endif
997 
998 /*
999  *	Misc support functions
1000  */
1001 
1002 /* called with rcu_lock held */
1003 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1004 {
1005 	struct net_device *dev = res->nh->fib_nh_dev;
1006 
1007 	if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1008 		/* for copies of local routes, dst->dev needs to be the
1009 		 * device if it is a master device, the master device if
1010 		 * device is enslaved, and the loopback as the default
1011 		 */
1012 		if (netif_is_l3_slave(dev) &&
1013 		    !rt6_need_strict(&res->f6i->fib6_dst.addr))
1014 			dev = l3mdev_master_dev_rcu(dev);
1015 		else if (!netif_is_l3_master(dev))
1016 			dev = dev_net(dev)->loopback_dev;
1017 		/* last case is netif_is_l3_master(dev) is true in which
1018 		 * case we want dev returned to be dev
1019 		 */
1020 	}
1021 
1022 	return dev;
1023 }
1024 
1025 static const int fib6_prop[RTN_MAX + 1] = {
1026 	[RTN_UNSPEC]	= 0,
1027 	[RTN_UNICAST]	= 0,
1028 	[RTN_LOCAL]	= 0,
1029 	[RTN_BROADCAST]	= 0,
1030 	[RTN_ANYCAST]	= 0,
1031 	[RTN_MULTICAST]	= 0,
1032 	[RTN_BLACKHOLE]	= -EINVAL,
1033 	[RTN_UNREACHABLE] = -EHOSTUNREACH,
1034 	[RTN_PROHIBIT]	= -EACCES,
1035 	[RTN_THROW]	= -EAGAIN,
1036 	[RTN_NAT]	= -EINVAL,
1037 	[RTN_XRESOLVE]	= -EINVAL,
1038 };
1039 
1040 static int ip6_rt_type_to_error(u8 fib6_type)
1041 {
1042 	return fib6_prop[fib6_type];
1043 }
1044 
1045 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1046 {
1047 	unsigned short flags = 0;
1048 
1049 	if (rt->dst_nocount)
1050 		flags |= DST_NOCOUNT;
1051 	if (rt->dst_nopolicy)
1052 		flags |= DST_NOPOLICY;
1053 
1054 	return flags;
1055 }
1056 
1057 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1058 {
1059 	rt->dst.error = ip6_rt_type_to_error(fib6_type);
1060 
1061 	switch (fib6_type) {
1062 	case RTN_BLACKHOLE:
1063 		rt->dst.output = dst_discard_out;
1064 		rt->dst.input = dst_discard;
1065 		break;
1066 	case RTN_PROHIBIT:
1067 		rt->dst.output = ip6_pkt_prohibit_out;
1068 		rt->dst.input = ip6_pkt_prohibit;
1069 		break;
1070 	case RTN_THROW:
1071 	case RTN_UNREACHABLE:
1072 	default:
1073 		rt->dst.output = ip6_pkt_discard_out;
1074 		rt->dst.input = ip6_pkt_discard;
1075 		break;
1076 	}
1077 }
1078 
1079 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1080 {
1081 	struct fib6_info *f6i = res->f6i;
1082 
1083 	if (res->fib6_flags & RTF_REJECT) {
1084 		ip6_rt_init_dst_reject(rt, res->fib6_type);
1085 		return;
1086 	}
1087 
1088 	rt->dst.error = 0;
1089 	rt->dst.output = ip6_output;
1090 
1091 	if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1092 		rt->dst.input = ip6_input;
1093 	} else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1094 		rt->dst.input = ip6_mc_input;
1095 	} else {
1096 		rt->dst.input = ip6_forward;
1097 	}
1098 
1099 	if (res->nh->fib_nh_lws) {
1100 		rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1101 		lwtunnel_set_redirect(&rt->dst);
1102 	}
1103 
1104 	rt->dst.lastuse = jiffies;
1105 }
1106 
1107 /* Caller must already hold reference to @from */
1108 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1109 {
1110 	rt->rt6i_flags &= ~RTF_EXPIRES;
1111 	rcu_assign_pointer(rt->from, from);
1112 	ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1113 }
1114 
1115 /* Caller must already hold reference to f6i in result */
1116 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1117 {
1118 	const struct fib6_nh *nh = res->nh;
1119 	const struct net_device *dev = nh->fib_nh_dev;
1120 	struct fib6_info *f6i = res->f6i;
1121 
1122 	ip6_rt_init_dst(rt, res);
1123 
1124 	rt->rt6i_dst = f6i->fib6_dst;
1125 	rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1126 	rt->rt6i_flags = res->fib6_flags;
1127 	if (nh->fib_nh_gw_family) {
1128 		rt->rt6i_gateway = nh->fib_nh_gw6;
1129 		rt->rt6i_flags |= RTF_GATEWAY;
1130 	}
1131 	rt6_set_from(rt, f6i);
1132 #ifdef CONFIG_IPV6_SUBTREES
1133 	rt->rt6i_src = f6i->fib6_src;
1134 #endif
1135 }
1136 
1137 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1138 					struct in6_addr *saddr)
1139 {
1140 	struct fib6_node *pn, *sn;
1141 	while (1) {
1142 		if (fn->fn_flags & RTN_TL_ROOT)
1143 			return NULL;
1144 		pn = rcu_dereference(fn->parent);
1145 		sn = FIB6_SUBTREE(pn);
1146 		if (sn && sn != fn)
1147 			fn = fib6_node_lookup(sn, NULL, saddr);
1148 		else
1149 			fn = pn;
1150 		if (fn->fn_flags & RTN_RTINFO)
1151 			return fn;
1152 	}
1153 }
1154 
1155 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1156 {
1157 	struct rt6_info *rt = *prt;
1158 
1159 	if (dst_hold_safe(&rt->dst))
1160 		return true;
1161 	if (net) {
1162 		rt = net->ipv6.ip6_null_entry;
1163 		dst_hold(&rt->dst);
1164 	} else {
1165 		rt = NULL;
1166 	}
1167 	*prt = rt;
1168 	return false;
1169 }
1170 
1171 /* called with rcu_lock held */
1172 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1173 {
1174 	struct net_device *dev = res->nh->fib_nh_dev;
1175 	struct fib6_info *f6i = res->f6i;
1176 	unsigned short flags;
1177 	struct rt6_info *nrt;
1178 
1179 	if (!fib6_info_hold_safe(f6i))
1180 		goto fallback;
1181 
1182 	flags = fib6_info_dst_flags(f6i);
1183 	nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1184 	if (!nrt) {
1185 		fib6_info_release(f6i);
1186 		goto fallback;
1187 	}
1188 
1189 	ip6_rt_copy_init(nrt, res);
1190 	return nrt;
1191 
1192 fallback:
1193 	nrt = dev_net(dev)->ipv6.ip6_null_entry;
1194 	dst_hold(&nrt->dst);
1195 	return nrt;
1196 }
1197 
1198 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1199 					     struct fib6_table *table,
1200 					     struct flowi6 *fl6,
1201 					     const struct sk_buff *skb,
1202 					     int flags)
1203 {
1204 	struct fib6_result res = {};
1205 	struct fib6_node *fn;
1206 	struct rt6_info *rt;
1207 
1208 	if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1209 		flags &= ~RT6_LOOKUP_F_IFACE;
1210 
1211 	rcu_read_lock();
1212 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1213 restart:
1214 	res.f6i = rcu_dereference(fn->leaf);
1215 	if (!res.f6i)
1216 		res.f6i = net->ipv6.fib6_null_entry;
1217 	else
1218 		rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1219 				 flags);
1220 
1221 	if (res.f6i == net->ipv6.fib6_null_entry) {
1222 		fn = fib6_backtrack(fn, &fl6->saddr);
1223 		if (fn)
1224 			goto restart;
1225 
1226 		rt = net->ipv6.ip6_null_entry;
1227 		dst_hold(&rt->dst);
1228 		goto out;
1229 	} else if (res.fib6_flags & RTF_REJECT) {
1230 		goto do_create;
1231 	}
1232 
1233 	fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1234 			 fl6->flowi6_oif != 0, skb, flags);
1235 
1236 	/* Search through exception table */
1237 	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1238 	if (rt) {
1239 		if (ip6_hold_safe(net, &rt))
1240 			dst_use_noref(&rt->dst, jiffies);
1241 	} else {
1242 do_create:
1243 		rt = ip6_create_rt_rcu(&res);
1244 	}
1245 
1246 out:
1247 	trace_fib6_table_lookup(net, &res, table, fl6);
1248 
1249 	rcu_read_unlock();
1250 
1251 	return rt;
1252 }
1253 
1254 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1255 				   const struct sk_buff *skb, int flags)
1256 {
1257 	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1258 }
1259 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1260 
1261 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1262 			    const struct in6_addr *saddr, int oif,
1263 			    const struct sk_buff *skb, int strict)
1264 {
1265 	struct flowi6 fl6 = {
1266 		.flowi6_oif = oif,
1267 		.daddr = *daddr,
1268 	};
1269 	struct dst_entry *dst;
1270 	int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1271 
1272 	if (saddr) {
1273 		memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1274 		flags |= RT6_LOOKUP_F_HAS_SADDR;
1275 	}
1276 
1277 	dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1278 	if (dst->error == 0)
1279 		return (struct rt6_info *) dst;
1280 
1281 	dst_release(dst);
1282 
1283 	return NULL;
1284 }
1285 EXPORT_SYMBOL(rt6_lookup);
1286 
1287 /* ip6_ins_rt is called with FREE table->tb6_lock.
1288  * It takes new route entry, the addition fails by any reason the
1289  * route is released.
1290  * Caller must hold dst before calling it.
1291  */
1292 
1293 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1294 			struct netlink_ext_ack *extack)
1295 {
1296 	int err;
1297 	struct fib6_table *table;
1298 
1299 	table = rt->fib6_table;
1300 	spin_lock_bh(&table->tb6_lock);
1301 	err = fib6_add(&table->tb6_root, rt, info, extack);
1302 	spin_unlock_bh(&table->tb6_lock);
1303 
1304 	return err;
1305 }
1306 
1307 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1308 {
1309 	struct nl_info info = {	.nl_net = net, };
1310 
1311 	return __ip6_ins_rt(rt, &info, NULL);
1312 }
1313 
1314 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1315 					   const struct in6_addr *daddr,
1316 					   const struct in6_addr *saddr)
1317 {
1318 	struct fib6_info *f6i = res->f6i;
1319 	struct net_device *dev;
1320 	struct rt6_info *rt;
1321 
1322 	/*
1323 	 *	Clone the route.
1324 	 */
1325 
1326 	if (!fib6_info_hold_safe(f6i))
1327 		return NULL;
1328 
1329 	dev = ip6_rt_get_dev_rcu(res);
1330 	rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1331 	if (!rt) {
1332 		fib6_info_release(f6i);
1333 		return NULL;
1334 	}
1335 
1336 	ip6_rt_copy_init(rt, res);
1337 	rt->rt6i_flags |= RTF_CACHE;
1338 	rt->rt6i_dst.addr = *daddr;
1339 	rt->rt6i_dst.plen = 128;
1340 
1341 	if (!rt6_is_gw_or_nonexthop(res)) {
1342 		if (f6i->fib6_dst.plen != 128 &&
1343 		    ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1344 			rt->rt6i_flags |= RTF_ANYCAST;
1345 #ifdef CONFIG_IPV6_SUBTREES
1346 		if (rt->rt6i_src.plen && saddr) {
1347 			rt->rt6i_src.addr = *saddr;
1348 			rt->rt6i_src.plen = 128;
1349 		}
1350 #endif
1351 	}
1352 
1353 	return rt;
1354 }
1355 
1356 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1357 {
1358 	struct fib6_info *f6i = res->f6i;
1359 	unsigned short flags = fib6_info_dst_flags(f6i);
1360 	struct net_device *dev;
1361 	struct rt6_info *pcpu_rt;
1362 
1363 	if (!fib6_info_hold_safe(f6i))
1364 		return NULL;
1365 
1366 	rcu_read_lock();
1367 	dev = ip6_rt_get_dev_rcu(res);
1368 	pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1369 	rcu_read_unlock();
1370 	if (!pcpu_rt) {
1371 		fib6_info_release(f6i);
1372 		return NULL;
1373 	}
1374 	ip6_rt_copy_init(pcpu_rt, res);
1375 	pcpu_rt->rt6i_flags |= RTF_PCPU;
1376 
1377 	if (f6i->nh)
1378 		pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1379 
1380 	return pcpu_rt;
1381 }
1382 
1383 static bool rt6_is_valid(const struct rt6_info *rt6)
1384 {
1385 	return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1386 }
1387 
1388 /* It should be called with rcu_read_lock() acquired */
1389 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1390 {
1391 	struct rt6_info *pcpu_rt;
1392 
1393 	pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1394 
1395 	if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1396 		struct rt6_info *prev, **p;
1397 
1398 		p = this_cpu_ptr(res->nh->rt6i_pcpu);
1399 		prev = xchg(p, NULL);
1400 		if (prev) {
1401 			dst_dev_put(&prev->dst);
1402 			dst_release(&prev->dst);
1403 		}
1404 
1405 		pcpu_rt = NULL;
1406 	}
1407 
1408 	return pcpu_rt;
1409 }
1410 
1411 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1412 					    const struct fib6_result *res)
1413 {
1414 	struct rt6_info *pcpu_rt, *prev, **p;
1415 
1416 	pcpu_rt = ip6_rt_pcpu_alloc(res);
1417 	if (!pcpu_rt)
1418 		return NULL;
1419 
1420 	p = this_cpu_ptr(res->nh->rt6i_pcpu);
1421 	prev = cmpxchg(p, NULL, pcpu_rt);
1422 	BUG_ON(prev);
1423 
1424 	if (res->f6i->fib6_destroying) {
1425 		struct fib6_info *from;
1426 
1427 		from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1428 		fib6_info_release(from);
1429 	}
1430 
1431 	return pcpu_rt;
1432 }
1433 
1434 /* exception hash table implementation
1435  */
1436 static DEFINE_SPINLOCK(rt6_exception_lock);
1437 
1438 /* Remove rt6_ex from hash table and free the memory
1439  * Caller must hold rt6_exception_lock
1440  */
1441 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1442 				 struct rt6_exception *rt6_ex)
1443 {
1444 	struct fib6_info *from;
1445 	struct net *net;
1446 
1447 	if (!bucket || !rt6_ex)
1448 		return;
1449 
1450 	net = dev_net(rt6_ex->rt6i->dst.dev);
1451 	net->ipv6.rt6_stats->fib_rt_cache--;
1452 
1453 	/* purge completely the exception to allow releasing the held resources:
1454 	 * some [sk] cache may keep the dst around for unlimited time
1455 	 */
1456 	from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1457 	fib6_info_release(from);
1458 	dst_dev_put(&rt6_ex->rt6i->dst);
1459 
1460 	hlist_del_rcu(&rt6_ex->hlist);
1461 	dst_release(&rt6_ex->rt6i->dst);
1462 	kfree_rcu(rt6_ex, rcu);
1463 	WARN_ON_ONCE(!bucket->depth);
1464 	bucket->depth--;
1465 }
1466 
1467 /* Remove oldest rt6_ex in bucket and free the memory
1468  * Caller must hold rt6_exception_lock
1469  */
1470 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1471 {
1472 	struct rt6_exception *rt6_ex, *oldest = NULL;
1473 
1474 	if (!bucket)
1475 		return;
1476 
1477 	hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1478 		if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1479 			oldest = rt6_ex;
1480 	}
1481 	rt6_remove_exception(bucket, oldest);
1482 }
1483 
1484 static u32 rt6_exception_hash(const struct in6_addr *dst,
1485 			      const struct in6_addr *src)
1486 {
1487 	static u32 seed __read_mostly;
1488 	u32 val;
1489 
1490 	net_get_random_once(&seed, sizeof(seed));
1491 	val = jhash2((const u32 *)dst, sizeof(*dst)/sizeof(u32), seed);
1492 
1493 #ifdef CONFIG_IPV6_SUBTREES
1494 	if (src)
1495 		val = jhash2((const u32 *)src, sizeof(*src)/sizeof(u32), val);
1496 #endif
1497 	return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1498 }
1499 
1500 /* Helper function to find the cached rt in the hash table
1501  * and update bucket pointer to point to the bucket for this
1502  * (daddr, saddr) pair
1503  * Caller must hold rt6_exception_lock
1504  */
1505 static struct rt6_exception *
1506 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1507 			      const struct in6_addr *daddr,
1508 			      const struct in6_addr *saddr)
1509 {
1510 	struct rt6_exception *rt6_ex;
1511 	u32 hval;
1512 
1513 	if (!(*bucket) || !daddr)
1514 		return NULL;
1515 
1516 	hval = rt6_exception_hash(daddr, saddr);
1517 	*bucket += hval;
1518 
1519 	hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1520 		struct rt6_info *rt6 = rt6_ex->rt6i;
1521 		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1522 
1523 #ifdef CONFIG_IPV6_SUBTREES
1524 		if (matched && saddr)
1525 			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1526 #endif
1527 		if (matched)
1528 			return rt6_ex;
1529 	}
1530 	return NULL;
1531 }
1532 
1533 /* Helper function to find the cached rt in the hash table
1534  * and update bucket pointer to point to the bucket for this
1535  * (daddr, saddr) pair
1536  * Caller must hold rcu_read_lock()
1537  */
1538 static struct rt6_exception *
1539 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1540 			 const struct in6_addr *daddr,
1541 			 const struct in6_addr *saddr)
1542 {
1543 	struct rt6_exception *rt6_ex;
1544 	u32 hval;
1545 
1546 	WARN_ON_ONCE(!rcu_read_lock_held());
1547 
1548 	if (!(*bucket) || !daddr)
1549 		return NULL;
1550 
1551 	hval = rt6_exception_hash(daddr, saddr);
1552 	*bucket += hval;
1553 
1554 	hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1555 		struct rt6_info *rt6 = rt6_ex->rt6i;
1556 		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1557 
1558 #ifdef CONFIG_IPV6_SUBTREES
1559 		if (matched && saddr)
1560 			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1561 #endif
1562 		if (matched)
1563 			return rt6_ex;
1564 	}
1565 	return NULL;
1566 }
1567 
1568 static unsigned int fib6_mtu(const struct fib6_result *res)
1569 {
1570 	const struct fib6_nh *nh = res->nh;
1571 	unsigned int mtu;
1572 
1573 	if (res->f6i->fib6_pmtu) {
1574 		mtu = res->f6i->fib6_pmtu;
1575 	} else {
1576 		struct net_device *dev = nh->fib_nh_dev;
1577 		struct inet6_dev *idev;
1578 
1579 		rcu_read_lock();
1580 		idev = __in6_dev_get(dev);
1581 		mtu = idev->cnf.mtu6;
1582 		rcu_read_unlock();
1583 	}
1584 
1585 	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1586 
1587 	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1588 }
1589 
1590 #define FIB6_EXCEPTION_BUCKET_FLUSHED  0x1UL
1591 
1592 /* used when the flushed bit is not relevant, only access to the bucket
1593  * (ie., all bucket users except rt6_insert_exception);
1594  *
1595  * called under rcu lock; sometimes called with rt6_exception_lock held
1596  */
1597 static
1598 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1599 						       spinlock_t *lock)
1600 {
1601 	struct rt6_exception_bucket *bucket;
1602 
1603 	if (lock)
1604 		bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1605 						   lockdep_is_held(lock));
1606 	else
1607 		bucket = rcu_dereference(nh->rt6i_exception_bucket);
1608 
1609 	/* remove bucket flushed bit if set */
1610 	if (bucket) {
1611 		unsigned long p = (unsigned long)bucket;
1612 
1613 		p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1614 		bucket = (struct rt6_exception_bucket *)p;
1615 	}
1616 
1617 	return bucket;
1618 }
1619 
1620 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1621 {
1622 	unsigned long p = (unsigned long)bucket;
1623 
1624 	return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1625 }
1626 
1627 /* called with rt6_exception_lock held */
1628 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1629 					      spinlock_t *lock)
1630 {
1631 	struct rt6_exception_bucket *bucket;
1632 	unsigned long p;
1633 
1634 	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1635 					   lockdep_is_held(lock));
1636 
1637 	p = (unsigned long)bucket;
1638 	p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1639 	bucket = (struct rt6_exception_bucket *)p;
1640 	rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1641 }
1642 
1643 static int rt6_insert_exception(struct rt6_info *nrt,
1644 				const struct fib6_result *res)
1645 {
1646 	struct net *net = dev_net(nrt->dst.dev);
1647 	struct rt6_exception_bucket *bucket;
1648 	struct fib6_info *f6i = res->f6i;
1649 	struct in6_addr *src_key = NULL;
1650 	struct rt6_exception *rt6_ex;
1651 	struct fib6_nh *nh = res->nh;
1652 	int err = 0;
1653 
1654 	spin_lock_bh(&rt6_exception_lock);
1655 
1656 	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1657 					  lockdep_is_held(&rt6_exception_lock));
1658 	if (!bucket) {
1659 		bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1660 				 GFP_ATOMIC);
1661 		if (!bucket) {
1662 			err = -ENOMEM;
1663 			goto out;
1664 		}
1665 		rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1666 	} else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1667 		err = -EINVAL;
1668 		goto out;
1669 	}
1670 
1671 #ifdef CONFIG_IPV6_SUBTREES
1672 	/* fib6_src.plen != 0 indicates f6i is in subtree
1673 	 * and exception table is indexed by a hash of
1674 	 * both fib6_dst and fib6_src.
1675 	 * Otherwise, the exception table is indexed by
1676 	 * a hash of only fib6_dst.
1677 	 */
1678 	if (f6i->fib6_src.plen)
1679 		src_key = &nrt->rt6i_src.addr;
1680 #endif
1681 	/* rt6_mtu_change() might lower mtu on f6i.
1682 	 * Only insert this exception route if its mtu
1683 	 * is less than f6i's mtu value.
