xref: /linux/net/ipv6/ndisc.c (revision af2d6148d2a159e1a0862bce5a2c88c1618a2b27)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Neighbour Discovery for IPv6
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Mike Shaver		<shaver@ingenia.com>
9  */
10 
11 /*
12  *	Changes:
13  *
14  *	Alexey I. Froloff		:	RFC6106 (DNSSL) support
15  *	Pierre Ynard			:	export userland ND options
16  *						through netlink (RDNSS support)
17  *	Lars Fenneberg			:	fixed MTU setting on receipt
18  *						of an RA.
19  *	Janos Farkas			:	kmalloc failure checks
20  *	Alexey Kuznetsov		:	state machine reworked
21  *						and moved to net/core.
22  *	Pekka Savola			:	RFC2461 validation
23  *	YOSHIFUJI Hideaki @USAGI	:	Verify ND options properly
24  */
25 
26 #define pr_fmt(fmt) "ICMPv6: " fmt
27 
28 #include <linux/module.h>
29 #include <linux/errno.h>
30 #include <linux/types.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/sched.h>
34 #include <linux/net.h>
35 #include <linux/in6.h>
36 #include <linux/route.h>
37 #include <linux/init.h>
38 #include <linux/rcupdate.h>
39 #include <linux/slab.h>
40 #ifdef CONFIG_SYSCTL
41 #include <linux/sysctl.h>
42 #endif
43 
44 #include <linux/if_addr.h>
45 #include <linux/if_ether.h>
46 #include <linux/if_arp.h>
47 #include <linux/ipv6.h>
48 #include <linux/icmpv6.h>
49 #include <linux/jhash.h>
50 
51 #include <net/sock.h>
52 #include <net/snmp.h>
53 
54 #include <net/ipv6.h>
55 #include <net/protocol.h>
56 #include <net/ndisc.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #include <net/icmp.h>
60 
61 #include <net/netlink.h>
62 #include <linux/rtnetlink.h>
63 
64 #include <net/flow.h>
65 #include <net/ip6_checksum.h>
66 #include <net/inet_common.h>
67 #include <linux/proc_fs.h>
68 
69 #include <linux/netfilter.h>
70 #include <linux/netfilter_ipv6.h>
71 
72 static u32 ndisc_hash(const void *pkey,
73 		      const struct net_device *dev,
74 		      __u32 *hash_rnd);
75 static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
76 static bool ndisc_allow_add(const struct net_device *dev,
77 			    struct netlink_ext_ack *extack);
78 static int ndisc_constructor(struct neighbour *neigh);
79 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
80 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
81 static int pndisc_constructor(struct pneigh_entry *n);
82 static void pndisc_destructor(struct pneigh_entry *n);
83 static void pndisc_redo(struct sk_buff *skb);
84 static int ndisc_is_multicast(const void *pkey);
85 
86 static const struct neigh_ops ndisc_generic_ops = {
87 	.family =		AF_INET6,
88 	.solicit =		ndisc_solicit,
89 	.error_report =		ndisc_error_report,
90 	.output =		neigh_resolve_output,
91 	.connected_output =	neigh_connected_output,
92 };
93 
94 static const struct neigh_ops ndisc_hh_ops = {
95 	.family =		AF_INET6,
96 	.solicit =		ndisc_solicit,
97 	.error_report =		ndisc_error_report,
98 	.output =		neigh_resolve_output,
99 	.connected_output =	neigh_resolve_output,
100 };
101 
102 
103 static const struct neigh_ops ndisc_direct_ops = {
104 	.family =		AF_INET6,
105 	.output =		neigh_direct_output,
106 	.connected_output =	neigh_direct_output,
107 };
108 
109 struct neigh_table nd_tbl = {
110 	.family =	AF_INET6,
111 	.key_len =	sizeof(struct in6_addr),
112 	.protocol =	cpu_to_be16(ETH_P_IPV6),
113 	.hash =		ndisc_hash,
114 	.key_eq =	ndisc_key_eq,
115 	.constructor =	ndisc_constructor,
116 	.pconstructor =	pndisc_constructor,
117 	.pdestructor =	pndisc_destructor,
118 	.proxy_redo =	pndisc_redo,
119 	.is_multicast =	ndisc_is_multicast,
120 	.allow_add  =   ndisc_allow_add,
121 	.id =		"ndisc_cache",
122 	.parms = {
123 		.tbl			= &nd_tbl,
124 		.reachable_time		= ND_REACHABLE_TIME,
125 		.data = {
126 			[NEIGH_VAR_MCAST_PROBES] = 3,
127 			[NEIGH_VAR_UCAST_PROBES] = 3,
128 			[NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER,
129 			[NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME,
130 			[NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
131 			[NEIGH_VAR_INTERVAL_PROBE_TIME_MS] = 5 * HZ,
132 			[NEIGH_VAR_GC_STALETIME] = 60 * HZ,
133 			[NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
134 			[NEIGH_VAR_PROXY_QLEN] = 64,
135 			[NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
136 			[NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
137 		},
138 	},
139 	.gc_interval =	  30 * HZ,
140 	.gc_thresh1 =	 128,
141 	.gc_thresh2 =	 512,
142 	.gc_thresh3 =	1024,
143 };
144 EXPORT_SYMBOL_GPL(nd_tbl);
145 
146 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, const void *data,
147 			      int data_len, int pad)
148 {
149 	int space = __ndisc_opt_addr_space(data_len, pad);
150 	u8 *opt = skb_put(skb, space);
151 
152 	opt[0] = type;
153 	opt[1] = space>>3;
154 
155 	memset(opt + 2, 0, pad);
156 	opt   += pad;
157 	space -= pad;
158 
159 	memcpy(opt+2, data, data_len);
160 	data_len += 2;
161 	opt += data_len;
162 	space -= data_len;
163 	if (space > 0)
164 		memset(opt, 0, space);
165 }
166 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
167 
168 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
169 					  const void *data, u8 icmp6_type)
170 {
171 	__ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
172 				 ndisc_addr_option_pad(skb->dev->type));
173 	ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
174 }
175 
176 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
177 						   void *ha,
178 						   const u8 *ops_data)
179 {
180 	ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
181 	ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
182 }
183 
184 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
185 					    struct nd_opt_hdr *end)
186 {
187 	int type;
188 	if (!cur || !end || cur >= end)
189 		return NULL;
190 	type = cur->nd_opt_type;
191 	do {
192 		cur = ((void *)cur) + (cur->nd_opt_len << 3);
193 	} while (cur < end && cur->nd_opt_type != type);
194 	return cur <= end && cur->nd_opt_type == type ? cur : NULL;
195 }
196 
197 static inline int ndisc_is_useropt(const struct net_device *dev,
198 				   struct nd_opt_hdr *opt)
199 {
200 	return opt->nd_opt_type == ND_OPT_PREFIX_INFO ||
201 		opt->nd_opt_type == ND_OPT_RDNSS ||
202 		opt->nd_opt_type == ND_OPT_DNSSL ||
203 		opt->nd_opt_type == ND_OPT_6CO ||
204 		opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
205 		opt->nd_opt_type == ND_OPT_PREF64;
206 }
207 
208 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
209 					     struct nd_opt_hdr *cur,
210 					     struct nd_opt_hdr *end)
211 {
212 	if (!cur || !end || cur >= end)
213 		return NULL;
214 	do {
215 		cur = ((void *)cur) + (cur->nd_opt_len << 3);
216 	} while (cur < end && !ndisc_is_useropt(dev, cur));
217 	return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
218 }
219 
220 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
221 					  u8 *opt, int opt_len,
222 					  struct ndisc_options *ndopts)
223 {
224 	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
225 
226 	if (!nd_opt || opt_len < 0 || !ndopts)
227 		return NULL;
228 	memset(ndopts, 0, sizeof(*ndopts));
229 	while (opt_len) {
230 		bool unknown = false;
231 		int l;
232 		if (opt_len < sizeof(struct nd_opt_hdr))
233 			return NULL;
234 		l = nd_opt->nd_opt_len << 3;
235 		if (opt_len < l || l == 0)
236 			return NULL;
237 		if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
238 			goto next_opt;
239 		switch (nd_opt->nd_opt_type) {
240 		case ND_OPT_SOURCE_LL_ADDR:
241 		case ND_OPT_TARGET_LL_ADDR:
242 		case ND_OPT_MTU:
243 		case ND_OPT_NONCE:
244 		case ND_OPT_REDIRECT_HDR:
245 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
246 				net_dbg_ratelimited("%s: duplicated ND6 option found: type=%d\n",
247 						    __func__, nd_opt->nd_opt_type);
248 			} else {
249 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
250 			}
251 			break;
252 		case ND_OPT_PREFIX_INFO:
253 			ndopts->nd_opts_pi_end = nd_opt;
254 			if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
255 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
256 			break;
257 #ifdef CONFIG_IPV6_ROUTE_INFO
258 		case ND_OPT_ROUTE_INFO:
259 			ndopts->nd_opts_ri_end = nd_opt;
260 			if (!ndopts->nd_opts_ri)
261 				ndopts->nd_opts_ri = nd_opt;
262 			break;
263 #endif
264 		default:
265 			unknown = true;
266 		}
267 		if (ndisc_is_useropt(dev, nd_opt)) {
268 			ndopts->nd_useropts_end = nd_opt;
269 			if (!ndopts->nd_useropts)
270 				ndopts->nd_useropts = nd_opt;
271 		} else if (unknown) {
272 			/*
273 			 * Unknown options must be silently ignored,
274 			 * to accommodate future extension to the
275 			 * protocol.
