xref: /openbmc/linux/net/ipv6/route.c (revision 1634f2bf)
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 
65 #ifdef CONFIG_SYSCTL
66 #include <linux/sysctl.h>
67 #endif
68 
69 static int ip6_rt_type_to_error(u8 fib6_type);
70 
71 #define CREATE_TRACE_POINTS
72 #include <trace/events/fib6.h>
73 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
74 #undef CREATE_TRACE_POINTS
75 
76 enum rt6_nud_state {
77 	RT6_NUD_FAIL_HARD = -3,
78 	RT6_NUD_FAIL_PROBE = -2,
79 	RT6_NUD_FAIL_DO_RR = -1,
80 	RT6_NUD_SUCCEED = 1
81 };
82 
83 static struct dst_entry	*ip6_dst_check(struct dst_entry *dst, u32 cookie);
84 static unsigned int	 ip6_default_advmss(const struct dst_entry *dst);
85 static unsigned int	 ip6_mtu(const struct dst_entry *dst);
86 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87 static void		ip6_dst_destroy(struct dst_entry *);
88 static void		ip6_dst_ifdown(struct dst_entry *,
89 				       struct net_device *dev, int how);
90 static int		 ip6_dst_gc(struct dst_ops *ops);
91 
92 static int		ip6_pkt_discard(struct sk_buff *skb);
93 static int		ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
94 static int		ip6_pkt_prohibit(struct sk_buff *skb);
95 static int		ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
96 static void		ip6_link_failure(struct sk_buff *skb);
97 static void		ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
98 					   struct sk_buff *skb, u32 mtu);
99 static void		rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
100 					struct sk_buff *skb);
101 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
102 			   int strict);
103 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
104 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
105 			 struct fib6_info *rt, struct dst_entry *dst,
106 			 struct in6_addr *dest, struct in6_addr *src,
107 			 int iif, int type, u32 portid, u32 seq,
108 			 unsigned int flags);
109 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
110 					   const struct in6_addr *daddr,
111 					   const struct in6_addr *saddr);
112 
113 #ifdef CONFIG_IPV6_ROUTE_INFO
114 static struct fib6_info *rt6_add_route_info(struct net *net,
115 					   const struct in6_addr *prefix, int prefixlen,
116 					   const struct in6_addr *gwaddr,
117 					   struct net_device *dev,
118 					   unsigned int pref);
119 static struct fib6_info *rt6_get_route_info(struct net *net,
120 					   const struct in6_addr *prefix, int prefixlen,
121 					   const struct in6_addr *gwaddr,
122 					   struct net_device *dev);
123 #endif
124 
125 struct uncached_list {
126 	spinlock_t		lock;
127 	struct list_head	head;
128 };
129 
130 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
131 
132 void rt6_uncached_list_add(struct rt6_info *rt)
133 {
134 	struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
135 
136 	rt->rt6i_uncached_list = ul;
137 
138 	spin_lock_bh(&ul->lock);
139 	list_add_tail(&rt->rt6i_uncached, &ul->head);
140 	spin_unlock_bh(&ul->lock);
141 }
142 
143 void rt6_uncached_list_del(struct rt6_info *rt)
144 {
145 	if (!list_empty(&rt->rt6i_uncached)) {
146 		struct uncached_list *ul = rt->rt6i_uncached_list;
147 		struct net *net = dev_net(rt->dst.dev);
148 
149 		spin_lock_bh(&ul->lock);
150 		list_del(&rt->rt6i_uncached);
151 		atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
152 		spin_unlock_bh(&ul->lock);
153 	}
154 }
155 
156 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
157 {
158 	struct net_device *loopback_dev = net->loopback_dev;
159 	int cpu;
160 
161 	if (dev == loopback_dev)
162 		return;
163 
164 	for_each_possible_cpu(cpu) {
165 		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166 		struct rt6_info *rt;
167 
168 		spin_lock_bh(&ul->lock);
169 		list_for_each_entry(rt, &ul->head, rt6i_uncached) {
170 			struct inet6_dev *rt_idev = rt->rt6i_idev;
171 			struct net_device *rt_dev = rt->dst.dev;
172 
173 			if (rt_idev->dev == dev) {
174 				rt->rt6i_idev = in6_dev_get(loopback_dev);
175 				in6_dev_put(rt_idev);
176 			}
177 
178 			if (rt_dev == dev) {
179 				rt->dst.dev = blackhole_netdev;
180 				dev_hold(rt->dst.dev);
181 				dev_put(rt_dev);
182 			}
183 		}
184 		spin_unlock_bh(&ul->lock);
185 	}
186 }
187 
188 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
189 					     struct sk_buff *skb,
190 					     const void *daddr)
191 {
192 	if (!ipv6_addr_any(p))
193 		return (const void *) p;
194 	else if (skb)
195 		return &ipv6_hdr(skb)->daddr;
196 	return daddr;
197 }
198 
199 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
200 				   struct net_device *dev,
201 				   struct sk_buff *skb,
202 				   const void *daddr)
203 {
204 	struct neighbour *n;
205 
206 	daddr = choose_neigh_daddr(gw, skb, daddr);
207 	n = __ipv6_neigh_lookup(dev, daddr);
208 	if (n)
209 		return n;
210 
211 	n = neigh_create(&nd_tbl, daddr, dev);
212 	return IS_ERR(n) ? NULL : n;
213 }
214 
215 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
216 					      struct sk_buff *skb,
217 					      const void *daddr)
218 {
219 	const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
220 
221 	return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
222 				dst->dev, skb, daddr);
223 }
224 
225 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
226 {
227 	struct net_device *dev = dst->dev;
228 	struct rt6_info *rt = (struct rt6_info *)dst;
229 
230 	daddr = choose_neigh_daddr(&rt->rt6i_gateway, NULL, daddr);
231 	if (!daddr)
232 		return;
233 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
234 		return;
235 	if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
236 		return;
237 	__ipv6_confirm_neigh(dev, daddr);
238 }
239 
240 static struct dst_ops ip6_dst_ops_template = {
241 	.family			=	AF_INET6,
242 	.gc			=	ip6_dst_gc,
243 	.gc_thresh		=	1024,
244 	.check			=	ip6_dst_check,
245 	.default_advmss		=	ip6_default_advmss,
246 	.mtu			=	ip6_mtu,
247 	.cow_metrics		=	dst_cow_metrics_generic,
248 	.destroy		=	ip6_dst_destroy,
249 	.ifdown			=	ip6_dst_ifdown,
250 	.negative_advice	=	ip6_negative_advice,
251 	.link_failure		=	ip6_link_failure,
252 	.update_pmtu		=	ip6_rt_update_pmtu,
253 	.redirect		=	rt6_do_redirect,
254 	.local_out		=	__ip6_local_out,
255 	.neigh_lookup		=	ip6_dst_neigh_lookup,
256 	.confirm_neigh		=	ip6_confirm_neigh,
257 };
258 
259 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
260 {
261 	unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
262 
263 	return mtu ? : dst->dev->mtu;
264 }
265 
266 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
267 					 struct sk_buff *skb, u32 mtu)
268 {
269 }
270 
271 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
272 				      struct sk_buff *skb)
273 {
274 }
275 
276 static struct dst_ops ip6_dst_blackhole_ops = {
277 	.family			=	AF_INET6,
278 	.destroy		=	ip6_dst_destroy,
279 	.check			=	ip6_dst_check,
280 	.mtu			=	ip6_blackhole_mtu,
281 	.default_advmss		=	ip6_default_advmss,
282 	.update_pmtu		=	ip6_rt_blackhole_update_pmtu,
283 	.redirect		=	ip6_rt_blackhole_redirect,
284 	.cow_metrics		=	dst_cow_metrics_generic,
285 	.neigh_lookup		=	ip6_dst_neigh_lookup,
286 };
287 
288 static const u32 ip6_template_metrics[RTAX_MAX] = {
289 	[RTAX_HOPLIMIT - 1] = 0,
290 };
291 
292 static const struct fib6_info fib6_null_entry_template = {
293 	.fib6_flags	= (RTF_REJECT | RTF_NONEXTHOP),
294 	.fib6_protocol  = RTPROT_KERNEL,
295 	.fib6_metric	= ~(u32)0,
296 	.fib6_ref	= REFCOUNT_INIT(1),
297 	.fib6_type	= RTN_UNREACHABLE,
298 	.fib6_metrics	= (struct dst_metrics *)&dst_default_metrics,
299 };
300 
301 static const struct rt6_info ip6_null_entry_template = {
302 	.dst = {
303 		.__refcnt	= ATOMIC_INIT(1),
304 		.__use		= 1,
305 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
306 		.error		= -ENETUNREACH,
307 		.input		= ip6_pkt_discard,
308 		.output		= ip6_pkt_discard_out,
309 	},
310 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
311 };
312 
313 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
314 
315 static const struct rt6_info ip6_prohibit_entry_template = {
316 	.dst = {
317 		.__refcnt	= ATOMIC_INIT(1),
318 		.__use		= 1,
319 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
320 		.error		= -EACCES,
321 		.input		= ip6_pkt_prohibit,
322 		.output		= ip6_pkt_prohibit_out,
323 	},
324 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
325 };
326 
327 static const struct rt6_info ip6_blk_hole_entry_template = {
328 	.dst = {
329 		.__refcnt	= ATOMIC_INIT(1),
330 		.__use		= 1,
331 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
332 		.error		= -EINVAL,
333 		.input		= dst_discard,
334 		.output		= dst_discard_out,
335 	},
336 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
337 };
338 
339 #endif
340 
341 static void rt6_info_init(struct rt6_info *rt)
342 {
343 	struct dst_entry *dst = &rt->dst;
344 
345 	memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
346 	INIT_LIST_HEAD(&rt->rt6i_uncached);
347 }
348 
349 /* allocate dst with ip6_dst_ops */
350 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
351 			       int flags)
352 {
353 	struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
354 					1, DST_OBSOLETE_FORCE_CHK, flags);
355 
356 	if (rt) {
357 		rt6_info_init(rt);
358 		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
359 	}
360 
361 	return rt;
362 }
363 EXPORT_SYMBOL(ip6_dst_alloc);
364 
365 static void ip6_dst_destroy(struct dst_entry *dst)
366 {
367 	struct rt6_info *rt = (struct rt6_info *)dst;
368 	struct fib6_info *from;
369 	struct inet6_dev *idev;
370 
371 	ip_dst_metrics_put(dst);
372 	rt6_uncached_list_del(rt);
373 
374 	idev = rt->rt6i_idev;
375 	if (idev) {
376 		rt->rt6i_idev = NULL;
377 		in6_dev_put(idev);
378 	}
379 
380 	from = xchg((__force struct fib6_info **)&rt->from, NULL);
381 	fib6_info_release(from);
382 }
383 
384 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
385 			   int how)
386 {
387 	struct rt6_info *rt = (struct rt6_info *)dst;
388 	struct inet6_dev *idev = rt->rt6i_idev;
389 	struct net_device *loopback_dev =
390 		dev_net(dev)->loopback_dev;
391 
392 	if (idev && idev->dev != loopback_dev) {
393 		struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
394 		if (loopback_idev) {
395 			rt->rt6i_idev = loopback_idev;
396 			in6_dev_put(idev);
397 		}
398 	}
399 }
400 
401 static bool __rt6_check_expired(const struct rt6_info *rt)
402 {
403 	if (rt->rt6i_flags & RTF_EXPIRES)
404 		return time_after(jiffies, rt->dst.expires);
405 	else
406 		return false;
407 }
408 
409 static bool rt6_check_expired(const struct rt6_info *rt)
410 {
411 	struct fib6_info *from;
412 
413 	from = rcu_dereference(rt->from);
414 
415 	if (rt->rt6i_flags & RTF_EXPIRES) {
416 		if (time_after(jiffies, rt->dst.expires))
417 			return true;
418 	} else if (from) {
419 		return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
420 			fib6_check_expired(from);
421 	}
422 	return false;
423 }
424 
425 void fib6_select_path(const struct net *net, struct fib6_result *res,
426 		      struct flowi6 *fl6, int oif, bool have_oif_match,
427 		      const struct sk_buff *skb, int strict)
428 {
429 	struct fib6_info *sibling, *next_sibling;
430 	struct fib6_info *match = res->f6i;
431 
432 	if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
433 		goto out;
434 
435 	/* We might have already computed the hash for ICMPv6 errors. In such
436 	 * case it will always be non-zero. Otherwise now is the time to do it.
437 	 */
438 	if (!fl6->mp_hash &&
439 	    (!match->nh || nexthop_is_multipath(match->nh)))
440 		fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
441 
442 	if (unlikely(match->nh)) {
443 		nexthop_path_fib6_result(res, fl6->mp_hash);
444 		return;
445 	}
446 
447 	if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
448 		goto out;
449 
450 	list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
451 				 fib6_siblings) {
452 		const struct fib6_nh *nh = sibling->fib6_nh;
453 		int nh_upper_bound;
454 
455 		nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
456 		if (fl6->mp_hash > nh_upper_bound)
457 			continue;
458 		if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
459 			break;
460 		match = sibling;
461 		break;
462 	}
463 
464 out:
465 	res->f6i = match;
466 	res->nh = match->fib6_nh;
467 }
468 
469 /*
470  *	Route lookup. rcu_read_lock() should be held.
471  */
472 
473 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
474 			       const struct in6_addr *saddr, int oif, int flags)
475 {
476 	const struct net_device *dev;
477 
478 	if (nh->fib_nh_flags & RTNH_F_DEAD)
479 		return false;
480 
481 	dev = nh->fib_nh_dev;
482 	if (oif) {
483 		if (dev->ifindex == oif)
484 			return true;
485 	} else {
486 		if (ipv6_chk_addr(net, saddr, dev,
487 				  flags & RT6_LOOKUP_F_IFACE))
488 			return true;
489 	}
490 
491 	return false;
492 }
493 
494 struct fib6_nh_dm_arg {
495 	struct net		*net;
496 	const struct in6_addr	*saddr;
497 	int			oif;
498 	int			flags;
499 	struct fib6_nh		*nh;
500 };
501 
502 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
503 {
504 	struct fib6_nh_dm_arg *arg = _arg;
505 
506 	arg->nh = nh;
507 	return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
508 				  arg->flags);
509 }
510 
511 /* returns fib6_nh from nexthop or NULL */
512 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
513 					struct fib6_result *res,
514 					const struct in6_addr *saddr,
515 					int oif, int flags)
516 {
517 	struct fib6_nh_dm_arg arg = {
518 		.net   = net,
519 		.saddr = saddr,
520 		.oif   = oif,
521 		.flags = flags,
522 	};
523 
524 	if (nexthop_is_blackhole(nh))
525 		return NULL;
526 
527 	if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
528 		return arg.nh;
529 
530 	return NULL;
531 }
532 
533 static void rt6_device_match(struct net *net, struct fib6_result *res,
534 			     const struct in6_addr *saddr, int oif, int flags)
535 {
536 	struct fib6_info *f6i = res->f6i;
537 	struct fib6_info *spf6i;
538 	struct fib6_nh *nh;
539 
540 	if (!oif && ipv6_addr_any(saddr)) {
541 		if (unlikely(f6i->nh)) {
542 			nh = nexthop_fib6_nh(f6i->nh);
543 			if (nexthop_is_blackhole(f6i->nh))
544 				goto out_blackhole;
545 		} else {
546 			nh = f6i->fib6_nh;
547 		}
548 		if (!(nh->fib_nh_flags & RTNH_F_DEAD))
549 			goto out;
550 	}
551 
552 	for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
553 		bool matched = false;
554 
555 		if (unlikely(spf6i->nh)) {
556 			nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
557 					      oif, flags);
558 			if (nh)
559 				matched = true;
560 		} else {
561 			nh = spf6i->fib6_nh;
562 			if (__rt6_device_match(net, nh, saddr, oif, flags))
563 				matched = true;
564 		}
565 		if (matched) {
566 			res->f6i = spf6i;
567 			goto out;
568 		}
569 	}
570 
571 	if (oif && flags & RT6_LOOKUP_F_IFACE) {
572 		res->f6i = net->ipv6.fib6_null_entry;
573 		nh = res->f6i->fib6_nh;
574 		goto out;
575 	}
576 
577 	if (unlikely(f6i->nh)) {
578 		nh = nexthop_fib6_nh(f6i->nh);
579 		if (nexthop_is_blackhole(f6i->nh))
580 			goto out_blackhole;
581 	} else {
582 		nh = f6i->fib6_nh;
583 	}
584 
585 	if (nh->fib_nh_flags & RTNH_F_DEAD) {
586 		res->f6i = net->ipv6.fib6_null_entry;
587 		nh = res->f6i->fib6_nh;
588 	}
589 out:
590 	res->nh = nh;
591 	res->fib6_type = res->f6i->fib6_type;
592 	res->fib6_flags = res->f6i->fib6_flags;
593 	return;
594 
595 out_blackhole:
596 	res->fib6_flags |= RTF_REJECT;
597 	res->fib6_type = RTN_BLACKHOLE;
598 	res->nh = nh;
599 }
600 
601 #ifdef CONFIG_IPV6_ROUTER_PREF
602 struct __rt6_probe_work {
603 	struct work_struct work;
604 	struct in6_addr target;
605 	struct net_device *dev;
606 };
607 
608 static void rt6_probe_deferred(struct work_struct *w)
609 {
610 	struct in6_addr mcaddr;
611 	struct __rt6_probe_work *work =
612 		container_of(w, struct __rt6_probe_work, work);
613 
614 	addrconf_addr_solict_mult(&work->target, &mcaddr);
615 	ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
616 	dev_put(work->dev);
617 	kfree(work);
618 }
619 
620 static void rt6_probe(struct fib6_nh *fib6_nh)
621 {
622 	struct __rt6_probe_work *work = NULL;
623 	const struct in6_addr *nh_gw;
624 	struct neighbour *neigh;
625 	struct net_device *dev;
626 	struct inet6_dev *idev;
627 
628 	/*
629 	 * Okay, this does not seem to be appropriate
630 	 * for now, however, we need to check if it
631 	 * is really so; aka Router Reachability Probing.
632 	 *
633 	 * Router Reachability Probe MUST be rate-limited
634 	 * to no more than one per minute.
635 	 */
636 	if (fib6_nh->fib_nh_gw_family)
637 		return;
638 
639 	nh_gw = &fib6_nh->fib_nh_gw6;
640 	dev = fib6_nh->fib_nh_dev;
641 	rcu_read_lock_bh();
642 	idev = __in6_dev_get(dev);
643 	neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
644 	if (neigh) {
645 		if (neigh->nud_state & NUD_VALID)
646 			goto out;
647 
648 		write_lock(&neigh->lock);
649 		if (!(neigh->nud_state & NUD_VALID) &&
650 		    time_after(jiffies,
651 			       neigh->updated + idev->cnf.rtr_probe_interval)) {
652 			work = kmalloc(sizeof(*work), GFP_ATOMIC);
653 			if (work)
654 				__neigh_set_probe_once(neigh);
655 		}
656 		write_unlock(&neigh->lock);
657 	} else if (time_after(jiffies, fib6_nh->last_probe +
658 				       idev->cnf.rtr_probe_interval)) {
659 		work = kmalloc(sizeof(*work), GFP_ATOMIC);
660 	}
661 
662 	if (work) {
663 		fib6_nh->last_probe = jiffies;
664 		INIT_WORK(&work->work, rt6_probe_deferred);
665 		work->target = *nh_gw;
666 		dev_hold(dev);
667 		work->dev = dev;
668 		schedule_work(&work->work);
669 	}
670 
671 out:
672 	rcu_read_unlock_bh();
673 }
674 #else
675 static inline void rt6_probe(struct fib6_nh *fib6_nh)
676 {
677 }
678 #endif
679 
680 /*
681  * Default Router Selection (RFC 2461 6.3.6)
682  */
683 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
684 {
685 	enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
686 	struct neighbour *neigh;
687 
688 	rcu_read_lock_bh();
689 	neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
690 					  &fib6_nh->fib_nh_gw6);
691 	if (neigh) {
692 		read_lock(&neigh->lock);
693 		if (neigh->nud_state & NUD_VALID)
694 			ret = RT6_NUD_SUCCEED;
695 #ifdef CONFIG_IPV6_ROUTER_PREF
696 		else if (!(neigh->nud_state & NUD_FAILED))
697 			ret = RT6_NUD_SUCCEED;
698 		else
699 			ret = RT6_NUD_FAIL_PROBE;
700 #endif
701 		read_unlock(&neigh->lock);
702 	} else {
703 		ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
704 		      RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
705 	}
706 	rcu_read_unlock_bh();
707 
708 	return ret;
709 }
710 
711 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
712 			   int strict)
713 {
714 	int m = 0;
715 
716 	if (!oif || nh->fib_nh_dev->ifindex == oif)
717 		m = 2;
718 
719 	if (!m && (strict & RT6_LOOKUP_F_IFACE))
720 		return RT6_NUD_FAIL_HARD;
721 #ifdef CONFIG_IPV6_ROUTER_PREF
722 	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
723 #endif
724 	if ((strict & RT6_LOOKUP_F_REACHABLE) &&
725 	    !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
726 		int n = rt6_check_neigh(nh);
727 		if (n < 0)
728 			return n;
729 	}
730 	return m;
731 }
732 
733 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
734 		       int oif, int strict, int *mpri, bool *do_rr)
735 {
736 	bool match_do_rr = false;
737 	bool rc = false;
738 	int m;
739 
740 	if (nh->fib_nh_flags & RTNH_F_DEAD)
741 		goto out;
742 
743 	if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
744 	    nh->fib_nh_flags & RTNH_F_LINKDOWN &&
745 	    !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
746 		goto out;
747 
748 	m = rt6_score_route(nh, fib6_flags, oif, strict);
749 	if (m == RT6_NUD_FAIL_DO_RR) {
750 		match_do_rr = true;
751 		m = 0; /* lowest valid score */
752 	} else if (m == RT6_NUD_FAIL_HARD) {
753 		goto out;
754 	}
755 
756 	if (strict & RT6_LOOKUP_F_REACHABLE)
757 		rt6_probe(nh);
758 
759 	/* note that m can be RT6_NUD_FAIL_PROBE at this point */
760 	if (m > *mpri) {
761 		*do_rr = match_do_rr;
762 		*mpri = m;
763 		rc = true;
764 	}
765 out:
766 	return rc;
767 }
768 
769 struct fib6_nh_frl_arg {
770 	u32		flags;
771 	int		oif;
772 	int		strict;
773 	int		*mpri;
774 	bool		*do_rr;
775 	struct fib6_nh	*nh;
776 };
777 
778 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
779 {
780 	struct fib6_nh_frl_arg *arg = _arg;
781 
782 	arg->nh = nh;
783 	return find_match(nh, arg->flags, arg->oif, arg->strict,
784 			  arg->mpri, arg->do_rr);
785 }
786 
787 static void __find_rr_leaf(struct fib6_info *f6i_start,
788 			   struct fib6_info *nomatch, u32 metric,
789 			   struct fib6_result *res, struct fib6_info **cont,
790 			   int oif, int strict, bool *do_rr, int *mpri)
791 {
792 	struct fib6_info *f6i;
793 
794 	for (f6i = f6i_start;
795 	     f6i && f6i != nomatch;
796 	     f6i = rcu_dereference(f6i->fib6_next)) {
797 		bool matched = false;
798 		struct fib6_nh *nh;
799 
800 		if (cont && f6i->fib6_metric != metric) {
801 			*cont = f6i;
802 			return;
803 		}
804 
805 		if (fib6_check_expired(f6i))
806 			continue;
807 
808 		if (unlikely(f6i->nh)) {
809 			struct fib6_nh_frl_arg arg = {
810 				.flags  = f6i->fib6_flags,
811 				.oif    = oif,
812 				.strict = strict,
813 				.mpri   = mpri,
814 				.do_rr  = do_rr
815 			};
816 
817 			if (nexthop_is_blackhole(f6i->nh)) {
818 				res->fib6_flags = RTF_REJECT;
819 				res->fib6_type = RTN_BLACKHOLE;
820 				res->f6i = f6i;
821 				res->nh = nexthop_fib6_nh(f6i->nh);
822 				return;
823 			}
824 			if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
825 						     &arg)) {
826 				matched = true;
827 				nh = arg.nh;
828 			}
829 		} else {
830 			nh = f6i->fib6_nh;
831 			if (find_match(nh, f6i->fib6_flags, oif, strict,
832 				       mpri, do_rr))
833 				matched = true;
834 		}
835 		if (matched) {
836 			res->f6i = f6i;
837 			res->nh = nh;
838 			res->fib6_flags = f6i->fib6_flags;
839 			res->fib6_type = f6i->fib6_type;
840 		}
841 	}
842 }
843 
844 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
845 			 struct fib6_info *rr_head, int oif, int strict,
846 			 bool *do_rr, struct fib6_result *res)
847 {
848 	u32 metric = rr_head->fib6_metric;
849 	struct fib6_info *cont = NULL;
850 	int mpri = -1;
851 
852 	__find_rr_leaf(rr_head, NULL, metric, res, &cont,
853 		       oif, strict, do_rr, &mpri);
854 
855 	__find_rr_leaf(leaf, rr_head, metric, res, &cont,
856 		       oif, strict, do_rr, &mpri);
857 
858 	if (res->f6i || !cont)
859 		return;
860 
861 	__find_rr_leaf(cont, NULL, metric, res, NULL,
862 		       oif, strict, do_rr, &mpri);
863 }
864 
865 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
866 		       struct fib6_result *res, int strict)
867 {
868 	struct fib6_info *leaf = rcu_dereference(fn->leaf);
869 	struct fib6_info *rt0;
870 	bool do_rr = false;
871 	int key_plen;
872 
873 	/* make sure this function or its helpers sets f6i */
874 	res->f6i = NULL;
875 
876 	if (!leaf || leaf == net->ipv6.fib6_null_entry)
877 		goto out;
878 
879 	rt0 = rcu_dereference(fn->rr_ptr);
880 	if (!rt0)
881 		rt0 = leaf;
882 
883 	/* Double check to make sure fn is not an intermediate node
884 	 * and fn->leaf does not points to its child's leaf
885 	 * (This might happen if all routes under fn are deleted from
886 	 * the tree and fib6_repair_tree() is called on the node.)
