xref: /linux/net/ipv6/route.c (revision 5fc31936081919a8572a3d644f3fbb258038f337)
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 <linux/siphash.h>
45 #include <net/net_namespace.h>
46 #include <net/snmp.h>
47 #include <net/ipv6.h>
48 #include <net/ip6_fib.h>
49 #include <net/ip6_route.h>
50 #include <net/ndisc.h>
51 #include <net/addrconf.h>
52 #include <net/tcp.h>
53 #include <linux/rtnetlink.h>
54 #include <net/dst.h>
55 #include <net/dst_metadata.h>
56 #include <net/xfrm.h>
57 #include <net/netevent.h>
58 #include <net/netlink.h>
59 #include <net/rtnh.h>
60 #include <net/lwtunnel.h>
61 #include <net/ip_tunnels.h>
62 #include <net/l3mdev.h>
63 #include <net/ip.h>
64 #include <linux/uaccess.h>
65 #include <linux/btf_ids.h>
66 
67 #ifdef CONFIG_SYSCTL
68 #include <linux/sysctl.h>
69 #endif
70 
71 static int ip6_rt_type_to_error(u8 fib6_type);
72 
73 #define CREATE_TRACE_POINTS
74 #include <trace/events/fib6.h>
75 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
76 #undef CREATE_TRACE_POINTS
77 
78 enum rt6_nud_state {
79 	RT6_NUD_FAIL_HARD = -3,
80 	RT6_NUD_FAIL_PROBE = -2,
81 	RT6_NUD_FAIL_DO_RR = -1,
82 	RT6_NUD_SUCCEED = 1
83 };
84 
85 INDIRECT_CALLABLE_SCOPE
86 struct dst_entry	*ip6_dst_check(struct dst_entry *dst, u32 cookie);
87 static unsigned int	 ip6_default_advmss(const struct dst_entry *dst);
88 INDIRECT_CALLABLE_SCOPE
89 unsigned int		ip6_mtu(const struct dst_entry *dst);
90 static void		ip6_negative_advice(struct sock *sk,
91 					    struct dst_entry *dst);
92 static void		ip6_dst_destroy(struct dst_entry *);
93 static void		ip6_dst_ifdown(struct dst_entry *,
94 				       struct net_device *dev);
95 static void		 ip6_dst_gc(struct dst_ops *ops);
96 
97 static int		ip6_pkt_discard(struct sk_buff *skb);
98 static int		ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
99 static int		ip6_pkt_prohibit(struct sk_buff *skb);
100 static int		ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
101 static void		ip6_link_failure(struct sk_buff *skb);
102 static void		ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
103 					   struct sk_buff *skb, u32 mtu,
104 					   bool confirm_neigh);
105 static void		rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
106 					struct sk_buff *skb);
107 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
108 			   int strict);
109 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
110 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
111 			 struct fib6_info *rt, struct dst_entry *dst,
112 			 struct in6_addr *dest, struct in6_addr *src,
113 			 int iif, int type, u32 portid, u32 seq,
114 			 unsigned int flags);
115 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
116 					   const struct in6_addr *daddr,
117 					   const struct in6_addr *saddr);
118 
119 #ifdef CONFIG_IPV6_ROUTE_INFO
120 static struct fib6_info *rt6_add_route_info(struct net *net,
121 					   const struct in6_addr *prefix, int prefixlen,
122 					   const struct in6_addr *gwaddr,
123 					   struct net_device *dev,
124 					   unsigned int pref);
125 static struct fib6_info *rt6_get_route_info(struct net *net,
126 					   const struct in6_addr *prefix, int prefixlen,
127 					   const struct in6_addr *gwaddr,
128 					   struct net_device *dev);
129 #endif
130 
131 struct uncached_list {
132 	spinlock_t		lock;
133 	struct list_head	head;
134 };
135 
136 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
137 
rt6_uncached_list_add(struct rt6_info * rt)138 void rt6_uncached_list_add(struct rt6_info *rt)
139 {
140 	struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
141 
142 	rt->dst.rt_uncached_list = ul;
143 
144 	spin_lock_bh(&ul->lock);
145 	list_add_tail(&rt->dst.rt_uncached, &ul->head);
146 	spin_unlock_bh(&ul->lock);
147 }
148 
rt6_uncached_list_del(struct rt6_info * rt)149 void rt6_uncached_list_del(struct rt6_info *rt)
150 {
151 	if (!list_empty(&rt->dst.rt_uncached)) {
152 		struct uncached_list *ul = rt->dst.rt_uncached_list;
153 
154 		spin_lock_bh(&ul->lock);
155 		list_del_init(&rt->dst.rt_uncached);
156 		spin_unlock_bh(&ul->lock);
157 	}
158 }
159 
rt6_uncached_list_flush_dev(struct net_device * dev)160 static void rt6_uncached_list_flush_dev(struct net_device *dev)
161 {
162 	int cpu;
163 
164 	for_each_possible_cpu(cpu) {
165 		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166 		struct rt6_info *rt, *safe;
167 
168 		if (list_empty(&ul->head))
169 			continue;
170 
171 		spin_lock_bh(&ul->lock);
172 		list_for_each_entry_safe(rt, safe, &ul->head, dst.rt_uncached) {
173 			struct inet6_dev *rt_idev = rt->rt6i_idev;
174 			struct net_device *rt_dev = rt->dst.dev;
175 			bool handled = false;
176 
177 			if (rt_idev && rt_idev->dev == dev) {
178 				rt->rt6i_idev = in6_dev_get(blackhole_netdev);
179 				in6_dev_put(rt_idev);
180 				handled = true;
181 			}
182 
183 			if (rt_dev == dev) {
184 				rt->dst.dev = blackhole_netdev;
185 				netdev_ref_replace(rt_dev, blackhole_netdev,
186 						   &rt->dst.dev_tracker,
187 						   GFP_ATOMIC);
188 				handled = true;
189 			}
190 			if (handled)
191 				list_del_init(&rt->dst.rt_uncached);
192 		}
193 		spin_unlock_bh(&ul->lock);
194 	}
195 }
196 
choose_neigh_daddr(const struct in6_addr * p,struct sk_buff * skb,const void * daddr)197 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
198 					     struct sk_buff *skb,
199 					     const void *daddr)
200 {
201 	if (!ipv6_addr_any(p))
202 		return (const void *) p;
203 	else if (skb)
204 		return &ipv6_hdr(skb)->daddr;
205 	return daddr;
206 }
207 
ip6_neigh_lookup(const struct in6_addr * gw,struct net_device * dev,struct sk_buff * skb,const void * daddr)208 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
209 				   struct net_device *dev,
210 				   struct sk_buff *skb,
211 				   const void *daddr)
212 {
213 	struct neighbour *n;
214 
215 	daddr = choose_neigh_daddr(gw, skb, daddr);
216 	n = __ipv6_neigh_lookup(dev, daddr);
217 	if (n)
218 		return n;
219 
220 	n = neigh_create(&nd_tbl, daddr, dev);
221 	return IS_ERR(n) ? NULL : n;
222 }
223 
ip6_dst_neigh_lookup(const struct dst_entry * dst,struct sk_buff * skb,const void * daddr)224 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
225 					      struct sk_buff *skb,
226 					      const void *daddr)
227 {
228 	const struct rt6_info *rt = dst_rt6_info(dst);
229 
230 	return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
231 				dst->dev, skb, daddr);
232 }
233 
ip6_confirm_neigh(const struct dst_entry * dst,const void * daddr)234 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
235 {
236 	const struct rt6_info *rt = dst_rt6_info(dst);
237 	struct net_device *dev = dst->dev;
238 
239 	daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
240 	if (!daddr)
241 		return;
242 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
243 		return;
244 	if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
245 		return;
246 	__ipv6_confirm_neigh(dev, daddr);
247 }
248 
249 static struct dst_ops ip6_dst_ops_template = {
250 	.family			=	AF_INET6,
251 	.gc			=	ip6_dst_gc,
252 	.gc_thresh		=	1024,
253 	.check			=	ip6_dst_check,
254 	.default_advmss		=	ip6_default_advmss,
255 	.mtu			=	ip6_mtu,
256 	.cow_metrics		=	dst_cow_metrics_generic,
257 	.destroy		=	ip6_dst_destroy,
258 	.ifdown			=	ip6_dst_ifdown,
259 	.negative_advice	=	ip6_negative_advice,
260 	.link_failure		=	ip6_link_failure,
261 	.update_pmtu		=	ip6_rt_update_pmtu,
262 	.redirect		=	rt6_do_redirect,
263 	.local_out		=	__ip6_local_out,
264 	.neigh_lookup		=	ip6_dst_neigh_lookup,
265 	.confirm_neigh		=	ip6_confirm_neigh,
266 };
267 
268 static struct dst_ops ip6_dst_blackhole_ops = {
269 	.family			= AF_INET6,
270 	.default_advmss		= ip6_default_advmss,
271 	.neigh_lookup		= ip6_dst_neigh_lookup,
272 	.check			= ip6_dst_check,
273 	.destroy		= ip6_dst_destroy,
274 	.cow_metrics		= dst_cow_metrics_generic,
275 	.update_pmtu		= dst_blackhole_update_pmtu,
276 	.redirect		= dst_blackhole_redirect,
277 	.mtu			= dst_blackhole_mtu,
278 };
279 
280 static const u32 ip6_template_metrics[RTAX_MAX] = {
281 	[RTAX_HOPLIMIT - 1] = 0,
282 };
283 
284 static const struct fib6_info fib6_null_entry_template = {
285 	.fib6_flags	= (RTF_REJECT | RTF_NONEXTHOP),
286 	.fib6_protocol  = RTPROT_KERNEL,
287 	.fib6_metric	= ~(u32)0,
288 	.fib6_ref	= REFCOUNT_INIT(1),
289 	.fib6_type	= RTN_UNREACHABLE,
290 	.fib6_metrics	= (struct dst_metrics *)&dst_default_metrics,
291 };
292 
293 static const struct rt6_info ip6_null_entry_template = {
294 	.dst = {
295 		.__rcuref	= RCUREF_INIT(1),
296 		.__use		= 1,
297 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
298 		.error		= -ENETUNREACH,
299 		.input		= ip6_pkt_discard,
300 		.output		= ip6_pkt_discard_out,
301 	},
302 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
303 };
304 
305 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
306 
307 static const struct rt6_info ip6_prohibit_entry_template = {
308 	.dst = {
309 		.__rcuref	= RCUREF_INIT(1),
310 		.__use		= 1,
311 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
312 		.error		= -EACCES,
313 		.input		= ip6_pkt_prohibit,
314 		.output		= ip6_pkt_prohibit_out,
315 	},
316 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
317 };
318 
319 static const struct rt6_info ip6_blk_hole_entry_template = {
320 	.dst = {
321 		.__rcuref	= RCUREF_INIT(1),
322 		.__use		= 1,
323 		.obsolete	= DST_OBSOLETE_FORCE_CHK,
324 		.error		= -EINVAL,
325 		.input		= dst_discard,
326 		.output		= dst_discard_out,
327 	},
328 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
329 };
330 
331 #endif
332 
rt6_info_init(struct rt6_info * rt)333 static void rt6_info_init(struct rt6_info *rt)
334 {
335 	memset_after(rt, 0, dst);
336 }
337 
338 /* allocate dst with ip6_dst_ops */
ip6_dst_alloc(struct net * net,struct net_device * dev,int flags)339 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
340 			       int flags)
341 {
342 	struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
343 					DST_OBSOLETE_FORCE_CHK, flags);
344 
345 	if (rt) {
346 		rt6_info_init(rt);
347 		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
348 	}
349 
350 	return rt;
351 }
352 EXPORT_SYMBOL(ip6_dst_alloc);
353 
ip6_dst_destroy(struct dst_entry * dst)354 static void ip6_dst_destroy(struct dst_entry *dst)
355 {
356 	struct rt6_info *rt = dst_rt6_info(dst);
357 	struct fib6_info *from;
358 	struct inet6_dev *idev;
359 
360 	ip_dst_metrics_put(dst);
361 	rt6_uncached_list_del(rt);
362 
363 	idev = rt->rt6i_idev;
364 	if (idev) {
365 		rt->rt6i_idev = NULL;
366 		in6_dev_put(idev);
367 	}
368 
369 	from = unrcu_pointer(xchg(&rt->from, NULL));
370 	fib6_info_release(from);
371 }
372 
ip6_dst_ifdown(struct dst_entry * dst,struct net_device * dev)373 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
374 {
375 	struct rt6_info *rt = dst_rt6_info(dst);
376 	struct inet6_dev *idev = rt->rt6i_idev;
377 	struct fib6_info *from;
378 
379 	if (idev && idev->dev != blackhole_netdev) {
380 		struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);
381 
382 		if (blackhole_idev) {
383 			rt->rt6i_idev = blackhole_idev;
384 			in6_dev_put(idev);
385 		}
386 	}
387 	from = unrcu_pointer(xchg(&rt->from, NULL));
388 	fib6_info_release(from);
389 }
390 
__rt6_check_expired(const struct rt6_info * rt)391 static bool __rt6_check_expired(const struct rt6_info *rt)
392 {
393 	if (rt->rt6i_flags & RTF_EXPIRES)
394 		return time_after(jiffies, rt->dst.expires);
395 	else
396 		return false;
397 }
398 
rt6_check_expired(const struct rt6_info * rt)399 static bool rt6_check_expired(const struct rt6_info *rt)
400 {
401 	struct fib6_info *from;
402 
403 	from = rcu_dereference(rt->from);
404 
405 	if (rt->rt6i_flags & RTF_EXPIRES) {
406 		if (time_after(jiffies, rt->dst.expires))
407 			return true;
408 	} else if (from) {
409 		return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
410 			fib6_check_expired(from);
411 	}
412 	return false;
413 }
414 
fib6_select_path(const struct net * net,struct fib6_result * res,struct flowi6 * fl6,int oif,bool have_oif_match,const struct sk_buff * skb,int strict)415 void fib6_select_path(const struct net *net, struct fib6_result *res,
416 		      struct flowi6 *fl6, int oif, bool have_oif_match,
417 		      const struct sk_buff *skb, int strict)
418 {
419 	struct fib6_info *match = res->f6i;
420 	struct fib6_info *sibling;
421 
422 	if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
423 		goto out;
424 
425 	if (match->nh && have_oif_match && res->nh)
426 		return;
427 
428 	if (skb)
429 		IP6CB(skb)->flags |= IP6SKB_MULTIPATH;
430 
431 	/* We might have already computed the hash for ICMPv6 errors. In such
432 	 * case it will always be non-zero. Otherwise now is the time to do it.
433 	 */
434 	if (!fl6->mp_hash &&
435 	    (!match->nh || nexthop_is_multipath(match->nh)))
436 		fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
437 
438 	if (unlikely(match->nh)) {
439 		nexthop_path_fib6_result(res, fl6->mp_hash);
440 		return;
441 	}
442 
443 	if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
444 		goto out;
445 
446 	list_for_each_entry_rcu(sibling, &match->fib6_siblings,
447 				fib6_siblings) {
448 		const struct fib6_nh *nh = sibling->fib6_nh;
449 		int nh_upper_bound;
450 
451 		nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
452 		if (fl6->mp_hash > nh_upper_bound)
453 			continue;
454 		if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
455 			break;
456 		match = sibling;
457 		break;
458 	}
459 
460 out:
461 	res->f6i = match;
462 	res->nh = match->fib6_nh;
463 }
464 
465 /*
466  *	Route lookup. rcu_read_lock() should be held.
467  */
468 
__rt6_device_match(struct net * net,const struct fib6_nh * nh,const struct in6_addr * saddr,int oif,int flags)469 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
470 			       const struct in6_addr *saddr, int oif, int flags)
471 {
472 	const struct net_device *dev;
473 
474 	if (nh->fib_nh_flags & RTNH_F_DEAD)
475 		return false;
476 
477 	dev = nh->fib_nh_dev;
478 	if (oif) {
479 		if (dev->ifindex == oif)
480 			return true;
481 	} else {
482 		if (ipv6_chk_addr(net, saddr, dev,
483 				  flags & RT6_LOOKUP_F_IFACE))
484 			return true;
485 	}
486 
487 	return false;
488 }
489 
490 struct fib6_nh_dm_arg {
491 	struct net		*net;
492 	const struct in6_addr	*saddr;
493 	int			oif;
494 	int			flags;
495 	struct fib6_nh		*nh;
496 };
497 
__rt6_nh_dev_match(struct fib6_nh * nh,void * _arg)498 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
499 {
500 	struct fib6_nh_dm_arg *arg = _arg;
501 
502 	arg->nh = nh;
503 	return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
504 				  arg->flags);
505 }
506 
507 /* returns fib6_nh from nexthop or NULL */
rt6_nh_dev_match(struct net * net,struct nexthop * nh,struct fib6_result * res,const struct in6_addr * saddr,int oif,int flags)508 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
509 					struct fib6_result *res,
510 					const struct in6_addr *saddr,
511 					int oif, int flags)
512 {
513 	struct fib6_nh_dm_arg arg = {
514 		.net   = net,
515 		.saddr = saddr,
516 		.oif   = oif,
517 		.flags = flags,
518 	};
519 
520 	if (nexthop_is_blackhole(nh))
521 		return NULL;
522 
523 	if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
524 		return arg.nh;
525 
526 	return NULL;
527 }
528 
rt6_device_match(struct net * net,struct fib6_result * res,const struct in6_addr * saddr,int oif,int flags)529 static void rt6_device_match(struct net *net, struct fib6_result *res,
530 			     const struct in6_addr *saddr, int oif, int flags)
531 {
532 	struct fib6_info *f6i = res->f6i;
533 	struct fib6_info *spf6i;
534 	struct fib6_nh *nh;
535 
536 	if (!oif && ipv6_addr_any(saddr)) {
537 		if (unlikely(f6i->nh)) {
538 			nh = nexthop_fib6_nh(f6i->nh);
539 			if (nexthop_is_blackhole(f6i->nh))
540 				goto out_blackhole;
541 		} else {
542 			nh = f6i->fib6_nh;
543 		}
544 		if (!(nh->fib_nh_flags & RTNH_F_DEAD))
545 			goto out;
546 	}
547 
548 	for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
549 		bool matched = false;
550 
551 		if (unlikely(spf6i->nh)) {
552 			nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
553 					      oif, flags);
554 			if (nh)
555 				matched = true;
556 		} else {
557 			nh = spf6i->fib6_nh;
558 			if (__rt6_device_match(net, nh, saddr, oif, flags))
559 				matched = true;
560 		}
561 		if (matched) {
562 			res->f6i = spf6i;
563 			goto out;
564 		}
565 	}
566 
567 	if (oif && flags & RT6_LOOKUP_F_IFACE) {
568 		res->f6i = net->ipv6.fib6_null_entry;
569 		nh = res->f6i->fib6_nh;
570 		goto out;
571 	}
572 
573 	if (unlikely(f6i->nh)) {
574 		nh = nexthop_fib6_nh(f6i->nh);
575 		if (nexthop_is_blackhole(f6i->nh))
576 			goto out_blackhole;
577 	} else {
578 		nh = f6i->fib6_nh;
579 	}
580 
581 	if (nh->fib_nh_flags & RTNH_F_DEAD) {
582 		res->f6i = net->ipv6.fib6_null_entry;
583 		nh = res->f6i->fib6_nh;
584 	}
585 out:
586 	res->nh = nh;
587 	res->fib6_type = res->f6i->fib6_type;
588 	res->fib6_flags = res->f6i->fib6_flags;
589 	return;
590 
591 out_blackhole:
592 	res->fib6_flags |= RTF_REJECT;
593 	res->fib6_type = RTN_BLACKHOLE;
594 	res->nh = nh;
595 }
596 
597 #ifdef CONFIG_IPV6_ROUTER_PREF
598 struct __rt6_probe_work {
599 	struct work_struct work;
600 	struct in6_addr target;
601 	struct net_device *dev;
602 	netdevice_tracker dev_tracker;
603 };
604 
rt6_probe_deferred(struct work_struct * w)605 static void rt6_probe_deferred(struct work_struct *w)
606 {
607 	struct in6_addr mcaddr;
608 	struct __rt6_probe_work *work =
609 		container_of(w, struct __rt6_probe_work, work);
610 
611 	addrconf_addr_solict_mult(&work->target, &mcaddr);
612 	ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
613 	netdev_put(work->dev, &work->dev_tracker);
614 	kfree(work);
615 }
616 
rt6_probe(struct fib6_nh * fib6_nh)617 static void rt6_probe(struct fib6_nh *fib6_nh)
618 {
619 	struct __rt6_probe_work *work = NULL;
620 	const struct in6_addr *nh_gw;
621 	unsigned long last_probe;
622 	struct neighbour *neigh;
623 	struct net_device *dev;
624 	struct inet6_dev *idev;
625 
626 	/*
627 	 * Okay, this does not seem to be appropriate
628 	 * for now, however, we need to check if it
629 	 * is really so; aka Router Reachability Probing.
630 	 *
631 	 * Router Reachability Probe MUST be rate-limited
632 	 * to no more than one per minute.
633 	 */
634 	if (!fib6_nh->fib_nh_gw_family)
635 		return;
636 
637 	nh_gw = &fib6_nh->fib_nh_gw6;
638 	dev = fib6_nh->fib_nh_dev;
639 	rcu_read_lock();
640 	last_probe = READ_ONCE(fib6_nh->last_probe);
641 	idev = __in6_dev_get(dev);
642 	if (!idev)
643 		goto out;
644 	neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
645 	if (neigh) {
646 		if (READ_ONCE(neigh->nud_state) & NUD_VALID)
647 			goto out;
648 
649 		write_lock_bh(&neigh->lock);
650 		if (!(neigh->nud_state & NUD_VALID) &&
651 		    time_after(jiffies,
652 			       neigh->updated +
653 			       READ_ONCE(idev->cnf.rtr_probe_interval))) {
654 			work = kmalloc(sizeof(*work), GFP_ATOMIC);
655 			if (work)
656 				__neigh_set_probe_once(neigh);
657 		}
658 		write_unlock_bh(&neigh->lock);
659 	} else if (time_after(jiffies, last_probe +
660 				       READ_ONCE(idev->cnf.rtr_probe_interval))) {
661 		work = kmalloc(sizeof(*work), GFP_ATOMIC);
662 	}
663 
664 	if (!work || cmpxchg(&fib6_nh->last_probe,
665 			     last_probe, jiffies) != last_probe) {
666 		kfree(work);
667 	} else {
668 		INIT_WORK(&work->work, rt6_probe_deferred);
669 		work->target = *nh_gw;
670 		netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
671 		work->dev = dev;
672 		schedule_work(&work->work);
673 	}
674 
675 out:
676 	rcu_read_unlock();
677 }
678 #else
rt6_probe(struct fib6_nh * fib6_nh)679 static inline void rt6_probe(struct fib6_nh *fib6_nh)
680 {
681 }
682 #endif
683 
684 /*
685  * Default Router Selection (RFC 2461 6.3.6)
686  */
rt6_check_neigh(const struct fib6_nh * fib6_nh)687 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
688 {
689 	enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
690 	struct neighbour *neigh;
691 
692 	rcu_read_lock();
693 	neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
694 					  &fib6_nh->fib_nh_gw6);
695 	if (neigh) {
696 		u8 nud_state = READ_ONCE(neigh->nud_state);
697 
698 		if (nud_state & NUD_VALID)
699 			ret = RT6_NUD_SUCCEED;
700 #ifdef CONFIG_IPV6_ROUTER_PREF
701 		else if (!(nud_state & NUD_FAILED))
702 			ret = RT6_NUD_SUCCEED;
703 		else
704 			ret = RT6_NUD_FAIL_PROBE;
705 #endif
706 	} else {
707 		ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
708 		      RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
709 	}
710 	rcu_read_unlock();
711 
712 	return ret;
713 }
714 
rt6_score_route(const struct fib6_nh * nh,u32 fib6_flags,int oif,int strict)715 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
716 			   int strict)
717 {
718 	int m = 0;
719 
720 	if (!oif || nh->fib_nh_dev->ifindex == oif)
721 		m = 2;
722 
723 	if (!m && (strict & RT6_LOOKUP_F_IFACE))
724 		return RT6_NUD_FAIL_HARD;
725 #ifdef CONFIG_IPV6_ROUTER_PREF
726 	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
727 #endif
728 	if ((strict & RT6_LOOKUP_F_REACHABLE) &&
729 	    !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
730 		int n = rt6_check_neigh(nh);
731 		if (n < 0)
732 			return n;
733 	}
734 	return m;
735 }
736 
find_match(struct fib6_nh * nh,u32 fib6_flags,int oif,int strict,int * mpri,bool * do_rr)737 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
738 		       int oif, int strict, int *mpri, bool *do_rr)
739 {
740 	bool match_do_rr = false;
741 	bool rc = false;
742 	int m;
743 
744 	if (nh->fib_nh_flags & RTNH_F_DEAD)
745 		goto out;
746 
747 	if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
748 	    nh->fib_nh_flags & RTNH_F_LINKDOWN &&
749 	    !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
750 		goto out;
751 
752 	m = rt6_score_route(nh, fib6_flags, oif, strict);
753 	if (m == RT6_NUD_FAIL_DO_RR) {
754 		match_do_rr = true;
755 		m = 0; /* lowest valid score */
756 	} else if (m == RT6_NUD_FAIL_HARD) {
757 		goto out;
758 	}
759 
760 	if (strict & RT6_LOOKUP_F_REACHABLE)
761 		rt6_probe(nh);
762 
763 	/* note that m can be RT6_NUD_FAIL_PROBE at this point */
764 	if (m > *mpri) {
765 		*do_rr = match_do_rr;
766 		*mpri = m;
767 		rc = true;
768 	}
769 out:
770 	return rc;
771 }
772 
773 struct fib6_nh_frl_arg {
774 	u32		flags;
775 	int		oif;
776 	int		strict;
777 	int		*mpri;
778 	bool		*do_rr;
779 	struct fib6_nh	*nh;
780 };
781 
rt6_nh_find_match(struct fib6_nh * nh,void * _arg)782 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
783 {
784 	struct fib6_nh_frl_arg *arg = _arg;
785 
786 	arg->nh = nh;
787 	return find_match(nh, arg->flags, arg->oif, arg->strict,
788 			  arg->mpri, arg->do_rr);
789 }
790 
__find_rr_leaf(struct fib6_info * f6i_start,struct fib6_info * nomatch,u32 metric,struct fib6_result * res,struct fib6_info ** cont,int oif,int strict,bool * do_rr,int * mpri)791 static void __find_rr_leaf(struct fib6_info *f6i_start,
792 			   struct fib6_info *nomatch, u32 metric,
793 			   struct fib6_result *res, struct fib6_info **cont,
794 			   int oif, int strict, bool *do_rr, int *mpri)
795 {
796 	struct fib6_info *f6i;
797 
798 	for (f6i = f6i_start;
799 	     f6i && f6i != nomatch;
800 	     f6i = rcu_dereference(f6i->fib6_next)) {
801 		bool matched = false;
802 		struct fib6_nh *nh;
803 
804 		if (cont && f6i->fib6_metric != metric) {
805 			*cont = f6i;
806 			return;
807 		}
808 
809 		if (fib6_check_expired(f6i))
810 			continue;
811 
812 		if (unlikely(f6i->nh)) {
813 			struct fib6_nh_frl_arg arg = {
814 				.flags  = f6i->fib6_flags,
815 				.oif    = oif,
816 				.strict = strict,
817 				.mpri   = mpri,
818 				.do_rr  = do_rr
819 			};
820 
821 			if (nexthop_is_blackhole(f6i->nh)) {
822 				res->fib6_flags = RTF_REJECT;
823 				res->fib6_type = RTN_BLACKHOLE;
824 				res->f6i = f6i;
825 				res->nh = nexthop_fib6_nh(f6i->nh);
826 				return;
827 			}
828 			if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
829 						     &arg)) {
830 				matched = true;
831 				nh = arg.nh;
832 			}
833 		} else {
834 			nh = f6i->fib6_nh;
835 			if (find_match(nh, f6i->fib6_flags, oif, strict,
836 				       mpri, do_rr))
837 				matched = true;
838 		}
839 		if (matched) {
840 			res->f6i = f6i;
841 			res->nh = nh;
842 			res->fib6_flags = f6i->fib6_flags;
843 			res->fib6_type = f6i->fib6_type;
844 		}
845 	}
846 }
847 
find_rr_leaf(struct fib6_node * fn,struct fib6_info * leaf,struct fib6_info * rr_head,int oif,int strict,bool * do_rr,struct fib6_result * res)848 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
849 			 struct fib6_info *rr_head, int oif, int strict,
850 			 bool *do_rr, struct fib6_result *res)
851 {
852 	u32 metric = rr_head->fib6_metric;
853 	struct fib6_info *cont = NULL;
854 	int mpri = -1;
855 
856 	__find_rr_leaf(rr_head, NULL, metric, res, &cont,
857 		       oif, strict, do_rr, &mpri);
858 
859 	__find_rr_leaf(leaf, rr_head, metric, res, &cont,
860 		       oif, strict, do_rr, &mpri);
861 
862 	if (res->f6i || !cont)
863 		return;
864 
865 	__find_rr_leaf(cont, NULL, metric, res, NULL,
866 		       oif, strict, do_rr, &mpri);
867 }
868 
rt6_select(struct net * net,struct fib6_node * fn,int oif,struct fib6_result * res,int strict)869 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
870 		       struct fib6_result *res, int strict)
871 {
872 	struct fib6_info *leaf = rcu_dereference(fn->leaf);
873 	struct fib6_info *rt0;
874 	bool do_rr = false;
875 	int key_plen;
876 
877 	/* make sure this function or its helpers sets f6i */
878 	res->f6i = NULL;
879 
880 	if (!leaf || leaf == net->ipv6.fib6_null_entry)
881 		goto out;
882 
883 	rt0 = rcu_dereference(fn->rr_ptr);
884 	if (!rt0)
885 		rt0 = leaf;
886 
887 	/* Double check to make sure fn is not an intermediate node
888 	 * and fn->leaf does not points to its child's leaf
889 	 * (This might happen if all routes under fn are deleted from
890 	 * the tree and fib6_repair_tree() is called on the node.)