276 			 */
277 			net_dbg_ratelimited("%s: ignored unsupported option; type=%d, len=%d\n",
278 					    __func__, nd_opt->nd_opt_type, nd_opt->nd_opt_len);
279 		}
280 next_opt:
281 		opt_len -= l;
282 		nd_opt = ((void *)nd_opt) + l;
283 	}
284 	return ndopts;
285 }
286 
287 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
288 {
289 	switch (dev->type) {
290 	case ARPHRD_ETHER:
291 	case ARPHRD_IEEE802:	/* Not sure. Check it later. --ANK */
292 	case ARPHRD_FDDI:
293 		ipv6_eth_mc_map(addr, buf);
294 		return 0;
295 	case ARPHRD_ARCNET:
296 		ipv6_arcnet_mc_map(addr, buf);
297 		return 0;
298 	case ARPHRD_INFINIBAND:
299 		ipv6_ib_mc_map(addr, dev->broadcast, buf);
300 		return 0;
301 	case ARPHRD_IPGRE:
302 		return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
303 	default:
304 		if (dir) {
305 			memcpy(buf, dev->broadcast, dev->addr_len);
306 			return 0;
307 		}
308 	}
309 	return -EINVAL;
310 }
311 EXPORT_SYMBOL(ndisc_mc_map);
312 
313 static u32 ndisc_hash(const void *pkey,
314 		      const struct net_device *dev,
315 		      __u32 *hash_rnd)
316 {
317 	return ndisc_hashfn(pkey, dev, hash_rnd);
318 }
319 
320 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
321 {
322 	return neigh_key_eq128(n, pkey);
323 }
324 
325 static int ndisc_constructor(struct neighbour *neigh)
326 {
327 	struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
328 	struct net_device *dev = neigh->dev;
329 	struct inet6_dev *in6_dev;
330 	struct neigh_parms *parms;
331 	bool is_multicast = ipv6_addr_is_multicast(addr);
332 
333 	in6_dev = in6_dev_get(dev);
334 	if (!in6_dev) {
335 		return -EINVAL;
336 	}
337 
338 	parms = in6_dev->nd_parms;
339 	__neigh_parms_put(neigh->parms);
340 	neigh->parms = neigh_parms_clone(parms);
341 
342 	neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
343 	if (!dev->header_ops) {
344 		neigh->nud_state = NUD_NOARP;
345 		neigh->ops = &ndisc_direct_ops;
346 		neigh->output = neigh_direct_output;
347 	} else {
348 		if (is_multicast) {
349 			neigh->nud_state = NUD_NOARP;
350 			ndisc_mc_map(addr, neigh->ha, dev, 1);
351 		} else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
352 			neigh->nud_state = NUD_NOARP;
353 			memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
354 			if (dev->flags&IFF_LOOPBACK)
355 				neigh->type = RTN_LOCAL;
356 		} else if (dev->flags&IFF_POINTOPOINT) {
357 			neigh->nud_state = NUD_NOARP;
358 			memcpy(neigh->ha, dev->broadcast, dev->addr_len);
359 		}
360 		if (dev->header_ops->cache)
361 			neigh->ops = &ndisc_hh_ops;
362 		else
363 			neigh->ops = &ndisc_generic_ops;
364 		if (neigh->nud_state&NUD_VALID)
365 			neigh->output = neigh->ops->connected_output;
366 		else
367 			neigh->output = neigh->ops->output;
368 	}
369 	in6_dev_put(in6_dev);
370 	return 0;
371 }
372 
373 static int pndisc_constructor(struct pneigh_entry *n)
374 {
375 	struct in6_addr *addr = (struct in6_addr *)&n->key;
376 	struct net_device *dev = n->dev;
377 	struct in6_addr maddr;
378 
379 	if (!dev)
380 		return -EINVAL;
381 
382 	addrconf_addr_solict_mult(addr, &maddr);
383 	return ipv6_dev_mc_inc(dev, &maddr);
384 }
385 
386 static void pndisc_destructor(struct pneigh_entry *n)
387 {
388 	struct in6_addr *addr = (struct in6_addr *)&n->key;
389 	struct net_device *dev = n->dev;
390 	struct in6_addr maddr;
391 
392 	if (!dev)
393 		return;
394 
395 	addrconf_addr_solict_mult(addr, &maddr);
396 	ipv6_dev_mc_dec(dev, &maddr);
397 }
398 
399 /* called with rtnl held */
400 static bool ndisc_allow_add(const struct net_device *dev,
401 			    struct netlink_ext_ack *extack)
402 {
403 	struct inet6_dev *idev = __in6_dev_get(dev);
404 
405 	if (!idev || idev->cnf.disable_ipv6) {
406 		NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
407 		return false;
408 	}
409 
410 	return true;
411 }
412 
413 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
414 				       int len)
415 {
416 	int hlen = LL_RESERVED_SPACE(dev);
417 	int tlen = dev->needed_tailroom;
418 	struct sk_buff *skb;
419 
420 	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
421 	if (!skb)
422 		return NULL;
423 
424 	skb->protocol = htons(ETH_P_IPV6);
425 	skb->dev = dev;
426 
427 	skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
428 	skb_reset_transport_header(skb);
429 
430 	/* Manually assign socket ownership as we avoid calling
431 	 * sock_alloc_send_pskb() to bypass wmem buffer limits
432 	 */
433 	rcu_read_lock();
434 	skb_set_owner_w(skb, dev_net_rcu(dev)->ipv6.ndisc_sk);
435 	rcu_read_unlock();
436 
437 	return skb;
438 }
439 
440 static void ip6_nd_hdr(struct sk_buff *skb,
441 		       const struct in6_addr *saddr,
442 		       const struct in6_addr *daddr,
443 		       int hop_limit, int len)
444 {
445 	struct ipv6hdr *hdr;
446 	struct inet6_dev *idev;
447 	unsigned tclass;
448 
449 	rcu_read_lock();
450 	idev = __in6_dev_get(skb->dev);
451 	tclass = idev ? READ_ONCE(idev->cnf.ndisc_tclass) : 0;
452 	rcu_read_unlock();
453 
454 	skb_push(skb, sizeof(*hdr));
455 	skb_reset_network_header(skb);
456 	hdr = ipv6_hdr(skb);
457 
458 	ip6_flow_hdr(hdr, tclass, 0);
459 
460 	hdr->payload_len = htons(len);
461 	hdr->nexthdr = IPPROTO_ICMPV6;
462 	hdr->hop_limit = hop_limit;
463 
464 	hdr->saddr = *saddr;
465 	hdr->daddr = *daddr;
466 }
467 
468 void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
469 		    const struct in6_addr *saddr)
470 {
471 	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
472 	struct dst_entry *dst = skb_dst(skb);
473 	struct net_device *dev;
474 	struct inet6_dev *idev;
475 	struct net *net;
476 	struct sock *sk;
477 	int err;
478 	u8 type;
479 
480 	type = icmp6h->icmp6_type;
481 
482 	rcu_read_lock();
483 
484 	net = dev_net_rcu(skb->dev);
485 	sk = net->ipv6.ndisc_sk;
486 	if (!dst) {
487 		struct flowi6 fl6;
488 		int oif = skb->dev->ifindex;
489 
490 		icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
491 		dst = icmp6_dst_alloc(skb->dev, &fl6);
492 		if (IS_ERR(dst)) {
493 			rcu_read_unlock();
494 			kfree_skb(skb);
495 			return;
496 		}
497 
498 		skb_dst_set(skb, dst);
499 	}
500 
501 	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
502 					      IPPROTO_ICMPV6,
503 					      csum_partial(icmp6h,
504 							   skb->len, 0));
505 
506 	ip6_nd_hdr(skb, saddr, daddr, READ_ONCE(inet6_sk(sk)->hop_limit), skb->len);
507 
508 	dev = dst_dev(dst);
509 	idev = __in6_dev_get(dev);
510 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
511 
512 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
513 		      net, sk, skb, NULL, dev,
514 		      dst_output);
515 	if (!err) {
516 		ICMP6MSGOUT_INC_STATS(net, idev, type);
517 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
518 	}
519 
520 	rcu_read_unlock();
521 }
522 EXPORT_SYMBOL(ndisc_send_skb);
523 
524 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
525 		   const struct in6_addr *solicited_addr,
526 		   bool router, bool solicited, bool override, bool inc_opt)
527 {
528 	struct sk_buff *skb;
529 	struct in6_addr tmpaddr;
530 	struct inet6_ifaddr *ifp;
531 	const struct in6_addr *src_addr;
532 	struct nd_msg *msg;
533 	int optlen = 0;
534 
535 	/* for anycast or proxy, solicited_addr != src_addr */
536 	ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
537 	if (ifp) {
538 		src_addr = solicited_addr;
539 		if (ifp->flags & IFA_F_OPTIMISTIC)
540 			override = false;
541 		inc_opt |= READ_ONCE(ifp->idev->cnf.force_tllao);
542 		in6_ifa_put(ifp);
543 	} else {
544 		if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
545 				       inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
546 				       &tmpaddr))
547 			return;
