xref: /linux/net/ipv6/route.c (revision 9410645520e9b820069761f3450ef6661418e279)
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 
378 	if (idev && idev->dev != blackhole_netdev) {
379 		struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);
380 
381 		if (blackhole_idev) {
382 			rt->rt6i_idev = blackhole_idev;
383 			in6_dev_put(idev);
384 		}
385 	}
386 }
387 
__rt6_check_expired(const struct rt6_info * rt)388 static bool __rt6_check_expired(const struct rt6_info *rt)
389 {
390 	if (rt->rt6i_flags & RTF_EXPIRES)
391 		return time_after(jiffies, rt->dst.expires);
392 	else
393 		return false;
394 }
395 
rt6_check_expired(const struct rt6_info * rt)396 static bool rt6_check_expired(const struct rt6_info *rt)
397 {
398 	struct fib6_info *from;
399 
400 	from = rcu_dereference(rt->from);
401 
402 	if (rt->rt6i_flags & RTF_EXPIRES) {
403 		if (time_after(jiffies, rt->dst.expires))
404 			return true;
405 	} else if (from) {
406 		return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
407 			fib6_check_expired(from);
408 	}
409 	return false;
410 }
411 
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)412 void fib6_select_path(const struct net *net, struct fib6_result *res,
413 		      struct flowi6 *fl6, int oif, bool have_oif_match,
414 		      const struct sk_buff *skb, int strict)
415 {
416 	struct fib6_info *sibling, *next_sibling;
417 	struct fib6_info *match = res->f6i;
418 
419 	if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
420 		goto out;
421 
422 	if (match->nh && have_oif_match && res->nh)
423 		return;
424 
425 	if (skb)
426 		IP6CB(skb)->flags |= IP6SKB_MULTIPATH;
427 
428 	/* We might have already computed the hash for ICMPv6 errors. In such
429 	 * case it will always be non-zero. Otherwise now is the time to do it.
430 	 */
431 	if (!fl6->mp_hash &&
432 	    (!match->nh || nexthop_is_multipath(match->nh)))
433 		fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
434 
435 	if (unlikely(match->nh)) {
436 		nexthop_path_fib6_result(res, fl6->mp_hash);
437 		return;
438 	}
439 
440 	if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
441 		goto out;
442 
443 	list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
444 				 fib6_siblings) {
445 		const struct fib6_nh *nh = sibling->fib6_nh;
446 		int nh_upper_bound;
447 
448 		nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
449 		if (fl6->mp_hash > nh_upper_bound)
450 			continue;
451 		if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
452 			break;
453 		match = sibling;
454 		break;
455 	}
456 
457 out:
458 	res->f6i = match;
459 	res->nh = match->fib6_nh;
460 }
461 
462 /*
463  *	Route lookup. rcu_read_lock() should be held.
464  */
465 
__rt6_device_match(struct net * net,const struct fib6_nh * nh,const struct in6_addr * saddr,int oif,int flags)466 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
467 			       const struct in6_addr *saddr, int oif, int flags)
468 {
469 	const struct net_device *dev;
470 
471 	if (nh->fib_nh_flags & RTNH_F_DEAD)
472 		return false;
473 
474 	dev = nh->fib_nh_dev;
475 	if (oif) {
476 		if (dev->ifindex == oif)
477 			return true;
478 	} else {
479 		if (ipv6_chk_addr(net, saddr, dev,
480 				  flags & RT6_LOOKUP_F_IFACE))
481 			return true;
482 	}
483 
484 	return false;
485 }
486 
487 struct fib6_nh_dm_arg {
488 	struct net		*net;
489 	const struct in6_addr	*saddr;
490 	int			oif;
491 	int			flags;
492 	struct fib6_nh		*nh;
493 };
494 
__rt6_nh_dev_match(struct fib6_nh * nh,void * _arg)495 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
496 {
497 	struct fib6_nh_dm_arg *arg = _arg;
498 
499 	arg->nh = nh;
500 	return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
501 				  arg->flags);
502 }
503 
504 /* 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)505 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
506 					struct fib6_result *res,
507 					const struct in6_addr *saddr,
508 					int oif, int flags)
509 {
510 	struct fib6_nh_dm_arg arg = {
511 		.net   = net,
512 		.saddr = saddr,
513 		.oif   = oif,
514 		.flags = flags,
515 	};
516 
517 	if (nexthop_is_blackhole(nh))
518 		return NULL;
519 
520 	if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
521 		return arg.nh;
522 
523 	return NULL;
524 }
525 
rt6_device_match(struct net * net,struct fib6_result * res,const struct in6_addr * saddr,int oif,int flags)526 static void rt6_device_match(struct net *net, struct fib6_result *res,
527 			     const struct in6_addr *saddr, int oif, int flags)
528 {
529 	struct fib6_info *f6i = res->f6i;
530 	struct fib6_info *spf6i;
531 	struct fib6_nh *nh;
532 
533 	if (!oif && ipv6_addr_any(saddr)) {
534 		if (unlikely(f6i->nh)) {
535 			nh = nexthop_fib6_nh(f6i->nh);
536 			if (nexthop_is_blackhole(f6i->nh))
537 				goto out_blackhole;
538 		} else {
539 			nh = f6i->fib6_nh;
540 		}
541 		if (!(nh->fib_nh_flags & RTNH_F_DEAD))
542 			goto out;
543 	}
544 
545 	for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
546 		bool matched = false;
547 
548 		if (unlikely(spf6i->nh)) {
549 			nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
550 					      oif, flags);
551 			if (nh)
552 				matched = true;
553 		} else {
554 			nh = spf6i->fib6_nh;
555 			if (__rt6_device_match(net, nh, saddr, oif, flags))
556 				matched = true;
557 		}
558 		if (matched) {
559 			res->f6i = spf6i;
560 			goto out;
561 		}
562 	}
563 
564 	if (oif && flags & RT6_LOOKUP_F_IFACE) {
565 		res->f6i = net->ipv6.fib6_null_entry;
566 		nh = res->f6i->fib6_nh;
567 		goto out;
568 	}
569 
570 	if (unlikely(f6i->nh)) {
571 		nh = nexthop_fib6_nh(f6i->nh);
572 		if (nexthop_is_blackhole(f6i->nh))
573 			goto out_blackhole;
574 	} else {
575 		nh = f6i->fib6_nh;
576 	}
577 
578 	if (nh->fib_nh_flags & RTNH_F_DEAD) {
579 		res->f6i = net->ipv6.fib6_null_entry;
580 		nh = res->f6i->fib6_nh;
581 	}
582 out:
583 	res->nh = nh;
584 	res->fib6_type = res->f6i->fib6_type;
585 	res->fib6_flags = res->f6i->fib6_flags;
586 	return;
587 
588 out_blackhole:
589 	res->fib6_flags |= RTF_REJECT;
590 	res->fib6_type = RTN_BLACKHOLE;
591 	res->nh = nh;
592 }
593 
594 #ifdef CONFIG_IPV6_ROUTER_PREF
595 struct __rt6_probe_work {
596 	struct work_struct work;
597 	struct in6_addr target;
598 	struct net_device *dev;
599 	netdevice_tracker dev_tracker;
600 };
601 
rt6_probe_deferred(struct work_struct * w)602 static void rt6_probe_deferred(struct work_struct *w)
603 {
604 	struct in6_addr mcaddr;
605 	struct __rt6_probe_work *work =
606 		container_of(w, struct __rt6_probe_work, work);
607 
608 	addrconf_addr_solict_mult(&work->target, &mcaddr);
609 	ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
610 	netdev_put(work->dev, &work->dev_tracker);
611 	kfree(work);
612 }
613 
rt6_probe(struct fib6_nh * fib6_nh)614 static void rt6_probe(struct fib6_nh *fib6_nh)
615 {
616 	struct __rt6_probe_work *work = NULL;
617 	const struct in6_addr *nh_gw;
618 	unsigned long last_probe;
619 	struct neighbour *neigh;
620 	struct net_device *dev;
621 	struct inet6_dev *idev;
622 
623 	/*
624 	 * Okay, this does not seem to be appropriate
625 	 * for now, however, we need to check if it
626 	 * is really so; aka Router Reachability Probing.
627 	 *
628 	 * Router Reachability Probe MUST be rate-limited
629 	 * to no more than one per minute.
630 	 */
631 	if (!fib6_nh->fib_nh_gw_family)
632 		return;
633 
634 	nh_gw = &fib6_nh->fib_nh_gw6;
635 	dev = fib6_nh->fib_nh_dev;
636 	rcu_read_lock();
637 	last_probe = READ_ONCE(fib6_nh->last_probe);
638 	idev = __in6_dev_get(dev);
639 	if (!idev)
640 		goto out;
641 	neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
642 	if (neigh) {
643 		if (READ_ONCE(neigh->nud_state) & NUD_VALID)
644 			goto out;
645 
646 		write_lock_bh(&neigh->lock);
647 		if (!(neigh->nud_state & NUD_VALID) &&
648 		    time_after(jiffies,
649 			       neigh->updated +
650 			       READ_ONCE(idev->cnf.rtr_probe_interval))) {
651 			work = kmalloc(sizeof(*work), GFP_ATOMIC);
652 			if (work)
653 				__neigh_set_probe_once(neigh);
654 		}
655 		write_unlock_bh(&neigh->lock);
656 	} else if (time_after(jiffies, last_probe +
657 				       READ_ONCE(idev->cnf.rtr_probe_interval))) {
658 		work = kmalloc(sizeof(*work), GFP_ATOMIC);
659 	}
660 
661 	if (!work || cmpxchg(&fib6_nh->last_probe,
662 			     last_probe, jiffies) != last_probe) {
663 		kfree(work);
664 	} else {
665 		INIT_WORK(&work->work, rt6_probe_deferred);
666 		work->target = *nh_gw;
667 		netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
668 		work->dev = dev;
669 		schedule_work(&work->work);
670 	}
671 
672 out:
673 	rcu_read_unlock();
674 }
675 #else
rt6_probe(struct fib6_nh * fib6_nh)676 static inline void rt6_probe(struct fib6_nh *fib6_nh)
677 {
678 }
679 #endif
680 
681 /*
682  * Default Router Selection (RFC 2461 6.3.6)
683  */
rt6_check_neigh(const struct fib6_nh * fib6_nh)684 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
685 {
686 	enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
687 	struct neighbour *neigh;
688 
689 	rcu_read_lock();
690 	neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
691 					  &fib6_nh->fib_nh_gw6);
692 	if (neigh) {
693 		u8 nud_state = READ_ONCE(neigh->nud_state);
694 
695 		if (nud_state & NUD_VALID)
696 			ret = RT6_NUD_SUCCEED;
697 #ifdef CONFIG_IPV6_ROUTER_PREF
698 		else if (!(nud_state & NUD_FAILED))
699 			ret = RT6_NUD_SUCCEED;
700 		else
701 			ret = RT6_NUD_FAIL_PROBE;
702 #endif
703 	} else {
704 		ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
705 		      RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
706 	}
707 	rcu_read_unlock();
708 
709 	return ret;
710 }
711 
rt6_score_route(const struct fib6_nh * nh,u32 fib6_flags,int oif,int strict)712 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
713 			   int strict)
714 {
715 	int m = 0;
716 
717 	if (!oif || nh->fib_nh_dev->ifindex == oif)
718 		m = 2;
719 
720 	if (!m && (strict & RT6_LOOKUP_F_IFACE))
721 		return RT6_NUD_FAIL_HARD;
722 #ifdef CONFIG_IPV6_ROUTER_PREF
723 	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
724 #endif
725 	if ((strict & RT6_LOOKUP_F_REACHABLE) &&
726 	    !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
727 		int n = rt6_check_neigh(nh);
728 		if (n < 0)
729 			return n;
730 	}
731 	return m;
732 }
733 
find_match(struct fib6_nh * nh,u32 fib6_flags,int oif,int strict,int * mpri,bool * do_rr)734 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
735 		       int oif, int strict, int *mpri, bool *do_rr)
736 {
737 	bool match_do_rr = false;
738 	bool rc = false;
739 	int m;
740 
741 	if (nh->fib_nh_flags & RTNH_F_DEAD)
742 		goto out;
743 
744 	if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
745 	    nh->fib_nh_flags & RTNH_F_LINKDOWN &&
746 	    !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
747 		goto out;
748 
749 	m = rt6_score_route(nh, fib6_flags, oif, strict);
750 	if (m == RT6_NUD_FAIL_DO_RR) {
751 		match_do_rr = true;
752 		m = 0; /* lowest valid score */
753 	} else if (m == RT6_NUD_FAIL_HARD) {
754 		goto out;
755 	}
756 
757 	if (strict & RT6_LOOKUP_F_REACHABLE)
758 		rt6_probe(nh);
759 
760 	/* note that m can be RT6_NUD_FAIL_PROBE at this point */
761 	if (m > *mpri) {
762 		*do_rr = match_do_rr;
763 		*mpri = m;
764 		rc = true;
765 	}
766 out:
767 	return rc;
768 }
769 
770 struct fib6_nh_frl_arg {
771 	u32		flags;
772 	int		oif;
773 	int		strict;
774 	int		*mpri;
775 	bool		*do_rr;
776 	struct fib6_nh	*nh;
777 };
778 
rt6_nh_find_match(struct fib6_nh * nh,void * _arg)779 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
780 {
781 	struct fib6_nh_frl_arg *arg = _arg;
782 
783 	arg->nh = nh;
784 	return find_match(nh, arg->flags, arg->oif, arg->strict,
785 			  arg->mpri, arg->do_rr);
786 }
787 
__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)788 static void __find_rr_leaf(struct fib6_info *f6i_start,
789 			   struct fib6_info *nomatch, u32 metric,
790 			   struct fib6_result *res, struct fib6_info **cont,
791 			   int oif, int strict, bool *do_rr, int *mpri)
792 {
793 	struct fib6_info *f6i;
794 
795 	for (f6i = f6i_start;
796 	     f6i && f6i != nomatch;
797 	     f6i = rcu_dereference(f6i->fib6_next)) {
798 		bool matched = false;
799 		struct fib6_nh *nh;
800 
801 		if (cont && f6i->fib6_metric != metric) {
802 			*cont = f6i;
803 			return;
804 		}
805 
806 		if (fib6_check_expired(f6i))
807 			continue;
808 
809 		if (unlikely(f6i->nh)) {
810 			struct fib6_nh_frl_arg arg = {
811 				.flags  = f6i->fib6_flags,
812 				.oif    = oif,
813 				.strict = strict,
814 				.mpri   = mpri,
815 				.do_rr  = do_rr
816 			};
817 
818 			if (nexthop_is_blackhole(f6i->nh)) {
819 				res->fib6_flags = RTF_REJECT;
820 				res->fib6_type = RTN_BLACKHOLE;
821 				res->f6i = f6i;
822 				res->nh = nexthop_fib6_nh(f6i->nh);
823 				return;
824 			}
825 			if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
826 						     &arg)) {
827 				matched = true;
828 				nh = arg.nh;
829 			}
830 		} else {
831 			nh = f6i->fib6_nh;
832 			if (find_match(nh, f6i->fib6_flags, oif, strict,
833 				       mpri, do_rr))
834 				matched = true;
835 		}
836 		if (matched) {
837 			res->f6i = f6i;
838 			res->nh = nh;
839 			res->fib6_flags = f6i->fib6_flags;
840 			res->fib6_type = f6i->fib6_type;
841 		}
842 	}
843 }
844 
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)845 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
846 			 struct fib6_info *rr_head, int oif, int strict,
847 			 bool *do_rr, struct fib6_result *res)
848 {
849 	u32 metric = rr_head->fib6_metric;
850 	struct fib6_info *cont = NULL;
851 	int mpri = -1;
852 
853 	__find_rr_leaf(rr_head, NULL, metric, res, &cont,
854 		       oif, strict, do_rr, &mpri);
855 
856 	__find_rr_leaf(leaf, rr_head, metric, res, &cont,
857 		       oif, strict, do_rr, &mpri);
858 
859 	if (res->f6i || !cont)
860 		return;
861 
862 	__find_rr_leaf(cont, NULL, metric, res, NULL,
863 		       oif, strict, do_rr, &mpri);
864 }
865 
rt6_select(struct net * net,struct fib6_node * fn,int oif,struct fib6_result * res,int strict)866 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
867 		       struct fib6_result *res, int strict)
868 {
869 	struct fib6_info *leaf = rcu_dereference(fn->leaf);
870 	struct fib6_info *rt0;
871 	bool do_rr = false;
872 	int key_plen;
873 
874 	/* make sure this function or its helpers sets f6i */
875 	res->f6i = NULL;
876 
877 	if (!leaf || leaf == net->ipv6.fib6_null_entry)
878 		goto out;
879 
880 	rt0 = rcu_dereference(fn->rr_ptr);
881 	if (!rt0)
882 		rt0 = leaf;
883 
884 	/* Double check to make sure fn is not an intermediate node
885 	 * and fn->leaf does not points to its child's leaf
886 	 * (This might happen if all routes under fn are deleted from
887 	 * the tree and fib6_repair_tree() is called on the node.)