891 	 */
892 	key_plen = rt0->fib6_dst.plen;
893 #ifdef CONFIG_IPV6_SUBTREES
894 	if (rt0->fib6_src.plen)
895 		key_plen = rt0->fib6_src.plen;
896 #endif
897 	if (fn->fn_bit != key_plen)
898 		goto out;
899 
900 	find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
901 	if (do_rr) {
902 		struct fib6_info *next = rcu_dereference(rt0->fib6_next);
903 
904 		/* no entries matched; do round-robin */
905 		if (!next || next->fib6_metric != rt0->fib6_metric)
906 			next = leaf;
907 
908 		if (next != rt0) {
909 			spin_lock_bh(&leaf->fib6_table->tb6_lock);
910 			/* make sure next is not being deleted from the tree */
911 			if (next->fib6_node)
912 				rcu_assign_pointer(fn->rr_ptr, next);
913 			spin_unlock_bh(&leaf->fib6_table->tb6_lock);
914 		}
915 	}
916 
917 out:
918 	if (!res->f6i) {
919 		res->f6i = net->ipv6.fib6_null_entry;
920 		res->nh = res->f6i->fib6_nh;
921 		res->fib6_flags = res->f6i->fib6_flags;
922 		res->fib6_type = res->f6i->fib6_type;
923 	}
924 }
925 
rt6_is_gw_or_nonexthop(const struct fib6_result * res)926 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
927 {
928 	return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
929 	       res->nh->fib_nh_gw_family;
930 }
931 
932 #ifdef CONFIG_IPV6_ROUTE_INFO
rt6_route_rcv(struct net_device * dev,u8 * opt,int len,const struct in6_addr * gwaddr)933 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
934 		  const struct in6_addr *gwaddr)
935 {
936 	struct net *net = dev_net(dev);
937 	struct route_info *rinfo = (struct route_info *) opt;
938 	struct in6_addr prefix_buf, *prefix;
939 	struct fib6_table *table;
940 	unsigned int pref;
941 	unsigned long lifetime;
942 	struct fib6_info *rt;
943 
944 	if (len < sizeof(struct route_info)) {
945 		return -EINVAL;
946 	}
947 
948 	/* Sanity check for prefix_len and length */
949 	if (rinfo->length > 3) {
950 		return -EINVAL;
951 	} else if (rinfo->prefix_len > 128) {
952 		return -EINVAL;
953 	} else if (rinfo->prefix_len > 64) {
954 		if (rinfo->length < 2) {
955 			return -EINVAL;
956 		}
957 	} else if (rinfo->prefix_len > 0) {
958 		if (rinfo->length < 1) {
959 			return -EINVAL;
960 		}
961 	}
962 
963 	pref = rinfo->route_pref;
964 	if (pref == ICMPV6_ROUTER_PREF_INVALID)
965 		return -EINVAL;
966 
967 	lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
968 
969 	if (rinfo->length == 3)
970 		prefix = (struct in6_addr *)rinfo->prefix;
971 	else {
972 		/* this function is safe */
973 		ipv6_addr_prefix(&prefix_buf,
974 				 (struct in6_addr *)rinfo->prefix,
975 				 rinfo->prefix_len);
976 		prefix = &prefix_buf;
977 	}
978 
979 	if (rinfo->prefix_len == 0)
980 		rt = rt6_get_dflt_router(net, gwaddr, dev);
981 	else
982 		rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
983 					gwaddr, dev);
984 
985 	if (rt && !lifetime) {
986 		ip6_del_rt(net, rt, false);
987 		rt = NULL;
988 	}
989 
990 	if (!rt && lifetime)
991 		rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
992 					dev, pref);
993 	else if (rt)
994 		rt->fib6_flags = RTF_ROUTEINFO |
995 				 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
996 
997 	if (rt) {
998 		table = rt->fib6_table;
999 		spin_lock_bh(&table->tb6_lock);
1000 
1001 		if (!addrconf_finite_timeout(lifetime)) {
1002 			fib6_clean_expires(rt);
1003 			fib6_remove_gc_list(rt);
1004 		} else {
1005 			fib6_set_expires(rt, jiffies + HZ * lifetime);
1006 			fib6_add_gc_list(rt);
1007 		}
1008 
1009 		spin_unlock_bh(&table->tb6_lock);
1010 
1011 		fib6_info_release(rt);
1012 	}
1013 	return 0;
1014 }
1015 #endif
1016 
1017 /*
1018  *	Misc support functions
1019  */
1020 
1021 /* called with rcu_lock held */
ip6_rt_get_dev_rcu(const struct fib6_result * res)1022 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1023 {
1024 	struct net_device *dev = res->nh->fib_nh_dev;
1025 
1026 	if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1027 		/* for copies of local routes, dst->dev needs to be the
1028 		 * device if it is a master device, the master device if
1029 		 * device is enslaved, and the loopback as the default
1030 		 */
1031 		if (netif_is_l3_slave(dev) &&
1032 		    !rt6_need_strict(&res->f6i->fib6_dst.addr))
1033 			dev = l3mdev_master_dev_rcu(dev);
1034 		else if (!netif_is_l3_master(dev))
1035 			dev = dev_net(dev)->loopback_dev;
1036 		/* last case is netif_is_l3_master(dev) is true in which
1037 		 * case we want dev returned to be dev
1038 		 */
1039 	}
1040 
1041 	return dev;
1042 }
1043 
1044 static const int fib6_prop[RTN_MAX + 1] = {
1045 	[RTN_UNSPEC]	= 0,
1046 	[RTN_UNICAST]	= 0,
1047 	[RTN_LOCAL]	= 0,
1048 	[RTN_BROADCAST]	= 0,
1049 	[RTN_ANYCAST]	= 0,
1050 	[RTN_MULTICAST]	= 0,
1051 	[RTN_BLACKHOLE]	= -EINVAL,
1052 	[RTN_UNREACHABLE] = -EHOSTUNREACH,
1053 	[RTN_PROHIBIT]	= -EACCES,
1054 	[RTN_THROW]	= -EAGAIN,
1055 	[RTN_NAT]	= -EINVAL,
1056 	[RTN_XRESOLVE]	= -EINVAL,
1057 };
1058 
ip6_rt_type_to_error(u8 fib6_type)1059 static int ip6_rt_type_to_error(u8 fib6_type)
1060 {
1061 	return fib6_prop[fib6_type];
1062 }
1063 
fib6_info_dst_flags(struct fib6_info * rt)1064 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1065 {
1066 	unsigned short flags = 0;
1067 
1068 	if (rt->dst_nocount)
1069 		flags |= DST_NOCOUNT;
1070 	if (rt->dst_nopolicy)
1071 		flags |= DST_NOPOLICY;
1072 
1073 	return flags;
1074 }
1075 
ip6_rt_init_dst_reject(struct rt6_info * rt,u8 fib6_type)1076 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1077 {
1078 	rt->dst.error = ip6_rt_type_to_error(fib6_type);
1079 
1080 	switch (fib6_type) {
1081 	case RTN_BLACKHOLE:
1082 		rt->dst.output = dst_discard_out;
1083 		rt->dst.input = dst_discard;
1084 		break;
1085 	case RTN_PROHIBIT:
1086 		rt->dst.output = ip6_pkt_prohibit_out;
1087 		rt->dst.input = ip6_pkt_prohibit;
1088 		break;
1089 	case RTN_THROW:
1090 	case RTN_UNREACHABLE:
1091 	default:
1092 		rt->dst.output = ip6_pkt_discard_out;
1093 		rt->dst.input = ip6_pkt_discard;
1094 		break;
1095 	}
1096 }
1097 
ip6_rt_init_dst(struct rt6_info * rt,const struct fib6_result * res)1098 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1099 {
1100 	struct fib6_info *f6i = res->f6i;
1101 
1102 	if (res->fib6_flags & RTF_REJECT) {
1103 		ip6_rt_init_dst_reject(rt, res->fib6_type);
1104 		return;
1105 	}
1106 
1107 	rt->dst.error = 0;
1108 	rt->dst.output = ip6_output;
1109 
1110 	if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1111 		rt->dst.input = ip6_input;
1112 	} else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1113 		rt->dst.input = ip6_mc_input;
1114 	} else {
1115 		rt->dst.input = ip6_forward;
1116 	}
1117 
1118 	if (res->nh->fib_nh_lws) {
1119 		rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1120 		lwtunnel_set_redirect(&rt->dst);
1121 	}
1122 
1123 	rt->dst.lastuse = jiffies;
1124 }
1125 
1126 /* Caller must already hold reference to @from */
rt6_set_from(struct rt6_info * rt,struct fib6_info * from)1127 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1128 {
1129 	rt->rt6i_flags &= ~RTF_EXPIRES;
1130 	rcu_assign_pointer(rt->from, from);
1131 	ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1132 }
1133 
1134 /* Caller must already hold reference to f6i in result */
ip6_rt_copy_init(struct rt6_info * rt,const struct fib6_result * res)1135 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1136 {
1137 	const struct fib6_nh *nh = res->nh;
1138 	const struct net_device *dev = nh->fib_nh_dev;
1139 	struct fib6_info *f6i = res->f6i;
1140 
1141 	ip6_rt_init_dst(rt, res);
1142 
1143 	rt->rt6i_dst = f6i->fib6_dst;
1144 	rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1145 	rt->rt6i_flags = res->fib6_flags;
1146 	if (nh->fib_nh_gw_family) {
1147 		rt->rt6i_gateway = nh->fib_nh_gw6;
1148 		rt->rt6i_flags |= RTF_GATEWAY;
1149 	}
1150 	rt6_set_from(rt, f6i);
1151 #ifdef CONFIG_IPV6_SUBTREES
1152 	rt->rt6i_src = f6i->fib6_src;
1153 #endif
1154 }
1155 
fib6_backtrack(struct fib6_node * fn,struct in6_addr * saddr)1156 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1157 					struct in6_addr *saddr)
1158 {
1159 	struct fib6_node *pn, *sn;
1160 	while (1) {
1161 		if (fn->fn_flags & RTN_TL_ROOT)
1162 			return NULL;
1163 		pn = rcu_dereference(fn->parent);
1164 		sn = FIB6_SUBTREE(pn);
1165 		if (sn && sn != fn)
1166 			fn = fib6_node_lookup(sn, NULL, saddr);
1167 		else
1168 			fn = pn;
1169 		if (fn->fn_flags & RTN_RTINFO)
1170 			return fn;
1171 	}
1172 }
1173 
ip6_hold_safe(struct net * net,struct rt6_info ** prt)1174 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1175 {
1176 	struct rt6_info *rt = *prt;
1177 
1178 	if (dst_hold_safe(&rt->dst))
1179 		return true;
1180 	if (net) {
1181 		rt = net->ipv6.ip6_null_entry;
1182 		dst_hold(&rt->dst);
1183 	} else {
1184 		rt = NULL;
1185 	}
1186 	*prt = rt;
1187 	return false;
1188 }
1189 
1190 /* called with rcu_lock held */
ip6_create_rt_rcu(const struct fib6_result * res)1191 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1192 {
1193 	struct net_device *dev = res->nh->fib_nh_dev;
1194 	struct fib6_info *f6i = res->f6i;
1195 	unsigned short flags;
1196 	struct rt6_info *nrt;
1197 
1198 	if (!fib6_info_hold_safe(f6i))
1199 		goto fallback;
1200 
1201 	flags = fib6_info_dst_flags(f6i);
1202 	nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1203 	if (!nrt) {
1204 		fib6_info_release(f6i);
1205 		goto fallback;
1206 	}
1207 
1208 	ip6_rt_copy_init(nrt, res);
1209 	return nrt;
1210 
1211 fallback:
1212 	nrt = dev_net(dev)->ipv6.ip6_null_entry;
1213 	dst_hold(&nrt->dst);
1214 	return nrt;
1215 }
1216 
ip6_pol_route_lookup(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)1217 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1218 					     struct fib6_table *table,
1219 					     struct flowi6 *fl6,
1220 					     const struct sk_buff *skb,
1221 					     int flags)
1222 {
1223 	struct fib6_result res = {};
1224 	struct fib6_node *fn;
1225 	struct rt6_info *rt;
1226 
1227 	rcu_read_lock();
1228 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1229 restart:
1230 	res.f6i = rcu_dereference(fn->leaf);
1231 	if (!res.f6i)
1232 		res.f6i = net->ipv6.fib6_null_entry;
1233 	else
1234 		rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1235 				 flags);
1236 
1237 	if (res.f6i == net->ipv6.fib6_null_entry) {
1238 		fn = fib6_backtrack(fn, &fl6->saddr);
1239 		if (fn)
1240 			goto restart;
1241 
1242 		rt = net->ipv6.ip6_null_entry;
1243 		dst_hold(&rt->dst);
1244 		goto out;
1245 	} else if (res.fib6_flags & RTF_REJECT) {
1246 		goto do_create;
1247 	}
1248 
1249 	fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1250 			 fl6->flowi6_oif != 0, skb, flags);
1251 
1252 	/* Search through exception table */
1253 	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1254 	if (rt) {
1255 		if (ip6_hold_safe(net, &rt))
1256 			dst_use_noref(&rt->dst, jiffies);
1257 	} else {
1258 do_create:
1259 		rt = ip6_create_rt_rcu(&res);
1260 	}
1261 
1262 out:
1263 	trace_fib6_table_lookup(net, &res, table, fl6);
1264 
1265 	rcu_read_unlock();
1266 
1267 	return rt;
1268 }
1269 
ip6_route_lookup(struct net * net,struct flowi6 * fl6,const struct sk_buff * skb,int flags)1270 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1271 				   const struct sk_buff *skb, int flags)
1272 {
1273 	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1274 }
1275 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1276 
rt6_lookup(struct net * net,const struct in6_addr * daddr,const struct in6_addr * saddr,int oif,const struct sk_buff * skb,int strict)1277 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1278 			    const struct in6_addr *saddr, int oif,
1279 			    const struct sk_buff *skb, int strict)
1280 {
1281 	struct flowi6 fl6 = {
1282 		.flowi6_oif = oif,
1283 		.daddr = *daddr,
1284 	};
1285 	struct dst_entry *dst;
1286 	int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1287 
1288 	if (saddr) {
1289 		memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1290 		flags |= RT6_LOOKUP_F_HAS_SADDR;
1291 	}
1292 
1293 	dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1294 	if (dst->error == 0)
1295 		return dst_rt6_info(dst);
1296 
1297 	dst_release(dst);
1298 
1299 	return NULL;
1300 }
1301 EXPORT_SYMBOL(rt6_lookup);
1302 
1303 /* ip6_ins_rt is called with FREE table->tb6_lock.
1304  * It takes new route entry, the addition fails by any reason the
1305  * route is released.
1306  * Caller must hold dst before calling it.
1307  */
1308 
__ip6_ins_rt(struct fib6_info * rt,struct nl_info * info,struct netlink_ext_ack * extack)1309 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1310 			struct netlink_ext_ack *extack)
1311 {
1312 	int err;
1313 	struct fib6_table *table;
1314 
1315 	table = rt->fib6_table;
1316 	spin_lock_bh(&table->tb6_lock);
1317 	err = fib6_add(&table->tb6_root, rt, info, extack);
1318 	spin_unlock_bh(&table->tb6_lock);
1319 
1320 	return err;
1321 }
1322 
ip6_ins_rt(struct net * net,struct fib6_info * rt)1323 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1324 {
1325 	struct nl_info info = {	.nl_net = net, };
1326 
1327 	return __ip6_ins_rt(rt, &info, NULL);
1328 }
1329 
ip6_rt_cache_alloc(const struct fib6_result * res,const struct in6_addr * daddr,const struct in6_addr * saddr)1330 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1331 					   const struct in6_addr *daddr,
1332 					   const struct in6_addr *saddr)
1333 {
1334 	struct fib6_info *f6i = res->f6i;
1335 	struct net_device *dev;
1336 	struct rt6_info *rt;
1337 
1338 	/*
1339 	 *	Clone the route.
1340 	 */
1341 
1342 	if (!fib6_info_hold_safe(f6i))
1343 		return NULL;
1344 
1345 	dev = ip6_rt_get_dev_rcu(res);
1346 	rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1347 	if (!rt) {
1348 		fib6_info_release(f6i);
1349 		return NULL;
1350 	}
1351 
1352 	ip6_rt_copy_init(rt, res);
1353 	rt->rt6i_flags |= RTF_CACHE;
1354 	rt->rt6i_dst.addr = *daddr;
1355 	rt->rt6i_dst.plen = 128;
1356 
1357 	if (!rt6_is_gw_or_nonexthop(res)) {
1358 		if (f6i->fib6_dst.plen != 128 &&
1359 		    ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1360 			rt->rt6i_flags |= RTF_ANYCAST;
1361 #ifdef CONFIG_IPV6_SUBTREES
1362 		if (rt->rt6i_src.plen && saddr) {
1363 			rt->rt6i_src.addr = *saddr;
1364 			rt->rt6i_src.plen = 128;
1365 		}
1366 #endif
1367 	}
1368 
1369 	return rt;
1370 }
1371 
ip6_rt_pcpu_alloc(const struct fib6_result * res)1372 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1373 {
1374 	struct fib6_info *f6i = res->f6i;
1375 	unsigned short flags = fib6_info_dst_flags(f6i);
1376 	struct net_device *dev;
1377 	struct rt6_info *pcpu_rt;
1378 
1379 	if (!fib6_info_hold_safe(f6i))
1380 		return NULL;
1381 
1382 	rcu_read_lock();
1383 	dev = ip6_rt_get_dev_rcu(res);
1384 	pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1385 	rcu_read_unlock();
1386 	if (!pcpu_rt) {
1387 		fib6_info_release(f6i);
1388 		return NULL;
1389 	}
1390 	ip6_rt_copy_init(pcpu_rt, res);
1391 	pcpu_rt->rt6i_flags |= RTF_PCPU;
1392 
1393 	if (f6i->nh)
1394 		pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1395 
1396 	return pcpu_rt;
1397 }
1398 
rt6_is_valid(const struct rt6_info * rt6)1399 static bool rt6_is_valid(const struct rt6_info *rt6)
1400 {
1401 	return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1402 }
1403 
1404 /* It should be called with rcu_read_lock() acquired */
rt6_get_pcpu_route(const struct fib6_result * res)1405 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1406 {
1407 	struct rt6_info *pcpu_rt;
1408 
1409 	pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1410 
1411 	if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1412 		struct rt6_info *prev, **p;
1413 
1414 		p = this_cpu_ptr(res->nh->rt6i_pcpu);
1415 		/* Paired with READ_ONCE() in __fib6_drop_pcpu_from() */
1416 		prev = xchg(p, NULL);
1417 		if (prev) {
1418 			dst_dev_put(&prev->dst);
1419 			dst_release(&prev->dst);
1420 		}
1421 
1422 		pcpu_rt = NULL;
1423 	}
1424 
1425 	return pcpu_rt;
1426 }
1427 
rt6_make_pcpu_route(struct net * net,const struct fib6_result * res)1428 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1429 					    const struct fib6_result *res)
1430 {
1431 	struct rt6_info *pcpu_rt, *prev, **p;
1432 
1433 	pcpu_rt = ip6_rt_pcpu_alloc(res);
1434 	if (!pcpu_rt)
1435 		return NULL;
1436 
1437 	p = this_cpu_ptr(res->nh->rt6i_pcpu);
1438 	prev = cmpxchg(p, NULL, pcpu_rt);
1439 	BUG_ON(prev);
1440 
1441 	if (res->f6i->fib6_destroying) {
1442 		struct fib6_info *from;
1443 
1444 		from = unrcu_pointer(xchg(&pcpu_rt->from, NULL));
1445 		fib6_info_release(from);
1446 	}
1447 
1448 	return pcpu_rt;
1449 }
1450 
1451 /* exception hash table implementation
1452  */
1453 static DEFINE_SPINLOCK(rt6_exception_lock);
1454 
1455 /* Remove rt6_ex from hash table and free the memory
1456  * Caller must hold rt6_exception_lock
1457  */
rt6_remove_exception(struct rt6_exception_bucket * bucket,struct rt6_exception * rt6_ex)1458 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1459 				 struct rt6_exception *rt6_ex)
1460 {
1461 	struct net *net;
1462 
1463 	if (!bucket || !rt6_ex)
1464 		return;
1465 
1466 	net = dev_net(rt6_ex->rt6i->dst.dev);
1467 	net->ipv6.rt6_stats->fib_rt_cache--;
1468 
1469 	/* purge completely the exception to allow releasing the held resources:
1470 	 * some [sk] cache may keep the dst around for unlimited time
1471 	 */
1472 	dst_dev_put(&rt6_ex->rt6i->dst);
1473 
1474 	hlist_del_rcu(&rt6_ex->hlist);
1475 	dst_release(&rt6_ex->rt6i->dst);
1476 	kfree_rcu(rt6_ex, rcu);
1477 	WARN_ON_ONCE(!bucket->depth);
1478 	bucket->depth--;
1479 }
1480 
1481 /* Remove oldest rt6_ex in bucket and free the memory
1482  * Caller must hold rt6_exception_lock
1483  */
rt6_exception_remove_oldest(struct rt6_exception_bucket * bucket)1484 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1485 {
1486 	struct rt6_exception *rt6_ex, *oldest = NULL;
1487 
1488 	if (!bucket)
1489 		return;
1490 
1491 	hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1492 		if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1493 			oldest = rt6_ex;
1494 	}
1495 	rt6_remove_exception(bucket, oldest);
1496 }
1497 
rt6_exception_hash(const struct in6_addr * dst,const struct in6_addr * src)1498 static u32 rt6_exception_hash(const struct in6_addr *dst,
1499 			      const struct in6_addr *src)
1500 {
1501 	static siphash_aligned_key_t rt6_exception_key;
1502 	struct {
1503 		struct in6_addr dst;
1504 		struct in6_addr src;
1505 	} __aligned(SIPHASH_ALIGNMENT) combined = {
1506 		.dst = *dst,
1507 	};
1508 	u64 val;
1509 
1510 	net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1511 
1512 #ifdef CONFIG_IPV6_SUBTREES
1513 	if (src)
1514 		combined.src = *src;
1515 #endif
1516 	val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1517 
1518 	return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1519 }
1520 
1521 /* Helper function to find the cached rt in the hash table
1522  * and update bucket pointer to point to the bucket for this
1523  * (daddr, saddr) pair
1524  * Caller must hold rt6_exception_lock
1525  */
1526 static struct rt6_exception *
__rt6_find_exception_spinlock(struct rt6_exception_bucket ** bucket,const struct in6_addr * daddr,const struct in6_addr * saddr)1527 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1528 			      const struct in6_addr *daddr,
1529 			      const struct in6_addr *saddr)
1530 {
1531 	struct rt6_exception *rt6_ex;
1532 	u32 hval;
1533 
1534 	if (!(*bucket) || !daddr)
1535 		return NULL;
1536 
1537 	hval = rt6_exception_hash(daddr, saddr);
1538 	*bucket += hval;
1539 
1540 	hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1541 		struct rt6_info *rt6 = rt6_ex->rt6i;
1542 		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1543 
1544 #ifdef CONFIG_IPV6_SUBTREES
1545 		if (matched && saddr)
1546 			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1547 #endif
1548 		if (matched)
1549 			return rt6_ex;
1550 	}
1551 	return NULL;
1552 }
1553 
1554 /* Helper function to find the cached rt in the hash table
1555  * and update bucket pointer to point to the bucket for this
1556  * (daddr, saddr) pair
1557  * Caller must hold rcu_read_lock()
1558  */
1559 static struct rt6_exception *
__rt6_find_exception_rcu(struct rt6_exception_bucket ** bucket,const struct in6_addr * daddr,const struct in6_addr * saddr)1560 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1561 			 const struct in6_addr *daddr,
1562 			 const struct in6_addr *saddr)
1563 {
1564 	struct rt6_exception *rt6_ex;
1565 	u32 hval;
1566 
1567 	WARN_ON_ONCE(!rcu_read_lock_held());
1568 
1569 	if (!(*bucket) || !daddr)
1570 		return NULL;
1571 
1572 	hval = rt6_exception_hash(daddr, saddr);
1573 	*bucket += hval;
1574 
1575 	hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1576 		struct rt6_info *rt6 = rt6_ex->rt6i;
1577 		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1578 
1579 #ifdef CONFIG_IPV6_SUBTREES
1580 		if (matched && saddr)
1581 			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1582 #endif
1583 		if (matched)
1584 			return rt6_ex;
1585 	}
1586 	return NULL;
1587 }
1588 
fib6_mtu(const struct fib6_result * res)1589 static unsigned int fib6_mtu(const struct fib6_result *res)
1590 {
1591 	const struct fib6_nh *nh = res->nh;
1592 	unsigned int mtu;
1593 
1594 	if (res->f6i->fib6_pmtu) {
1595 		mtu = res->f6i->fib6_pmtu;
1596 	} else {
1597 		struct net_device *dev = nh->fib_nh_dev;
1598 		struct inet6_dev *idev;
1599 
1600 		rcu_read_lock();
1601 		idev = __in6_dev_get(dev);
1602 		mtu = READ_ONCE(idev->cnf.mtu6);
1603 		rcu_read_unlock();
1604 	}
1605 
1606 	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1607 
1608 	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1609 }
1610 
1611 #define FIB6_EXCEPTION_BUCKET_FLUSHED  0x1UL
1612 
1613 /* used when the flushed bit is not relevant, only access to the bucket
1614  * (ie., all bucket users except rt6_insert_exception);
1615  *
1616  * called under rcu lock; sometimes called with rt6_exception_lock held
1617  */
1618 static
fib6_nh_get_excptn_bucket(const struct fib6_nh * nh,spinlock_t * lock)1619 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1620 						       spinlock_t *lock)
1621 {
1622 	struct rt6_exception_bucket *bucket;
1623 
1624 	if (lock)
1625 		bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1626 						   lockdep_is_held(lock));
1627 	else
1628 		bucket = rcu_dereference(nh->rt6i_exception_bucket);
1629 
1630 	/* remove bucket flushed bit if set */
1631 	if (bucket) {
1632 		unsigned long p = (unsigned long)bucket;
1633 
1634 		p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1635 		bucket = (struct rt6_exception_bucket *)p;
1636 	}
1637 
1638 	return bucket;
1639 }
1640 
fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket * bucket)1641 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1642 {
1643 	unsigned long p = (unsigned long)bucket;
1644 
1645 	return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1646 }
1647 
1648 /* called with rt6_exception_lock held */
fib6_nh_excptn_bucket_set_flushed(struct fib6_nh * nh,spinlock_t * lock)1649 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1650 					      spinlock_t *lock)
1651 {
1652 	struct rt6_exception_bucket *bucket;
1653 	unsigned long p;
1654 
1655 	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1656 					   lockdep_is_held(lock));
1657 
1658 	p = (unsigned long)bucket;
1659 	p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1660 	bucket = (struct rt6_exception_bucket *)p;
1661 	rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1662 }
1663 
rt6_insert_exception(struct rt6_info * nrt,const struct fib6_result * res)1664 static int rt6_insert_exception(struct rt6_info *nrt,
1665 				const struct fib6_result *res)
1666 {
1667 	struct net *net = dev_net(nrt->dst.dev);
1668 	struct rt6_exception_bucket *bucket;
1669 	struct fib6_info *f6i = res->f6i;
1670 	struct in6_addr *src_key = NULL;
1671 	struct rt6_exception *rt6_ex;
1672 	struct fib6_nh *nh = res->nh;
1673 	int max_depth;
1674 	int err = 0;
1675 
1676 	spin_lock_bh(&rt6_exception_lock);
1677 
1678 	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1679 					  lockdep_is_held(&rt6_exception_lock));
1680 	if (!bucket) {
1681 		bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1682 				 GFP_ATOMIC);
1683 		if (!bucket) {
1684 			err = -ENOMEM;
1685 			goto out;
1686 		}
1687 		rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1688 	} else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1689 		err = -EINVAL;
1690 		goto out;
1691 	}
1692 
1693 #ifdef CONFIG_IPV6_SUBTREES
1694 	/* fib6_src.plen != 0 indicates f6i is in subtree
1695 	 * and exception table is indexed by a hash of
1696 	 * both fib6_dst and fib6_src.