887 	 */
888 	key_plen = rt0->fib6_dst.plen;
889 #ifdef CONFIG_IPV6_SUBTREES
890 	if (rt0->fib6_src.plen)
891 		key_plen = rt0->fib6_src.plen;
892 #endif
893 	if (fn->fn_bit != key_plen)
894 		goto out;
895 
896 	find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
897 	if (do_rr) {
898 		struct fib6_info *next = rcu_dereference(rt0->fib6_next);
899 
900 		/* no entries matched; do round-robin */
901 		if (!next || next->fib6_metric != rt0->fib6_metric)
902 			next = leaf;
903 
904 		if (next != rt0) {
905 			spin_lock_bh(&leaf->fib6_table->tb6_lock);
906 			/* make sure next is not being deleted from the tree */
907 			if (next->fib6_node)
908 				rcu_assign_pointer(fn->rr_ptr, next);
909 			spin_unlock_bh(&leaf->fib6_table->tb6_lock);
910 		}
911 	}
912 
913 out:
914 	if (!res->f6i) {
915 		res->f6i = net->ipv6.fib6_null_entry;
916 		res->nh = res->f6i->fib6_nh;
917 		res->fib6_flags = res->f6i->fib6_flags;
918 		res->fib6_type = res->f6i->fib6_type;
919 	}
920 }
921 
922 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
923 {
924 	return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
925 	       res->nh->fib_nh_gw_family;
926 }
927 
928 #ifdef CONFIG_IPV6_ROUTE_INFO
929 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
930 		  const struct in6_addr *gwaddr)
931 {
932 	struct net *net = dev_net(dev);
933 	struct route_info *rinfo = (struct route_info *) opt;
934 	struct in6_addr prefix_buf, *prefix;
935 	unsigned int pref;
936 	unsigned long lifetime;
937 	struct fib6_info *rt;
938 
939 	if (len < sizeof(struct route_info)) {
940 		return -EINVAL;
941 	}
942 
943 	/* Sanity check for prefix_len and length */
944 	if (rinfo->length > 3) {
945 		return -EINVAL;
946 	} else if (rinfo->prefix_len > 128) {
947 		return -EINVAL;
948 	} else if (rinfo->prefix_len > 64) {
949 		if (rinfo->length < 2) {
950 			return -EINVAL;
951 		}
952 	} else if (rinfo->prefix_len > 0) {
953 		if (rinfo->length < 1) {
954 			return -EINVAL;
955 		}
956 	}
957 
958 	pref = rinfo->route_pref;
959 	if (pref == ICMPV6_ROUTER_PREF_INVALID)
960 		return -EINVAL;
961 
962 	lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
963 
964 	if (rinfo->length == 3)
965 		prefix = (struct in6_addr *)rinfo->prefix;
966 	else {
967 		/* this function is safe */
968 		ipv6_addr_prefix(&prefix_buf,
969 				 (struct in6_addr *)rinfo->prefix,
970 				 rinfo->prefix_len);
971 		prefix = &prefix_buf;
972 	}
973 
974 	if (rinfo->prefix_len == 0)
975 		rt = rt6_get_dflt_router(net, gwaddr, dev);
976 	else
977 		rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
978 					gwaddr, dev);
979 
980 	if (rt && !lifetime) {
981 		ip6_del_rt(net, rt);
982 		rt = NULL;
983 	}
984 
985 	if (!rt && lifetime)
986 		rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
987 					dev, pref);
988 	else if (rt)
989 		rt->fib6_flags = RTF_ROUTEINFO |
990 				 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
991 
992 	if (rt) {
993 		if (!addrconf_finite_timeout(lifetime))
994 			fib6_clean_expires(rt);
995 		else
996 			fib6_set_expires(rt, jiffies + HZ * lifetime);
997 
998 		fib6_info_release(rt);
999 	}
1000 	return 0;
1001 }
1002 #endif
1003 
1004 /*
1005  *	Misc support functions
1006  */
1007 
1008 /* called with rcu_lock held */
1009 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1010 {
1011 	struct net_device *dev = res->nh->fib_nh_dev;
1012 
1013 	if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1014 		/* for copies of local routes, dst->dev needs to be the
1015 		 * device if it is a master device, the master device if
1016 		 * device is enslaved, and the loopback as the default
1017 		 */
1018 		if (netif_is_l3_slave(dev) &&
1019 		    !rt6_need_strict(&res->f6i->fib6_dst.addr))
1020 			dev = l3mdev_master_dev_rcu(dev);
1021 		else if (!netif_is_l3_master(dev))
1022 			dev = dev_net(dev)->loopback_dev;
1023 		/* last case is netif_is_l3_master(dev) is true in which
1024 		 * case we want dev returned to be dev
1025 		 */
1026 	}
1027 
1028 	return dev;
1029 }
1030 
1031 static const int fib6_prop[RTN_MAX + 1] = {
1032 	[RTN_UNSPEC]	= 0,
1033 	[RTN_UNICAST]	= 0,
1034 	[RTN_LOCAL]	= 0,
1035 	[RTN_BROADCAST]	= 0,
1036 	[RTN_ANYCAST]	= 0,
1037 	[RTN_MULTICAST]	= 0,
1038 	[RTN_BLACKHOLE]	= -EINVAL,
1039 	[RTN_UNREACHABLE] = -EHOSTUNREACH,
1040 	[RTN_PROHIBIT]	= -EACCES,
1041 	[RTN_THROW]	= -EAGAIN,
1042 	[RTN_NAT]	= -EINVAL,
1043 	[RTN_XRESOLVE]	= -EINVAL,
1044 };
1045 
1046 static int ip6_rt_type_to_error(u8 fib6_type)
1047 {
1048 	return fib6_prop[fib6_type];
1049 }
1050 
1051 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1052 {
1053 	unsigned short flags = 0;
1054 
1055 	if (rt->dst_nocount)
1056 		flags |= DST_NOCOUNT;
1057 	if (rt->dst_nopolicy)
1058 		flags |= DST_NOPOLICY;
1059 	if (rt->dst_host)
1060 		flags |= DST_HOST;
1061 
1062 	return flags;
1063 }
1064 
1065 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1066 {
1067 	rt->dst.error = ip6_rt_type_to_error(fib6_type);
1068 
1069 	switch (fib6_type) {
1070 	case RTN_BLACKHOLE:
1071 		rt->dst.output = dst_discard_out;
1072 		rt->dst.input = dst_discard;
1073 		break;
1074 	case RTN_PROHIBIT:
1075 		rt->dst.output = ip6_pkt_prohibit_out;
1076 		rt->dst.input = ip6_pkt_prohibit;
1077 		break;
1078 	case RTN_THROW:
1079 	case RTN_UNREACHABLE:
1080 	default:
1081 		rt->dst.output = ip6_pkt_discard_out;
1082 		rt->dst.input = ip6_pkt_discard;
1083 		break;
1084 	}
1085 }
1086 
1087 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1088 {
1089 	struct fib6_info *f6i = res->f6i;
1090 
1091 	if (res->fib6_flags & RTF_REJECT) {
1092 		ip6_rt_init_dst_reject(rt, res->fib6_type);
1093 		return;
1094 	}
1095 
1096 	rt->dst.error = 0;
1097 	rt->dst.output = ip6_output;
1098 
1099 	if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1100 		rt->dst.input = ip6_input;
1101 	} else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1102 		rt->dst.input = ip6_mc_input;
1103 	} else {
1104 		rt->dst.input = ip6_forward;
1105 	}
1106 
1107 	if (res->nh->fib_nh_lws) {
1108 		rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1109 		lwtunnel_set_redirect(&rt->dst);
1110 	}
1111 
1112 	rt->dst.lastuse = jiffies;
1113 }
1114 
1115 /* Caller must already hold reference to @from */
1116 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1117 {
1118 	rt->rt6i_flags &= ~RTF_EXPIRES;
1119 	rcu_assign_pointer(rt->from, from);
1120 	ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1121 }
1122 
1123 /* Caller must already hold reference to f6i in result */
1124 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1125 {
1126 	const struct fib6_nh *nh = res->nh;
1127 	const struct net_device *dev = nh->fib_nh_dev;
1128 	struct fib6_info *f6i = res->f6i;
1129 
1130 	ip6_rt_init_dst(rt, res);
1131 
1132 	rt->rt6i_dst = f6i->fib6_dst;
1133 	rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1134 	rt->rt6i_flags = res->fib6_flags;
1135 	if (nh->fib_nh_gw_family) {
1136 		rt->rt6i_gateway = nh->fib_nh_gw6;
1137 		rt->rt6i_flags |= RTF_GATEWAY;
1138 	}
1139 	rt6_set_from(rt, f6i);
1140 #ifdef CONFIG_IPV6_SUBTREES
1141 	rt->rt6i_src = f6i->fib6_src;
1142 #endif
1143 }
1144 
1145 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1146 					struct in6_addr *saddr)
1147 {
1148 	struct fib6_node *pn, *sn;
1149 	while (1) {
1150 		if (fn->fn_flags & RTN_TL_ROOT)
1151 			return NULL;
1152 		pn = rcu_dereference(fn->parent);
1153 		sn = FIB6_SUBTREE(pn);
1154 		if (sn && sn != fn)
1155 			fn = fib6_node_lookup(sn, NULL, saddr);
1156 		else
1157 			fn = pn;
1158 		if (fn->fn_flags & RTN_RTINFO)
1159 			return fn;
1160 	}
1161 }
1162 
1163 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1164 {
1165 	struct rt6_info *rt = *prt;
1166 
1167 	if (dst_hold_safe(&rt->dst))
1168 		return true;
1169 	if (net) {
1170 		rt = net->ipv6.ip6_null_entry;
1171 		dst_hold(&rt->dst);
1172 	} else {
1173 		rt = NULL;
1174 	}
1175 	*prt = rt;
1176 	return false;
1177 }
1178 
1179 /* called with rcu_lock held */
1180 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1181 {
1182 	struct net_device *dev = res->nh->fib_nh_dev;
1183 	struct fib6_info *f6i = res->f6i;
1184 	unsigned short flags;
1185 	struct rt6_info *nrt;
1186 
1187 	if (!fib6_info_hold_safe(f6i))
1188 		goto fallback;
1189 
1190 	flags = fib6_info_dst_flags(f6i);
1191 	nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1192 	if (!nrt) {
1193 		fib6_info_release(f6i);
1194 		goto fallback;
1195 	}
1196 
1197 	ip6_rt_copy_init(nrt, res);
1198 	return nrt;
1199 
1200 fallback:
1201 	nrt = dev_net(dev)->ipv6.ip6_null_entry;
1202 	dst_hold(&nrt->dst);
1203 	return nrt;
1204 }
1205 
1206 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1207 					     struct fib6_table *table,
1208 					     struct flowi6 *fl6,
1209 					     const struct sk_buff *skb,
1210 					     int flags)
1211 {
1212 	struct fib6_result res = {};
1213 	struct fib6_node *fn;
1214 	struct rt6_info *rt;
1215 
1216 	if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1217 		flags &= ~RT6_LOOKUP_F_IFACE;
1218 
1219 	rcu_read_lock();
1220 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1221 restart:
1222 	res.f6i = rcu_dereference(fn->leaf);
1223 	if (!res.f6i)
1224 		res.f6i = net->ipv6.fib6_null_entry;
1225 	else
1226 		rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1227 				 flags);
1228 
1229 	if (res.f6i == net->ipv6.fib6_null_entry) {
1230 		fn = fib6_backtrack(fn, &fl6->saddr);
1231 		if (fn)
1232 			goto restart;
1233 
1234 		rt = net->ipv6.ip6_null_entry;
1235 		dst_hold(&rt->dst);
1236 		goto out;
1237 	} else if (res.fib6_flags & RTF_REJECT) {
1238 		goto do_create;
1239 	}
1240 
1241 	fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1242 			 fl6->flowi6_oif != 0, skb, flags);
1243 
1244 	/* Search through exception table */
1245 	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1246 	if (rt) {
1247 		if (ip6_hold_safe(net, &rt))
1248 			dst_use_noref(&rt->dst, jiffies);
1249 	} else {
1250 do_create:
1251 		rt = ip6_create_rt_rcu(&res);
1252 	}
1253 
1254 out:
1255 	trace_fib6_table_lookup(net, &res, table, fl6);
1256 
1257 	rcu_read_unlock();
1258 
1259 	return rt;
1260 }
1261 
1262 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1263 				   const struct sk_buff *skb, int flags)
1264 {
1265 	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1266 }
1267 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1268 
1269 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1270 			    const struct in6_addr *saddr, int oif,
1271 			    const struct sk_buff *skb, int strict)
1272 {
1273 	struct flowi6 fl6 = {
1274 		.flowi6_oif = oif,
1275 		.daddr = *daddr,
1276 	};
1277 	struct dst_entry *dst;
1278 	int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1279 
1280 	if (saddr) {
1281 		memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1282 		flags |= RT6_LOOKUP_F_HAS_SADDR;
1283 	}
1284 
1285 	dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1286 	if (dst->error == 0)
1287 		return (struct rt6_info *) dst;
1288 
1289 	dst_release(dst);
1290 
1291 	return NULL;
1292 }
1293 EXPORT_SYMBOL(rt6_lookup);
1294 
1295 /* ip6_ins_rt is called with FREE table->tb6_lock.
1296  * It takes new route entry, the addition fails by any reason the
1297  * route is released.
1298  * Caller must hold dst before calling it.
1299  */
1300 
1301 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1302 			struct netlink_ext_ack *extack)
1303 {
1304 	int err;
1305 	struct fib6_table *table;
1306 
1307 	table = rt->fib6_table;
1308 	spin_lock_bh(&table->tb6_lock);
1309 	err = fib6_add(&table->tb6_root, rt, info, extack);
1310 	spin_unlock_bh(&table->tb6_lock);
1311 
1312 	return err;
1313 }
1314 
1315 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1316 {
1317 	struct nl_info info = {	.nl_net = net, };
1318 
1319 	return __ip6_ins_rt(rt, &info, NULL);
1320 }
1321 
1322 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1323 					   const struct in6_addr *daddr,
1324 					   const struct in6_addr *saddr)
1325 {
1326 	struct fib6_info *f6i = res->f6i;
1327 	struct net_device *dev;
1328 	struct rt6_info *rt;
1329 
1330 	/*
1331 	 *	Clone the route.
1332 	 */
1333 
1334 	if (!fib6_info_hold_safe(f6i))
1335 		return NULL;
1336 
1337 	dev = ip6_rt_get_dev_rcu(res);
1338 	rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1339 	if (!rt) {
1340 		fib6_info_release(f6i);
1341 		return NULL;
1342 	}
1343 
1344 	ip6_rt_copy_init(rt, res);
1345 	rt->rt6i_flags |= RTF_CACHE;
1346 	rt->dst.flags |= DST_HOST;
1347 	rt->rt6i_dst.addr = *daddr;
1348 	rt->rt6i_dst.plen = 128;
1349 
1350 	if (!rt6_is_gw_or_nonexthop(res)) {
1351 		if (f6i->fib6_dst.plen != 128 &&
1352 		    ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1353 			rt->rt6i_flags |= RTF_ANYCAST;
1354 #ifdef CONFIG_IPV6_SUBTREES
1355 		if (rt->rt6i_src.plen && saddr) {
1356 			rt->rt6i_src.addr = *saddr;
1357 			rt->rt6i_src.plen = 128;
1358 		}
1359 #endif
1360 	}
1361 
1362 	return rt;
1363 }
1364 
1365 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1366 {
1367 	struct fib6_info *f6i = res->f6i;
1368 	unsigned short flags = fib6_info_dst_flags(f6i);
1369 	struct net_device *dev;
1370 	struct rt6_info *pcpu_rt;
1371 
1372 	if (!fib6_info_hold_safe(f6i))
1373 		return NULL;
1374 
1375 	rcu_read_lock();
1376 	dev = ip6_rt_get_dev_rcu(res);
1377 	pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
1378 	rcu_read_unlock();
1379 	if (!pcpu_rt) {
1380 		fib6_info_release(f6i);
1381 		return NULL;
1382 	}
1383 	ip6_rt_copy_init(pcpu_rt, res);
1384 	pcpu_rt->rt6i_flags |= RTF_PCPU;
1385 	return pcpu_rt;
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 	return pcpu_rt;
1396 }
1397 
1398 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1399 					    const struct fib6_result *res)
1400 {
1401 	struct rt6_info *pcpu_rt, *prev, **p;
1402 
1403 	pcpu_rt = ip6_rt_pcpu_alloc(res);
1404 	if (!pcpu_rt)
1405 		return NULL;
1406 
1407 	p = this_cpu_ptr(res->nh->rt6i_pcpu);
1408 	prev = cmpxchg(p, NULL, pcpu_rt);
1409 	BUG_ON(prev);
1410 
1411 	if (res->f6i->fib6_destroying) {
1412 		struct fib6_info *from;
1413 
1414 		from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1415 		fib6_info_release(from);
1416 	}
1417 
1418 	return pcpu_rt;
1419 }
1420 
1421 /* exception hash table implementation
1422  */
1423 static DEFINE_SPINLOCK(rt6_exception_lock);
1424 
1425 /* Remove rt6_ex from hash table and free the memory
1426  * Caller must hold rt6_exception_lock
1427  */
1428 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1429 				 struct rt6_exception *rt6_ex)
1430 {
1431 	struct fib6_info *from;
1432 	struct net *net;
1433 
1434 	if (!bucket || !rt6_ex)
1435 		return;
1436 
1437 	net = dev_net(rt6_ex->rt6i->dst.dev);
1438 	net->ipv6.rt6_stats->fib_rt_cache--;
1439 
1440 	/* purge completely the exception to allow releasing the held resources:
1441 	 * some [sk] cache may keep the dst around for unlimited time
1442 	 */
1443 	from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1444 	fib6_info_release(from);
1445 	dst_dev_put(&rt6_ex->rt6i->dst);
1446 
1447 	hlist_del_rcu(&rt6_ex->hlist);
1448 	dst_release(&rt6_ex->rt6i->dst);
1449 	kfree_rcu(rt6_ex, rcu);
1450 	WARN_ON_ONCE(!bucket->depth);
1451 	bucket->depth--;
1452 }
1453 
1454 /* Remove oldest rt6_ex in bucket and free the memory
1455  * Caller must hold rt6_exception_lock
1456  */
1457 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1458 {
1459 	struct rt6_exception *rt6_ex, *oldest = NULL;
1460 
1461 	if (!bucket)
1462 		return;
1463 
1464 	hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1465 		if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1466 			oldest = rt6_ex;
1467 	}
1468 	rt6_remove_exception(bucket, oldest);
1469 }
1470 
1471 static u32 rt6_exception_hash(const struct in6_addr *dst,
1472 			      const struct in6_addr *src)
1473 {
1474 	static u32 seed __read_mostly;
1475 	u32 val;
1476 
1477 	net_get_random_once(&seed, sizeof(seed));
1478 	val = jhash(dst, sizeof(*dst), seed);
1479 
1480 #ifdef CONFIG_IPV6_SUBTREES
1481 	if (src)
1482 		val = jhash(src, sizeof(*src), val);
1483 #endif
1484 	return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1485 }
1486 
1487 /* Helper function to find the cached rt in the hash table
1488  * and update bucket pointer to point to the bucket for this
1489  * (daddr, saddr) pair
1490  * Caller must hold rt6_exception_lock
1491  */
1492 static struct rt6_exception *
1493 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1494 			      const struct in6_addr *daddr,
1495 			      const struct in6_addr *saddr)
1496 {
1497 	struct rt6_exception *rt6_ex;
1498 	u32 hval;
1499 
1500 	if (!(*bucket) || !daddr)
1501 		return NULL;
1502 
1503 	hval = rt6_exception_hash(daddr, saddr);
1504 	*bucket += hval;
1505 
1506 	hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1507 		struct rt6_info *rt6 = rt6_ex->rt6i;
1508 		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1509 
1510 #ifdef CONFIG_IPV6_SUBTREES
1511 		if (matched && saddr)
1512 			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1513 #endif
1514 		if (matched)
1515 			return rt6_ex;
1516 	}
1517 	return NULL;
1518 }
1519 
1520 /* Helper function to find the cached rt in the hash table
1521  * and update bucket pointer to point to the bucket for this
1522  * (daddr, saddr) pair
1523  * Caller must hold rcu_read_lock()
1524  */
1525 static struct rt6_exception *
1526 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1527 			 const struct in6_addr *daddr,
1528 			 const struct in6_addr *saddr)
1529 {
1530 	struct rt6_exception *rt6_ex;
1531 	u32 hval;
1532 
1533 	WARN_ON_ONCE(!rcu_read_lock_held());
1534 
1535 	if (!(*bucket) || !daddr)
1536 		return NULL;
1537 
1538 	hval = rt6_exception_hash(daddr, saddr);
1539 	*bucket += hval;
1540 
1541 	hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1542 		struct rt6_info *rt6 = rt6_ex->rt6i;
1543 		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1544 
1545 #ifdef CONFIG_IPV6_SUBTREES
1546 		if (matched && saddr)
1547 			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1548 #endif
1549 		if (matched)
1550 			return rt6_ex;
1551 	}
1552 	return NULL;
1553 }
1554 
1555 static unsigned int fib6_mtu(const struct fib6_result *res)
1556 {
1557 	const struct fib6_nh *nh = res->nh;
1558 	unsigned int mtu;
1559 
1560 	if (res->f6i->fib6_pmtu) {
1561 		mtu = res->f6i->fib6_pmtu;
1562 	} else {
1563 		struct net_device *dev = nh->fib_nh_dev;
1564 		struct inet6_dev *idev;
1565 
1566 		rcu_read_lock();
1567 		idev = __in6_dev_get(dev);
1568 		mtu = idev->cnf.mtu6;
1569 		rcu_read_unlock();
1570 	}
1571 
1572 	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1573 
1574 	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1575 }
1576 
1577 #define FIB6_EXCEPTION_BUCKET_FLUSHED  0x1UL
1578 
1579 /* used when the flushed bit is not relevant, only access to the bucket
1580  * (ie., all bucket users except rt6_insert_exception);
1581  *
1582  * called under rcu lock; sometimes called with rt6_exception_lock held
1583  */
1584 static
1585 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1586 						       spinlock_t *lock)
1587 {
1588 	struct rt6_exception_bucket *bucket;
1589 
1590 	if (lock)
1591 		bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1592 						   lockdep_is_held(lock));
1593 	else
1594 		bucket = rcu_dereference(nh->rt6i_exception_bucket);
1595 
1596 	/* remove bucket flushed bit if set */
1597 	if (bucket) {
1598 		unsigned long p = (unsigned long)bucket;
1599 
1600 		p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1601 		bucket = (struct rt6_exception_bucket *)p;
1602 	}
1603 
1604 	return bucket;
1605 }
1606 
1607 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1608 {
1609 	unsigned long p = (unsigned long)bucket;
1610 
1611 	return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1612 }
1613 
1614 /* called with rt6_exception_lock held */
1615 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1616 					      spinlock_t *lock)
1617 {
1618 	struct rt6_exception_bucket *bucket;
1619 	unsigned long p;
1620 
1621 	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1622 					   lockdep_is_held(lock));
1623 
1624 	p = (unsigned long)bucket;
1625 	p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1626 	bucket = (struct rt6_exception_bucket *)p;
1627 	rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1628 }
1629 
1630 static int rt6_insert_exception(struct rt6_info *nrt,
1631 				const struct fib6_result *res)
1632 {
1633 	struct net *net = dev_net(nrt->dst.dev);
1634 	struct rt6_exception_bucket *bucket;
1635 	struct fib6_info *f6i = res->f6i;
1636 	struct in6_addr *src_key = NULL;
1637 	struct rt6_exception *rt6_ex;
1638 	struct fib6_nh *nh = res->nh;
1639 	int err = 0;
1640 
1641 	spin_lock_bh(&rt6_exception_lock);
1642 
1643 	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1644 					  lockdep_is_held(&rt6_exception_lock));
1645 	if (!bucket) {
1646 		bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1647 				 GFP_ATOMIC);
1648 		if (!bucket) {
1649 			err = -ENOMEM;
1650 			goto out;
1651 		}
1652 		rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1653 	} else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1654 		err = -EINVAL;
1655 		goto out;
1656 	}
1657 
1658 #ifdef CONFIG_IPV6_SUBTREES
1659 	/* fib6_src.plen != 0 indicates f6i is in subtree
1660 	 * and exception table is indexed by a hash of
1661 	 * both fib6_dst and fib6_src.
