xref: /openbmc/linux/net/ipv6/ndisc.c (revision 67dffd3d)
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_RDNSS ||
201 		opt->nd_opt_type == ND_OPT_DNSSL ||
202 		opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
203 		opt->nd_opt_type == ND_OPT_PREF64 ||
204 		ndisc_ops_is_useropt(dev, opt->nd_opt_type);
205 }
206 
207 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
208 					     struct nd_opt_hdr *cur,
209 					     struct nd_opt_hdr *end)
210 {
211 	if (!cur || !end || cur >= end)
212 		return NULL;
213 	do {
214 		cur = ((void *)cur) + (cur->nd_opt_len << 3);
215 	} while (cur < end && !ndisc_is_useropt(dev, cur));
216 	return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
217 }
218 
219 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
220 					  u8 *opt, int opt_len,
221 					  struct ndisc_options *ndopts)
222 {
223 	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
224 
225 	if (!nd_opt || opt_len < 0 || !ndopts)
226 		return NULL;
227 	memset(ndopts, 0, sizeof(*ndopts));
228 	while (opt_len) {
229 		int l;
230 		if (opt_len < sizeof(struct nd_opt_hdr))
231 			return NULL;
232 		l = nd_opt->nd_opt_len << 3;
233 		if (opt_len < l || l == 0)
234 			return NULL;
235 		if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
236 			goto next_opt;
237 		switch (nd_opt->nd_opt_type) {
238 		case ND_OPT_SOURCE_LL_ADDR:
239 		case ND_OPT_TARGET_LL_ADDR:
240 		case ND_OPT_MTU:
241 		case ND_OPT_NONCE:
242 		case ND_OPT_REDIRECT_HDR:
243 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
244 				ND_PRINTK(2, warn,
245 					  "%s: duplicated ND6 option found: type=%d\n",
246 					  __func__, nd_opt->nd_opt_type);
247 			} else {
248 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
249 			}
250 			break;
251 		case ND_OPT_PREFIX_INFO:
252 			ndopts->nd_opts_pi_end = nd_opt;
253 			if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
254 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
255 			break;
256 #ifdef CONFIG_IPV6_ROUTE_INFO
257 		case ND_OPT_ROUTE_INFO:
258 			ndopts->nd_opts_ri_end = nd_opt;
259 			if (!ndopts->nd_opts_ri)
260 				ndopts->nd_opts_ri = nd_opt;
261 			break;
262 #endif
263 		default:
264 			if (ndisc_is_useropt(dev, nd_opt)) {
265 				ndopts->nd_useropts_end = nd_opt;
266 				if (!ndopts->nd_useropts)
267 					ndopts->nd_useropts = nd_opt;
268 			} else {
269 				/*
270 				 * Unknown options must be silently ignored,
271 				 * to accommodate future extension to the
272 				 * protocol.
273 				 */
274 				ND_PRINTK(2, notice,
275 					  "%s: ignored unsupported option; type=%d, len=%d\n",
276 					  __func__,
277 					  nd_opt->nd_opt_type,
278 					  nd_opt->nd_opt_len);
279 			}
280 		}
281 next_opt:
282 		opt_len -= l;
283 		nd_opt = ((void *)nd_opt) + l;
284 	}
285 	return ndopts;
286 }
287 
288 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
289 {
290 	switch (dev->type) {
291 	case ARPHRD_ETHER:
292 	case ARPHRD_IEEE802:	/* Not sure. Check it later. --ANK */
293 	case ARPHRD_FDDI:
294 		ipv6_eth_mc_map(addr, buf);
295 		return 0;
296 	case ARPHRD_ARCNET:
297 		ipv6_arcnet_mc_map(addr, buf);
298 		return 0;
299 	case ARPHRD_INFINIBAND:
300 		ipv6_ib_mc_map(addr, dev->broadcast, buf);
301 		return 0;
302 	case ARPHRD_IPGRE:
303 		return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
304 	default:
305 		if (dir) {
306 			memcpy(buf, dev->broadcast, dev->addr_len);
307 			return 0;
308 		}
309 	}
310 	return -EINVAL;
311 }
312 EXPORT_SYMBOL(ndisc_mc_map);
313 
314 static u32 ndisc_hash(const void *pkey,
315 		      const struct net_device *dev,
316 		      __u32 *hash_rnd)
317 {
318 	return ndisc_hashfn(pkey, dev, hash_rnd);
319 }
320 
321 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
322 {
323 	return neigh_key_eq128(n, pkey);
324 }
325 
326 static int ndisc_constructor(struct neighbour *neigh)
327 {
328 	struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
329 	struct net_device *dev = neigh->dev;
330 	struct inet6_dev *in6_dev;
331 	struct neigh_parms *parms;
332 	bool is_multicast = ipv6_addr_is_multicast(addr);
333 
334 	in6_dev = in6_dev_get(dev);
335 	if (!in6_dev) {
336 		return -EINVAL;