1684 	 */
1685 	if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1686 		err = -EINVAL;
1687 		goto out;
1688 	}
1689 
1690 	rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1691 					       src_key);
1692 	if (rt6_ex)
1693 		rt6_remove_exception(bucket, rt6_ex);
1694 
1695 	rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1696 	if (!rt6_ex) {
1697 		err = -ENOMEM;
1698 		goto out;
1699 	}
1700 	rt6_ex->rt6i = nrt;
1701 	rt6_ex->stamp = jiffies;
1702 	hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1703 	bucket->depth++;
1704 	net->ipv6.rt6_stats->fib_rt_cache++;
1705 
1706 	if (bucket->depth > FIB6_MAX_DEPTH)
1707 		rt6_exception_remove_oldest(bucket);
1708 
1709 out:
1710 	spin_unlock_bh(&rt6_exception_lock);
1711 
1712 	/* Update fn->fn_sernum to invalidate all cached dst */
1713 	if (!err) {
1714 		spin_lock_bh(&f6i->fib6_table->tb6_lock);
1715 		fib6_update_sernum(net, f6i);
1716 		spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1717 		fib6_force_start_gc(net);
1718 	}
1719 
1720 	return err;
1721 }
1722 
1723 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1724 {
1725 	struct rt6_exception_bucket *bucket;
1726 	struct rt6_exception *rt6_ex;
1727 	struct hlist_node *tmp;
1728 	int i;
1729 
1730 	spin_lock_bh(&rt6_exception_lock);
1731 
1732 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1733 	if (!bucket)
1734 		goto out;
1735 
1736 	/* Prevent rt6_insert_exception() to recreate the bucket list */
1737 	if (!from)
1738 		fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1739 
1740 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1741 		hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1742 			if (!from ||
1743 			    rcu_access_pointer(rt6_ex->rt6i->from) == from)
1744 				rt6_remove_exception(bucket, rt6_ex);
1745 		}
1746 		WARN_ON_ONCE(!from && bucket->depth);
1747 		bucket++;
1748 	}
1749 out:
1750 	spin_unlock_bh(&rt6_exception_lock);
1751 }
1752 
1753 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1754 {
1755 	struct fib6_info *f6i = arg;
1756 
1757 	fib6_nh_flush_exceptions(nh, f6i);
1758 
1759 	return 0;
1760 }
1761 
1762 void rt6_flush_exceptions(struct fib6_info *f6i)
1763 {
1764 	if (f6i->nh)
1765 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1766 					 f6i);
1767 	else
1768 		fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1769 }
1770 
1771 /* Find cached rt in the hash table inside passed in rt
1772  * Caller has to hold rcu_read_lock()
1773  */
1774 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1775 					   const struct in6_addr *daddr,
1776 					   const struct in6_addr *saddr)
1777 {
1778 	const struct in6_addr *src_key = NULL;
1779 	struct rt6_exception_bucket *bucket;
1780 	struct rt6_exception *rt6_ex;
1781 	struct rt6_info *ret = NULL;
1782 
1783 #ifdef CONFIG_IPV6_SUBTREES
1784 	/* fib6i_src.plen != 0 indicates f6i is in subtree
1785 	 * and exception table is indexed by a hash of
1786 	 * both fib6_dst and fib6_src.
1787 	 * However, the src addr used to create the hash
1788 	 * might not be exactly the passed in saddr which
1789 	 * is a /128 addr from the flow.
1790 	 * So we need to use f6i->fib6_src to redo lookup
1791 	 * if the passed in saddr does not find anything.
1792 	 * (See the logic in ip6_rt_cache_alloc() on how
1793 	 * rt->rt6i_src is updated.)
1794 	 */
1795 	if (res->f6i->fib6_src.plen)
1796 		src_key = saddr;
1797 find_ex:
1798 #endif
1799 	bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1800 	rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1801 
1802 	if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1803 		ret = rt6_ex->rt6i;
1804 
1805 #ifdef CONFIG_IPV6_SUBTREES
1806 	/* Use fib6_src as src_key and redo lookup */
1807 	if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1808 		src_key = &res->f6i->fib6_src.addr;
1809 		goto find_ex;
1810 	}
1811 #endif
1812 
1813 	return ret;
1814 }
1815 
1816 /* Remove the passed in cached rt from the hash table that contains it */
1817 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1818 				    const struct rt6_info *rt)
1819 {
1820 	const struct in6_addr *src_key = NULL;
1821 	struct rt6_exception_bucket *bucket;
1822 	struct rt6_exception *rt6_ex;
1823 	int err;
1824 
1825 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1826 		return -ENOENT;
1827 
1828 	spin_lock_bh(&rt6_exception_lock);
1829 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1830 
1831 #ifdef CONFIG_IPV6_SUBTREES
1832 	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1833 	 * and exception table is indexed by a hash of
1834 	 * both rt6i_dst and rt6i_src.
1835 	 * Otherwise, the exception table is indexed by
1836 	 * a hash of only rt6i_dst.
1837 	 */
1838 	if (plen)
1839 		src_key = &rt->rt6i_src.addr;
1840 #endif
1841 	rt6_ex = __rt6_find_exception_spinlock(&bucket,
1842 					       &rt->rt6i_dst.addr,
1843 					       src_key);
1844 	if (rt6_ex) {
1845 		rt6_remove_exception(bucket, rt6_ex);
1846 		err = 0;
1847 	} else {
1848 		err = -ENOENT;
1849 	}
1850 
1851 	spin_unlock_bh(&rt6_exception_lock);
1852 	return err;
1853 }
1854 
1855 struct fib6_nh_excptn_arg {
1856 	struct rt6_info	*rt;
1857 	int		plen;
1858 };
1859 
1860 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1861 {
1862 	struct fib6_nh_excptn_arg *arg = _arg;
1863 	int err;
1864 
1865 	err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1866 	if (err == 0)
1867 		return 1;
1868 
1869 	return 0;
1870 }
1871 
1872 static int rt6_remove_exception_rt(struct rt6_info *rt)
1873 {
1874 	struct fib6_info *from;
1875 
1876 	from = rcu_dereference(rt->from);
1877 	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1878 		return -EINVAL;
1879 
1880 	if (from->nh) {
1881 		struct fib6_nh_excptn_arg arg = {
1882 			.rt = rt,
1883 			.plen = from->fib6_src.plen
1884 		};
1885 		int rc;
1886 
1887 		/* rc = 1 means an entry was found */
1888 		rc = nexthop_for_each_fib6_nh(from->nh,
1889 					      rt6_nh_remove_exception_rt,
1890 					      &arg);
1891 		return rc ? 0 : -ENOENT;
1892 	}
1893 
1894 	return fib6_nh_remove_exception(from->fib6_nh,
1895 					from->fib6_src.plen, rt);
1896 }
1897 
1898 /* Find rt6_ex which contains the passed in rt cache and
1899  * refresh its stamp
1900  */
1901 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1902 				     const struct rt6_info *rt)
1903 {
1904 	const struct in6_addr *src_key = NULL;
1905 	struct rt6_exception_bucket *bucket;
1906 	struct rt6_exception *rt6_ex;
1907 
1908 	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1909 #ifdef CONFIG_IPV6_SUBTREES
1910 	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1911 	 * and exception table is indexed by a hash of
1912 	 * both rt6i_dst and rt6i_src.
1913 	 * Otherwise, the exception table is indexed by
1914 	 * a hash of only rt6i_dst.
1915 	 */
1916 	if (plen)
1917 		src_key = &rt->rt6i_src.addr;
1918 #endif
1919 	rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1920 	if (rt6_ex)
1921 		rt6_ex->stamp = jiffies;
1922 }
1923 
1924 struct fib6_nh_match_arg {
1925 	const struct net_device *dev;
1926 	const struct in6_addr	*gw;
1927 	struct fib6_nh		*match;
1928 };
1929 
1930 /* determine if fib6_nh has given device and gateway */
1931 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1932 {
1933 	struct fib6_nh_match_arg *arg = _arg;
1934 
1935 	if (arg->dev != nh->fib_nh_dev ||
1936 	    (arg->gw && !nh->fib_nh_gw_family) ||
1937 	    (!arg->gw && nh->fib_nh_gw_family) ||
1938 	    (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1939 		return 0;
1940 
1941 	arg->match = nh;
1942 
1943 	/* found a match, break the loop */
1944 	return 1;
1945 }
1946 
1947 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1948 {
1949 	struct fib6_info *from;
1950 	struct fib6_nh *fib6_nh;
1951 
1952 	rcu_read_lock();
1953 
1954 	from = rcu_dereference(rt->from);
1955 	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1956 		goto unlock;
1957 
1958 	if (from->nh) {
1959 		struct fib6_nh_match_arg arg = {
1960 			.dev = rt->dst.dev,
1961 			.gw = &rt->rt6i_gateway,
1962 		};
1963 
1964 		nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1965 
1966 		if (!arg.match)
1967 			goto unlock;
1968 		fib6_nh = arg.match;
1969 	} else {
1970 		fib6_nh = from->fib6_nh;
1971 	}
1972 	fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1973 unlock:
1974 	rcu_read_unlock();
1975 }
1976 
1977 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1978 					 struct rt6_info *rt, int mtu)
1979 {
1980 	/* If the new MTU is lower than the route PMTU, this new MTU will be the
1981 	 * lowest MTU in the path: always allow updating the route PMTU to
1982 	 * reflect PMTU decreases.
1983 	 *
1984 	 * If the new MTU is higher, and the route PMTU is equal to the local
1985 	 * MTU, this means the old MTU is the lowest in the path, so allow
1986 	 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1987 	 * handle this.
1988 	 */
1989 
1990 	if (dst_mtu(&rt->dst) >= mtu)
1991 		return true;
1992 
1993 	if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1994 		return true;
1995 
1996 	return false;
1997 }
1998 
1999 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2000 				       const struct fib6_nh *nh, int mtu)
2001 {
2002 	struct rt6_exception_bucket *bucket;
2003 	struct rt6_exception *rt6_ex;
2004 	int i;
2005 
2006 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2007 	if (!bucket)
2008 		return;
2009 
2010 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2011 		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2012 			struct rt6_info *entry = rt6_ex->rt6i;
2013 
2014 			/* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2015 			 * route), the metrics of its rt->from have already
2016 			 * been updated.
2017 			 */
2018 			if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2019 			    rt6_mtu_change_route_allowed(idev, entry, mtu))
2020 				dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2021 		}
2022 		bucket++;
2023 	}
2024 }
2025 
2026 #define RTF_CACHE_GATEWAY	(RTF_GATEWAY | RTF_CACHE)
2027 
2028 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2029 					    const struct in6_addr *gateway)
2030 {
2031 	struct rt6_exception_bucket *bucket;
2032 	struct rt6_exception *rt6_ex;
2033 	struct hlist_node *tmp;
2034 	int i;
2035 
2036 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2037 		return;
2038 
2039 	spin_lock_bh(&rt6_exception_lock);
2040 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2041 	if (bucket) {
2042 		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2043 			hlist_for_each_entry_safe(rt6_ex, tmp,
2044 						  &bucket->chain, hlist) {
2045 				struct rt6_info *entry = rt6_ex->rt6i;
2046 
2047 				if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2048 				    RTF_CACHE_GATEWAY &&
2049 				    ipv6_addr_equal(gateway,
2050 						    &entry->rt6i_gateway)) {
2051 					rt6_remove_exception(bucket, rt6_ex);
2052 				}
2053 			}
2054 			bucket++;
2055 		}
2056 	}
2057 
2058 	spin_unlock_bh(&rt6_exception_lock);
2059 }
2060 
2061 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2062 				      struct rt6_exception *rt6_ex,
2063 				      struct fib6_gc_args *gc_args,
2064 				      unsigned long now)
2065 {
2066 	struct rt6_info *rt = rt6_ex->rt6i;
2067 
2068 	/* we are pruning and obsoleting aged-out and non gateway exceptions
2069 	 * even if others have still references to them, so that on next
2070 	 * dst_check() such references can be dropped.
2071 	 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2072 	 * expired, independently from their aging, as per RFC 8201 section 4
2073 	 */
2074 	if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2075 		if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2076 			RT6_TRACE("aging clone %p\n", rt);
2077 			rt6_remove_exception(bucket, rt6_ex);
2078 			return;
2079 		}
2080 	} else if (time_after(jiffies, rt->dst.expires)) {
2081 		RT6_TRACE("purging expired route %p\n", rt);
2082 		rt6_remove_exception(bucket, rt6_ex);
2083 		return;
2084 	}
2085 
2086 	if (rt->rt6i_flags & RTF_GATEWAY) {
2087 		struct neighbour *neigh;
2088 
2089 		neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2090 
2091 		if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2092 			RT6_TRACE("purging route %p via non-router but gateway\n",
2093 				  rt);
2094 			rt6_remove_exception(bucket, rt6_ex);
2095 			return;
2096 		}
2097 	}
2098 
2099 	gc_args->more++;
2100 }
2101 
2102 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2103 				   struct fib6_gc_args *gc_args,
2104 				   unsigned long now)
2105 {
2106 	struct rt6_exception_bucket *bucket;
2107 	struct rt6_exception *rt6_ex;
2108 	struct hlist_node *tmp;
2109 	int i;
2110 
2111 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2112 		return;
2113 
2114 	rcu_read_lock_bh();
2115 	spin_lock(&rt6_exception_lock);
2116 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2117 	if (bucket) {
2118 		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2119 			hlist_for_each_entry_safe(rt6_ex, tmp,
2120 						  &bucket->chain, hlist) {
2121 				rt6_age_examine_exception(bucket, rt6_ex,
2122 							  gc_args, now);
2123 			}
2124 			bucket++;
2125 		}
2126 	}
2127 	spin_unlock(&rt6_exception_lock);
2128 	rcu_read_unlock_bh();
2129 }
2130 
2131 struct fib6_nh_age_excptn_arg {
2132 	struct fib6_gc_args	*gc_args;
2133 	unsigned long		now;
2134 };
2135 
2136 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2137 {
2138 	struct fib6_nh_age_excptn_arg *arg = _arg;
2139 
2140 	fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2141 	return 0;
2142 }
2143 
2144 void rt6_age_exceptions(struct fib6_info *f6i,
2145 			struct fib6_gc_args *gc_args,
2146 			unsigned long now)
2147 {
2148 	if (f6i->nh) {
2149 		struct fib6_nh_age_excptn_arg arg = {
2150 			.gc_args = gc_args,
2151 			.now = now
2152 		};
2153 
2154 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2155 					 &arg);
2156 	} else {
2157 		fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2158 	}
2159 }
2160 
2161 /* must be called with rcu lock held */
2162 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2163 		      struct flowi6 *fl6, struct fib6_result *res, int strict)
2164 {
2165 	struct fib6_node *fn, *saved_fn;
2166 
2167 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2168 	saved_fn = fn;
2169 
2170 	if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2171 		oif = 0;
2172 
2173 redo_rt6_select:
2174 	rt6_select(net, fn, oif, res, strict);
2175 	if (res->f6i == net->ipv6.fib6_null_entry) {
2176 		fn = fib6_backtrack(fn, &fl6->saddr);
2177 		if (fn)
2178 			goto redo_rt6_select;
2179 		else if (strict & RT6_LOOKUP_F_REACHABLE) {
2180 			/* also consider unreachable route */
2181 			strict &= ~RT6_LOOKUP_F_REACHABLE;
2182 			fn = saved_fn;
2183 			goto redo_rt6_select;
2184 		}
2185 	}
2186 
2187 	trace_fib6_table_lookup(net, res, table, fl6);
2188 
2189 	return 0;
2190 }
2191 
2192 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2193 			       int oif, struct flowi6 *fl6,
2194 			       const struct sk_buff *skb, int flags)
2195 {
2196 	struct fib6_result res = {};
2197 	struct rt6_info *rt = NULL;
2198 	int strict = 0;
2199 
2200 	WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2201 		     !rcu_read_lock_held());
2202 
2203 	strict |= flags & RT6_LOOKUP_F_IFACE;
2204 	strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2205 	if (net->ipv6.devconf_all->forwarding == 0)
2206 		strict |= RT6_LOOKUP_F_REACHABLE;
2207 
2208 	rcu_read_lock();
2209 
2210 	fib6_table_lookup(net, table, oif, fl6, &res, strict);
2211 	if (res.f6i == net->ipv6.fib6_null_entry)
2212 		goto out;
2213 
2214 	fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2215 
2216 	/*Search through exception table */
2217 	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2218 	if (rt) {
2219 		goto out;
2220 	} else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2221 			    !res.nh->fib_nh_gw_family)) {
2222 		/* Create a RTF_CACHE clone which will not be
2223 		 * owned by the fib6 tree.  It is for the special case where
2224 		 * the daddr in the skb during the neighbor look-up is different
2225 		 * from the fl6->daddr used to look-up route here.
2226 		 */
2227 		rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2228 
2229 		if (rt) {
2230 			/* 1 refcnt is taken during ip6_rt_cache_alloc().
2231 			 * As rt6_uncached_list_add() does not consume refcnt,
2232 			 * this refcnt is always returned to the caller even
2233 			 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2234 			 */
2235 			rt6_uncached_list_add(rt);
2236 			atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2237 			rcu_read_unlock();
2238 
2239 			return rt;
2240 		}
2241 	} else {
2242 		/* Get a percpu copy */
2243 		local_bh_disable();
2244 		rt = rt6_get_pcpu_route(&res);
2245 
2246 		if (!rt)
2247 			rt = rt6_make_pcpu_route(net, &res);
2248 
2249 		local_bh_enable();
2250 	}
2251 out:
2252 	if (!rt)
2253 		rt = net->ipv6.ip6_null_entry;
2254 	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2255 		ip6_hold_safe(net, &rt);
2256 	rcu_read_unlock();
2257 
2258 	return rt;
2259 }
2260 EXPORT_SYMBOL_GPL(ip6_pol_route);
2261 
2262 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2263 					    struct fib6_table *table,
2264 					    struct flowi6 *fl6,
2265 					    const struct sk_buff *skb,
2266 					    int flags)
2267 {
2268 	return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2269 }
2270 
2271 struct dst_entry *ip6_route_input_lookup(struct net *net,
2272 					 struct net_device *dev,
2273 					 struct flowi6 *fl6,
2274 					 const struct sk_buff *skb,
2275 					 int flags)
2276 {
2277 	if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2278 		flags |= RT6_LOOKUP_F_IFACE;
2279 
2280 	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2281 }
2282 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2283 
2284 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2285 				  struct flow_keys *keys,
2286 				  struct flow_keys *flkeys)
2287 {
2288 	const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2289 	const struct ipv6hdr *key_iph = outer_iph;
2290 	struct flow_keys *_flkeys = flkeys;
2291 	const struct ipv6hdr *inner_iph;
2292 	const struct icmp6hdr *icmph;
2293 	struct ipv6hdr _inner_iph;
2294 	struct icmp6hdr _icmph;
2295 
2296 	if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2297 		goto out;
2298 
2299 	icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2300 				   sizeof(_icmph), &_icmph);
2301 	if (!icmph)
2302 		goto out;
2303 
2304 	if (!icmpv6_is_err(icmph->icmp6_type))
2305 		goto out;
2306 
2307 	inner_iph = skb_header_pointer(skb,
2308 				       skb_transport_offset(skb) + sizeof(*icmph),
2309 				       sizeof(_inner_iph), &_inner_iph);
2310 	if (!inner_iph)
2311 		goto out;
2312 
2313 	key_iph = inner_iph;
2314 	_flkeys = NULL;
2315 out:
2316 	if (_flkeys) {
2317 		keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2318 		keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2319 		keys->tags.flow_label = _flkeys->tags.flow_label;
2320 		keys->basic.ip_proto = _flkeys->basic.ip_proto;
2321 	} else {
2322 		keys->addrs.v6addrs.src = key_iph->saddr;
2323 		keys->addrs.v6addrs.dst = key_iph->daddr;
2324 		keys->tags.flow_label = ip6_flowlabel(key_iph);
2325 		keys->basic.ip_proto = key_iph->nexthdr;
2326 	}
2327 }
2328 
2329 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2330 					   const struct sk_buff *skb,
2331 					   bool *p_has_inner)
2332 {
2333 	u32 hash_fields = ip6_multipath_hash_fields(net);
2334 	struct flow_keys keys, hash_keys;
2335 
2336 	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2337 		return 0;
2338 
2339 	memset(&hash_keys, 0, sizeof(hash_keys));
2340 	skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2341 
2342 	hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2343 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2344 		hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2345 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2346 		hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2347 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2348 		hash_keys.basic.ip_proto = keys.basic.ip_proto;
2349 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2350 		hash_keys.tags.flow_label = keys.tags.flow_label;
2351 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2352 		hash_keys.ports.src = keys.ports.src;
2353 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2354 		hash_keys.ports.dst = keys.ports.dst;
2355 
2356 	*p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2357 	return flow_hash_from_keys(&hash_keys);
2358 }
2359 
2360 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2361 					   const struct sk_buff *skb,
2362 					   bool has_inner)
2363 {
2364 	u32 hash_fields = ip6_multipath_hash_fields(net);
2365 	struct flow_keys keys, hash_keys;
2366 
2367 	/* We assume the packet carries an encapsulation, but if none was
2368 	 * encountered during dissection of the outer flow, then there is no
2369 	 * point in calling the flow dissector again.