1662 	 * Otherwise, the exception table is indexed by
1663 	 * a hash of only fib6_dst.
1664 	 */
1665 	if (f6i->fib6_src.plen)
1666 		src_key = &nrt->rt6i_src.addr;
1667 #endif
1668 	/* rt6_mtu_change() might lower mtu on f6i.
1669 	 * Only insert this exception route if its mtu
1670 	 * is less than f6i's mtu value.
1671 	 */
1672 	if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1673 		err = -EINVAL;
1674 		goto out;
1675 	}
1676 
1677 	rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1678 					       src_key);
1679 	if (rt6_ex)
1680 		rt6_remove_exception(bucket, rt6_ex);
1681 
1682 	rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1683 	if (!rt6_ex) {
1684 		err = -ENOMEM;
1685 		goto out;
1686 	}
1687 	rt6_ex->rt6i = nrt;
1688 	rt6_ex->stamp = jiffies;
1689 	hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1690 	bucket->depth++;
1691 	net->ipv6.rt6_stats->fib_rt_cache++;
1692 
1693 	if (bucket->depth > FIB6_MAX_DEPTH)
1694 		rt6_exception_remove_oldest(bucket);
1695 
1696 out:
1697 	spin_unlock_bh(&rt6_exception_lock);
1698 
1699 	/* Update fn->fn_sernum to invalidate all cached dst */
1700 	if (!err) {
1701 		spin_lock_bh(&f6i->fib6_table->tb6_lock);
1702 		fib6_update_sernum(net, f6i);
1703 		spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1704 		fib6_force_start_gc(net);
1705 	}
1706 
1707 	return err;
1708 }
1709 
1710 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1711 {
1712 	struct rt6_exception_bucket *bucket;
1713 	struct rt6_exception *rt6_ex;
1714 	struct hlist_node *tmp;
1715 	int i;
1716 
1717 	spin_lock_bh(&rt6_exception_lock);
1718 
1719 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1720 	if (!bucket)
1721 		goto out;
1722 
1723 	/* Prevent rt6_insert_exception() to recreate the bucket list */
1724 	if (!from)
1725 		fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1726 
1727 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1728 		hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1729 			if (!from ||
1730 			    rcu_access_pointer(rt6_ex->rt6i->from) == from)
1731 				rt6_remove_exception(bucket, rt6_ex);
1732 		}
1733 		WARN_ON_ONCE(!from && bucket->depth);
1734 		bucket++;
1735 	}
1736 out:
1737 	spin_unlock_bh(&rt6_exception_lock);
1738 }
1739 
1740 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1741 {
1742 	struct fib6_info *f6i = arg;
1743 
1744 	fib6_nh_flush_exceptions(nh, f6i);
1745 
1746 	return 0;
1747 }
1748 
1749 void rt6_flush_exceptions(struct fib6_info *f6i)
1750 {
1751 	if (f6i->nh)
1752 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1753 					 f6i);
1754 	else
1755 		fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1756 }
1757 
1758 /* Find cached rt in the hash table inside passed in rt
1759  * Caller has to hold rcu_read_lock()
1760  */
1761 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1762 					   const struct in6_addr *daddr,
1763 					   const struct in6_addr *saddr)
1764 {
1765 	const struct in6_addr *src_key = NULL;
1766 	struct rt6_exception_bucket *bucket;
1767 	struct rt6_exception *rt6_ex;
1768 	struct rt6_info *ret = NULL;
1769 
1770 #ifdef CONFIG_IPV6_SUBTREES
1771 	/* fib6i_src.plen != 0 indicates f6i is in subtree
1772 	 * and exception table is indexed by a hash of
1773 	 * both fib6_dst and fib6_src.
1774 	 * However, the src addr used to create the hash
1775 	 * might not be exactly the passed in saddr which
1776 	 * is a /128 addr from the flow.
1777 	 * So we need to use f6i->fib6_src to redo lookup
1778 	 * if the passed in saddr does not find anything.
1779 	 * (See the logic in ip6_rt_cache_alloc() on how
1780 	 * rt->rt6i_src is updated.)
1781 	 */
1782 	if (res->f6i->fib6_src.plen)
1783 		src_key = saddr;
1784 find_ex:
1785 #endif
1786 	bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1787 	rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1788 
1789 	if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1790 		ret = rt6_ex->rt6i;
1791 
1792 #ifdef CONFIG_IPV6_SUBTREES
1793 	/* Use fib6_src as src_key and redo lookup */
1794 	if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1795 		src_key = &res->f6i->fib6_src.addr;
1796 		goto find_ex;
1797 	}
1798 #endif
1799 
1800 	return ret;
1801 }
1802 
1803 /* Remove the passed in cached rt from the hash table that contains it */
1804 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1805 				    const struct rt6_info *rt)
1806 {
1807 	const struct in6_addr *src_key = NULL;
1808 	struct rt6_exception_bucket *bucket;
1809 	struct rt6_exception *rt6_ex;
1810 	int err;
1811 
1812 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1813 		return -ENOENT;
1814 
1815 	spin_lock_bh(&rt6_exception_lock);
1816 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1817 
1818 #ifdef CONFIG_IPV6_SUBTREES
1819 	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1820 	 * and exception table is indexed by a hash of
1821 	 * both rt6i_dst and rt6i_src.
1822 	 * Otherwise, the exception table is indexed by
1823 	 * a hash of only rt6i_dst.
1824 	 */
1825 	if (plen)
1826 		src_key = &rt->rt6i_src.addr;
1827 #endif
1828 	rt6_ex = __rt6_find_exception_spinlock(&bucket,
1829 					       &rt->rt6i_dst.addr,
1830 					       src_key);
1831 	if (rt6_ex) {
1832 		rt6_remove_exception(bucket, rt6_ex);
1833 		err = 0;
1834 	} else {
1835 		err = -ENOENT;
1836 	}
1837 
1838 	spin_unlock_bh(&rt6_exception_lock);
1839 	return err;
1840 }
1841 
1842 struct fib6_nh_excptn_arg {
1843 	struct rt6_info	*rt;
1844 	int		plen;
1845 };
1846 
1847 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1848 {
1849 	struct fib6_nh_excptn_arg *arg = _arg;
1850 	int err;
1851 
1852 	err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1853 	if (err == 0)
1854 		return 1;
1855 
1856 	return 0;
1857 }
1858 
1859 static int rt6_remove_exception_rt(struct rt6_info *rt)
1860 {
1861 	struct fib6_info *from;
1862 
1863 	from = rcu_dereference(rt->from);
1864 	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1865 		return -EINVAL;
1866 
1867 	if (from->nh) {
1868 		struct fib6_nh_excptn_arg arg = {
1869 			.rt = rt,
1870 			.plen = from->fib6_src.plen
1871 		};
1872 		int rc;
1873 
1874 		/* rc = 1 means an entry was found */
1875 		rc = nexthop_for_each_fib6_nh(from->nh,
1876 					      rt6_nh_remove_exception_rt,
1877 					      &arg);
1878 		return rc ? 0 : -ENOENT;
1879 	}
1880 
1881 	return fib6_nh_remove_exception(from->fib6_nh,
1882 					from->fib6_src.plen, rt);
1883 }
1884 
1885 /* Find rt6_ex which contains the passed in rt cache and
1886  * refresh its stamp
1887  */
1888 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1889 				     const struct rt6_info *rt)
1890 {
1891 	const struct in6_addr *src_key = NULL;
1892 	struct rt6_exception_bucket *bucket;
1893 	struct rt6_exception *rt6_ex;
1894 
1895 	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1896 #ifdef CONFIG_IPV6_SUBTREES
1897 	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1898 	 * and exception table is indexed by a hash of
1899 	 * both rt6i_dst and rt6i_src.
1900 	 * Otherwise, the exception table is indexed by
1901 	 * a hash of only rt6i_dst.
1902 	 */
1903 	if (plen)
1904 		src_key = &rt->rt6i_src.addr;
1905 #endif
1906 	rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1907 	if (rt6_ex)
1908 		rt6_ex->stamp = jiffies;
1909 }
1910 
1911 struct fib6_nh_match_arg {
1912 	const struct net_device *dev;
1913 	const struct in6_addr	*gw;
1914 	struct fib6_nh		*match;
1915 };
1916 
1917 /* determine if fib6_nh has given device and gateway */
1918 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1919 {
1920 	struct fib6_nh_match_arg *arg = _arg;
1921 
1922 	if (arg->dev != nh->fib_nh_dev ||
1923 	    (arg->gw && !nh->fib_nh_gw_family) ||
1924 	    (!arg->gw && nh->fib_nh_gw_family) ||
1925 	    (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1926 		return 0;
1927 
1928 	arg->match = nh;
1929 
1930 	/* found a match, break the loop */
1931 	return 1;
1932 }
1933 
1934 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1935 {
1936 	struct fib6_info *from;
1937 	struct fib6_nh *fib6_nh;
1938 
1939 	rcu_read_lock();
1940 
1941 	from = rcu_dereference(rt->from);
1942 	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1943 		goto unlock;
1944 
1945 	if (from->nh) {
1946 		struct fib6_nh_match_arg arg = {
1947 			.dev = rt->dst.dev,
1948 			.gw = &rt->rt6i_gateway,
1949 		};
1950 
1951 		nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1952 
1953 		if (!arg.match)
1954 			goto unlock;
1955 		fib6_nh = arg.match;
1956 	} else {
1957 		fib6_nh = from->fib6_nh;
1958 	}
1959 	fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1960 unlock:
1961 	rcu_read_unlock();
1962 }
1963 
1964 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1965 					 struct rt6_info *rt, int mtu)
1966 {
1967 	/* If the new MTU is lower than the route PMTU, this new MTU will be the
1968 	 * lowest MTU in the path: always allow updating the route PMTU to
1969 	 * reflect PMTU decreases.
1970 	 *
1971 	 * If the new MTU is higher, and the route PMTU is equal to the local
1972 	 * MTU, this means the old MTU is the lowest in the path, so allow
1973 	 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1974 	 * handle this.
1975 	 */
1976 
1977 	if (dst_mtu(&rt->dst) >= mtu)
1978 		return true;
1979 
1980 	if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1981 		return true;
1982 
1983 	return false;
1984 }
1985 
1986 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
1987 				       const struct fib6_nh *nh, int mtu)
1988 {
1989 	struct rt6_exception_bucket *bucket;
1990 	struct rt6_exception *rt6_ex;
1991 	int i;
1992 
1993 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1994 	if (!bucket)
1995 		return;
1996 
1997 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1998 		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1999 			struct rt6_info *entry = rt6_ex->rt6i;
2000 
2001 			/* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2002 			 * route), the metrics of its rt->from have already
2003 			 * been updated.
2004 			 */
2005 			if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2006 			    rt6_mtu_change_route_allowed(idev, entry, mtu))
2007 				dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2008 		}
2009 		bucket++;
2010 	}
2011 }
2012 
2013 #define RTF_CACHE_GATEWAY	(RTF_GATEWAY | RTF_CACHE)
2014 
2015 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2016 					    const struct in6_addr *gateway)
2017 {
2018 	struct rt6_exception_bucket *bucket;
2019 	struct rt6_exception *rt6_ex;
2020 	struct hlist_node *tmp;
2021 	int i;
2022 
2023 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2024 		return;
2025 
2026 	spin_lock_bh(&rt6_exception_lock);
2027 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2028 	if (bucket) {
2029 		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2030 			hlist_for_each_entry_safe(rt6_ex, tmp,
2031 						  &bucket->chain, hlist) {
2032 				struct rt6_info *entry = rt6_ex->rt6i;
2033 
2034 				if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2035 				    RTF_CACHE_GATEWAY &&
2036 				    ipv6_addr_equal(gateway,
2037 						    &entry->rt6i_gateway)) {
2038 					rt6_remove_exception(bucket, rt6_ex);
2039 				}
2040 			}
2041 			bucket++;
2042 		}
2043 	}
2044 
2045 	spin_unlock_bh(&rt6_exception_lock);
2046 }
2047 
2048 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2049 				      struct rt6_exception *rt6_ex,
2050 				      struct fib6_gc_args *gc_args,
2051 				      unsigned long now)
2052 {
2053 	struct rt6_info *rt = rt6_ex->rt6i;
2054 
2055 	/* we are pruning and obsoleting aged-out and non gateway exceptions
2056 	 * even if others have still references to them, so that on next
2057 	 * dst_check() such references can be dropped.
2058 	 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2059 	 * expired, independently from their aging, as per RFC 8201 section 4
2060 	 */
2061 	if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2062 		if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2063 			RT6_TRACE("aging clone %p\n", rt);
2064 			rt6_remove_exception(bucket, rt6_ex);
2065 			return;
2066 		}
2067 	} else if (time_after(jiffies, rt->dst.expires)) {
2068 		RT6_TRACE("purging expired route %p\n", rt);
2069 		rt6_remove_exception(bucket, rt6_ex);
2070 		return;
2071 	}
2072 
2073 	if (rt->rt6i_flags & RTF_GATEWAY) {
2074 		struct neighbour *neigh;
2075 		__u8 neigh_flags = 0;
2076 
2077 		neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2078 		if (neigh)
2079 			neigh_flags = neigh->flags;
2080 
2081 		if (!(neigh_flags & NTF_ROUTER)) {
2082 			RT6_TRACE("purging route %p via non-router but gateway\n",
2083 				  rt);
2084 			rt6_remove_exception(bucket, rt6_ex);
2085 			return;
2086 		}
2087 	}
2088 
2089 	gc_args->more++;
2090 }
2091 
2092 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2093 				   struct fib6_gc_args *gc_args,
2094 				   unsigned long now)
2095 {
2096 	struct rt6_exception_bucket *bucket;
2097 	struct rt6_exception *rt6_ex;
2098 	struct hlist_node *tmp;
2099 	int i;
2100 
2101 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2102 		return;
2103 
2104 	rcu_read_lock_bh();
2105 	spin_lock(&rt6_exception_lock);
2106 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2107 	if (bucket) {
2108 		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2109 			hlist_for_each_entry_safe(rt6_ex, tmp,
2110 						  &bucket->chain, hlist) {
2111 				rt6_age_examine_exception(bucket, rt6_ex,
2112 							  gc_args, now);
2113 			}
2114 			bucket++;
2115 		}
2116 	}
2117 	spin_unlock(&rt6_exception_lock);
2118 	rcu_read_unlock_bh();
2119 }
2120 
2121 struct fib6_nh_age_excptn_arg {
2122 	struct fib6_gc_args	*gc_args;
2123 	unsigned long		now;
2124 };
2125 
2126 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2127 {
2128 	struct fib6_nh_age_excptn_arg *arg = _arg;
2129 
2130 	fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2131 	return 0;
2132 }
2133 
2134 void rt6_age_exceptions(struct fib6_info *f6i,
2135 			struct fib6_gc_args *gc_args,
2136 			unsigned long now)
2137 {
2138 	if (f6i->nh) {
2139 		struct fib6_nh_age_excptn_arg arg = {
2140 			.gc_args = gc_args,
2141 			.now = now
2142 		};
2143 
2144 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2145 					 &arg);
2146 	} else {
2147 		fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2148 	}
2149 }
2150 
2151 /* must be called with rcu lock held */
2152 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2153 		      struct flowi6 *fl6, struct fib6_result *res, int strict)
2154 {
2155 	struct fib6_node *fn, *saved_fn;
2156 
2157 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2158 	saved_fn = fn;
2159 
2160 	if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2161 		oif = 0;
2162 
2163 redo_rt6_select:
2164 	rt6_select(net, fn, oif, res, strict);
2165 	if (res->f6i == net->ipv6.fib6_null_entry) {
2166 		fn = fib6_backtrack(fn, &fl6->saddr);
2167 		if (fn)
2168 			goto redo_rt6_select;
2169 		else if (strict & RT6_LOOKUP_F_REACHABLE) {
2170 			/* also consider unreachable route */
2171 			strict &= ~RT6_LOOKUP_F_REACHABLE;
2172 			fn = saved_fn;
2173 			goto redo_rt6_select;
2174 		}
2175 	}
2176 
2177 	trace_fib6_table_lookup(net, res, table, fl6);
2178 
2179 	return 0;
2180 }
2181 
2182 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2183 			       int oif, struct flowi6 *fl6,
2184 			       const struct sk_buff *skb, int flags)
2185 {
2186 	struct fib6_result res = {};
2187 	struct rt6_info *rt = NULL;
2188 	int strict = 0;
2189 
2190 	WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2191 		     !rcu_read_lock_held());
2192 
2193 	strict |= flags & RT6_LOOKUP_F_IFACE;
2194 	strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2195 	if (net->ipv6.devconf_all->forwarding == 0)
2196 		strict |= RT6_LOOKUP_F_REACHABLE;
2197 
2198 	rcu_read_lock();
2199 
2200 	fib6_table_lookup(net, table, oif, fl6, &res, strict);
2201 	if (res.f6i == net->ipv6.fib6_null_entry)
2202 		goto out;
2203 
2204 	fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2205 
2206 	/*Search through exception table */
2207 	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2208 	if (rt) {
2209 		goto out;
2210 	} else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2211 			    !res.nh->fib_nh_gw_family)) {
2212 		/* Create a RTF_CACHE clone which will not be
2213 		 * owned by the fib6 tree.  It is for the special case where
2214 		 * the daddr in the skb during the neighbor look-up is different
2215 		 * from the fl6->daddr used to look-up route here.
2216 		 */
2217 		rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2218 
2219 		if (rt) {
2220 			/* 1 refcnt is taken during ip6_rt_cache_alloc().
2221 			 * As rt6_uncached_list_add() does not consume refcnt,
2222 			 * this refcnt is always returned to the caller even
2223 			 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2224 			 */
2225 			rt6_uncached_list_add(rt);
2226 			atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2227 			rcu_read_unlock();
2228 
2229 			return rt;
2230 		}
2231 	} else {
2232 		/* Get a percpu copy */
2233 		local_bh_disable();
2234 		rt = rt6_get_pcpu_route(&res);
2235 
2236 		if (!rt)
2237 			rt = rt6_make_pcpu_route(net, &res);
2238 
2239 		local_bh_enable();
2240 	}
2241 out:
2242 	if (!rt)
2243 		rt = net->ipv6.ip6_null_entry;
2244 	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2245 		ip6_hold_safe(net, &rt);
2246 	rcu_read_unlock();
2247 
2248 	return rt;
2249 }
2250 EXPORT_SYMBOL_GPL(ip6_pol_route);
2251 
2252 static struct rt6_info *ip6_pol_route_input(struct net *net,
2253 					    struct fib6_table *table,
2254 					    struct flowi6 *fl6,
2255 					    const struct sk_buff *skb,
2256 					    int flags)
2257 {
2258 	return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2259 }
2260 
2261 struct dst_entry *ip6_route_input_lookup(struct net *net,
2262 					 struct net_device *dev,
2263 					 struct flowi6 *fl6,
2264 					 const struct sk_buff *skb,
2265 					 int flags)
2266 {
2267 	if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2268 		flags |= RT6_LOOKUP_F_IFACE;
2269 
2270 	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2271 }
2272 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2273 
2274 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2275 				  struct flow_keys *keys,
2276 				  struct flow_keys *flkeys)
2277 {
2278 	const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2279 	const struct ipv6hdr *key_iph = outer_iph;
2280 	struct flow_keys *_flkeys = flkeys;
2281 	const struct ipv6hdr *inner_iph;
2282 	const struct icmp6hdr *icmph;
2283 	struct ipv6hdr _inner_iph;
2284 	struct icmp6hdr _icmph;
2285 
2286 	if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2287 		goto out;
2288 
2289 	icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2290 				   sizeof(_icmph), &_icmph);
2291 	if (!icmph)
2292 		goto out;
2293 
2294 	if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
2295 	    icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
2296 	    icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
2297 	    icmph->icmp6_type != ICMPV6_PARAMPROB)
2298 		goto out;
2299 
2300 	inner_iph = skb_header_pointer(skb,
2301 				       skb_transport_offset(skb) + sizeof(*icmph),
2302 				       sizeof(_inner_iph), &_inner_iph);
2303 	if (!inner_iph)
2304 		goto out;
2305 
2306 	key_iph = inner_iph;
2307 	_flkeys = NULL;
2308 out:
2309 	if (_flkeys) {
2310 		keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2311 		keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2312 		keys->tags.flow_label = _flkeys->tags.flow_label;
2313 		keys->basic.ip_proto = _flkeys->basic.ip_proto;
2314 	} else {
2315 		keys->addrs.v6addrs.src = key_iph->saddr;
2316 		keys->addrs.v6addrs.dst = key_iph->daddr;
2317 		keys->tags.flow_label = ip6_flowlabel(key_iph);
2318 		keys->basic.ip_proto = key_iph->nexthdr;
2319 	}
2320 }
2321 
2322 /* if skb is set it will be used and fl6 can be NULL */
2323 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2324 		       const struct sk_buff *skb, struct flow_keys *flkeys)
2325 {
2326 	struct flow_keys hash_keys;
2327 	u32 mhash;
2328 
2329 	switch (ip6_multipath_hash_policy(net)) {
2330 	case 0:
2331 		memset(&hash_keys, 0, sizeof(hash_keys));
2332 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2333 		if (skb) {
2334 			ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2335 		} else {
2336 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2337 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2338 			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2339 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2340 		}
2341 		break;
2342 	case 1:
2343 		if (skb) {
2344 			unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2345 			struct flow_keys keys;
2346 
2347 			/* short-circuit if we already have L4 hash present */
2348 			if (skb->l4_hash)
2349 				return skb_get_hash_raw(skb) >> 1;
2350 
2351 			memset(&hash_keys, 0, sizeof(hash_keys));
2352 
2353                         if (!flkeys) {
2354 				skb_flow_dissect_flow_keys(skb, &keys, flag);
2355 				flkeys = &keys;
2356 			}
2357 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2358 			hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2359 			hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2360 			hash_keys.ports.src = flkeys->ports.src;
2361 			hash_keys.ports.dst = flkeys->ports.dst;
2362 			hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2363 		} else {
2364 			memset(&hash_keys, 0, sizeof(hash_keys));
2365 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2366 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2367 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2368 			hash_keys.ports.src = fl6->fl6_sport;
2369 			hash_keys.ports.dst = fl6->fl6_dport;
2370 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2371 		}
2372 		break;
2373 	case 2:
2374 		memset(&hash_keys, 0, sizeof(hash_keys));
2375 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2376 		if (skb) {
2377 			struct flow_keys keys;
2378 
2379 			if (!flkeys) {
2380 				skb_flow_dissect_flow_keys(skb, &keys, 0);
2381 				flkeys = &keys;
2382 			}
2383 
2384 			/* Inner can be v4 or v6 */
2385 			if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2386 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2387 				hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2388 				hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2389 			} else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2390 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2391 				hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2392 				hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2393 				hash_keys.tags.flow_label = flkeys->tags.flow_label;
2394 				hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2395 			} else {
2396 				/* Same as case 0 */
2397 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2398 				ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2399 			}
2400 		} else {
2401 			/* Same as case 0 */
2402 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2403 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2404 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2405 			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2406 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2407 		}
2408 		break;
2409 	}
2410 	mhash = flow_hash_from_keys(&hash_keys);
2411 
2412 	return mhash >> 1;
2413 }
2414 
2415 /* Called with rcu held */
2416 void ip6_route_input(struct sk_buff *skb)
2417 {
2418 	const struct ipv6hdr *iph = ipv6_hdr(skb);
2419 	struct net *net = dev_net(skb->dev);
2420 	int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2421 	struct ip_tunnel_info *tun_info;
2422 	struct flowi6 fl6 = {
2423 		.flowi6_iif = skb->dev->ifindex,
2424 		.daddr = iph->daddr,
2425 		.saddr = iph->saddr,
2426 		.flowlabel = ip6_flowinfo(iph),
2427 		.flowi6_mark = skb->mark,
2428 		.flowi6_proto = iph->nexthdr,
2429 	};
2430 	struct flow_keys *flkeys = NULL, _flkeys;
2431 
2432 	tun_info = skb_tunnel_info(skb);
2433 	if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2434 		fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2435 
2436 	if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2437 		flkeys = &_flkeys;
2438 
2439 	if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2440 		fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2441 	skb_dst_drop(skb);
2442 	skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2443 						      &fl6, skb, flags));
2444 }
2445 
2446 static struct rt6_info *ip6_pol_route_output(struct net *net,
2447 					     struct fib6_table *table,
2448 					     struct flowi6 *fl6,
2449 					     const struct sk_buff *skb,
2450 					     int flags)
2451 {
2452 	return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2453 }
2454 
2455 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2456 					       const struct sock *sk,
2457 					       struct flowi6 *fl6, int flags)
2458 {
2459 	bool any_src;
2460 
2461 	if (ipv6_addr_type(&fl6->daddr) &
2462 	    (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2463 		struct dst_entry *dst;
2464 
2465 		/* This function does not take refcnt on the dst */
2466 		dst = l3mdev_link_scope_lookup(net, fl6);
2467 		if (dst)
2468 			return dst;
2469 	}
2470 
2471 	fl6->flowi6_iif = LOOPBACK_IFINDEX;
2472 
2473 	flags |= RT6_LOOKUP_F_DST_NOREF;
2474 	any_src = ipv6_addr_any(&fl6->saddr);
2475 	if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2476 	    (fl6->flowi6_oif && any_src))
2477 		flags |= RT6_LOOKUP_F_IFACE;
2478 
2479 	if (!any_src)
2480 		flags |= RT6_LOOKUP_F_HAS_SADDR;
2481 	else if (sk)
2482 		flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2483 
2484 	return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2485 }
2486 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2487 
2488 struct dst_entry *ip6_route_output_flags(struct net *net,
2489 					 const struct sock *sk,
2490 					 struct flowi6 *fl6,
2491 					 int flags)
2492 {
2493         struct dst_entry *dst;
2494         struct rt6_info *rt6;
2495 
2496         rcu_read_lock();
2497         dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2498         rt6 = (struct rt6_info *)dst;
2499         /* For dst cached in uncached_list, refcnt is already taken. */
2500         if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2501                 dst = &net->ipv6.ip6_null_entry->dst;
2502                 dst_hold(dst);
2503         }
2504         rcu_read_unlock();
2505 
2506         return dst;
2507 }
2508 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2509 
2510 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2511 {
2512 	struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2513 	struct net_device *loopback_dev = net->loopback_dev;
2514 	struct dst_entry *new = NULL;
2515 
2516 	rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2517 		       DST_OBSOLETE_DEAD, 0);
2518 	if (rt) {
2519 		rt6_info_init(rt);
2520 		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2521 
2522 		new = &rt->dst;
2523 		new->__use = 1;
2524 		new->input = dst_discard;
2525 		new->output = dst_discard_out;
2526 
2527 		dst_copy_metrics(new, &ort->dst);
2528 
2529 		rt->rt6i_idev = in6_dev_get(loopback_dev);
2530 		rt->rt6i_gateway = ort->rt6i_gateway;
2531 		rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2532 
2533 		memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2534 #ifdef CONFIG_IPV6_SUBTREES
2535 		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2536 #endif
2537 	}
2538 
2539 	dst_release(dst_orig);
2540 	return new ? new : ERR_PTR(-ENOMEM);
2541 }
2542 
2543 /*
2544  *	Destination cache support functions
2545  */
2546 
2547 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2548 {
2549 	u32 rt_cookie = 0;
2550 
2551 	if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2552 		return false;
2553 
2554 	if (fib6_check_expired(f6i))
2555 		return false;
2556 
2557 	return true;
2558 }
2559 
2560 static struct dst_entry *rt6_check(struct rt6_info *rt,
2561 				   struct fib6_info *from,
2562 				   u32 cookie)
2563 {
2564 	u32 rt_cookie = 0;
2565 
2566 	if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2567 	    rt_cookie != cookie)
2568 		return NULL;
2569 
2570 	if (rt6_check_expired(rt))
2571 		return NULL;
2572 
2573 	return &rt->dst;
2574 }
2575 
2576 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2577 					    struct fib6_info *from,
2578 					    u32 cookie)
2579 {
2580 	if (!__rt6_check_expired(rt) &&
2581 	    rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2582 	    fib6_check(from, cookie))
2583 		return &rt->dst;
2584 	else
2585 		return NULL;
2586 }
2587 
2588 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2589 {
2590 	struct dst_entry *dst_ret;
2591 	struct fib6_info *from;
2592 	struct rt6_info *rt;
2593 
2594 	rt = container_of(dst, struct rt6_info, dst);
2595 
2596 	rcu_read_lock();
2597 
2598 	/* All IPV6 dsts are created with ->obsolete set to the value
2599 	 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2600 	 * into this function always.