1697 	 * Otherwise, the exception table is indexed by
1698 	 * a hash of only fib6_dst.
1699 	 */
1700 	if (f6i->fib6_src.plen)
1701 		src_key = &nrt->rt6i_src.addr;
1702 #endif
1703 	/* rt6_mtu_change() might lower mtu on f6i.
1704 	 * Only insert this exception route if its mtu
1705 	 * is less than f6i's mtu value.
1706 	 */
1707 	if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1708 		err = -EINVAL;
1709 		goto out;
1710 	}
1711 
1712 	rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1713 					       src_key);
1714 	if (rt6_ex)
1715 		rt6_remove_exception(bucket, rt6_ex);
1716 
1717 	rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1718 	if (!rt6_ex) {
1719 		err = -ENOMEM;
1720 		goto out;
1721 	}
1722 	rt6_ex->rt6i = nrt;
1723 	rt6_ex->stamp = jiffies;
1724 	hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1725 	bucket->depth++;
1726 	net->ipv6.rt6_stats->fib_rt_cache++;
1727 
1728 	/* Randomize max depth to avoid some side channels attacks. */
1729 	max_depth = FIB6_MAX_DEPTH + get_random_u32_below(FIB6_MAX_DEPTH);
1730 	while (bucket->depth > max_depth)
1731 		rt6_exception_remove_oldest(bucket);
1732 
1733 out:
1734 	spin_unlock_bh(&rt6_exception_lock);
1735 
1736 	/* Update fn->fn_sernum to invalidate all cached dst */
1737 	if (!err) {
1738 		spin_lock_bh(&f6i->fib6_table->tb6_lock);
1739 		fib6_update_sernum(net, f6i);
1740 		spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1741 		fib6_force_start_gc(net);
1742 	}
1743 
1744 	return err;
1745 }
1746 
fib6_nh_flush_exceptions(struct fib6_nh * nh,struct fib6_info * from)1747 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1748 {
1749 	struct rt6_exception_bucket *bucket;
1750 	struct rt6_exception *rt6_ex;
1751 	struct hlist_node *tmp;
1752 	int i;
1753 
1754 	spin_lock_bh(&rt6_exception_lock);
1755 
1756 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1757 	if (!bucket)
1758 		goto out;
1759 
1760 	/* Prevent rt6_insert_exception() to recreate the bucket list */
1761 	if (!from)
1762 		fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1763 
1764 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1765 		hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1766 			if (!from ||
1767 			    rcu_access_pointer(rt6_ex->rt6i->from) == from)
1768 				rt6_remove_exception(bucket, rt6_ex);
1769 		}
1770 		WARN_ON_ONCE(!from && bucket->depth);
1771 		bucket++;
1772 	}
1773 out:
1774 	spin_unlock_bh(&rt6_exception_lock);
1775 }
1776 
rt6_nh_flush_exceptions(struct fib6_nh * nh,void * arg)1777 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1778 {
1779 	struct fib6_info *f6i = arg;
1780 
1781 	fib6_nh_flush_exceptions(nh, f6i);
1782 
1783 	return 0;
1784 }
1785 
rt6_flush_exceptions(struct fib6_info * f6i)1786 void rt6_flush_exceptions(struct fib6_info *f6i)
1787 {
1788 	if (f6i->nh)
1789 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1790 					 f6i);
1791 	else
1792 		fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1793 }
1794 
1795 /* Find cached rt in the hash table inside passed in rt
1796  * Caller has to hold rcu_read_lock()
1797  */
rt6_find_cached_rt(const struct fib6_result * res,const struct in6_addr * daddr,const struct in6_addr * saddr)1798 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1799 					   const struct in6_addr *daddr,
1800 					   const struct in6_addr *saddr)
1801 {
1802 	const struct in6_addr *src_key = NULL;
1803 	struct rt6_exception_bucket *bucket;
1804 	struct rt6_exception *rt6_ex;
1805 	struct rt6_info *ret = NULL;
1806 
1807 #ifdef CONFIG_IPV6_SUBTREES
1808 	/* fib6i_src.plen != 0 indicates f6i is in subtree
1809 	 * and exception table is indexed by a hash of
1810 	 * both fib6_dst and fib6_src.
1811 	 * However, the src addr used to create the hash
1812 	 * might not be exactly the passed in saddr which
1813 	 * is a /128 addr from the flow.
1814 	 * So we need to use f6i->fib6_src to redo lookup
1815 	 * if the passed in saddr does not find anything.
1816 	 * (See the logic in ip6_rt_cache_alloc() on how
1817 	 * rt->rt6i_src is updated.)
1818 	 */
1819 	if (res->f6i->fib6_src.plen)
1820 		src_key = saddr;
1821 find_ex:
1822 #endif
1823 	bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1824 	rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1825 
1826 	if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1827 		ret = rt6_ex->rt6i;
1828 
1829 #ifdef CONFIG_IPV6_SUBTREES
1830 	/* Use fib6_src as src_key and redo lookup */
1831 	if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1832 		src_key = &res->f6i->fib6_src.addr;
1833 		goto find_ex;
1834 	}
1835 #endif
1836 
1837 	return ret;
1838 }
1839 
1840 /* Remove the passed in cached rt from the hash table that contains it */
fib6_nh_remove_exception(const struct fib6_nh * nh,int plen,const struct rt6_info * rt)1841 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1842 				    const struct rt6_info *rt)
1843 {
1844 	const struct in6_addr *src_key = NULL;
1845 	struct rt6_exception_bucket *bucket;
1846 	struct rt6_exception *rt6_ex;
1847 	int err;
1848 
1849 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1850 		return -ENOENT;
1851 
1852 	spin_lock_bh(&rt6_exception_lock);
1853 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1854 
1855 #ifdef CONFIG_IPV6_SUBTREES
1856 	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1857 	 * and exception table is indexed by a hash of
1858 	 * both rt6i_dst and rt6i_src.
1859 	 * Otherwise, the exception table is indexed by
1860 	 * a hash of only rt6i_dst.
1861 	 */
1862 	if (plen)
1863 		src_key = &rt->rt6i_src.addr;
1864 #endif
1865 	rt6_ex = __rt6_find_exception_spinlock(&bucket,
1866 					       &rt->rt6i_dst.addr,
1867 					       src_key);
1868 	if (rt6_ex) {
1869 		rt6_remove_exception(bucket, rt6_ex);
1870 		err = 0;
1871 	} else {
1872 		err = -ENOENT;
1873 	}
1874 
1875 	spin_unlock_bh(&rt6_exception_lock);
1876 	return err;
1877 }
1878 
1879 struct fib6_nh_excptn_arg {
1880 	struct rt6_info	*rt;
1881 	int		plen;
1882 };
1883 
rt6_nh_remove_exception_rt(struct fib6_nh * nh,void * _arg)1884 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1885 {
1886 	struct fib6_nh_excptn_arg *arg = _arg;
1887 	int err;
1888 
1889 	err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1890 	if (err == 0)
1891 		return 1;
1892 
1893 	return 0;
1894 }
1895 
rt6_remove_exception_rt(struct rt6_info * rt)1896 static int rt6_remove_exception_rt(struct rt6_info *rt)
1897 {
1898 	struct fib6_info *from;
1899 
1900 	from = rcu_dereference(rt->from);
1901 	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1902 		return -EINVAL;
1903 
1904 	if (from->nh) {
1905 		struct fib6_nh_excptn_arg arg = {
1906 			.rt = rt,
1907 			.plen = from->fib6_src.plen
1908 		};
1909 		int rc;
1910 
1911 		/* rc = 1 means an entry was found */
1912 		rc = nexthop_for_each_fib6_nh(from->nh,
1913 					      rt6_nh_remove_exception_rt,
1914 					      &arg);
1915 		return rc ? 0 : -ENOENT;
1916 	}
1917 
1918 	return fib6_nh_remove_exception(from->fib6_nh,
1919 					from->fib6_src.plen, rt);
1920 }
1921 
1922 /* Find rt6_ex which contains the passed in rt cache and
1923  * refresh its stamp
1924  */
fib6_nh_update_exception(const struct fib6_nh * nh,int plen,const struct rt6_info * rt)1925 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1926 				     const struct rt6_info *rt)
1927 {
1928 	const struct in6_addr *src_key = NULL;
1929 	struct rt6_exception_bucket *bucket;
1930 	struct rt6_exception *rt6_ex;
1931 
1932 	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1933 #ifdef CONFIG_IPV6_SUBTREES
1934 	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1935 	 * and exception table is indexed by a hash of
1936 	 * both rt6i_dst and rt6i_src.
1937 	 * Otherwise, the exception table is indexed by
1938 	 * a hash of only rt6i_dst.
1939 	 */
1940 	if (plen)
1941 		src_key = &rt->rt6i_src.addr;
1942 #endif
1943 	rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1944 	if (rt6_ex)
1945 		rt6_ex->stamp = jiffies;
1946 }
1947 
1948 struct fib6_nh_match_arg {
1949 	const struct net_device *dev;
1950 	const struct in6_addr	*gw;
1951 	struct fib6_nh		*match;
1952 };
1953 
1954 /* determine if fib6_nh has given device and gateway */
fib6_nh_find_match(struct fib6_nh * nh,void * _arg)1955 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1956 {
1957 	struct fib6_nh_match_arg *arg = _arg;
1958 
1959 	if (arg->dev != nh->fib_nh_dev ||
1960 	    (arg->gw && !nh->fib_nh_gw_family) ||
1961 	    (!arg->gw && nh->fib_nh_gw_family) ||
1962 	    (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1963 		return 0;
1964 
1965 	arg->match = nh;
1966 
1967 	/* found a match, break the loop */
1968 	return 1;
1969 }
1970 
rt6_update_exception_stamp_rt(struct rt6_info * rt)1971 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1972 {
1973 	struct fib6_info *from;
1974 	struct fib6_nh *fib6_nh;
1975 
1976 	rcu_read_lock();
1977 
1978 	from = rcu_dereference(rt->from);
1979 	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1980 		goto unlock;
1981 
1982 	if (from->nh) {
1983 		struct fib6_nh_match_arg arg = {
1984 			.dev = rt->dst.dev,
1985 			.gw = &rt->rt6i_gateway,
1986 		};
1987 
1988 		nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1989 
1990 		if (!arg.match)
1991 			goto unlock;
1992 		fib6_nh = arg.match;
1993 	} else {
1994 		fib6_nh = from->fib6_nh;
1995 	}
1996 	fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1997 unlock:
1998 	rcu_read_unlock();
1999 }
2000 
rt6_mtu_change_route_allowed(struct inet6_dev * idev,struct rt6_info * rt,int mtu)2001 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
2002 					 struct rt6_info *rt, int mtu)
2003 {
2004 	/* If the new MTU is lower than the route PMTU, this new MTU will be the
2005 	 * lowest MTU in the path: always allow updating the route PMTU to
2006 	 * reflect PMTU decreases.
2007 	 *
2008 	 * If the new MTU is higher, and the route PMTU is equal to the local
2009 	 * MTU, this means the old MTU is the lowest in the path, so allow
2010 	 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2011 	 * handle this.
2012 	 */
2013 
2014 	if (dst_mtu(&rt->dst) >= mtu)
2015 		return true;
2016 
2017 	if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2018 		return true;
2019 
2020 	return false;
2021 }
2022 
rt6_exceptions_update_pmtu(struct inet6_dev * idev,const struct fib6_nh * nh,int mtu)2023 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2024 				       const struct fib6_nh *nh, int mtu)
2025 {
2026 	struct rt6_exception_bucket *bucket;
2027 	struct rt6_exception *rt6_ex;
2028 	int i;
2029 
2030 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2031 	if (!bucket)
2032 		return;
2033 
2034 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2035 		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2036 			struct rt6_info *entry = rt6_ex->rt6i;
2037 
2038 			/* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2039 			 * route), the metrics of its rt->from have already
2040 			 * been updated.
2041 			 */
2042 			if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2043 			    rt6_mtu_change_route_allowed(idev, entry, mtu))
2044 				dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2045 		}
2046 		bucket++;
2047 	}
2048 }
2049 
2050 #define RTF_CACHE_GATEWAY	(RTF_GATEWAY | RTF_CACHE)
2051 
fib6_nh_exceptions_clean_tohost(const struct fib6_nh * nh,const struct in6_addr * gateway)2052 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2053 					    const struct in6_addr *gateway)
2054 {
2055 	struct rt6_exception_bucket *bucket;
2056 	struct rt6_exception *rt6_ex;
2057 	struct hlist_node *tmp;
2058 	int i;
2059 
2060 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2061 		return;
2062 
2063 	spin_lock_bh(&rt6_exception_lock);
2064 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2065 	if (bucket) {
2066 		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2067 			hlist_for_each_entry_safe(rt6_ex, tmp,
2068 						  &bucket->chain, hlist) {
2069 				struct rt6_info *entry = rt6_ex->rt6i;
2070 
2071 				if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2072 				    RTF_CACHE_GATEWAY &&
2073 				    ipv6_addr_equal(gateway,
2074 						    &entry->rt6i_gateway)) {
2075 					rt6_remove_exception(bucket, rt6_ex);
2076 				}
2077 			}
2078 			bucket++;
2079 		}
2080 	}
2081 
2082 	spin_unlock_bh(&rt6_exception_lock);
2083 }
2084 
rt6_age_examine_exception(struct rt6_exception_bucket * bucket,struct rt6_exception * rt6_ex,struct fib6_gc_args * gc_args,unsigned long now)2085 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2086 				      struct rt6_exception *rt6_ex,
2087 				      struct fib6_gc_args *gc_args,
2088 				      unsigned long now)
2089 {
2090 	struct rt6_info *rt = rt6_ex->rt6i;
2091 
2092 	/* we are pruning and obsoleting aged-out and non gateway exceptions
2093 	 * even if others have still references to them, so that on next
2094 	 * dst_check() such references can be dropped.
2095 	 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2096 	 * expired, independently from their aging, as per RFC 8201 section 4
2097 	 */
2098 	if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2099 		if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2100 			pr_debug("aging clone %p\n", rt);
2101 			rt6_remove_exception(bucket, rt6_ex);
2102 			return;
2103 		}
2104 	} else if (time_after(jiffies, rt->dst.expires)) {
2105 		pr_debug("purging expired route %p\n", rt);
2106 		rt6_remove_exception(bucket, rt6_ex);
2107 		return;
2108 	}
2109 
2110 	if (rt->rt6i_flags & RTF_GATEWAY) {
2111 		struct neighbour *neigh;
2112 
2113 		neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2114 
2115 		if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2116 			pr_debug("purging route %p via non-router but gateway\n",
2117 				 rt);
2118 			rt6_remove_exception(bucket, rt6_ex);
2119 			return;
2120 		}
2121 	}
2122 
2123 	gc_args->more++;
2124 }
2125 
fib6_nh_age_exceptions(const struct fib6_nh * nh,struct fib6_gc_args * gc_args,unsigned long now)2126 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2127 				   struct fib6_gc_args *gc_args,
2128 				   unsigned long now)
2129 {
2130 	struct rt6_exception_bucket *bucket;
2131 	struct rt6_exception *rt6_ex;
2132 	struct hlist_node *tmp;
2133 	int i;
2134 
2135 	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2136 		return;
2137 
2138 	rcu_read_lock_bh();
2139 	spin_lock(&rt6_exception_lock);
2140 	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2141 	if (bucket) {
2142 		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2143 			hlist_for_each_entry_safe(rt6_ex, tmp,
2144 						  &bucket->chain, hlist) {
2145 				rt6_age_examine_exception(bucket, rt6_ex,
2146 							  gc_args, now);
2147 			}
2148 			bucket++;
2149 		}
2150 	}
2151 	spin_unlock(&rt6_exception_lock);
2152 	rcu_read_unlock_bh();
2153 }
2154 
2155 struct fib6_nh_age_excptn_arg {
2156 	struct fib6_gc_args	*gc_args;
2157 	unsigned long		now;
2158 };
2159 
rt6_nh_age_exceptions(struct fib6_nh * nh,void * _arg)2160 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2161 {
2162 	struct fib6_nh_age_excptn_arg *arg = _arg;
2163 
2164 	fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2165 	return 0;
2166 }
2167 
rt6_age_exceptions(struct fib6_info * f6i,struct fib6_gc_args * gc_args,unsigned long now)2168 void rt6_age_exceptions(struct fib6_info *f6i,
2169 			struct fib6_gc_args *gc_args,
2170 			unsigned long now)
2171 {
2172 	if (f6i->nh) {
2173 		struct fib6_nh_age_excptn_arg arg = {
2174 			.gc_args = gc_args,
2175 			.now = now
2176 		};
2177 
2178 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2179 					 &arg);
2180 	} else {
2181 		fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2182 	}
2183 }
2184 
2185 /* must be called with rcu lock held */
fib6_table_lookup(struct net * net,struct fib6_table * table,int oif,struct flowi6 * fl6,struct fib6_result * res,int strict)2186 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2187 		      struct flowi6 *fl6, struct fib6_result *res, int strict)
2188 {
2189 	struct fib6_node *fn, *saved_fn;
2190 
2191 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2192 	saved_fn = fn;
2193 
2194 redo_rt6_select:
2195 	rt6_select(net, fn, oif, res, strict);
2196 	if (res->f6i == net->ipv6.fib6_null_entry) {
2197 		fn = fib6_backtrack(fn, &fl6->saddr);
2198 		if (fn)
2199 			goto redo_rt6_select;
2200 		else if (strict & RT6_LOOKUP_F_REACHABLE) {
2201 			/* also consider unreachable route */
2202 			strict &= ~RT6_LOOKUP_F_REACHABLE;
2203 			fn = saved_fn;
2204 			goto redo_rt6_select;
2205 		}
2206 	}
2207 
2208 	trace_fib6_table_lookup(net, res, table, fl6);
2209 
2210 	return 0;
2211 }
2212 
ip6_pol_route(struct net * net,struct fib6_table * table,int oif,struct flowi6 * fl6,const struct sk_buff * skb,int flags)2213 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2214 			       int oif, struct flowi6 *fl6,
2215 			       const struct sk_buff *skb, int flags)
2216 {
2217 	struct fib6_result res = {};
2218 	struct rt6_info *rt = NULL;
2219 	int strict = 0;
2220 
2221 	WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2222 		     !rcu_read_lock_held());
2223 
2224 	strict |= flags & RT6_LOOKUP_F_IFACE;
2225 	strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2226 	if (READ_ONCE(net->ipv6.devconf_all->forwarding) == 0)
2227 		strict |= RT6_LOOKUP_F_REACHABLE;
2228 
2229 	rcu_read_lock();
2230 
2231 	fib6_table_lookup(net, table, oif, fl6, &res, strict);
2232 	if (res.f6i == net->ipv6.fib6_null_entry)
2233 		goto out;
2234 
2235 	fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2236 
2237 	/*Search through exception table */
2238 	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2239 	if (rt) {
2240 		goto out;
2241 	} else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2242 			    !res.nh->fib_nh_gw_family)) {
2243 		/* Create a RTF_CACHE clone which will not be
2244 		 * owned by the fib6 tree.  It is for the special case where
2245 		 * the daddr in the skb during the neighbor look-up is different
2246 		 * from the fl6->daddr used to look-up route here.
2247 		 */
2248 		rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2249 
2250 		if (rt) {
2251 			/* 1 refcnt is taken during ip6_rt_cache_alloc().
2252 			 * As rt6_uncached_list_add() does not consume refcnt,
2253 			 * this refcnt is always returned to the caller even
2254 			 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2255 			 */
2256 			rt6_uncached_list_add(rt);
2257 			rcu_read_unlock();
2258 
2259 			return rt;
2260 		}
2261 	} else {
2262 		/* Get a percpu copy */
2263 		local_bh_disable();
2264 		rt = rt6_get_pcpu_route(&res);
2265 
2266 		if (!rt)
2267 			rt = rt6_make_pcpu_route(net, &res);
2268 
2269 		local_bh_enable();
2270 	}
2271 out:
2272 	if (!rt)
2273 		rt = net->ipv6.ip6_null_entry;
2274 	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2275 		ip6_hold_safe(net, &rt);
2276 	rcu_read_unlock();
2277 
2278 	return rt;
2279 }
2280 EXPORT_SYMBOL_GPL(ip6_pol_route);
2281 
ip6_pol_route_input(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)2282 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2283 					    struct fib6_table *table,
2284 					    struct flowi6 *fl6,
2285 					    const struct sk_buff *skb,
2286 					    int flags)
2287 {
2288 	return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2289 }
2290 
ip6_route_input_lookup(struct net * net,struct net_device * dev,struct flowi6 * fl6,const struct sk_buff * skb,int flags)2291 struct dst_entry *ip6_route_input_lookup(struct net *net,
2292 					 struct net_device *dev,
2293 					 struct flowi6 *fl6,
2294 					 const struct sk_buff *skb,
2295 					 int flags)
2296 {
2297 	if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2298 		flags |= RT6_LOOKUP_F_IFACE;
2299 
2300 	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2301 }
2302 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2303 
ip6_multipath_l3_keys(const struct sk_buff * skb,struct flow_keys * keys,struct flow_keys * flkeys)2304 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2305 				  struct flow_keys *keys,
2306 				  struct flow_keys *flkeys)
2307 {
2308 	const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2309 	const struct ipv6hdr *key_iph = outer_iph;
2310 	struct flow_keys *_flkeys = flkeys;
2311 	const struct ipv6hdr *inner_iph;
2312 	const struct icmp6hdr *icmph;
2313 	struct ipv6hdr _inner_iph;
2314 	struct icmp6hdr _icmph;
2315 
2316 	if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2317 		goto out;
2318 
2319 	icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2320 				   sizeof(_icmph), &_icmph);
2321 	if (!icmph)
2322 		goto out;
2323 
2324 	if (!icmpv6_is_err(icmph->icmp6_type))
2325 		goto out;
2326 
2327 	inner_iph = skb_header_pointer(skb,
2328 				       skb_transport_offset(skb) + sizeof(*icmph),
2329 				       sizeof(_inner_iph), &_inner_iph);
2330 	if (!inner_iph)
2331 		goto out;
2332 
2333 	key_iph = inner_iph;
2334 	_flkeys = NULL;
2335 out:
2336 	if (_flkeys) {
2337 		keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2338 		keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2339 		keys->tags.flow_label = _flkeys->tags.flow_label;
2340 		keys->basic.ip_proto = _flkeys->basic.ip_proto;
2341 	} else {
2342 		keys->addrs.v6addrs.src = key_iph->saddr;
2343 		keys->addrs.v6addrs.dst = key_iph->daddr;
2344 		keys->tags.flow_label = ip6_flowlabel(key_iph);
2345 		keys->basic.ip_proto = key_iph->nexthdr;
2346 	}
2347 }
2348 
rt6_multipath_custom_hash_outer(const struct net * net,const struct sk_buff * skb,bool * p_has_inner)2349 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2350 					   const struct sk_buff *skb,
2351 					   bool *p_has_inner)
2352 {
2353 	u32 hash_fields = ip6_multipath_hash_fields(net);
2354 	struct flow_keys keys, hash_keys;
2355 
2356 	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2357 		return 0;
2358 
2359 	memset(&hash_keys, 0, sizeof(hash_keys));
2360 	skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2361 
2362 	hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2363 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2364 		hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2365 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2366 		hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2367 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2368 		hash_keys.basic.ip_proto = keys.basic.ip_proto;
2369 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2370 		hash_keys.tags.flow_label = keys.tags.flow_label;
2371 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2372 		hash_keys.ports.src = keys.ports.src;
2373 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2374 		hash_keys.ports.dst = keys.ports.dst;
2375 
2376 	*p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2377 	return fib_multipath_hash_from_keys(net, &hash_keys);
2378 }
2379 
rt6_multipath_custom_hash_inner(const struct net * net,const struct sk_buff * skb,bool has_inner)2380 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2381 					   const struct sk_buff *skb,
2382 					   bool has_inner)
2383 {
2384 	u32 hash_fields = ip6_multipath_hash_fields(net);
2385 	struct flow_keys keys, hash_keys;
2386 
2387 	/* We assume the packet carries an encapsulation, but if none was
2388 	 * encountered during dissection of the outer flow, then there is no
2389 	 * point in calling the flow dissector again.