337 	}
338 
339 	parms = in6_dev->nd_parms;
340 	__neigh_parms_put(neigh->parms);
341 	neigh->parms = neigh_parms_clone(parms);
342 
343 	neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
344 	if (!dev->header_ops) {
345 		neigh->nud_state = NUD_NOARP;
346 		neigh->ops = &ndisc_direct_ops;
347 		neigh->output = neigh_direct_output;
348 	} else {
349 		if (is_multicast) {
350 			neigh->nud_state = NUD_NOARP;
351 			ndisc_mc_map(addr, neigh->ha, dev, 1);
352 		} else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
353 			neigh->nud_state = NUD_NOARP;
354 			memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
355 			if (dev->flags&IFF_LOOPBACK)
356 				neigh->type = RTN_LOCAL;
357 		} else if (dev->flags&IFF_POINTOPOINT) {
358 			neigh->nud_state = NUD_NOARP;
359 			memcpy(neigh->ha, dev->broadcast, dev->addr_len);
360 		}
361 		if (dev->header_ops->cache)
362 			neigh->ops = &ndisc_hh_ops;
363 		else
364 			neigh->ops = &ndisc_generic_ops;
365 		if (neigh->nud_state&NUD_VALID)
366 			neigh->output = neigh->ops->connected_output;
367 		else
368 			neigh->output = neigh->ops->output;
369 	}
370 	in6_dev_put(in6_dev);
371 	return 0;
372 }
373 
374 static int pndisc_constructor(struct pneigh_entry *n)
375 {
376 	struct in6_addr *addr = (struct in6_addr *)&n->key;
377 	struct in6_addr maddr;
378 	struct net_device *dev = n->dev;
379 
380 	if (!dev || !__in6_dev_get(dev))
381 		return -EINVAL;
382 	addrconf_addr_solict_mult(addr, &maddr);
383 	ipv6_dev_mc_inc(dev, &maddr);
384 	return 0;
385 }
386 
387 static void pndisc_destructor(struct pneigh_entry *n)
388 {
389 	struct in6_addr *addr = (struct in6_addr *)&n->key;
390 	struct in6_addr maddr;
391 	struct net_device *dev = n->dev;
392 
393 	if (!dev || !__in6_dev_get(dev))
394 		return;
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 sock *sk = dev_net(dev)->ipv6.ndisc_sk;
419 	struct sk_buff *skb;
420 
421 	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
422 	if (!skb) {
423 		ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
424 			  __func__);
425 		return NULL;
426 	}
427 
428 	skb->protocol = htons(ETH_P_IPV6);
429 	skb->dev = dev;
430 
431 	skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
432 	skb_reset_transport_header(skb);
433 
434 	/* Manually assign socket ownership as we avoid calling
435 	 * sock_alloc_send_pskb() to bypass wmem buffer limits
436 	 */
437 	skb_set_owner_w(skb, sk);
438 
439 	return skb;
440 }
441 
442 static void ip6_nd_hdr(struct sk_buff *skb,
443 		       const struct in6_addr *saddr,
444 		       const struct in6_addr *daddr,
445 		       int hop_limit, int len)
446 {
447 	struct ipv6hdr *hdr;
448 	struct inet6_dev *idev;
449 	unsigned tclass;
450 
451 	rcu_read_lock();
452 	idev = __in6_dev_get(skb->dev);
453 	tclass = idev ? idev->cnf.ndisc_tclass : 0;
454 	rcu_read_unlock();
455 
456 	skb_push(skb, sizeof(*hdr));
457 	skb_reset_network_header(skb);
458 	hdr = ipv6_hdr(skb);
459 
460 	ip6_flow_hdr(hdr, tclass, 0);
461 
462 	hdr->payload_len = htons(len);
463 	hdr->nexthdr = IPPROTO_ICMPV6;
464 	hdr->hop_limit = hop_limit;
465 
466 	hdr->saddr = *saddr;
467 	hdr->daddr = *daddr;
468 }
469 
470 void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
471 		    const struct in6_addr *saddr)
472 {
473 	struct dst_entry *dst = skb_dst(skb);
474 	struct net *net = dev_net(skb->dev);
475 	struct sock *sk = net->ipv6.ndisc_sk;
476 	struct inet6_dev *idev;
477 	int err;
478 	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
479 	u8 type;
480 
481 	type = icmp6h->icmp6_type;
482 
483 	if (!dst) {
484 		struct flowi6 fl6;
485 		int oif = skb->dev->ifindex;
486 
487 		icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
488 		dst = icmp6_dst_alloc(skb->dev, &fl6);
489 		if (IS_ERR(dst)) {
490 			kfree_skb(skb);
491 			return;
492 		}
493 
494 		skb_dst_set(skb, dst);
495 	}
496 
497 	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
498 					      IPPROTO_ICMPV6,
499 					      csum_partial(icmp6h,
500 							   skb->len, 0));
501 
502 	ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
503 
504 	rcu_read_lock();
505 	idev = __in6_dev_get(dst->dev);