548 		src_addr = &tmpaddr;
549 	}
550 
551 	if (!dev->addr_len)
552 		inc_opt = false;
553 	if (inc_opt)
554 		optlen += ndisc_opt_addr_space(dev,
555 					       NDISC_NEIGHBOUR_ADVERTISEMENT);
556 
557 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
558 	if (!skb)
559 		return;
560 
561 	msg = skb_put(skb, sizeof(*msg));
562 	*msg = (struct nd_msg) {
563 		.icmph = {
564 			.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
565 			.icmp6_router = router,
566 			.icmp6_solicited = solicited,
567 			.icmp6_override = override,
568 		},
569 		.target = *solicited_addr,
570 	};
571 
572 	if (inc_opt)
573 		ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
574 				       dev->dev_addr,
575 				       NDISC_NEIGHBOUR_ADVERTISEMENT);
576 
577 	ndisc_send_skb(skb, daddr, src_addr);
578 }
579 
580 static void ndisc_send_unsol_na(struct net_device *dev)
581 {
582 	struct inet6_dev *idev;
583 	struct inet6_ifaddr *ifa;
584 
585 	idev = in6_dev_get(dev);
586 	if (!idev)
587 		return;
588 
589 	read_lock_bh(&idev->lock);
590 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
591 		/* skip tentative addresses until dad completes */
592 		if (ifa->flags & IFA_F_TENTATIVE &&
593 		    !(ifa->flags & IFA_F_OPTIMISTIC))
594 			continue;
595 
596 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
597 			      /*router=*/ !!idev->cnf.forwarding,
598 			      /*solicited=*/ false, /*override=*/ true,
599 			      /*inc_opt=*/ true);
600 	}
601 	read_unlock_bh(&idev->lock);
602 
603 	in6_dev_put(idev);
604 }
605 
606 struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
607 				const struct in6_addr *saddr, u64 nonce)
608 {
609 	int inc_opt = dev->addr_len;
610 	struct sk_buff *skb;
611 	struct nd_msg *msg;
612 	int optlen = 0;
613 
614 	if (!saddr)
615 		return NULL;
616 
617 	if (ipv6_addr_any(saddr))
618 		inc_opt = false;
619 	if (inc_opt)
620 		optlen += ndisc_opt_addr_space(dev,
621 					       NDISC_NEIGHBOUR_SOLICITATION);
622 	if (nonce != 0)
623 		optlen += 8;
624 
625 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
626 	if (!skb)
627 		return NULL;
628 
629 	msg = skb_put(skb, sizeof(*msg));
630 	*msg = (struct nd_msg) {
631 		.icmph = {
632 			.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
633 		},
634 		.target = *solicit,
635 	};
636 
637 	if (inc_opt)
638 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
639 				       dev->dev_addr,
640 				       NDISC_NEIGHBOUR_SOLICITATION);
641 	if (nonce != 0) {
642 		u8 *opt = skb_put(skb, 8);
643 
644 		opt[0] = ND_OPT_NONCE;
645 		opt[1] = 8 >> 3;
646 		memcpy(opt + 2, &nonce, 6);
647 	}
648 
649 	return skb;
650 }
651 EXPORT_SYMBOL(ndisc_ns_create);
652 
653 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
654 		   const struct in6_addr *daddr, const struct in6_addr *saddr,
655 		   u64 nonce)
656 {
657 	struct in6_addr addr_buf;
658 	struct sk_buff *skb;
659 
660 	if (!saddr) {
661 		if (ipv6_get_lladdr(dev, &addr_buf,
662 				    (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
663 			return;
664 		saddr = &addr_buf;
665 	}
666 
667 	skb = ndisc_ns_create(dev, solicit, saddr, nonce);
668 
669 	if (skb)
670 		ndisc_send_skb(skb, daddr, saddr);
671 }
672 
673 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
674 		   const struct in6_addr *daddr)
675 {
676 	struct sk_buff *skb;
677 	struct rs_msg *msg;
678 	int send_sllao = dev->addr_len;
679 	int optlen = 0;
680 
681 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
682 	/*
683 	 * According to section 2.2 of RFC 4429, we must not
684 	 * send router solicitations with a sllao from
685 	 * optimistic addresses, but we may send the solicitation
686 	 * if we don't include the sllao.  So here we check
687 	 * if our address is optimistic, and if so, we
688 	 * suppress the inclusion of the sllao.
689 	 */
690 	if (send_sllao) {
691 		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
692 							   dev, 1);
693 		if (ifp) {
694 			if (ifp->flags & IFA_F_OPTIMISTIC)  {
695 				send_sllao = 0;
696 			}
697 			in6_ifa_put(ifp);
698 		} else {
699 			send_sllao = 0;
700 		}
701 	}
702 #endif
703 	if (send_sllao)
704 		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
705 
706 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
707 	if (!skb)
708 		return;
709 
710 	msg = skb_put(skb, sizeof(*msg));
711 	*msg = (struct rs_msg) {
712 		.icmph = {
713 			.icmp6_type = NDISC_ROUTER_SOLICITATION,
714 		},
715 	};
716 
717 	if (send_sllao)
718 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
719 				       dev->dev_addr,
720 				       NDISC_ROUTER_SOLICITATION);
721 
722 	ndisc_send_skb(skb, daddr, saddr);
723 }
724 
725 
726 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
727 {
728 	/*
729 	 *	"The sender MUST return an ICMP
730 	 *	 destination unreachable"
731 	 */
732 	dst_link_failure(skb);
733 	kfree_skb(skb);
734 }
735 
736 /* Called with locked neigh: either read or both */
737 
738 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
739 {
740 	struct in6_addr *saddr = NULL;
741 	struct in6_addr mcaddr;
742 	struct net_device *dev = neigh->dev;
743 	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
744 	int probes = atomic_read(&neigh->probes);
745 
746 	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
747 					   dev, false, 1,
748 					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
749 		saddr = &ipv6_hdr(skb)->saddr;
750 	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
751 	if (probes < 0) {
752 		if (!(READ_ONCE(neigh->nud_state) & NUD_VALID)) {
753 			net_dbg_ratelimited("%s: trying to ucast probe in NUD_INVALID: %pI6\n",
754 					    __func__, target);
755 		}
756 		ndisc_send_ns(dev, target, target, saddr, 0);
757 	} else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
758 		neigh_app_ns(neigh);
759 	} else {
760 		addrconf_addr_solict_mult(target, &mcaddr);
761 		ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
762 	}
763 }
764 
765 static int pndisc_is_router(const void *pkey,
766 			    struct net_device *dev)
767 {
768 	struct pneigh_entry *n;
769 	int ret = -1;
770 
771 	read_lock_bh(&nd_tbl.lock);
772 	n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
773 	if (n)
774 		ret = !!(n->flags & NTF_ROUTER);
775 	read_unlock_bh(&nd_tbl.lock);
776 
777 	return ret;
778 }
779 
780 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
781 		  const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
782 		  struct ndisc_options *ndopts)
783 {
784 	neigh_update(neigh, lladdr, new, flags, 0);
785 	/* report ndisc ops about neighbour update */
786 	ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
787 }
788 
789 static enum skb_drop_reason ndisc_recv_ns(struct sk_buff *skb)
790 {
791 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
792 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
793 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
794 	u8 *lladdr = NULL;
795 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
796 				    offsetof(struct nd_msg, opt));
797 	struct ndisc_options ndopts;
798 	struct net_device *dev = skb->dev;
799 	struct inet6_ifaddr *ifp;
800 	struct inet6_dev *idev = NULL;
801 	struct neighbour *neigh;
802 	int dad = ipv6_addr_any(saddr);
803 	int is_router = -1;
804 	SKB_DR(reason);
805 	u64 nonce = 0;
806 	bool inc;
807 
808 	if (skb->len < sizeof(struct nd_msg))
809 		return SKB_DROP_REASON_PKT_TOO_SMALL;
810 
811 	if (ipv6_addr_is_multicast(&msg->target)) {
812 		net_dbg_ratelimited("NS: multicast target address\n");
813 		return reason;
814 	}
815 
816 	/*
817 	 * RFC2461 7.1.1:
818 	 * DAD has to be destined for solicited node multicast address.
