xref: /linux/net/ipv6/ndisc.c (revision 527a0f2bdcfe77fce22f006b97e42e4da3137c86)
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_CAPTIVE_PORTAL ||
204 		opt->nd_opt_type == ND_OPT_PREF64 ||
205 		ndisc_ops_is_useropt(dev, opt->nd_opt_type);
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 				ND_PRINTK(2, warn,
247 					  "%s: duplicated ND6 option found: type=%d\n",
248 					  __func__, nd_opt->nd_opt_type);
249 			} else {
250 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
251 			}
252 			break;
253 		case ND_OPT_PREFIX_INFO:
254 			ndopts->nd_opts_pi_end = nd_opt;
255 			if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
256 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
257 			break;
258 #ifdef CONFIG_IPV6_ROUTE_INFO
259 		case ND_OPT_ROUTE_INFO:
260 			ndopts->nd_opts_ri_end = nd_opt;
261 			if (!ndopts->nd_opts_ri)
262 				ndopts->nd_opts_ri = nd_opt;
263 			break;
264 #endif
265 		default:
266 			unknown = true;
267 		}
268 		if (ndisc_is_useropt(dev, nd_opt)) {
269 			ndopts->nd_useropts_end = nd_opt;
270 			if (!ndopts->nd_useropts)
271 				ndopts->nd_useropts = nd_opt;
272 		} else if (unknown) {
273 			/*
274 			 * Unknown options must be silently ignored,
275 			 * to accommodate future extension to the
276 			 * protocol.
277 			 */
278 			ND_PRINTK(2, notice,
279 				  "%s: ignored unsupported option; type=%d, len=%d\n",
280 				  __func__,
281 				  nd_opt->nd_opt_type,
282 				  nd_opt->nd_opt_len);
283 		}
284 next_opt:
285 		opt_len -= l;
286 		nd_opt = ((void *)nd_opt) + l;
287 	}
288 	return ndopts;
289 }
290 
291 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
292 {
293 	switch (dev->type) {
294 	case ARPHRD_ETHER:
295 	case ARPHRD_IEEE802:	/* Not sure. Check it later. --ANK */
296 	case ARPHRD_FDDI:
297 		ipv6_eth_mc_map(addr, buf);
298 		return 0;
299 	case ARPHRD_ARCNET:
300 		ipv6_arcnet_mc_map(addr, buf);
301 		return 0;
302 	case ARPHRD_INFINIBAND:
303 		ipv6_ib_mc_map(addr, dev->broadcast, buf);
304 		return 0;
305 	case ARPHRD_IPGRE:
306 		return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
307 	default:
308 		if (dir) {
309 			memcpy(buf, dev->broadcast, dev->addr_len);
310 			return 0;
311 		}
312 	}
313 	return -EINVAL;
314 }
315 EXPORT_SYMBOL(ndisc_mc_map);
316 
317 static u32 ndisc_hash(const void *pkey,
318 		      const struct net_device *dev,
319 		      __u32 *hash_rnd)
320 {
321 	return ndisc_hashfn(pkey, dev, hash_rnd);
322 }
323 
324 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
325 {
326 	return neigh_key_eq128(n, pkey);
327 }
328 
329 static int ndisc_constructor(struct neighbour *neigh)
330 {
331 	struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
332 	struct net_device *dev = neigh->dev;
333 	struct inet6_dev *in6_dev;
334 	struct neigh_parms *parms;
335 	bool is_multicast = ipv6_addr_is_multicast(addr);
336 
337 	in6_dev = in6_dev_get(dev);
338 	if (!in6_dev) {
339 		return -EINVAL;
340 	}
341 
342 	parms = in6_dev->nd_parms;
343 	__neigh_parms_put(neigh->parms);
344 	neigh->parms = neigh_parms_clone(parms);
345 
346 	neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
347 	if (!dev->header_ops) {
348 		neigh->nud_state = NUD_NOARP;
349 		neigh->ops = &ndisc_direct_ops;
350 		neigh->output = neigh_direct_output;
351 	} else {
352 		if (is_multicast) {
353 			neigh->nud_state = NUD_NOARP;
354 			ndisc_mc_map(addr, neigh->ha, dev, 1);
355 		} else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
356 			neigh->nud_state = NUD_NOARP;
357 			memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
358 			if (dev->flags&IFF_LOOPBACK)
359 				neigh->type = RTN_LOCAL;
360 		} else if (dev->flags&IFF_POINTOPOINT) {
361 			neigh->nud_state = NUD_NOARP;
362 			memcpy(neigh->ha, dev->broadcast, dev->addr_len);
363 		}
364 		if (dev->header_ops->cache)
365 			neigh->ops = &ndisc_hh_ops;
366 		else
367 			neigh->ops = &ndisc_generic_ops;
368 		if (neigh->nud_state&NUD_VALID)
369 			neigh->output = neigh->ops->connected_output;
370 		else
371 			neigh->output = neigh->ops->output;
372 	}
373 	in6_dev_put(in6_dev);
374 	return 0;
375 }
376 
377 static int pndisc_constructor(struct pneigh_entry *n)
378 {
379 	struct in6_addr *addr = (struct in6_addr *)&n->key;
380 	struct in6_addr maddr;
381 	struct net_device *dev = n->dev;
382 
383 	if (!dev || !__in6_dev_get(dev))
384 		return -EINVAL;
385 	addrconf_addr_solict_mult(addr, &maddr);
386 	ipv6_dev_mc_inc(dev, &maddr);
387 	return 0;
388 }
389 
390 static void pndisc_destructor(struct pneigh_entry *n)
391 {
392 	struct in6_addr *addr = (struct in6_addr *)&n->key;
393 	struct in6_addr maddr;
394 	struct net_device *dev = n->dev;
395 
396 	if (!dev || !__in6_dev_get(dev))
397 		return;
398 	addrconf_addr_solict_mult(addr, &maddr);
399 	ipv6_dev_mc_dec(dev, &maddr);
400 }
401 
402 /* called with rtnl held */
403 static bool ndisc_allow_add(const struct net_device *dev,
404 			    struct netlink_ext_ack *extack)
405 {
406 	struct inet6_dev *idev = __in6_dev_get(dev);
407 
408 	if (!idev || idev->cnf.disable_ipv6) {
409 		NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
410 		return false;
411 	}
412 
413 	return true;
414 }
415 
416 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
417 				       int len)
418 {
419 	int hlen = LL_RESERVED_SPACE(dev);
420 	int tlen = dev->needed_tailroom;
421 	struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
422 	struct sk_buff *skb;
423 
424 	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
425 	if (!skb) {
426 		ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
427 			  __func__);
428 		return NULL;
429 	}
430 
431 	skb->protocol = htons(ETH_P_IPV6);
432 	skb->dev = dev;
433 
434 	skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
435 	skb_reset_transport_header(skb);
436 
437 	/* Manually assign socket ownership as we avoid calling
438 	 * sock_alloc_send_pskb() to bypass wmem buffer limits
439 	 */
440 	skb_set_owner_w(skb, sk);
441 
442 	return skb;
443 }
444 
445 static void ip6_nd_hdr(struct sk_buff *skb,
446 		       const struct in6_addr *saddr,
447 		       const struct in6_addr *daddr,
448 		       int hop_limit, int len)
449 {
450 	struct ipv6hdr *hdr;
451 	struct inet6_dev *idev;
452 	unsigned tclass;
453 
454 	rcu_read_lock();
455 	idev = __in6_dev_get(skb->dev);