506 	IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
507 
508 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
509 		      net, sk, skb, NULL, dst->dev,
510 		      dst_output);
511 	if (!err) {
512 		ICMP6MSGOUT_INC_STATS(net, idev, type);
513 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
514 	}
515 
516 	rcu_read_unlock();
517 }
518 EXPORT_SYMBOL(ndisc_send_skb);
519 
520 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
521 		   const struct in6_addr *solicited_addr,
522 		   bool router, bool solicited, bool override, bool inc_opt)
523 {
524 	struct sk_buff *skb;
525 	struct in6_addr tmpaddr;
526 	struct inet6_ifaddr *ifp;
527 	const struct in6_addr *src_addr;
528 	struct nd_msg *msg;
529 	int optlen = 0;
530 
531 	/* for anycast or proxy, solicited_addr != src_addr */
532 	ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
533 	if (ifp) {
534 		src_addr = solicited_addr;
535 		if (ifp->flags & IFA_F_OPTIMISTIC)
536 			override = false;
537 		inc_opt |= ifp->idev->cnf.force_tllao;
538 		in6_ifa_put(ifp);
539 	} else {
540 		if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
541 				       inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
542 				       &tmpaddr))
543 			return;
544 		src_addr = &tmpaddr;
545 	}
546 
547 	if (!dev->addr_len)
548 		inc_opt = false;
549 	if (inc_opt)
550 		optlen += ndisc_opt_addr_space(dev,
551 					       NDISC_NEIGHBOUR_ADVERTISEMENT);
552 
553 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
554 	if (!skb)
555 		return;
556 
557 	msg = skb_put(skb, sizeof(*msg));
558 	*msg = (struct nd_msg) {
559 		.icmph = {
560 			.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
561 			.icmp6_router = router,
562 			.icmp6_solicited = solicited,
563 			.icmp6_override = override,
564 		},
565 		.target = *solicited_addr,
566 	};
567 
568 	if (inc_opt)
569 		ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
570 				       dev->dev_addr,
571 				       NDISC_NEIGHBOUR_ADVERTISEMENT);
572 
573 	ndisc_send_skb(skb, daddr, src_addr);
574 }
575 
576 static void ndisc_send_unsol_na(struct net_device *dev)
577 {
578 	struct inet6_dev *idev;
579 	struct inet6_ifaddr *ifa;
580 
581 	idev = in6_dev_get(dev);
582 	if (!idev)
583 		return;
584 
585 	read_lock_bh(&idev->lock);
586 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
587 		/* skip tentative addresses until dad completes */
588 		if (ifa->flags & IFA_F_TENTATIVE &&
589 		    !(ifa->flags & IFA_F_OPTIMISTIC))
590 			continue;
591 
592 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
593 			      /*router=*/ !!idev->cnf.forwarding,
594 			      /*solicited=*/ false, /*override=*/ true,
595 			      /*inc_opt=*/ true);
596 	}
597 	read_unlock_bh(&idev->lock);
598 
599 	in6_dev_put(idev);
600 }
601 
602 struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
603 				const struct in6_addr *saddr, u64 nonce)
604 {
605 	int inc_opt = dev->addr_len;
606 	struct sk_buff *skb;
607 	struct nd_msg *msg;
608 	int optlen = 0;
609 
610 	if (!saddr)
611 		return NULL;
612 
613 	if (ipv6_addr_any(saddr))
614 		inc_opt = false;
615 	if (inc_opt)
616 		optlen += ndisc_opt_addr_space(dev,
617 					       NDISC_NEIGHBOUR_SOLICITATION);
618 	if (nonce != 0)
619 		optlen += 8;
620 
621 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
622 	if (!skb)
623 		return NULL;
624 
625 	msg = skb_put(skb, sizeof(*msg));
626 	*msg = (struct nd_msg) {
627 		.icmph = {
628 			.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
629 		},
630 		.target = *solicit,
631 	};
632 
633 	if (inc_opt)
634 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
635 				       dev->dev_addr,
636 				       NDISC_NEIGHBOUR_SOLICITATION);
637 	if (nonce != 0) {
638 		u8 *opt = skb_put(skb, 8);
639 
640 		opt[0] = ND_OPT_NONCE;
641 		opt[1] = 8 >> 3;
642 		memcpy(opt + 2, &nonce, 6);
643 	}
644 
645 	return skb;
646 }
647 EXPORT_SYMBOL(ndisc_ns_create);
648 
649 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
650 		   const struct in6_addr *daddr, const struct in6_addr *saddr,
651 		   u64 nonce)
652 {
653 	struct in6_addr addr_buf;
654 	struct sk_buff *skb;
655 
656 	if (!saddr) {