2370 	 */
2371 	if (!has_inner)
2372 		return 0;
2373 
2374 	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2375 		return 0;
2376 
2377 	memset(&hash_keys, 0, sizeof(hash_keys));
2378 	skb_flow_dissect_flow_keys(skb, &keys, 0);
2379 
2380 	if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2381 		return 0;
2382 
2383 	if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2384 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2385 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2386 			hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2387 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2388 			hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2389 	} else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2390 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2391 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2392 			hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2393 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2394 			hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2395 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2396 			hash_keys.tags.flow_label = keys.tags.flow_label;
2397 	}
2398 
2399 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2400 		hash_keys.basic.ip_proto = keys.basic.ip_proto;
2401 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2402 		hash_keys.ports.src = keys.ports.src;
2403 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2404 		hash_keys.ports.dst = keys.ports.dst;
2405 
2406 	return flow_hash_from_keys(&hash_keys);
2407 }
2408 
2409 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2410 					 const struct sk_buff *skb)
2411 {
2412 	u32 mhash, mhash_inner;
2413 	bool has_inner = true;
2414 
2415 	mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2416 	mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2417 
2418 	return jhash_2words(mhash, mhash_inner, 0);
2419 }
2420 
2421 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2422 					 const struct flowi6 *fl6)
2423 {
2424 	u32 hash_fields = ip6_multipath_hash_fields(net);
2425 	struct flow_keys hash_keys;
2426 
2427 	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2428 		return 0;
2429 
2430 	memset(&hash_keys, 0, sizeof(hash_keys));
2431 	hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2432 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2433 		hash_keys.addrs.v6addrs.src = fl6->saddr;
2434 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2435 		hash_keys.addrs.v6addrs.dst = fl6->daddr;
2436 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2437 		hash_keys.basic.ip_proto = fl6->flowi6_proto;
2438 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2439 		hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2440 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2441 		hash_keys.ports.src = fl6->fl6_sport;
2442 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2443 		hash_keys.ports.dst = fl6->fl6_dport;
2444 
2445 	return flow_hash_from_keys(&hash_keys);
2446 }
2447 
2448 /* if skb is set it will be used and fl6 can be NULL */
2449 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2450 		       const struct sk_buff *skb, struct flow_keys *flkeys)
2451 {
2452 	struct flow_keys hash_keys;
2453 	u32 mhash = 0;
2454 
2455 	switch (ip6_multipath_hash_policy(net)) {
2456 	case 0:
2457 		memset(&hash_keys, 0, sizeof(hash_keys));
2458 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2459 		if (skb) {
2460 			ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2461 		} else {
2462 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2463 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2464 			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2465 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2466 		}
2467 		mhash = flow_hash_from_keys(&hash_keys);
2468 		break;
2469 	case 1:
2470 		if (skb) {
2471 			unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2472 			struct flow_keys keys;
2473 
2474 			/* short-circuit if we already have L4 hash present */
2475 			if (skb->l4_hash)
2476 				return skb_get_hash_raw(skb) >> 1;
2477 
2478 			memset(&hash_keys, 0, sizeof(hash_keys));
2479 
2480 			if (!flkeys) {
2481 				skb_flow_dissect_flow_keys(skb, &keys, flag);
2482 				flkeys = &keys;
2483 			}
2484 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2485 			hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2486 			hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2487 			hash_keys.ports.src = flkeys->ports.src;
2488 			hash_keys.ports.dst = flkeys->ports.dst;
2489 			hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2490 		} else {
2491 			memset(&hash_keys, 0, sizeof(hash_keys));
2492 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2493 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2494 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2495 			hash_keys.ports.src = fl6->fl6_sport;
2496 			hash_keys.ports.dst = fl6->fl6_dport;
2497 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2498 		}
2499 		mhash = flow_hash_from_keys(&hash_keys);
2500 		break;
2501 	case 2:
2502 		memset(&hash_keys, 0, sizeof(hash_keys));
2503 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2504 		if (skb) {
2505 			struct flow_keys keys;
2506 
2507 			if (!flkeys) {
2508 				skb_flow_dissect_flow_keys(skb, &keys, 0);
2509 				flkeys = &keys;
2510 			}
2511 
2512 			/* Inner can be v4 or v6 */
2513 			if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2514 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2515 				hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2516 				hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2517 			} else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2518 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2519 				hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2520 				hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2521 				hash_keys.tags.flow_label = flkeys->tags.flow_label;
2522 				hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2523 			} else {
2524 				/* Same as case 0 */
2525 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2526 				ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2527 			}
2528 		} else {
2529 			/* Same as case 0 */
2530 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2531 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2532 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2533 			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2534 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2535 		}
2536 		mhash = flow_hash_from_keys(&hash_keys);
2537 		break;
2538 	case 3:
2539 		if (skb)
2540 			mhash = rt6_multipath_custom_hash_skb(net, skb);
2541 		else
2542 			mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2543 		break;
2544 	}
2545 
2546 	return mhash >> 1;
2547 }
2548 
2549 /* Called with rcu held */
2550 void ip6_route_input(struct sk_buff *skb)
2551 {
2552 	const struct ipv6hdr *iph = ipv6_hdr(skb);
2553 	struct net *net = dev_net(skb->dev);
2554 	int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2555 	struct ip_tunnel_info *tun_info;
2556 	struct flowi6 fl6 = {
2557 		.flowi6_iif = skb->dev->ifindex,
2558 		.daddr = iph->daddr,
2559 		.saddr = iph->saddr,
2560 		.flowlabel = ip6_flowinfo(iph),
2561 		.flowi6_mark = skb->mark,
2562 		.flowi6_proto = iph->nexthdr,
2563 	};
2564 	struct flow_keys *flkeys = NULL, _flkeys;
2565 
2566 	tun_info = skb_tunnel_info(skb);
2567 	if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2568 		fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2569 
2570 	if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2571 		flkeys = &_flkeys;
2572 
2573 	if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2574 		fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2575 	skb_dst_drop(skb);
2576 	skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2577 						      &fl6, skb, flags));
2578 }
2579 
2580 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2581 					     struct fib6_table *table,
2582 					     struct flowi6 *fl6,
2583 					     const struct sk_buff *skb,
2584 					     int flags)
2585 {
2586 	return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2587 }
2588 
2589 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2590 					       const struct sock *sk,
2591 					       struct flowi6 *fl6, int flags)
2592 {
2593 	bool any_src;
2594 
2595 	if (ipv6_addr_type(&fl6->daddr) &
2596 	    (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2597 		struct dst_entry *dst;
2598 
2599 		/* This function does not take refcnt on the dst */
2600 		dst = l3mdev_link_scope_lookup(net, fl6);
2601 		if (dst)
2602 			return dst;
2603 	}
2604 
2605 	fl6->flowi6_iif = LOOPBACK_IFINDEX;
2606 
2607 	flags |= RT6_LOOKUP_F_DST_NOREF;
2608 	any_src = ipv6_addr_any(&fl6->saddr);
2609 	if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2610 	    (fl6->flowi6_oif && any_src))
2611 		flags |= RT6_LOOKUP_F_IFACE;
2612 
2613 	if (!any_src)
2614 		flags |= RT6_LOOKUP_F_HAS_SADDR;
2615 	else if (sk)
2616 		flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2617 
2618 	return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2619 }
2620 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2621 
2622 struct dst_entry *ip6_route_output_flags(struct net *net,
2623 					 const struct sock *sk,
2624 					 struct flowi6 *fl6,
2625 					 int flags)
2626 {
2627 	struct dst_entry *dst;
2628 	struct rt6_info *rt6;
2629 
2630 	rcu_read_lock();
2631 	dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2632 	rt6 = (struct rt6_info *)dst;
2633 	/* For dst cached in uncached_list, refcnt is already taken. */
2634 	if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2635 		dst = &net->ipv6.ip6_null_entry->dst;
2636 		dst_hold(dst);
2637 	}
2638 	rcu_read_unlock();
2639 
2640 	return dst;
2641 }
2642 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2643 
2644 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2645 {
2646 	struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2647 	struct net_device *loopback_dev = net->loopback_dev;
2648 	struct dst_entry *new = NULL;
2649 
2650 	rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2651 		       DST_OBSOLETE_DEAD, 0);
2652 	if (rt) {
2653 		rt6_info_init(rt);
2654 		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2655 
2656 		new = &rt->dst;
2657 		new->__use = 1;
2658 		new->input = dst_discard;
2659 		new->output = dst_discard_out;
2660 
2661 		dst_copy_metrics(new, &ort->dst);
2662 
2663 		rt->rt6i_idev = in6_dev_get(loopback_dev);
2664 		rt->rt6i_gateway = ort->rt6i_gateway;
2665 		rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2666 
2667 		memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2668 #ifdef CONFIG_IPV6_SUBTREES
2669 		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2670 #endif
2671 	}
2672 
2673 	dst_release(dst_orig);
2674 	return new ? new : ERR_PTR(-ENOMEM);
2675 }
2676 
2677 /*
2678  *	Destination cache support functions
2679  */
2680 
2681 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2682 {
2683 	u32 rt_cookie = 0;
2684 
2685 	if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2686 		return false;
2687 
2688 	if (fib6_check_expired(f6i))
2689 		return false;
2690 
2691 	return true;
2692 }
2693 
2694 static struct dst_entry *rt6_check(struct rt6_info *rt,
2695 				   struct fib6_info *from,
2696 				   u32 cookie)
2697 {
2698 	u32 rt_cookie = 0;
2699 
2700 	if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2701 	    rt_cookie != cookie)
2702 		return NULL;
2703 
2704 	if (rt6_check_expired(rt))
2705 		return NULL;
2706 
2707 	return &rt->dst;
2708 }
2709 
2710 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2711 					    struct fib6_info *from,
2712 					    u32 cookie)
2713 {
2714 	if (!__rt6_check_expired(rt) &&
2715 	    rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2716 	    fib6_check(from, cookie))
2717 		return &rt->dst;
2718 	else
2719 		return NULL;
2720 }
2721 
2722 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2723 							u32 cookie)
2724 {
2725 	struct dst_entry *dst_ret;
2726 	struct fib6_info *from;
2727 	struct rt6_info *rt;
2728 
2729 	rt = container_of(dst, struct rt6_info, dst);
2730 
2731 	if (rt->sernum)
2732 		return rt6_is_valid(rt) ? dst : NULL;
2733 
2734 	rcu_read_lock();
2735 
2736 	/* All IPV6 dsts are created with ->obsolete set to the value
2737 	 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2738 	 * into this function always.
2739 	 */
2740 
2741 	from = rcu_dereference(rt->from);
2742 
2743 	if (from && (rt->rt6i_flags & RTF_PCPU ||
2744 	    unlikely(!list_empty(&rt->rt6i_uncached))))
2745 		dst_ret = rt6_dst_from_check(rt, from, cookie);
2746 	else
2747 		dst_ret = rt6_check(rt, from, cookie);
2748 
2749 	rcu_read_unlock();
2750 
2751 	return dst_ret;
2752 }
2753 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2754 
2755 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2756 {
2757 	struct rt6_info *rt = (struct rt6_info *) dst;
2758 
2759 	if (rt) {
2760 		if (rt->rt6i_flags & RTF_CACHE) {
2761 			rcu_read_lock();
2762 			if (rt6_check_expired(rt)) {
2763 				rt6_remove_exception_rt(rt);
2764 				dst = NULL;
2765 			}
2766 			rcu_read_unlock();
2767 		} else {
2768 			dst_release(dst);
2769 			dst = NULL;
2770 		}
2771 	}
2772 	return dst;
2773 }
2774 
2775 static void ip6_link_failure(struct sk_buff *skb)
2776 {
2777 	struct rt6_info *rt;
2778 
2779 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2780 
2781 	rt = (struct rt6_info *) skb_dst(skb);
2782 	if (rt) {
2783 		rcu_read_lock();
2784 		if (rt->rt6i_flags & RTF_CACHE) {
2785 			rt6_remove_exception_rt(rt);
2786 		} else {
2787 			struct fib6_info *from;
2788 			struct fib6_node *fn;
2789 
2790 			from = rcu_dereference(rt->from);
2791 			if (from) {
2792 				fn = rcu_dereference(from->fib6_node);
2793 				if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2794 					fn->fn_sernum = -1;
2795 			}
2796 		}
2797 		rcu_read_unlock();
2798 	}
2799 }
2800 
2801 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2802 {
2803 	if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2804 		struct fib6_info *from;
2805 
2806 		rcu_read_lock();
2807 		from = rcu_dereference(rt0->from);
2808 		if (from)
2809 			rt0->dst.expires = from->expires;
2810 		rcu_read_unlock();
2811 	}
2812 
2813 	dst_set_expires(&rt0->dst, timeout);
2814 	rt0->rt6i_flags |= RTF_EXPIRES;
2815 }
2816 
2817 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2818 {
2819 	struct net *net = dev_net(rt->dst.dev);
2820 
2821 	dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2822 	rt->rt6i_flags |= RTF_MODIFIED;
2823 	rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2824 }
2825 
2826 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2827 {
2828 	return !(rt->rt6i_flags & RTF_CACHE) &&
2829 		(rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2830 }
2831 
2832 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2833 				 const struct ipv6hdr *iph, u32 mtu,
2834 				 bool confirm_neigh)
2835 {
2836 	const struct in6_addr *daddr, *saddr;
2837 	struct rt6_info *rt6 = (struct rt6_info *)dst;
2838 
2839 	/* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2840 	 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2841 	 * [see also comment in rt6_mtu_change_route()]
2842 	 */
2843 
2844 	if (iph) {
2845 		daddr = &iph->daddr;
2846 		saddr = &iph->saddr;
2847 	} else if (sk) {
2848 		daddr = &sk->sk_v6_daddr;
2849 		saddr = &inet6_sk(sk)->saddr;
2850 	} else {
2851 		daddr = NULL;
2852 		saddr = NULL;
2853 	}
2854 
2855 	if (confirm_neigh)
2856 		dst_confirm_neigh(dst, daddr);
2857 
2858 	if (mtu < IPV6_MIN_MTU)
2859 		return;
2860 	if (mtu >= dst_mtu(dst))
2861 		return;
2862 
2863 	if (!rt6_cache_allowed_for_pmtu(rt6)) {
2864 		rt6_do_update_pmtu(rt6, mtu);
2865 		/* update rt6_ex->stamp for cache */
2866 		if (rt6->rt6i_flags & RTF_CACHE)
2867 			rt6_update_exception_stamp_rt(rt6);
2868 	} else if (daddr) {
2869 		struct fib6_result res = {};
2870 		struct rt6_info *nrt6;
2871 
2872 		rcu_read_lock();
2873 		res.f6i = rcu_dereference(rt6->from);
2874 		if (!res.f6i)
2875 			goto out_unlock;
2876 
2877 		res.fib6_flags = res.f6i->fib6_flags;
2878 		res.fib6_type = res.f6i->fib6_type;
2879 
2880 		if (res.f6i->nh) {
2881 			struct fib6_nh_match_arg arg = {
2882 				.dev = dst->dev,
2883 				.gw = &rt6->rt6i_gateway,
2884 			};
2885 
2886 			nexthop_for_each_fib6_nh(res.f6i->nh,
2887 						 fib6_nh_find_match, &arg);
2888 
2889 			/* fib6_info uses a nexthop that does not have fib6_nh
2890 			 * using the dst->dev + gw. Should be impossible.
2891 			 */
2892 			if (!arg.match)
2893 				goto out_unlock;
2894 
2895 			res.nh = arg.match;
2896 		} else {
2897 			res.nh = res.f6i->fib6_nh;
2898 		}
2899 
2900 		nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2901 		if (nrt6) {
2902 			rt6_do_update_pmtu(nrt6, mtu);
2903 			if (rt6_insert_exception(nrt6, &res))
2904 				dst_release_immediate(&nrt6->dst);
2905 		}
2906 out_unlock:
2907 		rcu_read_unlock();
2908 	}
2909 }
2910 
2911 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2912 			       struct sk_buff *skb, u32 mtu,
2913 			       bool confirm_neigh)
2914 {
2915 	__ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2916 			     confirm_neigh);
2917 }
2918 
2919 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2920 		     int oif, u32 mark, kuid_t uid)
2921 {
2922 	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2923 	struct dst_entry *dst;
2924 	struct flowi6 fl6 = {
2925 		.flowi6_oif = oif,
2926 		.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2927 		.daddr = iph->daddr,
2928 		.saddr = iph->saddr,
2929 		.flowlabel = ip6_flowinfo(iph),
2930 		.flowi6_uid = uid,
2931 	};
2932 
2933 	dst = ip6_route_output(net, NULL, &fl6);
2934 	if (!dst->error)
2935 		__ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2936 	dst_release(dst);
2937 }
2938 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2939 
2940 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2941 {
2942 	int oif = sk->sk_bound_dev_if;
2943 	struct dst_entry *dst;
2944 
2945 	if (!oif && skb->dev)
2946 		oif = l3mdev_master_ifindex(skb->dev);
2947 
2948 	ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2949 
2950 	dst = __sk_dst_get(sk);
2951 	if (!dst || !dst->obsolete ||
2952 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2953 		return;
2954 
2955 	bh_lock_sock(sk);
2956 	if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2957 		ip6_datagram_dst_update(sk, false);
2958 	bh_unlock_sock(sk);
2959 }
2960 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2961 
2962 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2963 			   const struct flowi6 *fl6)
2964 {
2965 #ifdef CONFIG_IPV6_SUBTREES
2966 	struct ipv6_pinfo *np = inet6_sk(sk);
2967 #endif
2968 
2969 	ip6_dst_store(sk, dst,
2970 		      ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2971 		      &sk->sk_v6_daddr : NULL,
2972 #ifdef CONFIG_IPV6_SUBTREES
2973 		      ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2974 		      &np->saddr :
2975 #endif
2976 		      NULL);
2977 }
2978 
2979 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2980 				  struct flowi6 *fl6,
2981 				  const struct in6_addr *gw,
2982 				  struct rt6_info **ret)
2983 {
2984 	const struct fib6_nh *nh = res->nh;
2985 
2986 	if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2987 	    fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2988 		return false;
2989 
2990 	/* rt_cache's gateway might be different from its 'parent'
2991 	 * in the case of an ip redirect.
2992 	 * So we keep searching in the exception table if the gateway
2993 	 * is different.
2994 	 */
2995 	if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2996 		struct rt6_info *rt_cache;
2997 
2998 		rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
2999 		if (rt_cache &&
3000 		    ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3001 			*ret = rt_cache;
3002 			return true;
3003 		}
3004 		return false;
3005 	}
3006 	return true;
3007 }
3008 
3009 struct fib6_nh_rd_arg {
3010 	struct fib6_result	*res;
3011 	struct flowi6		*fl6;
3012 	const struct in6_addr	*gw;
3013 	struct rt6_info		**ret;
3014 };
3015 
3016 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3017 {
3018 	struct fib6_nh_rd_arg *arg = _arg;
3019 
3020 	arg->res->nh = nh;
3021 	return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3022 }
3023 
3024 /* Handle redirects */
3025 struct ip6rd_flowi {
3026 	struct flowi6 fl6;
3027 	struct in6_addr gateway;
3028 };
3029 
3030 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3031 					     struct fib6_table *table,
3032 					     struct flowi6 *fl6,
3033 					     const struct sk_buff *skb,
3034 					     int flags)
3035 {
3036 	struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3037 	struct rt6_info *ret = NULL;
3038 	struct fib6_result res = {};
3039 	struct fib6_nh_rd_arg arg = {
3040 		.res = &res,
3041 		.fl6 = fl6,
3042 		.gw  = &rdfl->gateway,
3043 		.ret = &ret
3044 	};
3045 	struct fib6_info *rt;
3046 	struct fib6_node *fn;
3047 
3048 	/* l3mdev_update_flow overrides oif if the device is enslaved; in
3049 	 * this case we must match on the real ingress device, so reset it
3050 	 */
3051 	if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
3052 		fl6->flowi6_oif = skb->dev->ifindex;
3053 
3054 	/* Get the "current" route for this destination and
3055 	 * check if the redirect has come from appropriate router.
3056 	 *
3057 	 * RFC 4861 specifies that redirects should only be
3058 	 * accepted if they come from the nexthop to the target.
3059 	 * Due to the way the routes are chosen, this notion
3060 	 * is a bit fuzzy and one might need to check all possible
3061 	 * routes.