456 	tclass = idev ? READ_ONCE(idev->cnf.ndisc_tclass) : 0;
457 	rcu_read_unlock();
458 
459 	skb_push(skb, sizeof(*hdr));
460 	skb_reset_network_header(skb);
461 	hdr = ipv6_hdr(skb);
462 
463 	ip6_flow_hdr(hdr, tclass, 0);
464 
465 	hdr->payload_len = htons(len);
466 	hdr->nexthdr = IPPROTO_ICMPV6;
467 	hdr->hop_limit = hop_limit;
468 
469 	hdr->saddr = *saddr;
470 	hdr->daddr = *daddr;
471 }
472 
473 void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
474 		    const struct in6_addr *saddr)
475 {
476 	struct dst_entry *dst = skb_dst(skb);
477 	struct net *net = dev_net(skb->dev);
478 	struct sock *sk = net->ipv6.ndisc_sk;
479 	struct inet6_dev *idev;
480 	int err;
481 	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
482 	u8 type;
483 
484 	type = icmp6h->icmp6_type;
485 
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 			kfree_skb(skb);
494 			return;
495 		}
496 
497 		skb_dst_set(skb, dst);
498 	}
499 
500 	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
501 					      IPPROTO_ICMPV6,
502 					      csum_partial(icmp6h,
503 							   skb->len, 0));
504 
505 	ip6_nd_hdr(skb, saddr, daddr, READ_ONCE(inet6_sk(sk)->hop_limit), skb->len);
506 
507 	rcu_read_lock();
508 	idev = __in6_dev_get(dst->dev);
509 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
510 
511 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
512 		      net, sk, skb, NULL, dst->dev,
513 		      dst_output);
514 	if (!err) {
515 		ICMP6MSGOUT_INC_STATS(net, idev, type);
516 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
517 	}
518 
519 	rcu_read_unlock();
520 }
521 EXPORT_SYMBOL(ndisc_send_skb);
522 
523 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
524 		   const struct in6_addr *solicited_addr,
525 		   bool router, bool solicited, bool override, bool inc_opt)
526 {
527 	struct sk_buff *skb;
528 	struct in6_addr tmpaddr;
529 	struct inet6_ifaddr *ifp;
530 	const struct in6_addr *src_addr;
531 	struct nd_msg *msg;
532 	int optlen = 0;
533 
534 	/* for anycast or proxy, solicited_addr != src_addr */
535 	ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
536 	if (ifp) {
537 		src_addr = solicited_addr;
538 		if (ifp->flags & IFA_F_OPTIMISTIC)
539 			override = false;
540 		inc_opt |= READ_ONCE(ifp->idev->cnf.force_tllao);
541 		in6_ifa_put(ifp);
542 	} else {
543 		if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
544 				       inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
545 				       &tmpaddr))
546 			return;
547 		src_addr = &tmpaddr;
548 	}
549 
550 	if (!dev->addr_len)
551 		inc_opt = false;
552 	if (inc_opt)
553 		optlen += ndisc_opt_addr_space(dev,
554 					       NDISC_NEIGHBOUR_ADVERTISEMENT);
555 
556 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
557 	if (!skb)
558 		return;
559 
560 	msg = skb_put(skb, sizeof(*msg));
561 	*msg = (struct nd_msg) {
562 		.icmph = {
563 			.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
564 			.icmp6_router = router,
565 			.icmp6_solicited = solicited,
566 			.icmp6_override = override,
567 		},
568 		.target = *solicited_addr,
569 	};
570 
571 	if (inc_opt)
572 		ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
573 				       dev->dev_addr,
574 				       NDISC_NEIGHBOUR_ADVERTISEMENT);
575 
576 	ndisc_send_skb(skb, daddr, src_addr);
577 }
578 
579 static void ndisc_send_unsol_na(struct net_device *dev)
580 {
581 	struct inet6_dev *idev;
582 	struct inet6_ifaddr *ifa;
583 
584 	idev = in6_dev_get(dev);
585 	if (!idev)
586 		return;
587 
588 	read_lock_bh(&idev->lock);
589 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
590 		/* skip tentative addresses until dad completes */
591 		if (ifa->flags & IFA_F_TENTATIVE &&
592 		    !(ifa->flags & IFA_F_OPTIMISTIC))
593 			continue;
594 
595 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
596 			      /*router=*/ !!idev->cnf.forwarding,
597 			      /*solicited=*/ false, /*override=*/ true,
598 			      /*inc_opt=*/ true);
599 	}
600 	read_unlock_bh(&idev->lock);
601 
602 	in6_dev_put(idev);
603 }
604 
605 struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
606 				const struct in6_addr *saddr, u64 nonce)
607 {
608 	int inc_opt = dev->addr_len;
609 	struct sk_buff *skb;
610 	struct nd_msg *msg;
611 	int optlen = 0;
612 
613 	if (!saddr)
614 		return NULL;
615 
616 	if (ipv6_addr_any(saddr))
617 		inc_opt = false;
618 	if (inc_opt)
619 		optlen += ndisc_opt_addr_space(dev,
620 					       NDISC_NEIGHBOUR_SOLICITATION);
621 	if (nonce != 0)
622 		optlen += 8;
623 
624 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
625 	if (!skb)
626 		return NULL;
627 
628 	msg = skb_put(skb, sizeof(*msg));
629 	*msg = (struct nd_msg) {
630 		.icmph = {
631 			.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
632 		},
633 		.target = *solicit,
634 	};
635 
636 	if (inc_opt)
637 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
638 				       dev->dev_addr,
639 				       NDISC_NEIGHBOUR_SOLICITATION);
640 	if (nonce != 0) {
641 		u8 *opt = skb_put(skb, 8);
642 
643 		opt[0] = ND_OPT_NONCE;
644 		opt[1] = 8 >> 3;
645 		memcpy(opt + 2, &nonce, 6);
646 	}
647 
648 	return skb;
649 }
650 EXPORT_SYMBOL(ndisc_ns_create);
651 
652 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
653 		   const struct in6_addr *daddr, const struct in6_addr *saddr,
654 		   u64 nonce)
655 {
656 	struct in6_addr addr_buf;
657 	struct sk_buff *skb;
658 
659 	if (!saddr) {
660 		if (ipv6_get_lladdr(dev, &addr_buf,
661 				    (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
662 			return;
663 		saddr = &addr_buf;
664 	}
665 
666 	skb = ndisc_ns_create(dev, solicit, saddr, nonce);
667 
668 	if (skb)
669 		ndisc_send_skb(skb, daddr, saddr);
670 }
671 
672 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
673 		   const struct in6_addr *daddr)
674 {
675 	struct sk_buff *skb;
676 	struct rs_msg *msg;
677 	int send_sllao = dev->addr_len;
678 	int optlen = 0;
679 
680 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
681 	/*
682 	 * According to section 2.2 of RFC 4429, we must not
683 	 * send router solicitations with a sllao from
684 	 * optimistic addresses, but we may send the solicitation
685 	 * if we don't include the sllao.  So here we check
686 	 * if our address is optimistic, and if so, we
687 	 * suppress the inclusion of the sllao.