657 		if (ipv6_get_lladdr(dev, &addr_buf,
658 				    (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
659 			return;
660 		saddr = &addr_buf;
661 	}
662 
663 	skb = ndisc_ns_create(dev, solicit, saddr, nonce);
664 
665 	if (skb)
666 		ndisc_send_skb(skb, daddr, saddr);
667 }
668 
669 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
670 		   const struct in6_addr *daddr)
671 {
672 	struct sk_buff *skb;
673 	struct rs_msg *msg;
674 	int send_sllao = dev->addr_len;
675 	int optlen = 0;
676 
677 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
678 	/*
679 	 * According to section 2.2 of RFC 4429, we must not
680 	 * send router solicitations with a sllao from
681 	 * optimistic addresses, but we may send the solicitation
682 	 * if we don't include the sllao.  So here we check
683 	 * if our address is optimistic, and if so, we
684 	 * suppress the inclusion of the sllao.
685 	 */
686 	if (send_sllao) {
687 		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
688 							   dev, 1);
689 		if (ifp) {
690 			if (ifp->flags & IFA_F_OPTIMISTIC)  {
691 				send_sllao = 0;
692 			}
693 			in6_ifa_put(ifp);
694 		} else {
695 			send_sllao = 0;
696 		}
697 	}
698 #endif
699 	if (send_sllao)
700 		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
701 
702 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
703 	if (!skb)
704 		return;
705 
706 	msg = skb_put(skb, sizeof(*msg));
707 	*msg = (struct rs_msg) {
708 		.icmph = {
709 			.icmp6_type = NDISC_ROUTER_SOLICITATION,
710 		},
711 	};
712 
713 	if (send_sllao)
714 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
715 				       dev->dev_addr,
716 				       NDISC_ROUTER_SOLICITATION);
717 
718 	ndisc_send_skb(skb, daddr, saddr);
719 }
720 
721 
722 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
723 {
724 	/*
725 	 *	"The sender MUST return an ICMP
726 	 *	 destination unreachable"
727 	 */
728 	dst_link_failure(skb);
729 	kfree_skb(skb);
730 }
731 
732 /* Called with locked neigh: either read or both */
733 
734 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
735 {
736 	struct in6_addr *saddr = NULL;
737 	struct in6_addr mcaddr;
738 	struct net_device *dev = neigh->dev;
739 	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
740 	int probes = atomic_read(&neigh->probes);
741 
742 	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
743 					   dev, false, 1,
744 					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
745 		saddr = &ipv6_hdr(skb)->saddr;
746 	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
747 	if (probes < 0) {
748 		if (!(neigh->nud_state & NUD_VALID)) {
749 			ND_PRINTK(1, dbg,
750 				  "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
751 				  __func__, target);
752 		}
753 		ndisc_send_ns(dev, target, target, saddr, 0);
754 	} else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
755 		neigh_app_ns(neigh);
756 	} else {
757 		addrconf_addr_solict_mult(target, &mcaddr);
758 		ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
759 	}
760 }
761 
762 static int pndisc_is_router(const void *pkey,
763 			    struct net_device *dev)
764 {
765 	struct pneigh_entry *n;
766 	int ret = -1;
767 
768 	read_lock_bh(&nd_tbl.lock);
769 	n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
770 	if (n)
771 		ret = !!(n->flags & NTF_ROUTER);
772 	read_unlock_bh(&nd_tbl.lock);
773 
774 	return ret;
775 }
776 
777 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
778 		  const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
779 		  struct ndisc_options *ndopts)
780 {
781 	neigh_update(neigh, lladdr, new, flags, 0);
782 	/* report ndisc ops about neighbour update */
783 	ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
784 }
785 
786 static void ndisc_recv_ns(struct sk_buff *skb)
787 {
788 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
789 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
790 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
791 	u8 *lladdr = NULL;
792 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
793 				    offsetof(struct nd_msg, opt));
794 	struct ndisc_options ndopts;
795 	struct net_device *dev = skb->dev;