888 	 */
889 	key_plen = rt0->fib6_dst.plen;
890 #ifdef CONFIG_IPV6_SUBTREES
891 	if (rt0->fib6_src.plen)
892 		key_plen = rt0->fib6_src.plen;
893 #endif
894 	if (fn->fn_bit != key_plen)
895 		goto out;
896 
897 	find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
898 	if (do_rr) {
899 		struct fib6_info *next = rcu_dereference(rt0->fib6_next);
900 
901 		/* no entries matched; do round-robin */
902 		if (!next || next->fib6_metric != rt0->fib6_metric)
903 			next = leaf;
904 
905 		if (next != rt0) {
906 			spin_lock_bh(&leaf->fib6_table->tb6_lock);
907 			/* make sure next is not being deleted from the tree */
908 			if (next->fib6_node)
909 				rcu_assign_pointer(fn->rr_ptr, next);
910 			spin_unlock_bh(&leaf->fib6_table->tb6_lock);
911 		}
912 	}
913 
914 out:
915 	if (!res->f6i) {
916 		res->f6i = net->ipv6.fib6_null_entry;
917 		res->nh = res->f6i->fib6_nh;
918 		res->fib6_flags = res->f6i->fib6_flags;
919 		res->fib6_type = res->f6i->fib6_type;
920 	}
921 }
922 
rt6_is_gw_or_nonexthop(const struct fib6_result * res)923 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
924 {
925 	return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
926 	       res->nh->fib_nh_gw_family;
927 }
928 
929 #ifdef CONFIG_IPV6_ROUTE_INFO
rt6_route_rcv(struct net_device * dev,u8 * opt,int len,const struct in6_addr * gwaddr)930 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
931 		  const struct in6_addr *gwaddr)
932 {
933 	struct net *net = dev_net(dev);
934 	struct route_info *rinfo = (struct route_info *) opt;
935 	struct in6_addr prefix_buf, *prefix;
936 	struct fib6_table *table;
937 	unsigned int pref;
938 	unsigned long lifetime;
939 	struct fib6_info *rt;
940 
941 	if (len < sizeof(struct route_info)) {
942 		return -EINVAL;
943 	}
944 
945 	/* Sanity check for prefix_len and length */
946 	if (rinfo->length > 3) {
947 		return -EINVAL;
948 	} else if (rinfo->prefix_len > 128) {
949 		return -EINVAL;
950 	} else if (rinfo->prefix_len > 64) {
951 		if (rinfo->length < 2) {
952 			return -EINVAL;
953 		}
954 	} else if (rinfo->prefix_len > 0) {
955 		if (rinfo->length < 1) {
956 			return -EINVAL;
957 		}
958 	}
959 
960 	pref = rinfo->route_pref;
961 	if (pref == ICMPV6_ROUTER_PREF_INVALID)
962 		return -EINVAL;
963 
964 	lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
965 
966 	if (rinfo->length == 3)
967 		prefix = (struct in6_addr *)rinfo->prefix;
968 	else {
969 		/* this function is safe */
970 		ipv6_addr_prefix(&prefix_buf,
971 				 (struct in6_addr *)rinfo->prefix,
972 				 rinfo->prefix_len);
973 		prefix = &prefix_buf;
974 	}
975 
976 	if (rinfo->prefix_len == 0)
977 		rt = rt6_get_dflt_router(net, gwaddr, dev);
978 	else
979 		rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
980 					gwaddr, dev);
981 
982 	if (rt && !lifetime) {
983 		ip6_del_rt(net, rt, false);
984 		rt = NULL;
985 	}
986 
987 	if (!rt && lifetime)
988 		rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
989 					dev, pref);
990 	else if (rt)
991 		rt->fib6_flags = RTF_ROUTEINFO |
992 				 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
993 
994 	if (rt) {
995 		table = rt->fib6_table;
996 		spin_lock_bh(&table->tb6_lock);
997 
998 		if (!addrconf_finite_timeout(lifetime)) {
999 			fib6_clean_expires(rt);
1000 			fib6_remove_gc_list(rt);
1001 		} else {
1002 			fib6_set_expires(rt, jiffies + HZ * lifetime);
1003 			fib6_add_gc_list(rt);
1004 		}
1005 
1006 		spin_unlock_bh(&table->tb6_lock);
1007 
1008 		fib6_info_release(rt);
1009 	}
1010 	return 0;
1011 }
1012 #endif
1013 
1014 /*
1015  *	Misc support functions
1016  */
1017 
1018 /* called with rcu_lock held */
ip6_rt_get_dev_rcu(const struct fib6_result * res)1019 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1020 {
1021 	struct net_device *dev = res->nh->fib_nh_dev;
1022 
1023 	if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1024 		/* for copies of local routes, dst->dev needs to be the
1025 		 * device if it is a master device, the master device if
1026 		 * device is enslaved, and the loopback as the default
1027 		 */
1028 		if (netif_is_l3_slave(dev) &&
1029 		    !rt6_need_strict(&res->f6i->fib6_dst.addr))
1030 			dev = l3mdev_master_dev_rcu(dev);
1031 		else if (!netif_is_l3_master(dev))
1032 			dev = dev_net(dev)->loopback_dev;
1033 		/* last case is netif_is_l3_master(dev) is true in which
1034 		 * case we want dev returned to be dev
1035 		 */
1036 	}
1037 
1038 	return dev;
1039 }
1040 
1041 static const int fib6_prop[RTN_MAX + 1] = {
1042 	[RTN_UNSPEC]	= 0,
1043 	[RTN_UNICAST]	= 0,
1044 	[RTN_LOCAL]	= 0,
1045 	[RTN_BROADCAST]	= 0,
1046 	[RTN_ANYCAST]	= 0,
1047 	[RTN_MULTICAST]	= 0,
1048 	[RTN_BLACKHOLE]	= -EINVAL,
1049 	[RTN_UNREACHABLE] = -EHOSTUNREACH,
1050 	[RTN_PROHIBIT]	= -EACCES,
1051 	[RTN_THROW]	= -EAGAIN,
1052 	[RTN_NAT]	= -EINVAL,
1053 	[RTN_XRESOLVE]	= -EINVAL,
1054 };
1055 
ip6_rt_type_to_error(u8 fib6_type)1056 static int ip6_rt_type_to_error(u8 fib6_type)
1057 {
1058 	return fib6_prop[fib6_type];
1059 }
1060 
fib6_info_dst_flags(struct fib6_info * rt)1061 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1062 {
1063 	unsigned short flags = 0;
1064 
1065 	if (rt->dst_nocount)
1066 		flags |= DST_NOCOUNT;
1067 	if (rt->dst_nopolicy)
1068 		flags |= DST_NOPOLICY;
1069 
1070 	return flags;
1071 }
1072 
ip6_rt_init_dst_reject(struct rt6_info * rt,u8 fib6_type)1073 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1074 {
1075 	rt->dst.error = ip6_rt_type_to_error(fib6_type);
1076 
1077 	switch (fib6_type) {
1078 	case RTN_BLACKHOLE:
1079 		rt->dst.output = dst_discard_out;
1080 		rt->dst.input = dst_discard;
1081 		break;
1082 	case RTN_PROHIBIT:
1083 		rt->dst.output = ip6_pkt_prohibit_out;
1084 		rt->dst.input = ip6_pkt_prohibit;
1085 		break;
1086 	case RTN_THROW:
1087 	case RTN_UNREACHABLE:
1088 	default:
1089 		rt->dst.output = ip6_pkt_discard_out;
1090 		rt->dst.input = ip6_pkt_discard;
1091 		break;
1092 	}
1093 }
1094 
ip6_rt_init_dst(struct rt6_info * rt,const struct fib6_result * res)1095 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1096 {
1097 	struct fib6_info *f6i = res->f6i;
1098 
1099 	if (res->fib6_flags & RTF_REJECT) {
1100 		ip6_rt_init_dst_reject(rt, res->fib6_type);
1101 		return;
1102 	}
1103 
1104 	rt->dst.error = 0;
1105 	rt->dst.output = ip6_output;
1106 
1107 	if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1108 		rt->dst.input = ip6_input;
1109 	} else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1110 		rt->dst.input = ip6_mc_input;
1111 	} else {
1112 		rt->dst.input = ip6_forward;
1113 	}
1114 
1115 	if (res->nh->fib_nh_lws) {
1116 		rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1117 		lwtunnel_set_redirect(&rt->dst);
1118 	}
1119 
1120 	rt->dst.lastuse = jiffies;
1121 }
1122 
1123 /* Caller must already hold reference to @from */
rt6_set_from(struct rt6_info * rt,struct fib6_info * from)1124 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1125 {
1126 	rt->rt6i_flags &= ~RTF_EXPIRES;
1127 	rcu_assign_pointer(rt->from, from);
1128 	ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1129 }
1130 
1131 /* Caller must already hold reference to f6i in result */
ip6_rt_copy_init(struct rt6_info * rt,const struct fib6_result * res)1132 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1133 {
1134 	const struct fib6_nh *nh = res->nh;
1135 	const struct net_device *dev = nh->fib_nh_dev;
1136 	struct fib6_info *f6i = res->f6i;
1137 
1138 	ip6_rt_init_dst(rt, res);
1139 
1140 	rt->rt6i_dst = f6i->fib6_dst;
1141 	rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1142 	rt->rt6i_flags = res->fib6_flags;
1143 	if (nh->fib_nh_gw_family) {
1144 		rt->rt6i_gateway = nh->fib_nh_gw6;
1145 		rt->rt6i_flags |= RTF_GATEWAY;
1146 	}
1147 	rt6_set_from(rt, f6i);
1148 #ifdef CONFIG_IPV6_SUBTREES
1149 	rt->rt6i_src = f6i->fib6_src;
1150 #endif
1151 }
1152 
fib6_backtrack(struct fib6_node * fn,struct in6_addr * saddr)1153 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1154 					struct in6_addr *saddr)
1155 {
1156 	struct fib6_node *pn, *sn;
1157 	while (1) {
1158 		if (fn->fn_flags & RTN_TL_ROOT)
1159 			return NULL;
1160 		pn = rcu_dereference(fn->parent);
1161 		sn = FIB6_SUBTREE(pn);
1162 		if (sn && sn != fn)
1163 			fn = fib6_node_lookup(sn, NULL, saddr);
1164 		else
1165 			fn = pn;
1166 		if (fn->fn_flags & RTN_RTINFO)
1167 			return fn;
1168 	}
1169 }
1170 
ip6_hold_safe(struct net * net,struct rt6_info ** prt)1171 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1172 {
1173 	struct rt6_info *rt = *prt;
1174 
1175 	if (dst_hold_safe(&rt->dst))
1176 		return true;
1177 	if (net) {
1178 		rt = net->ipv6.ip6_null_entry;
1179 		dst_hold(&rt->dst);
1180 	} else {
1181 		rt = NULL;
1182 	}
1183 	*prt = rt;
1184 	return false;
1185 }
1186 
1187 /* called with rcu_lock held */
ip6_create_rt_rcu(const struct fib6_result * res)1188 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1189 {
1190 	struct net_device *dev = res->nh->fib_nh_dev;
1191 	struct fib6_info *f6i = res->f6i;
1192 	unsigned short flags;
1193 	struct rt6_info *nrt;
1194 
1195 	if (!fib6_info_hold_safe(f6i))
1196 		goto fallback;
1197 
1198 	flags = fib6_info_dst_flags(f6i);
1199 	nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1200 	if (!nrt) {
1201 		fib6_info_release(f6i);
1202 		goto fallback;
1203 	}
1204 
1205 	ip6_rt_copy_init(nrt, res);
1206 	return nrt;
1207 
1208 fallback:
1209 	nrt = dev_net(dev)->ipv6.ip6_null_entry;
1210 	dst_hold(&nrt->dst);
1211 	return nrt;
1212 }
1213 
ip6_pol_route_lookup(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)1214 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1215 					     struct fib6_table *table,
1216 					     struct flowi6 *fl6,
1217 					     const struct sk_buff *skb,
1218 					     int flags)
1219 {
1220 	struct fib6_result res = {};
1221 	struct fib6_node *fn;
1222 	struct rt6_info *rt;
1223 
1224 	rcu_read_lock();
1225 	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1226 restart:
1227 	res.f6i = rcu_dereference(fn->leaf);
1228 	if (!res.f6i)
1229 		res.f6i = net->ipv6.fib6_null_entry;
1230 	else
1231 		rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1232 				 flags);
1233 
1234 	if (res.f6i == net->ipv6.fib6_null_entry) {
1235 		fn = fib6_backtrack(fn, &fl6->saddr);
1236 		if (fn)
1237 			goto restart;
1238 
1239 		rt = net->ipv6.ip6_null_entry;
1240 		dst_hold(&rt->dst);
1241 		goto out;
1242 	} else if (res.fib6_flags & RTF_REJECT) {
1243 		goto do_create;
1244 	}
1245 
1246 	fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1247 			 fl6->flowi6_oif != 0, skb, flags);
1248 
1249 	/* Search through exception table */
1250 	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1251 	if (rt) {
1252 		if (ip6_hold_safe(net, &rt))
1253 			dst_use_noref(&rt->dst, jiffies);
1254 	} else {
1255 do_create:
1256 		rt = ip6_create_rt_rcu(&res);
1257 	}
1258 
1259 out:
1260 	trace_fib6_table_lookup(net, &res, table, fl6);
1261 
1262 	rcu_read_unlock();
1263 
1264 	return rt;
1265 }
1266 
ip6_route_lookup(struct net * net,struct flowi6 * fl6,const struct sk_buff * skb,int flags)1267 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1268 				   const struct sk_buff *skb, int flags)
1269 {
1270 	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1271 }
1272 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1273 
rt6_lookup(struct net * net,const struct in6_addr * daddr,const struct in6_addr * saddr,int oif,const struct sk_buff * skb,int strict)1274 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1275 			    const struct in6_addr *saddr, int oif,
1276 			    const struct sk_buff *skb, int strict)
1277 {
1278 	struct flowi6 fl6 = {
1279 		.flowi6_oif = oif,
1280 		.daddr = *daddr,
1281 	};
1282 	struct dst_entry *dst;
1283 	int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1284 
1285 	if (saddr) {
1286 		memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1287 		flags |= RT6_LOOKUP_F_HAS_SADDR;
1288 	}
1289 
1290 	dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1291 	if (dst->error == 0)
1292 		return dst_rt6_info(dst);
1293 
1294 	dst_release(dst);
1295 
1296 	return NULL;
1297 }
1298 EXPORT_SYMBOL(rt6_lookup);
1299 
1300 /* ip6_ins_rt is called with FREE table->tb6_lock.
1301  * It takes new route entry, the addition fails by any reason the
1302  * route is released.
1303  * Caller must hold dst before calling it.
1304  */
1305 
__ip6_ins_rt(struct fib6_info * rt,struct nl_info * info,struct netlink_ext_ack * extack)1306 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1307 			struct netlink_ext_ack *extack)
1308 {
1309 	int err;
1310 	struct fib6_table *table;
1311 
1312 	table = rt->fib6_table;
1313 	spin_lock_bh(&table->tb6_lock);
1314 	err = fib6_add(&table->tb6_root, rt, info, extack);
1315 	spin_unlock_bh(&table->tb6_lock);
1316 
1317 	return err;
1318 }
1319 
ip6_ins_rt(struct net * net,struct fib6_info * rt)1320 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1321 {
1322 	struct nl_info info = {	.nl_net = net, };
1323 
1324 	return __ip6_ins_rt(rt, &info, NULL);
1325 }
1326 
ip6_rt_cache_alloc(const struct fib6_result * res,const struct in6_addr * daddr,const struct in6_addr * saddr)1327 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1328 					   const struct in6_addr *daddr,
1329 					   const struct in6_addr *saddr)
1330 {
1331 	struct fib6_info *f6i = res->f6i;
1332 	struct net_device *dev;
1333 	struct rt6_info *rt;
1334 
1335 	/*
1336 	 *	Clone the route.
1337 	 */
1338 
1339 	if (!fib6_info_hold_safe(f6i))
1340 		return NULL;
1341 
1342 	dev = ip6_rt_get_dev_rcu(res);
1343 	rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1344 	if (!rt) {
1345 		fib6_info_release(f6i);
1346 		return NULL;
1347 	}
1348 
1349 	ip6_rt_copy_init(rt, res);
1350 	rt->rt6i_flags |= RTF_CACHE;
1351 	rt->rt6i_dst.addr = *daddr;
1352 	rt->rt6i_dst.plen = 128;
1353 
1354 	if (!rt6_is_gw_or_nonexthop(res)) {
1355 		if (f6i->fib6_dst.plen != 128 &&
1356 		    ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1357 			rt->rt6i_flags |= RTF_ANYCAST;
1358 #ifdef CONFIG_IPV6_SUBTREES
1359 		if (rt->rt6i_src.plen && saddr) {
1360 			rt->rt6i_src.addr = *saddr;
1361 			rt->rt6i_src.plen = 128;
1362 		}
1363 #endif
1364 	}
1365 
1366 	return rt;
1367 }
1368 
ip6_rt_pcpu_alloc(const struct fib6_result * res)1369 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1370 {
1371 	struct fib6_info *f6i = res->f6i;
1372 	unsigned short flags = fib6_info_dst_flags(f6i);
1373 	struct net_device *dev;
1374 	struct rt6_info *pcpu_rt;
1375 
1376 	if (!fib6_info_hold_safe(f6i))
1377 		return NULL;
1378 
1379 	rcu_read_lock();
1380 	dev = ip6_rt_get_dev_rcu(res);
1381 	pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1382 	rcu_read_unlock();
1383 	if (!pcpu_rt) {
1384 		fib6_info_release(f6i);
1385 		return NULL;
1386 	}
1387 	ip6_rt_copy_init(pcpu_rt, res);
1388 	pcpu_rt->rt6i_flags |= RTF_PCPU;
1389 
1390 	if (f6i->nh)
1391 		pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1392 
1393 	return pcpu_rt;
1394 }
1395 
rt6_is_valid(const struct rt6_info * rt6)1396 static bool rt6_is_valid(const struct rt6_info *rt6)
1397 {
1398 	return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1399 }
1400 
1401 /* It should be called with rcu_read_lock() acquired */
rt6_get_pcpu_route(const struct fib6_result * res)1402 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1403 {
1404 	struct rt6_info *pcpu_rt;
1405 
1406 	pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1407 
1408 	if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1409 		struct rt6_info *prev, **p;
1410 
1411 		p = this_cpu_ptr(res->nh->rt6i_pcpu);
1412 		/* Paired with READ_ONCE() in __fib6_drop_pcpu_from() */
1413 		prev = xchg(p, NULL);
1414 		if (prev) {
1415 			dst_dev_put(&prev->dst);
1416 			dst_release(&prev->dst);
1417 		}
1418 
1419 		pcpu_rt = NULL;
1420 	}
1421 
1422 	return pcpu_rt;
1423 }
1424 
rt6_make_pcpu_route(struct net * net,const struct fib6_result * res)1425 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1426 					    const struct fib6_result *res)
1427 {
1428 	struct rt6_info *pcpu_rt, *prev, **p;
1429 
1430 	pcpu_rt = ip6_rt_pcpu_alloc(res);
1431 	if (!pcpu_rt)
1432 		return NULL;
1433 
1434 	p = this_cpu_ptr(res->nh->rt6i_pcpu);
1435 	prev = cmpxchg(p, NULL, pcpu_rt);
1436 	BUG_ON(prev);
1437 
1438 	if (res->f6i->fib6_destroying) {
1439 		struct fib6_info *from;
1440 
1441 		from = unrcu_pointer(xchg(&pcpu_rt->from, NULL));
1442 		fib6_info_release(from);
1443 	}
1444 
1445 	return pcpu_rt;
1446 }
1447 
1448 /* exception hash table implementation
1449  */
1450 static DEFINE_SPINLOCK(rt6_exception_lock);
1451 
1452 /* Remove rt6_ex from hash table and free the memory
1453  * Caller must hold rt6_exception_lock
1454  */
rt6_remove_exception(struct rt6_exception_bucket * bucket,struct rt6_exception * rt6_ex)1455 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1456 				 struct rt6_exception *rt6_ex)
1457 {
1458 	struct fib6_info *from;
1459 	struct net *net;
1460 
1461 	if (!bucket || !rt6_ex)
1462 		return;
1463 
1464 	net = dev_net(rt6_ex->rt6i->dst.dev);
1465 	net->ipv6.rt6_stats->fib_rt_cache--;
1466 
1467 	/* purge completely the exception to allow releasing the held resources:
1468 	 * some [sk] cache may keep the dst around for unlimited time
1469 	 */
1470 	from = unrcu_pointer(xchg(&rt6_ex->rt6i->from, NULL));
1471 	fib6_info_release(from);
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 			/* counteract the dst_release() in sk_dst_reset() */
2784 			dst_hold(dst);
2785 			sk_dst_reset(sk);
2786 
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 = dev_net(dev);
3200 
3201 	mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3202 
3203 	if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3204 		mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3205 
3206 	/*
3207 	 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3208 	 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3209 	 * IPV6_MAXPLEN is also valid and means: "any MSS,
3210 	 * rely only on pmtu discovery"
3211 	 */
3212 	if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3213 		mtu = IPV6_MAXPLEN;
3214 	return mtu;
3215 }
3216 
ip6_mtu(const struct dst_entry * dst)3217 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3218 {
3219 	return ip6_dst_mtu_maybe_forward(dst, false);
3220 }
3221 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3222 
3223 /* MTU selection:
3224  * 1. mtu on route is locked - use it
3225  * 2. mtu from nexthop exception
3226  * 3. mtu from egress device
3227  *
3228  * based on ip6_dst_mtu_forward and exception logic of
3229  * rt6_find_cached_rt; called with rcu_read_lock
3230  */
ip6_mtu_from_fib6(const struct fib6_result * res,const struct in6_addr * daddr,const struct in6_addr * saddr)3231 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3232 		      const struct in6_addr *daddr,
3233 		      const struct in6_addr *saddr)
3234 {
3235 	const struct fib6_nh *nh = res->nh;
3236 	struct fib6_info *f6i = res->f6i;
3237 	struct inet6_dev *idev;
3238 	struct rt6_info *rt;
3239 	u32 mtu = 0;
3240 
3241 	if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3242 		mtu = f6i->fib6_pmtu;
3243 		if (mtu)
3244 			goto out;
3245 	}
3246 
3247 	rt = rt6_find_cached_rt(res, daddr, saddr);
3248 	if (unlikely(rt)) {
3249 		mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3250 	} else {
3251 		struct net_device *dev = nh->fib_nh_dev;
3252 
3253 		mtu = IPV6_MIN_MTU;
3254 		idev = __in6_dev_get(dev);
3255 		if (idev)
3256 			mtu = max_t(u32, mtu, READ_ONCE(idev->cnf.mtu6));
3257 	}
3258 
3259 	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3260 out:
3261 	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3262 }
3263 
icmp6_dst_alloc(struct net_device * dev,struct flowi6 * fl6)3264 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3265 				  struct flowi6 *fl6)
3266 {
3267 	struct dst_entry *dst;
3268 	struct rt6_info *rt;
3269 	struct inet6_dev *idev = in6_dev_get(dev);
3270 	struct net *net = dev_net(dev);
3271 
3272 	if (unlikely(!idev))
3273 		return ERR_PTR(-ENODEV);
3274 
3275 	rt = ip6_dst_alloc(net, dev, 0);
3276 	if (unlikely(!rt)) {
3277 		in6_dev_put(idev);
3278 		dst = ERR_PTR(-ENOMEM);
3279 		goto out;
3280 	}
3281 
3282 	rt->dst.input = ip6_input;
3283 	rt->dst.output  = ip6_output;
3284 	rt->rt6i_gateway  = fl6->daddr;
3285 	rt->rt6i_dst.addr = fl6->daddr;
3286 	rt->rt6i_dst.plen = 128;
3287 	rt->rt6i_idev     = idev;
3288 	dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3289 
3290 	/* Add this dst into uncached_list so that rt6_disable_ip() can
3291 	 * do proper release of the net_device
3292 	 */
3293 	rt6_uncached_list_add(rt);
3294 
3295 	dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3296 
3297 out:
3298 	return dst;
3299 }
3300 
ip6_dst_gc(struct dst_ops * ops)3301 static void ip6_dst_gc(struct dst_ops *ops)
3302 {
3303 	struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3304 	int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3305 	int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3306 	int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3307 	unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3308 	unsigned int val;
3309 	int entries;
3310 
3311 	if (time_after(rt_last_gc + rt_min_interval, jiffies))
3312 		goto out;
3313 
3314 	fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3315 	entries = dst_entries_get_slow(ops);
3316 	if (entries < ops->gc_thresh)
3317 		atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3318 out:
3319 	val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3320 	atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3321 }
3322 
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)3323 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3324 			       const struct in6_addr *gw_addr, u32 tbid,
3325 			       int flags, struct fib6_result *res)
3326 {
3327 	struct flowi6 fl6 = {
3328 		.flowi6_oif = cfg->fc_ifindex,
3329 		.daddr = *gw_addr,
3330 		.saddr = cfg->fc_prefsrc,
3331 	};
3332 	struct fib6_table *table;
3333 	int err;
3334 
3335 	table = fib6_get_table(net, tbid);
3336 	if (!table)
3337 		return -EINVAL;
3338 
3339 	if (!ipv6_addr_any(&cfg->fc_prefsrc))
3340 		flags |= RT6_LOOKUP_F_HAS_SADDR;
3341 
3342 	flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3343 
3344 	err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3345 	if (!err && res->f6i != net->ipv6.fib6_null_entry)
3346 		fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3347 				 cfg->fc_ifindex != 0, NULL, flags);
3348 
3349 	return err;
3350 }
3351 
ip6_route_check_nh_onlink(struct net * net,struct fib6_config * cfg,const struct net_device * dev,struct netlink_ext_ack * extack)3352 static int ip6_route_check_nh_onlink(struct net *net,
3353 				     struct fib6_config *cfg,
3354 				     const struct net_device *dev,
3355 				     struct netlink_ext_ack *extack)
3356 {
3357 	u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3358 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3359 	struct fib6_result res = {};
3360 	int err;
3361 
3362 	err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3363 	if (!err && !(res.fib6_flags & RTF_REJECT) &&
3364 	    /* ignore match if it is the default route */
3365 	    !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3366 	    (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3367 		NL_SET_ERR_MSG(extack,
3368 			       "Nexthop has invalid gateway or device mismatch");
3369 		err = -EINVAL;
3370 	}
3371 
3372 	return err;
3373 }
3374 
ip6_route_check_nh(struct net * net,struct fib6_config * cfg,struct net_device ** _dev,netdevice_tracker * dev_tracker,struct inet6_dev ** idev)3375 static int ip6_route_check_nh(struct net *net,
3376 			      struct fib6_config *cfg,
3377 			      struct net_device **_dev,
3378 			      netdevice_tracker *dev_tracker,
3379 			      struct inet6_dev **idev)
3380 {
3381 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3382 	struct net_device *dev = _dev ? *_dev : NULL;
3383 	int flags = RT6_LOOKUP_F_IFACE;
3384 	struct fib6_result res = {};
3385 	int err = -EHOSTUNREACH;
3386 
3387 	if (cfg->fc_table) {
3388 		err = ip6_nh_lookup_table(net, cfg, gw_addr,
3389 					  cfg->fc_table, flags, &res);
3390 		/* gw_addr can not require a gateway or resolve to a reject
3391 		 * route. If a device is given, it must match the result.