2601 	 */
2602 
2603 	from = rcu_dereference(rt->from);
2604 
2605 	if (from && (rt->rt6i_flags & RTF_PCPU ||
2606 	    unlikely(!list_empty(&rt->rt6i_uncached))))
2607 		dst_ret = rt6_dst_from_check(rt, from, cookie);
2608 	else
2609 		dst_ret = rt6_check(rt, from, cookie);
2610 
2611 	rcu_read_unlock();
2612 
2613 	return dst_ret;
2614 }
2615 
2616 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2617 {
2618 	struct rt6_info *rt = (struct rt6_info *) dst;
2619 
2620 	if (rt) {
2621 		if (rt->rt6i_flags & RTF_CACHE) {
2622 			rcu_read_lock();
2623 			if (rt6_check_expired(rt)) {
2624 				rt6_remove_exception_rt(rt);
2625 				dst = NULL;
2626 			}
2627 			rcu_read_unlock();
2628 		} else {
2629 			dst_release(dst);
2630 			dst = NULL;
2631 		}
2632 	}
2633 	return dst;
2634 }
2635 
2636 static void ip6_link_failure(struct sk_buff *skb)
2637 {
2638 	struct rt6_info *rt;
2639 
2640 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2641 
2642 	rt = (struct rt6_info *) skb_dst(skb);
2643 	if (rt) {
2644 		rcu_read_lock();
2645 		if (rt->rt6i_flags & RTF_CACHE) {
2646 			rt6_remove_exception_rt(rt);
2647 		} else {
2648 			struct fib6_info *from;
2649 			struct fib6_node *fn;
2650 
2651 			from = rcu_dereference(rt->from);
2652 			if (from) {
2653 				fn = rcu_dereference(from->fib6_node);
2654 				if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2655 					fn->fn_sernum = -1;
2656 			}
2657 		}
2658 		rcu_read_unlock();
2659 	}
2660 }
2661 
2662 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2663 {
2664 	if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2665 		struct fib6_info *from;
2666 
2667 		rcu_read_lock();
2668 		from = rcu_dereference(rt0->from);
2669 		if (from)
2670 			rt0->dst.expires = from->expires;
2671 		rcu_read_unlock();
2672 	}
2673 
2674 	dst_set_expires(&rt0->dst, timeout);
2675 	rt0->rt6i_flags |= RTF_EXPIRES;
2676 }
2677 
2678 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2679 {
2680 	struct net *net = dev_net(rt->dst.dev);
2681 
2682 	dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2683 	rt->rt6i_flags |= RTF_MODIFIED;
2684 	rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2685 }
2686 
2687 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2688 {
2689 	return !(rt->rt6i_flags & RTF_CACHE) &&
2690 		(rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2691 }
2692 
2693 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2694 				 const struct ipv6hdr *iph, u32 mtu)
2695 {
2696 	const struct in6_addr *daddr, *saddr;
2697 	struct rt6_info *rt6 = (struct rt6_info *)dst;
2698 
2699 	if (dst_metric_locked(dst, RTAX_MTU))
2700 		return;
2701 
2702 	if (iph) {
2703 		daddr = &iph->daddr;
2704 		saddr = &iph->saddr;
2705 	} else if (sk) {
2706 		daddr = &sk->sk_v6_daddr;
2707 		saddr = &inet6_sk(sk)->saddr;
2708 	} else {
2709 		daddr = NULL;
2710 		saddr = NULL;
2711 	}
2712 	dst_confirm_neigh(dst, daddr);
2713 	mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2714 	if (mtu >= dst_mtu(dst))
2715 		return;
2716 
2717 	if (!rt6_cache_allowed_for_pmtu(rt6)) {
2718 		rt6_do_update_pmtu(rt6, mtu);
2719 		/* update rt6_ex->stamp for cache */
2720 		if (rt6->rt6i_flags & RTF_CACHE)
2721 			rt6_update_exception_stamp_rt(rt6);
2722 	} else if (daddr) {
2723 		struct fib6_result res = {};
2724 		struct rt6_info *nrt6;
2725 
2726 		rcu_read_lock();
2727 		res.f6i = rcu_dereference(rt6->from);
2728 		if (!res.f6i) {
2729 			rcu_read_unlock();
2730 			return;
2731 		}
2732 		res.fib6_flags = res.f6i->fib6_flags;
2733 		res.fib6_type = res.f6i->fib6_type;
2734 
2735 		if (res.f6i->nh) {
2736 			struct fib6_nh_match_arg arg = {
2737 				.dev = dst->dev,
2738 				.gw = &rt6->rt6i_gateway,
2739 			};
2740 
2741 			nexthop_for_each_fib6_nh(res.f6i->nh,
2742 						 fib6_nh_find_match, &arg);
2743 
2744 			/* fib6_info uses a nexthop that does not have fib6_nh
2745 			 * using the dst->dev + gw. Should be impossible.
2746 			 */
2747 			if (!arg.match) {
2748 				rcu_read_unlock();
2749 				return;
2750 			}
2751 
2752 			res.nh = arg.match;
2753 		} else {
2754 			res.nh = res.f6i->fib6_nh;
2755 		}
2756 
2757 		nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2758 		if (nrt6) {
2759 			rt6_do_update_pmtu(nrt6, mtu);
2760 			if (rt6_insert_exception(nrt6, &res))
2761 				dst_release_immediate(&nrt6->dst);
2762 		}
2763 		rcu_read_unlock();
2764 	}
2765 }
2766 
2767 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2768 			       struct sk_buff *skb, u32 mtu)
2769 {
2770 	__ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2771 }
2772 
2773 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2774 		     int oif, u32 mark, kuid_t uid)
2775 {
2776 	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2777 	struct dst_entry *dst;
2778 	struct flowi6 fl6 = {
2779 		.flowi6_oif = oif,
2780 		.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2781 		.daddr = iph->daddr,
2782 		.saddr = iph->saddr,
2783 		.flowlabel = ip6_flowinfo(iph),
2784 		.flowi6_uid = uid,
2785 	};
2786 
2787 	dst = ip6_route_output(net, NULL, &fl6);
2788 	if (!dst->error)
2789 		__ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
2790 	dst_release(dst);
2791 }
2792 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2793 
2794 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2795 {
2796 	int oif = sk->sk_bound_dev_if;
2797 	struct dst_entry *dst;
2798 
2799 	if (!oif && skb->dev)
2800 		oif = l3mdev_master_ifindex(skb->dev);
2801 
2802 	ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2803 
2804 	dst = __sk_dst_get(sk);
2805 	if (!dst || !dst->obsolete ||
2806 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2807 		return;
2808 
2809 	bh_lock_sock(sk);
2810 	if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2811 		ip6_datagram_dst_update(sk, false);
2812 	bh_unlock_sock(sk);
2813 }
2814 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2815 
2816 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2817 			   const struct flowi6 *fl6)
2818 {
2819 #ifdef CONFIG_IPV6_SUBTREES
2820 	struct ipv6_pinfo *np = inet6_sk(sk);
2821 #endif
2822 
2823 	ip6_dst_store(sk, dst,
2824 		      ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2825 		      &sk->sk_v6_daddr : NULL,
2826 #ifdef CONFIG_IPV6_SUBTREES
2827 		      ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2828 		      &np->saddr :
2829 #endif
2830 		      NULL);
2831 }
2832 
2833 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2834 				  struct flowi6 *fl6,
2835 				  const struct in6_addr *gw,
2836 				  struct rt6_info **ret)
2837 {
2838 	const struct fib6_nh *nh = res->nh;
2839 
2840 	if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2841 	    fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2842 		return false;
2843 
2844 	/* rt_cache's gateway might be different from its 'parent'
2845 	 * in the case of an ip redirect.
2846 	 * So we keep searching in the exception table if the gateway
2847 	 * is different.
2848 	 */
2849 	if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2850 		struct rt6_info *rt_cache;
2851 
2852 		rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
2853 		if (rt_cache &&
2854 		    ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2855 			*ret = rt_cache;
2856 			return true;
2857 		}
2858 		return false;
2859 	}
2860 	return true;
2861 }
2862 
2863 struct fib6_nh_rd_arg {
2864 	struct fib6_result	*res;
2865 	struct flowi6		*fl6;
2866 	const struct in6_addr	*gw;
2867 	struct rt6_info		**ret;
2868 };
2869 
2870 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2871 {
2872 	struct fib6_nh_rd_arg *arg = _arg;
2873 
2874 	arg->res->nh = nh;
2875 	return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2876 }
2877 
2878 /* Handle redirects */
2879 struct ip6rd_flowi {
2880 	struct flowi6 fl6;
2881 	struct in6_addr gateway;
2882 };
2883 
2884 static struct rt6_info *__ip6_route_redirect(struct net *net,
2885 					     struct fib6_table *table,
2886 					     struct flowi6 *fl6,
2887 					     const struct sk_buff *skb,
2888 					     int flags)
2889 {
2890 	struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
2891 	struct rt6_info *ret = NULL;
2892 	struct fib6_result res = {};
2893 	struct fib6_nh_rd_arg arg = {
2894 		.res = &res,
2895 		.fl6 = fl6,
2896 		.gw  = &rdfl->gateway,
2897 		.ret = &ret
2898 	};
2899 	struct fib6_info *rt;
2900 	struct fib6_node *fn;
2901 
2902 	/* l3mdev_update_flow overrides oif if the device is enslaved; in
2903 	 * this case we must match on the real ingress device, so reset it
2904 	 */
2905 	if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2906 		fl6->flowi6_oif = skb->dev->ifindex;
2907 
2908 	/* Get the "current" route for this destination and
2909 	 * check if the redirect has come from appropriate router.
2910 	 *
2911 	 * RFC 4861 specifies that redirects should only be
2912 	 * accepted if they come from the nexthop to the target.
2913 	 * Due to the way the routes are chosen, this notion
2914 	 * is a bit fuzzy and one might need to check all possible
2915 	 * routes.
2916 	 */
2917 
2918 	rcu_read_lock();
2919 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2920 restart:
2921 	for_each_fib6_node_rt_rcu(fn) {
2922 		res.f6i = rt;
2923 		if (fib6_check_expired(rt))
2924 			continue;
2925 		if (rt->fib6_flags & RTF_REJECT)
2926 			break;
2927 		if (unlikely(rt->nh)) {
2928 			if (nexthop_is_blackhole(rt->nh))
2929 				continue;
2930 			/* on match, res->nh is filled in and potentially ret */
2931 			if (nexthop_for_each_fib6_nh(rt->nh,
2932 						     fib6_nh_redirect_match,
2933 						     &arg))
2934 				goto out;
2935 		} else {
2936 			res.nh = rt->fib6_nh;
2937 			if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2938 						  &ret))
2939 				goto out;
2940 		}
2941 	}
2942 
2943 	if (!rt)
2944 		rt = net->ipv6.fib6_null_entry;
2945 	else if (rt->fib6_flags & RTF_REJECT) {
2946 		ret = net->ipv6.ip6_null_entry;
2947 		goto out;
2948 	}
2949 
2950 	if (rt == net->ipv6.fib6_null_entry) {
2951 		fn = fib6_backtrack(fn, &fl6->saddr);
2952 		if (fn)
2953 			goto restart;
2954 	}
2955 
2956 	res.f6i = rt;
2957 	res.nh = rt->fib6_nh;
2958 out:
2959 	if (ret) {
2960 		ip6_hold_safe(net, &ret);
2961 	} else {
2962 		res.fib6_flags = res.f6i->fib6_flags;
2963 		res.fib6_type = res.f6i->fib6_type;
2964 		ret = ip6_create_rt_rcu(&res);
2965 	}
2966 
2967 	rcu_read_unlock();
2968 
2969 	trace_fib6_table_lookup(net, &res, table, fl6);
2970 	return ret;
2971 };
2972 
2973 static struct dst_entry *ip6_route_redirect(struct net *net,
2974 					    const struct flowi6 *fl6,
2975 					    const struct sk_buff *skb,
2976 					    const struct in6_addr *gateway)
2977 {
2978 	int flags = RT6_LOOKUP_F_HAS_SADDR;
2979 	struct ip6rd_flowi rdfl;
2980 
2981 	rdfl.fl6 = *fl6;
2982 	rdfl.gateway = *gateway;
2983 
2984 	return fib6_rule_lookup(net, &rdfl.fl6, skb,
2985 				flags, __ip6_route_redirect);
2986 }
2987 
2988 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2989 		  kuid_t uid)
2990 {
2991 	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2992 	struct dst_entry *dst;
2993 	struct flowi6 fl6 = {
2994 		.flowi6_iif = LOOPBACK_IFINDEX,
2995 		.flowi6_oif = oif,
2996 		.flowi6_mark = mark,
2997 		.daddr = iph->daddr,
2998 		.saddr = iph->saddr,
2999 		.flowlabel = ip6_flowinfo(iph),
3000 		.flowi6_uid = uid,
3001 	};
3002 
3003 	dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3004 	rt6_do_redirect(dst, NULL, skb);
3005 	dst_release(dst);
3006 }
3007 EXPORT_SYMBOL_GPL(ip6_redirect);
3008 
3009 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3010 {
3011 	const struct ipv6hdr *iph = ipv6_hdr(skb);
3012 	const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3013 	struct dst_entry *dst;
3014 	struct flowi6 fl6 = {
3015 		.flowi6_iif = LOOPBACK_IFINDEX,
3016 		.flowi6_oif = oif,
3017 		.daddr = msg->dest,
3018 		.saddr = iph->daddr,
3019 		.flowi6_uid = sock_net_uid(net, NULL),
3020 	};
3021 
3022 	dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3023 	rt6_do_redirect(dst, NULL, skb);
3024 	dst_release(dst);
3025 }
3026 
3027 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3028 {
3029 	ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3030 		     sk->sk_uid);
3031 }
3032 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3033 
3034 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3035 {
3036 	struct net_device *dev = dst->dev;
3037 	unsigned int mtu = dst_mtu(dst);
3038 	struct net *net = dev_net(dev);
3039 
3040 	mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3041 
3042 	if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3043 		mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3044 
3045 	/*
3046 	 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3047 	 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3048 	 * IPV6_MAXPLEN is also valid and means: "any MSS,
3049 	 * rely only on pmtu discovery"
3050 	 */
3051 	if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3052 		mtu = IPV6_MAXPLEN;
3053 	return mtu;
3054 }
3055 
3056 static unsigned int ip6_mtu(const struct dst_entry *dst)
3057 {
3058 	struct inet6_dev *idev;
3059 	unsigned int mtu;
3060 
3061 	mtu = dst_metric_raw(dst, RTAX_MTU);
3062 	if (mtu)
3063 		goto out;
3064 
3065 	mtu = IPV6_MIN_MTU;
3066 
3067 	rcu_read_lock();
3068 	idev = __in6_dev_get(dst->dev);
3069 	if (idev)
3070 		mtu = idev->cnf.mtu6;
3071 	rcu_read_unlock();
3072 
3073 out:
3074 	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3075 
3076 	return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
3077 }
3078 
3079 /* MTU selection:
3080  * 1. mtu on route is locked - use it
3081  * 2. mtu from nexthop exception
3082  * 3. mtu from egress device
3083  *
3084  * based on ip6_dst_mtu_forward and exception logic of
3085  * rt6_find_cached_rt; called with rcu_read_lock
3086  */
3087 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3088 		      const struct in6_addr *daddr,
3089 		      const struct in6_addr *saddr)
3090 {
3091 	const struct fib6_nh *nh = res->nh;
3092 	struct fib6_info *f6i = res->f6i;
3093 	struct inet6_dev *idev;
3094 	struct rt6_info *rt;
3095 	u32 mtu = 0;
3096 
3097 	if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3098 		mtu = f6i->fib6_pmtu;
3099 		if (mtu)
3100 			goto out;
3101 	}
3102 
3103 	rt = rt6_find_cached_rt(res, daddr, saddr);
3104 	if (unlikely(rt)) {
3105 		mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3106 	} else {
3107 		struct net_device *dev = nh->fib_nh_dev;
3108 
3109 		mtu = IPV6_MIN_MTU;
3110 		idev = __in6_dev_get(dev);
3111 		if (idev && idev->cnf.mtu6 > mtu)
3112 			mtu = idev->cnf.mtu6;
3113 	}
3114 
3115 	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3116 out:
3117 	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3118 }
3119 
3120 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3121 				  struct flowi6 *fl6)
3122 {
3123 	struct dst_entry *dst;
3124 	struct rt6_info *rt;
3125 	struct inet6_dev *idev = in6_dev_get(dev);
3126 	struct net *net = dev_net(dev);
3127 
3128 	if (unlikely(!idev))
3129 		return ERR_PTR(-ENODEV);
3130 
3131 	rt = ip6_dst_alloc(net, dev, 0);
3132 	if (unlikely(!rt)) {
3133 		in6_dev_put(idev);
3134 		dst = ERR_PTR(-ENOMEM);
3135 		goto out;
3136 	}
3137 
3138 	rt->dst.flags |= DST_HOST;
3139 	rt->dst.input = ip6_input;
3140 	rt->dst.output  = ip6_output;
3141 	rt->rt6i_gateway  = fl6->daddr;
3142 	rt->rt6i_dst.addr = fl6->daddr;
3143 	rt->rt6i_dst.plen = 128;
3144 	rt->rt6i_idev     = idev;
3145 	dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3146 
3147 	/* Add this dst into uncached_list so that rt6_disable_ip() can
3148 	 * do proper release of the net_device
3149 	 */
3150 	rt6_uncached_list_add(rt);
3151 	atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
3152 
3153 	dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3154 
3155 out:
3156 	return dst;
3157 }
3158 
3159 static int ip6_dst_gc(struct dst_ops *ops)
3160 {
3161 	struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3162 	int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3163 	int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3164 	int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3165 	int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3166 	unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3167 	int entries;
3168 
3169 	entries = dst_entries_get_fast(ops);
3170 	if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3171 	    entries <= rt_max_size)
3172 		goto out;
3173 
3174 	net->ipv6.ip6_rt_gc_expire++;
3175 	fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
3176 	entries = dst_entries_get_slow(ops);
3177 	if (entries < ops->gc_thresh)
3178 		net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
3179 out:
3180 	net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
3181 	return entries > rt_max_size;
3182 }
3183 
3184 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3185 			       const struct in6_addr *gw_addr, u32 tbid,
3186 			       int flags, struct fib6_result *res)
3187 {
3188 	struct flowi6 fl6 = {
3189 		.flowi6_oif = cfg->fc_ifindex,
3190 		.daddr = *gw_addr,
3191 		.saddr = cfg->fc_prefsrc,
3192 	};
3193 	struct fib6_table *table;
3194 	int err;
3195 
3196 	table = fib6_get_table(net, tbid);
3197 	if (!table)
3198 		return -EINVAL;
3199 
3200 	if (!ipv6_addr_any(&cfg->fc_prefsrc))
3201 		flags |= RT6_LOOKUP_F_HAS_SADDR;
3202 
3203 	flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3204 
3205 	err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3206 	if (!err && res->f6i != net->ipv6.fib6_null_entry)
3207 		fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3208 				 cfg->fc_ifindex != 0, NULL, flags);
3209 
3210 	return err;
3211 }
3212 
3213 static int ip6_route_check_nh_onlink(struct net *net,
3214 				     struct fib6_config *cfg,
3215 				     const struct net_device *dev,
3216 				     struct netlink_ext_ack *extack)
3217 {
3218 	u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3219 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3220 	struct fib6_result res = {};
3221 	int err;
3222 
3223 	err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3224 	if (!err && !(res.fib6_flags & RTF_REJECT) &&
3225 	    /* ignore match if it is the default route */
3226 	    !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3227 	    (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3228 		NL_SET_ERR_MSG(extack,
3229 			       "Nexthop has invalid gateway or device mismatch");
3230 		err = -EINVAL;
3231 	}
3232 
3233 	return err;
3234 }
3235 
3236 static int ip6_route_check_nh(struct net *net,
3237 			      struct fib6_config *cfg,
3238 			      struct net_device **_dev,
3239 			      struct inet6_dev **idev)
3240 {
3241 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3242 	struct net_device *dev = _dev ? *_dev : NULL;
3243 	int flags = RT6_LOOKUP_F_IFACE;
3244 	struct fib6_result res = {};
3245 	int err = -EHOSTUNREACH;
3246 
3247 	if (cfg->fc_table) {
3248 		err = ip6_nh_lookup_table(net, cfg, gw_addr,
3249 					  cfg->fc_table, flags, &res);
3250 		/* gw_addr can not require a gateway or resolve to a reject
3251 		 * route. If a device is given, it must match the result.