688 	 */
689 	if (send_sllao) {
690 		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
691 							   dev, 1);
692 		if (ifp) {
693 			if (ifp->flags & IFA_F_OPTIMISTIC)  {
694 				send_sllao = 0;
695 			}
696 			in6_ifa_put(ifp);
697 		} else {
698 			send_sllao = 0;
699 		}
700 	}
701 #endif
702 	if (send_sllao)
703 		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
704 
705 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
706 	if (!skb)
707 		return;
708 
709 	msg = skb_put(skb, sizeof(*msg));
710 	*msg = (struct rs_msg) {
711 		.icmph = {
712 			.icmp6_type = NDISC_ROUTER_SOLICITATION,
713 		},
714 	};
715 
716 	if (send_sllao)
717 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
718 				       dev->dev_addr,
719 				       NDISC_ROUTER_SOLICITATION);
720 
721 	ndisc_send_skb(skb, daddr, saddr);
722 }
723 
724 
725 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
726 {
727 	/*
728 	 *	"The sender MUST return an ICMP
729 	 *	 destination unreachable"
730 	 */
731 	dst_link_failure(skb);
732 	kfree_skb(skb);
733 }
734 
735 /* Called with locked neigh: either read or both */
736 
737 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
738 {
739 	struct in6_addr *saddr = NULL;
740 	struct in6_addr mcaddr;
741 	struct net_device *dev = neigh->dev;
742 	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
743 	int probes = atomic_read(&neigh->probes);
744 
745 	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
746 					   dev, false, 1,
747 					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
748 		saddr = &ipv6_hdr(skb)->saddr;
749 	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
750 	if (probes < 0) {
751 		if (!(READ_ONCE(neigh->nud_state) & NUD_VALID)) {
752 			ND_PRINTK(1, dbg,
753 				  "%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 		ND_PRINTK(2, warn, "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 		ND_PRINTK(2, warn, "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 			ND_PRINTK(2, warn,
832 				  "NS: invalid link-layer address length\n");
833 			return reason;
834 		}
835 
836 		/* RFC2461 7.1.1:
837 		 *	If the IP source address is the unspecified address,
838 		 *	there MUST NOT be source link-layer address option
839 		 *	in the message.
840 		 */
841 		if (dad) {
842 			ND_PRINTK(2, warn,
843 				  "NS: bad DAD packet (link-layer address option)\n");
844 			return reason;
845 		}
846 	}
847 	if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
848 		memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
849 
850 	inc = ipv6_addr_is_multicast(daddr);
851 
852 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
853 	if (ifp) {
854 have_ifp:
855 		if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
856 			if (dad) {
857 				if (nonce != 0 && ifp->dad_nonce == nonce) {
858 					u8 *np = (u8 *)&nonce;
859 					/* Matching nonce if looped back */
860 					ND_PRINTK(2, notice,
861 						  "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
862 						  ifp->idev->dev->name,
863 						  &ifp->addr, np);
864 					goto out;
865 				}
866 				/*
867 				 * We are colliding with another node
868 				 * who is doing DAD
869 				 * so fail our DAD process
870 				 */
871 				addrconf_dad_failure(skb, ifp);
872 				return reason;
873 			} else {
874 				/*
875 				 * This is not a dad solicitation.
876 				 * If we are an optimistic node,
877 				 * we should respond.
878 				 * Otherwise, we should ignore it.
879 				 */
880 				if (!(ifp->flags & IFA_F_OPTIMISTIC))
881 					goto out;
882 			}
883 		}
884 
885 		idev = ifp->idev;
886 	} else {
887 		struct net *net = dev_net(dev);
888 
889 		/* perhaps an address on the master device */
890 		if (netif_is_l3_slave(dev)) {
891 			struct net_device *mdev;
892 
893 			mdev = netdev_master_upper_dev_get_rcu(dev);
894 			if (mdev) {
895 				ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
896 				if (ifp)
897 					goto have_ifp;
898 			}
899 		}
900 
901 		idev = in6_dev_get(dev);
902 		if (!idev) {
903 			/* XXX: count this drop? */
904 			return reason;
905 		}
906 
907 		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
908 		    (READ_ONCE(idev->cnf.forwarding) &&
909 		     (READ_ONCE(net->ipv6.devconf_all->proxy_ndp) ||
910 		      READ_ONCE(idev->cnf.proxy_ndp)) &&
911 		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
912 			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
913 			    skb->pkt_type != PACKET_HOST &&
914 			    inc &&
915 			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
916 				/*
917 				 * for anycast or proxy,
918 				 * sender should delay its response
919 				 * by a random time between 0 and
920 				 * MAX_ANYCAST_DELAY_TIME seconds.
921 				 * (RFC2461) -- yoshfuji
922 				 */
923 				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
924 				if (n)
925 					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
926 				goto out;
927 			}
928 		} else {
929 			SKB_DR_SET(reason, IPV6_NDISC_NS_OTHERHOST);
930 			goto out;
931 		}
932 	}
933 
934 	if (is_router < 0)
935 		is_router = READ_ONCE(idev->cnf.forwarding);
936 
937 	if (dad) {
938 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
939 			      !!is_router, false, (ifp != NULL), true);
940 		goto out;
941 	}
942 
943 	if (inc)
944 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
945 	else
946 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
947 
948 	/*
949 	 *	update / create cache entry
950 	 *	for the source address
951 	 */
952 	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
953 			       !inc || lladdr || !dev->addr_len);
954 	if (neigh)
955 		ndisc_update(dev, neigh, lladdr, NUD_STALE,
956 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
957 			     NEIGH_UPDATE_F_OVERRIDE,
958 			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
959 	if (neigh || !dev->header_ops) {
960 		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
961 			      true, (ifp != NULL && inc), inc);
962 		if (neigh)
963 			neigh_release(neigh);
964 		reason = SKB_CONSUMED;
965 	}
966 
967 out:
968 	if (ifp)
969 		in6_ifa_put(ifp);
970 	else
971 		in6_dev_put(idev);
972 	return reason;
973 }
974 
975 static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
976 {
977 	struct inet6_dev *idev = __in6_dev_get(dev);
978 
979 	switch (READ_ONCE(idev->cnf.accept_untracked_na)) {
980 	case 0: /* Don't accept untracked na (absent in neighbor cache) */
981 		return 0;
982 	case 1: /* Create new entries from na if currently untracked */
983 		return 1;
984 	case 2: /* Create new entries from untracked na only if saddr is in the
985 		 * same subnet as an address configured on the interface that
986 		 * received the na
987 		 */
988 		return !!ipv6_chk_prefix(saddr, dev);
989 	default:
990 		return 0;
991 	}
992 }
993 
994 static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)
995 {
996 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
997 	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
998 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
999 	u8 *lladdr = NULL;
1000 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1001 				    offsetof(struct nd_msg, opt));
1002 	struct ndisc_options ndopts;
1003 	struct net_device *dev = skb->dev;
1004 	struct inet6_dev *idev = __in6_dev_get(dev);
1005 	struct inet6_ifaddr *ifp;
1006 	struct neighbour *neigh;
1007 	SKB_DR(reason);
1008 	u8 new_state;
1009 
1010 	if (skb->len < sizeof(struct nd_msg))
1011 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1012 
1013 	if (ipv6_addr_is_multicast(&msg->target)) {
1014 		ND_PRINTK(2, warn, "NA: target address is multicast\n");
1015 		return reason;
1016 	}
1017 
1018 	if (ipv6_addr_is_multicast(daddr) &&
1019 	    msg->icmph.icmp6_solicited) {
1020 		ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
1021 		return reason;
1022 	}
1023 
1024 	/* For some 802.11 wireless deployments (and possibly other networks),
1025 	 * there will be a NA proxy and unsolicitd packets are attacks
1026 	 * and thus should not be accepted.