2390 	 */
2391 	if (!has_inner)
2392 		return 0;
2393 
2394 	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2395 		return 0;
2396 
2397 	memset(&hash_keys, 0, sizeof(hash_keys));
2398 	skb_flow_dissect_flow_keys(skb, &keys, 0);
2399 
2400 	if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2401 		return 0;
2402 
2403 	if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2404 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2405 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2406 			hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2407 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2408 			hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2409 	} else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2410 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2411 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2412 			hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2413 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2414 			hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2415 		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2416 			hash_keys.tags.flow_label = keys.tags.flow_label;
2417 	}
2418 
2419 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2420 		hash_keys.basic.ip_proto = keys.basic.ip_proto;
2421 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2422 		hash_keys.ports.src = keys.ports.src;
2423 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2424 		hash_keys.ports.dst = keys.ports.dst;
2425 
2426 	return fib_multipath_hash_from_keys(net, &hash_keys);
2427 }
2428 
rt6_multipath_custom_hash_skb(const struct net * net,const struct sk_buff * skb)2429 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2430 					 const struct sk_buff *skb)
2431 {
2432 	u32 mhash, mhash_inner;
2433 	bool has_inner = true;
2434 
2435 	mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2436 	mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2437 
2438 	return jhash_2words(mhash, mhash_inner, 0);
2439 }
2440 
rt6_multipath_custom_hash_fl6(const struct net * net,const struct flowi6 * fl6)2441 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2442 					 const struct flowi6 *fl6)
2443 {
2444 	u32 hash_fields = ip6_multipath_hash_fields(net);
2445 	struct flow_keys hash_keys;
2446 
2447 	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2448 		return 0;
2449 
2450 	memset(&hash_keys, 0, sizeof(hash_keys));
2451 	hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2452 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2453 		hash_keys.addrs.v6addrs.src = fl6->saddr;
2454 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2455 		hash_keys.addrs.v6addrs.dst = fl6->daddr;
2456 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2457 		hash_keys.basic.ip_proto = fl6->flowi6_proto;
2458 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2459 		hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2460 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2461 		hash_keys.ports.src = fl6->fl6_sport;
2462 	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2463 		hash_keys.ports.dst = fl6->fl6_dport;
2464 
2465 	return fib_multipath_hash_from_keys(net, &hash_keys);
2466 }
2467 
2468 /* if skb is set it will be used and fl6 can be NULL */
rt6_multipath_hash(const struct net * net,const struct flowi6 * fl6,const struct sk_buff * skb,struct flow_keys * flkeys)2469 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2470 		       const struct sk_buff *skb, struct flow_keys *flkeys)
2471 {
2472 	struct flow_keys hash_keys;
2473 	u32 mhash = 0;
2474 
2475 	switch (ip6_multipath_hash_policy(net)) {
2476 	case 0:
2477 		memset(&hash_keys, 0, sizeof(hash_keys));
2478 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2479 		if (skb) {
2480 			ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2481 		} else {
2482 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2483 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2484 			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2485 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2486 		}
2487 		mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2488 		break;
2489 	case 1:
2490 		if (skb) {
2491 			unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2492 			struct flow_keys keys;
2493 
2494 			/* short-circuit if we already have L4 hash present */
2495 			if (skb->l4_hash)
2496 				return skb_get_hash_raw(skb) >> 1;
2497 
2498 			memset(&hash_keys, 0, sizeof(hash_keys));
2499 
2500 			if (!flkeys) {
2501 				skb_flow_dissect_flow_keys(skb, &keys, flag);
2502 				flkeys = &keys;
2503 			}
2504 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2505 			hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2506 			hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2507 			hash_keys.ports.src = flkeys->ports.src;
2508 			hash_keys.ports.dst = flkeys->ports.dst;
2509 			hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2510 		} else {
2511 			memset(&hash_keys, 0, sizeof(hash_keys));
2512 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2513 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2514 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2515 			hash_keys.ports.src = fl6->fl6_sport;
2516 			hash_keys.ports.dst = fl6->fl6_dport;
2517 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2518 		}
2519 		mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2520 		break;
2521 	case 2:
2522 		memset(&hash_keys, 0, sizeof(hash_keys));
2523 		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2524 		if (skb) {
2525 			struct flow_keys keys;
2526 
2527 			if (!flkeys) {
2528 				skb_flow_dissect_flow_keys(skb, &keys, 0);
2529 				flkeys = &keys;
2530 			}
2531 
2532 			/* Inner can be v4 or v6 */
2533 			if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2534 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2535 				hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2536 				hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2537 			} else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2538 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2539 				hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2540 				hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2541 				hash_keys.tags.flow_label = flkeys->tags.flow_label;
2542 				hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2543 			} else {
2544 				/* Same as case 0 */
2545 				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2546 				ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2547 			}
2548 		} else {
2549 			/* Same as case 0 */
2550 			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2551 			hash_keys.addrs.v6addrs.src = fl6->saddr;
2552 			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2553 			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2554 			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2555 		}
2556 		mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2557 		break;
2558 	case 3:
2559 		if (skb)
2560 			mhash = rt6_multipath_custom_hash_skb(net, skb);
2561 		else
2562 			mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2563 		break;
2564 	}
2565 
2566 	return mhash >> 1;
2567 }
2568 
2569 /* Called with rcu held */
ip6_route_input(struct sk_buff * skb)2570 void ip6_route_input(struct sk_buff *skb)
2571 {
2572 	const struct ipv6hdr *iph = ipv6_hdr(skb);
2573 	struct net *net = dev_net(skb->dev);
2574 	int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2575 	struct ip_tunnel_info *tun_info;
2576 	struct flowi6 fl6 = {
2577 		.flowi6_iif = skb->dev->ifindex,
2578 		.daddr = iph->daddr,
2579 		.saddr = iph->saddr,
2580 		.flowlabel = ip6_flowinfo(iph),
2581 		.flowi6_mark = skb->mark,
2582 		.flowi6_proto = iph->nexthdr,
2583 	};
2584 	struct flow_keys *flkeys = NULL, _flkeys;
2585 
2586 	tun_info = skb_tunnel_info(skb);
2587 	if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2588 		fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2589 
2590 	if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2591 		flkeys = &_flkeys;
2592 
2593 	if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2594 		fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2595 	skb_dst_drop(skb);
2596 	skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2597 						      &fl6, skb, flags));
2598 }
2599 
ip6_pol_route_output(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)2600 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2601 					     struct fib6_table *table,
2602 					     struct flowi6 *fl6,
2603 					     const struct sk_buff *skb,
2604 					     int flags)
2605 {
2606 	return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2607 }
2608 
ip6_route_output_flags_noref(struct net * net,const struct sock * sk,struct flowi6 * fl6,int flags)2609 static struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2610 						      const struct sock *sk,
2611 						      struct flowi6 *fl6,
2612 						      int flags)
2613 {
2614 	bool any_src;
2615 
2616 	if (ipv6_addr_type(&fl6->daddr) &
2617 	    (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2618 		struct dst_entry *dst;
2619 
2620 		/* This function does not take refcnt on the dst */
2621 		dst = l3mdev_link_scope_lookup(net, fl6);
2622 		if (dst)
2623 			return dst;
2624 	}
2625 
2626 	fl6->flowi6_iif = LOOPBACK_IFINDEX;
2627 
2628 	flags |= RT6_LOOKUP_F_DST_NOREF;
2629 	any_src = ipv6_addr_any(&fl6->saddr);
2630 	if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2631 	    (fl6->flowi6_oif && any_src))
2632 		flags |= RT6_LOOKUP_F_IFACE;
2633 
2634 	if (!any_src)
2635 		flags |= RT6_LOOKUP_F_HAS_SADDR;
2636 	else if (sk)
2637 		flags |= rt6_srcprefs2flags(READ_ONCE(inet6_sk(sk)->srcprefs));
2638 
2639 	return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2640 }
2641 
ip6_route_output_flags(struct net * net,const struct sock * sk,struct flowi6 * fl6,int flags)2642 struct dst_entry *ip6_route_output_flags(struct net *net,
2643 					 const struct sock *sk,
2644 					 struct flowi6 *fl6,
2645 					 int flags)
2646 {
2647 	struct dst_entry *dst;
2648 	struct rt6_info *rt6;
2649 
2650 	rcu_read_lock();
2651 	dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2652 	rt6 = dst_rt6_info(dst);
2653 	/* For dst cached in uncached_list, refcnt is already taken. */
2654 	if (list_empty(&rt6->dst.rt_uncached) && !dst_hold_safe(dst)) {
2655 		dst = &net->ipv6.ip6_null_entry->dst;
2656 		dst_hold(dst);
2657 	}
2658 	rcu_read_unlock();
2659 
2660 	return dst;
2661 }
2662 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2663 
ip6_blackhole_route(struct net * net,struct dst_entry * dst_orig)2664 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2665 {
2666 	struct rt6_info *rt, *ort = dst_rt6_info(dst_orig);
2667 	struct net_device *loopback_dev = net->loopback_dev;
2668 	struct dst_entry *new = NULL;
2669 
2670 	rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev,
2671 		       DST_OBSOLETE_DEAD, 0);
2672 	if (rt) {
2673 		rt6_info_init(rt);
2674 		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2675 
2676 		new = &rt->dst;
2677 		new->__use = 1;
2678 		new->input = dst_discard;
2679 		new->output = dst_discard_out;
2680 
2681 		dst_copy_metrics(new, &ort->dst);
2682 
2683 		rt->rt6i_idev = in6_dev_get(loopback_dev);
2684 		rt->rt6i_gateway = ort->rt6i_gateway;
2685 		rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2686 
2687 		memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2688 #ifdef CONFIG_IPV6_SUBTREES
2689 		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2690 #endif
2691 	}
2692 
2693 	dst_release(dst_orig);
2694 	return new ? new : ERR_PTR(-ENOMEM);
2695 }
2696 
2697 /*
2698  *	Destination cache support functions
2699  */
2700 
fib6_check(struct fib6_info * f6i,u32 cookie)2701 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2702 {
2703 	u32 rt_cookie = 0;
2704 
2705 	if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2706 		return false;
2707 
2708 	if (fib6_check_expired(f6i))
2709 		return false;
2710 
2711 	return true;
2712 }
2713 
rt6_check(struct rt6_info * rt,struct fib6_info * from,u32 cookie)2714 static struct dst_entry *rt6_check(struct rt6_info *rt,
2715 				   struct fib6_info *from,
2716 				   u32 cookie)
2717 {
2718 	u32 rt_cookie = 0;
2719 
2720 	if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2721 	    rt_cookie != cookie)
2722 		return NULL;
2723 
2724 	if (rt6_check_expired(rt))
2725 		return NULL;
2726 
2727 	return &rt->dst;
2728 }
2729 
rt6_dst_from_check(struct rt6_info * rt,struct fib6_info * from,u32 cookie)2730 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2731 					    struct fib6_info *from,
2732 					    u32 cookie)
2733 {
2734 	if (!__rt6_check_expired(rt) &&
2735 	    rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2736 	    fib6_check(from, cookie))
2737 		return &rt->dst;
2738 	else
2739 		return NULL;
2740 }
2741 
ip6_dst_check(struct dst_entry * dst,u32 cookie)2742 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2743 							u32 cookie)
2744 {
2745 	struct dst_entry *dst_ret;
2746 	struct fib6_info *from;
2747 	struct rt6_info *rt;
2748 
2749 	rt = dst_rt6_info(dst);
2750 
2751 	if (rt->sernum)
2752 		return rt6_is_valid(rt) ? dst : NULL;
2753 
2754 	rcu_read_lock();
2755 
2756 	/* All IPV6 dsts are created with ->obsolete set to the value
2757 	 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2758 	 * into this function always.
2759 	 */
2760 
2761 	from = rcu_dereference(rt->from);
2762 
2763 	if (from && (rt->rt6i_flags & RTF_PCPU ||
2764 	    unlikely(!list_empty(&rt->dst.rt_uncached))))
2765 		dst_ret = rt6_dst_from_check(rt, from, cookie);
2766 	else
2767 		dst_ret = rt6_check(rt, from, cookie);
2768 
2769 	rcu_read_unlock();
2770 
2771 	return dst_ret;
2772 }
2773 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2774 
ip6_negative_advice(struct sock * sk,struct dst_entry * dst)2775 static void ip6_negative_advice(struct sock *sk,
2776 				struct dst_entry *dst)
2777 {
2778 	struct rt6_info *rt = dst_rt6_info(dst);
2779 
2780 	if (rt->rt6i_flags & RTF_CACHE) {
2781 		rcu_read_lock();
2782 		if (rt6_check_expired(rt)) {
2783 			/* rt/dst can not be destroyed yet,
2784 			 * because of rcu_read_lock()
2785 			 */
2786 			sk_dst_reset(sk);
2787 			rt6_remove_exception_rt(rt);
2788 		}
2789 		rcu_read_unlock();
2790 		return;
2791 	}
2792 	sk_dst_reset(sk);
2793 }
2794 
ip6_link_failure(struct sk_buff * skb)2795 static void ip6_link_failure(struct sk_buff *skb)
2796 {
2797 	struct rt6_info *rt;
2798 
2799 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2800 
2801 	rt = dst_rt6_info(skb_dst(skb));
2802 	if (rt) {
2803 		rcu_read_lock();
2804 		if (rt->rt6i_flags & RTF_CACHE) {
2805 			rt6_remove_exception_rt(rt);
2806 		} else {
2807 			struct fib6_info *from;
2808 			struct fib6_node *fn;
2809 
2810 			from = rcu_dereference(rt->from);
2811 			if (from) {
2812 				fn = rcu_dereference(from->fib6_node);
2813 				if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2814 					WRITE_ONCE(fn->fn_sernum, -1);
2815 			}
2816 		}
2817 		rcu_read_unlock();
2818 	}
2819 }
2820 
rt6_update_expires(struct rt6_info * rt0,int timeout)2821 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2822 {
2823 	if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2824 		struct fib6_info *from;
2825 
2826 		rcu_read_lock();
2827 		from = rcu_dereference(rt0->from);
2828 		if (from)
2829 			rt0->dst.expires = from->expires;
2830 		rcu_read_unlock();
2831 	}
2832 
2833 	dst_set_expires(&rt0->dst, timeout);
2834 	rt0->rt6i_flags |= RTF_EXPIRES;
2835 }
2836 
rt6_do_update_pmtu(struct rt6_info * rt,u32 mtu)2837 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2838 {
2839 	struct net *net = dev_net(rt->dst.dev);
2840 
2841 	dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2842 	rt->rt6i_flags |= RTF_MODIFIED;
2843 	rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2844 }
2845 
rt6_cache_allowed_for_pmtu(const struct rt6_info * rt)2846 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2847 {
2848 	return !(rt->rt6i_flags & RTF_CACHE) &&
2849 		(rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2850 }
2851 
__ip6_rt_update_pmtu(struct dst_entry * dst,const struct sock * sk,const struct ipv6hdr * iph,u32 mtu,bool confirm_neigh)2852 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2853 				 const struct ipv6hdr *iph, u32 mtu,
2854 				 bool confirm_neigh)
2855 {
2856 	const struct in6_addr *daddr, *saddr;
2857 	struct rt6_info *rt6 = dst_rt6_info(dst);
2858 
2859 	/* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2860 	 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2861 	 * [see also comment in rt6_mtu_change_route()]
2862 	 */
2863 
2864 	if (iph) {
2865 		daddr = &iph->daddr;
2866 		saddr = &iph->saddr;
2867 	} else if (sk) {
2868 		daddr = &sk->sk_v6_daddr;
2869 		saddr = &inet6_sk(sk)->saddr;
2870 	} else {
2871 		daddr = NULL;
2872 		saddr = NULL;
2873 	}
2874 
2875 	if (confirm_neigh)
2876 		dst_confirm_neigh(dst, daddr);
2877 
2878 	if (mtu < IPV6_MIN_MTU)
2879 		return;
2880 	if (mtu >= dst_mtu(dst))
2881 		return;
2882 
2883 	if (!rt6_cache_allowed_for_pmtu(rt6)) {
2884 		rt6_do_update_pmtu(rt6, mtu);
2885 		/* update rt6_ex->stamp for cache */
2886 		if (rt6->rt6i_flags & RTF_CACHE)
2887 			rt6_update_exception_stamp_rt(rt6);
2888 	} else if (daddr) {
2889 		struct fib6_result res = {};
2890 		struct rt6_info *nrt6;
2891 
2892 		rcu_read_lock();
2893 		res.f6i = rcu_dereference(rt6->from);
2894 		if (!res.f6i)
2895 			goto out_unlock;
2896 
2897 		res.fib6_flags = res.f6i->fib6_flags;
2898 		res.fib6_type = res.f6i->fib6_type;
2899 
2900 		if (res.f6i->nh) {
2901 			struct fib6_nh_match_arg arg = {
2902 				.dev = dst->dev,
2903 				.gw = &rt6->rt6i_gateway,
2904 			};
2905 
2906 			nexthop_for_each_fib6_nh(res.f6i->nh,
2907 						 fib6_nh_find_match, &arg);
2908 
2909 			/* fib6_info uses a nexthop that does not have fib6_nh
2910 			 * using the dst->dev + gw. Should be impossible.
2911 			 */
2912 			if (!arg.match)
2913 				goto out_unlock;
2914 
2915 			res.nh = arg.match;
2916 		} else {
2917 			res.nh = res.f6i->fib6_nh;
2918 		}
2919 
2920 		nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2921 		if (nrt6) {
2922 			rt6_do_update_pmtu(nrt6, mtu);
2923 			if (rt6_insert_exception(nrt6, &res))
2924 				dst_release_immediate(&nrt6->dst);
2925 		}
2926 out_unlock:
2927 		rcu_read_unlock();
2928 	}
2929 }
2930 
ip6_rt_update_pmtu(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb,u32 mtu,bool confirm_neigh)2931 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2932 			       struct sk_buff *skb, u32 mtu,
2933 			       bool confirm_neigh)
2934 {
2935 	__ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2936 			     confirm_neigh);
2937 }
2938 
ip6_update_pmtu(struct sk_buff * skb,struct net * net,__be32 mtu,int oif,u32 mark,kuid_t uid)2939 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2940 		     int oif, u32 mark, kuid_t uid)
2941 {
2942 	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2943 	struct dst_entry *dst;
2944 	struct flowi6 fl6 = {
2945 		.flowi6_oif = oif,
2946 		.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2947 		.daddr = iph->daddr,
2948 		.saddr = iph->saddr,
2949 		.flowlabel = ip6_flowinfo(iph),
2950 		.flowi6_uid = uid,
2951 	};
2952 
2953 	dst = ip6_route_output(net, NULL, &fl6);
2954 	if (!dst->error)
2955 		__ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2956 	dst_release(dst);
2957 }
2958 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2959 
ip6_sk_update_pmtu(struct sk_buff * skb,struct sock * sk,__be32 mtu)2960 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2961 {
2962 	int oif = sk->sk_bound_dev_if;
2963 	struct dst_entry *dst;
2964 
2965 	if (!oif && skb->dev)
2966 		oif = l3mdev_master_ifindex(skb->dev);
2967 
2968 	ip6_update_pmtu(skb, sock_net(sk), mtu, oif, READ_ONCE(sk->sk_mark),
2969 			sk->sk_uid);
2970 
2971 	dst = __sk_dst_get(sk);
2972 	if (!dst || !dst->obsolete ||
2973 	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2974 		return;
2975 
2976 	bh_lock_sock(sk);
2977 	if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2978 		ip6_datagram_dst_update(sk, false);
2979 	bh_unlock_sock(sk);
2980 }
2981 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2982 
ip6_sk_dst_store_flow(struct sock * sk,struct dst_entry * dst,const struct flowi6 * fl6)2983 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2984 			   const struct flowi6 *fl6)
2985 {
2986 #ifdef CONFIG_IPV6_SUBTREES
2987 	struct ipv6_pinfo *np = inet6_sk(sk);
2988 #endif
2989 
2990 	ip6_dst_store(sk, dst,
2991 		      ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2992 		      &sk->sk_v6_daddr : NULL,
2993 #ifdef CONFIG_IPV6_SUBTREES
2994 		      ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2995 		      &np->saddr :
2996 #endif
2997 		      NULL);
2998 }
2999 
ip6_redirect_nh_match(const struct fib6_result * res,struct flowi6 * fl6,const struct in6_addr * gw,struct rt6_info ** ret)3000 static bool ip6_redirect_nh_match(const struct fib6_result *res,
3001 				  struct flowi6 *fl6,
3002 				  const struct in6_addr *gw,
3003 				  struct rt6_info **ret)
3004 {
3005 	const struct fib6_nh *nh = res->nh;
3006 
3007 	if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
3008 	    fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
3009 		return false;
3010 
3011 	/* rt_cache's gateway might be different from its 'parent'
3012 	 * in the case of an ip redirect.
3013 	 * So we keep searching in the exception table if the gateway
3014 	 * is different.
3015 	 */
3016 	if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3017 		struct rt6_info *rt_cache;
3018 
3019 		rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3020 		if (rt_cache &&
3021 		    ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3022 			*ret = rt_cache;
3023 			return true;
3024 		}
3025 		return false;
3026 	}
3027 	return true;
3028 }
3029 
3030 struct fib6_nh_rd_arg {
3031 	struct fib6_result	*res;
3032 	struct flowi6		*fl6;
3033 	const struct in6_addr	*gw;
3034 	struct rt6_info		**ret;
3035 };
3036 
fib6_nh_redirect_match(struct fib6_nh * nh,void * _arg)3037 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3038 {
3039 	struct fib6_nh_rd_arg *arg = _arg;
3040 
3041 	arg->res->nh = nh;
3042 	return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3043 }
3044 
3045 /* Handle redirects */
3046 struct ip6rd_flowi {
3047 	struct flowi6 fl6;
3048 	struct in6_addr gateway;
3049 };
3050 
__ip6_route_redirect(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)3051 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3052 					     struct fib6_table *table,
3053 					     struct flowi6 *fl6,
3054 					     const struct sk_buff *skb,
3055 					     int flags)
3056 {
3057 	struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3058 	struct rt6_info *ret = NULL;
3059 	struct fib6_result res = {};
3060 	struct fib6_nh_rd_arg arg = {
3061 		.res = &res,
3062 		.fl6 = fl6,
3063 		.gw  = &rdfl->gateway,
3064 		.ret = &ret
3065 	};
3066 	struct fib6_info *rt;
3067 	struct fib6_node *fn;
3068 
3069 	/* Get the "current" route for this destination and
3070 	 * check if the redirect has come from appropriate router.
3071 	 *
3072 	 * RFC 4861 specifies that redirects should only be
3073 	 * accepted if they come from the nexthop to the target.
3074 	 * Due to the way the routes are chosen, this notion
3075 	 * is a bit fuzzy and one might need to check all possible
3076 	 * routes.
3077 	 */
3078 
3079 	rcu_read_lock();
3080 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3081 restart:
3082 	for_each_fib6_node_rt_rcu(fn) {
3083 		res.f6i = rt;
3084 		if (fib6_check_expired(rt))
3085 			continue;
3086 		if (rt->fib6_flags & RTF_REJECT)
3087 			break;
3088 		if (unlikely(rt->nh)) {
3089 			if (nexthop_is_blackhole(rt->nh))
3090 				continue;
3091 			/* on match, res->nh is filled in and potentially ret */
3092 			if (nexthop_for_each_fib6_nh(rt->nh,
3093 						     fib6_nh_redirect_match,
3094 						     &arg))
3095 				goto out;
3096 		} else {
3097 			res.nh = rt->fib6_nh;
3098 			if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3099 						  &ret))
3100 				goto out;
3101 		}
3102 	}
3103 
3104 	if (!rt)
3105 		rt = net->ipv6.fib6_null_entry;
3106 	else if (rt->fib6_flags & RTF_REJECT) {
3107 		ret = net->ipv6.ip6_null_entry;
3108 		goto out;
3109 	}
3110 
3111 	if (rt == net->ipv6.fib6_null_entry) {
3112 		fn = fib6_backtrack(fn, &fl6->saddr);
3113 		if (fn)
3114 			goto restart;
3115 	}
3116 
3117 	res.f6i = rt;
3118 	res.nh = rt->fib6_nh;
3119 out:
3120 	if (ret) {
3121 		ip6_hold_safe(net, &ret);
3122 	} else {
3123 		res.fib6_flags = res.f6i->fib6_flags;
3124 		res.fib6_type = res.f6i->fib6_type;
3125 		ret = ip6_create_rt_rcu(&res);
3126 	}
3127 
3128 	rcu_read_unlock();
3129 
3130 	trace_fib6_table_lookup(net, &res, table, fl6);
3131 	return ret;
3132 };
3133 
ip6_route_redirect(struct net * net,const struct flowi6 * fl6,const struct sk_buff * skb,const struct in6_addr * gateway)3134 static struct dst_entry *ip6_route_redirect(struct net *net,
3135 					    const struct flowi6 *fl6,
3136 					    const struct sk_buff *skb,
3137 					    const struct in6_addr *gateway)
3138 {
3139 	int flags = RT6_LOOKUP_F_HAS_SADDR;
3140 	struct ip6rd_flowi rdfl;
3141 
3142 	rdfl.fl6 = *fl6;
3143 	rdfl.gateway = *gateway;
3144 
3145 	return fib6_rule_lookup(net, &rdfl.fl6, skb,
3146 				flags, __ip6_route_redirect);
3147 }
3148 
ip6_redirect(struct sk_buff * skb,struct net * net,int oif,u32 mark,kuid_t uid)3149 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3150 		  kuid_t uid)
3151 {
3152 	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3153 	struct dst_entry *dst;
3154 	struct flowi6 fl6 = {
3155 		.flowi6_iif = LOOPBACK_IFINDEX,
3156 		.flowi6_oif = oif,
3157 		.flowi6_mark = mark,
3158 		.daddr = iph->daddr,
3159 		.saddr = iph->saddr,
3160 		.flowlabel = ip6_flowinfo(iph),
3161 		.flowi6_uid = uid,
3162 	};
3163 
3164 	dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3165 	rt6_do_redirect(dst, NULL, skb);
3166 	dst_release(dst);
3167 }
3168 EXPORT_SYMBOL_GPL(ip6_redirect);
3169 
ip6_redirect_no_header(struct sk_buff * skb,struct net * net,int oif)3170 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3171 {
3172 	const struct ipv6hdr *iph = ipv6_hdr(skb);
3173 	const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3174 	struct dst_entry *dst;
3175 	struct flowi6 fl6 = {
3176 		.flowi6_iif = LOOPBACK_IFINDEX,
3177 		.flowi6_oif = oif,
3178 		.daddr = msg->dest,
3179 		.saddr = iph->daddr,
3180 		.flowi6_uid = sock_net_uid(net, NULL),
3181 	};
3182 
3183 	dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3184 	rt6_do_redirect(dst, NULL, skb);
3185 	dst_release(dst);
3186 }
3187 
ip6_sk_redirect(struct sk_buff * skb,struct sock * sk)3188 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3189 {
3190 	ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if,
3191 		     READ_ONCE(sk->sk_mark), sk->sk_uid);
3192 }
3193 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3194 
ip6_default_advmss(const struct dst_entry * dst)3195 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3196 {
3197 	struct net_device *dev = dst->dev;
3198 	unsigned int mtu = dst_mtu(dst);
3199 	struct net *net;
3200 
3201 	mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3202 
3203 	rcu_read_lock();
3204 
3205 	net = dev_net_rcu(dev);
3206 	if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3207 		mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3208 
3209 	rcu_read_unlock();
3210 
3211 	/*
3212 	 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3213 	 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3214 	 * IPV6_MAXPLEN is also valid and means: "any MSS,
3215 	 * rely only on pmtu discovery"
3216 	 */
3217 	if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3218 		mtu = IPV6_MAXPLEN;
3219 	return mtu;
3220 }
3221 
ip6_mtu(const struct dst_entry * dst)3222 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3223 {
3224 	return ip6_dst_mtu_maybe_forward(dst, false);
3225 }
3226 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3227 
3228 /* MTU selection:
3229  * 1. mtu on route is locked - use it
3230  * 2. mtu from nexthop exception
3231  * 3. mtu from egress device
3232  *
3233  * based on ip6_dst_mtu_forward and exception logic of
3234  * rt6_find_cached_rt; called with rcu_read_lock
3235  */
ip6_mtu_from_fib6(const struct fib6_result * res,const struct in6_addr * daddr,const struct in6_addr * saddr)3236 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3237 		      const struct in6_addr *daddr,
3238 		      const struct in6_addr *saddr)
3239 {
3240 	const struct fib6_nh *nh = res->nh;
3241 	struct fib6_info *f6i = res->f6i;
3242 	struct inet6_dev *idev;
3243 	struct rt6_info *rt;
3244 	u32 mtu = 0;
3245 
3246 	if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3247 		mtu = f6i->fib6_pmtu;
3248 		if (mtu)
3249 			goto out;
3250 	}
3251 
3252 	rt = rt6_find_cached_rt(res, daddr, saddr);
3253 	if (unlikely(rt)) {
3254 		mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3255 	} else {
3256 		struct net_device *dev = nh->fib_nh_dev;
3257 
3258 		mtu = IPV6_MIN_MTU;
3259 		idev = __in6_dev_get(dev);
3260 		if (idev)
3261 			mtu = max_t(u32, mtu, READ_ONCE(idev->cnf.mtu6));
3262 	}
3263 
3264 	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3265 out:
3266 	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3267 }
3268 
icmp6_dst_alloc(struct net_device * dev,struct flowi6 * fl6)3269 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3270 				  struct flowi6 *fl6)
3271 {
3272 	struct dst_entry *dst;
3273 	struct rt6_info *rt;
3274 	struct inet6_dev *idev = in6_dev_get(dev);
3275 	struct net *net = dev_net(dev);
3276 
3277 	if (unlikely(!idev))
3278 		return ERR_PTR(-ENODEV);
3279 
3280 	rt = ip6_dst_alloc(net, dev, 0);
3281 	if (unlikely(!rt)) {
3282 		in6_dev_put(idev);
3283 		dst = ERR_PTR(-ENOMEM);
3284 		goto out;
3285 	}
3286 
3287 	rt->dst.input = ip6_input;
3288 	rt->dst.output  = ip6_output;
3289 	rt->rt6i_gateway  = fl6->daddr;
3290 	rt->rt6i_dst.addr = fl6->daddr;
3291 	rt->rt6i_dst.plen = 128;
3292 	rt->rt6i_idev     = idev;
3293 	dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3294 
3295 	/* Add this dst into uncached_list so that rt6_disable_ip() can
3296 	 * do proper release of the net_device
3297 	 */
3298 	rt6_uncached_list_add(rt);
3299 
3300 	dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3301 
3302 out:
3303 	return dst;
3304 }
3305 
ip6_dst_gc(struct dst_ops * ops)3306 static void ip6_dst_gc(struct dst_ops *ops)
3307 {
3308 	struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3309 	int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3310 	int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3311 	int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3312 	unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3313 	unsigned int val;
3314 	int entries;
3315 
3316 	if (time_after(rt_last_gc + rt_min_interval, jiffies))
3317 		goto out;
3318 
3319 	fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3320 	entries = dst_entries_get_slow(ops);
3321 	if (entries < ops->gc_thresh)
3322 		atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3323 out:
3324 	val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3325 	atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3326 }
3327 
ip6_nh_lookup_table(struct net * net,struct fib6_config * cfg,const struct in6_addr * gw_addr,u32 tbid,int flags,struct fib6_result * res)3328 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3329 			       const struct in6_addr *gw_addr, u32 tbid,
3330 			       int flags, struct fib6_result *res)
3331 {
3332 	struct flowi6 fl6 = {
3333 		.flowi6_oif = cfg->fc_ifindex,
3334 		.daddr = *gw_addr,
3335 		.saddr = cfg->fc_prefsrc,
3336 	};
3337 	struct fib6_table *table;
3338 	int err;
3339 
3340 	table = fib6_get_table(net, tbid);
3341 	if (!table)
3342 		return -EINVAL;
3343 
3344 	if (!ipv6_addr_any(&cfg->fc_prefsrc))
3345 		flags |= RT6_LOOKUP_F_HAS_SADDR;
3346 
3347 	flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3348 
3349 	err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3350 	if (!err && res->f6i != net->ipv6.fib6_null_entry)
3351 		fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3352 				 cfg->fc_ifindex != 0, NULL, flags);
3353 
3354 	return err;
3355 }
3356 
ip6_route_check_nh_onlink(struct net * net,struct fib6_config * cfg,const struct net_device * dev,struct netlink_ext_ack * extack)3357 static int ip6_route_check_nh_onlink(struct net *net,
3358 				     struct fib6_config *cfg,
3359 				     const struct net_device *dev,
3360 				     struct netlink_ext_ack *extack)
3361 {
3362 	u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3363 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3364 	struct fib6_result res = {};
3365 	int err;
3366 
3367 	err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3368 	if (!err && !(res.fib6_flags & RTF_REJECT) &&
3369 	    /* ignore match if it is the default route */
3370 	    !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3371 	    (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3372 		NL_SET_ERR_MSG(extack,
3373 			       "Nexthop has invalid gateway or device mismatch");
3374 		err = -EINVAL;
3375 	}
3376 
3377 	return err;
3378 }
3379 
ip6_route_check_nh(struct net * net,struct fib6_config * cfg,struct net_device ** _dev,netdevice_tracker * dev_tracker,struct inet6_dev ** idev)3380 static int ip6_route_check_nh(struct net *net,
3381 			      struct fib6_config *cfg,
3382 			      struct net_device **_dev,
3383 			      netdevice_tracker *dev_tracker,
3384 			      struct inet6_dev **idev)
3385 {
3386 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3387 	struct net_device *dev = _dev ? *_dev : NULL;
3388 	int flags = RT6_LOOKUP_F_IFACE;
3389 	struct fib6_result res = {};
3390 	int err = -EHOSTUNREACH;
3391 
3392 	if (cfg->fc_table) {
3393 		err = ip6_nh_lookup_table(net, cfg, gw_addr,
3394 					  cfg->fc_table, flags, &res);
3395 		/* gw_addr can not require a gateway or resolve to a reject
3396 		 * route. If a device is given, it must match the result.