3252 		 */
3253 		if (err || res.fib6_flags & RTF_REJECT ||
3254 		    res.nh->fib_nh_gw_family ||
3255 		    (dev && dev != res.nh->fib_nh_dev))
3256 			err = -EHOSTUNREACH;
3257 	}
3258 
3259 	if (err < 0) {
3260 		struct flowi6 fl6 = {
3261 			.flowi6_oif = cfg->fc_ifindex,
3262 			.daddr = *gw_addr,
3263 		};
3264 
3265 		err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3266 		if (err || res.fib6_flags & RTF_REJECT ||
3267 		    res.nh->fib_nh_gw_family)
3268 			err = -EHOSTUNREACH;
3269 
3270 		if (err)
3271 			return err;
3272 
3273 		fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3274 				 cfg->fc_ifindex != 0, NULL, flags);
3275 	}
3276 
3277 	err = 0;
3278 	if (dev) {
3279 		if (dev != res.nh->fib_nh_dev)
3280 			err = -EHOSTUNREACH;
3281 	} else {
3282 		*_dev = dev = res.nh->fib_nh_dev;
3283 		dev_hold(dev);
3284 		*idev = in6_dev_get(dev);
3285 	}
3286 
3287 	return err;
3288 }
3289 
3290 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3291 			   struct net_device **_dev, struct inet6_dev **idev,
3292 			   struct netlink_ext_ack *extack)
3293 {
3294 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3295 	int gwa_type = ipv6_addr_type(gw_addr);
3296 	bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3297 	const struct net_device *dev = *_dev;
3298 	bool need_addr_check = !dev;
3299 	int err = -EINVAL;
3300 
3301 	/* if gw_addr is local we will fail to detect this in case
3302 	 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3303 	 * will return already-added prefix route via interface that
3304 	 * prefix route was assigned to, which might be non-loopback.
3305 	 */
3306 	if (dev &&
3307 	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3308 		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3309 		goto out;
3310 	}
3311 
3312 	if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3313 		/* IPv6 strictly inhibits using not link-local
3314 		 * addresses as nexthop address.
3315 		 * Otherwise, router will not able to send redirects.
3316 		 * It is very good, but in some (rare!) circumstances
3317 		 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3318 		 * some exceptions. --ANK
3319 		 * We allow IPv4-mapped nexthops to support RFC4798-type
3320 		 * addressing
3321 		 */
3322 		if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3323 			NL_SET_ERR_MSG(extack, "Invalid gateway address");
3324 			goto out;
3325 		}
3326 
3327 		rcu_read_lock();
3328 
3329 		if (cfg->fc_flags & RTNH_F_ONLINK)
3330 			err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3331 		else
3332 			err = ip6_route_check_nh(net, cfg, _dev, idev);
3333 
3334 		rcu_read_unlock();
3335 
3336 		if (err)
3337 			goto out;
3338 	}
3339 
3340 	/* reload in case device was changed */
3341 	dev = *_dev;
3342 
3343 	err = -EINVAL;
3344 	if (!dev) {
3345 		NL_SET_ERR_MSG(extack, "Egress device not specified");
3346 		goto out;
3347 	} else if (dev->flags & IFF_LOOPBACK) {
3348 		NL_SET_ERR_MSG(extack,
3349 			       "Egress device can not be loopback device for this route");
3350 		goto out;
3351 	}
3352 
3353 	/* if we did not check gw_addr above, do so now that the
3354 	 * egress device has been resolved.
3355 	 */
3356 	if (need_addr_check &&
3357 	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3358 		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3359 		goto out;
3360 	}
3361 
3362 	err = 0;
3363 out:
3364 	return err;
3365 }
3366 
3367 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3368 {
3369 	if ((flags & RTF_REJECT) ||
3370 	    (dev && (dev->flags & IFF_LOOPBACK) &&
3371 	     !(addr_type & IPV6_ADDR_LOOPBACK) &&
3372 	     !(flags & RTF_LOCAL)))
3373 		return true;
3374 
3375 	return false;
3376 }
3377 
3378 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3379 		 struct fib6_config *cfg, gfp_t gfp_flags,
3380 		 struct netlink_ext_ack *extack)
3381 {
3382 	struct net_device *dev = NULL;
3383 	struct inet6_dev *idev = NULL;
3384 	int addr_type;
3385 	int err;
3386 
3387 	fib6_nh->fib_nh_family = AF_INET6;
3388 
3389 	err = -ENODEV;
3390 	if (cfg->fc_ifindex) {
3391 		dev = dev_get_by_index(net, cfg->fc_ifindex);
3392 		if (!dev)
3393 			goto out;
3394 		idev = in6_dev_get(dev);
3395 		if (!idev)
3396 			goto out;
3397 	}
3398 
3399 	if (cfg->fc_flags & RTNH_F_ONLINK) {
3400 		if (!dev) {
3401 			NL_SET_ERR_MSG(extack,
3402 				       "Nexthop device required for onlink");
3403 			goto out;
3404 		}
3405 
3406 		if (!(dev->flags & IFF_UP)) {
3407 			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3408 			err = -ENETDOWN;
3409 			goto out;
3410 		}
3411 
3412 		fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3413 	}
3414 
3415 	fib6_nh->fib_nh_weight = 1;
3416 
3417 	/* We cannot add true routes via loopback here,
3418 	 * they would result in kernel looping; promote them to reject routes
3419 	 */
3420 	addr_type = ipv6_addr_type(&cfg->fc_dst);
3421 	if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3422 		/* hold loopback dev/idev if we haven't done so. */
3423 		if (dev != net->loopback_dev) {
3424 			if (dev) {
3425 				dev_put(dev);
3426 				in6_dev_put(idev);
3427 			}
3428 			dev = net->loopback_dev;
3429 			dev_hold(dev);
3430 			idev = in6_dev_get(dev);
3431 			if (!idev) {
3432 				err = -ENODEV;
3433 				goto out;
3434 			}
3435 		}
3436 		goto pcpu_alloc;
3437 	}
3438 
3439 	if (cfg->fc_flags & RTF_GATEWAY) {
3440 		err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3441 		if (err)
3442 			goto out;
3443 
3444 		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3445 		fib6_nh->fib_nh_gw_family = AF_INET6;
3446 	}
3447 
3448 	err = -ENODEV;
3449 	if (!dev)
3450 		goto out;
3451 
3452 	if (idev->cnf.disable_ipv6) {
3453 		NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3454 		err = -EACCES;
3455 		goto out;
3456 	}
3457 
3458 	if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3459 		NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3460 		err = -ENETDOWN;
3461 		goto out;
3462 	}
3463 
3464 	if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3465 	    !netif_carrier_ok(dev))
3466 		fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3467 
3468 	err = fib_nh_common_init(&fib6_nh->nh_common, cfg->fc_encap,
3469 				 cfg->fc_encap_type, cfg, gfp_flags, extack);
3470 	if (err)
3471 		goto out;
3472 
3473 pcpu_alloc:
3474 	fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3475 	if (!fib6_nh->rt6i_pcpu) {
3476 		err = -ENOMEM;
3477 		goto out;
3478 	}
3479 
3480 	fib6_nh->fib_nh_dev = dev;
3481 	fib6_nh->fib_nh_oif = dev->ifindex;
3482 	err = 0;
3483 out:
3484 	if (idev)
3485 		in6_dev_put(idev);
3486 
3487 	if (err) {
3488 		lwtstate_put(fib6_nh->fib_nh_lws);
3489 		fib6_nh->fib_nh_lws = NULL;
3490 		if (dev)
3491 			dev_put(dev);
3492 	}
3493 
3494 	return err;
3495 }
3496 
3497 void fib6_nh_release(struct fib6_nh *fib6_nh)
3498 {
3499 	struct rt6_exception_bucket *bucket;
3500 
3501 	rcu_read_lock();
3502 
3503 	fib6_nh_flush_exceptions(fib6_nh, NULL);
3504 	bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3505 	if (bucket) {
3506 		rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3507 		kfree(bucket);
3508 	}
3509 
3510 	rcu_read_unlock();
3511 
3512 	if (fib6_nh->rt6i_pcpu) {
3513 		int cpu;
3514 
3515 		for_each_possible_cpu(cpu) {
3516 			struct rt6_info **ppcpu_rt;
3517 			struct rt6_info *pcpu_rt;
3518 
3519 			ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3520 			pcpu_rt = *ppcpu_rt;
3521 			if (pcpu_rt) {
3522 				dst_dev_put(&pcpu_rt->dst);
3523 				dst_release(&pcpu_rt->dst);
3524 				*ppcpu_rt = NULL;
3525 			}
3526 		}
3527 
3528 		free_percpu(fib6_nh->rt6i_pcpu);
3529 	}
3530 
3531 	fib_nh_common_release(&fib6_nh->nh_common);
3532 }
3533 
3534 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3535 					      gfp_t gfp_flags,
3536 					      struct netlink_ext_ack *extack)
3537 {
3538 	struct net *net = cfg->fc_nlinfo.nl_net;
3539 	struct fib6_info *rt = NULL;
3540 	struct nexthop *nh = NULL;
3541 	struct fib6_table *table;
3542 	struct fib6_nh *fib6_nh;
3543 	int err = -EINVAL;
3544 	int addr_type;
3545 
3546 	/* RTF_PCPU is an internal flag; can not be set by userspace */
3547 	if (cfg->fc_flags & RTF_PCPU) {
3548 		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3549 		goto out;
3550 	}
3551 
3552 	/* RTF_CACHE is an internal flag; can not be set by userspace */
3553 	if (cfg->fc_flags & RTF_CACHE) {
3554 		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3555 		goto out;
3556 	}
3557 
3558 	if (cfg->fc_type > RTN_MAX) {
3559 		NL_SET_ERR_MSG(extack, "Invalid route type");
3560 		goto out;
3561 	}
3562 
3563 	if (cfg->fc_dst_len > 128) {
3564 		NL_SET_ERR_MSG(extack, "Invalid prefix length");
3565 		goto out;
3566 	}
3567 	if (cfg->fc_src_len > 128) {
3568 		NL_SET_ERR_MSG(extack, "Invalid source address length");
3569 		goto out;
3570 	}
3571 #ifndef CONFIG_IPV6_SUBTREES
3572 	if (cfg->fc_src_len) {
3573 		NL_SET_ERR_MSG(extack,
3574 			       "Specifying source address requires IPV6_SUBTREES to be enabled");
3575 		goto out;
3576 	}
3577 #endif
3578 	if (cfg->fc_nh_id) {
3579 		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3580 		if (!nh) {
3581 			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3582 			goto out;
3583 		}
3584 		err = fib6_check_nexthop(nh, cfg, extack);
3585 		if (err)
3586 			goto out;
3587 	}
3588 
3589 	err = -ENOBUFS;
3590 	if (cfg->fc_nlinfo.nlh &&
3591 	    !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3592 		table = fib6_get_table(net, cfg->fc_table);
3593 		if (!table) {
3594 			pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3595 			table = fib6_new_table(net, cfg->fc_table);
3596 		}
3597 	} else {
3598 		table = fib6_new_table(net, cfg->fc_table);
3599 	}
3600 
3601 	if (!table)
3602 		goto out;
3603 
3604 	err = -ENOMEM;
3605 	rt = fib6_info_alloc(gfp_flags, !nh);
3606 	if (!rt)
3607 		goto out;
3608 
3609 	rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3610 					       extack);
3611 	if (IS_ERR(rt->fib6_metrics)) {
3612 		err = PTR_ERR(rt->fib6_metrics);
3613 		/* Do not leave garbage there. */
3614 		rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3615 		goto out;
3616 	}
3617 
3618 	if (cfg->fc_flags & RTF_ADDRCONF)
3619 		rt->dst_nocount = true;
3620 
3621 	if (cfg->fc_flags & RTF_EXPIRES)
3622 		fib6_set_expires(rt, jiffies +
3623 				clock_t_to_jiffies(cfg->fc_expires));
3624 	else
3625 		fib6_clean_expires(rt);
3626 
3627 	if (cfg->fc_protocol == RTPROT_UNSPEC)
3628 		cfg->fc_protocol = RTPROT_BOOT;
3629 	rt->fib6_protocol = cfg->fc_protocol;
3630 
3631 	rt->fib6_table = table;
3632 	rt->fib6_metric = cfg->fc_metric;
3633 	rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3634 	rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3635 
3636 	ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3637 	rt->fib6_dst.plen = cfg->fc_dst_len;
3638 	if (rt->fib6_dst.plen == 128)
3639 		rt->dst_host = true;
3640 
3641 #ifdef CONFIG_IPV6_SUBTREES
3642 	ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3643 	rt->fib6_src.plen = cfg->fc_src_len;
3644 #endif
3645 	if (nh) {
3646 		if (!nexthop_get(nh)) {
3647 			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3648 			goto out;
3649 		}
3650 		if (rt->fib6_src.plen) {
3651 			NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3652 			goto out;
3653 		}
3654 		rt->nh = nh;
3655 		fib6_nh = nexthop_fib6_nh(rt->nh);
3656 	} else {
3657 		err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3658 		if (err)
3659 			goto out;
3660 
3661 		fib6_nh = rt->fib6_nh;
3662 
3663 		/* We cannot add true routes via loopback here, they would
3664 		 * result in kernel looping; promote them to reject routes
3665 		 */
3666 		addr_type = ipv6_addr_type(&cfg->fc_dst);
3667 		if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3668 				   addr_type))
3669 			rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3670 	}
3671 
3672 	if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3673 		struct net_device *dev = fib6_nh->fib_nh_dev;
3674 
3675 		if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3676 			NL_SET_ERR_MSG(extack, "Invalid source address");
3677 			err = -EINVAL;
3678 			goto out;
3679 		}
3680 		rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3681 		rt->fib6_prefsrc.plen = 128;
3682 	} else
3683 		rt->fib6_prefsrc.plen = 0;
3684 
3685 	return rt;
3686 out:
3687 	fib6_info_release(rt);
3688 	return ERR_PTR(err);
3689 }
3690 
3691 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3692 		  struct netlink_ext_ack *extack)
3693 {
3694 	struct fib6_info *rt;
3695 	int err;
3696 
3697 	rt = ip6_route_info_create(cfg, gfp_flags, extack);
3698 	if (IS_ERR(rt))
3699 		return PTR_ERR(rt);
3700 
3701 	err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3702 	fib6_info_release(rt);
3703 
3704 	return err;
3705 }
3706 
3707 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3708 {
3709 	struct net *net = info->nl_net;
3710 	struct fib6_table *table;
3711 	int err;
3712 
3713 	if (rt == net->ipv6.fib6_null_entry) {
3714 		err = -ENOENT;
3715 		goto out;
3716 	}
3717 
3718 	table = rt->fib6_table;
3719 	spin_lock_bh(&table->tb6_lock);
3720 	err = fib6_del(rt, info);
3721 	spin_unlock_bh(&table->tb6_lock);
3722 
3723 out:
3724 	fib6_info_release(rt);
3725 	return err;
3726 }
3727 
3728 int ip6_del_rt(struct net *net, struct fib6_info *rt)
3729 {
3730 	struct nl_info info = { .nl_net = net };
3731 
3732 	return __ip6_del_rt(rt, &info);
3733 }
3734 
3735 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3736 {
3737 	struct nl_info *info = &cfg->fc_nlinfo;
3738 	struct net *net = info->nl_net;
3739 	struct sk_buff *skb = NULL;
3740 	struct fib6_table *table;
3741 	int err = -ENOENT;
3742 
3743 	if (rt == net->ipv6.fib6_null_entry)
3744 		goto out_put;
3745 	table = rt->fib6_table;
3746 	spin_lock_bh(&table->tb6_lock);
3747 
3748 	if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3749 		struct fib6_info *sibling, *next_sibling;
3750 
3751 		/* prefer to send a single notification with all hops */
3752 		skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3753 		if (skb) {
3754 			u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3755 
3756 			if (rt6_fill_node(net, skb, rt, NULL,
3757 					  NULL, NULL, 0, RTM_DELROUTE,
3758 					  info->portid, seq, 0) < 0) {
3759 				kfree_skb(skb);
3760 				skb = NULL;
3761 			} else
3762 				info->skip_notify = 1;
3763 		}
3764 
3765 		info->skip_notify_kernel = 1;
3766 		call_fib6_multipath_entry_notifiers(net,
3767 						    FIB_EVENT_ENTRY_DEL,
3768 						    rt,
3769 						    rt->fib6_nsiblings,
3770 						    NULL);
3771 		list_for_each_entry_safe(sibling, next_sibling,
3772 					 &rt->fib6_siblings,
3773 					 fib6_siblings) {
3774 			err = fib6_del(sibling, info);
3775 			if (err)
3776 				goto out_unlock;
3777 		}
3778 	}
3779 
3780 	err = fib6_del(rt, info);
3781 out_unlock:
3782 	spin_unlock_bh(&table->tb6_lock);
3783 out_put:
3784 	fib6_info_release(rt);
3785 
3786 	if (skb) {
3787 		rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3788 			    info->nlh, gfp_any());
3789 	}
3790 	return err;
3791 }
3792 
3793 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3794 {
3795 	int rc = -ESRCH;
3796 
3797 	if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3798 		goto out;
3799 
3800 	if (cfg->fc_flags & RTF_GATEWAY &&
3801 	    !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3802 		goto out;
3803 
3804 	rc = rt6_remove_exception_rt(rt);
3805 out:
3806 	return rc;
3807 }
3808 
3809 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3810 			     struct fib6_nh *nh)
3811 {
3812 	struct fib6_result res = {
3813 		.f6i = rt,
3814 		.nh = nh,
3815 	};
3816 	struct rt6_info *rt_cache;
3817 
3818 	rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3819 	if (rt_cache)
3820 		return __ip6_del_cached_rt(rt_cache, cfg);
3821 
3822 	return 0;
3823 }
3824 
3825 struct fib6_nh_del_cached_rt_arg {
3826 	struct fib6_config *cfg;
3827 	struct fib6_info *f6i;
3828 };
3829 
3830 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3831 {
3832 	struct fib6_nh_del_cached_rt_arg *arg = _arg;
3833 	int rc;
3834 
3835 	rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3836 	return rc != -ESRCH ? rc : 0;
3837 }
3838 
3839 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3840 {
3841 	struct fib6_nh_del_cached_rt_arg arg = {
3842 		.cfg = cfg,
3843 		.f6i = f6i
3844 	};
3845 
3846 	return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3847 }
3848 
3849 static int ip6_route_del(struct fib6_config *cfg,
3850 			 struct netlink_ext_ack *extack)
3851 {
3852 	struct fib6_table *table;
3853 	struct fib6_info *rt;
3854 	struct fib6_node *fn;
3855 	int err = -ESRCH;
3856 
3857 	table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
3858 	if (!table) {
3859 		NL_SET_ERR_MSG(extack, "FIB table does not exist");
3860 		return err;
3861 	}
3862 
3863 	rcu_read_lock();
3864 
3865 	fn = fib6_locate(&table->tb6_root,
3866 			 &cfg->fc_dst, cfg->fc_dst_len,
3867 			 &cfg->fc_src, cfg->fc_src_len,
3868 			 !(cfg->fc_flags & RTF_CACHE));
3869 
3870 	if (fn) {
3871 		for_each_fib6_node_rt_rcu(fn) {
3872 			struct fib6_nh *nh;
3873 
3874 			if (rt->nh && cfg->fc_nh_id &&
3875 			    rt->nh->id != cfg->fc_nh_id)
3876 				continue;
3877 
3878 			if (cfg->fc_flags & RTF_CACHE) {
3879 				int rc = 0;
3880 
3881 				if (rt->nh) {
3882 					rc = ip6_del_cached_rt_nh(cfg, rt);
3883 				} else if (cfg->fc_nh_id) {
3884 					continue;
3885 				} else {
3886 					nh = rt->fib6_nh;
3887 					rc = ip6_del_cached_rt(cfg, rt, nh);
3888 				}
3889 				if (rc != -ESRCH) {
3890 					rcu_read_unlock();
3891 					return rc;
3892 				}
3893 				continue;
3894 			}
3895 
3896 			if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3897 				continue;
3898 			if (cfg->fc_protocol &&
3899 			    cfg->fc_protocol != rt->fib6_protocol)
3900 				continue;
3901 
3902 			if (rt->nh) {
3903 				if (!fib6_info_hold_safe(rt))
3904 					continue;
3905 				rcu_read_unlock();
3906 
3907 				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3908 			}
3909 			if (cfg->fc_nh_id)
3910 				continue;
3911 
3912 			nh = rt->fib6_nh;
3913 			if (cfg->fc_ifindex &&
3914 			    (!nh->fib_nh_dev ||
3915 			     nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
3916 				continue;
3917 			if (cfg->fc_flags & RTF_GATEWAY &&
3918 			    !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
3919 				continue;
3920 			if (!fib6_info_hold_safe(rt))
3921 				continue;
3922 			rcu_read_unlock();
3923 
3924 			/* if gateway was specified only delete the one hop */
3925 			if (cfg->fc_flags & RTF_GATEWAY)
3926 				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3927 
3928 			return __ip6_del_rt_siblings(rt, cfg);
3929 		}
3930 	}
3931 	rcu_read_unlock();
3932 
3933 	return err;
3934 }
3935 
3936 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
3937 {
3938 	struct netevent_redirect netevent;
3939 	struct rt6_info *rt, *nrt = NULL;
3940 	struct fib6_result res = {};
3941 	struct ndisc_options ndopts;
3942 	struct inet6_dev *in6_dev;
3943 	struct neighbour *neigh;
3944 	struct rd_msg *msg;
3945 	int optlen, on_link;
3946 	u8 *lladdr;
3947 
3948 	optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
3949 	optlen -= sizeof(*msg);
3950 
3951 	if (optlen < 0) {
3952 		net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
3953 		return;
3954 	}
3955 
3956 	msg = (struct rd_msg *)icmp6_hdr(skb);
3957 
3958 	if (ipv6_addr_is_multicast(&msg->dest)) {
3959 		net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
3960 		return;
3961 	}
3962 
3963 	on_link = 0;
3964 	if (ipv6_addr_equal(&msg->dest, &msg->target)) {
3965 		on_link = 1;
3966 	} else if (ipv6_addr_type(&msg->target) !=
3967 		   (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
3968 		net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
3969 		return;
3970 	}
3971 
3972 	in6_dev = __in6_dev_get(skb->dev);
3973 	if (!in6_dev)
3974 		return;
3975 	if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3976 		return;
3977 
3978 	/* RFC2461 8.1:
3979 	 *	The IP source address of the Redirect MUST be the same as the current
3980 	 *	first-hop router for the specified ICMP Destination Address.
3981 	 */
3982 
3983 	if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
3984 		net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3985 		return;
3986 	}
3987 
3988 	lladdr = NULL;
3989 	if (ndopts.nd_opts_tgt_lladdr) {
3990 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3991 					     skb->dev);
3992 		if (!lladdr) {
3993 			net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3994 			return;
3995 		}
3996 	}
3997 
3998 	rt = (struct rt6_info *) dst;
3999 	if (rt->rt6i_flags & RTF_REJECT) {
4000 		net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4001 		return;
4002 	}
4003 
4004 	/* Redirect received -> path was valid.
4005 	 * Look, redirects are sent only in response to data packets,
4006 	 * so that this nexthop apparently is reachable. --ANK
4007 	 */
4008 	dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4009 
4010 	neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4011 	if (!neigh)
4012 		return;
4013 
4014 	/*
4015 	 *	We have finally decided to accept it.