1027 	 * drop_unsolicited_na takes precedence over accept_untracked_na
1028 	 */
1029 	if (!msg->icmph.icmp6_solicited && idev &&
1030 	    READ_ONCE(idev->cnf.drop_unsolicited_na))
1031 		return reason;
1032 
1033 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
1034 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1035 
1036 	if (ndopts.nd_opts_tgt_lladdr) {
1037 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1038 		if (!lladdr) {
1039 			ND_PRINTK(2, warn,
1040 				  "NA: invalid link-layer address length\n");
1041 			return reason;
1042 		}
1043 	}
1044 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1045 	if (ifp) {
1046 		if (skb->pkt_type != PACKET_LOOPBACK
1047 		    && (ifp->flags & IFA_F_TENTATIVE)) {
1048 				addrconf_dad_failure(skb, ifp);
1049 				return reason;
1050 		}
1051 		/* What should we make now? The advertisement
1052 		   is invalid, but ndisc specs say nothing
1053 		   about it. It could be misconfiguration, or
1054 		   an smart proxy agent tries to help us :-)
1055 
1056 		   We should not print the error if NA has been
1057 		   received from loopback - it is just our own
1058 		   unsolicited advertisement.
1059 		 */
1060 		if (skb->pkt_type != PACKET_LOOPBACK)
1061 			ND_PRINTK(1, warn,
1062 				  "NA: %pM advertised our address %pI6c on %s!\n",
1063 				  eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1064 		in6_ifa_put(ifp);
1065 		return reason;
1066 	}
1067 
1068 	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1069 
1070 	/* RFC 9131 updates original Neighbour Discovery RFC 4861.
1071 	 * NAs with Target LL Address option without a corresponding
1072 	 * entry in the neighbour cache can now create a STALE neighbour
1073 	 * cache entry on routers.
1074 	 *
1075 	 *   entry accept  fwding  solicited        behaviour
1076 	 * ------- ------  ------  ---------    ----------------------
1077 	 * present      X       X         0     Set state to STALE
1078 	 * present      X       X         1     Set state to REACHABLE
1079 	 *  absent      0       X         X     Do nothing
1080 	 *  absent      1       0         X     Do nothing
1081 	 *  absent      1       1         X     Add a new STALE entry
1082 	 *
1083 	 * Note that we don't do a (daddr == all-routers-mcast) check.
1084 	 */
1085 	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1086 	if (!neigh && lladdr && idev && READ_ONCE(idev->cnf.forwarding)) {
1087 		if (accept_untracked_na(dev, saddr)) {
1088 			neigh = neigh_create(&nd_tbl, &msg->target, dev);
1089 			new_state = NUD_STALE;
1090 		}
1091 	}
1092 
1093 	if (neigh && !IS_ERR(neigh)) {
1094 		u8 old_flags = neigh->flags;
1095 		struct net *net = dev_net(dev);
1096 
1097 		if (READ_ONCE(neigh->nud_state) & NUD_FAILED)
1098 			goto out;
1099 
1100 		/*
1101 		 * Don't update the neighbor cache entry on a proxy NA from
1102 		 * ourselves because either the proxied node is off link or it
1103 		 * has already sent a NA to us.
1104 		 */
1105 		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1106 		    READ_ONCE(net->ipv6.devconf_all->forwarding) &&
1107 		    READ_ONCE(net->ipv6.devconf_all->proxy_ndp) &&
1108 		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1109 			/* XXX: idev->cnf.proxy_ndp */
1110 			goto out;
1111 		}
1112 
1113 		ndisc_update(dev, neigh, lladdr,
1114 			     new_state,
1115 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1116 			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1117 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1118 			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1119 			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1120 
1121 		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1122 			/*
1123 			 * Change: router to host
1124 			 */
1125 			rt6_clean_tohost(dev_net(dev),  saddr);
1126 		}
1127 		reason = SKB_CONSUMED;
1128 out:
1129 		neigh_release(neigh);
1130 	}
1131 	return reason;
1132 }
1133 
1134 static enum skb_drop_reason ndisc_recv_rs(struct sk_buff *skb)
1135 {
1136 	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1137 	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1138 	struct neighbour *neigh;
1139 	struct inet6_dev *idev;
1140 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1141 	struct ndisc_options ndopts;
1142 	u8 *lladdr = NULL;
1143 	SKB_DR(reason);
1144 
1145 	if (skb->len < sizeof(*rs_msg))
1146 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1147 
1148 	idev = __in6_dev_get(skb->dev);
1149 	if (!idev) {
1150 		ND_PRINTK(1, err, "RS: can't find in6 device\n");
1151 		return reason;
1152 	}
1153 
1154 	/* Don't accept RS if we're not in router mode */
1155 	if (!READ_ONCE(idev->cnf.forwarding))
1156 		goto out;
1157 
1158 	/*
1159 	 * Don't update NCE if src = ::;
1160 	 * this implies that the source node has no ip address assigned yet.