819 	 */
820 	if (dad && !ipv6_addr_is_solict_mult(daddr)) {
821 		net_dbg_ratelimited("NS: bad DAD packet (wrong destination)\n");
822 		return reason;
823 	}
824 
825 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
826 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
827 
828 	if (ndopts.nd_opts_src_lladdr) {
829 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
830 		if (!lladdr) {
831 			net_dbg_ratelimited("NS: invalid link-layer address length\n");
832 			return reason;
833 		}
834 
835 		/* RFC2461 7.1.1:
836 		 *	If the IP source address is the unspecified address,
837 		 *	there MUST NOT be source link-layer address option
838 		 *	in the message.
839 		 */
840 		if (dad) {
841 			net_dbg_ratelimited("NS: bad DAD packet (link-layer address option)\n");
842 			return reason;
843 		}
844 	}
845 	if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
846 		memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
847 
848 	inc = ipv6_addr_is_multicast(daddr);
849 
850 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
851 	if (ifp) {
852 have_ifp:
853 		if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
854 			if (dad) {
855 				if (nonce != 0 && ifp->dad_nonce == nonce) {
856 					u8 *np = (u8 *)&nonce;
857 					/* Matching nonce if looped back */
858 					net_dbg_ratelimited("%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
859 							    ifp->idev->dev->name, &ifp->addr, np);
860 					goto out;
861 				}
862 				/*
863 				 * We are colliding with another node
864 				 * who is doing DAD
865 				 * so fail our DAD process
866 				 */
867 				addrconf_dad_failure(skb, ifp);
868 				return reason;
869 			} else {
870 				/*
871 				 * This is not a dad solicitation.
872 				 * If we are an optimistic node,
873 				 * we should respond.
874 				 * Otherwise, we should ignore it.
875 				 */
876 				if (!(ifp->flags & IFA_F_OPTIMISTIC))
877 					goto out;
878 			}
879 		}
880 
881 		idev = ifp->idev;
882 	} else {
883 		struct net *net = dev_net(dev);
884 
885 		/* perhaps an address on the master device */
886 		if (netif_is_l3_slave(dev)) {
887 			struct net_device *mdev;
888 
889 			mdev = netdev_master_upper_dev_get_rcu(dev);
890 			if (mdev) {
891 				ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
892 				if (ifp)
893 					goto have_ifp;
894 			}
895 		}
896 
897 		idev = in6_dev_get(dev);
898 		if (!idev) {
899 			/* XXX: count this drop? */
900 			return reason;
901 		}
902 
903 		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
904 		    (READ_ONCE(idev->cnf.forwarding) &&
905 		     (READ_ONCE(net->ipv6.devconf_all->proxy_ndp) ||
906 		      READ_ONCE(idev->cnf.proxy_ndp)) &&
907 		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
908 			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
909 			    skb->pkt_type != PACKET_HOST &&
910 			    inc &&
911 			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
912 				/*
913 				 * for anycast or proxy,
914 				 * sender should delay its response
915 				 * by a random time between 0 and
916 				 * MAX_ANYCAST_DELAY_TIME seconds.
917 				 * (RFC2461) -- yoshfuji
918 				 */
919 				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
920 				if (n)
921 					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
922 				goto out;
923 			}
924 		} else {
925 			SKB_DR_SET(reason, IPV6_NDISC_NS_OTHERHOST);
926 			goto out;
927 		}
928 	}
929 
930 	if (is_router < 0)
931 		is_router = READ_ONCE(idev->cnf.forwarding);
932 
933 	if (dad) {
934 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
935 			      !!is_router, false, (ifp != NULL), true);
936 		goto out;
937 	}
938 
939 	if (inc)
940 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
941 	else
942 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
943 
944 	/*
945 	 *	update / create cache entry
946 	 *	for the source address
947 	 */
948 	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
949 			       !inc || lladdr || !dev->addr_len);
950 	if (neigh)
951 		ndisc_update(dev, neigh, lladdr, NUD_STALE,
952 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
953 			     NEIGH_UPDATE_F_OVERRIDE,
954 			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
955 	if (neigh || !dev->header_ops) {
956 		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
957 			      true, (ifp != NULL && inc), inc);
958 		if (neigh)
959 			neigh_release(neigh);
960 		reason = SKB_CONSUMED;
961 	}
962 
963 out:
964 	if (ifp)
965 		in6_ifa_put(ifp);
966 	else
967 		in6_dev_put(idev);
968 	return reason;
969 }
970 
971 static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
972 {
973 	struct inet6_dev *idev = __in6_dev_get(dev);
974 
975 	switch (READ_ONCE(idev->cnf.accept_untracked_na)) {
976 	case 0: /* Don't accept untracked na (absent in neighbor cache) */
977 		return 0;
978 	case 1: /* Create new entries from na if currently untracked */
979 		return 1;
980 	case 2: /* Create new entries from untracked na only if saddr is in the
981 		 * same subnet as an address configured on the interface that
982 		 * received the na
983 		 */
984 		return !!ipv6_chk_prefix(saddr, dev);
985 	default:
986 		return 0;
987 	}
988 }
989 
990 static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)
991 {
992 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
993 	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
994 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
995 	u8 *lladdr = NULL;
996 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
997 				    offsetof(struct nd_msg, opt));
998 	struct ndisc_options ndopts;
999 	struct net_device *dev = skb->dev;
1000 	struct inet6_dev *idev = __in6_dev_get(dev);
1001 	struct inet6_ifaddr *ifp;
1002 	struct neighbour *neigh;
1003 	SKB_DR(reason);
1004 	u8 new_state;
1005 
1006 	if (skb->len < sizeof(struct nd_msg))
1007 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1008 
1009 	if (ipv6_addr_is_multicast(&msg->target)) {
1010 		net_dbg_ratelimited("NA: target address is multicast\n");
1011 		return reason;
1012 	}
1013 
1014 	if (ipv6_addr_is_multicast(daddr) &&
1015 	    msg->icmph.icmp6_solicited) {
1016 		net_dbg_ratelimited("NA: solicited NA is multicasted\n");
1017 		return reason;
1018 	}
1019 
1020 	/* For some 802.11 wireless deployments (and possibly other networks),
1021 	 * there will be a NA proxy and unsolicitd packets are attacks
1022 	 * and thus should not be accepted.