3392 		 */
3393 		if (err || res.fib6_flags & RTF_REJECT ||
3394 		    res.nh->fib_nh_gw_family ||
3395 		    (dev && dev != res.nh->fib_nh_dev))
3396 			err = -EHOSTUNREACH;
3397 	}
3398 
3399 	if (err < 0) {
3400 		struct flowi6 fl6 = {
3401 			.flowi6_oif = cfg->fc_ifindex,
3402 			.daddr = *gw_addr,
3403 		};
3404 
3405 		err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3406 		if (err || res.fib6_flags & RTF_REJECT ||
3407 		    res.nh->fib_nh_gw_family)
3408 			err = -EHOSTUNREACH;
3409 
3410 		if (err)
3411 			return err;
3412 
3413 		fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3414 				 cfg->fc_ifindex != 0, NULL, flags);
3415 	}
3416 
3417 	err = 0;
3418 	if (dev) {
3419 		if (dev != res.nh->fib_nh_dev)
3420 			err = -EHOSTUNREACH;
3421 	} else {
3422 		*_dev = dev = res.nh->fib_nh_dev;
3423 		netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3424 		*idev = in6_dev_get(dev);
3425 	}
3426 
3427 	return err;
3428 }
3429 
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)3430 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3431 			   struct net_device **_dev,
3432 			   netdevice_tracker *dev_tracker,
3433 			   struct inet6_dev **idev,
3434 			   struct netlink_ext_ack *extack)
3435 {
3436 	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3437 	int gwa_type = ipv6_addr_type(gw_addr);
3438 	bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3439 	const struct net_device *dev = *_dev;
3440 	bool need_addr_check = !dev;
3441 	int err = -EINVAL;
3442 
3443 	/* if gw_addr is local we will fail to detect this in case
3444 	 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3445 	 * will return already-added prefix route via interface that
3446 	 * prefix route was assigned to, which might be non-loopback.
3447 	 */
3448 	if (dev &&
3449 	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3450 		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3451 		goto out;
3452 	}
3453 
3454 	if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3455 		/* IPv6 strictly inhibits using not link-local
3456 		 * addresses as nexthop address.
3457 		 * Otherwise, router will not able to send redirects.
3458 		 * It is very good, but in some (rare!) circumstances
3459 		 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3460 		 * some exceptions. --ANK
3461 		 * We allow IPv4-mapped nexthops to support RFC4798-type
3462 		 * addressing
3463 		 */
3464 		if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3465 			NL_SET_ERR_MSG(extack, "Invalid gateway address");
3466 			goto out;
3467 		}
3468 
3469 		rcu_read_lock();
3470 
3471 		if (cfg->fc_flags & RTNH_F_ONLINK)
3472 			err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3473 		else
3474 			err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3475 						 idev);
3476 
3477 		rcu_read_unlock();
3478 
3479 		if (err)
3480 			goto out;
3481 	}
3482 
3483 	/* reload in case device was changed */
3484 	dev = *_dev;
3485 
3486 	err = -EINVAL;
3487 	if (!dev) {
3488 		NL_SET_ERR_MSG(extack, "Egress device not specified");
3489 		goto out;
3490 	} else if (dev->flags & IFF_LOOPBACK) {
3491 		NL_SET_ERR_MSG(extack,
3492 			       "Egress device can not be loopback device for this route");
3493 		goto out;
3494 	}
3495 
3496 	/* if we did not check gw_addr above, do so now that the
3497 	 * egress device has been resolved.
3498 	 */
3499 	if (need_addr_check &&
3500 	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3501 		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3502 		goto out;
3503 	}
3504 
3505 	err = 0;
3506 out:
3507 	return err;
3508 }
3509 
fib6_is_reject(u32 flags,struct net_device * dev,int addr_type)3510 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3511 {
3512 	if ((flags & RTF_REJECT) ||
3513 	    (dev && (dev->flags & IFF_LOOPBACK) &&
3514 	     !(addr_type & IPV6_ADDR_LOOPBACK) &&
3515 	     !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3516 		return true;
3517 
3518 	return false;
3519 }
3520 
fib6_nh_init(struct net * net,struct fib6_nh * fib6_nh,struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3521 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3522 		 struct fib6_config *cfg, gfp_t gfp_flags,
3523 		 struct netlink_ext_ack *extack)
3524 {
3525 	netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
3526 	struct net_device *dev = NULL;
3527 	struct inet6_dev *idev = NULL;
3528 	int addr_type;
3529 	int err;
3530 
3531 	fib6_nh->fib_nh_family = AF_INET6;
3532 #ifdef CONFIG_IPV6_ROUTER_PREF
3533 	fib6_nh->last_probe = jiffies;
3534 #endif
3535 	if (cfg->fc_is_fdb) {
3536 		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3537 		fib6_nh->fib_nh_gw_family = AF_INET6;
3538 		return 0;
3539 	}
3540 
3541 	err = -ENODEV;
3542 	if (cfg->fc_ifindex) {
3543 		dev = netdev_get_by_index(net, cfg->fc_ifindex,
3544 					  dev_tracker, gfp_flags);
3545 		if (!dev)
3546 			goto out;
3547 		idev = in6_dev_get(dev);
3548 		if (!idev)
3549 			goto out;
3550 	}
3551 
3552 	if (cfg->fc_flags & RTNH_F_ONLINK) {
3553 		if (!dev) {
3554 			NL_SET_ERR_MSG(extack,
3555 				       "Nexthop device required for onlink");
3556 			goto out;
3557 		}
3558 
3559 		if (!(dev->flags & IFF_UP)) {
3560 			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3561 			err = -ENETDOWN;
3562 			goto out;
3563 		}
3564 
3565 		fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3566 	}
3567 
3568 	fib6_nh->fib_nh_weight = 1;
3569 
3570 	/* We cannot add true routes via loopback here,
3571 	 * they would result in kernel looping; promote them to reject routes
3572 	 */
3573 	addr_type = ipv6_addr_type(&cfg->fc_dst);
3574 	if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3575 		/* hold loopback dev/idev if we haven't done so. */
3576 		if (dev != net->loopback_dev) {
3577 			if (dev) {
3578 				netdev_put(dev, dev_tracker);
3579 				in6_dev_put(idev);
3580 			}
3581 			dev = net->loopback_dev;
3582 			netdev_hold(dev, dev_tracker, gfp_flags);
3583 			idev = in6_dev_get(dev);
3584 			if (!idev) {
3585 				err = -ENODEV;
3586 				goto out;
3587 			}
3588 		}
3589 		goto pcpu_alloc;
3590 	}
3591 
3592 	if (cfg->fc_flags & RTF_GATEWAY) {
3593 		err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
3594 				      &idev, extack);
3595 		if (err)
3596 			goto out;
3597 
3598 		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3599 		fib6_nh->fib_nh_gw_family = AF_INET6;
3600 	}
3601 
3602 	err = -ENODEV;
3603 	if (!dev)
3604 		goto out;
3605 
3606 	if (!idev || idev->cnf.disable_ipv6) {
3607 		NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3608 		err = -EACCES;
3609 		goto out;
3610 	}
3611 
3612 	if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3613 		NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3614 		err = -ENETDOWN;
3615 		goto out;
3616 	}
3617 
3618 	if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3619 	    !netif_carrier_ok(dev))
3620 		fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3621 
3622 	err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3623 				 cfg->fc_encap_type, cfg, gfp_flags, extack);
3624 	if (err)
3625 		goto out;
3626 
3627 pcpu_alloc:
3628 	fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3629 	if (!fib6_nh->rt6i_pcpu) {
3630 		err = -ENOMEM;
3631 		goto out;
3632 	}
3633 
3634 	fib6_nh->fib_nh_dev = dev;
3635 	fib6_nh->fib_nh_oif = dev->ifindex;
3636 	err = 0;
3637 out:
3638 	if (idev)
3639 		in6_dev_put(idev);
3640 
3641 	if (err) {
3642 		lwtstate_put(fib6_nh->fib_nh_lws);
3643 		fib6_nh->fib_nh_lws = NULL;
3644 		netdev_put(dev, dev_tracker);
3645 	}
3646 
3647 	return err;
3648 }
3649 
fib6_nh_release(struct fib6_nh * fib6_nh)3650 void fib6_nh_release(struct fib6_nh *fib6_nh)
3651 {
3652 	struct rt6_exception_bucket *bucket;
3653 
3654 	rcu_read_lock();
3655 
3656 	fib6_nh_flush_exceptions(fib6_nh, NULL);
3657 	bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3658 	if (bucket) {
3659 		rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3660 		kfree(bucket);
3661 	}
3662 
3663 	rcu_read_unlock();
3664 
3665 	fib6_nh_release_dsts(fib6_nh);
3666 	free_percpu(fib6_nh->rt6i_pcpu);
3667 
3668 	fib_nh_common_release(&fib6_nh->nh_common);
3669 }
3670 
fib6_nh_release_dsts(struct fib6_nh * fib6_nh)3671 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3672 {
3673 	int cpu;
3674 
3675 	if (!fib6_nh->rt6i_pcpu)
3676 		return;
3677 
3678 	for_each_possible_cpu(cpu) {
3679 		struct rt6_info *pcpu_rt, **ppcpu_rt;
3680 
3681 		ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3682 		pcpu_rt = xchg(ppcpu_rt, NULL);
3683 		if (pcpu_rt) {
3684 			dst_dev_put(&pcpu_rt->dst);
3685 			dst_release(&pcpu_rt->dst);
3686 		}
3687 	}
3688 }
3689 
ip6_route_info_create(struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3690 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3691 					      gfp_t gfp_flags,
3692 					      struct netlink_ext_ack *extack)
3693 {
3694 	struct net *net = cfg->fc_nlinfo.nl_net;
3695 	struct fib6_info *rt = NULL;
3696 	struct nexthop *nh = NULL;
3697 	struct fib6_table *table;
3698 	struct fib6_nh *fib6_nh;
3699 	int err = -EINVAL;
3700 	int addr_type;
3701 
3702 	/* RTF_PCPU is an internal flag; can not be set by userspace */
3703 	if (cfg->fc_flags & RTF_PCPU) {
3704 		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3705 		goto out;
3706 	}
3707 
3708 	/* RTF_CACHE is an internal flag; can not be set by userspace */
3709 	if (cfg->fc_flags & RTF_CACHE) {
3710 		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3711 		goto out;
3712 	}
3713 
3714 	if (cfg->fc_type > RTN_MAX) {
3715 		NL_SET_ERR_MSG(extack, "Invalid route type");
3716 		goto out;
3717 	}
3718 
3719 	if (cfg->fc_dst_len > 128) {
3720 		NL_SET_ERR_MSG(extack, "Invalid prefix length");
3721 		goto out;
3722 	}
3723 	if (cfg->fc_src_len > 128) {
3724 		NL_SET_ERR_MSG(extack, "Invalid source address length");
3725 		goto out;
3726 	}
3727 #ifndef CONFIG_IPV6_SUBTREES
3728 	if (cfg->fc_src_len) {
3729 		NL_SET_ERR_MSG(extack,
3730 			       "Specifying source address requires IPV6_SUBTREES to be enabled");
3731 		goto out;
3732 	}
3733 #endif
3734 	if (cfg->fc_nh_id) {
3735 		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3736 		if (!nh) {
3737 			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3738 			goto out;
3739 		}
3740 		err = fib6_check_nexthop(nh, cfg, extack);
3741 		if (err)
3742 			goto out;
3743 	}
3744 
3745 	err = -ENOBUFS;
3746 	if (cfg->fc_nlinfo.nlh &&
3747 	    !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3748 		table = fib6_get_table(net, cfg->fc_table);
3749 		if (!table) {
3750 			pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3751 			table = fib6_new_table(net, cfg->fc_table);
3752 		}
3753 	} else {
3754 		table = fib6_new_table(net, cfg->fc_table);
3755 	}
3756 
3757 	if (!table)
3758 		goto out;
3759 
3760 	err = -ENOMEM;
3761 	rt = fib6_info_alloc(gfp_flags, !nh);
3762 	if (!rt)
3763 		goto out;
3764 
3765 	rt->fib6_metrics = ip_fib_metrics_init(cfg->fc_mx, cfg->fc_mx_len,
3766 					       extack);
3767 	if (IS_ERR(rt->fib6_metrics)) {
3768 		err = PTR_ERR(rt->fib6_metrics);
3769 		/* Do not leave garbage there. */
3770 		rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3771 		goto out_free;
3772 	}
3773 
3774 	if (cfg->fc_flags & RTF_ADDRCONF)
3775 		rt->dst_nocount = true;
3776 
3777 	if (cfg->fc_flags & RTF_EXPIRES)
3778 		fib6_set_expires(rt, jiffies +
3779 				clock_t_to_jiffies(cfg->fc_expires));
3780 
3781 	if (cfg->fc_protocol == RTPROT_UNSPEC)
3782 		cfg->fc_protocol = RTPROT_BOOT;
3783 	rt->fib6_protocol = cfg->fc_protocol;
3784 
3785 	rt->fib6_table = table;
3786 	rt->fib6_metric = cfg->fc_metric;
3787 	rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3788 	rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3789 
3790 	ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3791 	rt->fib6_dst.plen = cfg->fc_dst_len;
3792 
3793 #ifdef CONFIG_IPV6_SUBTREES
3794 	ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3795 	rt->fib6_src.plen = cfg->fc_src_len;
3796 #endif
3797 	if (nh) {
3798 		if (rt->fib6_src.plen) {
3799 			NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3800 			goto out_free;
3801 		}
3802 		if (!nexthop_get(nh)) {
3803 			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3804 			goto out_free;
3805 		}
3806 		rt->nh = nh;
3807 		fib6_nh = nexthop_fib6_nh(rt->nh);
3808 	} else {
3809 		err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3810 		if (err)
3811 			goto out;
3812 
3813 		fib6_nh = rt->fib6_nh;
3814 
3815 		/* We cannot add true routes via loopback here, they would
3816 		 * result in kernel looping; promote them to reject routes
3817 		 */
3818 		addr_type = ipv6_addr_type(&cfg->fc_dst);
3819 		if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3820 				   addr_type))
3821 			rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3822 	}
3823 
3824 	if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3825 		struct net_device *dev = fib6_nh->fib_nh_dev;
3826 
3827 		if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3828 			NL_SET_ERR_MSG(extack, "Invalid source address");
3829 			err = -EINVAL;
3830 			goto out;
3831 		}
3832 		rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3833 		rt->fib6_prefsrc.plen = 128;
3834 	} else
3835 		rt->fib6_prefsrc.plen = 0;
3836 
3837 	return rt;
3838 out:
3839 	fib6_info_release(rt);
3840 	return ERR_PTR(err);
3841 out_free:
3842 	ip_fib_metrics_put(rt->fib6_metrics);
3843 	kfree(rt);
3844 	return ERR_PTR(err);
3845 }
3846 
ip6_route_add(struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3847 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3848 		  struct netlink_ext_ack *extack)
3849 {
3850 	struct fib6_info *rt;
3851 	int err;
3852 
3853 	rt = ip6_route_info_create(cfg, gfp_flags, extack);
3854 	if (IS_ERR(rt))
3855 		return PTR_ERR(rt);
3856 
3857 	err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3858 	fib6_info_release(rt);
3859 
3860 	return err;
3861 }
3862 
__ip6_del_rt(struct fib6_info * rt,struct nl_info * info)3863 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3864 {
3865 	struct net *net = info->nl_net;
3866 	struct fib6_table *table;
3867 	int err;
3868 
3869 	if (rt == net->ipv6.fib6_null_entry) {
3870 		err = -ENOENT;
3871 		goto out;
3872 	}
3873 
3874 	table = rt->fib6_table;
3875 	spin_lock_bh(&table->tb6_lock);
3876 	err = fib6_del(rt, info);
3877 	spin_unlock_bh(&table->tb6_lock);
3878 
3879 out:
3880 	fib6_info_release(rt);
3881 	return err;
3882 }
3883 
ip6_del_rt(struct net * net,struct fib6_info * rt,bool skip_notify)3884 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3885 {
3886 	struct nl_info info = {
3887 		.nl_net = net,
3888 		.skip_notify = skip_notify
3889 	};
3890 
3891 	return __ip6_del_rt(rt, &info);
3892 }
3893 
__ip6_del_rt_siblings(struct fib6_info * rt,struct fib6_config * cfg)3894 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3895 {
3896 	struct nl_info *info = &cfg->fc_nlinfo;
3897 	struct net *net = info->nl_net;
3898 	struct sk_buff *skb = NULL;
3899 	struct fib6_table *table;
3900 	int err = -ENOENT;
3901 
3902 	if (rt == net->ipv6.fib6_null_entry)
3903 		goto out_put;
3904 	table = rt->fib6_table;
3905 	spin_lock_bh(&table->tb6_lock);
3906 
3907 	if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3908 		struct fib6_info *sibling, *next_sibling;
3909 		struct fib6_node *fn;
3910 
3911 		/* prefer to send a single notification with all hops */
3912 		skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3913 		if (skb) {
3914 			u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3915 
3916 			if (rt6_fill_node(net, skb, rt, NULL,
3917 					  NULL, NULL, 0, RTM_DELROUTE,
3918 					  info->portid, seq, 0) < 0) {
3919 				kfree_skb(skb);
3920 				skb = NULL;
3921 			} else
3922 				info->skip_notify = 1;
3923 		}
3924 
3925 		/* 'rt' points to the first sibling route. If it is not the
3926 		 * leaf, then we do not need to send a notification. Otherwise,
3927 		 * we need to check if the last sibling has a next route or not
3928 		 * and emit a replace or delete notification, respectively.