1161 	 */
1162 	if (ipv6_addr_any(saddr))
1163 		goto out;
1164 
1165 	/* Parse ND options */
1166 	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts))
1167 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1168 
1169 	if (ndopts.nd_opts_src_lladdr) {
1170 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1171 					     skb->dev);
1172 		if (!lladdr)
1173 			goto out;
1174 	}
1175 
1176 	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1177 	if (neigh) {
1178 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1179 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1180 			     NEIGH_UPDATE_F_OVERRIDE|
1181 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1182 			     NDISC_ROUTER_SOLICITATION, &ndopts);
1183 		neigh_release(neigh);
1184 		reason = SKB_CONSUMED;
1185 	}
1186 out:
1187 	return reason;
1188 }
1189 
1190 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1191 {
1192 	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1193 	struct sk_buff *skb;
1194 	struct nlmsghdr *nlh;
1195 	struct nduseroptmsg *ndmsg;
1196 	struct net *net = dev_net(ra->dev);
1197 	int err;
1198 	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1199 				    + (opt->nd_opt_len << 3));
1200 	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1201 
1202 	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1203 	if (!skb) {
1204 		err = -ENOBUFS;
1205 		goto errout;
1206 	}
1207 
1208 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1209 	if (!nlh) {
1210 		goto nla_put_failure;
1211 	}
1212 
1213 	ndmsg = nlmsg_data(nlh);
1214 	ndmsg->nduseropt_family = AF_INET6;
1215 	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1216 	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1217 	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1218 	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1219 
1220 	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1221 
1222 	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1223 		goto nla_put_failure;
1224 	nlmsg_end(skb, nlh);
1225 
1226 	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1227 	return;
1228 
1229 nla_put_failure:
1230 	nlmsg_free(skb);
1231 	err = -EMSGSIZE;
1232 errout:
1233 	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1234 }
1235 
1236 static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb)
1237 {
1238 	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1239 	bool send_ifinfo_notify = false;
1240 	struct neighbour *neigh = NULL;
1241 	struct ndisc_options ndopts;
1242 	struct fib6_info *rt = NULL;
1243 	struct inet6_dev *in6_dev;
1244 	struct fib6_table *table;
1245 	u32 defrtr_usr_metric;
1246 	unsigned int pref = 0;
1247 	__u32 old_if_flags;
1248 	struct net *net;
1249 	SKB_DR(reason);
1250 	int lifetime;
1251 	int optlen;
1252 
1253 	__u8 *opt = (__u8 *)(ra_msg + 1);
1254 
1255 	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1256 		sizeof(struct ra_msg);
1257 
1258 	ND_PRINTK(2, info,
1259 		  "RA: %s, dev: %s\n",
1260 		  __func__, skb->dev->name);
1261 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1262 		ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1263 		return reason;
1264 	}
1265 	if (optlen < 0)
1266 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1267 
1268 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1269 	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1270 		ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1271 		return reason;
1272 	}
1273 #endif
1274 
1275 	in6_dev = __in6_dev_get(skb->dev);
1276 	if (!in6_dev) {
1277 		ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1278 			  skb->dev->name);
1279 		return reason;
1280 	}
1281 
1282 	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts))
1283 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1284 
1285 	if (!ipv6_accept_ra(in6_dev)) {
1286 		ND_PRINTK(2, info,
1287 			  "RA: %s, did not accept ra for dev: %s\n",
1288 			  __func__, skb->dev->name);
1289 		goto skip_linkparms;
1290 	}
1291 
1292 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1293 	/* skip link-specific parameters from interior routers */
1294 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1295 		ND_PRINTK(2, info,
1296 			  "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1297 			  __func__, skb->dev->name);
1298 		goto skip_linkparms;
1299 	}
1300 #endif
1301 
1302 	if (in6_dev->if_flags & IF_RS_SENT) {
1303 		/*
1304 		 *	flag that an RA was received after an RS was sent
1305 		 *	out on this interface.
1306 		 */
1307 		in6_dev->if_flags |= IF_RA_RCVD;
1308 	}
1309 
1310 	/*
1311 	 * Remember the managed/otherconf flags from most recently
1312 	 * received RA message (RFC 2462) -- yoshfuji
1313 	 */
1314 	old_if_flags = in6_dev->if_flags;
1315 	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1316 				IF_RA_OTHERCONF)) |
1317 				(ra_msg->icmph.icmp6_addrconf_managed ?
1318 					IF_RA_MANAGED : 0) |
1319 				(ra_msg->icmph.icmp6_addrconf_other ?
1320 					IF_RA_OTHERCONF : 0);
1321 
1322 	if (old_if_flags != in6_dev->if_flags)
1323 		send_ifinfo_notify = true;
1324 
1325 	if (!READ_ONCE(in6_dev->cnf.accept_ra_defrtr)) {
1326 		ND_PRINTK(2, info,
1327 			  "RA: %s, defrtr is false for dev: %s\n",
1328 			  __func__, skb->dev->name);
1329 		goto skip_defrtr;
1330 	}
1331 
1332 	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1333 	if (lifetime != 0 &&
1334 	    lifetime < READ_ONCE(in6_dev->cnf.accept_ra_min_lft)) {
1335 		ND_PRINTK(2, info,
1336 			  "RA: router lifetime (%ds) is too short: %s\n",
1337 			  lifetime, skb->dev->name);
1338 		goto skip_defrtr;
1339 	}
1340 
1341 	/* Do not accept RA with source-addr found on local machine unless
1342 	 * accept_ra_from_local is set to true.
1343 	 */
1344 	net = dev_net(in6_dev->dev);
1345 	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1346 	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1347 		ND_PRINTK(2, info,
1348 			  "RA from local address detected on dev: %s: default router ignored\n",
1349 			  skb->dev->name);
1350 		goto skip_defrtr;
1351 	}
1352 
1353 #ifdef CONFIG_IPV6_ROUTER_PREF
1354 	pref = ra_msg->icmph.icmp6_router_pref;
1355 	/* 10b is handled as if it were 00b (medium) */
1356 	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1357 	    !READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref))
1358 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1359 #endif
1360 	/* routes added from RAs do not use nexthop objects */
1361 	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1362 	if (rt) {