796 	struct inet6_ifaddr *ifp;
797 	struct inet6_dev *idev = NULL;
798 	struct neighbour *neigh;
799 	int dad = ipv6_addr_any(saddr);
800 	bool inc;
801 	int is_router = -1;
802 	u64 nonce = 0;
803 
804 	if (skb->len < sizeof(struct nd_msg)) {
805 		ND_PRINTK(2, warn, "NS: packet too short\n");
806 		return;
807 	}
808 
809 	if (ipv6_addr_is_multicast(&msg->target)) {
810 		ND_PRINTK(2, warn, "NS: multicast target address\n");
811 		return;
812 	}
813 
814 	/*
815 	 * RFC2461 7.1.1:
816 	 * DAD has to be destined for solicited node multicast address.
817 	 */
818 	if (dad && !ipv6_addr_is_solict_mult(daddr)) {
819 		ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
820 		return;
821 	}
822 
823 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
824 		ND_PRINTK(2, warn, "NS: invalid ND options\n");
825 		return;
826 	}
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;
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;
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;
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;
905 		}
906 
907 		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
908 		    (idev->cnf.forwarding &&
909 		     (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
910 		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
911 			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
912 			    skb->pkt_type != PACKET_HOST &&
913 			    inc &&
914 			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
915 				/*
916 				 * for anycast or proxy,
917 				 * sender should delay its response
918 				 * by a random time between 0 and
919 				 * MAX_ANYCAST_DELAY_TIME seconds.
920 				 * (RFC2461) -- yoshfuji
921 				 */
922 				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
923 				if (n)
924 					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
925 				goto out;
926 			}
927 		} else
928 			goto out;
929 	}
930 
931 	if (is_router < 0)
932 		is_router = idev->cnf.forwarding;
933 
934 	if (dad) {
935 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
936 			      !!is_router, false, (ifp != NULL), true);
937 		goto out;
938 	}
939 
940 	if (inc)
941 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
942 	else
943 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
944 
945 	/*
946 	 *	update / create cache entry
947 	 *	for the source address
948 	 */
949 	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
950 			       !inc || lladdr || !dev->addr_len);
951 	if (neigh)
952 		ndisc_update(dev, neigh, lladdr, NUD_STALE,
953 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
954 			     NEIGH_UPDATE_F_OVERRIDE,
955 			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
956 	if (neigh || !dev->header_ops) {
957 		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
958 			      true, (ifp != NULL && inc), inc);
959 		if (neigh)
960 			neigh_release(neigh);
961 	}
962 
963 out:
964 	if (ifp)
965 		in6_ifa_put(ifp);
966 	else
967 		in6_dev_put(idev);
968 }
969 
970 static void ndisc_recv_na(struct sk_buff *skb)
971 {
972 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
973 	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
974 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
975 	u8 *lladdr = NULL;
976 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
977 				    offsetof(struct nd_msg, opt));
978 	struct ndisc_options ndopts;
979 	struct net_device *dev = skb->dev;
980 	struct inet6_dev *idev = __in6_dev_get(dev);
981 	struct inet6_ifaddr *ifp;
982 	struct neighbour *neigh;
983 	u8 new_state;
984 
985 	if (skb->len < sizeof(struct nd_msg)) {
986 		ND_PRINTK(2, warn, "NA: packet too short\n");
987 		return;
988 	}
989 
990 	if (ipv6_addr_is_multicast(&msg->target)) {
991 		ND_PRINTK(2, warn, "NA: target address is multicast\n");
992 		return;
993 	}
994 
995 	if (ipv6_addr_is_multicast(daddr) &&
996 	    msg->icmph.icmp6_solicited) {
997 		ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
998 		return;
999 	}
1000 
1001 	/* For some 802.11 wireless deployments (and possibly other networks),
1002 	 * there will be a NA proxy and unsolicitd packets are attacks
1003 	 * and thus should not be accepted.