3397 		 */
3398 		if (err || res.fib6_flags & RTF_REJECT ||
3399 		    res.nh->fib_nh_gw_family ||
3400 		    (dev && dev != res.nh->fib_nh_dev))
3401 			err = -EHOSTUNREACH;
3402 	}
3403 
3404 	if (err < 0) {
3405 		struct flowi6 fl6 = {
3406 			.flowi6_oif = cfg->fc_ifindex,
3407 			.daddr = *gw_addr,
3408 		};
3409 
3410 		err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3411 		if (err || res.fib6_flags & RTF_REJECT ||
3412 		    res.nh->fib_nh_gw_family)
3413 			err = -EHOSTUNREACH;
3414 
3415 		if (err)
3416 			return err;
3417 
3418 		fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3419 				 cfg->fc_ifindex != 0, NULL, flags);
3420 	}
3421 
3422 	err = 0;
3423 	if (dev) {
3424 		if (dev != res.nh->fib_nh_dev)
3425 			err = -EHOSTUNREACH;
3426 	} else {
3427 		*_dev = dev = res.nh->fib_nh_dev;
3428 		netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3429 		*idev = in6_dev_get(dev);
3430 	}
3431 
3432 	return err;
3433 }
3434 
ip6_validate_gw(struct net * net,struct fib6_config * cfg,struct net_device ** _dev,netdevice_tracker * dev_tracker,struct inet6_dev ** idev,struct netlink_ext_ack * extack)3435 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3436 			   struct net_device **_dev,
3437 			   netdevice_tracker *dev_tracker,
3438 			   struct inet6_dev **idev,
3439 			   struct netlink_ext_ack *extack)
3440 {
3441 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3442 	int gwa_type = ipv6_addr_type(gw_addr);
3443 	bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3444 	const struct net_device *dev = *_dev;
3445 	bool need_addr_check = !dev;
3446 	int err = -EINVAL;
3447 
3448 	/* if gw_addr is local we will fail to detect this in case
3449 	 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3450 	 * will return already-added prefix route via interface that
3451 	 * prefix route was assigned to, which might be non-loopback.
3452 	 */
3453 	if (dev &&
3454 	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3455 		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3456 		goto out;
3457 	}
3458 
3459 	if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3460 		/* IPv6 strictly inhibits using not link-local
3461 		 * addresses as nexthop address.
3462 		 * Otherwise, router will not able to send redirects.
3463 		 * It is very good, but in some (rare!) circumstances
3464 		 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3465 		 * some exceptions. --ANK
3466 		 * We allow IPv4-mapped nexthops to support RFC4798-type
3467 		 * addressing
3468 		 */
3469 		if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3470 			NL_SET_ERR_MSG(extack, "Invalid gateway address");
3471 			goto out;
3472 		}
3473 
3474 		rcu_read_lock();
3475 
3476 		if (cfg->fc_flags & RTNH_F_ONLINK)
3477 			err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3478 		else
3479 			err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3480 						 idev);
3481 
3482 		rcu_read_unlock();
3483 
3484 		if (err)
3485 			goto out;
3486 	}
3487 
3488 	/* reload in case device was changed */
3489 	dev = *_dev;
3490 
3491 	err = -EINVAL;
3492 	if (!dev) {
3493 		NL_SET_ERR_MSG(extack, "Egress device not specified");
3494 		goto out;
3495 	} else if (dev->flags & IFF_LOOPBACK) {
3496 		NL_SET_ERR_MSG(extack,
3497 			       "Egress device can not be loopback device for this route");
3498 		goto out;
3499 	}
3500 
3501 	/* if we did not check gw_addr above, do so now that the
3502 	 * egress device has been resolved.
3503 	 */
3504 	if (need_addr_check &&
3505 	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3506 		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3507 		goto out;
3508 	}
3509 
3510 	err = 0;
3511 out:
3512 	return err;
3513 }
3514 
fib6_is_reject(u32 flags,struct net_device * dev,int addr_type)3515 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3516 {
3517 	if ((flags & RTF_REJECT) ||
3518 	    (dev && (dev->flags & IFF_LOOPBACK) &&
3519 	     !(addr_type & IPV6_ADDR_LOOPBACK) &&
3520 	     !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3521 		return true;
3522 
3523 	return false;
3524 }
3525 
fib6_nh_init(struct net * net,struct fib6_nh * fib6_nh,struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3526 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3527 		 struct fib6_config *cfg, gfp_t gfp_flags,
3528 		 struct netlink_ext_ack *extack)
3529 {
3530 	netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
3531 	struct net_device *dev = NULL;
3532 	struct inet6_dev *idev = NULL;
3533 	int addr_type;
3534 	int err;
3535 
3536 	fib6_nh->fib_nh_family = AF_INET6;
3537 #ifdef CONFIG_IPV6_ROUTER_PREF
3538 	fib6_nh->last_probe = jiffies;
3539 #endif
3540 	if (cfg->fc_is_fdb) {
3541 		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3542 		fib6_nh->fib_nh_gw_family = AF_INET6;
3543 		return 0;
3544 	}
3545 
3546 	err = -ENODEV;
3547 	if (cfg->fc_ifindex) {
3548 		dev = netdev_get_by_index(net, cfg->fc_ifindex,
3549 					  dev_tracker, gfp_flags);
3550 		if (!dev)
3551 			goto out;
3552 		idev = in6_dev_get(dev);
3553 		if (!idev)
3554 			goto out;
3555 	}
3556 
3557 	if (cfg->fc_flags & RTNH_F_ONLINK) {
3558 		if (!dev) {
3559 			NL_SET_ERR_MSG(extack,
3560 				       "Nexthop device required for onlink");
3561 			goto out;
3562 		}
3563 
3564 		if (!(dev->flags & IFF_UP)) {
3565 			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3566 			err = -ENETDOWN;
3567 			goto out;
3568 		}
3569 
3570 		fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3571 	}
3572 
3573 	fib6_nh->fib_nh_weight = 1;
3574 
3575 	/* We cannot add true routes via loopback here,
3576 	 * they would result in kernel looping; promote them to reject routes
3577 	 */
3578 	addr_type = ipv6_addr_type(&cfg->fc_dst);
3579 	if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3580 		/* hold loopback dev/idev if we haven't done so. */
3581 		if (dev != net->loopback_dev) {
3582 			if (dev) {
3583 				netdev_put(dev, dev_tracker);
3584 				in6_dev_put(idev);
3585 			}
3586 			dev = net->loopback_dev;
3587 			netdev_hold(dev, dev_tracker, gfp_flags);
3588 			idev = in6_dev_get(dev);
3589 			if (!idev) {
3590 				err = -ENODEV;
3591 				goto out;
3592 			}
3593 		}
3594 		goto pcpu_alloc;
3595 	}
3596 
3597 	if (cfg->fc_flags & RTF_GATEWAY) {
3598 		err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
3599 				      &idev, extack);
3600 		if (err)
3601 			goto out;
3602 
3603 		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3604 		fib6_nh->fib_nh_gw_family = AF_INET6;
3605 	}
3606 
3607 	err = -ENODEV;
3608 	if (!dev)
3609 		goto out;
3610 
3611 	if (!idev || idev->cnf.disable_ipv6) {
3612 		NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3613 		err = -EACCES;
3614 		goto out;
3615 	}
3616 
3617 	if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3618 		NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3619 		err = -ENETDOWN;
3620 		goto out;
3621 	}
3622 
3623 	if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3624 	    !netif_carrier_ok(dev))
3625 		fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3626 
3627 	err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3628 				 cfg->fc_encap_type, cfg, gfp_flags, extack);
3629 	if (err)
3630 		goto out;
3631 
3632 pcpu_alloc:
3633 	fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3634 	if (!fib6_nh->rt6i_pcpu) {
3635 		err = -ENOMEM;
3636 		goto out;
3637 	}
3638 
3639 	fib6_nh->fib_nh_dev = dev;
3640 	fib6_nh->fib_nh_oif = dev->ifindex;
3641 	err = 0;
3642 out:
3643 	if (idev)
3644 		in6_dev_put(idev);
3645 
3646 	if (err) {
3647 		fib_nh_common_release(&fib6_nh->nh_common);
3648 		fib6_nh->nh_common.nhc_pcpu_rth_output = NULL;
3649 		fib6_nh->fib_nh_lws = NULL;
3650 		netdev_put(dev, dev_tracker);
3651 	}
3652 
3653 	return err;
3654 }
3655 
fib6_nh_release(struct fib6_nh * fib6_nh)3656 void fib6_nh_release(struct fib6_nh *fib6_nh)
3657 {
3658 	struct rt6_exception_bucket *bucket;
3659 
3660 	rcu_read_lock();
3661 
3662 	fib6_nh_flush_exceptions(fib6_nh, NULL);
3663 	bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3664 	if (bucket) {
3665 		rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3666 		kfree(bucket);
3667 	}
3668 
3669 	rcu_read_unlock();
3670 
3671 	fib6_nh_release_dsts(fib6_nh);
3672 	free_percpu(fib6_nh->rt6i_pcpu);
3673 
3674 	fib_nh_common_release(&fib6_nh->nh_common);
3675 }
3676 
fib6_nh_release_dsts(struct fib6_nh * fib6_nh)3677 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3678 {
3679 	int cpu;
3680 
3681 	if (!fib6_nh->rt6i_pcpu)
3682 		return;
3683 
3684 	for_each_possible_cpu(cpu) {
3685 		struct rt6_info *pcpu_rt, **ppcpu_rt;
3686 
3687 		ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3688 		pcpu_rt = xchg(ppcpu_rt, NULL);
3689 		if (pcpu_rt) {
3690 			dst_dev_put(&pcpu_rt->dst);
3691 			dst_release(&pcpu_rt->dst);
3692 		}
3693 	}
3694 }
3695 
ip6_route_info_create(struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3696 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3697 					      gfp_t gfp_flags,
3698 					      struct netlink_ext_ack *extack)
3699 {
3700 	struct net *net = cfg->fc_nlinfo.nl_net;
3701 	struct fib6_info *rt = NULL;
3702 	struct nexthop *nh = NULL;
3703 	struct fib6_table *table;
3704 	struct fib6_nh *fib6_nh;
3705 	int err = -EINVAL;
3706 	int addr_type;
3707 
3708 	/* RTF_PCPU is an internal flag; can not be set by userspace */
3709 	if (cfg->fc_flags & RTF_PCPU) {
3710 		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3711 		goto out;
3712 	}
3713 
3714 	/* RTF_CACHE is an internal flag; can not be set by userspace */
3715 	if (cfg->fc_flags & RTF_CACHE) {
3716 		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3717 		goto out;
3718 	}
3719 
3720 	if (cfg->fc_type > RTN_MAX) {
3721 		NL_SET_ERR_MSG(extack, "Invalid route type");
3722 		goto out;
3723 	}
3724 
3725 	if (cfg->fc_dst_len > 128) {
3726 		NL_SET_ERR_MSG(extack, "Invalid prefix length");
3727 		goto out;
3728 	}
3729 	if (cfg->fc_src_len > 128) {
3730 		NL_SET_ERR_MSG(extack, "Invalid source address length");
3731 		goto out;
3732 	}
3733 #ifndef CONFIG_IPV6_SUBTREES
3734 	if (cfg->fc_src_len) {
3735 		NL_SET_ERR_MSG(extack,
3736 			       "Specifying source address requires IPV6_SUBTREES to be enabled");
3737 		goto out;
3738 	}
3739 #endif
3740 	if (cfg->fc_nh_id) {
3741 		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3742 		if (!nh) {
3743 			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3744 			goto out;
3745 		}
3746 		err = fib6_check_nexthop(nh, cfg, extack);
3747 		if (err)
3748 			goto out;
3749 	}
3750 
3751 	err = -ENOBUFS;
3752 	if (cfg->fc_nlinfo.nlh &&
3753 	    !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3754 		table = fib6_get_table(net, cfg->fc_table);
3755 		if (!table) {
3756 			pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3757 			table = fib6_new_table(net, cfg->fc_table);
3758 		}
3759 	} else {
3760 		table = fib6_new_table(net, cfg->fc_table);
3761 	}
3762 
3763 	if (!table)
3764 		goto out;
3765 
3766 	err = -ENOMEM;
3767 	rt = fib6_info_alloc(gfp_flags, !nh);
3768 	if (!rt)
3769 		goto out;
3770 
3771 	rt->fib6_metrics = ip_fib_metrics_init(cfg->fc_mx, cfg->fc_mx_len,
3772 					       extack);
3773 	if (IS_ERR(rt->fib6_metrics)) {
3774 		err = PTR_ERR(rt->fib6_metrics);
3775 		/* Do not leave garbage there. */
3776 		rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3777 		goto out_free;
3778 	}
3779 
3780 	if (cfg->fc_flags & RTF_ADDRCONF)
3781 		rt->dst_nocount = true;
3782 
3783 	if (cfg->fc_flags & RTF_EXPIRES)
3784 		fib6_set_expires(rt, jiffies +
3785 				clock_t_to_jiffies(cfg->fc_expires));
3786 
3787 	if (cfg->fc_protocol == RTPROT_UNSPEC)
3788 		cfg->fc_protocol = RTPROT_BOOT;
3789 	rt->fib6_protocol = cfg->fc_protocol;
3790 
3791 	rt->fib6_table = table;
3792 	rt->fib6_metric = cfg->fc_metric;
3793 	rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3794 	rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3795 
3796 	ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3797 	rt->fib6_dst.plen = cfg->fc_dst_len;
3798 
3799 #ifdef CONFIG_IPV6_SUBTREES
3800 	ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3801 	rt->fib6_src.plen = cfg->fc_src_len;
3802 #endif
3803 	if (nh) {
3804 		if (rt->fib6_src.plen) {
3805 			NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3806 			err = -EINVAL;
3807 			goto out_free;
3808 		}
3809 		if (!nexthop_get(nh)) {
3810 			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3811 			err = -ENOENT;
3812 			goto out_free;
3813 		}
3814 		rt->nh = nh;
3815 		fib6_nh = nexthop_fib6_nh(rt->nh);
3816 	} else {
3817 		err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3818 		if (err)
3819 			goto out;
3820 
3821 		fib6_nh = rt->fib6_nh;
3822 
3823 		/* We cannot add true routes via loopback here, they would
3824 		 * result in kernel looping; promote them to reject routes
3825 		 */
3826 		addr_type = ipv6_addr_type(&cfg->fc_dst);
3827 		if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3828 				   addr_type))
3829 			rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3830 	}
3831 
3832 	if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3833 		struct net_device *dev = fib6_nh->fib_nh_dev;
3834 
3835 		if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3836 			NL_SET_ERR_MSG(extack, "Invalid source address");
3837 			err = -EINVAL;
3838 			goto out;
3839 		}
3840 		rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3841 		rt->fib6_prefsrc.plen = 128;
3842 	} else
3843 		rt->fib6_prefsrc.plen = 0;
3844 
3845 	return rt;
3846 out:
3847 	fib6_info_release(rt);
3848 	return ERR_PTR(err);
3849 out_free:
3850 	ip_fib_metrics_put(rt->fib6_metrics);
3851 	kfree(rt);
3852 	return ERR_PTR(err);
3853 }
3854 
ip6_route_add(struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3855 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3856 		  struct netlink_ext_ack *extack)
3857 {
3858 	struct fib6_info *rt;
3859 	int err;
3860 
3861 	rt = ip6_route_info_create(cfg, gfp_flags, extack);
3862 	if (IS_ERR(rt))
3863 		return PTR_ERR(rt);
3864 
3865 	err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3866 	fib6_info_release(rt);
3867 
3868 	return err;
3869 }
3870 
__ip6_del_rt(struct fib6_info * rt,struct nl_info * info)3871 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3872 {
3873 	struct net *net = info->nl_net;
3874 	struct fib6_table *table;
3875 	int err;
3876 
3877 	if (rt == net->ipv6.fib6_null_entry) {
3878 		err = -ENOENT;
3879 		goto out;
3880 	}
3881 
3882 	table = rt->fib6_table;
3883 	spin_lock_bh(&table->tb6_lock);
3884 	err = fib6_del(rt, info);
3885 	spin_unlock_bh(&table->tb6_lock);
3886 
3887 out:
3888 	fib6_info_release(rt);
3889 	return err;
3890 }
3891 
ip6_del_rt(struct net * net,struct fib6_info * rt,bool skip_notify)3892 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3893 {
3894 	struct nl_info info = {
3895 		.nl_net = net,
3896 		.skip_notify = skip_notify
3897 	};
3898 
3899 	return __ip6_del_rt(rt, &info);
3900 }
3901 
__ip6_del_rt_siblings(struct fib6_info * rt,struct fib6_config * cfg)3902 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3903 {
3904 	struct nl_info *info = &cfg->fc_nlinfo;
3905 	struct net *net = info->nl_net;
3906 	struct sk_buff *skb = NULL;
3907 	struct fib6_table *table;
3908 	int err = -ENOENT;
3909 
3910 	if (rt == net->ipv6.fib6_null_entry)
3911 		goto out_put;
3912 	table = rt->fib6_table;
3913 	spin_lock_bh(&table->tb6_lock);
3914 
3915 	if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3916 		struct fib6_info *sibling, *next_sibling;
3917 		struct fib6_node *fn;
3918 
3919 		/* prefer to send a single notification with all hops */
3920 		skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3921 		if (skb) {
3922 			u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3923 
3924 			if (rt6_fill_node(net, skb, rt, NULL,
3925 					  NULL, NULL, 0, RTM_DELROUTE,
3926 					  info->portid, seq, 0) < 0) {
3927 				kfree_skb(skb);
3928 				skb = NULL;
3929 			} else
3930 				info->skip_notify = 1;
3931 		}
3932 
3933 		/* 'rt' points to the first sibling route. If it is not the
3934 		 * leaf, then we do not need to send a notification. Otherwise,
3935 		 * we need to check if the last sibling has a next route or not
3936 		 * and emit a replace or delete notification, respectively.
3937 		 */
3938 		info->skip_notify_kernel = 1;
3939 		fn = rcu_dereference_protected(rt->fib6_node,
3940 					    lockdep_is_held(&table->tb6_lock));
3941 		if (rcu_access_pointer(fn->leaf) == rt) {
3942 			struct fib6_info *last_sibling, *replace_rt;
3943 
3944 			last_sibling = list_last_entry(&rt->fib6_siblings,
3945 						       struct fib6_info,
3946 						       fib6_siblings);
3947 			replace_rt = rcu_dereference_protected(
3948 					    last_sibling->fib6_next,
3949 					    lockdep_is_held(&table->tb6_lock));
3950 			if (replace_rt)
3951 				call_fib6_entry_notifiers_replace(net,
3952 								  replace_rt);
3953 			else
3954 				call_fib6_multipath_entry_notifiers(net,
3955 						       FIB_EVENT_ENTRY_DEL,
3956 						       rt, rt->fib6_nsiblings,
3957 						       NULL);
3958 		}
3959 		list_for_each_entry_safe(sibling, next_sibling,
3960 					 &rt->fib6_siblings,
3961 					 fib6_siblings) {
3962 			err = fib6_del(sibling, info);
3963 			if (err)
3964 				goto out_unlock;
3965 		}
3966 	}
3967 
3968 	err = fib6_del(rt, info);
3969 out_unlock:
3970 	spin_unlock_bh(&table->tb6_lock);
3971 out_put:
3972 	fib6_info_release(rt);
3973 
3974 	if (skb) {
3975 		rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3976 			    info->nlh, gfp_any());
3977 	}
3978 	return err;
3979 }
3980 
__ip6_del_cached_rt(struct rt6_info * rt,struct fib6_config * cfg)3981 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3982 {
3983 	int rc = -ESRCH;
3984 
3985 	if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3986 		goto out;
3987 
3988 	if (cfg->fc_flags & RTF_GATEWAY &&
3989 	    !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3990 		goto out;
3991 
3992 	rc = rt6_remove_exception_rt(rt);
3993 out:
3994 	return rc;
3995 }
3996 
ip6_del_cached_rt(struct fib6_config * cfg,struct fib6_info * rt,struct fib6_nh * nh)3997 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3998 			     struct fib6_nh *nh)
3999 {
4000 	struct fib6_result res = {
4001 		.f6i = rt,
4002 		.nh = nh,
4003 	};
4004 	struct rt6_info *rt_cache;
4005 
4006 	rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
4007 	if (rt_cache)
4008 		return __ip6_del_cached_rt(rt_cache, cfg);
4009 
4010 	return 0;
4011 }
4012 
4013 struct fib6_nh_del_cached_rt_arg {
4014 	struct fib6_config *cfg;
4015 	struct fib6_info *f6i;
4016 };
4017 
fib6_nh_del_cached_rt(struct fib6_nh * nh,void * _arg)4018 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4019 {
4020 	struct fib6_nh_del_cached_rt_arg *arg = _arg;
4021 	int rc;
4022 
4023 	rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4024 	return rc != -ESRCH ? rc : 0;
4025 }
4026 
ip6_del_cached_rt_nh(struct fib6_config * cfg,struct fib6_info * f6i)4027 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4028 {
4029 	struct fib6_nh_del_cached_rt_arg arg = {
4030 		.cfg = cfg,
4031 		.f6i = f6i
4032 	};
4033 
4034 	return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4035 }
4036 
ip6_route_del(struct fib6_config * cfg,struct netlink_ext_ack * extack)4037 static int ip6_route_del(struct fib6_config *cfg,
4038 			 struct netlink_ext_ack *extack)
4039 {
4040 	struct fib6_table *table;
4041 	struct fib6_info *rt;
4042 	struct fib6_node *fn;
4043 	int err = -ESRCH;
4044 
4045 	table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4046 	if (!table) {
4047 		NL_SET_ERR_MSG(extack, "FIB table does not exist");
4048 		return err;
4049 	}
4050 
4051 	rcu_read_lock();
4052 
4053 	fn = fib6_locate(&table->tb6_root,
4054 			 &cfg->fc_dst, cfg->fc_dst_len,
4055 			 &cfg->fc_src, cfg->fc_src_len,
4056 			 !(cfg->fc_flags & RTF_CACHE));
4057 
4058 	if (fn) {
4059 		for_each_fib6_node_rt_rcu(fn) {
4060 			struct fib6_nh *nh;
4061 
4062 			if (rt->nh && cfg->fc_nh_id &&
4063 			    rt->nh->id != cfg->fc_nh_id)
4064 				continue;
4065 
4066 			if (cfg->fc_flags & RTF_CACHE) {
4067 				int rc = 0;
4068 
4069 				if (rt->nh) {
4070 					rc = ip6_del_cached_rt_nh(cfg, rt);
4071 				} else if (cfg->fc_nh_id) {
4072 					continue;
4073 				} else {
4074 					nh = rt->fib6_nh;
4075 					rc = ip6_del_cached_rt(cfg, rt, nh);
4076 				}
4077 				if (rc != -ESRCH) {
4078 					rcu_read_unlock();
4079 					return rc;
4080 				}
4081 				continue;
4082 			}
4083 
4084 			if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4085 				continue;
4086 			if (cfg->fc_protocol &&
4087 			    cfg->fc_protocol != rt->fib6_protocol)
4088 				continue;
4089 
4090 			if (rt->nh) {
4091 				if (!fib6_info_hold_safe(rt))
4092 					continue;
4093 				rcu_read_unlock();
4094 
4095 				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4096 			}
4097 			if (cfg->fc_nh_id)
4098 				continue;
4099 
4100 			nh = rt->fib6_nh;
4101 			if (cfg->fc_ifindex &&
4102 			    (!nh->fib_nh_dev ||
4103 			     nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4104 				continue;
4105 			if (cfg->fc_flags & RTF_GATEWAY &&
4106 			    !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4107 				continue;
4108 			if (!fib6_info_hold_safe(rt))
4109 				continue;
4110 			rcu_read_unlock();
4111 
4112 			/* if gateway was specified only delete the one hop */
4113 			if (cfg->fc_flags & RTF_GATEWAY)
4114 				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4115 
4116 			return __ip6_del_rt_siblings(rt, cfg);
4117 		}
4118 	}
4119 	rcu_read_unlock();
4120 
4121 	return err;
4122 }
4123 
rt6_do_redirect(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb)4124 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4125 {
4126 	struct netevent_redirect netevent;
4127 	struct rt6_info *rt, *nrt = NULL;
4128 	struct fib6_result res = {};
4129 	struct ndisc_options ndopts;
4130 	struct inet6_dev *in6_dev;
4131 	struct neighbour *neigh;
4132 	struct rd_msg *msg;
4133 	int optlen, on_link;
4134 	u8 *lladdr;
4135 
4136 	optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4137 	optlen -= sizeof(*msg);
4138 
4139 	if (optlen < 0) {
4140 		net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4141 		return;
4142 	}
4143 
4144 	msg = (struct rd_msg *)icmp6_hdr(skb);
4145 
4146 	if (ipv6_addr_is_multicast(&msg->dest)) {
4147 		net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4148 		return;
4149 	}
4150 
4151 	on_link = 0;
4152 	if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4153 		on_link = 1;
4154 	} else if (ipv6_addr_type(&msg->target) !=
4155 		   (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4156 		net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4157 		return;
4158 	}
4159 
4160 	in6_dev = __in6_dev_get(skb->dev);
4161 	if (!in6_dev)
4162 		return;
4163 	if (READ_ONCE(in6_dev->cnf.forwarding) ||
4164 	    !READ_ONCE(in6_dev->cnf.accept_redirects))
4165 		return;
4166 
4167 	/* RFC2461 8.1:
4168 	 *	The IP source address of the Redirect MUST be the same as the current
4169 	 *	first-hop router for the specified ICMP Destination Address.
4170 	 */
4171 
4172 	if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4173 		net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4174 		return;
4175 	}
4176 
4177 	lladdr = NULL;
4178 	if (ndopts.nd_opts_tgt_lladdr) {
4179 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4180 					     skb->dev);
4181 		if (!lladdr) {
4182 			net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4183 			return;
4184 		}
4185 	}
4186 
4187 	rt = dst_rt6_info(dst);
4188 	if (rt->rt6i_flags & RTF_REJECT) {
4189 		net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4190 		return;
4191 	}
4192 
4193 	/* Redirect received -> path was valid.
4194 	 * Look, redirects are sent only in response to data packets,
4195 	 * so that this nexthop apparently is reachable. --ANK
4196 	 */
4197 	dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4198 
4199 	neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4200 	if (!neigh)
4201 		return;
4202 
4203 	/*
4204 	 *	We have finally decided to accept it.