3929 		 */
3930 		info->skip_notify_kernel = 1;
3931 		fn = rcu_dereference_protected(rt->fib6_node,
3932 					    lockdep_is_held(&table->tb6_lock));
3933 		if (rcu_access_pointer(fn->leaf) == rt) {
3934 			struct fib6_info *last_sibling, *replace_rt;
3935 
3936 			last_sibling = list_last_entry(&rt->fib6_siblings,
3937 						       struct fib6_info,
3938 						       fib6_siblings);
3939 			replace_rt = rcu_dereference_protected(
3940 					    last_sibling->fib6_next,
3941 					    lockdep_is_held(&table->tb6_lock));
3942 			if (replace_rt)
3943 				call_fib6_entry_notifiers_replace(net,
3944 								  replace_rt);
3945 			else
3946 				call_fib6_multipath_entry_notifiers(net,
3947 						       FIB_EVENT_ENTRY_DEL,
3948 						       rt, rt->fib6_nsiblings,
3949 						       NULL);
3950 		}
3951 		list_for_each_entry_safe(sibling, next_sibling,
3952 					 &rt->fib6_siblings,
3953 					 fib6_siblings) {
3954 			err = fib6_del(sibling, info);
3955 			if (err)
3956 				goto out_unlock;
3957 		}
3958 	}
3959 
3960 	err = fib6_del(rt, info);
3961 out_unlock:
3962 	spin_unlock_bh(&table->tb6_lock);
3963 out_put:
3964 	fib6_info_release(rt);
3965 
3966 	if (skb) {
3967 		rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3968 			    info->nlh, gfp_any());
3969 	}
3970 	return err;
3971 }
3972 
__ip6_del_cached_rt(struct rt6_info * rt,struct fib6_config * cfg)3973 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3974 {
3975 	int rc = -ESRCH;
3976 
3977 	if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3978 		goto out;
3979 
3980 	if (cfg->fc_flags & RTF_GATEWAY &&
3981 	    !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3982 		goto out;
3983 
3984 	rc = rt6_remove_exception_rt(rt);
3985 out:
3986 	return rc;
3987 }
3988 
ip6_del_cached_rt(struct fib6_config * cfg,struct fib6_info * rt,struct fib6_nh * nh)3989 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3990 			     struct fib6_nh *nh)
3991 {
3992 	struct fib6_result res = {
3993 		.f6i = rt,
3994 		.nh = nh,
3995 	};
3996 	struct rt6_info *rt_cache;
3997 
3998 	rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3999 	if (rt_cache)
4000 		return __ip6_del_cached_rt(rt_cache, cfg);
4001 
4002 	return 0;
4003 }
4004 
4005 struct fib6_nh_del_cached_rt_arg {
4006 	struct fib6_config *cfg;
4007 	struct fib6_info *f6i;
4008 };
4009 
fib6_nh_del_cached_rt(struct fib6_nh * nh,void * _arg)4010 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4011 {
4012 	struct fib6_nh_del_cached_rt_arg *arg = _arg;
4013 	int rc;
4014 
4015 	rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4016 	return rc != -ESRCH ? rc : 0;
4017 }
4018 
ip6_del_cached_rt_nh(struct fib6_config * cfg,struct fib6_info * f6i)4019 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4020 {
4021 	struct fib6_nh_del_cached_rt_arg arg = {
4022 		.cfg = cfg,
4023 		.f6i = f6i
4024 	};
4025 
4026 	return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4027 }
4028 
ip6_route_del(struct fib6_config * cfg,struct netlink_ext_ack * extack)4029 static int ip6_route_del(struct fib6_config *cfg,
4030 			 struct netlink_ext_ack *extack)
4031 {
4032 	struct fib6_table *table;
4033 	struct fib6_info *rt;
4034 	struct fib6_node *fn;
4035 	int err = -ESRCH;
4036 
4037 	table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4038 	if (!table) {
4039 		NL_SET_ERR_MSG(extack, "FIB table does not exist");
4040 		return err;
4041 	}
4042 
4043 	rcu_read_lock();
4044 
4045 	fn = fib6_locate(&table->tb6_root,
4046 			 &cfg->fc_dst, cfg->fc_dst_len,
4047 			 &cfg->fc_src, cfg->fc_src_len,
4048 			 !(cfg->fc_flags & RTF_CACHE));
4049 
4050 	if (fn) {
4051 		for_each_fib6_node_rt_rcu(fn) {
4052 			struct fib6_nh *nh;
4053 
4054 			if (rt->nh && cfg->fc_nh_id &&
4055 			    rt->nh->id != cfg->fc_nh_id)
4056 				continue;
4057 
4058 			if (cfg->fc_flags & RTF_CACHE) {
4059 				int rc = 0;
4060 
4061 				if (rt->nh) {
4062 					rc = ip6_del_cached_rt_nh(cfg, rt);
4063 				} else if (cfg->fc_nh_id) {
4064 					continue;
4065 				} else {
4066 					nh = rt->fib6_nh;
4067 					rc = ip6_del_cached_rt(cfg, rt, nh);
4068 				}
4069 				if (rc != -ESRCH) {
4070 					rcu_read_unlock();
4071 					return rc;
4072 				}
4073 				continue;
4074 			}
4075 
4076 			if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4077 				continue;
4078 			if (cfg->fc_protocol &&
4079 			    cfg->fc_protocol != rt->fib6_protocol)
4080 				continue;
4081 
4082 			if (rt->nh) {
4083 				if (!fib6_info_hold_safe(rt))
4084 					continue;
4085 				rcu_read_unlock();
4086 
4087 				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4088 			}
4089 			if (cfg->fc_nh_id)
4090 				continue;
4091 
4092 			nh = rt->fib6_nh;
4093 			if (cfg->fc_ifindex &&
4094 			    (!nh->fib_nh_dev ||
4095 			     nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4096 				continue;
4097 			if (cfg->fc_flags & RTF_GATEWAY &&
4098 			    !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4099 				continue;
4100 			if (!fib6_info_hold_safe(rt))
4101 				continue;
4102 			rcu_read_unlock();
4103 
4104 			/* if gateway was specified only delete the one hop */
4105 			if (cfg->fc_flags & RTF_GATEWAY)
4106 				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4107 
4108 			return __ip6_del_rt_siblings(rt, cfg);
4109 		}
4110 	}
4111 	rcu_read_unlock();
4112 
4113 	return err;
4114 }
4115 
rt6_do_redirect(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb)4116 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4117 {
4118 	struct netevent_redirect netevent;
4119 	struct rt6_info *rt, *nrt = NULL;
4120 	struct fib6_result res = {};
4121 	struct ndisc_options ndopts;
4122 	struct inet6_dev *in6_dev;
4123 	struct neighbour *neigh;
4124 	struct rd_msg *msg;
4125 	int optlen, on_link;
4126 	u8 *lladdr;
4127 
4128 	optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4129 	optlen -= sizeof(*msg);
4130 
4131 	if (optlen < 0) {
4132 		net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4133 		return;
4134 	}
4135 
4136 	msg = (struct rd_msg *)icmp6_hdr(skb);
4137 
4138 	if (ipv6_addr_is_multicast(&msg->dest)) {
4139 		net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4140 		return;
4141 	}
4142 
4143 	on_link = 0;
4144 	if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4145 		on_link = 1;
4146 	} else if (ipv6_addr_type(&msg->target) !=
4147 		   (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4148 		net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4149 		return;
4150 	}
4151 
4152 	in6_dev = __in6_dev_get(skb->dev);
4153 	if (!in6_dev)
4154 		return;
4155 	if (READ_ONCE(in6_dev->cnf.forwarding) ||
4156 	    !READ_ONCE(in6_dev->cnf.accept_redirects))
4157 		return;
4158 
4159 	/* RFC2461 8.1:
4160 	 *	The IP source address of the Redirect MUST be the same as the current
4161 	 *	first-hop router for the specified ICMP Destination Address.
4162 	 */
4163 
4164 	if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4165 		net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4166 		return;
4167 	}
4168 
4169 	lladdr = NULL;
4170 	if (ndopts.nd_opts_tgt_lladdr) {
4171 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4172 					     skb->dev);
4173 		if (!lladdr) {
4174 			net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4175 			return;
4176 		}
4177 	}
4178 
4179 	rt = dst_rt6_info(dst);
4180 	if (rt->rt6i_flags & RTF_REJECT) {
4181 		net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4182 		return;
4183 	}
4184 
4185 	/* Redirect received -> path was valid.
4186 	 * Look, redirects are sent only in response to data packets,
4187 	 * so that this nexthop apparently is reachable. --ANK
4188 	 */
4189 	dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4190 
4191 	neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4192 	if (!neigh)
4193 		return;
4194 
4195 	/*
4196 	 *	We have finally decided to accept it.
4197 	 */
4198 
4199 	ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4200 		     NEIGH_UPDATE_F_WEAK_OVERRIDE|
4201 		     NEIGH_UPDATE_F_OVERRIDE|
4202 		     (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4203 				     NEIGH_UPDATE_F_ISROUTER)),
4204 		     NDISC_REDIRECT, &ndopts);
4205 
4206 	rcu_read_lock();
4207 	res.f6i = rcu_dereference(rt->from);
4208 	if (!res.f6i)
4209 		goto out;
4210 
4211 	if (res.f6i->nh) {
4212 		struct fib6_nh_match_arg arg = {
4213 			.dev = dst->dev,
4214 			.gw = &rt->rt6i_gateway,
4215 		};
4216 
4217 		nexthop_for_each_fib6_nh(res.f6i->nh,
4218 					 fib6_nh_find_match, &arg);
4219 
4220 		/* fib6_info uses a nexthop that does not have fib6_nh
4221 		 * using the dst->dev. Should be impossible
4222 		 */
4223 		if (!arg.match)
4224 			goto out;
4225 		res.nh = arg.match;
4226 	} else {
4227 		res.nh = res.f6i->fib6_nh;
4228 	}
4229 
4230 	res.fib6_flags = res.f6i->fib6_flags;
4231 	res.fib6_type = res.f6i->fib6_type;
4232 	nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4233 	if (!nrt)
4234 		goto out;
4235 
4236 	nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4237 	if (on_link)
4238 		nrt->rt6i_flags &= ~RTF_GATEWAY;
4239 
4240 	nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4241 
4242 	/* rt6_insert_exception() will take care of duplicated exceptions */
4243 	if (rt6_insert_exception(nrt, &res)) {
4244 		dst_release_immediate(&nrt->dst);
4245 		goto out;
4246 	}
4247 
4248 	netevent.old = &rt->dst;
4249 	netevent.new = &nrt->dst;
4250 	netevent.daddr = &msg->dest;
4251 	netevent.neigh = neigh;
4252 	call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4253 
4254 out:
4255 	rcu_read_unlock();
4256 	neigh_release(neigh);
4257 }
4258 
4259 #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)4260 static struct fib6_info *rt6_get_route_info(struct net *net,
4261 					   const struct in6_addr *prefix, int prefixlen,
4262 					   const struct in6_addr *gwaddr,
4263 					   struct net_device *dev)
4264 {
4265 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4266 	int ifindex = dev->ifindex;
4267 	struct fib6_node *fn;
4268 	struct fib6_info *rt = NULL;
4269 	struct fib6_table *table;
4270 
4271 	table = fib6_get_table(net, tb_id);
4272 	if (!table)
4273 		return NULL;
4274 
4275 	rcu_read_lock();
4276 	fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4277 	if (!fn)
4278 		goto out;
4279 
4280 	for_each_fib6_node_rt_rcu(fn) {
4281 		/* these routes do not use nexthops */
4282 		if (rt->nh)
4283 			continue;
4284 		if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4285 			continue;
4286 		if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4287 		    !rt->fib6_nh->fib_nh_gw_family)
4288 			continue;
4289 		if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4290 			continue;
4291 		if (!fib6_info_hold_safe(rt))
4292 			continue;
4293 		break;
4294 	}
4295 out:
4296 	rcu_read_unlock();
4297 	return rt;
4298 }
4299 
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)4300 static struct fib6_info *rt6_add_route_info(struct net *net,
4301 					   const struct in6_addr *prefix, int prefixlen,
4302 					   const struct in6_addr *gwaddr,
4303 					   struct net_device *dev,
4304 					   unsigned int pref)
4305 {
4306 	struct fib6_config cfg = {
4307 		.fc_metric	= IP6_RT_PRIO_USER,
4308 		.fc_ifindex	= dev->ifindex,
4309 		.fc_dst_len	= prefixlen,
4310 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4311 				  RTF_UP | RTF_PREF(pref),
4312 		.fc_protocol = RTPROT_RA,
4313 		.fc_type = RTN_UNICAST,
4314 		.fc_nlinfo.portid = 0,
4315 		.fc_nlinfo.nlh = NULL,
4316 		.fc_nlinfo.nl_net = net,
4317 	};
4318 
4319 	cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4320 	cfg.fc_dst = *prefix;
4321 	cfg.fc_gateway = *gwaddr;
4322 
4323 	/* We should treat it as a default route if prefix length is 0. */
4324 	if (!prefixlen)
4325 		cfg.fc_flags |= RTF_DEFAULT;
4326 
4327 	ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4328 
4329 	return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4330 }
4331 #endif
4332 
rt6_get_dflt_router(struct net * net,const struct in6_addr * addr,struct net_device * dev)4333 struct fib6_info *rt6_get_dflt_router(struct net *net,
4334 				     const struct in6_addr *addr,
4335 				     struct net_device *dev)
4336 {
4337 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4338 	struct fib6_info *rt;
4339 	struct fib6_table *table;
4340 
4341 	table = fib6_get_table(net, tb_id);
4342 	if (!table)
4343 		return NULL;
4344 
4345 	rcu_read_lock();
4346 	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4347 		struct fib6_nh *nh;
4348 
4349 		/* RA routes do not use nexthops */
4350 		if (rt->nh)
4351 			continue;
4352 
4353 		nh = rt->fib6_nh;
4354 		if (dev == nh->fib_nh_dev &&
4355 		    ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4356 		    ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4357 			break;
4358 	}
4359 	if (rt && !fib6_info_hold_safe(rt))
4360 		rt = NULL;
4361 	rcu_read_unlock();
4362 	return rt;
4363 }
4364 
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)4365 struct fib6_info *rt6_add_dflt_router(struct net *net,
4366 				     const struct in6_addr *gwaddr,
4367 				     struct net_device *dev,
4368 				     unsigned int pref,
4369 				     u32 defrtr_usr_metric,
4370 				     int lifetime)
4371 {
4372 	struct fib6_config cfg = {
4373 		.fc_table	= l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4374 		.fc_metric	= defrtr_usr_metric,
4375 		.fc_ifindex	= dev->ifindex,
4376 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4377 				  RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4378 		.fc_protocol = RTPROT_RA,
4379 		.fc_type = RTN_UNICAST,
4380 		.fc_nlinfo.portid = 0,
4381 		.fc_nlinfo.nlh = NULL,
4382 		.fc_nlinfo.nl_net = net,
4383 		.fc_expires = jiffies_to_clock_t(lifetime * HZ),
4384 	};
4385 
4386 	cfg.fc_gateway = *gwaddr;
4387 
4388 	if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4389 		struct fib6_table *table;
4390 
4391 		table = fib6_get_table(dev_net(dev), cfg.fc_table);
4392 		if (table)
4393 			table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4394 	}
4395 
4396 	return rt6_get_dflt_router(net, gwaddr, dev);
4397 }
4398 
__rt6_purge_dflt_routers(struct net * net,struct fib6_table * table)4399 static void __rt6_purge_dflt_routers(struct net *net,
4400 				     struct fib6_table *table)
4401 {
4402 	struct fib6_info *rt;
4403 
4404 restart:
4405 	rcu_read_lock();
4406 	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4407 		struct net_device *dev = fib6_info_nh_dev(rt);
4408 		struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4409 
4410 		if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4411 		    (!idev || idev->cnf.accept_ra != 2) &&
4412 		    fib6_info_hold_safe(rt)) {
4413 			rcu_read_unlock();
4414 			ip6_del_rt(net, rt, false);
4415 			goto restart;
4416 		}
4417 	}
4418 	rcu_read_unlock();
4419 
4420 	table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4421 }
4422 
rt6_purge_dflt_routers(struct net * net)4423 void rt6_purge_dflt_routers(struct net *net)
4424 {
4425 	struct fib6_table *table;
4426 	struct hlist_head *head;
4427 	unsigned int h;
4428 
4429 	rcu_read_lock();
4430 
4431 	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4432 		head = &net->ipv6.fib_table_hash[h];
4433 		hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4434 			if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4435 				__rt6_purge_dflt_routers(net, table);
4436 		}
4437 	}
4438 
4439 	rcu_read_unlock();
4440 }
4441 
rtmsg_to_fib6_config(struct net * net,struct in6_rtmsg * rtmsg,struct fib6_config * cfg)4442 static void rtmsg_to_fib6_config(struct net *net,
4443 				 struct in6_rtmsg *rtmsg,
4444 				 struct fib6_config *cfg)
4445 {
4446 	*cfg = (struct fib6_config){
4447 		.fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4448 			 : RT6_TABLE_MAIN,
4449 		.fc_ifindex = rtmsg->rtmsg_ifindex,
4450 		.fc_metric = rtmsg->rtmsg_metric,
4451 		.fc_expires = rtmsg->rtmsg_info,
4452 		.fc_dst_len = rtmsg->rtmsg_dst_len,
4453 		.fc_src_len = rtmsg->rtmsg_src_len,
4454 		.fc_flags = rtmsg->rtmsg_flags,
4455 		.fc_type = rtmsg->rtmsg_type,
4456 
4457 		.fc_nlinfo.nl_net = net,
4458 
4459 		.fc_dst = rtmsg->rtmsg_dst,
4460 		.fc_src = rtmsg->rtmsg_src,
4461 		.fc_gateway = rtmsg->rtmsg_gateway,
4462 	};
4463 }
4464 
ipv6_route_ioctl(struct net * net,unsigned int cmd,struct in6_rtmsg * rtmsg)4465 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4466 {
4467 	struct fib6_config cfg;
4468 	int err;
4469 
4470 	if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4471 		return -EINVAL;
4472 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4473 		return -EPERM;
4474 
4475 	rtmsg_to_fib6_config(net, rtmsg, &cfg);
4476 
4477 	rtnl_lock();
4478 	switch (cmd) {
4479 	case SIOCADDRT:
4480 		/* Only do the default setting of fc_metric in route adding */
4481 		if (cfg.fc_metric == 0)
4482 			cfg.fc_metric = IP6_RT_PRIO_USER;
4483 		err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4484 		break;
4485 	case SIOCDELRT:
4486 		err = ip6_route_del(&cfg, NULL);
4487 		break;
4488 	}
4489 	rtnl_unlock();
4490 	return err;
4491 }
4492 
4493 /*
4494  *	Drop the packet on the floor
4495  */
4496 
ip6_pkt_drop(struct sk_buff * skb,u8 code,int ipstats_mib_noroutes)4497 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4498 {
4499 	struct dst_entry *dst = skb_dst(skb);
4500 	struct net *net = dev_net(dst->dev);
4501 	struct inet6_dev *idev;
4502 	SKB_DR(reason);
4503 	int type;
4504 
4505 	if (netif_is_l3_master(skb->dev) ||
4506 	    dst->dev == net->loopback_dev)
4507 		idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4508 	else
4509 		idev = ip6_dst_idev(dst);
4510 
4511 	switch (ipstats_mib_noroutes) {
4512 	case IPSTATS_MIB_INNOROUTES:
4513 		type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4514 		if (type == IPV6_ADDR_ANY) {
4515 			SKB_DR_SET(reason, IP_INADDRERRORS);
4516 			IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4517 			break;
4518 		}
4519 		SKB_DR_SET(reason, IP_INNOROUTES);
4520 		fallthrough;
4521 	case IPSTATS_MIB_OUTNOROUTES:
4522 		SKB_DR_OR(reason, IP_OUTNOROUTES);
4523 		IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4524 		break;
4525 	}
4526 
4527 	/* Start over by dropping the dst for l3mdev case */
4528 	if (netif_is_l3_master(skb->dev))
4529 		skb_dst_drop(skb);
4530 
4531 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4532 	kfree_skb_reason(skb, reason);
4533 	return 0;
4534 }
4535 
ip6_pkt_discard(struct sk_buff * skb)4536 static int ip6_pkt_discard(struct sk_buff *skb)
4537 {
4538 	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4539 }
4540 
ip6_pkt_discard_out(struct net * net,struct sock * sk,struct sk_buff * skb)4541 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4542 {
4543 	skb->dev = skb_dst(skb)->dev;
4544 	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4545 }
4546 
ip6_pkt_prohibit(struct sk_buff * skb)4547 static int ip6_pkt_prohibit(struct sk_buff *skb)
4548 {
4549 	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4550 }
4551 
ip6_pkt_prohibit_out(struct net * net,struct sock * sk,struct sk_buff * skb)4552 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4553 {
4554 	skb->dev = skb_dst(skb)->dev;
4555 	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4556 }
4557 
4558 /*
4559  *	Allocate a dst for local (unicast / anycast) address.