3062 	 */
3063 
3064 	rcu_read_lock();
3065 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3066 restart:
3067 	for_each_fib6_node_rt_rcu(fn) {
3068 		res.f6i = rt;
3069 		if (fib6_check_expired(rt))
3070 			continue;
3071 		if (rt->fib6_flags & RTF_REJECT)
3072 			break;
3073 		if (unlikely(rt->nh)) {
3074 			if (nexthop_is_blackhole(rt->nh))
3075 				continue;
3076 			/* on match, res->nh is filled in and potentially ret */
3077 			if (nexthop_for_each_fib6_nh(rt->nh,
3078 						     fib6_nh_redirect_match,
3079 						     &arg))
3080 				goto out;
3081 		} else {
3082 			res.nh = rt->fib6_nh;
3083 			if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3084 						  &ret))
3085 				goto out;
3086 		}
3087 	}
3088 
3089 	if (!rt)
3090 		rt = net->ipv6.fib6_null_entry;
3091 	else if (rt->fib6_flags & RTF_REJECT) {
3092 		ret = net->ipv6.ip6_null_entry;
3093 		goto out;
3094 	}
3095 
3096 	if (rt == net->ipv6.fib6_null_entry) {
3097 		fn = fib6_backtrack(fn, &fl6->saddr);
3098 		if (fn)
3099 			goto restart;
3100 	}
3101 
3102 	res.f6i = rt;
3103 	res.nh = rt->fib6_nh;
3104 out:
3105 	if (ret) {
3106 		ip6_hold_safe(net, &ret);
3107 	} else {
3108 		res.fib6_flags = res.f6i->fib6_flags;
3109 		res.fib6_type = res.f6i->fib6_type;
3110 		ret = ip6_create_rt_rcu(&res);
3111 	}
3112 
3113 	rcu_read_unlock();
3114 
3115 	trace_fib6_table_lookup(net, &res, table, fl6);
3116 	return ret;
3117 };
3118 
3119 static struct dst_entry *ip6_route_redirect(struct net *net,
3120 					    const struct flowi6 *fl6,
3121 					    const struct sk_buff *skb,
3122 					    const struct in6_addr *gateway)
3123 {
3124 	int flags = RT6_LOOKUP_F_HAS_SADDR;
3125 	struct ip6rd_flowi rdfl;
3126 
3127 	rdfl.fl6 = *fl6;
3128 	rdfl.gateway = *gateway;
3129 
3130 	return fib6_rule_lookup(net, &rdfl.fl6, skb,
3131 				flags, __ip6_route_redirect);
3132 }
3133 
3134 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3135 		  kuid_t uid)
3136 {
3137 	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3138 	struct dst_entry *dst;
3139 	struct flowi6 fl6 = {
3140 		.flowi6_iif = LOOPBACK_IFINDEX,
3141 		.flowi6_oif = oif,
3142 		.flowi6_mark = mark,
3143 		.daddr = iph->daddr,
3144 		.saddr = iph->saddr,
3145 		.flowlabel = ip6_flowinfo(iph),
3146 		.flowi6_uid = uid,
3147 	};
3148 
3149 	dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3150 	rt6_do_redirect(dst, NULL, skb);
3151 	dst_release(dst);
3152 }
3153 EXPORT_SYMBOL_GPL(ip6_redirect);
3154 
3155 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3156 {
3157 	const struct ipv6hdr *iph = ipv6_hdr(skb);
3158 	const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3159 	struct dst_entry *dst;
3160 	struct flowi6 fl6 = {
3161 		.flowi6_iif = LOOPBACK_IFINDEX,
3162 		.flowi6_oif = oif,
3163 		.daddr = msg->dest,
3164 		.saddr = iph->daddr,
3165 		.flowi6_uid = sock_net_uid(net, NULL),
3166 	};
3167 
3168 	dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3169 	rt6_do_redirect(dst, NULL, skb);
3170 	dst_release(dst);
3171 }
3172 
3173 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3174 {
3175 	ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3176 		     sk->sk_uid);
3177 }
3178 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3179 
3180 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3181 {
3182 	struct net_device *dev = dst->dev;
3183 	unsigned int mtu = dst_mtu(dst);
3184 	struct net *net = dev_net(dev);
3185 
3186 	mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3187 
3188 	if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3189 		mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3190 
3191 	/*
3192 	 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3193 	 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3194 	 * IPV6_MAXPLEN is also valid and means: "any MSS,
3195 	 * rely only on pmtu discovery"
3196 	 */
3197 	if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3198 		mtu = IPV6_MAXPLEN;
3199 	return mtu;
3200 }
3201 
3202 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3203 {
3204 	return ip6_dst_mtu_maybe_forward(dst, false);
3205 }
3206 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3207 
3208 /* MTU selection:
3209  * 1. mtu on route is locked - use it
3210  * 2. mtu from nexthop exception
3211  * 3. mtu from egress device
3212  *
3213  * based on ip6_dst_mtu_forward and exception logic of
3214  * rt6_find_cached_rt; called with rcu_read_lock
3215  */
3216 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3217 		      const struct in6_addr *daddr,
3218 		      const struct in6_addr *saddr)
3219 {
3220 	const struct fib6_nh *nh = res->nh;
3221 	struct fib6_info *f6i = res->f6i;
3222 	struct inet6_dev *idev;
3223 	struct rt6_info *rt;
3224 	u32 mtu = 0;
3225 
3226 	if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3227 		mtu = f6i->fib6_pmtu;
3228 		if (mtu)
3229 			goto out;
3230 	}
3231 
3232 	rt = rt6_find_cached_rt(res, daddr, saddr);
3233 	if (unlikely(rt)) {
3234 		mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3235 	} else {
3236 		struct net_device *dev = nh->fib_nh_dev;
3237 
3238 		mtu = IPV6_MIN_MTU;
3239 		idev = __in6_dev_get(dev);
3240 		if (idev && idev->cnf.mtu6 > mtu)
3241 			mtu = idev->cnf.mtu6;
3242 	}
3243 
3244 	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3245 out:
3246 	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3247 }
3248 
3249 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3250 				  struct flowi6 *fl6)
3251 {
3252 	struct dst_entry *dst;
3253 	struct rt6_info *rt;
3254 	struct inet6_dev *idev = in6_dev_get(dev);
3255 	struct net *net = dev_net(dev);
3256 
3257 	if (unlikely(!idev))
3258 		return ERR_PTR(-ENODEV);
3259 
3260 	rt = ip6_dst_alloc(net, dev, 0);
3261 	if (unlikely(!rt)) {
3262 		in6_dev_put(idev);
3263 		dst = ERR_PTR(-ENOMEM);
3264 		goto out;
3265 	}
3266 
3267 	rt->dst.input = ip6_input;
3268 	rt->dst.output  = ip6_output;
3269 	rt->rt6i_gateway  = fl6->daddr;
3270 	rt->rt6i_dst.addr = fl6->daddr;
3271 	rt->rt6i_dst.plen = 128;
3272 	rt->rt6i_idev     = idev;
3273 	dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3274 
3275 	/* Add this dst into uncached_list so that rt6_disable_ip() can
3276 	 * do proper release of the net_device
3277 	 */
3278 	rt6_uncached_list_add(rt);
3279 	atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
3280 
3281 	dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3282 
3283 out:
3284 	return dst;
3285 }
3286 
3287 static int ip6_dst_gc(struct dst_ops *ops)
3288 {
3289 	struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3290 	int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3291 	int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3292 	int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3293 	int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3294 	unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3295 	int entries;
3296 
3297 	entries = dst_entries_get_fast(ops);
3298 	if (entries > rt_max_size)
3299 		entries = dst_entries_get_slow(ops);
3300 
3301 	if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3302 	    entries <= rt_max_size)
3303 		goto out;
3304 
3305 	net->ipv6.ip6_rt_gc_expire++;
3306 	fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
3307 	entries = dst_entries_get_slow(ops);
3308 	if (entries < ops->gc_thresh)
3309 		net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
3310 out:
3311 	net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
3312 	return entries > rt_max_size;
3313 }
3314 
3315 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3316 			       const struct in6_addr *gw_addr, u32 tbid,
3317 			       int flags, struct fib6_result *res)
3318 {
3319 	struct flowi6 fl6 = {
3320 		.flowi6_oif = cfg->fc_ifindex,
3321 		.daddr = *gw_addr,
3322 		.saddr = cfg->fc_prefsrc,
3323 	};
3324 	struct fib6_table *table;
3325 	int err;
3326 
3327 	table = fib6_get_table(net, tbid);
3328 	if (!table)
3329 		return -EINVAL;
3330 
3331 	if (!ipv6_addr_any(&cfg->fc_prefsrc))
3332 		flags |= RT6_LOOKUP_F_HAS_SADDR;
3333 
3334 	flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3335 
3336 	err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3337 	if (!err && res->f6i != net->ipv6.fib6_null_entry)
3338 		fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3339 				 cfg->fc_ifindex != 0, NULL, flags);
3340 
3341 	return err;
3342 }
3343 
3344 static int ip6_route_check_nh_onlink(struct net *net,
3345 				     struct fib6_config *cfg,
3346 				     const struct net_device *dev,
3347 				     struct netlink_ext_ack *extack)
3348 {
3349 	u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3350 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3351 	struct fib6_result res = {};
3352 	int err;
3353 
3354 	err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3355 	if (!err && !(res.fib6_flags & RTF_REJECT) &&
3356 	    /* ignore match if it is the default route */
3357 	    !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3358 	    (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3359 		NL_SET_ERR_MSG(extack,
3360 			       "Nexthop has invalid gateway or device mismatch");
3361 		err = -EINVAL;
3362 	}
3363 
3364 	return err;
3365 }
3366 
3367 static int ip6_route_check_nh(struct net *net,
3368 			      struct fib6_config *cfg,
3369 			      struct net_device **_dev,
3370 			      struct inet6_dev **idev)
3371 {
3372 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3373 	struct net_device *dev = _dev ? *_dev : NULL;
3374 	int flags = RT6_LOOKUP_F_IFACE;
3375 	struct fib6_result res = {};
3376 	int err = -EHOSTUNREACH;
3377 
3378 	if (cfg->fc_table) {
3379 		err = ip6_nh_lookup_table(net, cfg, gw_addr,
3380 					  cfg->fc_table, flags, &res);
3381 		/* gw_addr can not require a gateway or resolve to a reject
3382 		 * route. If a device is given, it must match the result.
3383 		 */
3384 		if (err || res.fib6_flags & RTF_REJECT ||
3385 		    res.nh->fib_nh_gw_family ||
3386 		    (dev && dev != res.nh->fib_nh_dev))
3387 			err = -EHOSTUNREACH;
3388 	}
3389 
3390 	if (err < 0) {
3391 		struct flowi6 fl6 = {
3392 			.flowi6_oif = cfg->fc_ifindex,
3393 			.daddr = *gw_addr,
3394 		};
3395 
3396 		err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3397 		if (err || res.fib6_flags & RTF_REJECT ||
3398 		    res.nh->fib_nh_gw_family)
3399 			err = -EHOSTUNREACH;
3400 
3401 		if (err)
3402 			return err;
3403 
3404 		fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3405 				 cfg->fc_ifindex != 0, NULL, flags);
3406 	}
3407 
3408 	err = 0;
3409 	if (dev) {
3410 		if (dev != res.nh->fib_nh_dev)
3411 			err = -EHOSTUNREACH;
3412 	} else {
3413 		*_dev = dev = res.nh->fib_nh_dev;
3414 		dev_hold(dev);
3415 		*idev = in6_dev_get(dev);
3416 	}
3417 
3418 	return err;
3419 }
3420 
3421 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3422 			   struct net_device **_dev, struct inet6_dev **idev,
3423 			   struct netlink_ext_ack *extack)
3424 {
3425 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3426 	int gwa_type = ipv6_addr_type(gw_addr);
3427 	bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3428 	const struct net_device *dev = *_dev;
3429 	bool need_addr_check = !dev;
3430 	int err = -EINVAL;
3431 
3432 	/* if gw_addr is local we will fail to detect this in case
3433 	 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3434 	 * will return already-added prefix route via interface that
3435 	 * prefix route was assigned to, which might be non-loopback.
3436 	 */
3437 	if (dev &&
3438 	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3439 		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3440 		goto out;
3441 	}
3442 
3443 	if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3444 		/* IPv6 strictly inhibits using not link-local
3445 		 * addresses as nexthop address.
3446 		 * Otherwise, router will not able to send redirects.
3447 		 * It is very good, but in some (rare!) circumstances
3448 		 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3449 		 * some exceptions. --ANK
3450 		 * We allow IPv4-mapped nexthops to support RFC4798-type
3451 		 * addressing
3452 		 */
3453 		if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3454 			NL_SET_ERR_MSG(extack, "Invalid gateway address");
3455 			goto out;
3456 		}
3457 
3458 		rcu_read_lock();
3459 
3460 		if (cfg->fc_flags & RTNH_F_ONLINK)
3461 			err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3462 		else
3463 			err = ip6_route_check_nh(net, cfg, _dev, idev);
3464 
3465 		rcu_read_unlock();
3466 
3467 		if (err)
3468 			goto out;
3469 	}
3470 
3471 	/* reload in case device was changed */
3472 	dev = *_dev;
3473 
3474 	err = -EINVAL;
3475 	if (!dev) {
3476 		NL_SET_ERR_MSG(extack, "Egress device not specified");
3477 		goto out;
3478 	} else if (dev->flags & IFF_LOOPBACK) {
3479 		NL_SET_ERR_MSG(extack,
3480 			       "Egress device can not be loopback device for this route");
3481 		goto out;
3482 	}
3483 
3484 	/* if we did not check gw_addr above, do so now that the
3485 	 * egress device has been resolved.
3486 	 */
3487 	if (need_addr_check &&
3488 	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3489 		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3490 		goto out;
3491 	}
3492 
3493 	err = 0;
3494 out:
3495 	return err;
3496 }
3497 
3498 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3499 {
3500 	if ((flags & RTF_REJECT) ||
3501 	    (dev && (dev->flags & IFF_LOOPBACK) &&
3502 	     !(addr_type & IPV6_ADDR_LOOPBACK) &&
3503 	     !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3504 		return true;
3505 
3506 	return false;
3507 }
3508 
3509 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3510 		 struct fib6_config *cfg, gfp_t gfp_flags,
3511 		 struct netlink_ext_ack *extack)
3512 {
3513 	struct net_device *dev = NULL;
3514 	struct inet6_dev *idev = NULL;
3515 	int addr_type;
3516 	int err;
3517 
3518 	fib6_nh->fib_nh_family = AF_INET6;
3519 #ifdef CONFIG_IPV6_ROUTER_PREF
3520 	fib6_nh->last_probe = jiffies;
3521 #endif
3522 	if (cfg->fc_is_fdb) {
3523 		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3524 		fib6_nh->fib_nh_gw_family = AF_INET6;
3525 		return 0;
3526 	}
3527 
3528 	err = -ENODEV;
3529 	if (cfg->fc_ifindex) {
3530 		dev = dev_get_by_index(net, cfg->fc_ifindex);
3531 		if (!dev)
3532 			goto out;
3533 		idev = in6_dev_get(dev);
3534 		if (!idev)
3535 			goto out;
3536 	}
3537 
3538 	if (cfg->fc_flags & RTNH_F_ONLINK) {
3539 		if (!dev) {
3540 			NL_SET_ERR_MSG(extack,
3541 				       "Nexthop device required for onlink");
3542 			goto out;
3543 		}
3544 
3545 		if (!(dev->flags & IFF_UP)) {
3546 			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3547 			err = -ENETDOWN;
3548 			goto out;
3549 		}
3550 
3551 		fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3552 	}
3553 
3554 	fib6_nh->fib_nh_weight = 1;
3555 
3556 	/* We cannot add true routes via loopback here,
3557 	 * they would result in kernel looping; promote them to reject routes
3558 	 */
3559 	addr_type = ipv6_addr_type(&cfg->fc_dst);
3560 	if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3561 		/* hold loopback dev/idev if we haven't done so. */
3562 		if (dev != net->loopback_dev) {
3563 			if (dev) {
3564 				dev_put(dev);
3565 				in6_dev_put(idev);
3566 			}
3567 			dev = net->loopback_dev;
3568 			dev_hold(dev);
3569 			idev = in6_dev_get(dev);
3570 			if (!idev) {
3571 				err = -ENODEV;
3572 				goto out;
3573 			}
3574 		}
3575 		goto pcpu_alloc;
3576 	}
3577 
3578 	if (cfg->fc_flags & RTF_GATEWAY) {
3579 		err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3580 		if (err)
3581 			goto out;
3582 
3583 		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3584 		fib6_nh->fib_nh_gw_family = AF_INET6;
3585 	}
3586 
3587 	err = -ENODEV;
3588 	if (!dev)
3589 		goto out;
3590 
3591 	if (idev->cnf.disable_ipv6) {
3592 		NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3593 		err = -EACCES;
3594 		goto out;
3595 	}
3596 
3597 	if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3598 		NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3599 		err = -ENETDOWN;
3600 		goto out;
3601 	}
3602 
3603 	if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3604 	    !netif_carrier_ok(dev))
3605 		fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3606 
3607 	err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3608 				 cfg->fc_encap_type, cfg, gfp_flags, extack);
3609 	if (err)
3610 		goto out;
3611 
3612 pcpu_alloc:
3613 	fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3614 	if (!fib6_nh->rt6i_pcpu) {
3615 		err = -ENOMEM;
3616 		goto out;
3617 	}
3618 
3619 	fib6_nh->fib_nh_dev = dev;
3620 	fib6_nh->fib_nh_oif = dev->ifindex;
3621 	err = 0;
3622 out:
3623 	if (idev)
3624 		in6_dev_put(idev);
3625 
3626 	if (err) {
3627 		lwtstate_put(fib6_nh->fib_nh_lws);
3628 		fib6_nh->fib_nh_lws = NULL;
3629 		if (dev)
3630 			dev_put(dev);
3631 	}
3632 
3633 	return err;
3634 }
3635 
3636 void fib6_nh_release(struct fib6_nh *fib6_nh)
3637 {
3638 	struct rt6_exception_bucket *bucket;
3639 
3640 	rcu_read_lock();
3641 
3642 	fib6_nh_flush_exceptions(fib6_nh, NULL);
3643 	bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3644 	if (bucket) {
3645 		rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3646 		kfree(bucket);
3647 	}
3648 
3649 	rcu_read_unlock();
3650 
3651 	if (fib6_nh->rt6i_pcpu) {
3652 		int cpu;
3653 
3654 		for_each_possible_cpu(cpu) {
3655 			struct rt6_info **ppcpu_rt;
3656 			struct rt6_info *pcpu_rt;
3657 
3658 			ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3659 			pcpu_rt = *ppcpu_rt;
3660 			if (pcpu_rt) {
3661 				dst_dev_put(&pcpu_rt->dst);
3662 				dst_release(&pcpu_rt->dst);
3663 				*ppcpu_rt = NULL;
3664 			}
3665 		}
3666 
3667 		free_percpu(fib6_nh->rt6i_pcpu);
3668 	}
3669 
3670 	fib_nh_common_release(&fib6_nh->nh_common);
3671 }
3672 
3673 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3674 					      gfp_t gfp_flags,
3675 					      struct netlink_ext_ack *extack)
3676 {
3677 	struct net *net = cfg->fc_nlinfo.nl_net;
3678 	struct fib6_info *rt = NULL;
3679 	struct nexthop *nh = NULL;
3680 	struct fib6_table *table;
3681 	struct fib6_nh *fib6_nh;
3682 	int err = -EINVAL;
3683 	int addr_type;
3684 
3685 	/* RTF_PCPU is an internal flag; can not be set by userspace */
3686 	if (cfg->fc_flags & RTF_PCPU) {
3687 		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3688 		goto out;
3689 	}
3690 
3691 	/* RTF_CACHE is an internal flag; can not be set by userspace */
3692 	if (cfg->fc_flags & RTF_CACHE) {
3693 		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3694 		goto out;
3695 	}
3696 
3697 	if (cfg->fc_type > RTN_MAX) {
3698 		NL_SET_ERR_MSG(extack, "Invalid route type");
3699 		goto out;
3700 	}
3701 
3702 	if (cfg->fc_dst_len > 128) {
3703 		NL_SET_ERR_MSG(extack, "Invalid prefix length");
3704 		goto out;
3705 	}
3706 	if (cfg->fc_src_len > 128) {
3707 		NL_SET_ERR_MSG(extack, "Invalid source address length");
3708 		goto out;
3709 	}
3710 #ifndef CONFIG_IPV6_SUBTREES
3711 	if (cfg->fc_src_len) {
3712 		NL_SET_ERR_MSG(extack,
3713 			       "Specifying source address requires IPV6_SUBTREES to be enabled");
3714 		goto out;
3715 	}
3716 #endif
3717 	if (cfg->fc_nh_id) {
3718 		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3719 		if (!nh) {
3720 			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3721 			goto out;
3722 		}
3723 		err = fib6_check_nexthop(nh, cfg, extack);
3724 		if (err)
3725 			goto out;
3726 	}
3727 
3728 	err = -ENOBUFS;
3729 	if (cfg->fc_nlinfo.nlh &&
3730 	    !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3731 		table = fib6_get_table(net, cfg->fc_table);
3732 		if (!table) {
3733 			pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3734 			table = fib6_new_table(net, cfg->fc_table);
3735 		}
3736 	} else {
3737 		table = fib6_new_table(net, cfg->fc_table);
3738 	}
3739 
3740 	if (!table)
3741 		goto out;
3742 
3743 	err = -ENOMEM;
3744 	rt = fib6_info_alloc(gfp_flags, !nh);
3745 	if (!rt)
3746 		goto out;
3747 
3748 	rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3749 					       extack);
3750 	if (IS_ERR(rt->fib6_metrics)) {
3751 		err = PTR_ERR(rt->fib6_metrics);
3752 		/* Do not leave garbage there. */
3753 		rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3754 		goto out_free;
3755 	}
3756 
3757 	if (cfg->fc_flags & RTF_ADDRCONF)
3758 		rt->dst_nocount = true;
3759 
3760 	if (cfg->fc_flags & RTF_EXPIRES)
3761 		fib6_set_expires(rt, jiffies +
3762 				clock_t_to_jiffies(cfg->fc_expires));
3763 	else
3764 		fib6_clean_expires(rt);
3765 
3766 	if (cfg->fc_protocol == RTPROT_UNSPEC)
3767 		cfg->fc_protocol = RTPROT_BOOT;
3768 	rt->fib6_protocol = cfg->fc_protocol;
3769 
3770 	rt->fib6_table = table;
3771 	rt->fib6_metric = cfg->fc_metric;
3772 	rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3773 	rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3774 
3775 	ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3776 	rt->fib6_dst.plen = cfg->fc_dst_len;
3777 
3778 #ifdef CONFIG_IPV6_SUBTREES
3779 	ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3780 	rt->fib6_src.plen = cfg->fc_src_len;
3781 #endif
3782 	if (nh) {
3783 		if (rt->fib6_src.plen) {
3784 			NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3785 			goto out_free;
3786 		}
3787 		if (!nexthop_get(nh)) {
3788 			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3789 			goto out_free;
3790 		}
3791 		rt->nh = nh;
3792 		fib6_nh = nexthop_fib6_nh(rt->nh);
3793 	} else {
3794 		err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3795 		if (err)
3796 			goto out;
3797 
3798 		fib6_nh = rt->fib6_nh;
3799 
3800 		/* We cannot add true routes via loopback here, they would
3801 		 * result in kernel looping; promote them to reject routes
3802 		 */
3803 		addr_type = ipv6_addr_type(&cfg->fc_dst);
3804 		if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3805 				   addr_type))
3806 			rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3807 	}
3808 
3809 	if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3810 		struct net_device *dev = fib6_nh->fib_nh_dev;
3811 
3812 		if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3813 			NL_SET_ERR_MSG(extack, "Invalid source address");
3814 			err = -EINVAL;
3815 			goto out;
3816 		}
3817 		rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3818 		rt->fib6_prefsrc.plen = 128;
3819 	} else
3820 		rt->fib6_prefsrc.plen = 0;
3821 
3822 	return rt;
3823 out:
3824 	fib6_info_release(rt);
3825 	return ERR_PTR(err);
3826 out_free:
3827 	ip_fib_metrics_put(rt->fib6_metrics);
3828 	kfree(rt);
3829 	return ERR_PTR(err);
3830 }
3831 
3832 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3833 		  struct netlink_ext_ack *extack)
3834 {
3835 	struct fib6_info *rt;
3836 	int err;
3837 
3838 	rt = ip6_route_info_create(cfg, gfp_flags, extack);
3839 	if (IS_ERR(rt))
3840 		return PTR_ERR(rt);
3841 
3842 	err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3843 	fib6_info_release(rt);
3844 
3845 	return err;
3846 }
3847 
3848 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3849 {
3850 	struct net *net = info->nl_net;
3851 	struct fib6_table *table;
3852 	int err;
3853 
3854 	if (rt == net->ipv6.fib6_null_entry) {
3855 		err = -ENOENT;
3856 		goto out;
3857 	}
3858 
3859 	table = rt->fib6_table;
3860 	spin_lock_bh(&table->tb6_lock);
3861 	err = fib6_del(rt, info);
3862 	spin_unlock_bh(&table->tb6_lock);
3863 
3864 out:
3865 	fib6_info_release(rt);
3866 	return err;
3867 }
3868 
3869 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3870 {
3871 	struct nl_info info = {
3872 		.nl_net = net,
3873 		.skip_notify = skip_notify
3874 	};
3875 
3876 	return __ip6_del_rt(rt, &info);
3877 }
3878 
3879 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3880 {
3881 	struct nl_info *info = &cfg->fc_nlinfo;
3882 	struct net *net = info->nl_net;
3883 	struct sk_buff *skb = NULL;
3884 	struct fib6_table *table;
3885 	int err = -ENOENT;
3886 
3887 	if (rt == net->ipv6.fib6_null_entry)
3888 		goto out_put;
3889 	table = rt->fib6_table;
3890 	spin_lock_bh(&table->tb6_lock);
3891 
3892 	if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3893 		struct fib6_info *sibling, *next_sibling;
3894 		struct fib6_node *fn;
3895 
3896 		/* prefer to send a single notification with all hops */
3897 		skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3898 		if (skb) {
3899 			u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3900 
3901 			if (rt6_fill_node(net, skb, rt, NULL,
3902 					  NULL, NULL, 0, RTM_DELROUTE,
3903 					  info->portid, seq, 0) < 0) {
3904 				kfree_skb(skb);
3905 				skb = NULL;
3906 			} else
3907 				info->skip_notify = 1;
3908 		}
3909 
3910 		/* 'rt' points to the first sibling route. If it is not the
3911 		 * leaf, then we do not need to send a notification. Otherwise,
3912 		 * we need to check if the last sibling has a next route or not
3913 		 * and emit a replace or delete notification, respectively.