1004 	 * drop_unsolicited_na takes precedence over accept_untracked_na
1005 	 */
1006 	if (!msg->icmph.icmp6_solicited && idev &&
1007 	    idev->cnf.drop_unsolicited_na)
1008 		return;
1009 
1010 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
1011 		ND_PRINTK(2, warn, "NS: invalid ND option\n");
1012 		return;
1013 	}
1014 	if (ndopts.nd_opts_tgt_lladdr) {
1015 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1016 		if (!lladdr) {
1017 			ND_PRINTK(2, warn,
1018 				  "NA: invalid link-layer address length\n");
1019 			return;
1020 		}
1021 	}
1022 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1023 	if (ifp) {
1024 		if (skb->pkt_type != PACKET_LOOPBACK
1025 		    && (ifp->flags & IFA_F_TENTATIVE)) {
1026 				addrconf_dad_failure(skb, ifp);
1027 				return;
1028 		}
1029 		/* What should we make now? The advertisement
1030 		   is invalid, but ndisc specs say nothing
1031 		   about it. It could be misconfiguration, or
1032 		   an smart proxy agent tries to help us :-)
1033 
1034 		   We should not print the error if NA has been
1035 		   received from loopback - it is just our own
1036 		   unsolicited advertisement.
1037 		 */
1038 		if (skb->pkt_type != PACKET_LOOPBACK)
1039 			ND_PRINTK(1, warn,
1040 				  "NA: %pM advertised our address %pI6c on %s!\n",
1041 				  eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1042 		in6_ifa_put(ifp);
1043 		return;
1044 	}
1045 
1046 	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1047 
1048 	/* RFC 9131 updates original Neighbour Discovery RFC 4861.
1049 	 * NAs with Target LL Address option without a corresponding
1050 	 * entry in the neighbour cache can now create a STALE neighbour
1051 	 * cache entry on routers.
1052 	 *
1053 	 *   entry accept  fwding  solicited        behaviour
1054 	 * ------- ------  ------  ---------    ----------------------
1055 	 * present      X       X         0     Set state to STALE
1056 	 * present      X       X         1     Set state to REACHABLE
1057 	 *  absent      0       X         X     Do nothing
1058 	 *  absent      1       0         X     Do nothing
1059 	 *  absent      1       1         X     Add a new STALE entry
1060 	 *
1061 	 * Note that we don't do a (daddr == all-routers-mcast) check.
1062 	 */
1063 	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1064 	if (!neigh && lladdr &&
1065 	    idev && idev->cnf.forwarding &&
1066 	    idev->cnf.accept_untracked_na) {
1067 		neigh = neigh_create(&nd_tbl, &msg->target, dev);
1068 		new_state = NUD_STALE;
1069 	}
1070 
1071 	if (neigh && !IS_ERR(neigh)) {
1072 		u8 old_flags = neigh->flags;
1073 		struct net *net = dev_net(dev);
1074 
1075 		if (neigh->nud_state & NUD_FAILED)
1076 			goto out;
1077 
1078 		/*
1079 		 * Don't update the neighbor cache entry on a proxy NA from
1080 		 * ourselves because either the proxied node is off link or it
1081 		 * has already sent a NA to us.