4205 	 */
4206 
4207 	ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4208 		     NEIGH_UPDATE_F_WEAK_OVERRIDE|
4209 		     NEIGH_UPDATE_F_OVERRIDE|
4210 		     (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4211 				     NEIGH_UPDATE_F_ISROUTER)),
4212 		     NDISC_REDIRECT, &ndopts);
4213 
4214 	rcu_read_lock();
4215 	res.f6i = rcu_dereference(rt->from);
4216 	if (!res.f6i)
4217 		goto out;
4218 
4219 	if (res.f6i->nh) {
4220 		struct fib6_nh_match_arg arg = {
4221 			.dev = dst->dev,
4222 			.gw = &rt->rt6i_gateway,
4223 		};
4224 
4225 		nexthop_for_each_fib6_nh(res.f6i->nh,
4226 					 fib6_nh_find_match, &arg);
4227 
4228 		/* fib6_info uses a nexthop that does not have fib6_nh
4229 		 * using the dst->dev. Should be impossible
4230 		 */
4231 		if (!arg.match)
4232 			goto out;
4233 		res.nh = arg.match;
4234 	} else {
4235 		res.nh = res.f6i->fib6_nh;
4236 	}
4237 
4238 	res.fib6_flags = res.f6i->fib6_flags;
4239 	res.fib6_type = res.f6i->fib6_type;
4240 	nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4241 	if (!nrt)
4242 		goto out;
4243 
4244 	nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4245 	if (on_link)
4246 		nrt->rt6i_flags &= ~RTF_GATEWAY;
4247 
4248 	nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4249 
4250 	/* rt6_insert_exception() will take care of duplicated exceptions */
4251 	if (rt6_insert_exception(nrt, &res)) {
4252 		dst_release_immediate(&nrt->dst);
4253 		goto out;
4254 	}
4255 
4256 	netevent.old = &rt->dst;
4257 	netevent.new = &nrt->dst;
4258 	netevent.daddr = &msg->dest;
4259 	netevent.neigh = neigh;
4260 	call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4261 
4262 out:
4263 	rcu_read_unlock();
4264 	neigh_release(neigh);
4265 }
4266 
4267 #ifdef CONFIG_IPV6_ROUTE_INFO
rt6_get_route_info(struct net * net,const struct in6_addr * prefix,int prefixlen,const struct in6_addr * gwaddr,struct net_device * dev)4268 static struct fib6_info *rt6_get_route_info(struct net *net,
4269 					   const struct in6_addr *prefix, int prefixlen,
4270 					   const struct in6_addr *gwaddr,
4271 					   struct net_device *dev)
4272 {
4273 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4274 	int ifindex = dev->ifindex;
4275 	struct fib6_node *fn;
4276 	struct fib6_info *rt = NULL;
4277 	struct fib6_table *table;
4278 
4279 	table = fib6_get_table(net, tb_id);
4280 	if (!table)
4281 		return NULL;
4282 
4283 	rcu_read_lock();
4284 	fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4285 	if (!fn)
4286 		goto out;
4287 
4288 	for_each_fib6_node_rt_rcu(fn) {
4289 		/* these routes do not use nexthops */
4290 		if (rt->nh)
4291 			continue;
4292 		if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4293 			continue;
4294 		if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4295 		    !rt->fib6_nh->fib_nh_gw_family)
4296 			continue;
4297 		if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4298 			continue;
4299 		if (!fib6_info_hold_safe(rt))
4300 			continue;
4301 		break;
4302 	}
4303 out:
4304 	rcu_read_unlock();
4305 	return rt;
4306 }
4307 
rt6_add_route_info(struct net * net,const struct in6_addr * prefix,int prefixlen,const struct in6_addr * gwaddr,struct net_device * dev,unsigned int pref)4308 static struct fib6_info *rt6_add_route_info(struct net *net,
4309 					   const struct in6_addr *prefix, int prefixlen,
4310 					   const struct in6_addr *gwaddr,
4311 					   struct net_device *dev,
4312 					   unsigned int pref)
4313 {
4314 	struct fib6_config cfg = {
4315 		.fc_metric	= IP6_RT_PRIO_USER,
4316 		.fc_ifindex	= dev->ifindex,
4317 		.fc_dst_len	= prefixlen,
4318 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4319 				  RTF_UP | RTF_PREF(pref),
4320 		.fc_protocol = RTPROT_RA,
4321 		.fc_type = RTN_UNICAST,
4322 		.fc_nlinfo.portid = 0,
4323 		.fc_nlinfo.nlh = NULL,
4324 		.fc_nlinfo.nl_net = net,
4325 	};
4326 
4327 	cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4328 	cfg.fc_dst = *prefix;
4329 	cfg.fc_gateway = *gwaddr;
4330 
4331 	/* We should treat it as a default route if prefix length is 0. */
4332 	if (!prefixlen)
4333 		cfg.fc_flags |= RTF_DEFAULT;
4334 
4335 	ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4336 
4337 	return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4338 }
4339 #endif
4340 
rt6_get_dflt_router(struct net * net,const struct in6_addr * addr,struct net_device * dev)4341 struct fib6_info *rt6_get_dflt_router(struct net *net,
4342 				     const struct in6_addr *addr,
4343 				     struct net_device *dev)
4344 {
4345 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4346 	struct fib6_info *rt;
4347 	struct fib6_table *table;
4348 
4349 	table = fib6_get_table(net, tb_id);
4350 	if (!table)
4351 		return NULL;
4352 
4353 	rcu_read_lock();
4354 	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4355 		struct fib6_nh *nh;
4356 
4357 		/* RA routes do not use nexthops */
4358 		if (rt->nh)
4359 			continue;
4360 
4361 		nh = rt->fib6_nh;
4362 		if (dev == nh->fib_nh_dev &&
4363 		    ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4364 		    ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4365 			break;
4366 	}
4367 	if (rt && !fib6_info_hold_safe(rt))
4368 		rt = NULL;
4369 	rcu_read_unlock();
4370 	return rt;
4371 }
4372 
rt6_add_dflt_router(struct net * net,const struct in6_addr * gwaddr,struct net_device * dev,unsigned int pref,u32 defrtr_usr_metric,int lifetime)4373 struct fib6_info *rt6_add_dflt_router(struct net *net,
4374 				     const struct in6_addr *gwaddr,
4375 				     struct net_device *dev,
4376 				     unsigned int pref,
4377 				     u32 defrtr_usr_metric,
4378 				     int lifetime)
4379 {
4380 	struct fib6_config cfg = {
4381 		.fc_table	= l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4382 		.fc_metric	= defrtr_usr_metric,
4383 		.fc_ifindex	= dev->ifindex,
4384 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4385 				  RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4386 		.fc_protocol = RTPROT_RA,
4387 		.fc_type = RTN_UNICAST,
4388 		.fc_nlinfo.portid = 0,
4389 		.fc_nlinfo.nlh = NULL,
4390 		.fc_nlinfo.nl_net = net,
4391 		.fc_expires = jiffies_to_clock_t(lifetime * HZ),
4392 	};
4393 
4394 	cfg.fc_gateway = *gwaddr;
4395 
4396 	if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4397 		struct fib6_table *table;
4398 
4399 		table = fib6_get_table(dev_net(dev), cfg.fc_table);
4400 		if (table)
4401 			table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4402 	}
4403 
4404 	return rt6_get_dflt_router(net, gwaddr, dev);
4405 }
4406 
__rt6_purge_dflt_routers(struct net * net,struct fib6_table * table)4407 static void __rt6_purge_dflt_routers(struct net *net,
4408 				     struct fib6_table *table)
4409 {
4410 	struct fib6_info *rt;
4411 
4412 restart:
4413 	rcu_read_lock();
4414 	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4415 		struct net_device *dev = fib6_info_nh_dev(rt);
4416 		struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4417 
4418 		if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4419 		    (!idev || idev->cnf.accept_ra != 2) &&
4420 		    fib6_info_hold_safe(rt)) {
4421 			rcu_read_unlock();
4422 			ip6_del_rt(net, rt, false);
4423 			goto restart;
4424 		}
4425 	}
4426 	rcu_read_unlock();
4427 
4428 	table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4429 }
4430 
rt6_purge_dflt_routers(struct net * net)4431 void rt6_purge_dflt_routers(struct net *net)
4432 {
4433 	struct fib6_table *table;
4434 	struct hlist_head *head;
4435 	unsigned int h;
4436 
4437 	rcu_read_lock();
4438 
4439 	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4440 		head = &net->ipv6.fib_table_hash[h];
4441 		hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4442 			if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4443 				__rt6_purge_dflt_routers(net, table);
4444 		}
4445 	}
4446 
4447 	rcu_read_unlock();
4448 }
4449 
rtmsg_to_fib6_config(struct net * net,struct in6_rtmsg * rtmsg,struct fib6_config * cfg)4450 static void rtmsg_to_fib6_config(struct net *net,
4451 				 struct in6_rtmsg *rtmsg,
4452 				 struct fib6_config *cfg)
4453 {
4454 	*cfg = (struct fib6_config){
4455 		.fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4456 			 : RT6_TABLE_MAIN,
4457 		.fc_ifindex = rtmsg->rtmsg_ifindex,
4458 		.fc_metric = rtmsg->rtmsg_metric,
4459 		.fc_expires = rtmsg->rtmsg_info,
4460 		.fc_dst_len = rtmsg->rtmsg_dst_len,
4461 		.fc_src_len = rtmsg->rtmsg_src_len,
4462 		.fc_flags = rtmsg->rtmsg_flags,
4463 		.fc_type = rtmsg->rtmsg_type,
4464 
4465 		.fc_nlinfo.nl_net = net,
4466 
4467 		.fc_dst = rtmsg->rtmsg_dst,
4468 		.fc_src = rtmsg->rtmsg_src,
4469 		.fc_gateway = rtmsg->rtmsg_gateway,
4470 	};
4471 }
4472 
ipv6_route_ioctl(struct net * net,unsigned int cmd,struct in6_rtmsg * rtmsg)4473 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4474 {
4475 	struct fib6_config cfg;
4476 	int err;
4477 
4478 	if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4479 		return -EINVAL;
4480 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4481 		return -EPERM;
4482 
4483 	rtmsg_to_fib6_config(net, rtmsg, &cfg);
4484 
4485 	rtnl_lock();
4486 	switch (cmd) {
4487 	case SIOCADDRT:
4488 		/* Only do the default setting of fc_metric in route adding */
4489 		if (cfg.fc_metric == 0)
4490 			cfg.fc_metric = IP6_RT_PRIO_USER;
4491 		err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4492 		break;
4493 	case SIOCDELRT:
4494 		err = ip6_route_del(&cfg, NULL);
4495 		break;
4496 	}
4497 	rtnl_unlock();
4498 	return err;
4499 }
4500 
4501 /*
4502  *	Drop the packet on the floor
4503  */
4504 
ip6_pkt_drop(struct sk_buff * skb,u8 code,int ipstats_mib_noroutes)4505 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4506 {
4507 	struct dst_entry *dst = skb_dst(skb);
4508 	struct net *net = dev_net(dst->dev);
4509 	struct inet6_dev *idev;
4510 	SKB_DR(reason);
4511 	int type;
4512 
4513 	if (netif_is_l3_master(skb->dev) ||
4514 	    dst->dev == net->loopback_dev)
4515 		idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4516 	else
4517 		idev = ip6_dst_idev(dst);
4518 
4519 	switch (ipstats_mib_noroutes) {
4520 	case IPSTATS_MIB_INNOROUTES:
4521 		type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4522 		if (type == IPV6_ADDR_ANY) {
4523 			SKB_DR_SET(reason, IP_INADDRERRORS);
4524 			IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4525 			break;
4526 		}
4527 		SKB_DR_SET(reason, IP_INNOROUTES);
4528 		fallthrough;
4529 	case IPSTATS_MIB_OUTNOROUTES:
4530 		SKB_DR_OR(reason, IP_OUTNOROUTES);
4531 		IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4532 		break;
4533 	}
4534 
4535 	/* Start over by dropping the dst for l3mdev case */
4536 	if (netif_is_l3_master(skb->dev))
4537 		skb_dst_drop(skb);
4538 
4539 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4540 	kfree_skb_reason(skb, reason);
4541 	return 0;
4542 }
4543 
ip6_pkt_discard(struct sk_buff * skb)4544 static int ip6_pkt_discard(struct sk_buff *skb)
4545 {
4546 	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4547 }
4548 
ip6_pkt_discard_out(struct net * net,struct sock * sk,struct sk_buff * skb)4549 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4550 {
4551 	skb->dev = skb_dst(skb)->dev;
4552 	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4553 }
4554 
ip6_pkt_prohibit(struct sk_buff * skb)4555 static int ip6_pkt_prohibit(struct sk_buff *skb)
4556 {
4557 	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4558 }
4559 
ip6_pkt_prohibit_out(struct net * net,struct sock * sk,struct sk_buff * skb)4560 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4561 {
4562 	skb->dev = skb_dst(skb)->dev;
4563 	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4564 }
4565 
4566 /*
4567  *	Allocate a dst for local (unicast / anycast) address.
4568  */
4569 
addrconf_f6i_alloc(struct net * net,struct inet6_dev * idev,const struct in6_addr * addr,bool anycast,gfp_t gfp_flags,struct netlink_ext_ack * extack)4570 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4571 				     struct inet6_dev *idev,
4572 				     const struct in6_addr *addr,
4573 				     bool anycast, gfp_t gfp_flags,
4574 				     struct netlink_ext_ack *extack)
4575 {
4576 	struct fib6_config cfg = {
4577 		.fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4578 		.fc_ifindex = idev->dev->ifindex,
4579 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
4580 		.fc_dst = *addr,
4581 		.fc_dst_len = 128,
4582 		.fc_protocol = RTPROT_KERNEL,
4583 		.fc_nlinfo.nl_net = net,
4584 		.fc_ignore_dev_down = true,
4585 	};
4586 	struct fib6_info *f6i;
4587 
4588 	if (anycast) {
4589 		cfg.fc_type = RTN_ANYCAST;
4590 		cfg.fc_flags |= RTF_ANYCAST;
4591 	} else {
4592 		cfg.fc_type = RTN_LOCAL;
4593 		cfg.fc_flags |= RTF_LOCAL;
4594 	}
4595 
4596 	f6i = ip6_route_info_create(&cfg, gfp_flags, extack);
4597 	if (!IS_ERR(f6i)) {
4598 		f6i->dst_nocount = true;
4599 
4600 		if (!anycast &&
4601 		    (READ_ONCE(net->ipv6.devconf_all->disable_policy) ||
4602 		     READ_ONCE(idev->cnf.disable_policy)))
4603 			f6i->dst_nopolicy = true;
4604 	}
4605 
4606 	return f6i;
4607 }
4608 
4609 /* remove deleted ip from prefsrc entries */
4610 struct arg_dev_net_ip {
4611 	struct net *net;
4612 	struct in6_addr *addr;
4613 };
4614 
fib6_remove_prefsrc(struct fib6_info * rt,void * arg)4615 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4616 {
4617 	struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4618 	struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4619 
4620 	if (!rt->nh &&
4621 	    rt != net->ipv6.fib6_null_entry &&
4622 	    ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) &&
4623 	    !ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) {
4624 		spin_lock_bh(&rt6_exception_lock);
4625 		/* remove prefsrc entry */
4626 		rt->fib6_prefsrc.plen = 0;
4627 		spin_unlock_bh(&rt6_exception_lock);
4628 	}
4629 	return 0;
4630 }
4631 
rt6_remove_prefsrc(struct inet6_ifaddr * ifp)4632 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4633 {
4634 	struct net *net = dev_net(ifp->idev->dev);
4635 	struct arg_dev_net_ip adni = {
4636 		.net = net,
4637 		.addr = &ifp->addr,
4638 	};
4639 	fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4640 }
4641 
4642 #define RTF_RA_ROUTER		(RTF_ADDRCONF | RTF_DEFAULT)
4643 
4644 /* Remove routers and update dst entries when gateway turn into host. */
fib6_clean_tohost(struct fib6_info * rt,void * arg)4645 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4646 {
4647 	struct in6_addr *gateway = (struct in6_addr *)arg;
4648 	struct fib6_nh *nh;
4649 
4650 	/* RA routes do not use nexthops */
4651 	if (rt->nh)
4652 		return 0;
4653 
4654 	nh = rt->fib6_nh;
4655 	if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4656 	    nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4657 		return -1;
4658 
4659 	/* Further clean up cached routes in exception table.
4660 	 * This is needed because cached route may have a different
4661 	 * gateway than its 'parent' in the case of an ip redirect.
4662 	 */
4663 	fib6_nh_exceptions_clean_tohost(nh, gateway);
4664 
4665 	return 0;
4666 }
4667 
rt6_clean_tohost(struct net * net,struct in6_addr * gateway)4668 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4669 {
4670 	fib6_clean_all(net, fib6_clean_tohost, gateway);
4671 }
4672 
4673 struct arg_netdev_event {
4674 	const struct net_device *dev;
4675 	union {
4676 		unsigned char nh_flags;
4677 		unsigned long event;
4678 	};
4679 };
4680 
rt6_multipath_first_sibling(const struct fib6_info * rt)4681 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4682 {
4683 	struct fib6_info *iter;
4684 	struct fib6_node *fn;
4685 
4686 	fn = rcu_dereference_protected(rt->fib6_node,
4687 			lockdep_is_held(&rt->fib6_table->tb6_lock));
4688 	iter = rcu_dereference_protected(fn->leaf,
4689 			lockdep_is_held(&rt->fib6_table->tb6_lock));
4690 	while (iter) {
4691 		if (iter->fib6_metric == rt->fib6_metric &&
4692 		    rt6_qualify_for_ecmp(iter))
4693 			return iter;
4694 		iter = rcu_dereference_protected(iter->fib6_next,
4695 				lockdep_is_held(&rt->fib6_table->tb6_lock));
4696 	}
4697 
4698 	return NULL;
4699 }
4700 
4701 /* only called for fib entries with builtin fib6_nh */
rt6_is_dead(const struct fib6_info * rt)4702 static bool rt6_is_dead(const struct fib6_info *rt)
4703 {
4704 	if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4705 	    (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4706 	     ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4707 		return true;
4708 
4709 	return false;
4710 }
4711 
rt6_multipath_total_weight(const struct fib6_info * rt)4712 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4713 {
4714 	struct fib6_info *iter;
4715 	int total = 0;
4716 
4717 	if (!rt6_is_dead(rt))
4718 		total += rt->fib6_nh->fib_nh_weight;
4719 
4720 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4721 		if (!rt6_is_dead(iter))
4722 			total += iter->fib6_nh->fib_nh_weight;
4723 	}
4724 
4725 	return total;
4726 }
4727 
rt6_upper_bound_set(struct fib6_info * rt,int * weight,int total)4728 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4729 {
4730 	int upper_bound = -1;
4731 
4732 	if (!rt6_is_dead(rt)) {
4733 		*weight += rt->fib6_nh->fib_nh_weight;
4734 		upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4735 						    total) - 1;
4736 	}
4737 	atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4738 }
4739 
rt6_multipath_upper_bound_set(struct fib6_info * rt,int total)4740 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4741 {
4742 	struct fib6_info *iter;
4743 	int weight = 0;
4744 
4745 	rt6_upper_bound_set(rt, &weight, total);
4746 
4747 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4748 		rt6_upper_bound_set(iter, &weight, total);
4749 }
4750 
rt6_multipath_rebalance(struct fib6_info * rt)4751 void rt6_multipath_rebalance(struct fib6_info *rt)
4752 {
4753 	struct fib6_info *first;
4754 	int total;
4755 
4756 	/* In case the entire multipath route was marked for flushing,
4757 	 * then there is no need to rebalance upon the removal of every
4758 	 * sibling route.
4759 	 */
4760 	if (!rt->fib6_nsiblings || rt->should_flush)
4761 		return;
4762 
4763 	/* During lookup routes are evaluated in order, so we need to
4764 	 * make sure upper bounds are assigned from the first sibling
4765 	 * onwards.
4766 	 */
4767 	first = rt6_multipath_first_sibling(rt);
4768 	if (WARN_ON_ONCE(!first))
4769 		return;
4770 
4771 	total = rt6_multipath_total_weight(first);
4772 	rt6_multipath_upper_bound_set(first, total);
4773 }
4774 
fib6_ifup(struct fib6_info * rt,void * p_arg)4775 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4776 {
4777 	const struct arg_netdev_event *arg = p_arg;
4778 	struct net *net = dev_net(arg->dev);
4779 
4780 	if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4781 	    rt->fib6_nh->fib_nh_dev == arg->dev) {
4782 		rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4783 		fib6_update_sernum_upto_root(net, rt);
4784 		rt6_multipath_rebalance(rt);
4785 	}
4786 
4787 	return 0;
4788 }
4789 
rt6_sync_up(struct net_device * dev,unsigned char nh_flags)4790 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4791 {
4792 	struct arg_netdev_event arg = {
4793 		.dev = dev,
4794 		{
4795 			.nh_flags = nh_flags,
4796 		},
4797 	};
4798 
4799 	if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4800 		arg.nh_flags |= RTNH_F_LINKDOWN;
4801 
4802 	fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4803 }
4804 
4805 /* only called for fib entries with inline fib6_nh */
rt6_multipath_uses_dev(const struct fib6_info * rt,const struct net_device * dev)4806 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4807 				   const struct net_device *dev)
4808 {
4809 	struct fib6_info *iter;
4810 
4811 	if (rt->fib6_nh->fib_nh_dev == dev)
4812 		return true;
4813 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4814 		if (iter->fib6_nh->fib_nh_dev == dev)
4815 			return true;
4816 
4817 	return false;
4818 }
4819 
rt6_multipath_flush(struct fib6_info * rt)4820 static void rt6_multipath_flush(struct fib6_info *rt)
4821 {
4822 	struct fib6_info *iter;
4823 
4824 	rt->should_flush = 1;
4825 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4826 		iter->should_flush = 1;
4827 }
4828 
rt6_multipath_dead_count(const struct fib6_info * rt,const struct net_device * down_dev)4829 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4830 					     const struct net_device *down_dev)
4831 {
4832 	struct fib6_info *iter;
4833 	unsigned int dead = 0;
4834 
4835 	if (rt->fib6_nh->fib_nh_dev == down_dev ||
4836 	    rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4837 		dead++;
4838 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4839 		if (iter->fib6_nh->fib_nh_dev == down_dev ||
4840 		    iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4841 			dead++;
4842 
4843 	return dead;
4844 }
4845 
rt6_multipath_nh_flags_set(struct fib6_info * rt,const struct net_device * dev,unsigned char nh_flags)4846 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4847 				       const struct net_device *dev,
4848 				       unsigned char nh_flags)
4849 {
4850 	struct fib6_info *iter;
4851 
4852 	if (rt->fib6_nh->fib_nh_dev == dev)
4853 		rt->fib6_nh->fib_nh_flags |= nh_flags;
4854 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4855 		if (iter->fib6_nh->fib_nh_dev == dev)
4856 			iter->fib6_nh->fib_nh_flags |= nh_flags;
4857 }
4858 
4859 /* called with write lock held for table with rt */
fib6_ifdown(struct fib6_info * rt,void * p_arg)4860 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4861 {
4862 	const struct arg_netdev_event *arg = p_arg;
4863 	const struct net_device *dev = arg->dev;
4864 	struct net *net = dev_net(dev);
4865 
4866 	if (rt == net->ipv6.fib6_null_entry || rt->nh)
4867 		return 0;
4868 
4869 	switch (arg->event) {
4870 	case NETDEV_UNREGISTER:
4871 		return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4872 	case NETDEV_DOWN:
4873 		if (rt->should_flush)
4874 			return -1;
4875 		if (!rt->fib6_nsiblings)
4876 			return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4877 		if (rt6_multipath_uses_dev(rt, dev)) {
4878 			unsigned int count;
4879 
4880 			count = rt6_multipath_dead_count(rt, dev);
4881 			if (rt->fib6_nsiblings + 1 == count) {
4882 				rt6_multipath_flush(rt);
4883 				return -1;
4884 			}
4885 			rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4886 						   RTNH_F_LINKDOWN);
4887 			fib6_update_sernum(net, rt);
4888 			rt6_multipath_rebalance(rt);
4889 		}
4890 		return -2;
4891 	case NETDEV_CHANGE:
4892 		if (rt->fib6_nh->fib_nh_dev != dev ||
4893 		    rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4894 			break;
4895 		rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4896 		rt6_multipath_rebalance(rt);
4897 		break;
4898 	}
4899 
4900 	return 0;
4901 }
4902 
rt6_sync_down_dev(struct net_device * dev,unsigned long event)4903 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4904 {
4905 	struct arg_netdev_event arg = {
4906 		.dev = dev,
4907 		{
4908 			.event = event,
4909 		},
4910 	};
4911 	struct net *net = dev_net(dev);
4912 
4913 	if (net->ipv6.sysctl.skip_notify_on_dev_down)
4914 		fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4915 	else
4916 		fib6_clean_all(net, fib6_ifdown, &arg);
4917 }
4918 
rt6_disable_ip(struct net_device * dev,unsigned long event)4919 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4920 {
4921 	rt6_sync_down_dev(dev, event);
4922 	rt6_uncached_list_flush_dev(dev);
4923 	neigh_ifdown(&nd_tbl, dev);
4924 }
4925 
4926 struct rt6_mtu_change_arg {
4927 	struct net_device *dev;
4928 	unsigned int mtu;
4929 	struct fib6_info *f6i;
4930 };
4931 
fib6_nh_mtu_change(struct fib6_nh * nh,void * _arg)4932 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4933 {
4934 	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4935 	struct fib6_info *f6i = arg->f6i;
4936 
4937 	/* For administrative MTU increase, there is no way to discover
4938 	 * IPv6 PMTU increase, so PMTU increase should be updated here.
4939 	 * Since RFC 1981 doesn't include administrative MTU increase
4940 	 * update PMTU increase is a MUST. (i.e. jumbo frame)
4941 	 */
4942 	if (nh->fib_nh_dev == arg->dev) {
4943 		struct inet6_dev *idev = __in6_dev_get(arg->dev);
4944 		u32 mtu = f6i->fib6_pmtu;
4945 
4946 		if (mtu >= arg->mtu ||
4947 		    (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4948 			fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4949 
4950 		spin_lock_bh(&rt6_exception_lock);
4951 		rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4952 		spin_unlock_bh(&rt6_exception_lock);
4953 	}
4954 
4955 	return 0;
4956 }
4957 
rt6_mtu_change_route(struct fib6_info * f6i,void * p_arg)4958 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4959 {
4960 	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4961 	struct inet6_dev *idev;
4962 
4963 	/* In IPv6 pmtu discovery is not optional,
4964 	   so that RTAX_MTU lock cannot disable it.
4965 	   We still use this lock to block changes
4966 	   caused by addrconf/ndisc.