3914 		 */
3915 		info->skip_notify_kernel = 1;
3916 		fn = rcu_dereference_protected(rt->fib6_node,
3917 					    lockdep_is_held(&table->tb6_lock));
3918 		if (rcu_access_pointer(fn->leaf) == rt) {
3919 			struct fib6_info *last_sibling, *replace_rt;
3920 
3921 			last_sibling = list_last_entry(&rt->fib6_siblings,
3922 						       struct fib6_info,
3923 						       fib6_siblings);
3924 			replace_rt = rcu_dereference_protected(
3925 					    last_sibling->fib6_next,
3926 					    lockdep_is_held(&table->tb6_lock));
3927 			if (replace_rt)
3928 				call_fib6_entry_notifiers_replace(net,
3929 								  replace_rt);
3930 			else
3931 				call_fib6_multipath_entry_notifiers(net,
3932 						       FIB_EVENT_ENTRY_DEL,
3933 						       rt, rt->fib6_nsiblings,
3934 						       NULL);
3935 		}
3936 		list_for_each_entry_safe(sibling, next_sibling,
3937 					 &rt->fib6_siblings,
3938 					 fib6_siblings) {
3939 			err = fib6_del(sibling, info);
3940 			if (err)
3941 				goto out_unlock;
3942 		}
3943 	}
3944 
3945 	err = fib6_del(rt, info);
3946 out_unlock:
3947 	spin_unlock_bh(&table->tb6_lock);
3948 out_put:
3949 	fib6_info_release(rt);
3950 
3951 	if (skb) {
3952 		rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3953 			    info->nlh, gfp_any());
3954 	}
3955 	return err;
3956 }
3957 
3958 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3959 {
3960 	int rc = -ESRCH;
3961 
3962 	if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3963 		goto out;
3964 
3965 	if (cfg->fc_flags & RTF_GATEWAY &&
3966 	    !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3967 		goto out;
3968 
3969 	rc = rt6_remove_exception_rt(rt);
3970 out:
3971 	return rc;
3972 }
3973 
3974 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3975 			     struct fib6_nh *nh)
3976 {
3977 	struct fib6_result res = {
3978 		.f6i = rt,
3979 		.nh = nh,
3980 	};
3981 	struct rt6_info *rt_cache;
3982 
3983 	rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3984 	if (rt_cache)
3985 		return __ip6_del_cached_rt(rt_cache, cfg);
3986 
3987 	return 0;
3988 }
3989 
3990 struct fib6_nh_del_cached_rt_arg {
3991 	struct fib6_config *cfg;
3992 	struct fib6_info *f6i;
3993 };
3994 
3995 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3996 {
3997 	struct fib6_nh_del_cached_rt_arg *arg = _arg;
3998 	int rc;
3999 
4000 	rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4001 	return rc != -ESRCH ? rc : 0;
4002 }
4003 
4004 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4005 {
4006 	struct fib6_nh_del_cached_rt_arg arg = {
4007 		.cfg = cfg,
4008 		.f6i = f6i
4009 	};
4010 
4011 	return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4012 }
4013 
4014 static int ip6_route_del(struct fib6_config *cfg,
4015 			 struct netlink_ext_ack *extack)
4016 {
4017 	struct fib6_table *table;
4018 	struct fib6_info *rt;
4019 	struct fib6_node *fn;
4020 	int err = -ESRCH;
4021 
4022 	table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4023 	if (!table) {
4024 		NL_SET_ERR_MSG(extack, "FIB table does not exist");
4025 		return err;
4026 	}
4027 
4028 	rcu_read_lock();
4029 
4030 	fn = fib6_locate(&table->tb6_root,
4031 			 &cfg->fc_dst, cfg->fc_dst_len,
4032 			 &cfg->fc_src, cfg->fc_src_len,
4033 			 !(cfg->fc_flags & RTF_CACHE));
4034 
4035 	if (fn) {
4036 		for_each_fib6_node_rt_rcu(fn) {
4037 			struct fib6_nh *nh;
4038 
4039 			if (rt->nh && cfg->fc_nh_id &&
4040 			    rt->nh->id != cfg->fc_nh_id)
4041 				continue;
4042 
4043 			if (cfg->fc_flags & RTF_CACHE) {
4044 				int rc = 0;
4045 
4046 				if (rt->nh) {
4047 					rc = ip6_del_cached_rt_nh(cfg, rt);
4048 				} else if (cfg->fc_nh_id) {
4049 					continue;
4050 				} else {
4051 					nh = rt->fib6_nh;
4052 					rc = ip6_del_cached_rt(cfg, rt, nh);
4053 				}
4054 				if (rc != -ESRCH) {
4055 					rcu_read_unlock();
4056 					return rc;
4057 				}
4058 				continue;
4059 			}
4060 
4061 			if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4062 				continue;
4063 			if (cfg->fc_protocol &&
4064 			    cfg->fc_protocol != rt->fib6_protocol)
4065 				continue;
4066 
4067 			if (rt->nh) {
4068 				if (!fib6_info_hold_safe(rt))
4069 					continue;
4070 				rcu_read_unlock();
4071 
4072 				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4073 			}
4074 			if (cfg->fc_nh_id)
4075 				continue;
4076 
4077 			nh = rt->fib6_nh;
4078 			if (cfg->fc_ifindex &&
4079 			    (!nh->fib_nh_dev ||
4080 			     nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4081 				continue;
4082 			if (cfg->fc_flags & RTF_GATEWAY &&
4083 			    !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4084 				continue;
4085 			if (!fib6_info_hold_safe(rt))
4086 				continue;
4087 			rcu_read_unlock();
4088 
4089 			/* if gateway was specified only delete the one hop */
4090 			if (cfg->fc_flags & RTF_GATEWAY)
4091 				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4092 
4093 			return __ip6_del_rt_siblings(rt, cfg);
4094 		}
4095 	}
4096 	rcu_read_unlock();
4097 
4098 	return err;
4099 }
4100 
4101 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4102 {
4103 	struct netevent_redirect netevent;
4104 	struct rt6_info *rt, *nrt = NULL;
4105 	struct fib6_result res = {};
4106 	struct ndisc_options ndopts;
4107 	struct inet6_dev *in6_dev;
4108 	struct neighbour *neigh;
4109 	struct rd_msg *msg;
4110 	int optlen, on_link;
4111 	u8 *lladdr;
4112 
4113 	optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4114 	optlen -= sizeof(*msg);
4115 
4116 	if (optlen < 0) {
4117 		net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4118 		return;
4119 	}
4120 
4121 	msg = (struct rd_msg *)icmp6_hdr(skb);
4122 
4123 	if (ipv6_addr_is_multicast(&msg->dest)) {
4124 		net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4125 		return;
4126 	}
4127 
4128 	on_link = 0;
4129 	if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4130 		on_link = 1;
4131 	} else if (ipv6_addr_type(&msg->target) !=
4132 		   (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4133 		net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4134 		return;
4135 	}
4136 
4137 	in6_dev = __in6_dev_get(skb->dev);
4138 	if (!in6_dev)
4139 		return;
4140 	if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4141 		return;
4142 
4143 	/* RFC2461 8.1:
4144 	 *	The IP source address of the Redirect MUST be the same as the current
4145 	 *	first-hop router for the specified ICMP Destination Address.
4146 	 */
4147 
4148 	if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4149 		net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4150 		return;
4151 	}
4152 
4153 	lladdr = NULL;
4154 	if (ndopts.nd_opts_tgt_lladdr) {
4155 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4156 					     skb->dev);
4157 		if (!lladdr) {
4158 			net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4159 			return;
4160 		}
4161 	}
4162 
4163 	rt = (struct rt6_info *) dst;
4164 	if (rt->rt6i_flags & RTF_REJECT) {
4165 		net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4166 		return;
4167 	}
4168 
4169 	/* Redirect received -> path was valid.
4170 	 * Look, redirects are sent only in response to data packets,
4171 	 * so that this nexthop apparently is reachable. --ANK
4172 	 */
4173 	dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4174 
4175 	neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4176 	if (!neigh)
4177 		return;
4178 
4179 	/*
4180 	 *	We have finally decided to accept it.
4181 	 */
4182 
4183 	ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4184 		     NEIGH_UPDATE_F_WEAK_OVERRIDE|
4185 		     NEIGH_UPDATE_F_OVERRIDE|
4186 		     (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4187 				     NEIGH_UPDATE_F_ISROUTER)),
4188 		     NDISC_REDIRECT, &ndopts);
4189 
4190 	rcu_read_lock();
4191 	res.f6i = rcu_dereference(rt->from);
4192 	if (!res.f6i)
4193 		goto out;
4194 
4195 	if (res.f6i->nh) {
4196 		struct fib6_nh_match_arg arg = {
4197 			.dev = dst->dev,
4198 			.gw = &rt->rt6i_gateway,
4199 		};
4200 
4201 		nexthop_for_each_fib6_nh(res.f6i->nh,
4202 					 fib6_nh_find_match, &arg);
4203 
4204 		/* fib6_info uses a nexthop that does not have fib6_nh
4205 		 * using the dst->dev. Should be impossible
4206 		 */
4207 		if (!arg.match)
4208 			goto out;
4209 		res.nh = arg.match;
4210 	} else {
4211 		res.nh = res.f6i->fib6_nh;
4212 	}
4213 
4214 	res.fib6_flags = res.f6i->fib6_flags;
4215 	res.fib6_type = res.f6i->fib6_type;
4216 	nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4217 	if (!nrt)
4218 		goto out;
4219 
4220 	nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4221 	if (on_link)
4222 		nrt->rt6i_flags &= ~RTF_GATEWAY;
4223 
4224 	nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4225 
4226 	/* rt6_insert_exception() will take care of duplicated exceptions */
4227 	if (rt6_insert_exception(nrt, &res)) {
4228 		dst_release_immediate(&nrt->dst);
4229 		goto out;
4230 	}
4231 
4232 	netevent.old = &rt->dst;
4233 	netevent.new = &nrt->dst;
4234 	netevent.daddr = &msg->dest;
4235 	netevent.neigh = neigh;
4236 	call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4237 
4238 out:
4239 	rcu_read_unlock();
4240 	neigh_release(neigh);
4241 }
4242 
4243 #ifdef CONFIG_IPV6_ROUTE_INFO
4244 static struct fib6_info *rt6_get_route_info(struct net *net,
4245 					   const struct in6_addr *prefix, int prefixlen,
4246 					   const struct in6_addr *gwaddr,
4247 					   struct net_device *dev)
4248 {
4249 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4250 	int ifindex = dev->ifindex;
4251 	struct fib6_node *fn;
4252 	struct fib6_info *rt = NULL;
4253 	struct fib6_table *table;
4254 
4255 	table = fib6_get_table(net, tb_id);
4256 	if (!table)
4257 		return NULL;
4258 
4259 	rcu_read_lock();
4260 	fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4261 	if (!fn)
4262 		goto out;
4263 
4264 	for_each_fib6_node_rt_rcu(fn) {
4265 		/* these routes do not use nexthops */
4266 		if (rt->nh)
4267 			continue;
4268 		if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4269 			continue;
4270 		if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4271 		    !rt->fib6_nh->fib_nh_gw_family)
4272 			continue;
4273 		if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4274 			continue;
4275 		if (!fib6_info_hold_safe(rt))
4276 			continue;
4277 		break;
4278 	}
4279 out:
4280 	rcu_read_unlock();
4281 	return rt;
4282 }
4283 
4284 static struct fib6_info *rt6_add_route_info(struct net *net,
4285 					   const struct in6_addr *prefix, int prefixlen,
4286 					   const struct in6_addr *gwaddr,
4287 					   struct net_device *dev,
4288 					   unsigned int pref)
4289 {
4290 	struct fib6_config cfg = {
4291 		.fc_metric	= IP6_RT_PRIO_USER,
4292 		.fc_ifindex	= dev->ifindex,
4293 		.fc_dst_len	= prefixlen,
4294 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4295 				  RTF_UP | RTF_PREF(pref),
4296 		.fc_protocol = RTPROT_RA,
4297 		.fc_type = RTN_UNICAST,
4298 		.fc_nlinfo.portid = 0,
4299 		.fc_nlinfo.nlh = NULL,
4300 		.fc_nlinfo.nl_net = net,
4301 	};
4302 
4303 	cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4304 	cfg.fc_dst = *prefix;
4305 	cfg.fc_gateway = *gwaddr;
4306 
4307 	/* We should treat it as a default route if prefix length is 0. */
4308 	if (!prefixlen)
4309 		cfg.fc_flags |= RTF_DEFAULT;
4310 
4311 	ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4312 
4313 	return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4314 }
4315 #endif
4316 
4317 struct fib6_info *rt6_get_dflt_router(struct net *net,
4318 				     const struct in6_addr *addr,
4319 				     struct net_device *dev)
4320 {
4321 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4322 	struct fib6_info *rt;
4323 	struct fib6_table *table;
4324 
4325 	table = fib6_get_table(net, tb_id);
4326 	if (!table)
4327 		return NULL;
4328 
4329 	rcu_read_lock();
4330 	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4331 		struct fib6_nh *nh;
4332 
4333 		/* RA routes do not use nexthops */
4334 		if (rt->nh)
4335 			continue;
4336 
4337 		nh = rt->fib6_nh;
4338 		if (dev == nh->fib_nh_dev &&
4339 		    ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4340 		    ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4341 			break;
4342 	}
4343 	if (rt && !fib6_info_hold_safe(rt))
4344 		rt = NULL;
4345 	rcu_read_unlock();
4346 	return rt;
4347 }
4348 
4349 struct fib6_info *rt6_add_dflt_router(struct net *net,
4350 				     const struct in6_addr *gwaddr,
4351 				     struct net_device *dev,
4352 				     unsigned int pref,
4353 				     u32 defrtr_usr_metric)
4354 {
4355 	struct fib6_config cfg = {
4356 		.fc_table	= l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4357 		.fc_metric	= defrtr_usr_metric,
4358 		.fc_ifindex	= dev->ifindex,
4359 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4360 				  RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4361 		.fc_protocol = RTPROT_RA,
4362 		.fc_type = RTN_UNICAST,
4363 		.fc_nlinfo.portid = 0,
4364 		.fc_nlinfo.nlh = NULL,
4365 		.fc_nlinfo.nl_net = net,
4366 	};
4367 
4368 	cfg.fc_gateway = *gwaddr;
4369 
4370 	if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4371 		struct fib6_table *table;
4372 
4373 		table = fib6_get_table(dev_net(dev), cfg.fc_table);
4374 		if (table)
4375 			table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4376 	}
4377 
4378 	return rt6_get_dflt_router(net, gwaddr, dev);
4379 }
4380 
4381 static void __rt6_purge_dflt_routers(struct net *net,
4382 				     struct fib6_table *table)
4383 {
4384 	struct fib6_info *rt;
4385 
4386 restart:
4387 	rcu_read_lock();
4388 	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4389 		struct net_device *dev = fib6_info_nh_dev(rt);
4390 		struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4391 
4392 		if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4393 		    (!idev || idev->cnf.accept_ra != 2) &&
4394 		    fib6_info_hold_safe(rt)) {
4395 			rcu_read_unlock();
4396 			ip6_del_rt(net, rt, false);
4397 			goto restart;
4398 		}
4399 	}
4400 	rcu_read_unlock();
4401 
4402 	table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4403 }
4404 
4405 void rt6_purge_dflt_routers(struct net *net)
4406 {
4407 	struct fib6_table *table;
4408 	struct hlist_head *head;
4409 	unsigned int h;
4410 
4411 	rcu_read_lock();
4412 
4413 	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4414 		head = &net->ipv6.fib_table_hash[h];
4415 		hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4416 			if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4417 				__rt6_purge_dflt_routers(net, table);
4418 		}
4419 	}
4420 
4421 	rcu_read_unlock();
4422 }
4423 
4424 static void rtmsg_to_fib6_config(struct net *net,
4425 				 struct in6_rtmsg *rtmsg,
4426 				 struct fib6_config *cfg)
4427 {
4428 	*cfg = (struct fib6_config){
4429 		.fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4430 			 : RT6_TABLE_MAIN,
4431 		.fc_ifindex = rtmsg->rtmsg_ifindex,
4432 		.fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4433 		.fc_expires = rtmsg->rtmsg_info,
4434 		.fc_dst_len = rtmsg->rtmsg_dst_len,
4435 		.fc_src_len = rtmsg->rtmsg_src_len,
4436 		.fc_flags = rtmsg->rtmsg_flags,
4437 		.fc_type = rtmsg->rtmsg_type,
4438 
4439 		.fc_nlinfo.nl_net = net,
4440 
4441 		.fc_dst = rtmsg->rtmsg_dst,
4442 		.fc_src = rtmsg->rtmsg_src,
4443 		.fc_gateway = rtmsg->rtmsg_gateway,
4444 	};
4445 }
4446 
4447 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4448 {
4449 	struct fib6_config cfg;
4450 	int err;
4451 
4452 	if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4453 		return -EINVAL;
4454 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4455 		return -EPERM;
4456 
4457 	rtmsg_to_fib6_config(net, rtmsg, &cfg);
4458 
4459 	rtnl_lock();
4460 	switch (cmd) {
4461 	case SIOCADDRT:
4462 		err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4463 		break;
4464 	case SIOCDELRT:
4465 		err = ip6_route_del(&cfg, NULL);
4466 		break;
4467 	}
4468 	rtnl_unlock();
4469 	return err;
4470 }
4471 
4472 /*
4473  *	Drop the packet on the floor
4474  */
4475 
4476 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4477 {
4478 	struct dst_entry *dst = skb_dst(skb);
4479 	struct net *net = dev_net(dst->dev);
4480 	struct inet6_dev *idev;
4481 	int type;
4482 
4483 	if (netif_is_l3_master(skb->dev) &&
4484 	    dst->dev == net->loopback_dev)
4485 		idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4486 	else
4487 		idev = ip6_dst_idev(dst);
4488 
4489 	switch (ipstats_mib_noroutes) {
4490 	case IPSTATS_MIB_INNOROUTES:
4491 		type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4492 		if (type == IPV6_ADDR_ANY) {
4493 			IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4494 			break;
4495 		}
4496 		fallthrough;
4497 	case IPSTATS_MIB_OUTNOROUTES:
4498 		IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4499 		break;
4500 	}
4501 
4502 	/* Start over by dropping the dst for l3mdev case */
4503 	if (netif_is_l3_master(skb->dev))
4504 		skb_dst_drop(skb);
4505 
4506 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4507 	kfree_skb(skb);
4508 	return 0;
4509 }
4510 
4511 static int ip6_pkt_discard(struct sk_buff *skb)
4512 {
4513 	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4514 }
4515 
4516 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4517 {
4518 	skb->dev = skb_dst(skb)->dev;
4519 	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4520 }
4521 
4522 static int ip6_pkt_prohibit(struct sk_buff *skb)
4523 {
4524 	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4525 }
4526 
4527 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4528 {
4529 	skb->dev = skb_dst(skb)->dev;
4530 	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4531 }
4532 
4533 /*
4534  *	Allocate a dst for local (unicast / anycast) address.
4535  */
4536 
4537 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4538 				     struct inet6_dev *idev,
4539 				     const struct in6_addr *addr,
4540 				     bool anycast, gfp_t gfp_flags)
4541 {
4542 	struct fib6_config cfg = {
4543 		.fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4544 		.fc_ifindex = idev->dev->ifindex,
4545 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
4546 		.fc_dst = *addr,
4547 		.fc_dst_len = 128,
4548 		.fc_protocol = RTPROT_KERNEL,
4549 		.fc_nlinfo.nl_net = net,
4550 		.fc_ignore_dev_down = true,
4551 	};
4552 	struct fib6_info *f6i;
4553 
4554 	if (anycast) {
4555 		cfg.fc_type = RTN_ANYCAST;
4556 		cfg.fc_flags |= RTF_ANYCAST;
4557 	} else {
4558 		cfg.fc_type = RTN_LOCAL;
4559 		cfg.fc_flags |= RTF_LOCAL;
4560 	}
4561 
4562 	f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4563 	if (!IS_ERR(f6i))
4564 		f6i->dst_nocount = true;
4565 	return f6i;
4566 }
4567 
4568 /* remove deleted ip from prefsrc entries */
4569 struct arg_dev_net_ip {
4570 	struct net_device *dev;
4571 	struct net *net;
4572 	struct in6_addr *addr;
4573 };
4574 
4575 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4576 {
4577 	struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4578 	struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4579 	struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4580 
4581 	if (!rt->nh &&
4582 	    ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4583 	    rt != net->ipv6.fib6_null_entry &&
4584 	    ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4585 		spin_lock_bh(&rt6_exception_lock);
4586 		/* remove prefsrc entry */
4587 		rt->fib6_prefsrc.plen = 0;
4588 		spin_unlock_bh(&rt6_exception_lock);
4589 	}
4590 	return 0;
4591 }
4592 
4593 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4594 {
4595 	struct net *net = dev_net(ifp->idev->dev);
4596 	struct arg_dev_net_ip adni = {
4597 		.dev = ifp->idev->dev,
4598 		.net = net,
4599 		.addr = &ifp->addr,
4600 	};
4601 	fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4602 }
4603 
4604 #define RTF_RA_ROUTER		(RTF_ADDRCONF | RTF_DEFAULT)
4605 
4606 /* Remove routers and update dst entries when gateway turn into host. */
4607 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4608 {
4609 	struct in6_addr *gateway = (struct in6_addr *)arg;
4610 	struct fib6_nh *nh;
4611 
4612 	/* RA routes do not use nexthops */
4613 	if (rt->nh)
4614 		return 0;
4615 
4616 	nh = rt->fib6_nh;
4617 	if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4618 	    nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4619 		return -1;
4620 
4621 	/* Further clean up cached routes in exception table.