1023 	 * drop_unsolicited_na takes precedence over accept_untracked_na
1024 	 */
1025 	if (!msg->icmph.icmp6_solicited && idev &&
1026 	    READ_ONCE(idev->cnf.drop_unsolicited_na))
1027 		return reason;
1028 
1029 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
1030 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1031 
1032 	if (ndopts.nd_opts_tgt_lladdr) {
1033 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1034 		if (!lladdr) {
1035 			net_dbg_ratelimited("NA: invalid link-layer address length\n");
1036 			return reason;
1037 		}
1038 	}
1039 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1040 	if (ifp) {
1041 		if (skb->pkt_type != PACKET_LOOPBACK
1042 		    && (ifp->flags & IFA_F_TENTATIVE)) {
1043 				addrconf_dad_failure(skb, ifp);
1044 				return reason;
1045 		}
1046 		/* What should we make now? The advertisement
1047 		   is invalid, but ndisc specs say nothing
1048 		   about it. It could be misconfiguration, or
1049 		   an smart proxy agent tries to help us :-)
1050 
1051 		   We should not print the error if NA has been
1052 		   received from loopback - it is just our own
1053 		   unsolicited advertisement.
1054 		 */
1055 		if (skb->pkt_type != PACKET_LOOPBACK)
1056 			net_warn_ratelimited("NA: %pM advertised our address %pI6c on %s!\n",
1057 					     eth_hdr(skb)->h_source, &ifp->addr,
1058 					     ifp->idev->dev->name);
1059 		in6_ifa_put(ifp);
1060 		return reason;
1061 	}
1062 
1063 	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1064 
1065 	/* RFC 9131 updates original Neighbour Discovery RFC 4861.
1066 	 * NAs with Target LL Address option without a corresponding
1067 	 * entry in the neighbour cache can now create a STALE neighbour
1068 	 * cache entry on routers.
1069 	 *
1070 	 *   entry accept  fwding  solicited        behaviour
1071 	 * ------- ------  ------  ---------    ----------------------
1072 	 * present      X       X         0     Set state to STALE
1073 	 * present      X       X         1     Set state to REACHABLE
1074 	 *  absent      0       X         X     Do nothing
1075 	 *  absent      1       0         X     Do nothing
1076 	 *  absent      1       1         X     Add a new STALE entry
1077 	 *
1078 	 * Note that we don't do a (daddr == all-routers-mcast) check.
1079 	 */
1080 	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1081 	if (!neigh && lladdr && idev && READ_ONCE(idev->cnf.forwarding)) {
1082 		if (accept_untracked_na(dev, saddr)) {
1083 			neigh = neigh_create(&nd_tbl, &msg->target, dev);
1084 			new_state = NUD_STALE;
1085 		}
1086 	}
1087 
1088 	if (neigh && !IS_ERR(neigh)) {
1089 		u8 old_flags = neigh->flags;
1090 		struct net *net = dev_net(dev);
1091 
1092 		if (READ_ONCE(neigh->nud_state) & NUD_FAILED)
1093 			goto out;
1094 
1095 		/*
1096 		 * Don't update the neighbor cache entry on a proxy NA from
1097 		 * ourselves because either the proxied node is off link or it
1098 		 * has already sent a NA to us.
1099 		 */
1100 		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1101 		    READ_ONCE(net->ipv6.devconf_all->forwarding) &&
1102 		    READ_ONCE(net->ipv6.devconf_all->proxy_ndp) &&
1103 		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1104 			/* XXX: idev->cnf.proxy_ndp */
1105 			goto out;
1106 		}
1107 
1108 		ndisc_update(dev, neigh, lladdr,
1109 			     new_state,
1110 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1111 			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1112 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1113 			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1114 			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1115 
1116 		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1117 			/*
1118 			 * Change: router to host
1119 			 */
1120 			rt6_clean_tohost(dev_net(dev),  saddr);
1121 		}
1122 		reason = SKB_CONSUMED;
1123 out:
1124 		neigh_release(neigh);
1125 	}
1126 	return reason;
1127 }
1128 
1129 static enum skb_drop_reason ndisc_recv_rs(struct sk_buff *skb)
1130 {
1131 	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1132 	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1133 	struct neighbour *neigh;
1134 	struct inet6_dev *idev;
1135 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1136 	struct ndisc_options ndopts;
1137 	u8 *lladdr = NULL;
1138 	SKB_DR(reason);
1139 
1140 	if (skb->len < sizeof(*rs_msg))
1141 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1142 
1143 	idev = __in6_dev_get(skb->dev);
1144 	if (!idev) {
1145 		net_err_ratelimited("RS: can't find in6 device\n");
1146 		return reason;
1147 	}
1148 
1149 	/* Don't accept RS if we're not in router mode */
1150 	if (!READ_ONCE(idev->cnf.forwarding))
1151 		goto out;
1152 
1153 	/*
1154 	 * Don't update NCE if src = ::;
1155 	 * this implies that the source node has no ip address assigned yet.
1156 	 */
1157 	if (ipv6_addr_any(saddr))
1158 		goto out;
1159 
1160 	/* Parse ND options */
1161 	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts))
1162 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1163 
1164 	if (ndopts.nd_opts_src_lladdr) {
1165 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1166 					     skb->dev);
1167 		if (!lladdr)
1168 			goto out;
1169 	}
1170 
1171 	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1172 	if (neigh) {
1173 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1174 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1175 			     NEIGH_UPDATE_F_OVERRIDE|
1176 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1177 			     NDISC_ROUTER_SOLICITATION, &ndopts);
1178 		neigh_release(neigh);
1179 		reason = SKB_CONSUMED;
1180 	}
1181 out:
1182 	return reason;
1183 }
1184 
1185 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1186 {
1187 	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1188 	struct sk_buff *skb;
1189 	struct nlmsghdr *nlh;
1190 	struct nduseroptmsg *ndmsg;
1191 	struct net *net = dev_net(ra->dev);
1192 	int err;
1193 	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1194 				    + (opt->nd_opt_len << 3));
1195 	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1196 
1197 	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1198 	if (!skb) {
1199 		err = -ENOBUFS;
1200 		goto errout;
1201 	}
1202 
1203 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1204 	if (!nlh) {
1205 		goto nla_put_failure;
1206 	}
1207 
1208 	ndmsg = nlmsg_data(nlh);
1209 	ndmsg->nduseropt_family = AF_INET6;
1210 	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1211 	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1212 	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1213 	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1214 
1215 	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1216 
1217 	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1218 		goto nla_put_failure;
1219 	nlmsg_end(skb, nlh);
1220 
1221 	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1222 	return;
1223 
1224 nla_put_failure:
1225 	nlmsg_free(skb);
1226 	err = -EMSGSIZE;
1227 errout:
1228 	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1229 }
1230 
1231 static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb)
1232 {
1233 	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1234 	bool send_ifinfo_notify = false;
1235 	struct neighbour *neigh = NULL;
1236 	struct ndisc_options ndopts;
1237 	struct fib6_info *rt = NULL;
1238 	struct inet6_dev *in6_dev;
1239 	struct fib6_table *table;
1240 	u32 defrtr_usr_metric;
1241 	unsigned int pref = 0;
1242 	__u32 old_if_flags;
1243 	struct net *net;
1244 	SKB_DR(reason);
1245 	int lifetime;
1246 	int optlen;
1247 
1248 	__u8 *opt = (__u8 *)(ra_msg + 1);
1249 
1250 	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1251 		sizeof(struct ra_msg);
1252 
1253 	net_dbg_ratelimited("RA: %s, dev: %s\n", __func__, skb->dev->name);
1254 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1255 		net_dbg_ratelimited("RA: source address is not link-local\n");
1256 		return reason;
1257 	}
1258 	if (optlen < 0)
1259 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1260 
1261 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1262 	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1263 		net_dbg_ratelimited("RA: from host or unauthorized router\n");
1264 		return reason;
1265 	}
1266 #endif
1267 
1268 	in6_dev = __in6_dev_get(skb->dev);
1269 	if (!in6_dev) {
1270 		net_err_ratelimited("RA: can't find inet6 device for %s\n", skb->dev->name);
1271 		return reason;
1272 	}
1273 
1274 	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts))
1275 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1276 
1277 	if (!ipv6_accept_ra(in6_dev)) {
1278 		net_dbg_ratelimited("RA: %s, did not accept ra for dev: %s\n", __func__,
1279 				    skb->dev->name);
1280 		goto skip_linkparms;
1281 	}
1282 
1283 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1284 	/* skip link-specific parameters from interior routers */
1285 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1286 		net_dbg_ratelimited("RA: %s, nodetype is NODEFAULT, dev: %s\n", __func__,
1287 				    skb->dev->name);
1288 		goto skip_linkparms;
1289 	}
1290 #endif
1291 
1292 	if (in6_dev->if_flags & IF_RS_SENT) {
1293 		/*
1294 		 *	flag that an RA was received after an RS was sent
1295 		 *	out on this interface.