1363 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1364 					 rt->fib6_nh->fib_nh_dev, NULL,
1365 					  &ipv6_hdr(skb)->saddr);
1366 		if (!neigh) {
1367 			ND_PRINTK(0, err,
1368 				  "RA: %s got default router without neighbour\n",
1369 				  __func__);
1370 			fib6_info_release(rt);
1371 			return reason;
1372 		}
1373 	}
1374 	/* Set default route metric as specified by user */
1375 	defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1376 	/* delete the route if lifetime is 0 or if metric needs change */
1377 	if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1378 		ip6_del_rt(net, rt, false);
1379 		rt = NULL;
1380 	}
1381 
1382 	ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, metric: %d, for dev: %s\n",
1383 		  rt, lifetime, defrtr_usr_metric, skb->dev->name);
1384 	if (!rt && lifetime) {
1385 		ND_PRINTK(3, info, "RA: adding default router\n");
1386 
1387 		if (neigh)
1388 			neigh_release(neigh);
1389 
1390 		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1391 					 skb->dev, pref, defrtr_usr_metric,
1392 					 lifetime);
1393 		if (!rt) {
1394 			ND_PRINTK(0, err,
1395 				  "RA: %s failed to add default route\n",
1396 				  __func__);
1397 			return reason;
1398 		}
1399 
1400 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1401 					 rt->fib6_nh->fib_nh_dev, NULL,
1402 					  &ipv6_hdr(skb)->saddr);
1403 		if (!neigh) {
1404 			ND_PRINTK(0, err,
1405 				  "RA: %s got default router without neighbour\n",
1406 				  __func__);
1407 			fib6_info_release(rt);
1408 			return reason;
1409 		}
1410 		neigh->flags |= NTF_ROUTER;
1411 	} else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1412 		struct nl_info nlinfo = {
1413 			.nl_net = net,
1414 		};
1415 		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1416 		inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1417 	}
1418 
1419 	if (rt) {
1420 		table = rt->fib6_table;
1421 		spin_lock_bh(&table->tb6_lock);
1422 
1423 		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1424 		fib6_add_gc_list(rt);
1425 
1426 		spin_unlock_bh(&table->tb6_lock);
1427 	}
1428 	if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) < 256 &&
1429 	    ra_msg->icmph.icmp6_hop_limit) {
1430 		if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) <=
1431 		    ra_msg->icmph.icmp6_hop_limit) {
1432 			WRITE_ONCE(in6_dev->cnf.hop_limit,
1433 				   ra_msg->icmph.icmp6_hop_limit);
1434 			fib6_metric_set(rt, RTAX_HOPLIMIT,
1435 					ra_msg->icmph.icmp6_hop_limit);
1436 		} else {
1437 			ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1438 		}
1439 	}
1440 
1441 skip_defrtr:
1442 
1443 	/*
1444 	 *	Update Reachable Time and Retrans Timer
1445 	 */
1446 
1447 	if (in6_dev->nd_parms) {
1448 		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1449 
1450 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1451 			rtime = (rtime*HZ)/1000;
1452 			if (rtime < HZ/100)
1453 				rtime = HZ/100;
1454 			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1455 			in6_dev->tstamp = jiffies;
1456 			send_ifinfo_notify = true;
1457 		}
1458 
1459 		rtime = ntohl(ra_msg->reachable_time);
1460 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1461 			rtime = (rtime*HZ)/1000;
1462 
1463 			if (rtime < HZ/10)
1464 				rtime = HZ/10;
1465 
1466 			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1467 				NEIGH_VAR_SET(in6_dev->nd_parms,
1468 					      BASE_REACHABLE_TIME, rtime);
1469 				NEIGH_VAR_SET(in6_dev->nd_parms,
1470 					      GC_STALETIME, 3 * rtime);
1471 				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1472 				in6_dev->tstamp = jiffies;
1473 				send_ifinfo_notify = true;
1474 			}
1475 		}
1476 	}
1477 
1478 skip_linkparms:
1479 
1480 	/*
1481 	 *	Process options.
1482 	 */
1483 
1484 	if (!neigh)
1485 		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1486 				       skb->dev, 1);
1487 	if (neigh) {
1488 		u8 *lladdr = NULL;
1489 		if (ndopts.nd_opts_src_lladdr) {
1490 			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1491 						     skb->dev);
1492 			if (!lladdr) {
1493 				ND_PRINTK(2, warn,
1494 					  "RA: invalid link-layer address length\n");
1495 				goto out;
1496 			}
1497 		}
1498 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1499 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1500 			     NEIGH_UPDATE_F_OVERRIDE|
1501 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1502 			     NEIGH_UPDATE_F_ISROUTER,
1503 			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1504 		reason = SKB_CONSUMED;
1505 	}
1506 
1507 	if (!ipv6_accept_ra(in6_dev)) {
1508 		ND_PRINTK(2, info,
1509 			  "RA: %s, accept_ra is false for dev: %s\n",
1510 			  __func__, skb->dev->name);
1511 		goto out;
1512 	}
1513 
1514 #ifdef CONFIG_IPV6_ROUTE_INFO
1515 	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1516 	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1517 			  in6_dev->dev, 0)) {
1518 		ND_PRINTK(2, info,
1519 			  "RA from local address detected on dev: %s: router info ignored.\n",
1520 			  skb->dev->name);
1521 		goto skip_routeinfo;
1522 	}
1523 
1524 	if (READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref) && ndopts.nd_opts_ri) {
1525 		struct nd_opt_hdr *p;
1526 		for (p = ndopts.nd_opts_ri;
1527 		     p;
1528 		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1529 			struct route_info *ri = (struct route_info *)p;
1530 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1531 			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1532 			    ri->prefix_len == 0)
1533 				continue;
1534 #endif
1535 			if (ri->prefix_len == 0 &&
1536 			    !READ_ONCE(in6_dev->cnf.accept_ra_defrtr))
1537 				continue;
1538 			if (ri->lifetime != 0 &&
1539 			    ntohl(ri->lifetime) < READ_ONCE(in6_dev->cnf.accept_ra_min_lft))
1540 				continue;
1541 			if (ri->prefix_len < READ_ONCE(in6_dev->cnf.accept_ra_rt_info_min_plen))
1542 				continue;
1543 			if (ri->prefix_len > READ_ONCE(in6_dev->cnf.accept_ra_rt_info_max_plen))
1544 				continue;
1545 			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1546 				      &ipv6_hdr(skb)->saddr);
1547 		}
1548 	}
1549 
1550 skip_routeinfo:
1551 #endif
1552 
1553 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1554 	/* skip link-specific ndopts from interior routers */
1555 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1556 		ND_PRINTK(2, info,
1557 			  "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1558 			  __func__, skb->dev->name);
1559 		goto out;
1560 	}
1561 #endif
1562 
1563 	if (READ_ONCE(in6_dev->cnf.accept_ra_pinfo) && ndopts.nd_opts_pi) {
1564 		struct nd_opt_hdr *p;