4622 	 * This is needed because cached route may have a different
4623 	 * gateway than its 'parent' in the case of an ip redirect.
4624 	 */
4625 	fib6_nh_exceptions_clean_tohost(nh, gateway);
4626 
4627 	return 0;
4628 }
4629 
4630 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4631 {
4632 	fib6_clean_all(net, fib6_clean_tohost, gateway);
4633 }
4634 
4635 struct arg_netdev_event {
4636 	const struct net_device *dev;
4637 	union {
4638 		unsigned char nh_flags;
4639 		unsigned long event;
4640 	};
4641 };
4642 
4643 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4644 {
4645 	struct fib6_info *iter;
4646 	struct fib6_node *fn;
4647 
4648 	fn = rcu_dereference_protected(rt->fib6_node,
4649 			lockdep_is_held(&rt->fib6_table->tb6_lock));
4650 	iter = rcu_dereference_protected(fn->leaf,
4651 			lockdep_is_held(&rt->fib6_table->tb6_lock));
4652 	while (iter) {
4653 		if (iter->fib6_metric == rt->fib6_metric &&
4654 		    rt6_qualify_for_ecmp(iter))
4655 			return iter;
4656 		iter = rcu_dereference_protected(iter->fib6_next,
4657 				lockdep_is_held(&rt->fib6_table->tb6_lock));
4658 	}
4659 
4660 	return NULL;
4661 }
4662 
4663 /* only called for fib entries with builtin fib6_nh */
4664 static bool rt6_is_dead(const struct fib6_info *rt)
4665 {
4666 	if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4667 	    (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4668 	     ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4669 		return true;
4670 
4671 	return false;
4672 }
4673 
4674 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4675 {
4676 	struct fib6_info *iter;
4677 	int total = 0;
4678 
4679 	if (!rt6_is_dead(rt))
4680 		total += rt->fib6_nh->fib_nh_weight;
4681 
4682 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4683 		if (!rt6_is_dead(iter))
4684 			total += iter->fib6_nh->fib_nh_weight;
4685 	}
4686 
4687 	return total;
4688 }
4689 
4690 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4691 {
4692 	int upper_bound = -1;
4693 
4694 	if (!rt6_is_dead(rt)) {
4695 		*weight += rt->fib6_nh->fib_nh_weight;
4696 		upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4697 						    total) - 1;
4698 	}
4699 	atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4700 }
4701 
4702 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4703 {
4704 	struct fib6_info *iter;
4705 	int weight = 0;
4706 
4707 	rt6_upper_bound_set(rt, &weight, total);
4708 
4709 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4710 		rt6_upper_bound_set(iter, &weight, total);
4711 }
4712 
4713 void rt6_multipath_rebalance(struct fib6_info *rt)
4714 {
4715 	struct fib6_info *first;
4716 	int total;
4717 
4718 	/* In case the entire multipath route was marked for flushing,
4719 	 * then there is no need to rebalance upon the removal of every
4720 	 * sibling route.
4721 	 */
4722 	if (!rt->fib6_nsiblings || rt->should_flush)
4723 		return;
4724 
4725 	/* During lookup routes are evaluated in order, so we need to
4726 	 * make sure upper bounds are assigned from the first sibling
4727 	 * onwards.
4728 	 */
4729 	first = rt6_multipath_first_sibling(rt);
4730 	if (WARN_ON_ONCE(!first))
4731 		return;
4732 
4733 	total = rt6_multipath_total_weight(first);
4734 	rt6_multipath_upper_bound_set(first, total);
4735 }
4736 
4737 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4738 {
4739 	const struct arg_netdev_event *arg = p_arg;
4740 	struct net *net = dev_net(arg->dev);
4741 
4742 	if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4743 	    rt->fib6_nh->fib_nh_dev == arg->dev) {
4744 		rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4745 		fib6_update_sernum_upto_root(net, rt);
4746 		rt6_multipath_rebalance(rt);
4747 	}
4748 
4749 	return 0;
4750 }
4751 
4752 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4753 {
4754 	struct arg_netdev_event arg = {
4755 		.dev = dev,
4756 		{
4757 			.nh_flags = nh_flags,
4758 		},
4759 	};
4760 
4761 	if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4762 		arg.nh_flags |= RTNH_F_LINKDOWN;
4763 
4764 	fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4765 }
4766 
4767 /* only called for fib entries with inline fib6_nh */
4768 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4769 				   const struct net_device *dev)
4770 {
4771 	struct fib6_info *iter;
4772 
4773 	if (rt->fib6_nh->fib_nh_dev == dev)
4774 		return true;
4775 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4776 		if (iter->fib6_nh->fib_nh_dev == dev)
4777 			return true;
4778 
4779 	return false;
4780 }
4781 
4782 static void rt6_multipath_flush(struct fib6_info *rt)
4783 {
4784 	struct fib6_info *iter;
4785 
4786 	rt->should_flush = 1;
4787 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4788 		iter->should_flush = 1;
4789 }
4790 
4791 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4792 					     const struct net_device *down_dev)
4793 {
4794 	struct fib6_info *iter;
4795 	unsigned int dead = 0;
4796 
4797 	if (rt->fib6_nh->fib_nh_dev == down_dev ||
4798 	    rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4799 		dead++;
4800 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4801 		if (iter->fib6_nh->fib_nh_dev == down_dev ||
4802 		    iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4803 			dead++;
4804 
4805 	return dead;
4806 }
4807 
4808 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4809 				       const struct net_device *dev,
4810 				       unsigned char nh_flags)
4811 {
4812 	struct fib6_info *iter;
4813 
4814 	if (rt->fib6_nh->fib_nh_dev == dev)
4815 		rt->fib6_nh->fib_nh_flags |= nh_flags;
4816 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4817 		if (iter->fib6_nh->fib_nh_dev == dev)
4818 			iter->fib6_nh->fib_nh_flags |= nh_flags;
4819 }
4820 
4821 /* called with write lock held for table with rt */
4822 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4823 {
4824 	const struct arg_netdev_event *arg = p_arg;
4825 	const struct net_device *dev = arg->dev;
4826 	struct net *net = dev_net(dev);
4827 
4828 	if (rt == net->ipv6.fib6_null_entry || rt->nh)
4829 		return 0;
4830 
4831 	switch (arg->event) {
4832 	case NETDEV_UNREGISTER:
4833 		return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4834 	case NETDEV_DOWN:
4835 		if (rt->should_flush)
4836 			return -1;
4837 		if (!rt->fib6_nsiblings)
4838 			return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4839 		if (rt6_multipath_uses_dev(rt, dev)) {
4840 			unsigned int count;
4841 
4842 			count = rt6_multipath_dead_count(rt, dev);
4843 			if (rt->fib6_nsiblings + 1 == count) {
4844 				rt6_multipath_flush(rt);
4845 				return -1;
4846 			}
4847 			rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4848 						   RTNH_F_LINKDOWN);
4849 			fib6_update_sernum(net, rt);
4850 			rt6_multipath_rebalance(rt);
4851 		}
4852 		return -2;
4853 	case NETDEV_CHANGE:
4854 		if (rt->fib6_nh->fib_nh_dev != dev ||
4855 		    rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4856 			break;
4857 		rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4858 		rt6_multipath_rebalance(rt);
4859 		break;
4860 	}
4861 
4862 	return 0;
4863 }
4864 
4865 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4866 {
4867 	struct arg_netdev_event arg = {
4868 		.dev = dev,
4869 		{
4870 			.event = event,
4871 		},
4872 	};
4873 	struct net *net = dev_net(dev);
4874 
4875 	if (net->ipv6.sysctl.skip_notify_on_dev_down)
4876 		fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4877 	else
4878 		fib6_clean_all(net, fib6_ifdown, &arg);
4879 }
4880 
4881 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4882 {
4883 	rt6_sync_down_dev(dev, event);
4884 	rt6_uncached_list_flush_dev(dev_net(dev), dev);
4885 	neigh_ifdown(&nd_tbl, dev);
4886 }
4887 
4888 struct rt6_mtu_change_arg {
4889 	struct net_device *dev;
4890 	unsigned int mtu;
4891 	struct fib6_info *f6i;
4892 };
4893 
4894 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4895 {
4896 	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4897 	struct fib6_info *f6i = arg->f6i;
4898 
4899 	/* For administrative MTU increase, there is no way to discover
4900 	 * IPv6 PMTU increase, so PMTU increase should be updated here.
4901 	 * Since RFC 1981 doesn't include administrative MTU increase
4902 	 * update PMTU increase is a MUST. (i.e. jumbo frame)
4903 	 */
4904 	if (nh->fib_nh_dev == arg->dev) {
4905 		struct inet6_dev *idev = __in6_dev_get(arg->dev);
4906 		u32 mtu = f6i->fib6_pmtu;
4907 
4908 		if (mtu >= arg->mtu ||
4909 		    (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4910 			fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4911 
4912 		spin_lock_bh(&rt6_exception_lock);
4913 		rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4914 		spin_unlock_bh(&rt6_exception_lock);
4915 	}
4916 
4917 	return 0;
4918 }
4919 
4920 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4921 {
4922 	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4923 	struct inet6_dev *idev;
4924 
4925 	/* In IPv6 pmtu discovery is not optional,
4926 	   so that RTAX_MTU lock cannot disable it.
4927 	   We still use this lock to block changes
4928 	   caused by addrconf/ndisc.
4929 	*/
4930 
4931 	idev = __in6_dev_get(arg->dev);
4932 	if (!idev)
4933 		return 0;
4934 
4935 	if (fib6_metric_locked(f6i, RTAX_MTU))
4936 		return 0;
4937 
4938 	arg->f6i = f6i;
4939 	if (f6i->nh) {
4940 		/* fib6_nh_mtu_change only returns 0, so this is safe */
4941 		return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4942 						arg);
4943 	}
4944 
4945 	return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4946 }
4947 
4948 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4949 {
4950 	struct rt6_mtu_change_arg arg = {
4951 		.dev = dev,
4952 		.mtu = mtu,
4953 	};
4954 
4955 	fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4956 }
4957 
4958 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4959 	[RTA_UNSPEC]		= { .strict_start_type = RTA_DPORT + 1 },
4960 	[RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
4961 	[RTA_PREFSRC]		= { .len = sizeof(struct in6_addr) },
4962 	[RTA_OIF]               = { .type = NLA_U32 },
4963 	[RTA_IIF]		= { .type = NLA_U32 },
4964 	[RTA_PRIORITY]          = { .type = NLA_U32 },
4965 	[RTA_METRICS]           = { .type = NLA_NESTED },
4966 	[RTA_MULTIPATH]		= { .len = sizeof(struct rtnexthop) },
4967 	[RTA_PREF]              = { .type = NLA_U8 },
4968 	[RTA_ENCAP_TYPE]	= { .type = NLA_U16 },
4969 	[RTA_ENCAP]		= { .type = NLA_NESTED },
4970 	[RTA_EXPIRES]		= { .type = NLA_U32 },
4971 	[RTA_UID]		= { .type = NLA_U32 },
4972 	[RTA_MARK]		= { .type = NLA_U32 },
4973 	[RTA_TABLE]		= { .type = NLA_U32 },
4974 	[RTA_IP_PROTO]		= { .type = NLA_U8 },
4975 	[RTA_SPORT]		= { .type = NLA_U16 },
4976 	[RTA_DPORT]		= { .type = NLA_U16 },
4977 	[RTA_NH_ID]		= { .type = NLA_U32 },
4978 };
4979 
4980 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4981 			      struct fib6_config *cfg,
4982 			      struct netlink_ext_ack *extack)
4983 {
4984 	struct rtmsg *rtm;
4985 	struct nlattr *tb[RTA_MAX+1];
4986 	unsigned int pref;
4987 	int err;
4988 
4989 	err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4990 				     rtm_ipv6_policy, extack);
4991 	if (err < 0)
4992 		goto errout;
4993 
4994 	err = -EINVAL;
4995 	rtm = nlmsg_data(nlh);
4996 
4997 	*cfg = (struct fib6_config){
4998 		.fc_table = rtm->rtm_table,
4999 		.fc_dst_len = rtm->rtm_dst_len,
5000 		.fc_src_len = rtm->rtm_src_len,
5001 		.fc_flags = RTF_UP,
5002 		.fc_protocol = rtm->rtm_protocol,
5003 		.fc_type = rtm->rtm_type,
5004 
5005 		.fc_nlinfo.portid = NETLINK_CB(skb).portid,
5006 		.fc_nlinfo.nlh = nlh,
5007 		.fc_nlinfo.nl_net = sock_net(skb->sk),
5008 	};
5009 
5010 	if (rtm->rtm_type == RTN_UNREACHABLE ||
5011 	    rtm->rtm_type == RTN_BLACKHOLE ||
5012 	    rtm->rtm_type == RTN_PROHIBIT ||
5013 	    rtm->rtm_type == RTN_THROW)
5014 		cfg->fc_flags |= RTF_REJECT;
5015 
5016 	if (rtm->rtm_type == RTN_LOCAL)
5017 		cfg->fc_flags |= RTF_LOCAL;
5018 
5019 	if (rtm->rtm_flags & RTM_F_CLONED)
5020 		cfg->fc_flags |= RTF_CACHE;
5021 
5022 	cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5023 
5024 	if (tb[RTA_NH_ID]) {
5025 		if (tb[RTA_GATEWAY]   || tb[RTA_OIF] ||
5026 		    tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5027 			NL_SET_ERR_MSG(extack,
5028 				       "Nexthop specification and nexthop id are mutually exclusive");
5029 			goto errout;
5030 		}
5031 		cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5032 	}
5033 
5034 	if (tb[RTA_GATEWAY]) {
5035 		cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5036 		cfg->fc_flags |= RTF_GATEWAY;
5037 	}
5038 	if (tb[RTA_VIA]) {
5039 		NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5040 		goto errout;
5041 	}
5042 
5043 	if (tb[RTA_DST]) {
5044 		int plen = (rtm->rtm_dst_len + 7) >> 3;
5045 
5046 		if (nla_len(tb[RTA_DST]) < plen)
5047 			goto errout;
5048 
5049 		nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5050 	}
5051 
5052 	if (tb[RTA_SRC]) {
5053 		int plen = (rtm->rtm_src_len + 7) >> 3;
5054 
5055 		if (nla_len(tb[RTA_SRC]) < plen)
5056 			goto errout;
5057 
5058 		nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5059 	}
5060 
5061 	if (tb[RTA_PREFSRC])
5062 		cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5063 
5064 	if (tb[RTA_OIF])
5065 		cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5066 
5067 	if (tb[RTA_PRIORITY])
5068 		cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5069 
5070 	if (tb[RTA_METRICS]) {
5071 		cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5072 		cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5073 	}
5074 
5075 	if (tb[RTA_TABLE])
5076 		cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5077 
5078 	if (tb[RTA_MULTIPATH]) {
5079 		cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5080 		cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5081 
5082 		err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5083 						     cfg->fc_mp_len, extack);
5084 		if (err < 0)
5085 			goto errout;
5086 	}
5087 
5088 	if (tb[RTA_PREF]) {
5089 		pref = nla_get_u8(tb[RTA_PREF]);
5090 		if (pref != ICMPV6_ROUTER_PREF_LOW &&
5091 		    pref != ICMPV6_ROUTER_PREF_HIGH)
5092 			pref = ICMPV6_ROUTER_PREF_MEDIUM;
5093 		cfg->fc_flags |= RTF_PREF(pref);
5094 	}
5095 
5096 	if (tb[RTA_ENCAP])
5097 		cfg->fc_encap = tb[RTA_ENCAP];
5098 
5099 	if (tb[RTA_ENCAP_TYPE]) {
5100 		cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5101 
5102 		err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5103 		if (err < 0)
5104 			goto errout;
5105 	}
5106 
5107 	if (tb[RTA_EXPIRES]) {
5108 		unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5109 
5110 		if (addrconf_finite_timeout(timeout)) {
5111 			cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5112 			cfg->fc_flags |= RTF_EXPIRES;
5113 		}
5114 	}
5115 
5116 	err = 0;
5117 errout:
5118 	return err;
5119 }
5120 
5121 struct rt6_nh {
5122 	struct fib6_info *fib6_info;
5123 	struct fib6_config r_cfg;
5124 	struct list_head next;
5125 };
5126 
5127 static int ip6_route_info_append(struct net *net,
5128 				 struct list_head *rt6_nh_list,
5129 				 struct fib6_info *rt,
5130 				 struct fib6_config *r_cfg)
5131 {
5132 	struct rt6_nh *nh;
5133 	int err = -EEXIST;
5134 
5135 	list_for_each_entry(nh, rt6_nh_list, next) {
5136 		/* check if fib6_info already exists */
5137 		if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5138 			return err;
5139 	}
5140 
5141 	nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5142 	if (!nh)
5143 		return -ENOMEM;
5144 	nh->fib6_info = rt;
5145 	memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5146 	list_add_tail(&nh->next, rt6_nh_list);
5147 
5148 	return 0;
5149 }
5150 
5151 static void ip6_route_mpath_notify(struct fib6_info *rt,
5152 				   struct fib6_info *rt_last,
5153 				   struct nl_info *info,
5154 				   __u16 nlflags)
5155 {
5156 	/* if this is an APPEND route, then rt points to the first route
5157 	 * inserted and rt_last points to last route inserted. Userspace
5158 	 * wants a consistent dump of the route which starts at the first
5159 	 * nexthop. Since sibling routes are always added at the end of
5160 	 * the list, find the first sibling of the last route appended
5161 	 */
5162 	if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5163 		rt = list_first_entry(&rt_last->fib6_siblings,
5164 				      struct fib6_info,
5165 				      fib6_siblings);
5166 	}
5167 
5168 	if (rt)
5169 		inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5170 }
5171 
5172 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5173 {
5174 	bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5175 	bool should_notify = false;
5176 	struct fib6_info *leaf;
5177 	struct fib6_node *fn;
5178 
5179 	rcu_read_lock();
5180 	fn = rcu_dereference(rt->fib6_node);
5181 	if (!fn)
5182 		goto out;
5183 
5184 	leaf = rcu_dereference(fn->leaf);
5185 	if (!leaf)
5186 		goto out;
5187 
5188 	if (rt == leaf ||
5189 	    (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5190 	     rt6_qualify_for_ecmp(leaf)))
5191 		should_notify = true;
5192 out:
5193 	rcu_read_unlock();
5194 
5195 	return should_notify;
5196 }
5197 
5198 static int ip6_route_multipath_add(struct fib6_config *cfg,
5199 				   struct netlink_ext_ack *extack)
5200 {
5201 	struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5202 	struct nl_info *info = &cfg->fc_nlinfo;
5203 	struct fib6_config r_cfg;
5204 	struct rtnexthop *rtnh;
5205 	struct fib6_info *rt;
5206 	struct rt6_nh *err_nh;
5207 	struct rt6_nh *nh, *nh_safe;
5208 	__u16 nlflags;
5209 	int remaining;
5210 	int attrlen;
5211 	int err = 1;
5212 	int nhn = 0;
5213 	int replace = (cfg->fc_nlinfo.nlh &&
5214 		       (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5215 	LIST_HEAD(rt6_nh_list);
5216 
5217 	nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5218 	if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5219 		nlflags |= NLM_F_APPEND;
5220 
5221 	remaining = cfg->fc_mp_len;
5222 	rtnh = (struct rtnexthop *)cfg->fc_mp;
5223 
5224 	/* Parse a Multipath Entry and build a list (rt6_nh_list) of
5225 	 * fib6_info structs per nexthop
5226 	 */
5227 	while (rtnh_ok(rtnh, remaining)) {
5228 		memcpy(&r_cfg, cfg, sizeof(*cfg));
5229 		if (rtnh->rtnh_ifindex)
5230 			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5231 
5232 		attrlen = rtnh_attrlen(rtnh);
5233 		if (attrlen > 0) {
5234 			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5235 
5236 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5237 			if (nla) {
5238 				r_cfg.fc_gateway = nla_get_in6_addr(nla);
5239 				r_cfg.fc_flags |= RTF_GATEWAY;
5240 			}
5241 			r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5242 			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5243 			if (nla)
5244 				r_cfg.fc_encap_type = nla_get_u16(nla);
5245 		}
5246 
5247 		r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5248 		rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5249 		if (IS_ERR(rt)) {
5250 			err = PTR_ERR(rt);
5251 			rt = NULL;
5252 			goto cleanup;
5253 		}
5254 		if (!rt6_qualify_for_ecmp(rt)) {
5255 			err = -EINVAL;
5256 			NL_SET_ERR_MSG(extack,
5257 				       "Device only routes can not be added for IPv6 using the multipath API.");
5258 			fib6_info_release(rt);
5259 			goto cleanup;
5260 		}
5261 
5262 		rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5263 
5264 		err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5265 					    rt, &r_cfg);
5266 		if (err) {
5267 			fib6_info_release(rt);
5268 			goto cleanup;
5269 		}
5270 
5271 		rtnh = rtnh_next(rtnh, &remaining);
5272 	}
5273 
5274 	if (list_empty(&rt6_nh_list)) {
5275 		NL_SET_ERR_MSG(extack,
5276 			       "Invalid nexthop configuration - no valid nexthops");
5277 		return -EINVAL;
5278 	}
5279 
5280 	/* for add and replace send one notification with all nexthops.
5281 	 * Skip the notification in fib6_add_rt2node and send one with
5282 	 * the full route when done
5283 	 */
5284 	info->skip_notify = 1;
5285 
5286 	/* For add and replace, send one notification with all nexthops. For
5287 	 * append, send one notification with all appended nexthops.
5288 	 */
5289 	info->skip_notify_kernel = 1;
5290 
5291 	err_nh = NULL;
5292 	list_for_each_entry(nh, &rt6_nh_list, next) {
5293 		err = __ip6_ins_rt(nh->fib6_info, info, extack);
5294 		fib6_info_release(nh->fib6_info);
5295 
5296 		if (!err) {
5297 			/* save reference to last route successfully inserted */
5298 			rt_last = nh->fib6_info;
5299 
5300 			/* save reference to first route for notification */
5301 			if (!rt_notif)
5302 				rt_notif = nh->fib6_info;
5303 		}
5304 
5305 		/* nh->fib6_info is used or freed at this point, reset to NULL*/
5306 		nh->fib6_info = NULL;
5307 		if (err) {
5308 			if (replace && nhn)
5309 				NL_SET_ERR_MSG_MOD(extack,
5310 						   "multipath route replace failed (check consistency of installed routes)");
5311 			err_nh = nh;
5312 			goto add_errout;
5313 		}
5314 
5315 		/* Because each route is added like a single route we remove
5316 		 * these flags after the first nexthop: if there is a collision,
5317 		 * we have already failed to add the first nexthop:
5318 		 * fib6_add_rt2node() has rejected it; when replacing, old
5319 		 * nexthops have been replaced by first new, the rest should
5320 		 * be added to it.