4560  */
4561 
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)4562 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4563 				     struct inet6_dev *idev,
4564 				     const struct in6_addr *addr,
4565 				     bool anycast, gfp_t gfp_flags,
4566 				     struct netlink_ext_ack *extack)
4567 {
4568 	struct fib6_config cfg = {
4569 		.fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4570 		.fc_ifindex = idev->dev->ifindex,
4571 		.fc_flags = RTF_UP | RTF_NONEXTHOP,
4572 		.fc_dst = *addr,
4573 		.fc_dst_len = 128,
4574 		.fc_protocol = RTPROT_KERNEL,
4575 		.fc_nlinfo.nl_net = net,
4576 		.fc_ignore_dev_down = true,
4577 	};
4578 	struct fib6_info *f6i;
4579 
4580 	if (anycast) {
4581 		cfg.fc_type = RTN_ANYCAST;
4582 		cfg.fc_flags |= RTF_ANYCAST;
4583 	} else {
4584 		cfg.fc_type = RTN_LOCAL;
4585 		cfg.fc_flags |= RTF_LOCAL;
4586 	}
4587 
4588 	f6i = ip6_route_info_create(&cfg, gfp_flags, extack);
4589 	if (!IS_ERR(f6i)) {
4590 		f6i->dst_nocount = true;
4591 
4592 		if (!anycast &&
4593 		    (READ_ONCE(net->ipv6.devconf_all->disable_policy) ||
4594 		     READ_ONCE(idev->cnf.disable_policy)))
4595 			f6i->dst_nopolicy = true;
4596 	}
4597 
4598 	return f6i;
4599 }
4600 
4601 /* remove deleted ip from prefsrc entries */
4602 struct arg_dev_net_ip {
4603 	struct net *net;
4604 	struct in6_addr *addr;
4605 };
4606 
fib6_remove_prefsrc(struct fib6_info * rt,void * arg)4607 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4608 {
4609 	struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4610 	struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4611 
4612 	if (!rt->nh &&
4613 	    rt != net->ipv6.fib6_null_entry &&
4614 	    ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) &&
4615 	    !ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) {
4616 		spin_lock_bh(&rt6_exception_lock);
4617 		/* remove prefsrc entry */
4618 		rt->fib6_prefsrc.plen = 0;
4619 		spin_unlock_bh(&rt6_exception_lock);
4620 	}
4621 	return 0;
4622 }
4623 
rt6_remove_prefsrc(struct inet6_ifaddr * ifp)4624 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4625 {
4626 	struct net *net = dev_net(ifp->idev->dev);
4627 	struct arg_dev_net_ip adni = {
4628 		.net = net,
4629 		.addr = &ifp->addr,
4630 	};
4631 	fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4632 }
4633 
4634 #define RTF_RA_ROUTER		(RTF_ADDRCONF | RTF_DEFAULT)
4635 
4636 /* Remove routers and update dst entries when gateway turn into host. */
fib6_clean_tohost(struct fib6_info * rt,void * arg)4637 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4638 {
4639 	struct in6_addr *gateway = (struct in6_addr *)arg;
4640 	struct fib6_nh *nh;
4641 
4642 	/* RA routes do not use nexthops */
4643 	if (rt->nh)
4644 		return 0;
4645 
4646 	nh = rt->fib6_nh;
4647 	if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4648 	    nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4649 		return -1;
4650 
4651 	/* Further clean up cached routes in exception table.
4652 	 * This is needed because cached route may have a different
4653 	 * gateway than its 'parent' in the case of an ip redirect.
4654 	 */
4655 	fib6_nh_exceptions_clean_tohost(nh, gateway);
4656 
4657 	return 0;
4658 }
4659 
rt6_clean_tohost(struct net * net,struct in6_addr * gateway)4660 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4661 {
4662 	fib6_clean_all(net, fib6_clean_tohost, gateway);
4663 }
4664 
4665 struct arg_netdev_event {
4666 	const struct net_device *dev;
4667 	union {
4668 		unsigned char nh_flags;
4669 		unsigned long event;
4670 	};
4671 };
4672 
rt6_multipath_first_sibling(const struct fib6_info * rt)4673 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4674 {
4675 	struct fib6_info *iter;
4676 	struct fib6_node *fn;
4677 
4678 	fn = rcu_dereference_protected(rt->fib6_node,
4679 			lockdep_is_held(&rt->fib6_table->tb6_lock));
4680 	iter = rcu_dereference_protected(fn->leaf,
4681 			lockdep_is_held(&rt->fib6_table->tb6_lock));
4682 	while (iter) {
4683 		if (iter->fib6_metric == rt->fib6_metric &&
4684 		    rt6_qualify_for_ecmp(iter))
4685 			return iter;
4686 		iter = rcu_dereference_protected(iter->fib6_next,
4687 				lockdep_is_held(&rt->fib6_table->tb6_lock));
4688 	}
4689 
4690 	return NULL;
4691 }
4692 
4693 /* only called for fib entries with builtin fib6_nh */
rt6_is_dead(const struct fib6_info * rt)4694 static bool rt6_is_dead(const struct fib6_info *rt)
4695 {
4696 	if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4697 	    (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4698 	     ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4699 		return true;
4700 
4701 	return false;
4702 }
4703 
rt6_multipath_total_weight(const struct fib6_info * rt)4704 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4705 {
4706 	struct fib6_info *iter;
4707 	int total = 0;
4708 
4709 	if (!rt6_is_dead(rt))
4710 		total += rt->fib6_nh->fib_nh_weight;
4711 
4712 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4713 		if (!rt6_is_dead(iter))
4714 			total += iter->fib6_nh->fib_nh_weight;
4715 	}
4716 
4717 	return total;
4718 }
4719 
rt6_upper_bound_set(struct fib6_info * rt,int * weight,int total)4720 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4721 {
4722 	int upper_bound = -1;
4723 
4724 	if (!rt6_is_dead(rt)) {
4725 		*weight += rt->fib6_nh->fib_nh_weight;
4726 		upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4727 						    total) - 1;
4728 	}
4729 	atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4730 }
4731 
rt6_multipath_upper_bound_set(struct fib6_info * rt,int total)4732 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4733 {
4734 	struct fib6_info *iter;
4735 	int weight = 0;
4736 
4737 	rt6_upper_bound_set(rt, &weight, total);
4738 
4739 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4740 		rt6_upper_bound_set(iter, &weight, total);
4741 }
4742 
rt6_multipath_rebalance(struct fib6_info * rt)4743 void rt6_multipath_rebalance(struct fib6_info *rt)
4744 {
4745 	struct fib6_info *first;
4746 	int total;
4747 
4748 	/* In case the entire multipath route was marked for flushing,
4749 	 * then there is no need to rebalance upon the removal of every
4750 	 * sibling route.
4751 	 */
4752 	if (!rt->fib6_nsiblings || rt->should_flush)
4753 		return;
4754 
4755 	/* During lookup routes are evaluated in order, so we need to
4756 	 * make sure upper bounds are assigned from the first sibling
4757 	 * onwards.
4758 	 */
4759 	first = rt6_multipath_first_sibling(rt);
4760 	if (WARN_ON_ONCE(!first))
4761 		return;
4762 
4763 	total = rt6_multipath_total_weight(first);
4764 	rt6_multipath_upper_bound_set(first, total);
4765 }
4766 
fib6_ifup(struct fib6_info * rt,void * p_arg)4767 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4768 {
4769 	const struct arg_netdev_event *arg = p_arg;
4770 	struct net *net = dev_net(arg->dev);
4771 
4772 	if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4773 	    rt->fib6_nh->fib_nh_dev == arg->dev) {
4774 		rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4775 		fib6_update_sernum_upto_root(net, rt);
4776 		rt6_multipath_rebalance(rt);
4777 	}
4778 
4779 	return 0;
4780 }
4781 
rt6_sync_up(struct net_device * dev,unsigned char nh_flags)4782 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4783 {
4784 	struct arg_netdev_event arg = {
4785 		.dev = dev,
4786 		{
4787 			.nh_flags = nh_flags,
4788 		},
4789 	};
4790 
4791 	if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4792 		arg.nh_flags |= RTNH_F_LINKDOWN;
4793 
4794 	fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4795 }
4796 
4797 /* only called for fib entries with inline fib6_nh */
rt6_multipath_uses_dev(const struct fib6_info * rt,const struct net_device * dev)4798 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4799 				   const struct net_device *dev)
4800 {
4801 	struct fib6_info *iter;
4802 
4803 	if (rt->fib6_nh->fib_nh_dev == dev)
4804 		return true;
4805 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4806 		if (iter->fib6_nh->fib_nh_dev == dev)
4807 			return true;
4808 
4809 	return false;
4810 }
4811 
rt6_multipath_flush(struct fib6_info * rt)4812 static void rt6_multipath_flush(struct fib6_info *rt)
4813 {
4814 	struct fib6_info *iter;
4815 
4816 	rt->should_flush = 1;
4817 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4818 		iter->should_flush = 1;
4819 }
4820 
rt6_multipath_dead_count(const struct fib6_info * rt,const struct net_device * down_dev)4821 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4822 					     const struct net_device *down_dev)
4823 {
4824 	struct fib6_info *iter;
4825 	unsigned int dead = 0;
4826 
4827 	if (rt->fib6_nh->fib_nh_dev == down_dev ||
4828 	    rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4829 		dead++;
4830 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4831 		if (iter->fib6_nh->fib_nh_dev == down_dev ||
4832 		    iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4833 			dead++;
4834 
4835 	return dead;
4836 }
4837 
rt6_multipath_nh_flags_set(struct fib6_info * rt,const struct net_device * dev,unsigned char nh_flags)4838 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4839 				       const struct net_device *dev,
4840 				       unsigned char nh_flags)
4841 {
4842 	struct fib6_info *iter;
4843 
4844 	if (rt->fib6_nh->fib_nh_dev == dev)
4845 		rt->fib6_nh->fib_nh_flags |= nh_flags;
4846 	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4847 		if (iter->fib6_nh->fib_nh_dev == dev)
4848 			iter->fib6_nh->fib_nh_flags |= nh_flags;
4849 }
4850 
4851 /* called with write lock held for table with rt */
fib6_ifdown(struct fib6_info * rt,void * p_arg)4852 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4853 {
4854 	const struct arg_netdev_event *arg = p_arg;
4855 	const struct net_device *dev = arg->dev;
4856 	struct net *net = dev_net(dev);
4857 
4858 	if (rt == net->ipv6.fib6_null_entry || rt->nh)
4859 		return 0;
4860 
4861 	switch (arg->event) {
4862 	case NETDEV_UNREGISTER:
4863 		return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4864 	case NETDEV_DOWN:
4865 		if (rt->should_flush)
4866 			return -1;
4867 		if (!rt->fib6_nsiblings)
4868 			return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4869 		if (rt6_multipath_uses_dev(rt, dev)) {
4870 			unsigned int count;
4871 
4872 			count = rt6_multipath_dead_count(rt, dev);
4873 			if (rt->fib6_nsiblings + 1 == count) {
4874 				rt6_multipath_flush(rt);
4875 				return -1;
4876 			}
4877 			rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4878 						   RTNH_F_LINKDOWN);
4879 			fib6_update_sernum(net, rt);
4880 			rt6_multipath_rebalance(rt);
4881 		}
4882 		return -2;
4883 	case NETDEV_CHANGE:
4884 		if (rt->fib6_nh->fib_nh_dev != dev ||
4885 		    rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4886 			break;
4887 		rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4888 		rt6_multipath_rebalance(rt);
4889 		break;
4890 	}
4891 
4892 	return 0;
4893 }
4894 
rt6_sync_down_dev(struct net_device * dev,unsigned long event)4895 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4896 {
4897 	struct arg_netdev_event arg = {
4898 		.dev = dev,
4899 		{
4900 			.event = event,
4901 		},
4902 	};
4903 	struct net *net = dev_net(dev);
4904 
4905 	if (net->ipv6.sysctl.skip_notify_on_dev_down)
4906 		fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4907 	else
4908 		fib6_clean_all(net, fib6_ifdown, &arg);
4909 }
4910 
rt6_disable_ip(struct net_device * dev,unsigned long event)4911 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4912 {
4913 	rt6_sync_down_dev(dev, event);
4914 	rt6_uncached_list_flush_dev(dev);
4915 	neigh_ifdown(&nd_tbl, dev);
4916 }
4917 
4918 struct rt6_mtu_change_arg {
4919 	struct net_device *dev;
4920 	unsigned int mtu;
4921 	struct fib6_info *f6i;
4922 };
4923 
fib6_nh_mtu_change(struct fib6_nh * nh,void * _arg)4924 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4925 {
4926 	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4927 	struct fib6_info *f6i = arg->f6i;
4928 
4929 	/* For administrative MTU increase, there is no way to discover
4930 	 * IPv6 PMTU increase, so PMTU increase should be updated here.
4931 	 * Since RFC 1981 doesn't include administrative MTU increase
4932 	 * update PMTU increase is a MUST. (i.e. jumbo frame)
4933 	 */
4934 	if (nh->fib_nh_dev == arg->dev) {
4935 		struct inet6_dev *idev = __in6_dev_get(arg->dev);
4936 		u32 mtu = f6i->fib6_pmtu;
4937 
4938 		if (mtu >= arg->mtu ||
4939 		    (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4940 			fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4941 
4942 		spin_lock_bh(&rt6_exception_lock);
4943 		rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4944 		spin_unlock_bh(&rt6_exception_lock);
4945 	}
4946 
4947 	return 0;
4948 }
4949 
rt6_mtu_change_route(struct fib6_info * f6i,void * p_arg)4950 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4951 {
4952 	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4953 	struct inet6_dev *idev;
4954 
4955 	/* In IPv6 pmtu discovery is not optional,
4956 	   so that RTAX_MTU lock cannot disable it.
4957 	   We still use this lock to block changes
4958 	   caused by addrconf/ndisc.