1565 		for (p = ndopts.nd_opts_pi;
1566 		     p;
1567 		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1568 			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1569 					    (p->nd_opt_len) << 3,
1570 					    ndopts.nd_opts_src_lladdr != NULL);
1571 		}
1572 	}
1573 
1574 	if (ndopts.nd_opts_mtu && READ_ONCE(in6_dev->cnf.accept_ra_mtu)) {
1575 		__be32 n;
1576 		u32 mtu;
1577 
1578 		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1579 		mtu = ntohl(n);
1580 
1581 		if (in6_dev->ra_mtu != mtu) {
1582 			in6_dev->ra_mtu = mtu;
1583 			send_ifinfo_notify = true;
1584 		}
1585 
1586 		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1587 			ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1588 		} else if (READ_ONCE(in6_dev->cnf.mtu6) != mtu) {
1589 			WRITE_ONCE(in6_dev->cnf.mtu6, mtu);
1590 			fib6_metric_set(rt, RTAX_MTU, mtu);
1591 			rt6_mtu_change(skb->dev, mtu);
1592 		}
1593 	}
1594 
1595 	if (ndopts.nd_useropts) {
1596 		struct nd_opt_hdr *p;
1597 		for (p = ndopts.nd_useropts;
1598 		     p;
1599 		     p = ndisc_next_useropt(skb->dev, p,
1600 					    ndopts.nd_useropts_end)) {
1601 			ndisc_ra_useropt(skb, p);
1602 		}
1603 	}
1604 
1605 	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1606 		ND_PRINTK(2, warn, "RA: invalid RA options\n");
1607 	}
1608 out:
1609 	/* Send a notify if RA changed managed/otherconf flags or
1610 	 * timer settings or ra_mtu value
1611 	 */
1612 	if (send_ifinfo_notify)
1613 		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1614 
1615 	fib6_info_release(rt);
1616 	if (neigh)
1617 		neigh_release(neigh);
1618 	return reason;
1619 }
1620 
1621 static enum skb_drop_reason ndisc_redirect_rcv(struct sk_buff *skb)
1622 {
1623 	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1624 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1625 				    offsetof(struct rd_msg, opt));
1626 	struct ndisc_options ndopts;
1627 	SKB_DR(reason);
1628 	u8 *hdr;
1629 
1630 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1631 	switch (skb->ndisc_nodetype) {
1632 	case NDISC_NODETYPE_HOST:
1633 	case NDISC_NODETYPE_NODEFAULT:
1634 		ND_PRINTK(2, warn,
1635 			  "Redirect: from host or unauthorized router\n");
1636 		return reason;
1637 	}
1638 #endif
1639 
1640 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1641 		ND_PRINTK(2, warn,
1642 			  "Redirect: source address is not link-local\n");
1643 		return reason;
1644 	}
1645 
1646 	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1647 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1648 
1649 	if (!ndopts.nd_opts_rh) {
1650 		ip6_redirect_no_header(skb, dev_net(skb->dev),
1651 					skb->dev->ifindex);
1652 		return reason;
1653 	}
1654 
1655 	hdr = (u8 *)ndopts.nd_opts_rh;
1656 	hdr += 8;
1657 	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1658 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1659 
1660 	return icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1661 }
1662 
1663 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1664 					   struct sk_buff *orig_skb,
1665 					   int rd_len)
1666 {
1667 	u8 *opt = skb_put(skb, rd_len);
1668 
1669 	memset(opt, 0, 8);
1670 	*(opt++) = ND_OPT_REDIRECT_HDR;
1671 	*(opt++) = (rd_len >> 3);
1672 	opt += 6;
1673 
1674 	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1675 		      rd_len - 8);
1676 }
1677 
1678 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1679 {
1680 	struct net_device *dev = skb->dev;
1681 	struct net *net = dev_net(dev);
1682 	struct sock *sk = net->ipv6.ndisc_sk;
1683 	int optlen = 0;
1684 	struct inet_peer *peer;
1685 	struct sk_buff *buff;
1686 	struct rd_msg *msg;
1687 	struct in6_addr saddr_buf;
1688 	struct rt6_info *rt;
1689 	struct dst_entry *dst;
1690 	struct flowi6 fl6;
1691 	int rd_len;
1692 	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1693 	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1694 	bool ret;
1695 
1696 	if (netif_is_l3_master(skb->dev)) {
1697 		dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1698 		if (!dev)
1699 			return;
1700 	}
1701 
1702 	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1703 		ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1704 			  dev->name);
1705 		return;
1706 	}
1707 
1708 	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1709 	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1710 		ND_PRINTK(2, warn,
1711 			  "Redirect: target address is not link-local unicast\n");
1712 		return;
1713 	}
1714 
1715 	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1716 			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1717 
1718 	dst = ip6_route_output(net, NULL, &fl6);
1719 	if (dst->error) {
1720 		dst_release(dst);
1721 		return;
1722 	}
1723 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1724 	if (IS_ERR(dst))
1725 		return;
1726 
1727 	rt = dst_rt6_info(dst);
1728 
1729 	if (rt->rt6i_flags & RTF_GATEWAY) {
1730 		ND_PRINTK(2, warn,
1731 			  "Redirect: destination is not a neighbour\n");
1732 		goto release;
1733 	}
1734 	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1735 	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1736 	if (peer)
1737 		inet_putpeer(peer);
1738 	if (!ret)
1739 		goto release;
1740 
1741 	if (dev->addr_len) {
1742 		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1743 		if (!neigh) {
1744 			ND_PRINTK(2, warn,
1745 				  "Redirect: no neigh for target address\n");
1746 			goto release;
1747 		}
1748 
1749 		read_lock_bh(&neigh->lock);
1750 		if (neigh->nud_state & NUD_VALID) {
1751 			memcpy(ha_buf, neigh->ha, dev->addr_len);
1752 			read_unlock_bh(&neigh->lock);
1753 			ha = ha_buf;
1754 			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1755 								ops_data_buf,
1756 								&ops_data);
1757 		} else
1758 			read_unlock_bh(&neigh->lock);
1759 
1760 		neigh_release(neigh);
1761 	}
1762 
1763 	rd_len = min_t(unsigned int,
1764 		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1765 		       skb->len + 8);
1766 	rd_len &= ~0x7;
1767 	optlen += rd_len;
1768 
1769 	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1770 	if (!buff)
1771 		goto release;
1772 
1773 	msg = skb_put(buff, sizeof(*msg));
1774 	*msg = (struct rd_msg) {
1775 		.icmph = {
1776 			.icmp6_type = NDISC_REDIRECT,
1777 		},
1778 		.target = *target,
1779 		.dest = ipv6_hdr(skb)->daddr,
1780 	};
1781 
1782 	/*
1783 	 *	include target_address option
1784 	 */
1785 
1786 	if (ha)
1787 		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1788 
1789 	/*
1790 	 *	build redirect option and copy skb over to the new packet.