5321 		 */
5322 		if (cfg->fc_nlinfo.nlh) {
5323 			cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5324 							     NLM_F_REPLACE);
5325 			cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5326 		}
5327 		nhn++;
5328 	}
5329 
5330 	/* An in-kernel notification should only be sent in case the new
5331 	 * multipath route is added as the first route in the node, or if
5332 	 * it was appended to it. We pass 'rt_notif' since it is the first
5333 	 * sibling and might allow us to skip some checks in the replace case.
5334 	 */
5335 	if (ip6_route_mpath_should_notify(rt_notif)) {
5336 		enum fib_event_type fib_event;
5337 
5338 		if (rt_notif->fib6_nsiblings != nhn - 1)
5339 			fib_event = FIB_EVENT_ENTRY_APPEND;
5340 		else
5341 			fib_event = FIB_EVENT_ENTRY_REPLACE;
5342 
5343 		err = call_fib6_multipath_entry_notifiers(info->nl_net,
5344 							  fib_event, rt_notif,
5345 							  nhn - 1, extack);
5346 		if (err) {
5347 			/* Delete all the siblings that were just added */
5348 			err_nh = NULL;
5349 			goto add_errout;
5350 		}
5351 	}
5352 
5353 	/* success ... tell user about new route */
5354 	ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5355 	goto cleanup;
5356 
5357 add_errout:
5358 	/* send notification for routes that were added so that
5359 	 * the delete notifications sent by ip6_route_del are
5360 	 * coherent
5361 	 */
5362 	if (rt_notif)
5363 		ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5364 
5365 	/* Delete routes that were already added */
5366 	list_for_each_entry(nh, &rt6_nh_list, next) {
5367 		if (err_nh == nh)
5368 			break;
5369 		ip6_route_del(&nh->r_cfg, extack);
5370 	}
5371 
5372 cleanup:
5373 	list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5374 		if (nh->fib6_info)
5375 			fib6_info_release(nh->fib6_info);
5376 		list_del(&nh->next);
5377 		kfree(nh);
5378 	}
5379 
5380 	return err;
5381 }
5382 
5383 static int ip6_route_multipath_del(struct fib6_config *cfg,
5384 				   struct netlink_ext_ack *extack)
5385 {
5386 	struct fib6_config r_cfg;
5387 	struct rtnexthop *rtnh;
5388 	int last_err = 0;
5389 	int remaining;
5390 	int attrlen;
5391 	int err;
5392 
5393 	remaining = cfg->fc_mp_len;
5394 	rtnh = (struct rtnexthop *)cfg->fc_mp;
5395 
5396 	/* Parse a Multipath Entry */
5397 	while (rtnh_ok(rtnh, remaining)) {
5398 		memcpy(&r_cfg, cfg, sizeof(*cfg));
5399 		if (rtnh->rtnh_ifindex)
5400 			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5401 
5402 		attrlen = rtnh_attrlen(rtnh);
5403 		if (attrlen > 0) {
5404 			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5405 
5406 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5407 			if (nla) {
5408 				nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5409 				r_cfg.fc_flags |= RTF_GATEWAY;
5410 			}
5411 		}
5412 		err = ip6_route_del(&r_cfg, extack);
5413 		if (err)
5414 			last_err = err;
5415 
5416 		rtnh = rtnh_next(rtnh, &remaining);
5417 	}
5418 
5419 	return last_err;
5420 }
5421 
5422 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5423 			      struct netlink_ext_ack *extack)
5424 {
5425 	struct fib6_config cfg;
5426 	int err;
5427 
5428 	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5429 	if (err < 0)
5430 		return err;
5431 
5432 	if (cfg.fc_nh_id &&
5433 	    !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5434 		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5435 		return -EINVAL;
5436 	}
5437 
5438 	if (cfg.fc_mp)
5439 		return ip6_route_multipath_del(&cfg, extack);
5440 	else {
5441 		cfg.fc_delete_all_nh = 1;
5442 		return ip6_route_del(&cfg, extack);
5443 	}
5444 }
5445 
5446 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5447 			      struct netlink_ext_ack *extack)
5448 {
5449 	struct fib6_config cfg;
5450 	int err;
5451 
5452 	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5453 	if (err < 0)
5454 		return err;
5455 
5456 	if (cfg.fc_metric == 0)
5457 		cfg.fc_metric = IP6_RT_PRIO_USER;
5458 
5459 	if (cfg.fc_mp)
5460 		return ip6_route_multipath_add(&cfg, extack);
5461 	else
5462 		return ip6_route_add(&cfg, GFP_KERNEL, extack);
5463 }
5464 
5465 /* add the overhead of this fib6_nh to nexthop_len */
5466 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5467 {
5468 	int *nexthop_len = arg;
5469 
5470 	*nexthop_len += nla_total_size(0)	 /* RTA_MULTIPATH */
5471 		     + NLA_ALIGN(sizeof(struct rtnexthop))
5472 		     + nla_total_size(16); /* RTA_GATEWAY */
5473 
5474 	if (nh->fib_nh_lws) {
5475 		/* RTA_ENCAP_TYPE */
5476 		*nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5477 		/* RTA_ENCAP */
5478 		*nexthop_len += nla_total_size(2);
5479 	}
5480 
5481 	return 0;
5482 }
5483 
5484 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5485 {
5486 	int nexthop_len;
5487 
5488 	if (f6i->nh) {
5489 		nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5490 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5491 					 &nexthop_len);
5492 	} else {
5493 		struct fib6_nh *nh = f6i->fib6_nh;
5494 
5495 		nexthop_len = 0;
5496 		if (f6i->fib6_nsiblings) {
5497 			nexthop_len = nla_total_size(0)	 /* RTA_MULTIPATH */
5498 				    + NLA_ALIGN(sizeof(struct rtnexthop))
5499 				    + nla_total_size(16) /* RTA_GATEWAY */
5500 				    + lwtunnel_get_encap_size(nh->fib_nh_lws);
5501 
5502 			nexthop_len *= f6i->fib6_nsiblings;
5503 		}
5504 		nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5505 	}
5506 
5507 	return NLMSG_ALIGN(sizeof(struct rtmsg))
5508 	       + nla_total_size(16) /* RTA_SRC */
5509 	       + nla_total_size(16) /* RTA_DST */
5510 	       + nla_total_size(16) /* RTA_GATEWAY */
5511 	       + nla_total_size(16) /* RTA_PREFSRC */
5512 	       + nla_total_size(4) /* RTA_TABLE */
5513 	       + nla_total_size(4) /* RTA_IIF */
5514 	       + nla_total_size(4) /* RTA_OIF */
5515 	       + nla_total_size(4) /* RTA_PRIORITY */
5516 	       + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5517 	       + nla_total_size(sizeof(struct rta_cacheinfo))
5518 	       + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5519 	       + nla_total_size(1) /* RTA_PREF */
5520 	       + nexthop_len;
5521 }
5522 
5523 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5524 				 unsigned char *flags)
5525 {
5526 	if (nexthop_is_multipath(nh)) {
5527 		struct nlattr *mp;
5528 
5529 		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5530 		if (!mp)
5531 			goto nla_put_failure;
5532 
5533 		if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5534 			goto nla_put_failure;
5535 
5536 		nla_nest_end(skb, mp);
5537 	} else {
5538 		struct fib6_nh *fib6_nh;
5539 
5540 		fib6_nh = nexthop_fib6_nh(nh);
5541 		if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5542 				     flags, false) < 0)
5543 			goto nla_put_failure;
5544 	}
5545 
5546 	return 0;
5547 
5548 nla_put_failure:
5549 	return -EMSGSIZE;
5550 }
5551 
5552 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5553 			 struct fib6_info *rt, struct dst_entry *dst,
5554 			 struct in6_addr *dest, struct in6_addr *src,
5555 			 int iif, int type, u32 portid, u32 seq,
5556 			 unsigned int flags)
5557 {
5558 	struct rt6_info *rt6 = (struct rt6_info *)dst;
5559 	struct rt6key *rt6_dst, *rt6_src;
5560 	u32 *pmetrics, table, rt6_flags;
5561 	unsigned char nh_flags = 0;
5562 	struct nlmsghdr *nlh;
5563 	struct rtmsg *rtm;
5564 	long expires = 0;
5565 
5566 	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5567 	if (!nlh)
5568 		return -EMSGSIZE;
5569 
5570 	if (rt6) {
5571 		rt6_dst = &rt6->rt6i_dst;
5572 		rt6_src = &rt6->rt6i_src;
5573 		rt6_flags = rt6->rt6i_flags;
5574 	} else {
5575 		rt6_dst = &rt->fib6_dst;
5576 		rt6_src = &rt->fib6_src;
5577 		rt6_flags = rt->fib6_flags;
5578 	}
5579 
5580 	rtm = nlmsg_data(nlh);
5581 	rtm->rtm_family = AF_INET6;
5582 	rtm->rtm_dst_len = rt6_dst->plen;
5583 	rtm->rtm_src_len = rt6_src->plen;
5584 	rtm->rtm_tos = 0;
5585 	if (rt->fib6_table)
5586 		table = rt->fib6_table->tb6_id;
5587 	else
5588 		table = RT6_TABLE_UNSPEC;
5589 	rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5590 	if (nla_put_u32(skb, RTA_TABLE, table))
5591 		goto nla_put_failure;
5592 
5593 	rtm->rtm_type = rt->fib6_type;
5594 	rtm->rtm_flags = 0;
5595 	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5596 	rtm->rtm_protocol = rt->fib6_protocol;
5597 
5598 	if (rt6_flags & RTF_CACHE)
5599 		rtm->rtm_flags |= RTM_F_CLONED;
5600 
5601 	if (dest) {
5602 		if (nla_put_in6_addr(skb, RTA_DST, dest))
5603 			goto nla_put_failure;
5604 		rtm->rtm_dst_len = 128;
5605 	} else if (rtm->rtm_dst_len)
5606 		if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5607 			goto nla_put_failure;
5608 #ifdef CONFIG_IPV6_SUBTREES
5609 	if (src) {
5610 		if (nla_put_in6_addr(skb, RTA_SRC, src))
5611 			goto nla_put_failure;
5612 		rtm->rtm_src_len = 128;
5613 	} else if (rtm->rtm_src_len &&
5614 		   nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5615 		goto nla_put_failure;
5616 #endif
5617 	if (iif) {
5618 #ifdef CONFIG_IPV6_MROUTE
5619 		if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5620 			int err = ip6mr_get_route(net, skb, rtm, portid);
5621 
5622 			if (err == 0)
5623 				return 0;
5624 			if (err < 0)
5625 				goto nla_put_failure;
5626 		} else
5627 #endif
5628 			if (nla_put_u32(skb, RTA_IIF, iif))
5629 				goto nla_put_failure;
5630 	} else if (dest) {
5631 		struct in6_addr saddr_buf;
5632 		if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5633 		    nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5634 			goto nla_put_failure;
5635 	}
5636 
5637 	if (rt->fib6_prefsrc.plen) {
5638 		struct in6_addr saddr_buf;
5639 		saddr_buf = rt->fib6_prefsrc.addr;
5640 		if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5641 			goto nla_put_failure;
5642 	}
5643 
5644 	pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5645 	if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5646 		goto nla_put_failure;
5647 
5648 	if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5649 		goto nla_put_failure;
5650 
5651 	/* For multipath routes, walk the siblings list and add
5652 	 * each as a nexthop within RTA_MULTIPATH.
5653 	 */
5654 	if (rt6) {
5655 		if (rt6_flags & RTF_GATEWAY &&
5656 		    nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5657 			goto nla_put_failure;
5658 
5659 		if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5660 			goto nla_put_failure;
5661 
5662 		if (dst->lwtstate &&
5663 		    lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5664 			goto nla_put_failure;
5665 	} else if (rt->fib6_nsiblings) {
5666 		struct fib6_info *sibling, *next_sibling;
5667 		struct nlattr *mp;
5668 
5669 		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5670 		if (!mp)
5671 			goto nla_put_failure;
5672 
5673 		if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5674 				    rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
5675 			goto nla_put_failure;
5676 
5677 		list_for_each_entry_safe(sibling, next_sibling,
5678 					 &rt->fib6_siblings, fib6_siblings) {
5679 			if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5680 					    sibling->fib6_nh->fib_nh_weight,
5681 					    AF_INET6) < 0)
5682 				goto nla_put_failure;
5683 		}
5684 
5685 		nla_nest_end(skb, mp);
5686 	} else if (rt->nh) {
5687 		if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5688 			goto nla_put_failure;
5689 
5690 		if (nexthop_is_blackhole(rt->nh))
5691 			rtm->rtm_type = RTN_BLACKHOLE;
5692 
5693 		if (net->ipv4.sysctl_nexthop_compat_mode &&
5694 		    rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5695 			goto nla_put_failure;
5696 
5697 		rtm->rtm_flags |= nh_flags;
5698 	} else {
5699 		if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5700 				     &nh_flags, false) < 0)
5701 			goto nla_put_failure;
5702 
5703 		rtm->rtm_flags |= nh_flags;
5704 	}
5705 
5706 	if (rt6_flags & RTF_EXPIRES) {
5707 		expires = dst ? dst->expires : rt->expires;
5708 		expires -= jiffies;
5709 	}
5710 
5711 	if (!dst) {
5712 		if (rt->offload)
5713 			rtm->rtm_flags |= RTM_F_OFFLOAD;
5714 		if (rt->trap)
5715 			rtm->rtm_flags |= RTM_F_TRAP;
5716 		if (rt->offload_failed)
5717 			rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5718 	}
5719 
5720 	if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5721 		goto nla_put_failure;
5722 
5723 	if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5724 		goto nla_put_failure;
5725 
5726 
5727 	nlmsg_end(skb, nlh);
5728 	return 0;
5729 
5730 nla_put_failure:
5731 	nlmsg_cancel(skb, nlh);
5732 	return -EMSGSIZE;
5733 }
5734 
5735 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5736 {
5737 	const struct net_device *dev = arg;
5738 
5739 	if (nh->fib_nh_dev == dev)
5740 		return 1;
5741 
5742 	return 0;
5743 }
5744 
5745 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5746 			       const struct net_device *dev)
5747 {
5748 	if (f6i->nh) {
5749 		struct net_device *_dev = (struct net_device *)dev;
5750 
5751 		return !!nexthop_for_each_fib6_nh(f6i->nh,
5752 						  fib6_info_nh_uses_dev,
5753 						  _dev);
5754 	}
5755 
5756 	if (f6i->fib6_nh->fib_nh_dev == dev)
5757 		return true;
5758 
5759 	if (f6i->fib6_nsiblings) {
5760 		struct fib6_info *sibling, *next_sibling;
5761 
5762 		list_for_each_entry_safe(sibling, next_sibling,
5763 					 &f6i->fib6_siblings, fib6_siblings) {
5764 			if (sibling->fib6_nh->fib_nh_dev == dev)
5765 				return true;
5766 		}
5767 	}
5768 
5769 	return false;
5770 }
5771 
5772 struct fib6_nh_exception_dump_walker {
5773 	struct rt6_rtnl_dump_arg *dump;
5774 	struct fib6_info *rt;
5775 	unsigned int flags;
5776 	unsigned int skip;
5777 	unsigned int count;
5778 };
5779 
5780 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5781 {
5782 	struct fib6_nh_exception_dump_walker *w = arg;
5783 	struct rt6_rtnl_dump_arg *dump = w->dump;
5784 	struct rt6_exception_bucket *bucket;
5785 	struct rt6_exception *rt6_ex;
5786 	int i, err;
5787 
5788 	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5789 	if (!bucket)
5790 		return 0;
5791 
5792 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5793 		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5794 			if (w->skip) {
5795 				w->skip--;
5796 				continue;
5797 			}
5798 
5799 			/* Expiration of entries doesn't bump sernum, insertion
5800 			 * does. Removal is triggered by insertion, so we can
5801 			 * rely on the fact that if entries change between two
5802 			 * partial dumps, this node is scanned again completely,
5803 			 * see rt6_insert_exception() and fib6_dump_table().
5804 			 *
5805 			 * Count expired entries we go through as handled
5806 			 * entries that we'll skip next time, in case of partial
5807 			 * node dump. Otherwise, if entries expire meanwhile,
5808 			 * we'll skip the wrong amount.
5809 			 */
5810 			if (rt6_check_expired(rt6_ex->rt6i)) {
5811 				w->count++;
5812 				continue;
5813 			}
5814 
5815 			err = rt6_fill_node(dump->net, dump->skb, w->rt,
5816 					    &rt6_ex->rt6i->dst, NULL, NULL, 0,
5817 					    RTM_NEWROUTE,
5818 					    NETLINK_CB(dump->cb->skb).portid,
5819 					    dump->cb->nlh->nlmsg_seq, w->flags);
5820 			if (err)
5821 				return err;
5822 
5823 			w->count++;
5824 		}
5825 		bucket++;
5826 	}
5827 
5828 	return 0;
5829 }
5830 
5831 /* Return -1 if done with node, number of handled routes on partial dump */
5832 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5833 {
5834 	struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5835 	struct fib_dump_filter *filter = &arg->filter;
5836 	unsigned int flags = NLM_F_MULTI;
5837 	struct net *net = arg->net;
5838 	int count = 0;
5839 
5840 	if (rt == net->ipv6.fib6_null_entry)
5841 		return -1;
5842 
5843 	if ((filter->flags & RTM_F_PREFIX) &&
5844 	    !(rt->fib6_flags & RTF_PREFIX_RT)) {
5845 		/* success since this is not a prefix route */
5846 		return -1;
5847 	}
5848 	if (filter->filter_set &&
5849 	    ((filter->rt_type  && rt->fib6_type != filter->rt_type) ||
5850 	     (filter->dev      && !fib6_info_uses_dev(rt, filter->dev)) ||
5851 	     (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5852 		return -1;
5853 	}
5854 
5855 	if (filter->filter_set ||
5856 	    !filter->dump_routes || !filter->dump_exceptions) {
5857 		flags |= NLM_F_DUMP_FILTERED;
5858 	}
5859 
5860 	if (filter->dump_routes) {
5861 		if (skip) {
5862 			skip--;
5863 		} else {
5864 			if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5865 					  0, RTM_NEWROUTE,
5866 					  NETLINK_CB(arg->cb->skb).portid,
5867 					  arg->cb->nlh->nlmsg_seq, flags)) {
5868 				return 0;
5869 			}
5870 			count++;
5871 		}
5872 	}
5873 
5874 	if (filter->dump_exceptions) {
5875 		struct fib6_nh_exception_dump_walker w = { .dump = arg,
5876 							   .rt = rt,
5877 							   .flags = flags,
5878 							   .skip = skip,
5879 							   .count = 0 };
5880 		int err;
5881 
5882 		rcu_read_lock();
5883 		if (rt->nh) {
5884 			err = nexthop_for_each_fib6_nh(rt->nh,
5885 						       rt6_nh_dump_exceptions,
5886 						       &w);
5887 		} else {
5888 			err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5889 		}
5890 		rcu_read_unlock();
5891 
5892 		if (err)
5893 			return count += w.count;
5894 	}
5895 
5896 	return -1;
5897 }
5898 
5899 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5900 					const struct nlmsghdr *nlh,
5901 					struct nlattr **tb,
5902 					struct netlink_ext_ack *extack)
5903 {
5904 	struct rtmsg *rtm;
5905 	int i, err;
5906 
5907 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5908 		NL_SET_ERR_MSG_MOD(extack,
5909 				   "Invalid header for get route request");
5910 		return -EINVAL;
5911 	}
5912 
5913 	if (!netlink_strict_get_check(skb))
5914 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5915 					      rtm_ipv6_policy, extack);
5916 
5917 	rtm = nlmsg_data(nlh);
5918 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5919 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5920 	    rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5921 	    rtm->rtm_type) {
5922 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5923 		return -EINVAL;
5924 	}
5925 	if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5926 		NL_SET_ERR_MSG_MOD(extack,
5927 				   "Invalid flags for get route request");
5928 		return -EINVAL;
5929 	}
5930 
5931 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5932 					    rtm_ipv6_policy, extack);
5933 	if (err)
5934 		return err;
5935 
5936 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5937 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5938 		NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5939 		return -EINVAL;
5940 	}
5941 
5942 	for (i = 0; i <= RTA_MAX; i++) {
5943 		if (!tb[i])
5944 			continue;
5945 
5946 		switch (i) {
5947 		case RTA_SRC:
5948 		case RTA_DST:
5949 		case RTA_IIF:
5950 		case RTA_OIF:
5951 		case RTA_MARK:
5952 		case RTA_UID:
5953 		case RTA_SPORT:
5954 		case RTA_DPORT:
5955 		case RTA_IP_PROTO:
5956 			break;
5957 		default:
5958 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5959 			return -EINVAL;
5960 		}
5961 	}
5962 
5963 	return 0;
5964 }
5965 
5966 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5967 			      struct netlink_ext_ack *extack)
5968 {
5969 	struct net *net = sock_net(in_skb->sk);
5970 	struct nlattr *tb[RTA_MAX+1];
5971 	int err, iif = 0, oif = 0;
5972 	struct fib6_info *from;
5973 	struct dst_entry *dst;
5974 	struct rt6_info *rt;
5975 	struct sk_buff *skb;
5976 	struct rtmsg *rtm;
5977 	struct flowi6 fl6 = {};
5978 	bool fibmatch;
5979 
5980 	err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
5981 	if (err < 0)
5982 		goto errout;
5983 
5984 	err = -EINVAL;
5985 	rtm = nlmsg_data(nlh);
5986 	fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
5987 	fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
5988 
5989 	if (tb[RTA_SRC]) {
5990 		if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5991 			goto errout;
5992 
5993 		fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
5994 	}
5995 
5996 	if (tb[RTA_DST]) {
5997 		if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5998 			goto errout;
5999 
6000 		fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6001 	}
6002 
6003 	if (tb[RTA_IIF])
6004 		iif = nla_get_u32(tb[RTA_IIF]);
6005 
6006 	if (tb[RTA_OIF])
6007 		oif = nla_get_u32(tb[RTA_OIF]);
6008 
6009 	if (tb[RTA_MARK])
6010 		fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6011 
6012 	if (tb[RTA_UID])
6013 		fl6.flowi6_uid = make_kuid(current_user_ns(),
6014 					   nla_get_u32(tb[RTA_UID]));
6015 	else
6016 		fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6017 
6018 	if (tb[RTA_SPORT])
6019 		fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6020 
6021 	if (tb[RTA_DPORT])
6022 		fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6023 
6024 	if (tb[RTA_IP_PROTO]) {
6025 		err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6026 						  &fl6.flowi6_proto, AF_INET6,
6027 						  extack);
6028 		if (err)
6029 			goto errout;
6030 	}
6031 
6032 	if (iif) {
6033 		struct net_device *dev;
6034 		int flags = 0;
6035 
6036 		rcu_read_lock();
6037 
6038 		dev = dev_get_by_index_rcu(net, iif);
6039 		if (!dev) {
6040 			rcu_read_unlock();
6041 			err = -ENODEV;
6042 			goto errout;
6043 		}
6044 
6045 		fl6.flowi6_iif = iif;
6046 
6047 		if (!ipv6_addr_any(&fl6.saddr))
6048 			flags |= RT6_LOOKUP_F_HAS_SADDR;
6049 
6050 		dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6051 
6052 		rcu_read_unlock();
6053 	} else {
6054 		fl6.flowi6_oif = oif;
6055 
6056 		dst = ip6_route_output(net, NULL, &fl6);
6057 	}
6058 
6059 
6060 	rt = container_of(dst, struct rt6_info, dst);
6061 	if (rt->dst.error) {
6062 		err = rt->dst.error;
6063 		ip6_rt_put(rt);
6064 		goto errout;
6065 	}
6066 
6067 	if (rt == net->ipv6.ip6_null_entry) {
6068 		err = rt->dst.error;
6069 		ip6_rt_put(rt);
6070 		goto errout;
6071 	}
6072 
6073 	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6074 	if (!skb) {
6075 		ip6_rt_put(rt);
6076 		err = -ENOBUFS;
6077 		goto errout;
6078 	}
6079 
6080 	skb_dst_set(skb, &rt->dst);
6081 
6082 	rcu_read_lock();
6083 	from = rcu_dereference(rt->from);
6084 	if (from) {
6085 		if (fibmatch)
6086 			err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6087 					    iif, RTM_NEWROUTE,
6088 					    NETLINK_CB(in_skb).portid,
6089 					    nlh->nlmsg_seq, 0);
6090 		else
6091 			err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6092 					    &fl6.saddr, iif, RTM_NEWROUTE,
6093 					    NETLINK_CB(in_skb).portid,
6094 					    nlh->nlmsg_seq, 0);
6095 	} else {
6096 		err = -ENETUNREACH;
6097 	}
6098 	rcu_read_unlock();
6099 
6100 	if (err < 0) {
6101 		kfree_skb(skb);
6102 		goto errout;
6103 	}
6104 
6105 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6106 errout:
6107 	return err;
6108 }
6109 
6110 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6111 		     unsigned int nlm_flags)
6112 {
6113 	struct sk_buff *skb;
6114 	struct net *net = info->nl_net;
6115 	u32 seq;
6116 	int err;
6117 
6118 	err = -ENOBUFS;
6119 	seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6120 
6121 	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6122 	if (!skb)
6123 		goto errout;
6124 
6125 	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6126 			    event, info->portid, seq, nlm_flags);
6127 	if (err < 0) {
6128 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6129 		WARN_ON(err == -EMSGSIZE);
6130 		kfree_skb(skb);
6131 		goto errout;
6132 	}
6133 	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6134 		    info->nlh, gfp_any());
6135 	return;
6136 errout:
6137 	if (err < 0)
6138 		rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6139 }
6140 
6141 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6142 		    struct nl_info *info)
6143 {
6144 	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6145 	struct sk_buff *skb;
6146 	int err = -ENOBUFS;
6147 
6148 	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6149 	if (!skb)
6150 		goto errout;
6151 
6152 	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6153 			    RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6154 	if (err < 0) {
6155 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6156 		WARN_ON(err == -EMSGSIZE);
6157 		kfree_skb(skb);
6158 		goto errout;
6159 	}
6160 	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6161 		    info->nlh, gfp_any());
6162 	return;
6163 errout:
6164 	if (err < 0)
6165 		rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6166 }
6167 
6168 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6169 			    bool offload, bool trap, bool offload_failed)
6170 {
6171 	struct sk_buff *skb;
6172 	int err;
6173 
6174 	if (f6i->offload == offload && f6i->trap == trap &&
6175 	    f6i->offload_failed == offload_failed)
6176 		return;
6177 
6178 	f6i->offload = offload;
6179 	f6i->trap = trap;
6180 
6181 	/* 2 means send notifications only if offload_failed was changed. */
6182 	if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6183 	    f6i->offload_failed == offload_failed)
6184 		return;
6185 
6186 	f6i->offload_failed = offload_failed;
6187 
6188 	if (!rcu_access_pointer(f6i->fib6_node))
6189 		/* The route was removed from the tree, do not send
6190 		 * notification.