1082 		 */
1083 		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1084 		    net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1085 		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1086 			/* XXX: idev->cnf.proxy_ndp */
1087 			goto out;
1088 		}
1089 
1090 		ndisc_update(dev, neigh, lladdr,
1091 			     new_state,
1092 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1093 			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1094 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1095 			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1096 			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1097 
1098 		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1099 			/*
1100 			 * Change: router to host
1101 			 */
1102 			rt6_clean_tohost(dev_net(dev),  saddr);
1103 		}
1104 
1105 out:
1106 		neigh_release(neigh);
1107 	}
1108 }
1109 
1110 static void ndisc_recv_rs(struct sk_buff *skb)
1111 {
1112 	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1113 	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1114 	struct neighbour *neigh;
1115 	struct inet6_dev *idev;
1116 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1117 	struct ndisc_options ndopts;
1118 	u8 *lladdr = NULL;
1119 
1120 	if (skb->len < sizeof(*rs_msg))
1121 		return;
1122 
1123 	idev = __in6_dev_get(skb->dev);
1124 	if (!idev) {
1125 		ND_PRINTK(1, err, "RS: can't find in6 device\n");
1126 		return;
1127 	}
1128 
1129 	/* Don't accept RS if we're not in router mode */
1130 	if (!idev->cnf.forwarding)
1131 		goto out;
1132 
1133 	/*
1134 	 * Don't update NCE if src = ::;
1135 	 * this implies that the source node has no ip address assigned yet.
1136 	 */
1137 	if (ipv6_addr_any(saddr))
1138 		goto out;
1139 
1140 	/* Parse ND options */
1141 	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1142 		ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1143 		goto out;
1144 	}
1145 
1146 	if (ndopts.nd_opts_src_lladdr) {
1147 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1148 					     skb->dev);
1149 		if (!lladdr)
1150 			goto out;
1151 	}
1152 
1153 	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1154 	if (neigh) {
1155 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1156 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1157 			     NEIGH_UPDATE_F_OVERRIDE|
1158 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1159 			     NDISC_ROUTER_SOLICITATION, &ndopts);
1160 		neigh_release(neigh);
1161 	}
1162 out:
1163 	return;
1164 }
1165 
1166 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1167 {
1168 	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1169 	struct sk_buff *skb;
1170 	struct nlmsghdr *nlh;
1171 	struct nduseroptmsg *ndmsg;
1172 	struct net *net = dev_net(ra->dev);
1173 	int err;
1174 	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1175 				    + (opt->nd_opt_len << 3));
1176 	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1177 
1178 	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1179 	if (!skb) {
1180 		err = -ENOBUFS;
1181 		goto errout;
1182 	}
1183 
1184 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1185 	if (!nlh) {
1186 		goto nla_put_failure;
1187 	}
1188 
1189 	ndmsg = nlmsg_data(nlh);
1190 	ndmsg->nduseropt_family = AF_INET6;
1191 	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1192 	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1193 	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1194 	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1195 
1196 	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1197 
1198 	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1199 		goto nla_put_failure;
1200 	nlmsg_end(skb, nlh);
1201 
1202 	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1203 	return;
1204 
1205 nla_put_failure:
1206 	nlmsg_free(skb);
1207 	err = -EMSGSIZE;
1208 errout:
1209 	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1210 }
1211 
1212 static void ndisc_router_discovery(struct sk_buff *skb)
1213 {
1214 	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1215 	struct neighbour *neigh = NULL;
1216 	struct inet6_dev *in6_dev;
1217 	struct fib6_info *rt = NULL;
1218 	u32 defrtr_usr_metric;
1219 	struct net *net;
1220 	int lifetime;
1221 	struct ndisc_options ndopts;
1222 	int optlen;
1223 	unsigned int pref = 0;
1224 	__u32 old_if_flags;
1225 	bool send_ifinfo_notify = false;
1226 
1227 	__u8 *opt = (__u8 *)(ra_msg + 1);
1228 
1229 	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1230 		sizeof(struct ra_msg);
1231 
1232 	ND_PRINTK(2, info,
1233 		  "RA: %s, dev: %s\n",
1234 		  __func__, skb->dev->name);
1235 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1236 		ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1237 		return;
1238 	}
1239 	if (optlen < 0) {
1240 		ND_PRINTK(2, warn, "RA: packet too short\n");
1241 		return;
1242 	}
1243 
1244 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1245 	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1246 		ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1247 		return;
1248 	}
1249 #endif
1250 
1251 	/*
1252 	 *	set the RA_RECV flag in the interface
1253 	 */
1254 
1255 	in6_dev = __in6_dev_get(skb->dev);
1256 	if (!in6_dev) {
1257 		ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1258 			  skb->dev->name);
1259 		return;
1260 	}
1261 
1262 	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1263 		ND_PRINTK(2, warn, "RA: invalid ND options\n");
1264 		return;
1265 	}
1266 
1267 	if (!ipv6_accept_ra(in6_dev)) {
1268 		ND_PRINTK(2, info,
1269 			  "RA: %s, did not accept ra for dev: %s\n",
1270 			  __func__, skb->dev->name);
1271 		goto skip_linkparms;
1272 	}
1273 
1274 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1275 	/* skip link-specific parameters from interior routers */
1276 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1277 		ND_PRINTK(2, info,
1278 			  "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1279 			  __func__, skb->dev->name);
1280 		goto skip_linkparms;
1281 	}
1282 #endif
1283 
1284 	if (in6_dev->if_flags & IF_RS_SENT) {
1285 		/*
1286 		 *	flag that an RA was received after an RS was sent
1287 		 *	out on this interface.