1791 	 */
1792 
1793 	if (rd_len)
1794 		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1795 
1796 	skb_dst_set(buff, dst);
1797 	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1798 	return;
1799 
1800 release:
1801 	dst_release(dst);
1802 }
1803 
1804 static void pndisc_redo(struct sk_buff *skb)
1805 {
1806 	enum skb_drop_reason reason = ndisc_recv_ns(skb);
1807 
1808 	kfree_skb_reason(skb, reason);
1809 }
1810 
1811 static int ndisc_is_multicast(const void *pkey)
1812 {
1813 	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1814 }
1815 
1816 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1817 {
1818 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1819 
1820 	if (!idev)
1821 		return true;
1822 	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1823 	    READ_ONCE(idev->cnf.suppress_frag_ndisc)) {
1824 		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1825 		return true;
1826 	}
1827 	return false;
1828 }
1829 
1830 enum skb_drop_reason ndisc_rcv(struct sk_buff *skb)
1831 {
1832 	struct nd_msg *msg;
1833 	SKB_DR(reason);
1834 
1835 	if (ndisc_suppress_frag_ndisc(skb))
1836 		return SKB_DROP_REASON_IPV6_NDISC_FRAG;
1837 
1838 	if (skb_linearize(skb))
1839 		return SKB_DROP_REASON_NOMEM;
1840 
1841 	msg = (struct nd_msg *)skb_transport_header(skb);
1842 
1843 	__skb_push(skb, skb->data - skb_transport_header(skb));
1844 
1845 	if (ipv6_hdr(skb)->hop_limit != 255) {
1846 		ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1847 			  ipv6_hdr(skb)->hop_limit);
1848 		return SKB_DROP_REASON_IPV6_NDISC_HOP_LIMIT;
1849 	}
1850 
1851 	if (msg->icmph.icmp6_code != 0) {
1852 		ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1853 			  msg->icmph.icmp6_code);
1854 		return SKB_DROP_REASON_IPV6_NDISC_BAD_CODE;
1855 	}
1856 
1857 	switch (msg->icmph.icmp6_type) {
1858 	case NDISC_NEIGHBOUR_SOLICITATION:
1859 		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1860 		reason = ndisc_recv_ns(skb);
1861 		break;
1862 
1863 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1864 		reason = ndisc_recv_na(skb);
1865 		break;
1866 
1867 	case NDISC_ROUTER_SOLICITATION:
1868 		reason = ndisc_recv_rs(skb);
1869 		break;
1870 
1871 	case NDISC_ROUTER_ADVERTISEMENT:
1872 		reason = ndisc_router_discovery(skb);
1873 		break;
1874 
1875 	case NDISC_REDIRECT:
1876 		reason = ndisc_redirect_rcv(skb);
1877 		break;
1878 	}
1879 
1880 	return reason;
1881 }
1882 
1883 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1884 {
1885 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1886 	struct netdev_notifier_change_info *change_info;
1887 	struct net *net = dev_net(dev);
1888 	struct inet6_dev *idev;
1889 	bool evict_nocarrier;
1890 
1891 	switch (event) {
1892 	case NETDEV_CHANGEADDR:
1893 		neigh_changeaddr(&nd_tbl, dev);
1894 		fib6_run_gc(0, net, false);
1895 		fallthrough;
1896 	case NETDEV_UP:
1897 		idev = in6_dev_get(dev);
1898 		if (!idev)
1899 			break;
1900 		if (READ_ONCE(idev->cnf.ndisc_notify) ||
1901 		    READ_ONCE(net->ipv6.devconf_all->ndisc_notify))
1902 			ndisc_send_unsol_na(dev);
1903 		in6_dev_put(idev);
1904 		break;
1905 	case NETDEV_CHANGE:
1906 		idev = in6_dev_get(dev);
1907 		if (!idev)
1908 			evict_nocarrier = true;
1909 		else {
1910 			evict_nocarrier = READ_ONCE(idev->cnf.ndisc_evict_nocarrier) &&
1911 					  READ_ONCE(net->ipv6.devconf_all->ndisc_evict_nocarrier);
1912 			in6_dev_put(idev);
1913 		}
1914 
1915 		change_info = ptr;
1916 		if (change_info->flags_changed & IFF_NOARP)
1917 			neigh_changeaddr(&nd_tbl, dev);
1918 		if (evict_nocarrier && !netif_carrier_ok(dev))
1919 			neigh_carrier_down(&nd_tbl, dev);
1920 		break;
1921 	case NETDEV_DOWN:
1922 		neigh_ifdown(&nd_tbl, dev);
1923 		fib6_run_gc(0, net, false);
1924 		break;
1925 	case NETDEV_NOTIFY_PEERS:
1926 		ndisc_send_unsol_na(dev);
1927 		break;
1928 	default:
1929 		break;
1930 	}
1931 
1932 	return NOTIFY_DONE;
1933 }
1934 
1935 static struct notifier_block ndisc_netdev_notifier = {
1936 	.notifier_call = ndisc_netdev_event,
1937 	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1938 };
1939 
1940 #ifdef CONFIG_SYSCTL
1941 static void ndisc_warn_deprecated_sysctl(const struct ctl_table *ctl,
1942 					 const char *func, const char *dev_name)
1943 {
1944 	static char warncomm[TASK_COMM_LEN];
1945 	static int warned;
1946 	if (strcmp(warncomm, current->comm) && warned < 5) {
1947 		strcpy(warncomm, current->comm);
1948 		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1949 			warncomm, func,
1950 			dev_name, ctl->procname,
1951 			dev_name, ctl->procname);
1952 		warned++;
1953 	}
1954 }
1955 
1956 int ndisc_ifinfo_sysctl_change(const struct ctl_table *ctl, int write, void *buffer,
1957 		size_t *lenp, loff_t *ppos)
1958 {
1959 	struct net_device *dev = ctl->extra1;
1960 	struct inet6_dev *idev;
1961 	int ret;
1962 
1963 	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1964 	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1965 		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1966 
1967 	if (strcmp(ctl->procname, "retrans_time") == 0)
1968 		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1969 
1970 	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1971 		ret = neigh_proc_dointvec_jiffies(ctl, write,
1972 						  buffer, lenp, ppos);
1973 
1974 	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1975 		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1976 		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1977 						     buffer, lenp, ppos);
1978 	else
1979 		ret = -1;
1980 
1981 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1982 		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1983 			idev->nd_parms->reachable_time =
1984 					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1985 		WRITE_ONCE(idev->tstamp, jiffies);
1986 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1987 		in6_dev_put(idev);
1988 	}
1989 	return ret;
1990 }
1991 
1992 
1993 #endif
1994 
1995 static int __net_init ndisc_net_init(struct net *net)
1996 {
1997 	struct ipv6_pinfo *np;
1998 	struct sock *sk;
1999 	int err;
2000 
2001 	err = inet_ctl_sock_create(&sk, PF_INET6,
2002 				   SOCK_RAW, IPPROTO_ICMPV6, net);
2003 	if (err < 0) {
2004 		ND_PRINTK(0, err,
2005 			  "NDISC: Failed to initialize the control socket (err %d)\n",
2006 			  err);
2007 		return err;
2008 	}
2009 
2010 	net->ipv6.ndisc_sk = sk;
2011 
2012 	np = inet6_sk(sk);
2013 	np->hop_limit = 255;
2014 	/* Do not loopback ndisc messages */
2015 	inet6_clear_bit(MC6_LOOP, sk);
2016 
2017 	return 0;
2018 }
2019 
2020 static void __net_exit ndisc_net_exit(struct net *net)
2021 {
2022 	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
2023 }
2024 
2025 static struct pernet_operations ndisc_net_ops = {
2026 	.init = ndisc_net_init,
2027 	.exit = ndisc_net_exit,
2028 };
2029 
2030 int __init ndisc_init(void)
2031 {
2032 	int err;
2033 
2034 	err = register_pernet_subsys(&ndisc_net_ops);
2035 	if (err)
2036 		return err;
2037 	/*
2038 	 * Initialize the neighbour table
2039 	 */
2040 	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2041 
2042 #ifdef CONFIG_SYSCTL
2043 	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2044 				    ndisc_ifinfo_sysctl_change);
2045 	if (err)
2046 		goto out_unregister_pernet;
2047 out:
2048 #endif
2049 	return err;
2050 
2051 #ifdef CONFIG_SYSCTL
2052 out_unregister_pernet:
2053 	unregister_pernet_subsys(&ndisc_net_ops);
2054 	goto out;
2055 #endif
2056 }
2057 
2058 int __init ndisc_late_init(void)
2059 {
2060 	return register_netdevice_notifier(&ndisc_netdev_notifier);
2061 }
2062 
2063 void ndisc_late_cleanup(void)
2064 {
2065 	unregister_netdevice_notifier(&ndisc_netdev_notifier);
2066 }
2067 
2068 void ndisc_cleanup(void)
2069 {
2070 #ifdef CONFIG_SYSCTL
2071 	neigh_sysctl_unregister(&nd_tbl.parms);
2072 #endif
2073 	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2074 	unregister_pernet_subsys(&ndisc_net_ops);
2075 }
2076