4959 	*/
4960 
4961 	idev = __in6_dev_get(arg->dev);
4962 	if (!idev)
4963 		return 0;
4964 
4965 	if (fib6_metric_locked(f6i, RTAX_MTU))
4966 		return 0;
4967 
4968 	arg->f6i = f6i;
4969 	if (f6i->nh) {
4970 		/* fib6_nh_mtu_change only returns 0, so this is safe */
4971 		return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4972 						arg);
4973 	}
4974 
4975 	return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4976 }
4977 
rt6_mtu_change(struct net_device * dev,unsigned int mtu)4978 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4979 {
4980 	struct rt6_mtu_change_arg arg = {
4981 		.dev = dev,
4982 		.mtu = mtu,
4983 	};
4984 
4985 	fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4986 }
4987 
4988 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4989 	[RTA_UNSPEC]		= { .strict_start_type = RTA_DPORT + 1 },
4990 	[RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
4991 	[RTA_PREFSRC]		= { .len = sizeof(struct in6_addr) },
4992 	[RTA_OIF]               = { .type = NLA_U32 },
4993 	[RTA_IIF]		= { .type = NLA_U32 },
4994 	[RTA_PRIORITY]          = { .type = NLA_U32 },
4995 	[RTA_METRICS]           = { .type = NLA_NESTED },
4996 	[RTA_MULTIPATH]		= { .len = sizeof(struct rtnexthop) },
4997 	[RTA_PREF]              = { .type = NLA_U8 },
4998 	[RTA_ENCAP_TYPE]	= { .type = NLA_U16 },
4999 	[RTA_ENCAP]		= { .type = NLA_NESTED },
5000 	[RTA_EXPIRES]		= { .type = NLA_U32 },
5001 	[RTA_UID]		= { .type = NLA_U32 },
5002 	[RTA_MARK]		= { .type = NLA_U32 },
5003 	[RTA_TABLE]		= { .type = NLA_U32 },
5004 	[RTA_IP_PROTO]		= { .type = NLA_U8 },
5005 	[RTA_SPORT]		= { .type = NLA_U16 },
5006 	[RTA_DPORT]		= { .type = NLA_U16 },
5007 	[RTA_NH_ID]		= { .type = NLA_U32 },
5008 };
5009 
rtm_to_fib6_config(struct sk_buff * skb,struct nlmsghdr * nlh,struct fib6_config * cfg,struct netlink_ext_ack * extack)5010 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
5011 			      struct fib6_config *cfg,
5012 			      struct netlink_ext_ack *extack)
5013 {
5014 	struct rtmsg *rtm;
5015 	struct nlattr *tb[RTA_MAX+1];
5016 	unsigned int pref;
5017 	int err;
5018 
5019 	err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5020 				     rtm_ipv6_policy, extack);
5021 	if (err < 0)
5022 		goto errout;
5023 
5024 	err = -EINVAL;
5025 	rtm = nlmsg_data(nlh);
5026 
5027 	if (rtm->rtm_tos) {
5028 		NL_SET_ERR_MSG(extack,
5029 			       "Invalid dsfield (tos): option not available for IPv6");
5030 		goto errout;
5031 	}
5032 
5033 	*cfg = (struct fib6_config){
5034 		.fc_table = rtm->rtm_table,
5035 		.fc_dst_len = rtm->rtm_dst_len,
5036 		.fc_src_len = rtm->rtm_src_len,
5037 		.fc_flags = RTF_UP,
5038 		.fc_protocol = rtm->rtm_protocol,
5039 		.fc_type = rtm->rtm_type,
5040 
5041 		.fc_nlinfo.portid = NETLINK_CB(skb).portid,
5042 		.fc_nlinfo.nlh = nlh,
5043 		.fc_nlinfo.nl_net = sock_net(skb->sk),
5044 	};
5045 
5046 	if (rtm->rtm_type == RTN_UNREACHABLE ||
5047 	    rtm->rtm_type == RTN_BLACKHOLE ||
5048 	    rtm->rtm_type == RTN_PROHIBIT ||
5049 	    rtm->rtm_type == RTN_THROW)
5050 		cfg->fc_flags |= RTF_REJECT;
5051 
5052 	if (rtm->rtm_type == RTN_LOCAL)
5053 		cfg->fc_flags |= RTF_LOCAL;
5054 
5055 	if (rtm->rtm_flags & RTM_F_CLONED)
5056 		cfg->fc_flags |= RTF_CACHE;
5057 
5058 	cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5059 
5060 	if (tb[RTA_NH_ID]) {
5061 		if (tb[RTA_GATEWAY]   || tb[RTA_OIF] ||
5062 		    tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5063 			NL_SET_ERR_MSG(extack,
5064 				       "Nexthop specification and nexthop id are mutually exclusive");
5065 			goto errout;
5066 		}
5067 		cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5068 	}
5069 
5070 	if (tb[RTA_GATEWAY]) {
5071 		cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5072 		cfg->fc_flags |= RTF_GATEWAY;
5073 	}
5074 	if (tb[RTA_VIA]) {
5075 		NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5076 		goto errout;
5077 	}
5078 
5079 	if (tb[RTA_DST]) {
5080 		int plen = (rtm->rtm_dst_len + 7) >> 3;
5081 
5082 		if (nla_len(tb[RTA_DST]) < plen)
5083 			goto errout;
5084 
5085 		nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5086 	}
5087 
5088 	if (tb[RTA_SRC]) {
5089 		int plen = (rtm->rtm_src_len + 7) >> 3;
5090 
5091 		if (nla_len(tb[RTA_SRC]) < plen)
5092 			goto errout;
5093 
5094 		nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5095 	}
5096 
5097 	if (tb[RTA_PREFSRC])
5098 		cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5099 
5100 	if (tb[RTA_OIF])
5101 		cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5102 
5103 	if (tb[RTA_PRIORITY])
5104 		cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5105 
5106 	if (tb[RTA_METRICS]) {
5107 		cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5108 		cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5109 	}
5110 
5111 	if (tb[RTA_TABLE])
5112 		cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5113 
5114 	if (tb[RTA_MULTIPATH]) {
5115 		cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5116 		cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5117 
5118 		err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5119 						     cfg->fc_mp_len, extack);
5120 		if (err < 0)
5121 			goto errout;
5122 	}
5123 
5124 	if (tb[RTA_PREF]) {
5125 		pref = nla_get_u8(tb[RTA_PREF]);
5126 		if (pref != ICMPV6_ROUTER_PREF_LOW &&
5127 		    pref != ICMPV6_ROUTER_PREF_HIGH)
5128 			pref = ICMPV6_ROUTER_PREF_MEDIUM;
5129 		cfg->fc_flags |= RTF_PREF(pref);
5130 	}
5131 
5132 	if (tb[RTA_ENCAP])
5133 		cfg->fc_encap = tb[RTA_ENCAP];
5134 
5135 	if (tb[RTA_ENCAP_TYPE]) {
5136 		cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5137 
5138 		err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5139 		if (err < 0)
5140 			goto errout;
5141 	}
5142 
5143 	if (tb[RTA_EXPIRES]) {
5144 		unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5145 
5146 		if (addrconf_finite_timeout(timeout)) {
5147 			cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5148 			cfg->fc_flags |= RTF_EXPIRES;
5149 		}
5150 	}
5151 
5152 	err = 0;
5153 errout:
5154 	return err;
5155 }
5156 
5157 struct rt6_nh {
5158 	struct fib6_info *fib6_info;
5159 	struct fib6_config r_cfg;
5160 	struct list_head next;
5161 };
5162 
ip6_route_info_append(struct net * net,struct list_head * rt6_nh_list,struct fib6_info * rt,struct fib6_config * r_cfg)5163 static int ip6_route_info_append(struct net *net,
5164 				 struct list_head *rt6_nh_list,
5165 				 struct fib6_info *rt,
5166 				 struct fib6_config *r_cfg)
5167 {
5168 	struct rt6_nh *nh;
5169 	int err = -EEXIST;
5170 
5171 	list_for_each_entry(nh, rt6_nh_list, next) {
5172 		/* check if fib6_info already exists */
5173 		if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5174 			return err;
5175 	}
5176 
5177 	nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5178 	if (!nh)
5179 		return -ENOMEM;
5180 	nh->fib6_info = rt;
5181 	memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5182 	list_add_tail(&nh->next, rt6_nh_list);
5183 
5184 	return 0;
5185 }
5186 
ip6_route_mpath_notify(struct fib6_info * rt,struct fib6_info * rt_last,struct nl_info * info,__u16 nlflags)5187 static void ip6_route_mpath_notify(struct fib6_info *rt,
5188 				   struct fib6_info *rt_last,
5189 				   struct nl_info *info,
5190 				   __u16 nlflags)
5191 {
5192 	/* if this is an APPEND route, then rt points to the first route
5193 	 * inserted and rt_last points to last route inserted. Userspace
5194 	 * wants a consistent dump of the route which starts at the first
5195 	 * nexthop. Since sibling routes are always added at the end of
5196 	 * the list, find the first sibling of the last route appended
5197 	 */
5198 	if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5199 		rt = list_first_entry(&rt_last->fib6_siblings,
5200 				      struct fib6_info,
5201 				      fib6_siblings);
5202 	}
5203 
5204 	if (rt)
5205 		inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5206 }
5207 
ip6_route_mpath_should_notify(const struct fib6_info * rt)5208 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5209 {
5210 	bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5211 	bool should_notify = false;
5212 	struct fib6_info *leaf;
5213 	struct fib6_node *fn;
5214 
5215 	rcu_read_lock();
5216 	fn = rcu_dereference(rt->fib6_node);
5217 	if (!fn)
5218 		goto out;
5219 
5220 	leaf = rcu_dereference(fn->leaf);
5221 	if (!leaf)
5222 		goto out;
5223 
5224 	if (rt == leaf ||
5225 	    (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5226 	     rt6_qualify_for_ecmp(leaf)))
5227 		should_notify = true;
5228 out:
5229 	rcu_read_unlock();
5230 
5231 	return should_notify;
5232 }
5233 
fib6_gw_from_attr(struct in6_addr * gw,struct nlattr * nla,struct netlink_ext_ack * extack)5234 static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5235 			     struct netlink_ext_ack *extack)
5236 {
5237 	if (nla_len(nla) < sizeof(*gw)) {
5238 		NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5239 		return -EINVAL;
5240 	}
5241 
5242 	*gw = nla_get_in6_addr(nla);
5243 
5244 	return 0;
5245 }
5246 
ip6_route_multipath_add(struct fib6_config * cfg,struct netlink_ext_ack * extack)5247 static int ip6_route_multipath_add(struct fib6_config *cfg,
5248 				   struct netlink_ext_ack *extack)
5249 {
5250 	struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5251 	struct nl_info *info = &cfg->fc_nlinfo;
5252 	struct fib6_config r_cfg;
5253 	struct rtnexthop *rtnh;
5254 	struct fib6_info *rt;
5255 	struct rt6_nh *err_nh;
5256 	struct rt6_nh *nh, *nh_safe;
5257 	__u16 nlflags;
5258 	int remaining;
5259 	int attrlen;
5260 	int err = 1;
5261 	int nhn = 0;
5262 	int replace = (cfg->fc_nlinfo.nlh &&
5263 		       (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5264 	LIST_HEAD(rt6_nh_list);
5265 
5266 	nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5267 	if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5268 		nlflags |= NLM_F_APPEND;
5269 
5270 	remaining = cfg->fc_mp_len;
5271 	rtnh = (struct rtnexthop *)cfg->fc_mp;
5272 
5273 	/* Parse a Multipath Entry and build a list (rt6_nh_list) of
5274 	 * fib6_info structs per nexthop
5275 	 */
5276 	while (rtnh_ok(rtnh, remaining)) {
5277 		memcpy(&r_cfg, cfg, sizeof(*cfg));
5278 		if (rtnh->rtnh_ifindex)
5279 			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5280 
5281 		attrlen = rtnh_attrlen(rtnh);
5282 		if (attrlen > 0) {
5283 			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5284 
5285 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5286 			if (nla) {
5287 				err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5288 							extack);
5289 				if (err)
5290 					goto cleanup;
5291 
5292 				r_cfg.fc_flags |= RTF_GATEWAY;
5293 			}
5294 			r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5295 
5296 			/* RTA_ENCAP_TYPE length checked in
5297 			 * lwtunnel_valid_encap_type_attr
5298 			 */
5299 			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5300 			if (nla)
5301 				r_cfg.fc_encap_type = nla_get_u16(nla);
5302 		}
5303 
5304 		r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5305 		rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5306 		if (IS_ERR(rt)) {
5307 			err = PTR_ERR(rt);
5308 			rt = NULL;
5309 			goto cleanup;
5310 		}
5311 		if (!rt6_qualify_for_ecmp(rt)) {
5312 			err = -EINVAL;
5313 			NL_SET_ERR_MSG(extack,
5314 				       "Device only routes can not be added for IPv6 using the multipath API.");
5315 			fib6_info_release(rt);
5316 			goto cleanup;
5317 		}
5318 
5319 		rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5320 
5321 		err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5322 					    rt, &r_cfg);
5323 		if (err) {
5324 			fib6_info_release(rt);
5325 			goto cleanup;
5326 		}
5327 
5328 		rtnh = rtnh_next(rtnh, &remaining);
5329 	}
5330 
5331 	if (list_empty(&rt6_nh_list)) {
5332 		NL_SET_ERR_MSG(extack,
5333 			       "Invalid nexthop configuration - no valid nexthops");
5334 		return -EINVAL;
5335 	}
5336 
5337 	/* for add and replace send one notification with all nexthops.
5338 	 * Skip the notification in fib6_add_rt2node and send one with
5339 	 * the full route when done
5340 	 */
5341 	info->skip_notify = 1;
5342 
5343 	/* For add and replace, send one notification with all nexthops. For
5344 	 * append, send one notification with all appended nexthops.
5345 	 */
5346 	info->skip_notify_kernel = 1;
5347 
5348 	err_nh = NULL;
5349 	list_for_each_entry(nh, &rt6_nh_list, next) {
5350 		err = __ip6_ins_rt(nh->fib6_info, info, extack);
5351 
5352 		if (err) {
5353 			if (replace && nhn)
5354 				NL_SET_ERR_MSG_MOD(extack,
5355 						   "multipath route replace failed (check consistency of installed routes)");
5356 			err_nh = nh;
5357 			goto add_errout;
5358 		}
5359 		/* save reference to last route successfully inserted */
5360 		rt_last = nh->fib6_info;
5361 
5362 		/* save reference to first route for notification */
5363 		if (!rt_notif)
5364 			rt_notif = nh->fib6_info;
5365 
5366 		/* Because each route is added like a single route we remove
5367 		 * these flags after the first nexthop: if there is a collision,
5368 		 * we have already failed to add the first nexthop:
5369 		 * fib6_add_rt2node() has rejected it; when replacing, old
5370 		 * nexthops have been replaced by first new, the rest should
5371 		 * be added to it.
5372 		 */
5373 		if (cfg->fc_nlinfo.nlh) {
5374 			cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5375 							     NLM_F_REPLACE);
5376 			cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5377 		}
5378 		nhn++;
5379 	}
5380 
5381 	/* An in-kernel notification should only be sent in case the new
5382 	 * multipath route is added as the first route in the node, or if
5383 	 * it was appended to it. We pass 'rt_notif' since it is the first
5384 	 * sibling and might allow us to skip some checks in the replace case.
5385 	 */
5386 	if (ip6_route_mpath_should_notify(rt_notif)) {
5387 		enum fib_event_type fib_event;
5388 
5389 		if (rt_notif->fib6_nsiblings != nhn - 1)
5390 			fib_event = FIB_EVENT_ENTRY_APPEND;
5391 		else
5392 			fib_event = FIB_EVENT_ENTRY_REPLACE;
5393 
5394 		err = call_fib6_multipath_entry_notifiers(info->nl_net,
5395 							  fib_event, rt_notif,
5396 							  nhn - 1, extack);
5397 		if (err) {
5398 			/* Delete all the siblings that were just added */
5399 			err_nh = NULL;
5400 			goto add_errout;
5401 		}
5402 	}
5403 
5404 	/* success ... tell user about new route */
5405 	ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5406 	goto cleanup;
5407 
5408 add_errout:
5409 	/* send notification for routes that were added so that
5410 	 * the delete notifications sent by ip6_route_del are
5411 	 * coherent
5412 	 */
5413 	if (rt_notif)
5414 		ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5415 
5416 	/* Delete routes that were already added */
5417 	list_for_each_entry(nh, &rt6_nh_list, next) {
5418 		if (err_nh == nh)
5419 			break;
5420 		ip6_route_del(&nh->r_cfg, extack);
5421 	}
5422 
5423 cleanup:
5424 	list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5425 		fib6_info_release(nh->fib6_info);
5426 		list_del(&nh->next);
5427 		kfree(nh);
5428 	}
5429 
5430 	return err;
5431 }
5432 
ip6_route_multipath_del(struct fib6_config * cfg,struct netlink_ext_ack * extack)5433 static int ip6_route_multipath_del(struct fib6_config *cfg,
5434 				   struct netlink_ext_ack *extack)
5435 {
5436 	struct fib6_config r_cfg;
5437 	struct rtnexthop *rtnh;
5438 	int last_err = 0;
5439 	int remaining;
5440 	int attrlen;
5441 	int err;
5442 
5443 	remaining = cfg->fc_mp_len;
5444 	rtnh = (struct rtnexthop *)cfg->fc_mp;
5445 
5446 	/* Parse a Multipath Entry */
5447 	while (rtnh_ok(rtnh, remaining)) {
5448 		memcpy(&r_cfg, cfg, sizeof(*cfg));
5449 		if (rtnh->rtnh_ifindex)
5450 			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5451 
5452 		attrlen = rtnh_attrlen(rtnh);
5453 		if (attrlen > 0) {
5454 			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5455 
5456 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5457 			if (nla) {
5458 				err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5459 							extack);
5460 				if (err) {
5461 					last_err = err;
5462 					goto next_rtnh;
5463 				}
5464 
5465 				r_cfg.fc_flags |= RTF_GATEWAY;
5466 			}
5467 		}
5468 		err = ip6_route_del(&r_cfg, extack);
5469 		if (err)
5470 			last_err = err;
5471 
5472 next_rtnh:
5473 		rtnh = rtnh_next(rtnh, &remaining);
5474 	}
5475 
5476 	return last_err;
5477 }
5478 
inet6_rtm_delroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5479 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5480 			      struct netlink_ext_ack *extack)
5481 {
5482 	struct fib6_config cfg;
5483 	int err;
5484 
5485 	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5486 	if (err < 0)
5487 		return err;
5488 
5489 	if (cfg.fc_nh_id &&
5490 	    !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5491 		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5492 		return -EINVAL;
5493 	}
5494 
5495 	if (cfg.fc_mp)
5496 		return ip6_route_multipath_del(&cfg, extack);
5497 	else {
5498 		cfg.fc_delete_all_nh = 1;
5499 		return ip6_route_del(&cfg, extack);
5500 	}
5501 }
5502 
inet6_rtm_newroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5503 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5504 			      struct netlink_ext_ack *extack)
5505 {
5506 	struct fib6_config cfg;
5507 	int err;
5508 
5509 	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5510 	if (err < 0)
5511 		return err;
5512 
5513 	if (cfg.fc_metric == 0)
5514 		cfg.fc_metric = IP6_RT_PRIO_USER;
5515 
5516 	if (cfg.fc_mp)
5517 		return ip6_route_multipath_add(&cfg, extack);
5518 	else
5519 		return ip6_route_add(&cfg, GFP_KERNEL, extack);
5520 }
5521 
5522 /* add the overhead of this fib6_nh to nexthop_len */
rt6_nh_nlmsg_size(struct fib6_nh * nh,void * arg)5523 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5524 {
5525 	int *nexthop_len = arg;
5526 
5527 	*nexthop_len += nla_total_size(0)	 /* RTA_MULTIPATH */
5528 		     + NLA_ALIGN(sizeof(struct rtnexthop))
5529 		     + nla_total_size(16); /* RTA_GATEWAY */
5530 
5531 	if (nh->fib_nh_lws) {
5532 		/* RTA_ENCAP_TYPE */
5533 		*nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5534 		/* RTA_ENCAP */
5535 		*nexthop_len += nla_total_size(2);
5536 	}
5537 
5538 	return 0;
5539 }
5540 
rt6_nlmsg_size(struct fib6_info * f6i)5541 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5542 {
5543 	int nexthop_len;
5544 
5545 	if (f6i->nh) {
5546 		nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5547 		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5548 					 &nexthop_len);
5549 	} else {
5550 		struct fib6_info *sibling, *next_sibling;
5551 		struct fib6_nh *nh = f6i->fib6_nh;
5552 
5553 		nexthop_len = 0;
5554 		if (f6i->fib6_nsiblings) {
5555 			rt6_nh_nlmsg_size(nh, &nexthop_len);
5556 
5557 			list_for_each_entry_safe(sibling, next_sibling,
5558 						 &f6i->fib6_siblings, fib6_siblings) {
5559 				rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5560 			}
5561 		}
5562 		nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5563 	}
5564 
5565 	return NLMSG_ALIGN(sizeof(struct rtmsg))
5566 	       + nla_total_size(16) /* RTA_SRC */
5567 	       + nla_total_size(16) /* RTA_DST */
5568 	       + nla_total_size(16) /* RTA_GATEWAY */
5569 	       + nla_total_size(16) /* RTA_PREFSRC */
5570 	       + nla_total_size(4) /* RTA_TABLE */
5571 	       + nla_total_size(4) /* RTA_IIF */
5572 	       + nla_total_size(4) /* RTA_OIF */
5573 	       + nla_total_size(4) /* RTA_PRIORITY */
5574 	       + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5575 	       + nla_total_size(sizeof(struct rta_cacheinfo))
5576 	       + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5577 	       + nla_total_size(1) /* RTA_PREF */
5578 	       + nexthop_len;
5579 }
5580 
rt6_fill_node_nexthop(struct sk_buff * skb,struct nexthop * nh,unsigned char * flags)5581 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5582 				 unsigned char *flags)
5583 {
5584 	if (nexthop_is_multipath(nh)) {
5585 		struct nlattr *mp;
5586 
5587 		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5588 		if (!mp)
5589 			goto nla_put_failure;
5590 
5591 		if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5592 			goto nla_put_failure;
5593 
5594 		nla_nest_end(skb, mp);
5595 	} else {
5596 		struct fib6_nh *fib6_nh;
5597 
5598 		fib6_nh = nexthop_fib6_nh(nh);
5599 		if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5600 				     flags, false) < 0)
5601 			goto nla_put_failure;
5602 	}
5603 
5604 	return 0;
5605 
5606 nla_put_failure:
5607 	return -EMSGSIZE;
5608 }
5609 
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)5610 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5611 			 struct fib6_info *rt, struct dst_entry *dst,
5612 			 struct in6_addr *dest, struct in6_addr *src,
5613 			 int iif, int type, u32 portid, u32 seq,
5614 			 unsigned int flags)
5615 {
5616 	struct rt6_info *rt6 = dst_rt6_info(dst);
5617 	struct rt6key *rt6_dst, *rt6_src;
5618 	u32 *pmetrics, table, rt6_flags;
5619 	unsigned char nh_flags = 0;
5620 	struct nlmsghdr *nlh;
5621 	struct rtmsg *rtm;
5622 	long expires = 0;
5623 
5624 	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5625 	if (!nlh)
5626 		return -EMSGSIZE;
5627 
5628 	if (rt6) {
5629 		rt6_dst = &rt6->rt6i_dst;
5630 		rt6_src = &rt6->rt6i_src;
5631 		rt6_flags = rt6->rt6i_flags;
5632 	} else {
5633 		rt6_dst = &rt->fib6_dst;
5634 		rt6_src = &rt->fib6_src;
5635 		rt6_flags = rt->fib6_flags;
5636 	}
5637 
5638 	rtm = nlmsg_data(nlh);
5639 	rtm->rtm_family = AF_INET6;
5640 	rtm->rtm_dst_len = rt6_dst->plen;
5641 	rtm->rtm_src_len = rt6_src->plen;
5642 	rtm->rtm_tos = 0;
5643 	if (rt->fib6_table)
5644 		table = rt->fib6_table->tb6_id;
5645 	else
5646 		table = RT6_TABLE_UNSPEC;
5647 	rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5648 	if (nla_put_u32(skb, RTA_TABLE, table))
5649 		goto nla_put_failure;
5650 
5651 	rtm->rtm_type = rt->fib6_type;
5652 	rtm->rtm_flags = 0;
5653 	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5654 	rtm->rtm_protocol = rt->fib6_protocol;
5655 
5656 	if (rt6_flags & RTF_CACHE)
5657 		rtm->rtm_flags |= RTM_F_CLONED;
5658 
5659 	if (dest) {
5660 		if (nla_put_in6_addr(skb, RTA_DST, dest))
5661 			goto nla_put_failure;
5662 		rtm->rtm_dst_len = 128;
5663 	} else if (rtm->rtm_dst_len)
5664 		if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5665 			goto nla_put_failure;
5666 #ifdef CONFIG_IPV6_SUBTREES
5667 	if (src) {
5668 		if (nla_put_in6_addr(skb, RTA_SRC, src))
5669 			goto nla_put_failure;
5670 		rtm->rtm_src_len = 128;
5671 	} else if (rtm->rtm_src_len &&
5672 		   nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5673 		goto nla_put_failure;
5674 #endif
5675 	if (iif) {
5676 #ifdef CONFIG_IPV6_MROUTE
5677 		if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5678 			int err = ip6mr_get_route(net, skb, rtm, portid);
5679 
5680 			if (err == 0)
5681 				return 0;
5682 			if (err < 0)
5683 				goto nla_put_failure;
5684 		} else
5685 #endif
5686 			if (nla_put_u32(skb, RTA_IIF, iif))
5687 				goto nla_put_failure;
5688 	} else if (dest) {
5689 		struct in6_addr saddr_buf;
5690 		if (ip6_route_get_saddr(net, rt, dest, 0, 0, &saddr_buf) == 0 &&
5691 		    nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5692 			goto nla_put_failure;
5693 	}
5694 
5695 	if (rt->fib6_prefsrc.plen) {
5696 		struct in6_addr saddr_buf;
5697 		saddr_buf = rt->fib6_prefsrc.addr;
5698 		if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5699 			goto nla_put_failure;
5700 	}
5701 
5702 	pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5703 	if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5704 		goto nla_put_failure;
5705 
5706 	if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5707 		goto nla_put_failure;
5708 
5709 	/* For multipath routes, walk the siblings list and add
5710 	 * each as a nexthop within RTA_MULTIPATH.