4016 	 */
4017 
4018 	ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4019 		     NEIGH_UPDATE_F_WEAK_OVERRIDE|
4020 		     NEIGH_UPDATE_F_OVERRIDE|
4021 		     (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4022 				     NEIGH_UPDATE_F_ISROUTER)),
4023 		     NDISC_REDIRECT, &ndopts);
4024 
4025 	rcu_read_lock();
4026 	res.f6i = rcu_dereference(rt->from);
4027 	if (!res.f6i)
4028 		goto out;
4029 
4030 	if (res.f6i->nh) {
4031 		struct fib6_nh_match_arg arg = {
4032 			.dev = dst->dev,
4033 			.gw = &rt->rt6i_gateway,
4034 		};
4035 
4036 		nexthop_for_each_fib6_nh(res.f6i->nh,
4037 					 fib6_nh_find_match, &arg);
4038 
4039 		/* fib6_info uses a nexthop that does not have fib6_nh
4040 		 * using the dst->dev. Should be impossible
4041 		 */
4042 		if (!arg.match)
4043 			goto out;
4044 		res.nh = arg.match;
4045 	} else {
4046 		res.nh = res.f6i->fib6_nh;
4047 	}
4048 
4049 	res.fib6_flags = res.f6i->fib6_flags;
4050 	res.fib6_type = res.f6i->fib6_type;
4051 	nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4052 	if (!nrt)
4053 		goto out;
4054 
4055 	nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4056 	if (on_link)
4057 		nrt->rt6i_flags &= ~RTF_GATEWAY;
4058 
4059 	nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4060 
4061 	/* rt6_insert_exception() will take care of duplicated exceptions */
4062 	if (rt6_insert_exception(nrt, &res)) {
4063 		dst_release_immediate(&nrt->dst);
4064 		goto out;
4065 	}
4066 
4067 	netevent.old = &rt->dst;
4068 	netevent.new = &nrt->dst;
4069 	netevent.daddr = &msg->dest;
4070 	netevent.neigh = neigh;
4071 	call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4072 
4073 out:
4074 	rcu_read_unlock();
4075 	neigh_release(neigh);
4076 }
4077 
4078 #ifdef CONFIG_IPV6_ROUTE_INFO
4079 static struct fib6_info *rt6_get_route_info(struct net *net,
4080 					   const struct in6_addr *prefix, int prefixlen,
4081 					   const struct in6_addr *gwaddr,
4082 					   struct net_device *dev)
4083 {
4084 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4085 	int ifindex = dev->ifindex;
4086 	struct fib6_node *fn;
4087 	struct fib6_info *rt = NULL;
4088 	struct fib6_table *table;
4089 
4090 	table = fib6_get_table(net, tb_id);
4091 	if (!table)
4092 		return NULL;
4093 
4094 	rcu_read_lock();
4095 	fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4096 	if (!fn)
4097 		goto out;
4098 
4099 	for_each_fib6_node_rt_rcu(fn) {
4100 		/* these routes do not use nexthops */
4101 		if (rt->nh)
4102 			continue;
4103 		if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4104 			continue;
4105 		if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4106 		    !rt->fib6_nh->fib_nh_gw_family)
4107 			continue;
4108 		if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4109 			continue;
4110 		if (!fib6_info_hold_safe(rt))
4111 			continue;
4112 		break;
4113 	}
4114 out:
4115 	rcu_read_unlock();
4116 	return rt;
4117 }
4118 
4119 static struct fib6_info *rt6_add_route_info(struct net *net,
4120 					   const struct in6_addr *prefix, int prefixlen,
4121 					   const struct in6_addr *gwaddr,
4122 					   struct net_device *dev,
4123 					   unsigned int pref)
4124 {
4125 	struct fib6_config cfg = {
4126 		.fc_metric	= IP6_RT_PRIO_USER,
4127 		.fc_ifindex	= dev->ifindex,
4128 		.fc_dst_len	= prefixlen,
4129 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4130 				  RTF_UP | RTF_PREF(pref),
4131 		.fc_protocol = RTPROT_RA,
4132 		.fc_type = RTN_UNICAST,
4133 		.fc_nlinfo.portid = 0,
4134 		.fc_nlinfo.nlh = NULL,
4135 		.fc_nlinfo.nl_net = net,
4136 	};
4137 
4138 	cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4139 	cfg.fc_dst = *prefix;
4140 	cfg.fc_gateway = *gwaddr;
4141 
4142 	/* We should treat it as a default route if prefix length is 0. */
4143 	if (!prefixlen)
4144 		cfg.fc_flags |= RTF_DEFAULT;
4145 
4146 	ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4147 
4148 	return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4149 }
4150 #endif
4151 
4152 struct fib6_info *rt6_get_dflt_router(struct net *net,
4153 				     const struct in6_addr *addr,
4154 				     struct net_device *dev)
4155 {
4156 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4157 	struct fib6_info *rt;
4158 	struct fib6_table *table;
4159 
4160 	table = fib6_get_table(net, tb_id);
4161 	if (!table)
4162 		return NULL;
4163 
4164 	rcu_read_lock();
4165 	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4166 		struct fib6_nh *nh;
4167 
4168 		/* RA routes do not use nexthops */
4169 		if (rt->nh)
4170 			continue;
4171 
4172 		nh = rt->fib6_nh;
4173 		if (dev == nh->fib_nh_dev &&
4174 		    ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4175 		    ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4176 			break;
4177 	}
4178 	if (rt && !fib6_info_hold_safe(rt))
4179 		rt = NULL;
4180 	rcu_read_unlock();
4181 	return rt;
4182 }
4183 
4184 struct fib6_info *rt6_add_dflt_router(struct net *net,
4185 				     const struct in6_addr *gwaddr,
4186 				     struct net_device *dev,
4187 				     unsigned int pref)
4188 {
4189 	struct fib6_config cfg = {
4190 		.fc_table	= l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4191 		.fc_metric	= IP6_RT_PRIO_USER,
4192 		.fc_ifindex	= dev->ifindex,
4193 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4194 				  RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4195 		.fc_protocol = RTPROT_RA,
4196 		.fc_type = RTN_UNICAST,
4197 		.fc_nlinfo.portid = 0,
4198 		.fc_nlinfo.nlh = NULL,
4199 		.fc_nlinfo.nl_net = net,
4200 	};
4201 
4202 	cfg.fc_gateway = *gwaddr;
4203 
4204 	if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4205 		struct fib6_table *table;
4206 
4207 		table = fib6_get_table(dev_net(dev), cfg.fc_table);
4208 		if (table)
4209 			table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4210 	}
4211 
4212 	return rt6_get_dflt_router(net, gwaddr, dev);
4213 }
4214 
4215 static void __rt6_purge_dflt_routers(struct net *net,
4216 				     struct fib6_table *table)
4217 {
4218 	struct fib6_info *rt;
4219 
4220 restart:
4221 	rcu_read_lock();
4222 	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4223 		struct net_device *dev = fib6_info_nh_dev(rt);
4224 		struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4225 
4226 		if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4227 		    (!idev || idev->cnf.accept_ra != 2) &&
4228 		    fib6_info_hold_safe(rt)) {
4229 			rcu_read_unlock();
4230 			ip6_del_rt(net, rt);
4231 			goto restart;
4232 		}
4233 	}
4234 	rcu_read_unlock();
4235 
4236 	table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4237 }
4238 
4239 void rt6_purge_dflt_routers(struct net *net)
4240 {
4241 	struct fib6_table *table;
4242 	struct hlist_head *head;
4243 	unsigned int h;
4244 
4245 	rcu_read_lock();
4246 
4247 	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4248 		head = &net->ipv6.fib_table_hash[h];
4249 		hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4250 			if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4251 				__rt6_purge_dflt_routers(net, table);
4252 		}
4253 	}
4254 
4255 	rcu_read_unlock();
4256 }
4257 
4258 static void rtmsg_to_fib6_config(struct net *net,
4259 				 struct in6_rtmsg *rtmsg,
4260 				 struct fib6_config *cfg)
4261 {
4262 	*cfg = (struct fib6_config){
4263 		.fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4264 			 : RT6_TABLE_MAIN,
4265 		.fc_ifindex = rtmsg->rtmsg_ifindex,
4266 		.fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4267 		.fc_expires = rtmsg->rtmsg_info,
4268 		.fc_dst_len = rtmsg->rtmsg_dst_len,
4269 		.fc_src_len = rtmsg->rtmsg_src_len,
4270 		.fc_flags = rtmsg->rtmsg_flags,
4271 		.fc_type = rtmsg->rtmsg_type,
4272 
4273 		.fc_nlinfo.nl_net = net,
4274 
4275 		.fc_dst = rtmsg->rtmsg_dst,
4276 		.fc_src = rtmsg->rtmsg_src,
4277 		.fc_gateway = rtmsg->rtmsg_gateway,
4278 	};
4279 }
4280 
4281 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
4282 {
4283 	struct fib6_config cfg;
4284 	struct in6_rtmsg rtmsg;
4285 	int err;
4286 
4287 	switch (cmd) {
4288 	case SIOCADDRT:		/* Add a route */
4289 	case SIOCDELRT:		/* Delete a route */
4290 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4291 			return -EPERM;
4292 		err = copy_from_user(&rtmsg, arg,
4293 				     sizeof(struct in6_rtmsg));
4294 		if (err)
4295 			return -EFAULT;
4296 
4297 		rtmsg_to_fib6_config(net, &rtmsg, &cfg);
4298 
4299 		rtnl_lock();
4300 		switch (cmd) {
4301 		case SIOCADDRT:
4302 			err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4303 			break;
4304 		case SIOCDELRT:
4305 			err = ip6_route_del(&cfg, NULL);
4306 			break;
4307 		default:
4308 			err = -EINVAL;
4309 		}
4310 		rtnl_unlock();
4311 
4312 		return err;
4313 	}
4314 
4315 	return -EINVAL;
4316 }
4317 
4318 /*
4319  *	Drop the packet on the floor
4320  */
4321 
4322 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4323 {
4324 	struct dst_entry *dst = skb_dst(skb);
4325 	struct net *net = dev_net(dst->dev);
4326 	struct inet6_dev *idev;
4327 	int type;
4328 
4329 	if (netif_is_l3_master(skb->dev) &&
4330 	    dst->dev == net->loopback_dev)
4331 		idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4332 	else
4333 		idev = ip6_dst_idev(dst);
4334 
4335 	switch (ipstats_mib_noroutes) {
4336 	case IPSTATS_MIB_INNOROUTES:
4337 		type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4338 		if (type == IPV6_ADDR_ANY) {
4339 			IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4340 			break;
4341 		}
4342 		/* FALLTHROUGH */
4343 	case IPSTATS_MIB_OUTNOROUTES:
4344 		IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4345 		break;
4346 	}
4347 
4348 	/* Start over by dropping the dst for l3mdev case */
4349 	if (netif_is_l3_master(skb->dev))
4350 		skb_dst_drop(skb);
4351 
4352 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4353 	kfree_skb(skb);
4354 	return 0;
4355 }
4356 
4357 static int ip6_pkt_discard(struct sk_buff *skb)
4358 {
4359 	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4360 }
4361 
4362 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4363 {
4364 	skb->dev = skb_dst(skb)->dev;
4365 	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4366 }
4367 
4368 static int ip6_pkt_prohibit(struct sk_buff *skb)
4369 {
4370 	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4371 }
4372 
4373 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4374 {
4375 	skb->dev = skb_dst(skb)->dev;
4376 	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4377 }
4378 
4379 /*
4380  *	Allocate a dst for local (unicast / anycast) address.
4381  */
4382 
4383 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4384 				     struct inet6_dev *idev,
4385 				     const struct in6_addr *addr,
4386 				     bool anycast, gfp_t gfp_flags)
4387 {
4388 	struct fib6_config cfg = {
4389 		.fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4390 		.fc_ifindex = idev->dev->ifindex,
4391 		.fc_flags = RTF_UP | RTF_ADDRCONF | RTF_NONEXTHOP,
4392 		.fc_dst = *addr,
4393 		.fc_dst_len = 128,
4394 		.fc_protocol = RTPROT_KERNEL,
4395 		.fc_nlinfo.nl_net = net,
4396 		.fc_ignore_dev_down = true,
4397 	};
4398 
4399 	if (anycast) {
4400 		cfg.fc_type = RTN_ANYCAST;
4401 		cfg.fc_flags |= RTF_ANYCAST;
4402 	} else {
4403 		cfg.fc_type = RTN_LOCAL;
4404 		cfg.fc_flags |= RTF_LOCAL;
4405 	}
4406 
4407 	return ip6_route_info_create(&cfg, gfp_flags, NULL);
4408 }
4409 
4410 /* remove deleted ip from prefsrc entries */
4411 struct arg_dev_net_ip {
4412 	struct net_device *dev;
4413 	struct net *net;
4414 	struct in6_addr *addr;
4415 };
4416 
4417 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4418 {
4419 	struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4420 	struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4421 	struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4422 
4423 	if (!rt->nh &&
4424 	    ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4425 	    rt != net->ipv6.fib6_null_entry &&
4426 	    ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4427 		spin_lock_bh(&rt6_exception_lock);
4428 		/* remove prefsrc entry */
4429 		rt->fib6_prefsrc.plen = 0;
4430 		spin_unlock_bh(&rt6_exception_lock);
4431 	}
4432 	return 0;
4433 }
4434 
4435 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4436 {
4437 	struct net *net = dev_net(ifp->idev->dev);
4438 	struct arg_dev_net_ip adni = {
4439 		.dev = ifp->idev->dev,
4440 		.net = net,
4441 		.addr = &ifp->addr,
4442 	};
4443 	fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4444 }
4445 
4446 #define RTF_RA_ROUTER		(RTF_ADDRCONF | RTF_DEFAULT)
4447 
4448 /* Remove routers and update dst entries when gateway turn into host. */
4449 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4450 {
4451 	struct in6_addr *gateway = (struct in6_addr *)arg;
4452 	struct fib6_nh *nh;
4453 
4454 	/* RA routes do not use nexthops */
4455 	if (rt->nh)
4456 		return 0;
4457 
4458 	nh = rt->fib6_nh;
4459 	if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4460 	    nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4461 		return -1;
4462 
4463 	/* Further clean up cached routes in exception table.
4464 	 * This is needed because cached route may have a different
4465 	 * gateway than its 'parent' in the case of an ip redirect.
4466 	 */
4467 	fib6_nh_exceptions_clean_tohost(nh, gateway);
4468 
4469 	return 0;
4470 }
4471 
4472 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4473 {
4474 	fib6_clean_all(net, fib6_clean_tohost, gateway);
4475 }
4476 
4477 struct arg_netdev_event {
4478 	const struct net_device *dev;
4479 	union {
4480 		unsigned char nh_flags;
4481 		unsigned long event;
4482 	};
4483 };
4484 
4485 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4486 {
4487 	struct fib6_info *iter;
4488 	struct fib6_node *fn;
4489 
4490 	fn = rcu_dereference_protected(rt->fib6_node,
4491 			lockdep_is_held(&rt->fib6_table->tb6_lock));
4492 	iter = rcu_dereference_protected(fn->leaf,
4493 			lockdep_is_held(&rt->fib6_table->tb6_lock));
4494 	while (iter) {
4495 		if (iter->fib6_metric == rt->fib6_metric &&
4496 		    rt6_qualify_for_ecmp(iter))
4497 			return iter;
4498 		iter = rcu_dereference_protected(iter->fib6_next,
4499 				lockdep_is_held(&rt->fib6_table->tb6_lock));
4500 	}
4501 
4502 	return NULL;
4503 }
4504 
4505 /* only called for fib entries with builtin fib6_nh */
4506 static bool rt6_is_dead(const struct fib6_info *rt)
4507 {
4508 	if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4509 	    (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4510 	     ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4511 		return true;
4512 
4513 	return false;
4514 }
4515 
4516 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4517 {
4518 	struct fib6_info *iter;
4519 	int total = 0;
4520 
4521 	if (!rt6_is_dead(rt))
4522 		total += rt->fib6_nh->fib_nh_weight;
4523 
4524 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4525 		if (!rt6_is_dead(iter))
4526 			total += iter->fib6_nh->fib_nh_weight;
4527 	}
4528 
4529 	return total;
4530 }
4531 
4532 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4533 {
4534 	int upper_bound = -1;
4535 
4536 	if (!rt6_is_dead(rt)) {
4537 		*weight += rt->fib6_nh->fib_nh_weight;
4538 		upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4539 						    total) - 1;
4540 	}
4541 	atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4542 }
4543 
4544 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4545 {
4546 	struct fib6_info *iter;
4547 	int weight = 0;
4548 
4549 	rt6_upper_bound_set(rt, &weight, total);
4550 
4551 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4552 		rt6_upper_bound_set(iter, &weight, total);
4553 }
4554 
4555 void rt6_multipath_rebalance(struct fib6_info *rt)
4556 {
4557 	struct fib6_info *first;
4558 	int total;
4559 
4560 	/* In case the entire multipath route was marked for flushing,
4561 	 * then there is no need to rebalance upon the removal of every
4562 	 * sibling route.
4563 	 */
4564 	if (!rt->fib6_nsiblings || rt->should_flush)
4565 		return;
4566 
4567 	/* During lookup routes are evaluated in order, so we need to
4568 	 * make sure upper bounds are assigned from the first sibling
4569 	 * onwards.
4570 	 */
4571 	first = rt6_multipath_first_sibling(rt);
4572 	if (WARN_ON_ONCE(!first))
4573 		return;
4574 
4575 	total = rt6_multipath_total_weight(first);
4576 	rt6_multipath_upper_bound_set(first, total);
4577 }
4578 
4579 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4580 {
4581 	const struct arg_netdev_event *arg = p_arg;
4582 	struct net *net = dev_net(arg->dev);
4583 
4584 	if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4585 	    rt->fib6_nh->fib_nh_dev == arg->dev) {
4586 		rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4587 		fib6_update_sernum_upto_root(net, rt);
4588 		rt6_multipath_rebalance(rt);
4589 	}
4590 
4591 	return 0;
4592 }
4593 
4594 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4595 {
4596 	struct arg_netdev_event arg = {
4597 		.dev = dev,
4598 		{
4599 			.nh_flags = nh_flags,
4600 		},
4601 	};
4602 
4603 	if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4604 		arg.nh_flags |= RTNH_F_LINKDOWN;
4605 
4606 	fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4607 }
4608 
4609 /* only called for fib entries with inline fib6_nh */
4610 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4611 				   const struct net_device *dev)
4612 {
4613 	struct fib6_info *iter;
4614 
4615 	if (rt->fib6_nh->fib_nh_dev == dev)
4616 		return true;
4617 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4618 		if (iter->fib6_nh->fib_nh_dev == dev)
4619 			return true;
4620 
4621 	return false;
4622 }
4623 
4624 static void rt6_multipath_flush(struct fib6_info *rt)
4625 {
4626 	struct fib6_info *iter;
4627 
4628 	rt->should_flush = 1;
4629 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4630 		iter->should_flush = 1;
4631 }
4632 
4633 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4634 					     const struct net_device *down_dev)
4635 {
4636 	struct fib6_info *iter;
4637 	unsigned int dead = 0;
4638 
4639 	if (rt->fib6_nh->fib_nh_dev == down_dev ||
4640 	    rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4641 		dead++;
4642 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4643 		if (iter->fib6_nh->fib_nh_dev == down_dev ||
4644 		    iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4645 			dead++;
4646 
4647 	return dead;
4648 }
4649 
4650 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4651 				       const struct net_device *dev,
4652 				       unsigned char nh_flags)
4653 {
4654 	struct fib6_info *iter;
4655 
4656 	if (rt->fib6_nh->fib_nh_dev == dev)
4657 		rt->fib6_nh->fib_nh_flags |= nh_flags;
4658 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4659 		if (iter->fib6_nh->fib_nh_dev == dev)
4660 			iter->fib6_nh->fib_nh_flags |= nh_flags;
4661 }
4662 
4663 /* called with write lock held for table with rt */
4664 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4665 {
4666 	const struct arg_netdev_event *arg = p_arg;
4667 	const struct net_device *dev = arg->dev;
4668 	struct net *net = dev_net(dev);
4669 
4670 	if (rt == net->ipv6.fib6_null_entry || rt->nh)
4671 		return 0;
4672 
4673 	switch (arg->event) {
4674 	case NETDEV_UNREGISTER:
4675 		return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4676 	case NETDEV_DOWN:
4677 		if (rt->should_flush)
4678 			return -1;
4679 		if (!rt->fib6_nsiblings)
4680 			return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4681 		if (rt6_multipath_uses_dev(rt, dev)) {
4682 			unsigned int count;
4683 
4684 			count = rt6_multipath_dead_count(rt, dev);
4685 			if (rt->fib6_nsiblings + 1 == count) {
4686 				rt6_multipath_flush(rt);
4687 				return -1;
4688 			}
4689 			rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4690 						   RTNH_F_LINKDOWN);
4691 			fib6_update_sernum(net, rt);
4692 			rt6_multipath_rebalance(rt);
4693 		}
4694 		return -2;
4695 	case NETDEV_CHANGE:
4696 		if (rt->fib6_nh->fib_nh_dev != dev ||
4697 		    rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4698 			break;
4699 		rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4700 		rt6_multipath_rebalance(rt);
4701 		break;
4702 	}
4703 
4704 	return 0;
4705 }
4706 
4707 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4708 {
4709 	struct arg_netdev_event arg = {
4710 		.dev = dev,
4711 		{
4712 			.event = event,
4713 		},
4714 	};
4715 	struct net *net = dev_net(dev);
4716 
4717 	if (net->ipv6.sysctl.skip_notify_on_dev_down)
4718 		fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4719 	else
4720 		fib6_clean_all(net, fib6_ifdown, &arg);
4721 }
4722 
4723 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4724 {
4725 	rt6_sync_down_dev(dev, event);
4726 	rt6_uncached_list_flush_dev(dev_net(dev), dev);
4727 	neigh_ifdown(&nd_tbl, dev);
4728 }
4729 
4730 struct rt6_mtu_change_arg {
4731 	struct net_device *dev;
4732 	unsigned int mtu;
4733 	struct fib6_info *f6i;
4734 };
4735 
4736 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4737 {
4738 	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4739 	struct fib6_info *f6i = arg->f6i;
4740 
4741 	/* For administrative MTU increase, there is no way to discover
4742 	 * IPv6 PMTU increase, so PMTU increase should be updated here.
4743 	 * Since RFC 1981 doesn't include administrative MTU increase
4744 	 * update PMTU increase is a MUST. (i.e. jumbo frame)
4745 	 */
4746 	if (nh->fib_nh_dev == arg->dev) {
4747 		struct inet6_dev *idev = __in6_dev_get(arg->dev);
4748 		u32 mtu = f6i->fib6_pmtu;
4749 
4750 		if (mtu >= arg->mtu ||
4751 		    (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4752 			fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4753 
4754 		spin_lock_bh(&rt6_exception_lock);
4755 		rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4756 		spin_unlock_bh(&rt6_exception_lock);
4757 	}
4758 
4759 	return 0;
4760 }
4761 
4762 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4763 {
4764 	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4765 	struct inet6_dev *idev;
4766 
4767 	/* In IPv6 pmtu discovery is not optional,
4768 	   so that RTAX_MTU lock cannot disable it.
4769 	   We still use this lock to block changes
4770 	   caused by addrconf/ndisc.