1296 		 */
1297 		in6_dev->if_flags |= IF_RA_RCVD;
1298 	}
1299 
1300 	/*
1301 	 * Remember the managed/otherconf flags from most recently
1302 	 * received RA message (RFC 2462) -- yoshfuji
1303 	 */
1304 	old_if_flags = in6_dev->if_flags;
1305 	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1306 				IF_RA_OTHERCONF)) |
1307 				(ra_msg->icmph.icmp6_addrconf_managed ?
1308 					IF_RA_MANAGED : 0) |
1309 				(ra_msg->icmph.icmp6_addrconf_other ?
1310 					IF_RA_OTHERCONF : 0);
1311 
1312 	if (old_if_flags != in6_dev->if_flags)
1313 		send_ifinfo_notify = true;
1314 
1315 	if (!READ_ONCE(in6_dev->cnf.accept_ra_defrtr)) {
1316 		net_dbg_ratelimited("RA: %s, defrtr is false for dev: %s\n", __func__,
1317 				    skb->dev->name);
1318 		goto skip_defrtr;
1319 	}
1320 
1321 	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1322 	if (lifetime != 0 &&
1323 	    lifetime < READ_ONCE(in6_dev->cnf.accept_ra_min_lft)) {
1324 		net_dbg_ratelimited("RA: router lifetime (%ds) is too short: %s\n", lifetime,
1325 				    skb->dev->name);
1326 		goto skip_defrtr;
1327 	}
1328 
1329 	/* Do not accept RA with source-addr found on local machine unless
1330 	 * accept_ra_from_local is set to true.
1331 	 */
1332 	net = dev_net(in6_dev->dev);
1333 	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1334 	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1335 		net_dbg_ratelimited("RA from local address detected on dev: %s: default router ignored\n",
1336 				    skb->dev->name);
1337 		goto skip_defrtr;
1338 	}
1339 
1340 #ifdef CONFIG_IPV6_ROUTER_PREF
1341 	pref = ra_msg->icmph.icmp6_router_pref;
1342 	/* 10b is handled as if it were 00b (medium) */
1343 	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1344 	    !READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref))
1345 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1346 #endif
1347 	/* routes added from RAs do not use nexthop objects */
1348 	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1349 	if (rt) {
1350 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1351 					 rt->fib6_nh->fib_nh_dev, NULL,
1352 					  &ipv6_hdr(skb)->saddr);
1353 		if (!neigh) {
1354 			net_err_ratelimited("RA: %s got default router without neighbour\n",
1355 					    __func__);
1356 			fib6_info_release(rt);
1357 			return reason;
1358 		}
1359 	}
1360 	/* Set default route metric as specified by user */
1361 	defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1362 	/* delete the route if lifetime is 0 or if metric needs change */
1363 	if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1364 		ip6_del_rt(net, rt, false);
1365 		rt = NULL;
1366 	}
1367 
1368 	net_dbg_ratelimited("RA: rt: %p  lifetime: %d, metric: %d, for dev: %s\n", rt, lifetime,
1369 			    defrtr_usr_metric, skb->dev->name);
1370 	if (!rt && lifetime) {
1371 		net_dbg_ratelimited("RA: adding default router\n");
1372 
1373 		if (neigh)
1374 			neigh_release(neigh);
1375 
1376 		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1377 					 skb->dev, pref, defrtr_usr_metric,
1378 					 lifetime);
1379 		if (!rt) {
1380 			net_err_ratelimited("RA: %s failed to add default route\n", __func__);
1381 			return reason;
1382 		}
1383 
1384 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1385 					 rt->fib6_nh->fib_nh_dev, NULL,
1386 					  &ipv6_hdr(skb)->saddr);
1387 		if (!neigh) {
1388 			net_err_ratelimited("RA: %s got default router without neighbour\n",
1389 					    __func__);
1390 			fib6_info_release(rt);
1391 			return reason;
1392 		}
1393 		neigh->flags |= NTF_ROUTER;
1394 	} else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1395 		struct nl_info nlinfo = {
1396 			.nl_net = net,
1397 		};
1398 		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1399 		inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1400 	}
1401 
1402 	if (rt) {
1403 		table = rt->fib6_table;
1404 		spin_lock_bh(&table->tb6_lock);
1405 
1406 		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1407 		fib6_add_gc_list(rt);
1408 
1409 		spin_unlock_bh(&table->tb6_lock);
1410 	}
1411 	if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) < 256 &&
1412 	    ra_msg->icmph.icmp6_hop_limit) {
1413 		if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) <=
1414 		    ra_msg->icmph.icmp6_hop_limit) {
1415 			WRITE_ONCE(in6_dev->cnf.hop_limit,
1416 				   ra_msg->icmph.icmp6_hop_limit);
1417 			fib6_metric_set(rt, RTAX_HOPLIMIT,
1418 					ra_msg->icmph.icmp6_hop_limit);
1419 		} else {
1420 			net_dbg_ratelimited("RA: Got route advertisement with lower hop_limit than minimum\n");
1421 		}
1422 	}
1423 
1424 skip_defrtr:
1425 
1426 	/*
1427 	 *	Update Reachable Time and Retrans Timer
1428 	 */
1429 
1430 	if (in6_dev->nd_parms) {
1431 		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1432 
1433 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1434 			rtime = (rtime*HZ)/1000;
1435 			if (rtime < HZ/100)
1436 				rtime = HZ/100;
1437 			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1438 			in6_dev->tstamp = jiffies;
1439 			send_ifinfo_notify = true;
1440 		}
1441 
1442 		rtime = ntohl(ra_msg->reachable_time);
1443 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1444 			rtime = (rtime*HZ)/1000;
1445 
1446 			if (rtime < HZ/10)
1447 				rtime = HZ/10;
1448 
1449 			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1450 				NEIGH_VAR_SET(in6_dev->nd_parms,
1451 					      BASE_REACHABLE_TIME, rtime);
1452 				NEIGH_VAR_SET(in6_dev->nd_parms,
1453 					      GC_STALETIME, 3 * rtime);
1454 				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1455 				in6_dev->tstamp = jiffies;
1456 				send_ifinfo_notify = true;
1457 			}
1458 		}
1459 	}
1460 
1461 skip_linkparms:
1462 
1463 	/*
1464 	 *	Process options.