6191 		 */
6192 		return;
6193 
6194 	if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6195 		return;
6196 
6197 	skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6198 	if (!skb) {
6199 		err = -ENOBUFS;
6200 		goto errout;
6201 	}
6202 
6203 	err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6204 			    0, 0);
6205 	if (err < 0) {
6206 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6207 		WARN_ON(err == -EMSGSIZE);
6208 		kfree_skb(skb);
6209 		goto errout;
6210 	}
6211 
6212 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6213 	return;
6214 
6215 errout:
6216 	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6217 }
6218 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6219 
6220 static int ip6_route_dev_notify(struct notifier_block *this,
6221 				unsigned long event, void *ptr)
6222 {
6223 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6224 	struct net *net = dev_net(dev);
6225 
6226 	if (!(dev->flags & IFF_LOOPBACK))
6227 		return NOTIFY_OK;
6228 
6229 	if (event == NETDEV_REGISTER) {
6230 		net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6231 		net->ipv6.ip6_null_entry->dst.dev = dev;
6232 		net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6233 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6234 		net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6235 		net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6236 		net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6237 		net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6238 #endif
6239 	 } else if (event == NETDEV_UNREGISTER &&
6240 		    dev->reg_state != NETREG_UNREGISTERED) {
6241 		/* NETDEV_UNREGISTER could be fired for multiple times by
6242 		 * netdev_wait_allrefs(). Make sure we only call this once.
6243 		 */
6244 		in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6245 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6246 		in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6247 		in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6248 #endif
6249 	}
6250 
6251 	return NOTIFY_OK;
6252 }
6253 
6254 /*
6255  *	/proc
6256  */
6257 
6258 #ifdef CONFIG_PROC_FS
6259 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6260 {
6261 	struct net *net = (struct net *)seq->private;
6262 	seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6263 		   net->ipv6.rt6_stats->fib_nodes,
6264 		   net->ipv6.rt6_stats->fib_route_nodes,
6265 		   atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6266 		   net->ipv6.rt6_stats->fib_rt_entries,
6267 		   net->ipv6.rt6_stats->fib_rt_cache,
6268 		   dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6269 		   net->ipv6.rt6_stats->fib_discarded_routes);
6270 
6271 	return 0;
6272 }
6273 #endif	/* CONFIG_PROC_FS */
6274 
6275 #ifdef CONFIG_SYSCTL
6276 
6277 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6278 			      void *buffer, size_t *lenp, loff_t *ppos)
6279 {
6280 	struct net *net;
6281 	int delay;
6282 	int ret;
6283 	if (!write)
6284 		return -EINVAL;
6285 
6286 	net = (struct net *)ctl->extra1;
6287 	delay = net->ipv6.sysctl.flush_delay;
6288 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6289 	if (ret)
6290 		return ret;
6291 
6292 	fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6293 	return 0;
6294 }
6295 
6296 static struct ctl_table ipv6_route_table_template[] = {
6297 	{
6298 		.procname	=	"flush",
6299 		.data		=	&init_net.ipv6.sysctl.flush_delay,
6300 		.maxlen		=	sizeof(int),
6301 		.mode		=	0200,
6302 		.proc_handler	=	ipv6_sysctl_rtcache_flush
6303 	},
6304 	{
6305 		.procname	=	"gc_thresh",
6306 		.data		=	&ip6_dst_ops_template.gc_thresh,
6307 		.maxlen		=	sizeof(int),
6308 		.mode		=	0644,
6309 		.proc_handler	=	proc_dointvec,
6310 	},
6311 	{
6312 		.procname	=	"max_size",
6313 		.data		=	&init_net.ipv6.sysctl.ip6_rt_max_size,
6314 		.maxlen		=	sizeof(int),
6315 		.mode		=	0644,
6316 		.proc_handler	=	proc_dointvec,
6317 	},
6318 	{
6319 		.procname	=	"gc_min_interval",
6320 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6321 		.maxlen		=	sizeof(int),
6322 		.mode		=	0644,
6323 		.proc_handler	=	proc_dointvec_jiffies,
6324 	},
6325 	{
6326 		.procname	=	"gc_timeout",
6327 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6328 		.maxlen		=	sizeof(int),
6329 		.mode		=	0644,
6330 		.proc_handler	=	proc_dointvec_jiffies,
6331 	},
6332 	{
6333 		.procname	=	"gc_interval",
6334 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_interval,
6335 		.maxlen		=	sizeof(int),
6336 		.mode		=	0644,
6337 		.proc_handler	=	proc_dointvec_jiffies,
6338 	},
6339 	{
6340 		.procname	=	"gc_elasticity",
6341 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6342 		.maxlen		=	sizeof(int),
6343 		.mode		=	0644,
6344 		.proc_handler	=	proc_dointvec,
6345 	},
6346 	{
6347 		.procname	=	"mtu_expires",
6348 		.data		=	&init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6349 		.maxlen		=	sizeof(int),
6350 		.mode		=	0644,
6351 		.proc_handler	=	proc_dointvec_jiffies,
6352 	},
6353 	{
6354 		.procname	=	"min_adv_mss",
6355 		.data		=	&init_net.ipv6.sysctl.ip6_rt_min_advmss,
6356 		.maxlen		=	sizeof(int),
6357 		.mode		=	0644,
6358 		.proc_handler	=	proc_dointvec,
6359 	},
6360 	{
6361 		.procname	=	"gc_min_interval_ms",
6362 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6363 		.maxlen		=	sizeof(int),
6364 		.mode		=	0644,
6365 		.proc_handler	=	proc_dointvec_ms_jiffies,
6366 	},
6367 	{
6368 		.procname	=	"skip_notify_on_dev_down",
6369 		.data		=	&init_net.ipv6.sysctl.skip_notify_on_dev_down,
6370 		.maxlen		=	sizeof(int),
6371 		.mode		=	0644,
6372 		.proc_handler	=	proc_dointvec_minmax,
6373 		.extra1		=	SYSCTL_ZERO,
6374 		.extra2		=	SYSCTL_ONE,
6375 	},
6376 	{ }
6377 };
6378 
6379 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6380 {
6381 	struct ctl_table *table;
6382 
6383 	table = kmemdup(ipv6_route_table_template,
6384 			sizeof(ipv6_route_table_template),
6385 			GFP_KERNEL);
6386 
6387 	if (table) {
6388 		table[0].data = &net->ipv6.sysctl.flush_delay;
6389 		table[0].extra1 = net;
6390 		table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6391 		table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6392 		table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6393 		table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6394 		table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6395 		table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6396 		table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6397 		table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6398 		table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6399 		table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6400 
6401 		/* Don't export sysctls to unprivileged users */
6402 		if (net->user_ns != &init_user_ns)
6403 			table[0].procname = NULL;
6404 	}
6405 
6406 	return table;
6407 }
6408 #endif
6409 
6410 static int __net_init ip6_route_net_init(struct net *net)
6411 {
6412 	int ret = -ENOMEM;
6413 
6414 	memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6415 	       sizeof(net->ipv6.ip6_dst_ops));
6416 
6417 	if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6418 		goto out_ip6_dst_ops;
6419 
6420 	net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6421 	if (!net->ipv6.fib6_null_entry)
6422 		goto out_ip6_dst_entries;
6423 	memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6424 	       sizeof(*net->ipv6.fib6_null_entry));
6425 
6426 	net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6427 					   sizeof(*net->ipv6.ip6_null_entry),
6428 					   GFP_KERNEL);
6429 	if (!net->ipv6.ip6_null_entry)
6430 		goto out_fib6_null_entry;
6431 	net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6432 	dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6433 			 ip6_template_metrics, true);
6434 	INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6435 
6436 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6437 	net->ipv6.fib6_has_custom_rules = false;
6438 	net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6439 					       sizeof(*net->ipv6.ip6_prohibit_entry),
6440 					       GFP_KERNEL);
6441 	if (!net->ipv6.ip6_prohibit_entry)
6442 		goto out_ip6_null_entry;
6443 	net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6444 	dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6445 			 ip6_template_metrics, true);
6446 	INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6447 
6448 	net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6449 					       sizeof(*net->ipv6.ip6_blk_hole_entry),
6450 					       GFP_KERNEL);
6451 	if (!net->ipv6.ip6_blk_hole_entry)
6452 		goto out_ip6_prohibit_entry;
6453 	net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6454 	dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6455 			 ip6_template_metrics, true);
6456 	INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6457 #ifdef CONFIG_IPV6_SUBTREES
6458 	net->ipv6.fib6_routes_require_src = 0;
6459 #endif
6460 #endif
6461 
6462 	net->ipv6.sysctl.flush_delay = 0;
6463 	net->ipv6.sysctl.ip6_rt_max_size = 4096;
6464 	net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6465 	net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6466 	net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6467 	net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6468 	net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6469 	net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6470 	net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6471 
6472 	net->ipv6.ip6_rt_gc_expire = 30*HZ;
6473 
6474 	ret = 0;
6475 out:
6476 	return ret;
6477 
6478 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6479 out_ip6_prohibit_entry:
6480 	kfree(net->ipv6.ip6_prohibit_entry);
6481 out_ip6_null_entry:
6482 	kfree(net->ipv6.ip6_null_entry);
6483 #endif
6484 out_fib6_null_entry:
6485 	kfree(net->ipv6.fib6_null_entry);
6486 out_ip6_dst_entries:
6487 	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6488 out_ip6_dst_ops:
6489 	goto out;
6490 }
6491 
6492 static void __net_exit ip6_route_net_exit(struct net *net)
6493 {
6494 	kfree(net->ipv6.fib6_null_entry);
6495 	kfree(net->ipv6.ip6_null_entry);
6496 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6497 	kfree(net->ipv6.ip6_prohibit_entry);
6498 	kfree(net->ipv6.ip6_blk_hole_entry);
6499 #endif
6500 	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6501 }
6502 
6503 static int __net_init ip6_route_net_init_late(struct net *net)
6504 {
6505 #ifdef CONFIG_PROC_FS
6506 	proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6507 			sizeof(struct ipv6_route_iter));
6508 	proc_create_net_single("rt6_stats", 0444, net->proc_net,
6509 			rt6_stats_seq_show, NULL);
6510 #endif
6511 	return 0;
6512 }
6513 
6514 static void __net_exit ip6_route_net_exit_late(struct net *net)
6515 {
6516 #ifdef CONFIG_PROC_FS
6517 	remove_proc_entry("ipv6_route", net->proc_net);
6518 	remove_proc_entry("rt6_stats", net->proc_net);
6519 #endif
6520 }
6521 
6522 static struct pernet_operations ip6_route_net_ops = {
6523 	.init = ip6_route_net_init,
6524 	.exit = ip6_route_net_exit,
6525 };
6526 
6527 static int __net_init ipv6_inetpeer_init(struct net *net)
6528 {
6529 	struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6530 
6531 	if (!bp)
6532 		return -ENOMEM;
6533 	inet_peer_base_init(bp);
6534 	net->ipv6.peers = bp;
6535 	return 0;
6536 }
6537 
6538 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6539 {
6540 	struct inet_peer_base *bp = net->ipv6.peers;
6541 
6542 	net->ipv6.peers = NULL;
6543 	inetpeer_invalidate_tree(bp);
6544 	kfree(bp);
6545 }
6546 
6547 static struct pernet_operations ipv6_inetpeer_ops = {
6548 	.init	=	ipv6_inetpeer_init,
6549 	.exit	=	ipv6_inetpeer_exit,
6550 };
6551 
6552 static struct pernet_operations ip6_route_net_late_ops = {
6553 	.init = ip6_route_net_init_late,
6554 	.exit = ip6_route_net_exit_late,
6555 };
6556 
6557 static struct notifier_block ip6_route_dev_notifier = {
6558 	.notifier_call = ip6_route_dev_notify,
6559 	.priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6560 };
6561 
6562 void __init ip6_route_init_special_entries(void)
6563 {
6564 	/* Registering of the loopback is done before this portion of code,
6565 	 * the loopback reference in rt6_info will not be taken, do it
6566 	 * manually for init_net */
6567 	init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6568 	init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6569 	init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6570   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6571 	init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6572 	init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6573 	init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6574 	init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6575   #endif
6576 }
6577 
6578 #if IS_BUILTIN(CONFIG_IPV6)
6579 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6580 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6581 
6582 BTF_ID_LIST(btf_fib6_info_id)
6583 BTF_ID(struct, fib6_info)
6584 
6585 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6586 	.seq_ops		= &ipv6_route_seq_ops,
6587 	.init_seq_private	= bpf_iter_init_seq_net,
6588 	.fini_seq_private	= bpf_iter_fini_seq_net,
6589 	.seq_priv_size		= sizeof(struct ipv6_route_iter),
6590 };
6591 
6592 static struct bpf_iter_reg ipv6_route_reg_info = {
6593 	.target			= "ipv6_route",
6594 	.ctx_arg_info_size	= 1,
6595 	.ctx_arg_info		= {
6596 		{ offsetof(struct bpf_iter__ipv6_route, rt),
6597 		  PTR_TO_BTF_ID_OR_NULL },
6598 	},
6599 	.seq_info		= &ipv6_route_seq_info,
6600 };
6601 
6602 static int __init bpf_iter_register(void)
6603 {
6604 	ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6605 	return bpf_iter_reg_target(&ipv6_route_reg_info);
6606 }
6607 
6608 static void bpf_iter_unregister(void)
6609 {
6610 	bpf_iter_unreg_target(&ipv6_route_reg_info);
6611 }
6612 #endif
6613 #endif
6614 
6615 int __init ip6_route_init(void)
6616 {
6617 	int ret;
6618 	int cpu;
6619 
6620 	ret = -ENOMEM;
6621 	ip6_dst_ops_template.kmem_cachep =
6622 		kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6623 				  SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6624 	if (!ip6_dst_ops_template.kmem_cachep)
6625 		goto out;
6626 
6627 	ret = dst_entries_init(&ip6_dst_blackhole_ops);
6628 	if (ret)
6629 		goto out_kmem_cache;
6630 
6631 	ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6632 	if (ret)
6633 		goto out_dst_entries;
6634 
6635 	ret = register_pernet_subsys(&ip6_route_net_ops);
6636 	if (ret)
6637 		goto out_register_inetpeer;
6638 
6639 	ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6640 
6641 	ret = fib6_init();
6642 	if (ret)
6643 		goto out_register_subsys;
6644 
6645 	ret = xfrm6_init();
6646 	if (ret)
6647 		goto out_fib6_init;
6648 
6649 	ret = fib6_rules_init();
6650 	if (ret)
6651 		goto xfrm6_init;
6652 
6653 	ret = register_pernet_subsys(&ip6_route_net_late_ops);
6654 	if (ret)
6655 		goto fib6_rules_init;
6656 
6657 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6658 				   inet6_rtm_newroute, NULL, 0);
6659 	if (ret < 0)
6660 		goto out_register_late_subsys;
6661 
6662 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6663 				   inet6_rtm_delroute, NULL, 0);
6664 	if (ret < 0)
6665 		goto out_register_late_subsys;
6666 
6667 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6668 				   inet6_rtm_getroute, NULL,
6669 				   RTNL_FLAG_DOIT_UNLOCKED);
6670 	if (ret < 0)
6671 		goto out_register_late_subsys;
6672 
6673 	ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6674 	if (ret)
6675 		goto out_register_late_subsys;
6676 
6677 #if IS_BUILTIN(CONFIG_IPV6)
6678 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6679 	ret = bpf_iter_register();
6680 	if (ret)
6681 		goto out_register_late_subsys;
6682 #endif
6683 #endif
6684 
6685 	for_each_possible_cpu(cpu) {
6686 		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6687 
6688 		INIT_LIST_HEAD(&ul->head);
6689 		spin_lock_init(&ul->lock);
6690 	}
6691 
6692 out:
6693 	return ret;
6694 
6695 out_register_late_subsys:
6696 	rtnl_unregister_all(PF_INET6);
6697 	unregister_pernet_subsys(&ip6_route_net_late_ops);
6698 fib6_rules_init:
6699 	fib6_rules_cleanup();
6700 xfrm6_init:
6701 	xfrm6_fini();
6702 out_fib6_init:
6703 	fib6_gc_cleanup();
6704 out_register_subsys:
6705 	unregister_pernet_subsys(&ip6_route_net_ops);
6706 out_register_inetpeer:
6707 	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6708 out_dst_entries:
6709 	dst_entries_destroy(&ip6_dst_blackhole_ops);
6710 out_kmem_cache:
6711 	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6712 	goto out;
6713 }
6714 
6715 void ip6_route_cleanup(void)
6716 {
6717 #if IS_BUILTIN(CONFIG_IPV6)
6718 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6719 	bpf_iter_unregister();
6720 #endif
6721 #endif
6722 	unregister_netdevice_notifier(&ip6_route_dev_notifier);
6723 	unregister_pernet_subsys(&ip6_route_net_late_ops);
6724 	fib6_rules_cleanup();
6725 	xfrm6_fini();
6726 	fib6_gc_cleanup();
6727 	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6728 	unregister_pernet_subsys(&ip6_route_net_ops);
6729 	dst_entries_destroy(&ip6_dst_blackhole_ops);
6730 	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6731 }
6732