4771 	*/
4772 
4773 	idev = __in6_dev_get(arg->dev);
4774 	if (!idev)
4775 		return 0;
4776 
4777 	if (fib6_metric_locked(f6i, RTAX_MTU))
4778 		return 0;
4779 
4780 	arg->f6i = f6i;
4781 	if (f6i->nh) {
4782 		/* fib6_nh_mtu_change only returns 0, so this is safe */
4783 		return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4784 						arg);
4785 	}
4786 
4787 	return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4788 }
4789 
4790 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4791 {
4792 	struct rt6_mtu_change_arg arg = {
4793 		.dev = dev,
4794 		.mtu = mtu,
4795 	};
4796 
4797 	fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4798 }
4799 
4800 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4801 	[RTA_UNSPEC]		= { .strict_start_type = RTA_DPORT + 1 },
4802 	[RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
4803 	[RTA_PREFSRC]		= { .len = sizeof(struct in6_addr) },
4804 	[RTA_OIF]               = { .type = NLA_U32 },
4805 	[RTA_IIF]		= { .type = NLA_U32 },
4806 	[RTA_PRIORITY]          = { .type = NLA_U32 },
4807 	[RTA_METRICS]           = { .type = NLA_NESTED },
4808 	[RTA_MULTIPATH]		= { .len = sizeof(struct rtnexthop) },
4809 	[RTA_PREF]              = { .type = NLA_U8 },
4810 	[RTA_ENCAP_TYPE]	= { .type = NLA_U16 },
4811 	[RTA_ENCAP]		= { .type = NLA_NESTED },
4812 	[RTA_EXPIRES]		= { .type = NLA_U32 },
4813 	[RTA_UID]		= { .type = NLA_U32 },
4814 	[RTA_MARK]		= { .type = NLA_U32 },
4815 	[RTA_TABLE]		= { .type = NLA_U32 },
4816 	[RTA_IP_PROTO]		= { .type = NLA_U8 },
4817 	[RTA_SPORT]		= { .type = NLA_U16 },
4818 	[RTA_DPORT]		= { .type = NLA_U16 },
4819 	[RTA_NH_ID]		= { .type = NLA_U32 },
4820 };
4821 
4822 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4823 			      struct fib6_config *cfg,
4824 			      struct netlink_ext_ack *extack)
4825 {
4826 	struct rtmsg *rtm;
4827 	struct nlattr *tb[RTA_MAX+1];
4828 	unsigned int pref;
4829 	int err;
4830 
4831 	err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4832 				     rtm_ipv6_policy, extack);
4833 	if (err < 0)
4834 		goto errout;
4835 
4836 	err = -EINVAL;
4837 	rtm = nlmsg_data(nlh);
4838 
4839 	*cfg = (struct fib6_config){
4840 		.fc_table = rtm->rtm_table,
4841 		.fc_dst_len = rtm->rtm_dst_len,
4842 		.fc_src_len = rtm->rtm_src_len,
4843 		.fc_flags = RTF_UP,
4844 		.fc_protocol = rtm->rtm_protocol,
4845 		.fc_type = rtm->rtm_type,
4846 
4847 		.fc_nlinfo.portid = NETLINK_CB(skb).portid,
4848 		.fc_nlinfo.nlh = nlh,
4849 		.fc_nlinfo.nl_net = sock_net(skb->sk),
4850 	};
4851 
4852 	if (rtm->rtm_type == RTN_UNREACHABLE ||
4853 	    rtm->rtm_type == RTN_BLACKHOLE ||
4854 	    rtm->rtm_type == RTN_PROHIBIT ||
4855 	    rtm->rtm_type == RTN_THROW)
4856 		cfg->fc_flags |= RTF_REJECT;
4857 
4858 	if (rtm->rtm_type == RTN_LOCAL)
4859 		cfg->fc_flags |= RTF_LOCAL;
4860 
4861 	if (rtm->rtm_flags & RTM_F_CLONED)
4862 		cfg->fc_flags |= RTF_CACHE;
4863 
4864 	cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4865 
4866 	if (tb[RTA_NH_ID]) {
4867 		if (tb[RTA_GATEWAY]   || tb[RTA_OIF] ||
4868 		    tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4869 			NL_SET_ERR_MSG(extack,
4870 				       "Nexthop specification and nexthop id are mutually exclusive");
4871 			goto errout;
4872 		}
4873 		cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4874 	}
4875 
4876 	if (tb[RTA_GATEWAY]) {
4877 		cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
4878 		cfg->fc_flags |= RTF_GATEWAY;
4879 	}
4880 	if (tb[RTA_VIA]) {
4881 		NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4882 		goto errout;
4883 	}
4884 
4885 	if (tb[RTA_DST]) {
4886 		int plen = (rtm->rtm_dst_len + 7) >> 3;
4887 
4888 		if (nla_len(tb[RTA_DST]) < plen)
4889 			goto errout;
4890 
4891 		nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
4892 	}
4893 
4894 	if (tb[RTA_SRC]) {
4895 		int plen = (rtm->rtm_src_len + 7) >> 3;
4896 
4897 		if (nla_len(tb[RTA_SRC]) < plen)
4898 			goto errout;
4899 
4900 		nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
4901 	}
4902 
4903 	if (tb[RTA_PREFSRC])
4904 		cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
4905 
4906 	if (tb[RTA_OIF])
4907 		cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4908 
4909 	if (tb[RTA_PRIORITY])
4910 		cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4911 
4912 	if (tb[RTA_METRICS]) {
4913 		cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4914 		cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
4915 	}
4916 
4917 	if (tb[RTA_TABLE])
4918 		cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4919 
4920 	if (tb[RTA_MULTIPATH]) {
4921 		cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4922 		cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
4923 
4924 		err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
4925 						     cfg->fc_mp_len, extack);
4926 		if (err < 0)
4927 			goto errout;
4928 	}
4929 
4930 	if (tb[RTA_PREF]) {
4931 		pref = nla_get_u8(tb[RTA_PREF]);
4932 		if (pref != ICMPV6_ROUTER_PREF_LOW &&
4933 		    pref != ICMPV6_ROUTER_PREF_HIGH)
4934 			pref = ICMPV6_ROUTER_PREF_MEDIUM;
4935 		cfg->fc_flags |= RTF_PREF(pref);
4936 	}
4937 
4938 	if (tb[RTA_ENCAP])
4939 		cfg->fc_encap = tb[RTA_ENCAP];
4940 
4941 	if (tb[RTA_ENCAP_TYPE]) {
4942 		cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4943 
4944 		err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
4945 		if (err < 0)
4946 			goto errout;
4947 	}
4948 
4949 	if (tb[RTA_EXPIRES]) {
4950 		unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4951 
4952 		if (addrconf_finite_timeout(timeout)) {
4953 			cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4954 			cfg->fc_flags |= RTF_EXPIRES;
4955 		}
4956 	}
4957 
4958 	err = 0;
4959 errout:
4960 	return err;
4961 }
4962 
4963 struct rt6_nh {
4964 	struct fib6_info *fib6_info;
4965 	struct fib6_config r_cfg;
4966 	struct list_head next;
4967 };
4968 
4969 static int ip6_route_info_append(struct net *net,
4970 				 struct list_head *rt6_nh_list,
4971 				 struct fib6_info *rt,
4972 				 struct fib6_config *r_cfg)
4973 {
4974 	struct rt6_nh *nh;
4975 	int err = -EEXIST;
4976 
4977 	list_for_each_entry(nh, rt6_nh_list, next) {
4978 		/* check if fib6_info already exists */
4979 		if (rt6_duplicate_nexthop(nh->fib6_info, rt))
4980 			return err;
4981 	}
4982 
4983 	nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4984 	if (!nh)
4985 		return -ENOMEM;
4986 	nh->fib6_info = rt;
4987 	memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4988 	list_add_tail(&nh->next, rt6_nh_list);
4989 
4990 	return 0;
4991 }
4992 
4993 static void ip6_route_mpath_notify(struct fib6_info *rt,
4994 				   struct fib6_info *rt_last,
4995 				   struct nl_info *info,
4996 				   __u16 nlflags)
4997 {
4998 	/* if this is an APPEND route, then rt points to the first route
4999 	 * inserted and rt_last points to last route inserted. Userspace
5000 	 * wants a consistent dump of the route which starts at the first
5001 	 * nexthop. Since sibling routes are always added at the end of
5002 	 * the list, find the first sibling of the last route appended
5003 	 */
5004 	if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5005 		rt = list_first_entry(&rt_last->fib6_siblings,
5006 				      struct fib6_info,
5007 				      fib6_siblings);
5008 	}
5009 
5010 	if (rt)
5011 		inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5012 }
5013 
5014 static int ip6_route_multipath_add(struct fib6_config *cfg,
5015 				   struct netlink_ext_ack *extack)
5016 {
5017 	struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5018 	struct nl_info *info = &cfg->fc_nlinfo;
5019 	enum fib_event_type event_type;
5020 	struct fib6_config r_cfg;
5021 	struct rtnexthop *rtnh;
5022 	struct fib6_info *rt;
5023 	struct rt6_nh *err_nh;
5024 	struct rt6_nh *nh, *nh_safe;
5025 	__u16 nlflags;
5026 	int remaining;
5027 	int attrlen;
5028 	int err = 1;
5029 	int nhn = 0;
5030 	int replace = (cfg->fc_nlinfo.nlh &&
5031 		       (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5032 	LIST_HEAD(rt6_nh_list);
5033 
5034 	nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5035 	if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5036 		nlflags |= NLM_F_APPEND;
5037 
5038 	remaining = cfg->fc_mp_len;
5039 	rtnh = (struct rtnexthop *)cfg->fc_mp;
5040 
5041 	/* Parse a Multipath Entry and build a list (rt6_nh_list) of
5042 	 * fib6_info structs per nexthop
5043 	 */
5044 	while (rtnh_ok(rtnh, remaining)) {
5045 		memcpy(&r_cfg, cfg, sizeof(*cfg));
5046 		if (rtnh->rtnh_ifindex)
5047 			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5048 
5049 		attrlen = rtnh_attrlen(rtnh);
5050 		if (attrlen > 0) {
5051 			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5052 
5053 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5054 			if (nla) {
5055 				r_cfg.fc_gateway = nla_get_in6_addr(nla);
5056 				r_cfg.fc_flags |= RTF_GATEWAY;
5057 			}
5058 			r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5059 			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5060 			if (nla)
5061 				r_cfg.fc_encap_type = nla_get_u16(nla);
5062 		}
5063 
5064 		r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5065 		rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5066 		if (IS_ERR(rt)) {
5067 			err = PTR_ERR(rt);
5068 			rt = NULL;
5069 			goto cleanup;
5070 		}
5071 		if (!rt6_qualify_for_ecmp(rt)) {
5072 			err = -EINVAL;
5073 			NL_SET_ERR_MSG(extack,
5074 				       "Device only routes can not be added for IPv6 using the multipath API.");
5075 			fib6_info_release(rt);
5076 			goto cleanup;
5077 		}
5078 
5079 		rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5080 
5081 		err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5082 					    rt, &r_cfg);
5083 		if (err) {
5084 			fib6_info_release(rt);
5085 			goto cleanup;
5086 		}
5087 
5088 		rtnh = rtnh_next(rtnh, &remaining);
5089 	}
5090 
5091 	if (list_empty(&rt6_nh_list)) {
5092 		NL_SET_ERR_MSG(extack,
5093 			       "Invalid nexthop configuration - no valid nexthops");
5094 		return -EINVAL;
5095 	}
5096 
5097 	/* for add and replace send one notification with all nexthops.
5098 	 * Skip the notification in fib6_add_rt2node and send one with
5099 	 * the full route when done
5100 	 */
5101 	info->skip_notify = 1;
5102 
5103 	/* For add and replace, send one notification with all nexthops. For
5104 	 * append, send one notification with all appended nexthops.
5105 	 */
5106 	info->skip_notify_kernel = 1;
5107 
5108 	err_nh = NULL;
5109 	list_for_each_entry(nh, &rt6_nh_list, next) {
5110 		err = __ip6_ins_rt(nh->fib6_info, info, extack);
5111 		fib6_info_release(nh->fib6_info);
5112 
5113 		if (!err) {
5114 			/* save reference to last route successfully inserted */
5115 			rt_last = nh->fib6_info;
5116 
5117 			/* save reference to first route for notification */
5118 			if (!rt_notif)
5119 				rt_notif = nh->fib6_info;
5120 		}
5121 
5122 		/* nh->fib6_info is used or freed at this point, reset to NULL*/
5123 		nh->fib6_info = NULL;
5124 		if (err) {
5125 			if (replace && nhn)
5126 				NL_SET_ERR_MSG_MOD(extack,
5127 						   "multipath route replace failed (check consistency of installed routes)");
5128 			err_nh = nh;
5129 			goto add_errout;
5130 		}
5131 
5132 		/* Because each route is added like a single route we remove
5133 		 * these flags after the first nexthop: if there is a collision,
5134 		 * we have already failed to add the first nexthop:
5135 		 * fib6_add_rt2node() has rejected it; when replacing, old
5136 		 * nexthops have been replaced by first new, the rest should
5137 		 * be added to it.
5138 		 */
5139 		cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5140 						     NLM_F_REPLACE);
5141 		nhn++;
5142 	}
5143 
5144 	event_type = replace ? FIB_EVENT_ENTRY_REPLACE : FIB_EVENT_ENTRY_ADD;
5145 	err = call_fib6_multipath_entry_notifiers(info->nl_net, event_type,
5146 						  rt_notif, nhn - 1, extack);
5147 	if (err) {
5148 		/* Delete all the siblings that were just added */
5149 		err_nh = NULL;
5150 		goto add_errout;
5151 	}
5152 
5153 	/* success ... tell user about new route */
5154 	ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5155 	goto cleanup;
5156 
5157 add_errout:
5158 	/* send notification for routes that were added so that
5159 	 * the delete notifications sent by ip6_route_del are
5160 	 * coherent
5161 	 */
5162 	if (rt_notif)
5163 		ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5164 
5165 	/* Delete routes that were already added */
5166 	list_for_each_entry(nh, &rt6_nh_list, next) {
5167 		if (err_nh == nh)
5168 			break;
5169 		ip6_route_del(&nh->r_cfg, extack);
5170 	}
5171 
5172 cleanup:
5173 	list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5174 		if (nh->fib6_info)
5175 			fib6_info_release(nh->fib6_info);
5176 		list_del(&nh->next);
5177 		kfree(nh);
5178 	}
5179 
5180 	return err;
5181 }
5182 
5183 static int ip6_route_multipath_del(struct fib6_config *cfg,
5184 				   struct netlink_ext_ack *extack)
5185 {
5186 	struct fib6_config r_cfg;
5187 	struct rtnexthop *rtnh;
5188 	int remaining;
5189 	int attrlen;
5190 	int err = 1, last_err = 0;
5191 
5192 	remaining = cfg->fc_mp_len;
5193 	rtnh = (struct rtnexthop *)cfg->fc_mp;
5194 
5195 	/* Parse a Multipath Entry */
5196 	while (rtnh_ok(rtnh, remaining)) {
5197 		memcpy(&r_cfg, cfg, sizeof(*cfg));
5198 		if (rtnh->rtnh_ifindex)
5199 			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5200 
5201 		attrlen = rtnh_attrlen(rtnh);
5202 		if (attrlen > 0) {
5203 			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5204 
5205 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5206 			if (nla) {
5207 				nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5208 				r_cfg.fc_flags |= RTF_GATEWAY;
5209 			}
5210 		}
5211 		err = ip6_route_del(&r_cfg, extack);
5212 		if (err)
5213 			last_err = err;
5214 
5215 		rtnh = rtnh_next(rtnh, &remaining);
5216 	}
5217 
5218 	return last_err;
5219 }
5220 
5221 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5222 			      struct netlink_ext_ack *extack)
5223 {
5224 	struct fib6_config cfg;
5225 	int err;
5226 
5227 	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5228 	if (err < 0)
5229 		return err;
5230 
5231 	if (cfg.fc_nh_id &&
5232 	    !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5233 		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5234 		return -EINVAL;
5235 	}
5236 
5237 	if (cfg.fc_mp)
5238 		return ip6_route_multipath_del(&cfg, extack);
5239 	else {
5240 		cfg.fc_delete_all_nh = 1;
5241 		return ip6_route_del(&cfg, extack);
5242 	}
5243 }
5244 
5245 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5246 			      struct netlink_ext_ack *extack)
5247 {
5248 	struct fib6_config cfg;
5249 	int err;
5250 
5251 	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5252 	if (err < 0)
5253 		return err;
5254 
5255 	if (cfg.fc_metric == 0)
5256 		cfg.fc_metric = IP6_RT_PRIO_USER;
5257 
5258 	if (cfg.fc_mp)
5259 		return ip6_route_multipath_add(&cfg, extack);
5260 	else
5261 		return ip6_route_add(&cfg, GFP_KERNEL, extack);
5262 }
5263 
5264 /* add the overhead of this fib6_nh to nexthop_len */
5265 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5266 {
5267 	int *nexthop_len = arg;
5268 
5269 	*nexthop_len += nla_total_size(0)	 /* RTA_MULTIPATH */
5270 		     + NLA_ALIGN(sizeof(struct rtnexthop))
5271 		     + nla_total_size(16); /* RTA_GATEWAY */
5272 
5273 	if (nh->fib_nh_lws) {
5274 		/* RTA_ENCAP_TYPE */
5275 		*nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5276 		/* RTA_ENCAP */
5277 		*nexthop_len += nla_total_size(2);
5278 	}
5279 
5280 	return 0;
5281 }
5282 
5283 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5284 {
5285 	int nexthop_len;
5286 
5287 	if (f6i->nh) {
5288 		nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5289 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5290 					 &nexthop_len);
5291 	} else {
5292 		struct fib6_nh *nh = f6i->fib6_nh;
5293 
5294 		nexthop_len = 0;
5295 		if (f6i->fib6_nsiblings) {
5296 			nexthop_len = nla_total_size(0)	 /* RTA_MULTIPATH */
5297 				    + NLA_ALIGN(sizeof(struct rtnexthop))
5298 				    + nla_total_size(16) /* RTA_GATEWAY */
5299 				    + lwtunnel_get_encap_size(nh->fib_nh_lws);
5300 
5301 			nexthop_len *= f6i->fib6_nsiblings;
5302 		}
5303 		nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5304 	}
5305 
5306 	return NLMSG_ALIGN(sizeof(struct rtmsg))
5307 	       + nla_total_size(16) /* RTA_SRC */
5308 	       + nla_total_size(16) /* RTA_DST */
5309 	       + nla_total_size(16) /* RTA_GATEWAY */
5310 	       + nla_total_size(16) /* RTA_PREFSRC */
5311 	       + nla_total_size(4) /* RTA_TABLE */
5312 	       + nla_total_size(4) /* RTA_IIF */
5313 	       + nla_total_size(4) /* RTA_OIF */
5314 	       + nla_total_size(4) /* RTA_PRIORITY */
5315 	       + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5316 	       + nla_total_size(sizeof(struct rta_cacheinfo))
5317 	       + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5318 	       + nla_total_size(1) /* RTA_PREF */
5319 	       + nexthop_len;
5320 }
5321 
5322 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5323 				 unsigned char *flags)
5324 {
5325 	if (nexthop_is_multipath(nh)) {
5326 		struct nlattr *mp;
5327 
5328 		mp = nla_nest_start(skb, RTA_MULTIPATH);
5329 		if (!mp)
5330 			goto nla_put_failure;
5331 
5332 		if (nexthop_mpath_fill_node(skb, nh))
5333 			goto nla_put_failure;
5334 
5335 		nla_nest_end(skb, mp);
5336 	} else {
5337 		struct fib6_nh *fib6_nh;
5338 
5339 		fib6_nh = nexthop_fib6_nh(nh);
5340 		if (fib_nexthop_info(skb, &fib6_nh->nh_common,
5341 				     flags, false) < 0)
5342 			goto nla_put_failure;
5343 	}
5344 
5345 	return 0;
5346 
5347 nla_put_failure:
5348 	return -EMSGSIZE;
5349 }
5350 
5351 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5352 			 struct fib6_info *rt, struct dst_entry *dst,
5353 			 struct in6_addr *dest, struct in6_addr *src,
5354 			 int iif, int type, u32 portid, u32 seq,
5355 			 unsigned int flags)
5356 {
5357 	struct rt6_info *rt6 = (struct rt6_info *)dst;
5358 	struct rt6key *rt6_dst, *rt6_src;
5359 	u32 *pmetrics, table, rt6_flags;
5360 	unsigned char nh_flags = 0;
5361 	struct nlmsghdr *nlh;
5362 	struct rtmsg *rtm;
5363 	long expires = 0;
5364 
5365 	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5366 	if (!nlh)
5367 		return -EMSGSIZE;
5368 
5369 	if (rt6) {
5370 		rt6_dst = &rt6->rt6i_dst;
5371 		rt6_src = &rt6->rt6i_src;
5372 		rt6_flags = rt6->rt6i_flags;
5373 	} else {
5374 		rt6_dst = &rt->fib6_dst;
5375 		rt6_src = &rt->fib6_src;
5376 		rt6_flags = rt->fib6_flags;
5377 	}
5378 
5379 	rtm = nlmsg_data(nlh);
5380 	rtm->rtm_family = AF_INET6;
5381 	rtm->rtm_dst_len = rt6_dst->plen;
5382 	rtm->rtm_src_len = rt6_src->plen;
5383 	rtm->rtm_tos = 0;
5384 	if (rt->fib6_table)
5385 		table = rt->fib6_table->tb6_id;
5386 	else
5387 		table = RT6_TABLE_UNSPEC;
5388 	rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5389 	if (nla_put_u32(skb, RTA_TABLE, table))
5390 		goto nla_put_failure;
5391 
5392 	rtm->rtm_type = rt->fib6_type;
5393 	rtm->rtm_flags = 0;
5394 	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5395 	rtm->rtm_protocol = rt->fib6_protocol;
5396 
5397 	if (rt6_flags & RTF_CACHE)
5398 		rtm->rtm_flags |= RTM_F_CLONED;
5399 
5400 	if (dest) {
5401 		if (nla_put_in6_addr(skb, RTA_DST, dest))
5402 			goto nla_put_failure;
5403 		rtm->rtm_dst_len = 128;
5404 	} else if (rtm->rtm_dst_len)
5405 		if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5406 			goto nla_put_failure;
5407 #ifdef CONFIG_IPV6_SUBTREES
5408 	if (src) {
5409 		if (nla_put_in6_addr(skb, RTA_SRC, src))
5410 			goto nla_put_failure;
5411 		rtm->rtm_src_len = 128;
5412 	} else if (rtm->rtm_src_len &&
5413 		   nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5414 		goto nla_put_failure;
5415 #endif
5416 	if (iif) {
5417 #ifdef CONFIG_IPV6_MROUTE
5418 		if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5419 			int err = ip6mr_get_route(net, skb, rtm, portid);
5420 
5421 			if (err == 0)
5422 				return 0;
5423 			if (err < 0)
5424 				goto nla_put_failure;
5425 		} else
5426 #endif
5427 			if (nla_put_u32(skb, RTA_IIF, iif))
5428 				goto nla_put_failure;
5429 	} else if (dest) {
5430 		struct in6_addr saddr_buf;
5431 		if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5432 		    nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5433 			goto nla_put_failure;
5434 	}
5435 
5436 	if (rt->fib6_prefsrc.plen) {
5437 		struct in6_addr saddr_buf;
5438 		saddr_buf = rt->fib6_prefsrc.addr;
5439 		if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5440 			goto nla_put_failure;
5441 	}
5442 
5443 	pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5444 	if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5445 		goto nla_put_failure;
5446 
5447 	if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5448 		goto nla_put_failure;
5449 
5450 	/* For multipath routes, walk the siblings list and add
5451 	 * each as a nexthop within RTA_MULTIPATH.
5452 	 */
5453 	if (rt6) {
5454 		if (rt6_flags & RTF_GATEWAY &&
5455 		    nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5456 			goto nla_put_failure;
5457 
5458 		if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5459 			goto nla_put_failure;
5460 	} else if (rt->fib6_nsiblings) {
5461 		struct fib6_info *sibling, *next_sibling;
5462 		struct nlattr *mp;
5463 
5464 		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5465 		if (!mp)
5466 			goto nla_put_failure;
5467 
5468 		if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5469 				    rt->fib6_nh->fib_nh_weight) < 0)
5470 			goto nla_put_failure;
5471 
5472 		list_for_each_entry_safe(sibling, next_sibling,
5473 					 &rt->fib6_siblings, fib6_siblings) {
5474 			if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5475 					    sibling->fib6_nh->fib_nh_weight) < 0)
5476 				goto nla_put_failure;
5477 		}
5478 
5479 		nla_nest_end(skb, mp);
5480 	} else if (rt->nh) {
5481 		if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5482 			goto nla_put_failure;
5483 
5484 		if (nexthop_is_blackhole(rt->nh))
5485 			rtm->rtm_type = RTN_BLACKHOLE;
5486 
5487 		if (rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5488 			goto nla_put_failure;
5489 
5490 		rtm->rtm_flags |= nh_flags;
5491 	} else {
5492 		if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common,
5493 				     &nh_flags, false) < 0)
5494 			goto nla_put_failure;
5495 
5496 		rtm->rtm_flags |= nh_flags;
5497 	}
5498 
5499 	if (rt6_flags & RTF_EXPIRES) {
5500 		expires = dst ? dst->expires : rt->expires;
5501 		expires -= jiffies;
5502 	}
5503 
5504 	if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5505 		goto nla_put_failure;
5506 
5507 	if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5508 		goto nla_put_failure;
5509 
5510 
5511 	nlmsg_end(skb, nlh);
5512 	return 0;
5513 
5514 nla_put_failure:
5515 	nlmsg_cancel(skb, nlh);
5516 	return -EMSGSIZE;
5517 }
5518 
5519 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5520 {
5521 	const struct net_device *dev = arg;
5522 
5523 	if (nh->fib_nh_dev == dev)
5524 		return 1;
5525 
5526 	return 0;
5527 }
5528 
5529 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5530 			       const struct net_device *dev)
5531 {
5532 	if (f6i->nh) {
5533 		struct net_device *_dev = (struct net_device *)dev;
5534 
5535 		return !!nexthop_for_each_fib6_nh(f6i->nh,
5536 						  fib6_info_nh_uses_dev,
5537 						  _dev);
5538 	}
5539 
5540 	if (f6i->fib6_nh->fib_nh_dev == dev)
5541 		return true;
5542 
5543 	if (f6i->fib6_nsiblings) {
5544 		struct fib6_info *sibling, *next_sibling;
5545 
5546 		list_for_each_entry_safe(sibling, next_sibling,
5547 					 &f6i->fib6_siblings, fib6_siblings) {
5548 			if (sibling->fib6_nh->fib_nh_dev == dev)
5549 				return true;
5550 		}
5551 	}
5552 
5553 	return false;
5554 }
5555 
5556 struct fib6_nh_exception_dump_walker {
5557 	struct rt6_rtnl_dump_arg *dump;
5558 	struct fib6_info *rt;
5559 	unsigned int flags;
5560 	unsigned int skip;
5561 	unsigned int count;
5562 };
5563 
5564 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5565 {
5566 	struct fib6_nh_exception_dump_walker *w = arg;
5567 	struct rt6_rtnl_dump_arg *dump = w->dump;
5568 	struct rt6_exception_bucket *bucket;
5569 	struct rt6_exception *rt6_ex;
5570 	int i, err;
5571 
5572 	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5573 	if (!bucket)
5574 		return 0;
5575 
5576 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5577 		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5578 			if (w->skip) {
5579 				w->skip--;
5580 				continue;
5581 			}
5582 
5583 			/* Expiration of entries doesn't bump sernum, insertion
5584 			 * does. Removal is triggered by insertion, so we can
5585 			 * rely on the fact that if entries change between two
5586 			 * partial dumps, this node is scanned again completely,
5587 			 * see rt6_insert_exception() and fib6_dump_table().
5588 			 *
5589 			 * Count expired entries we go through as handled
5590 			 * entries that we'll skip next time, in case of partial
5591 			 * node dump. Otherwise, if entries expire meanwhile,
5592 			 * we'll skip the wrong amount.