4967 	*/
4968 
4969 	idev = __in6_dev_get(arg->dev);
4970 	if (!idev)
4971 		return 0;
4972 
4973 	if (fib6_metric_locked(f6i, RTAX_MTU))
4974 		return 0;
4975 
4976 	arg->f6i = f6i;
4977 	if (f6i->nh) {
4978 		/* fib6_nh_mtu_change only returns 0, so this is safe */
4979 		return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4980 						arg);
4981 	}
4982 
4983 	return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4984 }
4985 
rt6_mtu_change(struct net_device * dev,unsigned int mtu)4986 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4987 {
4988 	struct rt6_mtu_change_arg arg = {
4989 		.dev = dev,
4990 		.mtu = mtu,
4991 	};
4992 
4993 	fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4994 }
4995 
4996 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4997 	[RTA_UNSPEC]		= { .strict_start_type = RTA_DPORT + 1 },
4998 	[RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
4999 	[RTA_PREFSRC]		= { .len = sizeof(struct in6_addr) },
5000 	[RTA_OIF]               = { .type = NLA_U32 },
5001 	[RTA_IIF]		= { .type = NLA_U32 },
5002 	[RTA_PRIORITY]          = { .type = NLA_U32 },
5003 	[RTA_METRICS]           = { .type = NLA_NESTED },
5004 	[RTA_MULTIPATH]		= { .len = sizeof(struct rtnexthop) },
5005 	[RTA_PREF]              = { .type = NLA_U8 },
5006 	[RTA_ENCAP_TYPE]	= { .type = NLA_U16 },
5007 	[RTA_ENCAP]		= { .type = NLA_NESTED },
5008 	[RTA_EXPIRES]		= { .type = NLA_U32 },
5009 	[RTA_UID]		= { .type = NLA_U32 },
5010 	[RTA_MARK]		= { .type = NLA_U32 },
5011 	[RTA_TABLE]		= { .type = NLA_U32 },
5012 	[RTA_IP_PROTO]		= { .type = NLA_U8 },
5013 	[RTA_SPORT]		= { .type = NLA_U16 },
5014 	[RTA_DPORT]		= { .type = NLA_U16 },
5015 	[RTA_NH_ID]		= { .type = NLA_U32 },
5016 	[RTA_FLOWLABEL]		= { .type = NLA_BE32 },
5017 };
5018 
rtm_to_fib6_config(struct sk_buff * skb,struct nlmsghdr * nlh,struct fib6_config * cfg,struct netlink_ext_ack * extack)5019 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
5020 			      struct fib6_config *cfg,
5021 			      struct netlink_ext_ack *extack)
5022 {
5023 	struct rtmsg *rtm;
5024 	struct nlattr *tb[RTA_MAX+1];
5025 	unsigned int pref;
5026 	int err;
5027 
5028 	err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5029 				     rtm_ipv6_policy, extack);
5030 	if (err < 0)
5031 		goto errout;
5032 
5033 	err = -EINVAL;
5034 	rtm = nlmsg_data(nlh);
5035 
5036 	if (rtm->rtm_tos) {
5037 		NL_SET_ERR_MSG(extack,
5038 			       "Invalid dsfield (tos): option not available for IPv6");
5039 		goto errout;
5040 	}
5041 
5042 	if (tb[RTA_FLOWLABEL]) {
5043 		NL_SET_ERR_MSG_ATTR(extack, tb[RTA_FLOWLABEL],
5044 				    "Flow label cannot be specified for this operation");
5045 		goto errout;
5046 	}
5047 
5048 	*cfg = (struct fib6_config){
5049 		.fc_table = rtm->rtm_table,
5050 		.fc_dst_len = rtm->rtm_dst_len,
5051 		.fc_src_len = rtm->rtm_src_len,
5052 		.fc_flags = RTF_UP,
5053 		.fc_protocol = rtm->rtm_protocol,
5054 		.fc_type = rtm->rtm_type,
5055 
5056 		.fc_nlinfo.portid = NETLINK_CB(skb).portid,
5057 		.fc_nlinfo.nlh = nlh,
5058 		.fc_nlinfo.nl_net = sock_net(skb->sk),
5059 	};
5060 
5061 	if (rtm->rtm_type == RTN_UNREACHABLE ||
5062 	    rtm->rtm_type == RTN_BLACKHOLE ||
5063 	    rtm->rtm_type == RTN_PROHIBIT ||
5064 	    rtm->rtm_type == RTN_THROW)
5065 		cfg->fc_flags |= RTF_REJECT;
5066 
5067 	if (rtm->rtm_type == RTN_LOCAL)
5068 		cfg->fc_flags |= RTF_LOCAL;
5069 
5070 	if (rtm->rtm_flags & RTM_F_CLONED)
5071 		cfg->fc_flags |= RTF_CACHE;
5072 
5073 	cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5074 
5075 	if (tb[RTA_NH_ID]) {
5076 		if (tb[RTA_GATEWAY]   || tb[RTA_OIF] ||
5077 		    tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5078 			NL_SET_ERR_MSG(extack,
5079 				       "Nexthop specification and nexthop id are mutually exclusive");
5080 			goto errout;
5081 		}
5082 		cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5083 	}
5084 
5085 	if (tb[RTA_GATEWAY]) {
5086 		cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5087 		cfg->fc_flags |= RTF_GATEWAY;
5088 	}
5089 	if (tb[RTA_VIA]) {
5090 		NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5091 		goto errout;
5092 	}
5093 
5094 	if (tb[RTA_DST]) {
5095 		int plen = (rtm->rtm_dst_len + 7) >> 3;
5096 
5097 		if (nla_len(tb[RTA_DST]) < plen)
5098 			goto errout;
5099 
5100 		nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5101 	}
5102 
5103 	if (tb[RTA_SRC]) {
5104 		int plen = (rtm->rtm_src_len + 7) >> 3;
5105 
5106 		if (nla_len(tb[RTA_SRC]) < plen)
5107 			goto errout;
5108 
5109 		nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5110 	}
5111 
5112 	if (tb[RTA_PREFSRC])
5113 		cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5114 
5115 	if (tb[RTA_OIF])
5116 		cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5117 
5118 	if (tb[RTA_PRIORITY])
5119 		cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5120 
5121 	if (tb[RTA_METRICS]) {
5122 		cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5123 		cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5124 	}
5125 
5126 	if (tb[RTA_TABLE])
5127 		cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5128 
5129 	if (tb[RTA_MULTIPATH]) {
5130 		cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5131 		cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5132 
5133 		err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5134 						     cfg->fc_mp_len, extack);
5135 		if (err < 0)
5136 			goto errout;
5137 	}
5138 
5139 	if (tb[RTA_PREF]) {
5140 		pref = nla_get_u8(tb[RTA_PREF]);
5141 		if (pref != ICMPV6_ROUTER_PREF_LOW &&
5142 		    pref != ICMPV6_ROUTER_PREF_HIGH)
5143 			pref = ICMPV6_ROUTER_PREF_MEDIUM;
5144 		cfg->fc_flags |= RTF_PREF(pref);
5145 	}
5146 
5147 	if (tb[RTA_ENCAP])
5148 		cfg->fc_encap = tb[RTA_ENCAP];
5149 
5150 	if (tb[RTA_ENCAP_TYPE]) {
5151 		cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5152 
5153 		err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5154 		if (err < 0)
5155 			goto errout;
5156 	}
5157 
5158 	if (tb[RTA_EXPIRES]) {
5159 		unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5160 
5161 		if (addrconf_finite_timeout(timeout)) {
5162 			cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5163 			cfg->fc_flags |= RTF_EXPIRES;
5164 		}
5165 	}
5166 
5167 	err = 0;
5168 errout:
5169 	return err;
5170 }
5171 
5172 struct rt6_nh {
5173 	struct fib6_info *fib6_info;
5174 	struct fib6_config r_cfg;
5175 	struct list_head next;
5176 };
5177 
ip6_route_info_append(struct net * net,struct list_head * rt6_nh_list,struct fib6_info * rt,struct fib6_config * r_cfg)5178 static int ip6_route_info_append(struct net *net,
5179 				 struct list_head *rt6_nh_list,
5180 				 struct fib6_info *rt,
5181 				 struct fib6_config *r_cfg)
5182 {
5183 	struct rt6_nh *nh;
5184 	int err = -EEXIST;
5185 
5186 	list_for_each_entry(nh, rt6_nh_list, next) {
5187 		/* check if fib6_info already exists */
5188 		if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5189 			return err;
5190 	}
5191 
5192 	nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5193 	if (!nh)
5194 		return -ENOMEM;
5195 	nh->fib6_info = rt;
5196 	memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5197 	list_add_tail(&nh->next, rt6_nh_list);
5198 
5199 	return 0;
5200 }
5201 
ip6_route_mpath_notify(struct fib6_info * rt,struct fib6_info * rt_last,struct nl_info * info,__u16 nlflags)5202 static void ip6_route_mpath_notify(struct fib6_info *rt,
5203 				   struct fib6_info *rt_last,
5204 				   struct nl_info *info,
5205 				   __u16 nlflags)
5206 {
5207 	/* if this is an APPEND route, then rt points to the first route
5208 	 * inserted and rt_last points to last route inserted. Userspace
5209 	 * wants a consistent dump of the route which starts at the first
5210 	 * nexthop. Since sibling routes are always added at the end of
5211 	 * the list, find the first sibling of the last route appended
5212 	 */
5213 	rcu_read_lock();
5214 
5215 	if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5216 		rt = list_first_or_null_rcu(&rt_last->fib6_siblings,
5217 					    struct fib6_info,
5218 					    fib6_siblings);
5219 	}
5220 
5221 	if (rt)
5222 		inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5223 
5224 	rcu_read_unlock();
5225 }
5226 
ip6_route_mpath_should_notify(const struct fib6_info * rt)5227 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5228 {
5229 	bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5230 	bool should_notify = false;
5231 	struct fib6_info *leaf;
5232 	struct fib6_node *fn;
5233 
5234 	rcu_read_lock();
5235 	fn = rcu_dereference(rt->fib6_node);
5236 	if (!fn)
5237 		goto out;
5238 
5239 	leaf = rcu_dereference(fn->leaf);
5240 	if (!leaf)
5241 		goto out;
5242 
5243 	if (rt == leaf ||
5244 	    (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5245 	     rt6_qualify_for_ecmp(leaf)))
5246 		should_notify = true;
5247 out:
5248 	rcu_read_unlock();
5249 
5250 	return should_notify;
5251 }
5252 
fib6_gw_from_attr(struct in6_addr * gw,struct nlattr * nla,struct netlink_ext_ack * extack)5253 static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5254 			     struct netlink_ext_ack *extack)
5255 {
5256 	if (nla_len(nla) < sizeof(*gw)) {
5257 		NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5258 		return -EINVAL;
5259 	}
5260 
5261 	*gw = nla_get_in6_addr(nla);
5262 
5263 	return 0;
5264 }
5265 
ip6_route_multipath_add(struct fib6_config * cfg,struct netlink_ext_ack * extack)5266 static int ip6_route_multipath_add(struct fib6_config *cfg,
5267 				   struct netlink_ext_ack *extack)
5268 {
5269 	struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5270 	struct nl_info *info = &cfg->fc_nlinfo;
5271 	struct fib6_config r_cfg;
5272 	struct rtnexthop *rtnh;
5273 	struct fib6_info *rt;
5274 	struct rt6_nh *err_nh;
5275 	struct rt6_nh *nh, *nh_safe;
5276 	__u16 nlflags;
5277 	int remaining;
5278 	int attrlen;
5279 	int err = 1;
5280 	int nhn = 0;
5281 	int replace = (cfg->fc_nlinfo.nlh &&
5282 		       (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5283 	LIST_HEAD(rt6_nh_list);
5284 
5285 	nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5286 	if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5287 		nlflags |= NLM_F_APPEND;
5288 
5289 	remaining = cfg->fc_mp_len;
5290 	rtnh = (struct rtnexthop *)cfg->fc_mp;
5291 
5292 	/* Parse a Multipath Entry and build a list (rt6_nh_list) of
5293 	 * fib6_info structs per nexthop
5294 	 */
5295 	while (rtnh_ok(rtnh, remaining)) {
5296 		memcpy(&r_cfg, cfg, sizeof(*cfg));
5297 		if (rtnh->rtnh_ifindex)
5298 			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5299 
5300 		attrlen = rtnh_attrlen(rtnh);
5301 		if (attrlen > 0) {
5302 			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5303 
5304 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5305 			if (nla) {
5306 				err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5307 							extack);
5308 				if (err)
5309 					goto cleanup;
5310 
5311 				r_cfg.fc_flags |= RTF_GATEWAY;
5312 			}
5313 			r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5314 
5315 			/* RTA_ENCAP_TYPE length checked in
5316 			 * lwtunnel_valid_encap_type_attr
5317 			 */
5318 			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5319 			if (nla)
5320 				r_cfg.fc_encap_type = nla_get_u16(nla);
5321 		}
5322 
5323 		r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5324 		rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5325 		if (IS_ERR(rt)) {
5326 			err = PTR_ERR(rt);
5327 			rt = NULL;
5328 			goto cleanup;
5329 		}
5330 		if (!rt6_qualify_for_ecmp(rt)) {
5331 			err = -EINVAL;
5332 			NL_SET_ERR_MSG(extack,
5333 				       "Device only routes can not be added for IPv6 using the multipath API.");
5334 			fib6_info_release(rt);
5335 			goto cleanup;
5336 		}
5337 
5338 		rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5339 
5340 		err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5341 					    rt, &r_cfg);
5342 		if (err) {
5343 			fib6_info_release(rt);
5344 			goto cleanup;
5345 		}
5346 
5347 		rtnh = rtnh_next(rtnh, &remaining);
5348 	}
5349 
5350 	if (list_empty(&rt6_nh_list)) {
5351 		NL_SET_ERR_MSG(extack,
5352 			       "Invalid nexthop configuration - no valid nexthops");
5353 		return -EINVAL;
5354 	}
5355 
5356 	/* for add and replace send one notification with all nexthops.
5357 	 * Skip the notification in fib6_add_rt2node and send one with
5358 	 * the full route when done
5359 	 */
5360 	info->skip_notify = 1;
5361 
5362 	/* For add and replace, send one notification with all nexthops. For
5363 	 * append, send one notification with all appended nexthops.
5364 	 */
5365 	info->skip_notify_kernel = 1;
5366 
5367 	err_nh = NULL;
5368 	list_for_each_entry(nh, &rt6_nh_list, next) {
5369 		err = __ip6_ins_rt(nh->fib6_info, info, extack);
5370 
5371 		if (err) {
5372 			if (replace && nhn)
5373 				NL_SET_ERR_MSG_MOD(extack,
5374 						   "multipath route replace failed (check consistency of installed routes)");
5375 			err_nh = nh;
5376 			goto add_errout;
5377 		}
5378 		/* save reference to last route successfully inserted */
5379 		rt_last = nh->fib6_info;
5380 
5381 		/* save reference to first route for notification */
5382 		if (!rt_notif)
5383 			rt_notif = nh->fib6_info;
5384 
5385 		/* Because each route is added like a single route we remove
5386 		 * these flags after the first nexthop: if there is a collision,
5387 		 * we have already failed to add the first nexthop:
5388 		 * fib6_add_rt2node() has rejected it; when replacing, old
5389 		 * nexthops have been replaced by first new, the rest should
5390 		 * be added to it.
5391 		 */
5392 		if (cfg->fc_nlinfo.nlh) {
5393 			cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5394 							     NLM_F_REPLACE);
5395 			cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5396 		}
5397 		nhn++;
5398 	}
5399 
5400 	/* An in-kernel notification should only be sent in case the new
5401 	 * multipath route is added as the first route in the node, or if
5402 	 * it was appended to it. We pass 'rt_notif' since it is the first
5403 	 * sibling and might allow us to skip some checks in the replace case.
5404 	 */
5405 	if (ip6_route_mpath_should_notify(rt_notif)) {
5406 		enum fib_event_type fib_event;
5407 
5408 		if (rt_notif->fib6_nsiblings != nhn - 1)
5409 			fib_event = FIB_EVENT_ENTRY_APPEND;
5410 		else
5411 			fib_event = FIB_EVENT_ENTRY_REPLACE;
5412 
5413 		err = call_fib6_multipath_entry_notifiers(info->nl_net,
5414 							  fib_event, rt_notif,
5415 							  nhn - 1, extack);
5416 		if (err) {
5417 			/* Delete all the siblings that were just added */
5418 			err_nh = NULL;
5419 			goto add_errout;
5420 		}
5421 	}
5422 
5423 	/* success ... tell user about new route */
5424 	ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5425 	goto cleanup;
5426 
5427 add_errout:
5428 	/* send notification for routes that were added so that
5429 	 * the delete notifications sent by ip6_route_del are
5430 	 * coherent
5431 	 */
5432 	if (rt_notif)
5433 		ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5434 
5435 	/* Delete routes that were already added */
5436 	list_for_each_entry(nh, &rt6_nh_list, next) {
5437 		if (err_nh == nh)
5438 			break;
5439 		ip6_route_del(&nh->r_cfg, extack);
5440 	}
5441 
5442 cleanup:
5443 	list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5444 		fib6_info_release(nh->fib6_info);
5445 		list_del(&nh->next);
5446 		kfree(nh);
5447 	}
5448 
5449 	return err;
5450 }
5451 
ip6_route_multipath_del(struct fib6_config * cfg,struct netlink_ext_ack * extack)5452 static int ip6_route_multipath_del(struct fib6_config *cfg,
5453 				   struct netlink_ext_ack *extack)
5454 {
5455 	struct fib6_config r_cfg;
5456 	struct rtnexthop *rtnh;
5457 	int last_err = 0;
5458 	int remaining;
5459 	int attrlen;
5460 	int err;
5461 
5462 	remaining = cfg->fc_mp_len;
5463 	rtnh = (struct rtnexthop *)cfg->fc_mp;
5464 
5465 	/* Parse a Multipath Entry */
5466 	while (rtnh_ok(rtnh, remaining)) {
5467 		memcpy(&r_cfg, cfg, sizeof(*cfg));
5468 		if (rtnh->rtnh_ifindex)
5469 			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5470 
5471 		attrlen = rtnh_attrlen(rtnh);
5472 		if (attrlen > 0) {
5473 			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5474 
5475 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5476 			if (nla) {
5477 				err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5478 							extack);
5479 				if (err) {
5480 					last_err = err;
5481 					goto next_rtnh;
5482 				}
5483 
5484 				r_cfg.fc_flags |= RTF_GATEWAY;
5485 			}
5486 		}
5487 		err = ip6_route_del(&r_cfg, extack);
5488 		if (err)
5489 			last_err = err;
5490 
5491 next_rtnh:
5492 		rtnh = rtnh_next(rtnh, &remaining);
5493 	}
5494 
5495 	return last_err;
5496 }
5497 
inet6_rtm_delroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5498 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5499 			      struct netlink_ext_ack *extack)
5500 {
5501 	struct fib6_config cfg;
5502 	int err;
5503 
5504 	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5505 	if (err < 0)
5506 		return err;
5507 
5508 	if (cfg.fc_nh_id &&
5509 	    !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5510 		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5511 		return -EINVAL;
5512 	}
5513 
5514 	if (cfg.fc_mp)
5515 		return ip6_route_multipath_del(&cfg, extack);
5516 	else {
5517 		cfg.fc_delete_all_nh = 1;
5518 		return ip6_route_del(&cfg, extack);
5519 	}
5520 }
5521 
inet6_rtm_newroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5522 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5523 			      struct netlink_ext_ack *extack)
5524 {
5525 	struct fib6_config cfg;
5526 	int err;
5527 
5528 	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5529 	if (err < 0)
5530 		return err;
5531 
5532 	if (cfg.fc_metric == 0)
5533 		cfg.fc_metric = IP6_RT_PRIO_USER;
5534 
5535 	if (cfg.fc_mp)
5536 		return ip6_route_multipath_add(&cfg, extack);
5537 	else
5538 		return ip6_route_add(&cfg, GFP_KERNEL, extack);
5539 }
5540 
5541 /* add the overhead of this fib6_nh to nexthop_len */
rt6_nh_nlmsg_size(struct fib6_nh * nh,void * arg)5542 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5543 {
5544 	int *nexthop_len = arg;
5545 
5546 	*nexthop_len += nla_total_size(0)	 /* RTA_MULTIPATH */
5547 		     + NLA_ALIGN(sizeof(struct rtnexthop))
5548 		     + nla_total_size(16); /* RTA_GATEWAY */
5549 
5550 	if (nh->fib_nh_lws) {
5551 		/* RTA_ENCAP_TYPE */
5552 		*nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5553 		/* RTA_ENCAP */
5554 		*nexthop_len += nla_total_size(2);
5555 	}
5556 
5557 	return 0;
5558 }
5559 
rt6_nlmsg_size(struct fib6_info * f6i)5560 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5561 {
5562 	int nexthop_len;
5563 
5564 	if (f6i->nh) {
5565 		nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5566 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5567 					 &nexthop_len);
5568 	} else {
5569 		struct fib6_nh *nh = f6i->fib6_nh;
5570 		struct fib6_info *sibling;
5571 
5572 		nexthop_len = 0;
5573 		if (f6i->fib6_nsiblings) {
5574 			rt6_nh_nlmsg_size(nh, &nexthop_len);
5575 
5576 			rcu_read_lock();
5577 
5578 			list_for_each_entry_rcu(sibling, &f6i->fib6_siblings,
5579 						fib6_siblings) {
5580 				rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5581 			}
5582 
5583 			rcu_read_unlock();
5584 		}
5585 		nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5586 	}
5587 
5588 	return NLMSG_ALIGN(sizeof(struct rtmsg))
5589 	       + nla_total_size(16) /* RTA_SRC */
5590 	       + nla_total_size(16) /* RTA_DST */
5591 	       + nla_total_size(16) /* RTA_GATEWAY */
5592 	       + nla_total_size(16) /* RTA_PREFSRC */
5593 	       + nla_total_size(4) /* RTA_TABLE */
5594 	       + nla_total_size(4) /* RTA_IIF */
5595 	       + nla_total_size(4) /* RTA_OIF */
5596 	       + nla_total_size(4) /* RTA_PRIORITY */
5597 	       + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5598 	       + nla_total_size(sizeof(struct rta_cacheinfo))
5599 	       + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5600 	       + nla_total_size(1) /* RTA_PREF */
5601 	       + nexthop_len;
5602 }
5603 
rt6_fill_node_nexthop(struct sk_buff * skb,struct nexthop * nh,unsigned char * flags)5604 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5605 				 unsigned char *flags)
5606 {
5607 	if (nexthop_is_multipath(nh)) {
5608 		struct nlattr *mp;
5609 
5610 		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5611 		if (!mp)
5612 			goto nla_put_failure;
5613 
5614 		if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5615 			goto nla_put_failure;
5616 
5617 		nla_nest_end(skb, mp);
5618 	} else {
5619 		struct fib6_nh *fib6_nh;
5620 
5621 		fib6_nh = nexthop_fib6_nh(nh);
5622 		if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5623 				     flags, false) < 0)
5624 			goto nla_put_failure;
5625 	}
5626 
5627 	return 0;
5628 
5629 nla_put_failure:
5630 	return -EMSGSIZE;
5631 }
5632 
rt6_fill_node(struct net * net,struct sk_buff * skb,struct fib6_info * rt,struct dst_entry * dst,struct in6_addr * dest,struct in6_addr * src,int iif,int type,u32 portid,u32 seq,unsigned int flags)5633 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5634 			 struct fib6_info *rt, struct dst_entry *dst,
5635 			 struct in6_addr *dest, struct in6_addr *src,
5636 			 int iif, int type, u32 portid, u32 seq,
5637 			 unsigned int flags)
5638 {
5639 	struct rt6_info *rt6 = dst_rt6_info(dst);
5640 	struct rt6key *rt6_dst, *rt6_src;
5641 	u32 *pmetrics, table, rt6_flags;
5642 	unsigned char nh_flags = 0;
5643 	struct nlmsghdr *nlh;
5644 	struct rtmsg *rtm;
5645 	long expires = 0;
5646 
5647 	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5648 	if (!nlh)
5649 		return -EMSGSIZE;
5650 
5651 	if (rt6) {
5652 		rt6_dst = &rt6->rt6i_dst;
5653 		rt6_src = &rt6->rt6i_src;
5654 		rt6_flags = rt6->rt6i_flags;
5655 	} else {
5656 		rt6_dst = &rt->fib6_dst;
5657 		rt6_src = &rt->fib6_src;
5658 		rt6_flags = rt->fib6_flags;
5659 	}
5660 
5661 	rtm = nlmsg_data(nlh);
5662 	rtm->rtm_family = AF_INET6;
5663 	rtm->rtm_dst_len = rt6_dst->plen;
5664 	rtm->rtm_src_len = rt6_src->plen;
5665 	rtm->rtm_tos = 0;
5666 	if (rt->fib6_table)
5667 		table = rt->fib6_table->tb6_id;
5668 	else
5669 		table = RT6_TABLE_UNSPEC;
5670 	rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5671 	if (nla_put_u32(skb, RTA_TABLE, table))
5672 		goto nla_put_failure;
5673 
5674 	rtm->rtm_type = rt->fib6_type;
5675 	rtm->rtm_flags = 0;
5676 	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5677 	rtm->rtm_protocol = rt->fib6_protocol;
5678 
5679 	if (rt6_flags & RTF_CACHE)
5680 		rtm->rtm_flags |= RTM_F_CLONED;
5681 
5682 	if (dest) {
5683 		if (nla_put_in6_addr(skb, RTA_DST, dest))
5684 			goto nla_put_failure;
5685 		rtm->rtm_dst_len = 128;
5686 	} else if (rtm->rtm_dst_len)
5687 		if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5688 			goto nla_put_failure;
5689 #ifdef CONFIG_IPV6_SUBTREES
5690 	if (src) {
5691 		if (nla_put_in6_addr(skb, RTA_SRC, src))
5692 			goto nla_put_failure;
5693 		rtm->rtm_src_len = 128;
5694 	} else if (rtm->rtm_src_len &&
5695 		   nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5696 		goto nla_put_failure;
5697 #endif
5698 	if (iif) {
5699 #ifdef CONFIG_IPV6_MROUTE
5700 		if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5701 			int err = ip6mr_get_route(net, skb, rtm, portid);
5702 
5703 			if (err == 0)
5704 				return 0;
5705 			if (err < 0)
5706 				goto nla_put_failure;
5707 		} else
5708 #endif
5709 			if (nla_put_u32(skb, RTA_IIF, iif))
5710 				goto nla_put_failure;
5711 	} else if (dest) {
5712 		struct in6_addr saddr_buf;
5713 		if (ip6_route_get_saddr(net, rt, dest, 0, 0, &saddr_buf) == 0 &&
5714 		    nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5715 			goto nla_put_failure;
5716 	}
5717 
5718 	if (rt->fib6_prefsrc.plen) {
5719 		struct in6_addr saddr_buf;
5720 		saddr_buf = rt->fib6_prefsrc.addr;
5721 		if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5722 			goto nla_put_failure;
5723 	}
5724 
5725 	pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5726 	if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5727 		goto nla_put_failure;
5728 
5729 	if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5730 		goto nla_put_failure;
5731 
5732 	/* For multipath routes, walk the siblings list and add
5733 	 * each as a nexthop within RTA_MULTIPATH.
5734 	 */
5735 	if (rt6) {
5736 		if (rt6_flags & RTF_GATEWAY &&
5737 		    nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5738 			goto nla_put_failure;
5739 
5740 		if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5741 			goto nla_put_failure;
5742 
5743 		if (dst->lwtstate &&
5744 		    lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5745 			goto nla_put_failure;
5746 	} else if (rt->fib6_nsiblings) {
5747 		struct fib6_info *sibling;
5748 		struct nlattr *mp;
5749 
5750 		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5751 		if (!mp)
5752 			goto nla_put_failure;
5753 
5754 		if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5755 				    rt->fib6_nh->fib_nh_weight, AF_INET6,
5756 				    0) < 0)
5757 			goto nla_put_failure;
5758 
5759 		rcu_read_lock();
5760 
5761 		list_for_each_entry_rcu(sibling, &rt->fib6_siblings,
5762 					fib6_siblings) {
5763 			if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5764 					    sibling->fib6_nh->fib_nh_weight,
5765 					    AF_INET6, 0) < 0) {
5766 				rcu_read_unlock();
5767 
5768 				goto nla_put_failure;
5769 			}
5770 		}
5771 
5772 		rcu_read_unlock();
5773 
5774 		nla_nest_end(skb, mp);
5775 	} else if (rt->nh) {
5776 		if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5777 			goto nla_put_failure;
5778 
5779 		if (nexthop_is_blackhole(rt->nh))
5780 			rtm->rtm_type = RTN_BLACKHOLE;
5781 
5782 		if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5783 		    rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5784 			goto nla_put_failure;
5785 
5786 		rtm->rtm_flags |= nh_flags;
5787 	} else {
5788 		if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5789 				     &nh_flags, false) < 0)
5790 			goto nla_put_failure;
5791 
5792 		rtm->rtm_flags |= nh_flags;
5793 	}
5794 
5795 	if (rt6_flags & RTF_EXPIRES) {
5796 		expires = dst ? dst->expires : rt->expires;
5797 		expires -= jiffies;
5798 	}
5799 
5800 	if (!dst) {
5801 		if (READ_ONCE(rt->offload))
5802 			rtm->rtm_flags |= RTM_F_OFFLOAD;
5803 		if (READ_ONCE(rt->trap))
5804 			rtm->rtm_flags |= RTM_F_TRAP;
5805 		if (READ_ONCE(rt->offload_failed))
5806 			rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5807 	}
5808 
5809 	if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5810 		goto nla_put_failure;
5811 
5812 	if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5813 		goto nla_put_failure;
5814 
5815 
5816 	nlmsg_end(skb, nlh);
5817 	return 0;
5818 
5819 nla_put_failure:
5820 	nlmsg_cancel(skb, nlh);
5821 	return -EMSGSIZE;
5822 }
5823 
fib6_info_nh_uses_dev(struct fib6_nh * nh,void * arg)5824 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5825 {
5826 	const struct net_device *dev = arg;
5827 
5828 	if (nh->fib_nh_dev == dev)
5829 		return 1;
5830 
5831 	return 0;
5832 }
5833 
fib6_info_uses_dev(const struct fib6_info * f6i,const struct net_device * dev)5834 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5835 			       const struct net_device *dev)
5836 {
5837 	if (f6i->nh) {
5838 		struct net_device *_dev = (struct net_device *)dev;
5839 
5840 		return !!nexthop_for_each_fib6_nh(f6i->nh,
5841 						  fib6_info_nh_uses_dev,
5842 						  _dev);
5843 	}
5844 
5845 	if (f6i->fib6_nh->fib_nh_dev == dev)
5846 		return true;
5847 
5848 	if (f6i->fib6_nsiblings) {
5849 		struct fib6_info *sibling, *next_sibling;
5850 
5851 		list_for_each_entry_safe(sibling, next_sibling,
5852 					 &f6i->fib6_siblings, fib6_siblings) {
5853 			if (sibling->fib6_nh->fib_nh_dev == dev)
5854 				return true;
5855 		}
5856 	}
5857 
5858 	return false;
5859 }
5860 
5861 struct fib6_nh_exception_dump_walker {
5862 	struct rt6_rtnl_dump_arg *dump;
5863 	struct fib6_info *rt;
5864 	unsigned int flags;
5865 	unsigned int skip;
5866 	unsigned int count;
5867 };
5868 
rt6_nh_dump_exceptions(struct fib6_nh * nh,void * arg)5869 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5870 {
5871 	struct fib6_nh_exception_dump_walker *w = arg;
5872 	struct rt6_rtnl_dump_arg *dump = w->dump;
5873 	struct rt6_exception_bucket *bucket;
5874 	struct rt6_exception *rt6_ex;
5875 	int i, err;
5876 
5877 	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5878 	if (!bucket)
5879 		return 0;
5880 
5881 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5882 		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5883 			if (w->skip) {
5884 				w->skip--;
5885 				continue;
5886 			}
5887 
5888 			/* Expiration of entries doesn't bump sernum, insertion
5889 			 * does. Removal is triggered by insertion, so we can
5890 			 * rely on the fact that if entries change between two
5891 			 * partial dumps, this node is scanned again completely,
5892 			 * see rt6_insert_exception() and fib6_dump_table().
5893 			 *
5894 			 * Count expired entries we go through as handled
5895 			 * entries that we'll skip next time, in case of partial
5896 			 * node dump. Otherwise, if entries expire meanwhile,
5897 			 * we'll skip the wrong amount.