5711 	 */
5712 	if (rt6) {
5713 		if (rt6_flags & RTF_GATEWAY &&
5714 		    nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5715 			goto nla_put_failure;
5716 
5717 		if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5718 			goto nla_put_failure;
5719 
5720 		if (dst->lwtstate &&
5721 		    lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5722 			goto nla_put_failure;
5723 	} else if (rt->fib6_nsiblings) {
5724 		struct fib6_info *sibling, *next_sibling;
5725 		struct nlattr *mp;
5726 
5727 		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5728 		if (!mp)
5729 			goto nla_put_failure;
5730 
5731 		if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5732 				    rt->fib6_nh->fib_nh_weight, AF_INET6,
5733 				    0) < 0)
5734 			goto nla_put_failure;
5735 
5736 		list_for_each_entry_safe(sibling, next_sibling,
5737 					 &rt->fib6_siblings, fib6_siblings) {
5738 			if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5739 					    sibling->fib6_nh->fib_nh_weight,
5740 					    AF_INET6, 0) < 0)
5741 				goto nla_put_failure;
5742 		}
5743 
5744 		nla_nest_end(skb, mp);
5745 	} else if (rt->nh) {
5746 		if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5747 			goto nla_put_failure;
5748 
5749 		if (nexthop_is_blackhole(rt->nh))
5750 			rtm->rtm_type = RTN_BLACKHOLE;
5751 
5752 		if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5753 		    rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5754 			goto nla_put_failure;
5755 
5756 		rtm->rtm_flags |= nh_flags;
5757 	} else {
5758 		if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5759 				     &nh_flags, false) < 0)
5760 			goto nla_put_failure;
5761 
5762 		rtm->rtm_flags |= nh_flags;
5763 	}
5764 
5765 	if (rt6_flags & RTF_EXPIRES) {
5766 		expires = dst ? dst->expires : rt->expires;
5767 		expires -= jiffies;
5768 	}
5769 
5770 	if (!dst) {
5771 		if (READ_ONCE(rt->offload))
5772 			rtm->rtm_flags |= RTM_F_OFFLOAD;
5773 		if (READ_ONCE(rt->trap))
5774 			rtm->rtm_flags |= RTM_F_TRAP;
5775 		if (READ_ONCE(rt->offload_failed))
5776 			rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5777 	}
5778 
5779 	if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5780 		goto nla_put_failure;
5781 
5782 	if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5783 		goto nla_put_failure;
5784 
5785 
5786 	nlmsg_end(skb, nlh);
5787 	return 0;
5788 
5789 nla_put_failure:
5790 	nlmsg_cancel(skb, nlh);
5791 	return -EMSGSIZE;
5792 }
5793 
fib6_info_nh_uses_dev(struct fib6_nh * nh,void * arg)5794 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5795 {
5796 	const struct net_device *dev = arg;
5797 
5798 	if (nh->fib_nh_dev == dev)
5799 		return 1;
5800 
5801 	return 0;
5802 }
5803 
fib6_info_uses_dev(const struct fib6_info * f6i,const struct net_device * dev)5804 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5805 			       const struct net_device *dev)
5806 {
5807 	if (f6i->nh) {
5808 		struct net_device *_dev = (struct net_device *)dev;
5809 
5810 		return !!nexthop_for_each_fib6_nh(f6i->nh,
5811 						  fib6_info_nh_uses_dev,
5812 						  _dev);
5813 	}
5814 
5815 	if (f6i->fib6_nh->fib_nh_dev == dev)
5816 		return true;
5817 
5818 	if (f6i->fib6_nsiblings) {
5819 		struct fib6_info *sibling, *next_sibling;
5820 
5821 		list_for_each_entry_safe(sibling, next_sibling,
5822 					 &f6i->fib6_siblings, fib6_siblings) {
5823 			if (sibling->fib6_nh->fib_nh_dev == dev)
5824 				return true;
5825 		}
5826 	}
5827 
5828 	return false;
5829 }
5830 
5831 struct fib6_nh_exception_dump_walker {
5832 	struct rt6_rtnl_dump_arg *dump;
5833 	struct fib6_info *rt;
5834 	unsigned int flags;
5835 	unsigned int skip;
5836 	unsigned int count;
5837 };
5838 
rt6_nh_dump_exceptions(struct fib6_nh * nh,void * arg)5839 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5840 {
5841 	struct fib6_nh_exception_dump_walker *w = arg;
5842 	struct rt6_rtnl_dump_arg *dump = w->dump;
5843 	struct rt6_exception_bucket *bucket;
5844 	struct rt6_exception *rt6_ex;
5845 	int i, err;
5846 
5847 	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5848 	if (!bucket)
5849 		return 0;
5850 
5851 	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5852 		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5853 			if (w->skip) {
5854 				w->skip--;
5855 				continue;
5856 			}
5857 
5858 			/* Expiration of entries doesn't bump sernum, insertion
5859 			 * does. Removal is triggered by insertion, so we can
5860 			 * rely on the fact that if entries change between two
5861 			 * partial dumps, this node is scanned again completely,
5862 			 * see rt6_insert_exception() and fib6_dump_table().
5863 			 *
5864 			 * Count expired entries we go through as handled
5865 			 * entries that we'll skip next time, in case of partial
5866 			 * node dump. Otherwise, if entries expire meanwhile,
5867 			 * we'll skip the wrong amount.
5868 			 */
5869 			if (rt6_check_expired(rt6_ex->rt6i)) {
5870 				w->count++;
5871 				continue;
5872 			}
5873 
5874 			err = rt6_fill_node(dump->net, dump->skb, w->rt,
5875 					    &rt6_ex->rt6i->dst, NULL, NULL, 0,
5876 					    RTM_NEWROUTE,
5877 					    NETLINK_CB(dump->cb->skb).portid,
5878 					    dump->cb->nlh->nlmsg_seq, w->flags);
5879 			if (err)
5880 				return err;
5881 
5882 			w->count++;
5883 		}
5884 		bucket++;
5885 	}
5886 
5887 	return 0;
5888 }
5889 
5890 /* 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)5891 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5892 {
5893 	struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5894 	struct fib_dump_filter *filter = &arg->filter;
5895 	unsigned int flags = NLM_F_MULTI;
5896 	struct net *net = arg->net;
5897 	int count = 0;
5898 
5899 	if (rt == net->ipv6.fib6_null_entry)
5900 		return -1;
5901 
5902 	if ((filter->flags & RTM_F_PREFIX) &&
5903 	    !(rt->fib6_flags & RTF_PREFIX_RT)) {
5904 		/* success since this is not a prefix route */
5905 		return -1;
5906 	}
5907 	if (filter->filter_set &&
5908 	    ((filter->rt_type  && rt->fib6_type != filter->rt_type) ||
5909 	     (filter->dev      && !fib6_info_uses_dev(rt, filter->dev)) ||
5910 	     (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5911 		return -1;
5912 	}
5913 
5914 	if (filter->filter_set ||
5915 	    !filter->dump_routes || !filter->dump_exceptions) {
5916 		flags |= NLM_F_DUMP_FILTERED;
5917 	}
5918 
5919 	if (filter->dump_routes) {
5920 		if (skip) {
5921 			skip--;
5922 		} else {
5923 			if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5924 					  0, RTM_NEWROUTE,
5925 					  NETLINK_CB(arg->cb->skb).portid,
5926 					  arg->cb->nlh->nlmsg_seq, flags)) {
5927 				return 0;
5928 			}
5929 			count++;
5930 		}
5931 	}
5932 
5933 	if (filter->dump_exceptions) {
5934 		struct fib6_nh_exception_dump_walker w = { .dump = arg,
5935 							   .rt = rt,
5936 							   .flags = flags,
5937 							   .skip = skip,
5938 							   .count = 0 };
5939 		int err;
5940 
5941 		rcu_read_lock();
5942 		if (rt->nh) {
5943 			err = nexthop_for_each_fib6_nh(rt->nh,
5944 						       rt6_nh_dump_exceptions,
5945 						       &w);
5946 		} else {
5947 			err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5948 		}
5949 		rcu_read_unlock();
5950 
5951 		if (err)
5952 			return count + w.count;
5953 	}
5954 
5955 	return -1;
5956 }
5957 
inet6_rtm_valid_getroute_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5958 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5959 					const struct nlmsghdr *nlh,
5960 					struct nlattr **tb,
5961 					struct netlink_ext_ack *extack)
5962 {
5963 	struct rtmsg *rtm;
5964 	int i, err;
5965 
5966 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5967 		NL_SET_ERR_MSG_MOD(extack,
5968 				   "Invalid header for get route request");
5969 		return -EINVAL;
5970 	}
5971 
5972 	if (!netlink_strict_get_check(skb))
5973 		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5974 					      rtm_ipv6_policy, extack);
5975 
5976 	rtm = nlmsg_data(nlh);
5977 	if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5978 	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5979 	    rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5980 	    rtm->rtm_type) {
5981 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5982 		return -EINVAL;
5983 	}
5984 	if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5985 		NL_SET_ERR_MSG_MOD(extack,
5986 				   "Invalid flags for get route request");
5987 		return -EINVAL;
5988 	}
5989 
5990 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5991 					    rtm_ipv6_policy, extack);
5992 	if (err)
5993 		return err;
5994 
5995 	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5996 	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5997 		NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5998 		return -EINVAL;
5999 	}
6000 
6001 	for (i = 0; i <= RTA_MAX; i++) {
6002 		if (!tb[i])
6003 			continue;
6004 
6005 		switch (i) {
6006 		case RTA_SRC:
6007 		case RTA_DST:
6008 		case RTA_IIF:
6009 		case RTA_OIF:
6010 		case RTA_MARK:
6011 		case RTA_UID:
6012 		case RTA_SPORT:
6013 		case RTA_DPORT:
6014 		case RTA_IP_PROTO:
6015 			break;
6016 		default:
6017 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6018 			return -EINVAL;
6019 		}
6020 	}
6021 
6022 	return 0;
6023 }
6024 
inet6_rtm_getroute(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6025 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6026 			      struct netlink_ext_ack *extack)
6027 {
6028 	struct net *net = sock_net(in_skb->sk);
6029 	struct nlattr *tb[RTA_MAX+1];
6030 	int err, iif = 0, oif = 0;
6031 	struct fib6_info *from;
6032 	struct dst_entry *dst;
6033 	struct rt6_info *rt;
6034 	struct sk_buff *skb;
6035 	struct rtmsg *rtm;
6036 	struct flowi6 fl6 = {};
6037 	bool fibmatch;
6038 
6039 	err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6040 	if (err < 0)
6041 		goto errout;
6042 
6043 	err = -EINVAL;
6044 	rtm = nlmsg_data(nlh);
6045 	fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6046 	fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6047 
6048 	if (tb[RTA_SRC]) {
6049 		if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6050 			goto errout;
6051 
6052 		fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6053 	}
6054 
6055 	if (tb[RTA_DST]) {
6056 		if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6057 			goto errout;
6058 
6059 		fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6060 	}
6061 
6062 	if (tb[RTA_IIF])
6063 		iif = nla_get_u32(tb[RTA_IIF]);
6064 
6065 	if (tb[RTA_OIF])
6066 		oif = nla_get_u32(tb[RTA_OIF]);
6067 
6068 	if (tb[RTA_MARK])
6069 		fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6070 
6071 	if (tb[RTA_UID])
6072 		fl6.flowi6_uid = make_kuid(current_user_ns(),
6073 					   nla_get_u32(tb[RTA_UID]));
6074 	else
6075 		fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6076 
6077 	if (tb[RTA_SPORT])
6078 		fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6079 
6080 	if (tb[RTA_DPORT])
6081 		fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6082 
6083 	if (tb[RTA_IP_PROTO]) {
6084 		err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6085 						  &fl6.flowi6_proto, AF_INET6,
6086 						  extack);
6087 		if (err)
6088 			goto errout;
6089 	}
6090 
6091 	if (iif) {
6092 		struct net_device *dev;
6093 		int flags = 0;
6094 
6095 		rcu_read_lock();
6096 
6097 		dev = dev_get_by_index_rcu(net, iif);
6098 		if (!dev) {
6099 			rcu_read_unlock();
6100 			err = -ENODEV;
6101 			goto errout;
6102 		}
6103 
6104 		fl6.flowi6_iif = iif;
6105 
6106 		if (!ipv6_addr_any(&fl6.saddr))
6107 			flags |= RT6_LOOKUP_F_HAS_SADDR;
6108 
6109 		dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6110 
6111 		rcu_read_unlock();
6112 	} else {
6113 		fl6.flowi6_oif = oif;
6114 
6115 		dst = ip6_route_output(net, NULL, &fl6);
6116 	}
6117 
6118 
6119 	rt = dst_rt6_info(dst);
6120 	if (rt->dst.error) {
6121 		err = rt->dst.error;
6122 		ip6_rt_put(rt);
6123 		goto errout;
6124 	}
6125 
6126 	if (rt == net->ipv6.ip6_null_entry) {
6127 		err = rt->dst.error;
6128 		ip6_rt_put(rt);
6129 		goto errout;
6130 	}
6131 
6132 	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6133 	if (!skb) {
6134 		ip6_rt_put(rt);
6135 		err = -ENOBUFS;
6136 		goto errout;
6137 	}
6138 
6139 	skb_dst_set(skb, &rt->dst);
6140 
6141 	rcu_read_lock();
6142 	from = rcu_dereference(rt->from);
6143 	if (from) {
6144 		if (fibmatch)
6145 			err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6146 					    iif, RTM_NEWROUTE,
6147 					    NETLINK_CB(in_skb).portid,
6148 					    nlh->nlmsg_seq, 0);
6149 		else
6150 			err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6151 					    &fl6.saddr, iif, RTM_NEWROUTE,
6152 					    NETLINK_CB(in_skb).portid,
6153 					    nlh->nlmsg_seq, 0);
6154 	} else {
6155 		err = -ENETUNREACH;
6156 	}
6157 	rcu_read_unlock();
6158 
6159 	if (err < 0) {
6160 		kfree_skb(skb);
6161 		goto errout;
6162 	}
6163 
6164 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6165 errout:
6166 	return err;
6167 }
6168 
inet6_rt_notify(int event,struct fib6_info * rt,struct nl_info * info,unsigned int nlm_flags)6169 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6170 		     unsigned int nlm_flags)
6171 {
6172 	struct sk_buff *skb;
6173 	struct net *net = info->nl_net;
6174 	u32 seq;
6175 	int err;
6176 
6177 	err = -ENOBUFS;
6178 	seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6179 
6180 	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6181 	if (!skb)
6182 		goto errout;
6183 
6184 	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6185 			    event, info->portid, seq, nlm_flags);
6186 	if (err < 0) {
6187 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6188 		WARN_ON(err == -EMSGSIZE);
6189 		kfree_skb(skb);
6190 		goto errout;
6191 	}
6192 	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6193 		    info->nlh, gfp_any());
6194 	return;
6195 errout:
6196 	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6197 }
6198 
fib6_rt_update(struct net * net,struct fib6_info * rt,struct nl_info * info)6199 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6200 		    struct nl_info *info)
6201 {
6202 	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6203 	struct sk_buff *skb;
6204 	int err = -ENOBUFS;
6205 
6206 	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6207 	if (!skb)
6208 		goto errout;
6209 
6210 	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6211 			    RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6212 	if (err < 0) {
6213 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6214 		WARN_ON(err == -EMSGSIZE);
6215 		kfree_skb(skb);
6216 		goto errout;
6217 	}
6218 	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6219 		    info->nlh, gfp_any());
6220 	return;
6221 errout:
6222 	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6223 }
6224 
fib6_info_hw_flags_set(struct net * net,struct fib6_info * f6i,bool offload,bool trap,bool offload_failed)6225 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6226 			    bool offload, bool trap, bool offload_failed)
6227 {
6228 	struct sk_buff *skb;
6229 	int err;
6230 
6231 	if (READ_ONCE(f6i->offload) == offload &&
6232 	    READ_ONCE(f6i->trap) == trap &&
6233 	    READ_ONCE(f6i->offload_failed) == offload_failed)
6234 		return;
6235 
6236 	WRITE_ONCE(f6i->offload, offload);
6237 	WRITE_ONCE(f6i->trap, trap);
6238 
6239 	/* 2 means send notifications only if offload_failed was changed. */
6240 	if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6241 	    READ_ONCE(f6i->offload_failed) == offload_failed)
6242 		return;
6243 
6244 	WRITE_ONCE(f6i->offload_failed, offload_failed);
6245 
6246 	if (!rcu_access_pointer(f6i->fib6_node))
6247 		/* The route was removed from the tree, do not send
6248 		 * notification.