1465 	 */
1466 
1467 	if (!neigh)
1468 		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1469 				       skb->dev, 1);
1470 	if (neigh) {
1471 		u8 *lladdr = NULL;
1472 		if (ndopts.nd_opts_src_lladdr) {
1473 			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1474 						     skb->dev);
1475 			if (!lladdr) {
1476 				net_dbg_ratelimited("RA: invalid link-layer address length\n");
1477 				goto out;
1478 			}
1479 		}
1480 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1481 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1482 			     NEIGH_UPDATE_F_OVERRIDE|
1483 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1484 			     NEIGH_UPDATE_F_ISROUTER,
1485 			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1486 		reason = SKB_CONSUMED;
1487 	}
1488 
1489 	if (!ipv6_accept_ra(in6_dev)) {
1490 		net_dbg_ratelimited("RA: %s, accept_ra is false for dev: %s\n", __func__,
1491 				    skb->dev->name);
1492 		goto out;
1493 	}
1494 
1495 #ifdef CONFIG_IPV6_ROUTE_INFO
1496 	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1497 	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1498 			  in6_dev->dev, 0)) {
1499 		net_dbg_ratelimited("RA from local address detected on dev: %s: router info ignored.\n",
1500 				    skb->dev->name);
1501 		goto skip_routeinfo;
1502 	}
1503 
1504 	if (READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref) && ndopts.nd_opts_ri) {
1505 		struct nd_opt_hdr *p;
1506 		for (p = ndopts.nd_opts_ri;
1507 		     p;
1508 		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1509 			struct route_info *ri = (struct route_info *)p;
1510 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1511 			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1512 			    ri->prefix_len == 0)
1513 				continue;
1514 #endif
1515 			if (ri->prefix_len == 0 &&
1516 			    !READ_ONCE(in6_dev->cnf.accept_ra_defrtr))
1517 				continue;
1518 			if (ri->lifetime != 0 &&
1519 			    ntohl(ri->lifetime) < READ_ONCE(in6_dev->cnf.accept_ra_min_lft))
1520 				continue;
1521 			if (ri->prefix_len < READ_ONCE(in6_dev->cnf.accept_ra_rt_info_min_plen))
1522 				continue;
1523 			if (ri->prefix_len > READ_ONCE(in6_dev->cnf.accept_ra_rt_info_max_plen))
1524 				continue;
1525 			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1526 				      &ipv6_hdr(skb)->saddr);
1527 		}
1528 	}
1529 
1530 skip_routeinfo:
1531 #endif
1532 
1533 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1534 	/* skip link-specific ndopts from interior routers */
1535 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1536 		net_dbg_ratelimited("RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1537 				    __func__, skb->dev->name);
1538 		goto out;
1539 	}
1540 #endif
1541 
1542 	if (READ_ONCE(in6_dev->cnf.accept_ra_pinfo) && ndopts.nd_opts_pi) {
1543 		struct nd_opt_hdr *p;
1544 		for (p = ndopts.nd_opts_pi;
1545 		     p;
1546 		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1547 			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1548 					    (p->nd_opt_len) << 3,
1549 					    ndopts.nd_opts_src_lladdr != NULL);
1550 		}
1551 	}
1552 
1553 	if (ndopts.nd_opts_mtu && READ_ONCE(in6_dev->cnf.accept_ra_mtu)) {
1554 		__be32 n;
1555 		u32 mtu;
1556 
1557 		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1558 		mtu = ntohl(n);
1559 
1560 		if (in6_dev->ra_mtu != mtu) {
1561 			in6_dev->ra_mtu = mtu;
1562 			send_ifinfo_notify = true;
1563 		}
1564 
1565 		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1566 			net_dbg_ratelimited("RA: invalid mtu: %d\n", mtu);
1567 		} else if (READ_ONCE(in6_dev->cnf.mtu6) != mtu) {
1568 			WRITE_ONCE(in6_dev->cnf.mtu6, mtu);
1569 			fib6_metric_set(rt, RTAX_MTU, mtu);
1570 			rt6_mtu_change(skb->dev, mtu);
1571 		}
1572 	}
1573 
1574 	if (ndopts.nd_useropts) {
1575 		struct nd_opt_hdr *p;
1576 		for (p = ndopts.nd_useropts;
1577 		     p;
1578 		     p = ndisc_next_useropt(skb->dev, p,
1579 					    ndopts.nd_useropts_end)) {
1580 			ndisc_ra_useropt(skb, p);
1581 		}
1582 	}
1583 
1584 	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1585 		net_dbg_ratelimited("RA: invalid RA options\n");
1586 	}
1587 out:
1588 	/* Send a notify if RA changed managed/otherconf flags or
1589 	 * timer settings or ra_mtu value
1590 	 */
1591 	if (send_ifinfo_notify)
1592 		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1593 
1594 	fib6_info_release(rt);
1595 	if (neigh)
1596 		neigh_release(neigh);
1597 	return reason;
1598 }
1599 
1600 static enum skb_drop_reason ndisc_redirect_rcv(struct sk_buff *skb)
1601 {
1602 	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1603 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1604 				    offsetof(struct rd_msg, opt));
1605 	struct ndisc_options ndopts;
1606 	SKB_DR(reason);
1607 	u8 *hdr;
1608 
1609 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1610 	switch (skb->ndisc_nodetype) {
1611 	case NDISC_NODETYPE_HOST:
1612 	case NDISC_NODETYPE_NODEFAULT:
1613 		net_dbg_ratelimited("Redirect: from host or unauthorized router\n");
1614 		return reason;
1615 	}
1616 #endif
1617 
1618 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1619 		net_dbg_ratelimited("Redirect: source address is not link-local\n");
1620 		return reason;
1621 	}
1622 
1623 	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1624 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1625 
1626 	if (!ndopts.nd_opts_rh) {
1627 		ip6_redirect_no_header(skb, dev_net(skb->dev),
1628 					skb->dev->ifindex);
1629 		return reason;
1630 	}
1631 
1632 	hdr = (u8 *)ndopts.nd_opts_rh;
1633 	hdr += 8;
1634 	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1635 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1636 
1637 	return icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1638 }
1639 
1640 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1641 					   struct sk_buff *orig_skb,
1642 					   int rd_len)
1643 {
1644 	u8 *opt = skb_put(skb, rd_len);
1645 
1646 	memset(opt, 0, 8);
1647 	*(opt++) = ND_OPT_REDIRECT_HDR;
1648 	*(opt++) = (rd_len >> 3);
1649 	opt += 6;
1650 
1651 	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1652 		      rd_len - 8);
1653 }
1654 
1655 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1656 {
1657 	struct net_device *dev = skb->dev;
1658 	struct net *net = dev_net_rcu(dev);
1659 	struct sock *sk = net->ipv6.ndisc_sk;
1660 	int optlen = 0;
1661 	struct inet_peer *peer;
1662 	struct sk_buff *buff;
1663 	struct rd_msg *msg;
1664 	struct in6_addr saddr_buf;
1665 	struct rt6_info *rt;
1666 	struct dst_entry *dst;
1667 	struct flowi6 fl6;
1668 	int rd_len;
1669 	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1670 	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1671 	bool ret;
1672 
1673 	if (netif_is_l3_master(dev)) {
1674 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
1675 		if (!dev)
1676 			return;
1677 	}
1678 
1679 	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1680 		net_dbg_ratelimited("Redirect: no link-local address on %s\n", dev->name);
1681 		return;
1682 	}
1683 
1684 	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1685 	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1686 		net_dbg_ratelimited("Redirect: target address is not link-local unicast\n");
1687 		return;
1688 	}
1689 
1690 	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1691 			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1692 
1693 	dst = ip6_route_output(net, NULL, &fl6);
1694 	if (dst->error) {
1695 		dst_release(dst);
1696 		return;
1697 	}
1698 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1699 	if (IS_ERR(dst))
1700 		return;
1701 
1702 	rt = dst_rt6_info(dst);
1703 
1704 	if (rt->rt6i_flags & RTF_GATEWAY) {
1705 		net_dbg_ratelimited("Redirect: destination is not a neighbour\n");
1706 		goto release;
1707 	}
1708 
1709 	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr);
1710 	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1711 
1712 	if (!ret)
1713 		goto release;
1714 
1715 	if (dev->addr_len) {
1716 		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1717 		if (!neigh) {
1718 			net_dbg_ratelimited("Redirect: no neigh for target address\n");
1719 			goto release;
1720 		}
1721 
1722 		read_lock_bh(&neigh->lock);
1723 		if (neigh->nud_state & NUD_VALID) {
1724 			memcpy(ha_buf, neigh->ha, dev->addr_len);
1725 			read_unlock_bh(&neigh->lock);
1726 			ha = ha_buf;
1727 			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1728 								ops_data_buf,
1729 								&ops_data);
1730 		} else
1731 			read_unlock_bh(&neigh->lock);
1732 
1733 		neigh_release(neigh);
1734 	}
1735 
1736 	rd_len = min_t(unsigned int,
1737 		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1738 		       skb->len + 8);
1739 	rd_len &= ~0x7;
1740 	optlen += rd_len;
1741 
1742 	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1743 	if (!buff)
1744 		goto release;
1745 
1746 	msg = skb_put(buff, sizeof(*msg));
1747 	*msg = (struct rd_msg) {
1748 		.icmph = {
1749 			.icmp6_type = NDISC_REDIRECT,
1750 		},
1751 		.target = *target,
1752 		.dest = ipv6_hdr(skb)->daddr,
1753 	};
1754 
1755 	/*
1756 	 *	include target_address option
1757 	 */
1758 
1759 	if (ha)
1760 		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1761 
1762 	/*
1763 	 *	build redirect option and copy skb over to the new packet.