5898 			 */
5899 			if (rt6_check_expired(rt6_ex->rt6i)) {
5900 				w->count++;
5901 				continue;
5902 			}
5903 
5904 			err = rt6_fill_node(dump->net, dump->skb, w->rt,
5905 					    &rt6_ex->rt6i->dst, NULL, NULL, 0,
5906 					    RTM_NEWROUTE,
5907 					    NETLINK_CB(dump->cb->skb).portid,
5908 					    dump->cb->nlh->nlmsg_seq, w->flags);
5909 			if (err)
5910 				return err;
5911 
5912 			w->count++;
5913 		}
5914 		bucket++;
5915 	}
5916 
5917 	return 0;
5918 }
5919 
5920 /* Return -1 if done with node, number of handled routes on partial dump */
rt6_dump_route(struct fib6_info * rt,void * p_arg,unsigned int skip)5921 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5922 {
5923 	struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5924 	struct fib_dump_filter *filter = &arg->filter;
5925 	unsigned int flags = NLM_F_MULTI;
5926 	struct net *net = arg->net;
5927 	int count = 0;
5928 
5929 	if (rt == net->ipv6.fib6_null_entry)
5930 		return -1;
5931 
5932 	if ((filter->flags & RTM_F_PREFIX) &&
5933 	    !(rt->fib6_flags & RTF_PREFIX_RT)) {
5934 		/* success since this is not a prefix route */
5935 		return -1;
5936 	}
5937 	if (filter->filter_set &&
5938 	    ((filter->rt_type  && rt->fib6_type != filter->rt_type) ||
5939 	     (filter->dev      && !fib6_info_uses_dev(rt, filter->dev)) ||
5940 	     (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5941 		return -1;
5942 	}
5943 
5944 	if (filter->filter_set ||
5945 	    !filter->dump_routes || !filter->dump_exceptions) {
5946 		flags |= NLM_F_DUMP_FILTERED;
5947 	}
5948 
5949 	if (filter->dump_routes) {
5950 		if (skip) {
5951 			skip--;
5952 		} else {
5953 			if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5954 					  0, RTM_NEWROUTE,
5955 					  NETLINK_CB(arg->cb->skb).portid,
5956 					  arg->cb->nlh->nlmsg_seq, flags)) {
5957 				return 0;
5958 			}
5959 			count++;
5960 		}
5961 	}
5962 
5963 	if (filter->dump_exceptions) {
5964 		struct fib6_nh_exception_dump_walker w = { .dump = arg,
5965 							   .rt = rt,
5966 							   .flags = flags,
5967 							   .skip = skip,
5968 							   .count = 0 };
5969 		int err;
5970 
5971 		rcu_read_lock();
5972 		if (rt->nh) {
5973 			err = nexthop_for_each_fib6_nh(rt->nh,
5974 						       rt6_nh_dump_exceptions,
5975 						       &w);
5976 		} else {
5977 			err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5978 		}
5979 		rcu_read_unlock();
5980 
5981 		if (err)
5982 			return count + w.count;
5983 	}
5984 
5985 	return -1;
5986 }
5987 
inet6_rtm_valid_getroute_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5988 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5989 					const struct nlmsghdr *nlh,
5990 					struct nlattr **tb,
5991 					struct netlink_ext_ack *extack)
5992 {
5993 	struct rtmsg *rtm;
5994 	int i, err;
5995 
5996 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5997 		NL_SET_ERR_MSG_MOD(extack,
5998 				   "Invalid header for get route request");
5999 		return -EINVAL;
6000 	}
6001 
6002 	if (!netlink_strict_get_check(skb))
6003 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
6004 					      rtm_ipv6_policy, extack);
6005 
6006 	rtm = nlmsg_data(nlh);
6007 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
6008 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
6009 	    rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
6010 	    rtm->rtm_type) {
6011 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
6012 		return -EINVAL;
6013 	}
6014 	if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
6015 		NL_SET_ERR_MSG_MOD(extack,
6016 				   "Invalid flags for get route request");
6017 		return -EINVAL;
6018 	}
6019 
6020 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
6021 					    rtm_ipv6_policy, extack);
6022 	if (err)
6023 		return err;
6024 
6025 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
6026 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
6027 		NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
6028 		return -EINVAL;
6029 	}
6030 
6031 	if (tb[RTA_FLOWLABEL] &&
6032 	    (nla_get_be32(tb[RTA_FLOWLABEL]) & ~IPV6_FLOWLABEL_MASK)) {
6033 		NL_SET_ERR_MSG_ATTR(extack, tb[RTA_FLOWLABEL],
6034 				    "Invalid flow label");
6035 		return -EINVAL;
6036 	}
6037 
6038 	for (i = 0; i <= RTA_MAX; i++) {
6039 		if (!tb[i])
6040 			continue;
6041 
6042 		switch (i) {
6043 		case RTA_SRC:
6044 		case RTA_DST:
6045 		case RTA_IIF:
6046 		case RTA_OIF:
6047 		case RTA_MARK:
6048 		case RTA_UID:
6049 		case RTA_SPORT:
6050 		case RTA_DPORT:
6051 		case RTA_IP_PROTO:
6052 		case RTA_FLOWLABEL:
6053 			break;
6054 		default:
6055 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6056 			return -EINVAL;
6057 		}
6058 	}
6059 
6060 	return 0;
6061 }
6062 
inet6_rtm_getroute(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6063 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6064 			      struct netlink_ext_ack *extack)
6065 {
6066 	struct net *net = sock_net(in_skb->sk);
6067 	struct nlattr *tb[RTA_MAX+1];
6068 	int err, iif = 0, oif = 0;
6069 	struct fib6_info *from;
6070 	struct dst_entry *dst;
6071 	struct rt6_info *rt;
6072 	struct sk_buff *skb;
6073 	struct rtmsg *rtm;
6074 	struct flowi6 fl6 = {};
6075 	__be32 flowlabel;
6076 	bool fibmatch;
6077 
6078 	err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6079 	if (err < 0)
6080 		goto errout;
6081 
6082 	err = -EINVAL;
6083 	rtm = nlmsg_data(nlh);
6084 	fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6085 
6086 	if (tb[RTA_SRC]) {
6087 		if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6088 			goto errout;
6089 
6090 		fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6091 	}
6092 
6093 	if (tb[RTA_DST]) {
6094 		if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6095 			goto errout;
6096 
6097 		fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6098 	}
6099 
6100 	if (tb[RTA_IIF])
6101 		iif = nla_get_u32(tb[RTA_IIF]);
6102 
6103 	if (tb[RTA_OIF])
6104 		oif = nla_get_u32(tb[RTA_OIF]);
6105 
6106 	if (tb[RTA_MARK])
6107 		fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6108 
6109 	if (tb[RTA_UID])
6110 		fl6.flowi6_uid = make_kuid(current_user_ns(),
6111 					   nla_get_u32(tb[RTA_UID]));
6112 	else
6113 		fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6114 
6115 	if (tb[RTA_SPORT])
6116 		fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6117 
6118 	if (tb[RTA_DPORT])
6119 		fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6120 
6121 	if (tb[RTA_IP_PROTO]) {
6122 		err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6123 						  &fl6.flowi6_proto, AF_INET6,
6124 						  extack);
6125 		if (err)
6126 			goto errout;
6127 	}
6128 
6129 	flowlabel = nla_get_be32_default(tb[RTA_FLOWLABEL], 0);
6130 	fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, flowlabel);
6131 
6132 	if (iif) {
6133 		struct net_device *dev;
6134 		int flags = 0;
6135 
6136 		rcu_read_lock();
6137 
6138 		dev = dev_get_by_index_rcu(net, iif);
6139 		if (!dev) {
6140 			rcu_read_unlock();
6141 			err = -ENODEV;
6142 			goto errout;
6143 		}
6144 
6145 		fl6.flowi6_iif = iif;
6146 
6147 		if (!ipv6_addr_any(&fl6.saddr))
6148 			flags |= RT6_LOOKUP_F_HAS_SADDR;
6149 
6150 		dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6151 
6152 		rcu_read_unlock();
6153 	} else {
6154 		fl6.flowi6_oif = oif;
6155 
6156 		dst = ip6_route_output(net, NULL, &fl6);
6157 	}
6158 
6159 
6160 	rt = dst_rt6_info(dst);
6161 	if (rt->dst.error) {
6162 		err = rt->dst.error;
6163 		ip6_rt_put(rt);
6164 		goto errout;
6165 	}
6166 
6167 	if (rt == net->ipv6.ip6_null_entry) {
6168 		err = rt->dst.error;
6169 		ip6_rt_put(rt);
6170 		goto errout;
6171 	}
6172 
6173 	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6174 	if (!skb) {
6175 		ip6_rt_put(rt);
6176 		err = -ENOBUFS;
6177 		goto errout;
6178 	}
6179 
6180 	skb_dst_set(skb, &rt->dst);
6181 
6182 	rcu_read_lock();
6183 	from = rcu_dereference(rt->from);
6184 	if (from) {
6185 		if (fibmatch)
6186 			err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6187 					    iif, RTM_NEWROUTE,
6188 					    NETLINK_CB(in_skb).portid,
6189 					    nlh->nlmsg_seq, 0);
6190 		else
6191 			err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6192 					    &fl6.saddr, iif, RTM_NEWROUTE,
6193 					    NETLINK_CB(in_skb).portid,
6194 					    nlh->nlmsg_seq, 0);
6195 	} else {
6196 		err = -ENETUNREACH;
6197 	}
6198 	rcu_read_unlock();
6199 
6200 	if (err < 0) {
6201 		kfree_skb(skb);
6202 		goto errout;
6203 	}
6204 
6205 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6206 errout:
6207 	return err;
6208 }
6209 
inet6_rt_notify(int event,struct fib6_info * rt,struct nl_info * info,unsigned int nlm_flags)6210 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6211 		     unsigned int nlm_flags)
6212 {
6213 	struct sk_buff *skb;
6214 	struct net *net = info->nl_net;
6215 	u32 seq;
6216 	int err;
6217 
6218 	err = -ENOBUFS;
6219 	seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6220 
6221 	skb = nlmsg_new(rt6_nlmsg_size(rt), GFP_ATOMIC);
6222 	if (!skb)
6223 		goto errout;
6224 
6225 	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6226 			    event, info->portid, seq, nlm_flags);
6227 	if (err < 0) {
6228 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6229 		WARN_ON(err == -EMSGSIZE);
6230 		kfree_skb(skb);
6231 		goto errout;
6232 	}
6233 	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6234 		    info->nlh, GFP_ATOMIC);
6235 	return;
6236 errout:
6237 	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6238 }
6239 
fib6_rt_update(struct net * net,struct fib6_info * rt,struct nl_info * info)6240 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6241 		    struct nl_info *info)
6242 {
6243 	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6244 	struct sk_buff *skb;
6245 	int err = -ENOBUFS;
6246 
6247 	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6248 	if (!skb)
6249 		goto errout;
6250 
6251 	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6252 			    RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6253 	if (err < 0) {
6254 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6255 		WARN_ON(err == -EMSGSIZE);
6256 		kfree_skb(skb);
6257 		goto errout;
6258 	}
6259 	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6260 		    info->nlh, gfp_any());
6261 	return;
6262 errout:
6263 	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6264 }
6265 
fib6_info_hw_flags_set(struct net * net,struct fib6_info * f6i,bool offload,bool trap,bool offload_failed)6266 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6267 			    bool offload, bool trap, bool offload_failed)
6268 {
6269 	struct sk_buff *skb;
6270 	int err;
6271 
6272 	if (READ_ONCE(f6i->offload) == offload &&
6273 	    READ_ONCE(f6i->trap) == trap &&
6274 	    READ_ONCE(f6i->offload_failed) == offload_failed)
6275 		return;
6276 
6277 	WRITE_ONCE(f6i->offload, offload);
6278 	WRITE_ONCE(f6i->trap, trap);
6279 
6280 	/* 2 means send notifications only if offload_failed was changed. */
6281 	if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6282 	    READ_ONCE(f6i->offload_failed) == offload_failed)
6283 		return;
6284 
6285 	WRITE_ONCE(f6i->offload_failed, offload_failed);
6286 
6287 	if (!rcu_access_pointer(f6i->fib6_node))
6288 		/* The route was removed from the tree, do not send
6289 		 * notification.
6290 		 */
6291 		return;
6292 
6293 	if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6294 		return;
6295 
6296 	skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6297 	if (!skb) {
6298 		err = -ENOBUFS;
6299 		goto errout;
6300 	}
6301 
6302 	err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6303 			    0, 0);
6304 	if (err < 0) {
6305 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6306 		WARN_ON(err == -EMSGSIZE);
6307 		kfree_skb(skb);
6308 		goto errout;
6309 	}
6310 
6311 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6312 	return;
6313 
6314 errout:
6315 	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6316 }
6317 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6318 
ip6_route_dev_notify(struct notifier_block * this,unsigned long event,void * ptr)6319 static int ip6_route_dev_notify(struct notifier_block *this,
6320 				unsigned long event, void *ptr)
6321 {
6322 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6323 	struct net *net = dev_net(dev);
6324 
6325 	if (!(dev->flags & IFF_LOOPBACK))
6326 		return NOTIFY_OK;
6327 
6328 	if (event == NETDEV_REGISTER) {
6329 		net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6330 		net->ipv6.ip6_null_entry->dst.dev = dev;
6331 		net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6332 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6333 		net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6334 		net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6335 		net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6336 		net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6337 #endif
6338 	 } else if (event == NETDEV_UNREGISTER &&
6339 		    dev->reg_state != NETREG_UNREGISTERED) {
6340 		/* NETDEV_UNREGISTER could be fired for multiple times by
6341 		 * netdev_wait_allrefs(). Make sure we only call this once.
6342 		 */
6343 		in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6344 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6345 		in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6346 		in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6347 #endif
6348 	}
6349 
6350 	return NOTIFY_OK;
6351 }
6352 
6353 /*
6354  *	/proc
6355  */
6356 
6357 #ifdef CONFIG_PROC_FS
rt6_stats_seq_show(struct seq_file * seq,void * v)6358 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6359 {
6360 	struct net *net = (struct net *)seq->private;
6361 	seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6362 		   net->ipv6.rt6_stats->fib_nodes,
6363 		   net->ipv6.rt6_stats->fib_route_nodes,
6364 		   atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6365 		   net->ipv6.rt6_stats->fib_rt_entries,
6366 		   net->ipv6.rt6_stats->fib_rt_cache,
6367 		   dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6368 		   net->ipv6.rt6_stats->fib_discarded_routes);
6369 
6370 	return 0;
6371 }
6372 #endif	/* CONFIG_PROC_FS */
6373 
6374 #ifdef CONFIG_SYSCTL
6375 
ipv6_sysctl_rtcache_flush(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6376 static int ipv6_sysctl_rtcache_flush(const struct ctl_table *ctl, int write,
6377 			      void *buffer, size_t *lenp, loff_t *ppos)
6378 {
6379 	struct net *net;
6380 	int delay;
6381 	int ret;
6382 	if (!write)
6383 		return -EINVAL;
6384 
6385 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6386 	if (ret)
6387 		return ret;
6388 
6389 	net = (struct net *)ctl->extra1;
6390 	delay = net->ipv6.sysctl.flush_delay;
6391 	fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6392 	return 0;
6393 }
6394 
6395 static struct ctl_table ipv6_route_table_template[] = {
6396 	{
6397 		.procname	=	"max_size",
6398 		.data		=	&init_net.ipv6.sysctl.ip6_rt_max_size,
6399 		.maxlen		=	sizeof(int),
6400 		.mode		=	0644,
6401 		.proc_handler	=	proc_dointvec,
6402 	},
6403 	{
6404 		.procname	=	"gc_thresh",
6405 		.data		=	&ip6_dst_ops_template.gc_thresh,
6406 		.maxlen		=	sizeof(int),
6407 		.mode		=	0644,
6408 		.proc_handler	=	proc_dointvec,
6409 	},
6410 	{
6411 		.procname	=	"flush",
6412 		.data		=	&init_net.ipv6.sysctl.flush_delay,
6413 		.maxlen		=	sizeof(int),
6414 		.mode		=	0200,
6415 		.proc_handler	=	ipv6_sysctl_rtcache_flush
6416 	},
6417 	{
6418 		.procname	=	"gc_min_interval",
6419 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6420 		.maxlen		=	sizeof(int),
6421 		.mode		=	0644,
6422 		.proc_handler	=	proc_dointvec_jiffies,
6423 	},
6424 	{
6425 		.procname	=	"gc_timeout",
6426 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6427 		.maxlen		=	sizeof(int),
6428 		.mode		=	0644,
6429 		.proc_handler	=	proc_dointvec_jiffies,
6430 	},
6431 	{
6432 		.procname	=	"gc_interval",
6433 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_interval,
6434 		.maxlen		=	sizeof(int),
6435 		.mode		=	0644,
6436 		.proc_handler	=	proc_dointvec_jiffies,
6437 	},
6438 	{
6439 		.procname	=	"gc_elasticity",
6440 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6441 		.maxlen		=	sizeof(int),
6442 		.mode		=	0644,
6443 		.proc_handler	=	proc_dointvec,
6444 	},
6445 	{
6446 		.procname	=	"mtu_expires",
6447 		.data		=	&init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6448 		.maxlen		=	sizeof(int),
6449 		.mode		=	0644,
6450 		.proc_handler	=	proc_dointvec_jiffies,
6451 	},
6452 	{
6453 		.procname	=	"min_adv_mss",
6454 		.data		=	&init_net.ipv6.sysctl.ip6_rt_min_advmss,
6455 		.maxlen		=	sizeof(int),
6456 		.mode		=	0644,
6457 		.proc_handler	=	proc_dointvec,
6458 	},
6459 	{
6460 		.procname	=	"gc_min_interval_ms",
6461 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6462 		.maxlen		=	sizeof(int),
6463 		.mode		=	0644,
6464 		.proc_handler	=	proc_dointvec_ms_jiffies,
6465 	},
6466 	{
6467 		.procname	=	"skip_notify_on_dev_down",
6468 		.data		=	&init_net.ipv6.sysctl.skip_notify_on_dev_down,
6469 		.maxlen		=	sizeof(u8),
6470 		.mode		=	0644,
6471 		.proc_handler	=	proc_dou8vec_minmax,
6472 		.extra1		=	SYSCTL_ZERO,
6473 		.extra2		=	SYSCTL_ONE,
6474 	},
6475 };
6476 
ipv6_route_sysctl_init(struct net * net)6477 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6478 {
6479 	struct ctl_table *table;
6480 
6481 	table = kmemdup(ipv6_route_table_template,
6482 			sizeof(ipv6_route_table_template),
6483 			GFP_KERNEL);
6484 
6485 	if (table) {
6486 		table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6487 		table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6488 		table[2].data = &net->ipv6.sysctl.flush_delay;
6489 		table[2].extra1 = net;
6490 		table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6491 		table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6492 		table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6493 		table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6494 		table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6495 		table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6496 		table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6497 		table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6498 	}
6499 
6500 	return table;
6501 }
6502 
ipv6_route_sysctl_table_size(struct net * net)6503 size_t ipv6_route_sysctl_table_size(struct net *net)
6504 {
6505 	/* Don't export sysctls to unprivileged users */
6506 	if (net->user_ns != &init_user_ns)
6507 		return 1;
6508 
6509 	return ARRAY_SIZE(ipv6_route_table_template);
6510 }
6511 #endif
6512 
ip6_route_net_init(struct net * net)6513 static int __net_init ip6_route_net_init(struct net *net)
6514 {
6515 	int ret = -ENOMEM;
6516 
6517 	memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6518 	       sizeof(net->ipv6.ip6_dst_ops));
6519 
6520 	if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6521 		goto out_ip6_dst_ops;
6522 
6523 	net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6524 	if (!net->ipv6.fib6_null_entry)
6525 		goto out_ip6_dst_entries;
6526 	memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6527 	       sizeof(*net->ipv6.fib6_null_entry));
6528 
6529 	net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6530 					   sizeof(*net->ipv6.ip6_null_entry),
6531 					   GFP_KERNEL);
6532 	if (!net->ipv6.ip6_null_entry)
6533 		goto out_fib6_null_entry;
6534 	net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6535 	dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6536 			 ip6_template_metrics, true);
6537 	INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6538 
6539 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6540 	net->ipv6.fib6_has_custom_rules = false;
6541 	net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6542 					       sizeof(*net->ipv6.ip6_prohibit_entry),
6543 					       GFP_KERNEL);
6544 	if (!net->ipv6.ip6_prohibit_entry)
6545 		goto out_ip6_null_entry;
6546 	net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6547 	dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6548 			 ip6_template_metrics, true);
6549 	INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6550 
6551 	net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6552 					       sizeof(*net->ipv6.ip6_blk_hole_entry),
6553 					       GFP_KERNEL);
6554 	if (!net->ipv6.ip6_blk_hole_entry)
6555 		goto out_ip6_prohibit_entry;
6556 	net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6557 	dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6558 			 ip6_template_metrics, true);
6559 	INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6560 #ifdef CONFIG_IPV6_SUBTREES
6561 	net->ipv6.fib6_routes_require_src = 0;
6562 #endif
6563 #endif
6564 
6565 	net->ipv6.sysctl.flush_delay = 0;
6566 	net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6567 	net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6568 	net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6569 	net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6570 	net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6571 	net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6572 	net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6573 	net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6574 
6575 	atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6576 
6577 	ret = 0;
6578 out:
6579 	return ret;
6580 
6581 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6582 out_ip6_prohibit_entry:
6583 	kfree(net->ipv6.ip6_prohibit_entry);
6584 out_ip6_null_entry:
6585 	kfree(net->ipv6.ip6_null_entry);
6586 #endif
6587 out_fib6_null_entry:
6588 	kfree(net->ipv6.fib6_null_entry);
6589 out_ip6_dst_entries:
6590 	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6591 out_ip6_dst_ops:
6592 	goto out;
6593 }
6594 
ip6_route_net_exit(struct net * net)6595 static void __net_exit ip6_route_net_exit(struct net *net)
6596 {
6597 	kfree(net->ipv6.fib6_null_entry);
6598 	kfree(net->ipv6.ip6_null_entry);
6599 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6600 	kfree(net->ipv6.ip6_prohibit_entry);
6601 	kfree(net->ipv6.ip6_blk_hole_entry);
6602 #endif
6603 	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6604 }
6605 
ip6_route_net_init_late(struct net * net)6606 static int __net_init ip6_route_net_init_late(struct net *net)
6607 {
6608 #ifdef CONFIG_PROC_FS
6609 	if (!proc_create_net("ipv6_route", 0, net->proc_net,
6610 			     &ipv6_route_seq_ops,
6611 			     sizeof(struct ipv6_route_iter)))
6612 		return -ENOMEM;
6613 
6614 	if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6615 				    rt6_stats_seq_show, NULL)) {
6616 		remove_proc_entry("ipv6_route", net->proc_net);
6617 		return -ENOMEM;
6618 	}
6619 #endif
6620 	return 0;
6621 }
6622 
ip6_route_net_exit_late(struct net * net)6623 static void __net_exit ip6_route_net_exit_late(struct net *net)
6624 {
6625 #ifdef CONFIG_PROC_FS
6626 	remove_proc_entry("ipv6_route", net->proc_net);
6627 	remove_proc_entry("rt6_stats", net->proc_net);
6628 #endif
6629 }
6630 
6631 static struct pernet_operations ip6_route_net_ops = {
6632 	.init = ip6_route_net_init,
6633 	.exit = ip6_route_net_exit,
6634 };
6635 
ipv6_inetpeer_init(struct net * net)6636 static int __net_init ipv6_inetpeer_init(struct net *net)
6637 {
6638 	struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6639 
6640 	if (!bp)
6641 		return -ENOMEM;
6642 	inet_peer_base_init(bp);
6643 	net->ipv6.peers = bp;
6644 	return 0;
6645 }
6646 
ipv6_inetpeer_exit(struct net * net)6647 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6648 {
6649 	struct inet_peer_base *bp = net->ipv6.peers;
6650 
6651 	net->ipv6.peers = NULL;
6652 	inetpeer_invalidate_tree(bp);
6653 	kfree(bp);
6654 }
6655 
6656 static struct pernet_operations ipv6_inetpeer_ops = {
6657 	.init	=	ipv6_inetpeer_init,
6658 	.exit	=	ipv6_inetpeer_exit,
6659 };
6660 
6661 static struct pernet_operations ip6_route_net_late_ops = {
6662 	.init = ip6_route_net_init_late,
6663 	.exit = ip6_route_net_exit_late,
6664 };
6665 
6666 static struct notifier_block ip6_route_dev_notifier = {
6667 	.notifier_call = ip6_route_dev_notify,
6668 	.priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6669 };
6670 
ip6_route_init_special_entries(void)6671 void __init ip6_route_init_special_entries(void)
6672 {
6673 	/* Registering of the loopback is done before this portion of code,
6674 	 * the loopback reference in rt6_info will not be taken, do it
6675 	 * manually for init_net */
6676 	init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6677 	init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6678 	init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6679   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6680 	init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6681 	init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6682 	init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6683 	init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6684   #endif
6685 }
6686 
6687 #if IS_BUILTIN(CONFIG_IPV6)
6688 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6689 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6690 
6691 BTF_ID_LIST(btf_fib6_info_id)
6692 BTF_ID(struct, fib6_info)
6693 
6694 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6695 	.seq_ops		= &ipv6_route_seq_ops,
6696 	.init_seq_private	= bpf_iter_init_seq_net,
6697 	.fini_seq_private	= bpf_iter_fini_seq_net,
6698 	.seq_priv_size		= sizeof(struct ipv6_route_iter),
6699 };
6700 
6701 static struct bpf_iter_reg ipv6_route_reg_info = {
6702 	.target			= "ipv6_route",
6703 	.ctx_arg_info_size	= 1,
6704 	.ctx_arg_info		= {
6705 		{ offsetof(struct bpf_iter__ipv6_route, rt),
6706 		  PTR_TO_BTF_ID_OR_NULL },
6707 	},
6708 	.seq_info		= &ipv6_route_seq_info,
6709 };
6710 
bpf_iter_register(void)6711 static int __init bpf_iter_register(void)
6712 {
6713 	ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6714 	return bpf_iter_reg_target(&ipv6_route_reg_info);
6715 }
6716 
bpf_iter_unregister(void)6717 static void bpf_iter_unregister(void)
6718 {
6719 	bpf_iter_unreg_target(&ipv6_route_reg_info);
6720 }
6721 #endif
6722 #endif
6723 
6724 static const struct rtnl_msg_handler ip6_route_rtnl_msg_handlers[] __initconst_or_module = {
6725 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWROUTE,
6726 	 .doit = inet6_rtm_newroute},
6727 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELROUTE,
6728 	 .doit = inet6_rtm_delroute},
6729 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETROUTE,
6730 	 .doit = inet6_rtm_getroute, .flags = RTNL_FLAG_DOIT_UNLOCKED},
6731 };
6732 
ip6_route_init(void)6733 int __init ip6_route_init(void)
6734 {
6735 	int ret;
6736 	int cpu;
6737 
6738 	ret = -ENOMEM;
6739 	ip6_dst_ops_template.kmem_cachep =
6740 		kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6741 				  SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6742 	if (!ip6_dst_ops_template.kmem_cachep)
6743 		goto out;
6744 
6745 	ret = dst_entries_init(&ip6_dst_blackhole_ops);
6746 	if (ret)
6747 		goto out_kmem_cache;
6748 
6749 	ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6750 	if (ret)
6751 		goto out_dst_entries;
6752 
6753 	ret = register_pernet_subsys(&ip6_route_net_ops);
6754 	if (ret)
6755 		goto out_register_inetpeer;
6756 
6757 	ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6758 
6759 	ret = fib6_init();
6760 	if (ret)
6761 		goto out_register_subsys;
6762 
6763 	ret = xfrm6_init();
6764 	if (ret)
6765 		goto out_fib6_init;
6766 
6767 	ret = fib6_rules_init();
6768 	if (ret)
6769 		goto xfrm6_init;
6770 
6771 	ret = register_pernet_subsys(&ip6_route_net_late_ops);
6772 	if (ret)
6773 		goto fib6_rules_init;
6774 
6775 	ret = rtnl_register_many(ip6_route_rtnl_msg_handlers);
6776 	if (ret < 0)
6777 		goto out_register_late_subsys;
6778 
6779 	ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6780 	if (ret)
6781 		goto out_register_late_subsys;
6782 
6783 #if IS_BUILTIN(CONFIG_IPV6)
6784 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6785 	ret = bpf_iter_register();
6786 	if (ret)
6787 		goto out_register_late_subsys;
6788 #endif
6789 #endif
6790 
6791 	for_each_possible_cpu(cpu) {
6792 		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6793 
6794 		INIT_LIST_HEAD(&ul->head);
6795 		spin_lock_init(&ul->lock);
6796 	}
6797 
6798 out:
6799 	return ret;
6800 
6801 out_register_late_subsys:
6802 	rtnl_unregister_all(PF_INET6);
6803 	unregister_pernet_subsys(&ip6_route_net_late_ops);
6804 fib6_rules_init:
6805 	fib6_rules_cleanup();
6806 xfrm6_init:
6807 	xfrm6_fini();
6808 out_fib6_init:
6809 	fib6_gc_cleanup();
6810 out_register_subsys:
6811 	unregister_pernet_subsys(&ip6_route_net_ops);
6812 out_register_inetpeer:
6813 	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6814 out_dst_entries:
6815 	dst_entries_destroy(&ip6_dst_blackhole_ops);
6816 out_kmem_cache:
6817 	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6818 	goto out;
6819 }
6820 
ip6_route_cleanup(void)6821 void ip6_route_cleanup(void)
6822 {
6823 #if IS_BUILTIN(CONFIG_IPV6)
6824 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6825 	bpf_iter_unregister();
6826 #endif
6827 #endif
6828 	unregister_netdevice_notifier(&ip6_route_dev_notifier);
6829 	unregister_pernet_subsys(&ip6_route_net_late_ops);
6830 	fib6_rules_cleanup();
6831 	xfrm6_fini();
6832 	fib6_gc_cleanup();
6833 	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6834 	unregister_pernet_subsys(&ip6_route_net_ops);
6835 	dst_entries_destroy(&ip6_dst_blackhole_ops);
6836 	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6837 }
6838