6249 		 */
6250 		return;
6251 
6252 	if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6253 		return;
6254 
6255 	skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6256 	if (!skb) {
6257 		err = -ENOBUFS;
6258 		goto errout;
6259 	}
6260 
6261 	err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6262 			    0, 0);
6263 	if (err < 0) {
6264 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6265 		WARN_ON(err == -EMSGSIZE);
6266 		kfree_skb(skb);
6267 		goto errout;
6268 	}
6269 
6270 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6271 	return;
6272 
6273 errout:
6274 	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6275 }
6276 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6277 
ip6_route_dev_notify(struct notifier_block * this,unsigned long event,void * ptr)6278 static int ip6_route_dev_notify(struct notifier_block *this,
6279 				unsigned long event, void *ptr)
6280 {
6281 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6282 	struct net *net = dev_net(dev);
6283 
6284 	if (!(dev->flags & IFF_LOOPBACK))
6285 		return NOTIFY_OK;
6286 
6287 	if (event == NETDEV_REGISTER) {
6288 		net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6289 		net->ipv6.ip6_null_entry->dst.dev = dev;
6290 		net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6291 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6292 		net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6293 		net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6294 		net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6295 		net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6296 #endif
6297 	 } else if (event == NETDEV_UNREGISTER &&
6298 		    dev->reg_state != NETREG_UNREGISTERED) {
6299 		/* NETDEV_UNREGISTER could be fired for multiple times by
6300 		 * netdev_wait_allrefs(). Make sure we only call this once.
6301 		 */
6302 		in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6303 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6304 		in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6305 		in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6306 #endif
6307 	}
6308 
6309 	return NOTIFY_OK;
6310 }
6311 
6312 /*
6313  *	/proc
6314  */
6315 
6316 #ifdef CONFIG_PROC_FS
rt6_stats_seq_show(struct seq_file * seq,void * v)6317 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6318 {
6319 	struct net *net = (struct net *)seq->private;
6320 	seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6321 		   net->ipv6.rt6_stats->fib_nodes,
6322 		   net->ipv6.rt6_stats->fib_route_nodes,
6323 		   atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6324 		   net->ipv6.rt6_stats->fib_rt_entries,
6325 		   net->ipv6.rt6_stats->fib_rt_cache,
6326 		   dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6327 		   net->ipv6.rt6_stats->fib_discarded_routes);
6328 
6329 	return 0;
6330 }
6331 #endif	/* CONFIG_PROC_FS */
6332 
6333 #ifdef CONFIG_SYSCTL
6334 
ipv6_sysctl_rtcache_flush(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6335 static int ipv6_sysctl_rtcache_flush(const struct ctl_table *ctl, int write,
6336 			      void *buffer, size_t *lenp, loff_t *ppos)
6337 {
6338 	struct net *net;
6339 	int delay;
6340 	int ret;
6341 	if (!write)
6342 		return -EINVAL;
6343 
6344 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6345 	if (ret)
6346 		return ret;
6347 
6348 	net = (struct net *)ctl->extra1;
6349 	delay = net->ipv6.sysctl.flush_delay;
6350 	fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6351 	return 0;
6352 }
6353 
6354 static struct ctl_table ipv6_route_table_template[] = {
6355 	{
6356 		.procname	=	"max_size",
6357 		.data		=	&init_net.ipv6.sysctl.ip6_rt_max_size,
6358 		.maxlen		=	sizeof(int),
6359 		.mode		=	0644,
6360 		.proc_handler	=	proc_dointvec,
6361 	},
6362 	{
6363 		.procname	=	"gc_thresh",
6364 		.data		=	&ip6_dst_ops_template.gc_thresh,
6365 		.maxlen		=	sizeof(int),
6366 		.mode		=	0644,
6367 		.proc_handler	=	proc_dointvec,
6368 	},
6369 	{
6370 		.procname	=	"flush",
6371 		.data		=	&init_net.ipv6.sysctl.flush_delay,
6372 		.maxlen		=	sizeof(int),
6373 		.mode		=	0200,
6374 		.proc_handler	=	ipv6_sysctl_rtcache_flush
6375 	},
6376 	{
6377 		.procname	=	"gc_min_interval",
6378 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6379 		.maxlen		=	sizeof(int),
6380 		.mode		=	0644,
6381 		.proc_handler	=	proc_dointvec_jiffies,
6382 	},
6383 	{
6384 		.procname	=	"gc_timeout",
6385 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6386 		.maxlen		=	sizeof(int),
6387 		.mode		=	0644,
6388 		.proc_handler	=	proc_dointvec_jiffies,
6389 	},
6390 	{
6391 		.procname	=	"gc_interval",
6392 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_interval,
6393 		.maxlen		=	sizeof(int),
6394 		.mode		=	0644,
6395 		.proc_handler	=	proc_dointvec_jiffies,
6396 	},
6397 	{
6398 		.procname	=	"gc_elasticity",
6399 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6400 		.maxlen		=	sizeof(int),
6401 		.mode		=	0644,
6402 		.proc_handler	=	proc_dointvec,
6403 	},
6404 	{
6405 		.procname	=	"mtu_expires",
6406 		.data		=	&init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6407 		.maxlen		=	sizeof(int),
6408 		.mode		=	0644,
6409 		.proc_handler	=	proc_dointvec_jiffies,
6410 	},
6411 	{
6412 		.procname	=	"min_adv_mss",
6413 		.data		=	&init_net.ipv6.sysctl.ip6_rt_min_advmss,
6414 		.maxlen		=	sizeof(int),
6415 		.mode		=	0644,
6416 		.proc_handler	=	proc_dointvec,
6417 	},
6418 	{
6419 		.procname	=	"gc_min_interval_ms",
6420 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6421 		.maxlen		=	sizeof(int),
6422 		.mode		=	0644,
6423 		.proc_handler	=	proc_dointvec_ms_jiffies,
6424 	},
6425 	{
6426 		.procname	=	"skip_notify_on_dev_down",
6427 		.data		=	&init_net.ipv6.sysctl.skip_notify_on_dev_down,
6428 		.maxlen		=	sizeof(u8),
6429 		.mode		=	0644,
6430 		.proc_handler	=	proc_dou8vec_minmax,
6431 		.extra1		=	SYSCTL_ZERO,
6432 		.extra2		=	SYSCTL_ONE,
6433 	},
6434 };
6435 
ipv6_route_sysctl_init(struct net * net)6436 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6437 {
6438 	struct ctl_table *table;
6439 
6440 	table = kmemdup(ipv6_route_table_template,
6441 			sizeof(ipv6_route_table_template),
6442 			GFP_KERNEL);
6443 
6444 	if (table) {
6445 		table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6446 		table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6447 		table[2].data = &net->ipv6.sysctl.flush_delay;
6448 		table[2].extra1 = net;
6449 		table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6450 		table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6451 		table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6452 		table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6453 		table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6454 		table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6455 		table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6456 		table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6457 	}
6458 
6459 	return table;
6460 }
6461 
ipv6_route_sysctl_table_size(struct net * net)6462 size_t ipv6_route_sysctl_table_size(struct net *net)
6463 {
6464 	/* Don't export sysctls to unprivileged users */
6465 	if (net->user_ns != &init_user_ns)
6466 		return 1;
6467 
6468 	return ARRAY_SIZE(ipv6_route_table_template);
6469 }
6470 #endif
6471 
ip6_route_net_init(struct net * net)6472 static int __net_init ip6_route_net_init(struct net *net)
6473 {
6474 	int ret = -ENOMEM;
6475 
6476 	memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6477 	       sizeof(net->ipv6.ip6_dst_ops));
6478 
6479 	if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6480 		goto out_ip6_dst_ops;
6481 
6482 	net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6483 	if (!net->ipv6.fib6_null_entry)
6484 		goto out_ip6_dst_entries;
6485 	memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6486 	       sizeof(*net->ipv6.fib6_null_entry));
6487 
6488 	net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6489 					   sizeof(*net->ipv6.ip6_null_entry),
6490 					   GFP_KERNEL);
6491 	if (!net->ipv6.ip6_null_entry)
6492 		goto out_fib6_null_entry;
6493 	net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6494 	dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6495 			 ip6_template_metrics, true);
6496 	INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6497 
6498 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6499 	net->ipv6.fib6_has_custom_rules = false;
6500 	net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6501 					       sizeof(*net->ipv6.ip6_prohibit_entry),
6502 					       GFP_KERNEL);
6503 	if (!net->ipv6.ip6_prohibit_entry)
6504 		goto out_ip6_null_entry;
6505 	net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6506 	dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6507 			 ip6_template_metrics, true);
6508 	INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6509 
6510 	net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6511 					       sizeof(*net->ipv6.ip6_blk_hole_entry),
6512 					       GFP_KERNEL);
6513 	if (!net->ipv6.ip6_blk_hole_entry)
6514 		goto out_ip6_prohibit_entry;
6515 	net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6516 	dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6517 			 ip6_template_metrics, true);
6518 	INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6519 #ifdef CONFIG_IPV6_SUBTREES
6520 	net->ipv6.fib6_routes_require_src = 0;
6521 #endif
6522 #endif
6523 
6524 	net->ipv6.sysctl.flush_delay = 0;
6525 	net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6526 	net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6527 	net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6528 	net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6529 	net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6530 	net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6531 	net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6532 	net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6533 
6534 	atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6535 
6536 	ret = 0;
6537 out:
6538 	return ret;
6539 
6540 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6541 out_ip6_prohibit_entry:
6542 	kfree(net->ipv6.ip6_prohibit_entry);
6543 out_ip6_null_entry:
6544 	kfree(net->ipv6.ip6_null_entry);
6545 #endif
6546 out_fib6_null_entry:
6547 	kfree(net->ipv6.fib6_null_entry);
6548 out_ip6_dst_entries:
6549 	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6550 out_ip6_dst_ops:
6551 	goto out;
6552 }
6553 
ip6_route_net_exit(struct net * net)6554 static void __net_exit ip6_route_net_exit(struct net *net)
6555 {
6556 	kfree(net->ipv6.fib6_null_entry);
6557 	kfree(net->ipv6.ip6_null_entry);
6558 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6559 	kfree(net->ipv6.ip6_prohibit_entry);
6560 	kfree(net->ipv6.ip6_blk_hole_entry);
6561 #endif
6562 	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6563 }
6564 
ip6_route_net_init_late(struct net * net)6565 static int __net_init ip6_route_net_init_late(struct net *net)
6566 {
6567 #ifdef CONFIG_PROC_FS
6568 	if (!proc_create_net("ipv6_route", 0, net->proc_net,
6569 			     &ipv6_route_seq_ops,
6570 			     sizeof(struct ipv6_route_iter)))
6571 		return -ENOMEM;
6572 
6573 	if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6574 				    rt6_stats_seq_show, NULL)) {
6575 		remove_proc_entry("ipv6_route", net->proc_net);
6576 		return -ENOMEM;
6577 	}
6578 #endif
6579 	return 0;
6580 }
6581 
ip6_route_net_exit_late(struct net * net)6582 static void __net_exit ip6_route_net_exit_late(struct net *net)
6583 {
6584 #ifdef CONFIG_PROC_FS
6585 	remove_proc_entry("ipv6_route", net->proc_net);
6586 	remove_proc_entry("rt6_stats", net->proc_net);
6587 #endif
6588 }
6589 
6590 static struct pernet_operations ip6_route_net_ops = {
6591 	.init = ip6_route_net_init,
6592 	.exit = ip6_route_net_exit,
6593 };
6594 
ipv6_inetpeer_init(struct net * net)6595 static int __net_init ipv6_inetpeer_init(struct net *net)
6596 {
6597 	struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6598 
6599 	if (!bp)
6600 		return -ENOMEM;
6601 	inet_peer_base_init(bp);
6602 	net->ipv6.peers = bp;
6603 	return 0;
6604 }
6605 
ipv6_inetpeer_exit(struct net * net)6606 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6607 {
6608 	struct inet_peer_base *bp = net->ipv6.peers;
6609 
6610 	net->ipv6.peers = NULL;
6611 	inetpeer_invalidate_tree(bp);
6612 	kfree(bp);
6613 }
6614 
6615 static struct pernet_operations ipv6_inetpeer_ops = {
6616 	.init	=	ipv6_inetpeer_init,
6617 	.exit	=	ipv6_inetpeer_exit,
6618 };
6619 
6620 static struct pernet_operations ip6_route_net_late_ops = {
6621 	.init = ip6_route_net_init_late,
6622 	.exit = ip6_route_net_exit_late,
6623 };
6624 
6625 static struct notifier_block ip6_route_dev_notifier = {
6626 	.notifier_call = ip6_route_dev_notify,
6627 	.priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6628 };
6629 
ip6_route_init_special_entries(void)6630 void __init ip6_route_init_special_entries(void)
6631 {
6632 	/* Registering of the loopback is done before this portion of code,
6633 	 * the loopback reference in rt6_info will not be taken, do it
6634 	 * manually for init_net */
6635 	init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6636 	init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6637 	init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6638   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6639 	init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6640 	init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6641 	init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6642 	init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6643   #endif
6644 }
6645 
6646 #if IS_BUILTIN(CONFIG_IPV6)
6647 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6648 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6649 
6650 BTF_ID_LIST(btf_fib6_info_id)
6651 BTF_ID(struct, fib6_info)
6652 
6653 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6654 	.seq_ops		= &ipv6_route_seq_ops,
6655 	.init_seq_private	= bpf_iter_init_seq_net,
6656 	.fini_seq_private	= bpf_iter_fini_seq_net,
6657 	.seq_priv_size		= sizeof(struct ipv6_route_iter),
6658 };
6659 
6660 static struct bpf_iter_reg ipv6_route_reg_info = {
6661 	.target			= "ipv6_route",
6662 	.ctx_arg_info_size	= 1,
6663 	.ctx_arg_info		= {
6664 		{ offsetof(struct bpf_iter__ipv6_route, rt),
6665 		  PTR_TO_BTF_ID_OR_NULL },
6666 	},
6667 	.seq_info		= &ipv6_route_seq_info,
6668 };
6669 
bpf_iter_register(void)6670 static int __init bpf_iter_register(void)
6671 {
6672 	ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6673 	return bpf_iter_reg_target(&ipv6_route_reg_info);
6674 }
6675 
bpf_iter_unregister(void)6676 static void bpf_iter_unregister(void)
6677 {
6678 	bpf_iter_unreg_target(&ipv6_route_reg_info);
6679 }
6680 #endif
6681 #endif
6682 
ip6_route_init(void)6683 int __init ip6_route_init(void)
6684 {
6685 	int ret;
6686 	int cpu;
6687 
6688 	ret = -ENOMEM;
6689 	ip6_dst_ops_template.kmem_cachep =
6690 		kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6691 				  SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6692 	if (!ip6_dst_ops_template.kmem_cachep)
6693 		goto out;
6694 
6695 	ret = dst_entries_init(&ip6_dst_blackhole_ops);
6696 	if (ret)
6697 		goto out_kmem_cache;
6698 
6699 	ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6700 	if (ret)
6701 		goto out_dst_entries;
6702 
6703 	ret = register_pernet_subsys(&ip6_route_net_ops);
6704 	if (ret)
6705 		goto out_register_inetpeer;
6706 
6707 	ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6708 
6709 	ret = fib6_init();
6710 	if (ret)
6711 		goto out_register_subsys;
6712 
6713 	ret = xfrm6_init();
6714 	if (ret)
6715 		goto out_fib6_init;
6716 
6717 	ret = fib6_rules_init();
6718 	if (ret)
6719 		goto xfrm6_init;
6720 
6721 	ret = register_pernet_subsys(&ip6_route_net_late_ops);
6722 	if (ret)
6723 		goto fib6_rules_init;
6724 
6725 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6726 				   inet6_rtm_newroute, NULL, 0);
6727 	if (ret < 0)
6728 		goto out_register_late_subsys;
6729 
6730 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6731 				   inet6_rtm_delroute, NULL, 0);
6732 	if (ret < 0)
6733 		goto out_register_late_subsys;
6734 
6735 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6736 				   inet6_rtm_getroute, NULL,
6737 				   RTNL_FLAG_DOIT_UNLOCKED);
6738 	if (ret < 0)
6739 		goto out_register_late_subsys;
6740 
6741 	ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6742 	if (ret)
6743 		goto out_register_late_subsys;
6744 
6745 #if IS_BUILTIN(CONFIG_IPV6)
6746 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6747 	ret = bpf_iter_register();
6748 	if (ret)
6749 		goto out_register_late_subsys;
6750 #endif
6751 #endif
6752 
6753 	for_each_possible_cpu(cpu) {
6754 		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6755 
6756 		INIT_LIST_HEAD(&ul->head);
6757 		spin_lock_init(&ul->lock);
6758 	}
6759 
6760 out:
6761 	return ret;
6762 
6763 out_register_late_subsys:
6764 	rtnl_unregister_all(PF_INET6);
6765 	unregister_pernet_subsys(&ip6_route_net_late_ops);
6766 fib6_rules_init:
6767 	fib6_rules_cleanup();
6768 xfrm6_init:
6769 	xfrm6_fini();
6770 out_fib6_init:
6771 	fib6_gc_cleanup();
6772 out_register_subsys:
6773 	unregister_pernet_subsys(&ip6_route_net_ops);
6774 out_register_inetpeer:
6775 	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6776 out_dst_entries:
6777 	dst_entries_destroy(&ip6_dst_blackhole_ops);
6778 out_kmem_cache:
6779 	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6780 	goto out;
6781 }
6782 
ip6_route_cleanup(void)6783 void ip6_route_cleanup(void)
6784 {
6785 #if IS_BUILTIN(CONFIG_IPV6)
6786 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6787 	bpf_iter_unregister();
6788 #endif
6789 #endif
6790 	unregister_netdevice_notifier(&ip6_route_dev_notifier);
6791 	unregister_pernet_subsys(&ip6_route_net_late_ops);
6792 	fib6_rules_cleanup();
6793 	xfrm6_fini();
6794 	fib6_gc_cleanup();
6795 	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6796 	unregister_pernet_subsys(&ip6_route_net_ops);
6797 	dst_entries_destroy(&ip6_dst_blackhole_ops);
6798 	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6799 }
6800