1764 	 */
1765 
1766 	if (rd_len)
1767 		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1768 
1769 	skb_dst_set(buff, dst);
1770 	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1771 	return;
1772 
1773 release:
1774 	dst_release(dst);
1775 }
1776 
1777 static void pndisc_redo(struct sk_buff *skb)
1778 {
1779 	enum skb_drop_reason reason = ndisc_recv_ns(skb);
1780 
1781 	kfree_skb_reason(skb, reason);
1782 }
1783 
1784 static int ndisc_is_multicast(const void *pkey)
1785 {
1786 	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1787 }
1788 
1789 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1790 {
1791 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1792 
1793 	if (!idev)
1794 		return true;
1795 	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1796 	    READ_ONCE(idev->cnf.suppress_frag_ndisc)) {
1797 		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1798 		return true;
1799 	}
1800 	return false;
1801 }
1802 
1803 enum skb_drop_reason ndisc_rcv(struct sk_buff *skb)
1804 {
1805 	struct nd_msg *msg;
1806 	SKB_DR(reason);
1807 
1808 	if (ndisc_suppress_frag_ndisc(skb))
1809 		return SKB_DROP_REASON_IPV6_NDISC_FRAG;
1810 
1811 	if (skb_linearize(skb))
1812 		return SKB_DROP_REASON_NOMEM;
1813 
1814 	msg = (struct nd_msg *)skb_transport_header(skb);
1815 
1816 	__skb_push(skb, skb->data - skb_transport_header(skb));
1817 
1818 	if (ipv6_hdr(skb)->hop_limit != 255) {
1819 		net_dbg_ratelimited("NDISC: invalid hop-limit: %d\n", ipv6_hdr(skb)->hop_limit);
1820 		return SKB_DROP_REASON_IPV6_NDISC_HOP_LIMIT;
1821 	}
1822 
1823 	if (msg->icmph.icmp6_code != 0) {
1824 		net_dbg_ratelimited("NDISC: invalid ICMPv6 code: %d\n", msg->icmph.icmp6_code);
1825 		return SKB_DROP_REASON_IPV6_NDISC_BAD_CODE;
1826 	}
1827 
1828 	switch (msg->icmph.icmp6_type) {
1829 	case NDISC_NEIGHBOUR_SOLICITATION:
1830 		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1831 		reason = ndisc_recv_ns(skb);
1832 		break;
1833 
1834 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1835 		reason = ndisc_recv_na(skb);
1836 		break;
1837 
1838 	case NDISC_ROUTER_SOLICITATION:
1839 		reason = ndisc_recv_rs(skb);
1840 		break;
1841 
1842 	case NDISC_ROUTER_ADVERTISEMENT:
1843 		reason = ndisc_router_discovery(skb);
1844 		break;
1845 
1846 	case NDISC_REDIRECT:
1847 		reason = ndisc_redirect_rcv(skb);
1848 		break;
1849 	}
1850 
1851 	return reason;
1852 }
1853 
1854 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1855 {
1856 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1857 	struct netdev_notifier_change_info *change_info;
1858 	struct net *net = dev_net(dev);
1859 	struct inet6_dev *idev;
1860 	bool evict_nocarrier;
1861 
1862 	switch (event) {
1863 	case NETDEV_CHANGEADDR:
1864 		neigh_changeaddr(&nd_tbl, dev);
1865 		fib6_run_gc(0, net, false);
1866 		fallthrough;
1867 	case NETDEV_UP:
1868 		idev = in6_dev_get(dev);
1869 		if (!idev)
1870 			break;
1871 		if (READ_ONCE(idev->cnf.ndisc_notify) ||
1872 		    READ_ONCE(net->ipv6.devconf_all->ndisc_notify))
1873 			ndisc_send_unsol_na(dev);
1874 		in6_dev_put(idev);
1875 		break;
1876 	case NETDEV_CHANGE:
1877 		idev = in6_dev_get(dev);
1878 		if (!idev)
1879 			evict_nocarrier = true;
1880 		else {
1881 			evict_nocarrier = READ_ONCE(idev->cnf.ndisc_evict_nocarrier) &&
1882 					  READ_ONCE(net->ipv6.devconf_all->ndisc_evict_nocarrier);
1883 			in6_dev_put(idev);
1884 		}
1885 
1886 		change_info = ptr;
1887 		if (change_info->flags_changed & IFF_NOARP)
1888 			neigh_changeaddr(&nd_tbl, dev);
1889 		if (evict_nocarrier && !netif_carrier_ok(dev))
1890 			neigh_carrier_down(&nd_tbl, dev);
1891 		break;
1892 	case NETDEV_DOWN:
1893 		neigh_ifdown(&nd_tbl, dev);
1894 		fib6_run_gc(0, net, false);
1895 		break;
1896 	case NETDEV_NOTIFY_PEERS:
1897 		ndisc_send_unsol_na(dev);
1898 		break;
1899 	default:
1900 		break;
1901 	}
1902 
1903 	return NOTIFY_DONE;
1904 }
1905 
1906 static struct notifier_block ndisc_netdev_notifier = {
1907 	.notifier_call = ndisc_netdev_event,
1908 	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1909 };
1910 
1911 #ifdef CONFIG_SYSCTL
1912 static void ndisc_warn_deprecated_sysctl(const struct ctl_table *ctl,
1913 					 const char *func, const char *dev_name)
1914 {
1915 	static char warncomm[TASK_COMM_LEN];
1916 	static int warned;
1917 	if (strcmp(warncomm, current->comm) && warned < 5) {
1918 		strscpy(warncomm, current->comm);
1919 		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1920 			warncomm, func,
1921 			dev_name, ctl->procname,
1922 			dev_name, ctl->procname);
1923 		warned++;
1924 	}
1925 }
1926 
1927 int ndisc_ifinfo_sysctl_change(const struct ctl_table *ctl, int write, void *buffer,
1928 		size_t *lenp, loff_t *ppos)
1929 {
1930 	struct net_device *dev = ctl->extra1;
1931 	struct inet6_dev *idev;
1932 	int ret;
1933 
1934 	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1935 	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1936 		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1937 
1938 	if (strcmp(ctl->procname, "retrans_time") == 0)
1939 		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1940 
1941 	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1942 		ret = neigh_proc_dointvec_jiffies(ctl, write,
1943 						  buffer, lenp, ppos);
1944 
1945 	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1946 		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1947 		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1948 						     buffer, lenp, ppos);
1949 	else
1950 		ret = -1;
1951 
1952 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1953 		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1954 			idev->nd_parms->reachable_time =
1955 					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1956 		WRITE_ONCE(idev->tstamp, jiffies);
1957 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1958 		in6_dev_put(idev);
1959 	}
1960 	return ret;
1961 }
1962 
1963 
1964 #endif
1965 
1966 static int __net_init ndisc_net_init(struct net *net)
1967 {
1968 	struct ipv6_pinfo *np;
1969 	struct sock *sk;
1970 	int err;
1971 
1972 	err = inet_ctl_sock_create(&sk, PF_INET6,
1973 				   SOCK_RAW, IPPROTO_ICMPV6, net);
1974 	if (err < 0) {
1975 		net_err_ratelimited("NDISC: Failed to initialize the control socket (err %d)\n",
1976 				    err);
1977 		return err;
1978 	}
1979 
1980 	net->ipv6.ndisc_sk = sk;
1981 
1982 	np = inet6_sk(sk);
1983 	np->hop_limit = 255;
1984 	/* Do not loopback ndisc messages */
1985 	inet6_clear_bit(MC6_LOOP, sk);
1986 
1987 	return 0;
1988 }
1989 
1990 static void __net_exit ndisc_net_exit(struct net *net)
1991 {
1992 	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1993 }
1994 
1995 static struct pernet_operations ndisc_net_ops = {
1996 	.init = ndisc_net_init,
1997 	.exit = ndisc_net_exit,
1998 };
1999 
2000 int __init ndisc_init(void)
2001 {
2002 	int err;
2003 
2004 	err = register_pernet_subsys(&ndisc_net_ops);
2005 	if (err)
2006 		return err;
2007 	/*
2008 	 * Initialize the neighbour table
2009 	 */
2010 	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2011 
2012 #ifdef CONFIG_SYSCTL
2013 	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2014 				    ndisc_ifinfo_sysctl_change);
2015 	if (err)
2016 		goto out_unregister_pernet;
2017 out:
2018 #endif
2019 	return err;
2020 
2021 #ifdef CONFIG_SYSCTL
2022 out_unregister_pernet:
2023 	unregister_pernet_subsys(&ndisc_net_ops);
2024 	goto out;
2025 #endif
2026 }
2027 
2028 int __init ndisc_late_init(void)
2029 {
2030 	return register_netdevice_notifier(&ndisc_netdev_notifier);
2031 }
2032 
2033 void ndisc_late_cleanup(void)
2034 {
2035 	unregister_netdevice_notifier(&ndisc_netdev_notifier);
2036 }
2037 
2038 void ndisc_cleanup(void)
2039 {
2040 #ifdef CONFIG_SYSCTL
2041 	neigh_sysctl_unregister(&nd_tbl.parms);
2042 #endif
2043 	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2044 	unregister_pernet_subsys(&ndisc_net_ops);
2045 }
2046