5593 			 */
5594 			if (rt6_check_expired(rt6_ex->rt6i)) {
5595 				w->count++;
5596 				continue;
5597 			}
5598 
5599 			err = rt6_fill_node(dump->net, dump->skb, w->rt,
5600 					    &rt6_ex->rt6i->dst, NULL, NULL, 0,
5601 					    RTM_NEWROUTE,
5602 					    NETLINK_CB(dump->cb->skb).portid,
5603 					    dump->cb->nlh->nlmsg_seq, w->flags);
5604 			if (err)
5605 				return err;
5606 
5607 			w->count++;
5608 		}
5609 		bucket++;
5610 	}
5611 
5612 	return 0;
5613 }
5614 
5615 /* Return -1 if done with node, number of handled routes on partial dump */
5616 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5617 {
5618 	struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5619 	struct fib_dump_filter *filter = &arg->filter;
5620 	unsigned int flags = NLM_F_MULTI;
5621 	struct net *net = arg->net;
5622 	int count = 0;
5623 
5624 	if (rt == net->ipv6.fib6_null_entry)
5625 		return -1;
5626 
5627 	if ((filter->flags & RTM_F_PREFIX) &&
5628 	    !(rt->fib6_flags & RTF_PREFIX_RT)) {
5629 		/* success since this is not a prefix route */
5630 		return -1;
5631 	}
5632 	if (filter->filter_set &&
5633 	    ((filter->rt_type  && rt->fib6_type != filter->rt_type) ||
5634 	     (filter->dev      && !fib6_info_uses_dev(rt, filter->dev)) ||
5635 	     (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5636 		return -1;
5637 	}
5638 
5639 	if (filter->filter_set ||
5640 	    !filter->dump_routes || !filter->dump_exceptions) {
5641 		flags |= NLM_F_DUMP_FILTERED;
5642 	}
5643 
5644 	if (filter->dump_routes) {
5645 		if (skip) {
5646 			skip--;
5647 		} else {
5648 			if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5649 					  0, RTM_NEWROUTE,
5650 					  NETLINK_CB(arg->cb->skb).portid,
5651 					  arg->cb->nlh->nlmsg_seq, flags)) {
5652 				return 0;
5653 			}
5654 			count++;
5655 		}
5656 	}
5657 
5658 	if (filter->dump_exceptions) {
5659 		struct fib6_nh_exception_dump_walker w = { .dump = arg,
5660 							   .rt = rt,
5661 							   .flags = flags,
5662 							   .skip = skip,
5663 							   .count = 0 };
5664 		int err;
5665 
5666 		rcu_read_lock();
5667 		if (rt->nh) {
5668 			err = nexthop_for_each_fib6_nh(rt->nh,
5669 						       rt6_nh_dump_exceptions,
5670 						       &w);
5671 		} else {
5672 			err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5673 		}
5674 		rcu_read_unlock();
5675 
5676 		if (err)
5677 			return count += w.count;
5678 	}
5679 
5680 	return -1;
5681 }
5682 
5683 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5684 					const struct nlmsghdr *nlh,
5685 					struct nlattr **tb,
5686 					struct netlink_ext_ack *extack)
5687 {
5688 	struct rtmsg *rtm;
5689 	int i, err;
5690 
5691 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5692 		NL_SET_ERR_MSG_MOD(extack,
5693 				   "Invalid header for get route request");
5694 		return -EINVAL;
5695 	}
5696 
5697 	if (!netlink_strict_get_check(skb))
5698 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5699 					      rtm_ipv6_policy, extack);
5700 
5701 	rtm = nlmsg_data(nlh);
5702 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5703 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5704 	    rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5705 	    rtm->rtm_type) {
5706 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5707 		return -EINVAL;
5708 	}
5709 	if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5710 		NL_SET_ERR_MSG_MOD(extack,
5711 				   "Invalid flags for get route request");
5712 		return -EINVAL;
5713 	}
5714 
5715 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5716 					    rtm_ipv6_policy, extack);
5717 	if (err)
5718 		return err;
5719 
5720 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5721 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5722 		NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5723 		return -EINVAL;
5724 	}
5725 
5726 	for (i = 0; i <= RTA_MAX; i++) {
5727 		if (!tb[i])
5728 			continue;
5729 
5730 		switch (i) {
5731 		case RTA_SRC:
5732 		case RTA_DST:
5733 		case RTA_IIF:
5734 		case RTA_OIF:
5735 		case RTA_MARK:
5736 		case RTA_UID:
5737 		case RTA_SPORT:
5738 		case RTA_DPORT:
5739 		case RTA_IP_PROTO:
5740 			break;
5741 		default:
5742 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5743 			return -EINVAL;
5744 		}
5745 	}
5746 
5747 	return 0;
5748 }
5749 
5750 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5751 			      struct netlink_ext_ack *extack)
5752 {
5753 	struct net *net = sock_net(in_skb->sk);
5754 	struct nlattr *tb[RTA_MAX+1];
5755 	int err, iif = 0, oif = 0;
5756 	struct fib6_info *from;
5757 	struct dst_entry *dst;
5758 	struct rt6_info *rt;
5759 	struct sk_buff *skb;
5760 	struct rtmsg *rtm;
5761 	struct flowi6 fl6 = {};
5762 	bool fibmatch;
5763 
5764 	err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
5765 	if (err < 0)
5766 		goto errout;
5767 
5768 	err = -EINVAL;
5769 	rtm = nlmsg_data(nlh);
5770 	fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
5771 	fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
5772 
5773 	if (tb[RTA_SRC]) {
5774 		if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5775 			goto errout;
5776 
5777 		fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
5778 	}
5779 
5780 	if (tb[RTA_DST]) {
5781 		if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5782 			goto errout;
5783 
5784 		fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
5785 	}
5786 
5787 	if (tb[RTA_IIF])
5788 		iif = nla_get_u32(tb[RTA_IIF]);
5789 
5790 	if (tb[RTA_OIF])
5791 		oif = nla_get_u32(tb[RTA_OIF]);
5792 
5793 	if (tb[RTA_MARK])
5794 		fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5795 
5796 	if (tb[RTA_UID])
5797 		fl6.flowi6_uid = make_kuid(current_user_ns(),
5798 					   nla_get_u32(tb[RTA_UID]));
5799 	else
5800 		fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5801 
5802 	if (tb[RTA_SPORT])
5803 		fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5804 
5805 	if (tb[RTA_DPORT])
5806 		fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5807 
5808 	if (tb[RTA_IP_PROTO]) {
5809 		err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5810 						  &fl6.flowi6_proto, AF_INET6,
5811 						  extack);
5812 		if (err)
5813 			goto errout;
5814 	}
5815 
5816 	if (iif) {
5817 		struct net_device *dev;
5818 		int flags = 0;
5819 
5820 		rcu_read_lock();
5821 
5822 		dev = dev_get_by_index_rcu(net, iif);
5823 		if (!dev) {
5824 			rcu_read_unlock();
5825 			err = -ENODEV;
5826 			goto errout;
5827 		}
5828 
5829 		fl6.flowi6_iif = iif;
5830 
5831 		if (!ipv6_addr_any(&fl6.saddr))
5832 			flags |= RT6_LOOKUP_F_HAS_SADDR;
5833 
5834 		dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
5835 
5836 		rcu_read_unlock();
5837 	} else {
5838 		fl6.flowi6_oif = oif;
5839 
5840 		dst = ip6_route_output(net, NULL, &fl6);
5841 	}
5842 
5843 
5844 	rt = container_of(dst, struct rt6_info, dst);
5845 	if (rt->dst.error) {
5846 		err = rt->dst.error;
5847 		ip6_rt_put(rt);
5848 		goto errout;
5849 	}
5850 
5851 	if (rt == net->ipv6.ip6_null_entry) {
5852 		err = rt->dst.error;
5853 		ip6_rt_put(rt);
5854 		goto errout;
5855 	}
5856 
5857 	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
5858 	if (!skb) {
5859 		ip6_rt_put(rt);
5860 		err = -ENOBUFS;
5861 		goto errout;
5862 	}
5863 
5864 	skb_dst_set(skb, &rt->dst);
5865 
5866 	rcu_read_lock();
5867 	from = rcu_dereference(rt->from);
5868 	if (from) {
5869 		if (fibmatch)
5870 			err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5871 					    iif, RTM_NEWROUTE,
5872 					    NETLINK_CB(in_skb).portid,
5873 					    nlh->nlmsg_seq, 0);
5874 		else
5875 			err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5876 					    &fl6.saddr, iif, RTM_NEWROUTE,
5877 					    NETLINK_CB(in_skb).portid,
5878 					    nlh->nlmsg_seq, 0);
5879 	} else {
5880 		err = -ENETUNREACH;
5881 	}
5882 	rcu_read_unlock();
5883 
5884 	if (err < 0) {
5885 		kfree_skb(skb);
5886 		goto errout;
5887 	}
5888 
5889 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5890 errout:
5891 	return err;
5892 }
5893 
5894 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
5895 		     unsigned int nlm_flags)
5896 {
5897 	struct sk_buff *skb;
5898 	struct net *net = info->nl_net;
5899 	u32 seq;
5900 	int err;
5901 
5902 	err = -ENOBUFS;
5903 	seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5904 
5905 	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5906 	if (!skb)
5907 		goto errout;
5908 
5909 	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5910 			    event, info->portid, seq, nlm_flags);
5911 	if (err < 0) {
5912 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5913 		WARN_ON(err == -EMSGSIZE);
5914 		kfree_skb(skb);
5915 		goto errout;
5916 	}
5917 	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5918 		    info->nlh, gfp_any());
5919 	return;
5920 errout:
5921 	if (err < 0)
5922 		rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5923 }
5924 
5925 void fib6_rt_update(struct net *net, struct fib6_info *rt,
5926 		    struct nl_info *info)
5927 {
5928 	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5929 	struct sk_buff *skb;
5930 	int err = -ENOBUFS;
5931 
5932 	/* call_fib6_entry_notifiers will be removed when in-kernel notifier
5933 	 * is implemented and supported for nexthop objects
5934 	 */
5935 	call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
5936 
5937 	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5938 	if (!skb)
5939 		goto errout;
5940 
5941 	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5942 			    RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
5943 	if (err < 0) {
5944 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5945 		WARN_ON(err == -EMSGSIZE);
5946 		kfree_skb(skb);
5947 		goto errout;
5948 	}
5949 	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5950 		    info->nlh, gfp_any());
5951 	return;
5952 errout:
5953 	if (err < 0)
5954 		rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5955 }
5956 
5957 static int ip6_route_dev_notify(struct notifier_block *this,
5958 				unsigned long event, void *ptr)
5959 {
5960 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5961 	struct net *net = dev_net(dev);
5962 
5963 	if (!(dev->flags & IFF_LOOPBACK))
5964 		return NOTIFY_OK;
5965 
5966 	if (event == NETDEV_REGISTER) {
5967 		net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
5968 		net->ipv6.ip6_null_entry->dst.dev = dev;
5969 		net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5970 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5971 		net->ipv6.ip6_prohibit_entry->dst.dev = dev;
5972 		net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
5973 		net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
5974 		net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
5975 #endif
5976 	 } else if (event == NETDEV_UNREGISTER &&
5977 		    dev->reg_state != NETREG_UNREGISTERED) {
5978 		/* NETDEV_UNREGISTER could be fired for multiple times by
5979 		 * netdev_wait_allrefs(). Make sure we only call this once.
5980 		 */
5981 		in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
5982 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5983 		in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5984 		in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
5985 #endif
5986 	}
5987 
5988 	return NOTIFY_OK;
5989 }
5990 
5991 /*
5992  *	/proc
5993  */
5994 
5995 #ifdef CONFIG_PROC_FS
5996 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5997 {
5998 	struct net *net = (struct net *)seq->private;
5999 	seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6000 		   net->ipv6.rt6_stats->fib_nodes,
6001 		   net->ipv6.rt6_stats->fib_route_nodes,
6002 		   atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6003 		   net->ipv6.rt6_stats->fib_rt_entries,
6004 		   net->ipv6.rt6_stats->fib_rt_cache,
6005 		   dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6006 		   net->ipv6.rt6_stats->fib_discarded_routes);
6007 
6008 	return 0;
6009 }
6010 #endif	/* CONFIG_PROC_FS */
6011 
6012 #ifdef CONFIG_SYSCTL
6013 
6014 static
6015 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6016 			      void __user *buffer, size_t *lenp, loff_t *ppos)
6017 {
6018 	struct net *net;
6019 	int delay;
6020 	int ret;
6021 	if (!write)
6022 		return -EINVAL;
6023 
6024 	net = (struct net *)ctl->extra1;
6025 	delay = net->ipv6.sysctl.flush_delay;
6026 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6027 	if (ret)
6028 		return ret;
6029 
6030 	fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6031 	return 0;
6032 }
6033 
6034 static struct ctl_table ipv6_route_table_template[] = {
6035 	{
6036 		.procname	=	"flush",
6037 		.data		=	&init_net.ipv6.sysctl.flush_delay,
6038 		.maxlen		=	sizeof(int),
6039 		.mode		=	0200,
6040 		.proc_handler	=	ipv6_sysctl_rtcache_flush
6041 	},
6042 	{
6043 		.procname	=	"gc_thresh",
6044 		.data		=	&ip6_dst_ops_template.gc_thresh,
6045 		.maxlen		=	sizeof(int),
6046 		.mode		=	0644,
6047 		.proc_handler	=	proc_dointvec,
6048 	},
6049 	{
6050 		.procname	=	"max_size",
6051 		.data		=	&init_net.ipv6.sysctl.ip6_rt_max_size,
6052 		.maxlen		=	sizeof(int),
6053 		.mode		=	0644,
6054 		.proc_handler	=	proc_dointvec,
6055 	},
6056 	{
6057 		.procname	=	"gc_min_interval",
6058 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6059 		.maxlen		=	sizeof(int),
6060 		.mode		=	0644,
6061 		.proc_handler	=	proc_dointvec_jiffies,
6062 	},
6063 	{
6064 		.procname	=	"gc_timeout",
6065 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6066 		.maxlen		=	sizeof(int),
6067 		.mode		=	0644,
6068 		.proc_handler	=	proc_dointvec_jiffies,
6069 	},
6070 	{
6071 		.procname	=	"gc_interval",
6072 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_interval,
6073 		.maxlen		=	sizeof(int),
6074 		.mode		=	0644,
6075 		.proc_handler	=	proc_dointvec_jiffies,
6076 	},
6077 	{
6078 		.procname	=	"gc_elasticity",
6079 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6080 		.maxlen		=	sizeof(int),
6081 		.mode		=	0644,
6082 		.proc_handler	=	proc_dointvec,
6083 	},
6084 	{
6085 		.procname	=	"mtu_expires",
6086 		.data		=	&init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6087 		.maxlen		=	sizeof(int),
6088 		.mode		=	0644,
6089 		.proc_handler	=	proc_dointvec_jiffies,
6090 	},
6091 	{
6092 		.procname	=	"min_adv_mss",
6093 		.data		=	&init_net.ipv6.sysctl.ip6_rt_min_advmss,
6094 		.maxlen		=	sizeof(int),
6095 		.mode		=	0644,
6096 		.proc_handler	=	proc_dointvec,
6097 	},
6098 	{
6099 		.procname	=	"gc_min_interval_ms",
6100 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6101 		.maxlen		=	sizeof(int),
6102 		.mode		=	0644,
6103 		.proc_handler	=	proc_dointvec_ms_jiffies,
6104 	},
6105 	{
6106 		.procname	=	"skip_notify_on_dev_down",
6107 		.data		=	&init_net.ipv6.sysctl.skip_notify_on_dev_down,
6108 		.maxlen		=	sizeof(int),
6109 		.mode		=	0644,
6110 		.proc_handler	=	proc_dointvec_minmax,
6111 		.extra1		=	SYSCTL_ZERO,
6112 		.extra2		=	SYSCTL_ONE,
6113 	},
6114 	{ }
6115 };
6116 
6117 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6118 {
6119 	struct ctl_table *table;
6120 
6121 	table = kmemdup(ipv6_route_table_template,
6122 			sizeof(ipv6_route_table_template),
6123 			GFP_KERNEL);
6124 
6125 	if (table) {
6126 		table[0].data = &net->ipv6.sysctl.flush_delay;
6127 		table[0].extra1 = net;
6128 		table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6129 		table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6130 		table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6131 		table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6132 		table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6133 		table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6134 		table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6135 		table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6136 		table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6137 		table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6138 
6139 		/* Don't export sysctls to unprivileged users */
6140 		if (net->user_ns != &init_user_ns)
6141 			table[0].procname = NULL;
6142 	}
6143 
6144 	return table;
6145 }
6146 #endif
6147 
6148 static int __net_init ip6_route_net_init(struct net *net)
6149 {
6150 	int ret = -ENOMEM;
6151 
6152 	memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6153 	       sizeof(net->ipv6.ip6_dst_ops));
6154 
6155 	if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6156 		goto out_ip6_dst_ops;
6157 
6158 	net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6159 	if (!net->ipv6.fib6_null_entry)
6160 		goto out_ip6_dst_entries;
6161 	memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6162 	       sizeof(*net->ipv6.fib6_null_entry));
6163 
6164 	net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6165 					   sizeof(*net->ipv6.ip6_null_entry),
6166 					   GFP_KERNEL);
6167 	if (!net->ipv6.ip6_null_entry)
6168 		goto out_fib6_null_entry;
6169 	net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6170 	dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6171 			 ip6_template_metrics, true);
6172 	INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6173 
6174 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6175 	net->ipv6.fib6_has_custom_rules = false;
6176 	net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6177 					       sizeof(*net->ipv6.ip6_prohibit_entry),
6178 					       GFP_KERNEL);
6179 	if (!net->ipv6.ip6_prohibit_entry)
6180 		goto out_ip6_null_entry;
6181 	net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6182 	dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6183 			 ip6_template_metrics, true);
6184 	INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6185 
6186 	net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6187 					       sizeof(*net->ipv6.ip6_blk_hole_entry),
6188 					       GFP_KERNEL);
6189 	if (!net->ipv6.ip6_blk_hole_entry)
6190 		goto out_ip6_prohibit_entry;
6191 	net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6192 	dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6193 			 ip6_template_metrics, true);
6194 	INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6195 #endif
6196 
6197 	net->ipv6.sysctl.flush_delay = 0;
6198 	net->ipv6.sysctl.ip6_rt_max_size = 4096;
6199 	net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6200 	net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6201 	net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6202 	net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6203 	net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6204 	net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6205 	net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6206 
6207 	net->ipv6.ip6_rt_gc_expire = 30*HZ;
6208 
6209 	ret = 0;
6210 out:
6211 	return ret;
6212 
6213 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6214 out_ip6_prohibit_entry:
6215 	kfree(net->ipv6.ip6_prohibit_entry);
6216 out_ip6_null_entry:
6217 	kfree(net->ipv6.ip6_null_entry);
6218 #endif
6219 out_fib6_null_entry:
6220 	kfree(net->ipv6.fib6_null_entry);
6221 out_ip6_dst_entries:
6222 	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6223 out_ip6_dst_ops:
6224 	goto out;
6225 }
6226 
6227 static void __net_exit ip6_route_net_exit(struct net *net)
6228 {
6229 	kfree(net->ipv6.fib6_null_entry);
6230 	kfree(net->ipv6.ip6_null_entry);
6231 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6232 	kfree(net->ipv6.ip6_prohibit_entry);
6233 	kfree(net->ipv6.ip6_blk_hole_entry);
6234 #endif
6235 	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6236 }
6237 
6238 static int __net_init ip6_route_net_init_late(struct net *net)
6239 {
6240 #ifdef CONFIG_PROC_FS
6241 	proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6242 			sizeof(struct ipv6_route_iter));
6243 	proc_create_net_single("rt6_stats", 0444, net->proc_net,
6244 			rt6_stats_seq_show, NULL);
6245 #endif
6246 	return 0;
6247 }
6248 
6249 static void __net_exit ip6_route_net_exit_late(struct net *net)
6250 {
6251 #ifdef CONFIG_PROC_FS
6252 	remove_proc_entry("ipv6_route", net->proc_net);
6253 	remove_proc_entry("rt6_stats", net->proc_net);
6254 #endif
6255 }
6256 
6257 static struct pernet_operations ip6_route_net_ops = {
6258 	.init = ip6_route_net_init,
6259 	.exit = ip6_route_net_exit,
6260 };
6261 
6262 static int __net_init ipv6_inetpeer_init(struct net *net)
6263 {
6264 	struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6265 
6266 	if (!bp)
6267 		return -ENOMEM;
6268 	inet_peer_base_init(bp);
6269 	net->ipv6.peers = bp;
6270 	return 0;
6271 }
6272 
6273 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6274 {
6275 	struct inet_peer_base *bp = net->ipv6.peers;
6276 
6277 	net->ipv6.peers = NULL;
6278 	inetpeer_invalidate_tree(bp);
6279 	kfree(bp);
6280 }
6281 
6282 static struct pernet_operations ipv6_inetpeer_ops = {
6283 	.init	=	ipv6_inetpeer_init,
6284 	.exit	=	ipv6_inetpeer_exit,
6285 };
6286 
6287 static struct pernet_operations ip6_route_net_late_ops = {
6288 	.init = ip6_route_net_init_late,
6289 	.exit = ip6_route_net_exit_late,
6290 };
6291 
6292 static struct notifier_block ip6_route_dev_notifier = {
6293 	.notifier_call = ip6_route_dev_notify,
6294 	.priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6295 };
6296 
6297 void __init ip6_route_init_special_entries(void)
6298 {
6299 	/* Registering of the loopback is done before this portion of code,
6300 	 * the loopback reference in rt6_info will not be taken, do it
6301 	 * manually for init_net */
6302 	init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6303 	init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6304 	init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6305   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6306 	init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6307 	init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6308 	init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6309 	init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6310   #endif
6311 }
6312 
6313 int __init ip6_route_init(void)
6314 {
6315 	int ret;
6316 	int cpu;
6317 
6318 	ret = -ENOMEM;
6319 	ip6_dst_ops_template.kmem_cachep =
6320 		kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6321 				  SLAB_HWCACHE_ALIGN, NULL);
6322 	if (!ip6_dst_ops_template.kmem_cachep)
6323 		goto out;
6324 
6325 	ret = dst_entries_init(&ip6_dst_blackhole_ops);
6326 	if (ret)
6327 		goto out_kmem_cache;
6328 
6329 	ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6330 	if (ret)
6331 		goto out_dst_entries;
6332 
6333 	ret = register_pernet_subsys(&ip6_route_net_ops);
6334 	if (ret)
6335 		goto out_register_inetpeer;
6336 
6337 	ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6338 
6339 	ret = fib6_init();
6340 	if (ret)
6341 		goto out_register_subsys;
6342 
6343 	ret = xfrm6_init();
6344 	if (ret)
6345 		goto out_fib6_init;
6346 
6347 	ret = fib6_rules_init();
6348 	if (ret)
6349 		goto xfrm6_init;
6350 
6351 	ret = register_pernet_subsys(&ip6_route_net_late_ops);
6352 	if (ret)
6353 		goto fib6_rules_init;
6354 
6355 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6356 				   inet6_rtm_newroute, NULL, 0);
6357 	if (ret < 0)
6358 		goto out_register_late_subsys;
6359 
6360 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6361 				   inet6_rtm_delroute, NULL, 0);
6362 	if (ret < 0)
6363 		goto out_register_late_subsys;
6364 
6365 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6366 				   inet6_rtm_getroute, NULL,
6367 				   RTNL_FLAG_DOIT_UNLOCKED);
6368 	if (ret < 0)
6369 		goto out_register_late_subsys;
6370 
6371 	ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6372 	if (ret)
6373 		goto out_register_late_subsys;
6374 
6375 	for_each_possible_cpu(cpu) {
6376 		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6377 
6378 		INIT_LIST_HEAD(&ul->head);
6379 		spin_lock_init(&ul->lock);
6380 	}
6381 
6382 out:
6383 	return ret;
6384 
6385 out_register_late_subsys:
6386 	rtnl_unregister_all(PF_INET6);
6387 	unregister_pernet_subsys(&ip6_route_net_late_ops);
6388 fib6_rules_init:
6389 	fib6_rules_cleanup();
6390 xfrm6_init:
6391 	xfrm6_fini();
6392 out_fib6_init:
6393 	fib6_gc_cleanup();
6394 out_register_subsys:
6395 	unregister_pernet_subsys(&ip6_route_net_ops);
6396 out_register_inetpeer:
6397 	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6398 out_dst_entries:
6399 	dst_entries_destroy(&ip6_dst_blackhole_ops);
6400 out_kmem_cache:
6401 	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6402 	goto out;
6403 }
6404 
6405 void ip6_route_cleanup(void)
6406 {
6407 	unregister_netdevice_notifier(&ip6_route_dev_notifier);
6408 	unregister_pernet_subsys(&ip6_route_net_late_ops);
6409 	fib6_rules_cleanup();
6410 	xfrm6_fini();
6411 	fib6_gc_cleanup();
6412 	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6413 	unregister_pernet_subsys(&ip6_route_net_ops);
6414 	dst_entries_destroy(&ip6_dst_blackhole_ops);
6415 	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6416 }
6417