1288 		 */
1289 		in6_dev->if_flags |= IF_RA_RCVD;
1290 	}
1291 
1292 	/*
1293 	 * Remember the managed/otherconf flags from most recently
1294 	 * received RA message (RFC 2462) -- yoshfuji
1295 	 */
1296 	old_if_flags = in6_dev->if_flags;
1297 	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1298 				IF_RA_OTHERCONF)) |
1299 				(ra_msg->icmph.icmp6_addrconf_managed ?
1300 					IF_RA_MANAGED : 0) |
1301 				(ra_msg->icmph.icmp6_addrconf_other ?
1302 					IF_RA_OTHERCONF : 0);
1303 
1304 	if (old_if_flags != in6_dev->if_flags)
1305 		send_ifinfo_notify = true;
1306 
1307 	if (!in6_dev->cnf.accept_ra_defrtr) {
1308 		ND_PRINTK(2, info,
1309 			  "RA: %s, defrtr is false for dev: %s\n",
1310 			  __func__, skb->dev->name);
1311 		goto skip_defrtr;
1312 	}
1313 
1314 	/* Do not accept RA with source-addr found on local machine unless
1315 	 * accept_ra_from_local is set to true.
1316 	 */
1317 	net = dev_net(in6_dev->dev);
1318 	if (!in6_dev->cnf.accept_ra_from_local &&
1319 	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1320 		ND_PRINTK(2, info,
1321 			  "RA from local address detected on dev: %s: default router ignored\n",
1322 			  skb->dev->name);
1323 		goto skip_defrtr;
1324 	}
1325 
1326 	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1327 
1328 #ifdef CONFIG_IPV6_ROUTER_PREF
1329 	pref = ra_msg->icmph.icmp6_router_pref;
1330 	/* 10b is handled as if it were 00b (medium) */
1331 	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1332 	    !in6_dev->cnf.accept_ra_rtr_pref)
1333 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1334 #endif
1335 	/* routes added from RAs do not use nexthop objects */
1336 	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1337 	if (rt) {
1338 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1339 					 rt->fib6_nh->fib_nh_dev, NULL,
1340 					  &ipv6_hdr(skb)->saddr);
1341 		if (!neigh) {
1342 			ND_PRINTK(0, err,
1343 				  "RA: %s got default router without neighbour\n",
1344 				  __func__);
1345 			fib6_info_release(rt);
1346 			return;
1347 		}
1348 	}
1349 	/* Set default route metric as specified by user */
1350 	defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1351 	/* delete the route if lifetime is 0 or if metric needs change */
1352 	if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1353 		ip6_del_rt(net, rt, false);
1354 		rt = NULL;
1355 	}
1356 
1357 	ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, metric: %d, for dev: %s\n",
1358 		  rt, lifetime, defrtr_usr_metric, skb->dev->name);
1359 	if (!rt && lifetime) {
1360 		ND_PRINTK(3, info, "RA: adding default router\n");
1361 
1362 		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1363 					 skb->dev, pref, defrtr_usr_metric);
1364 		if (!rt) {
1365 			ND_PRINTK(0, err,
1366 				  "RA: %s failed to add default route\n",
1367 				  __func__);
1368 			return;
1369 		}
1370 
1371 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1372 					 rt->fib6_nh->fib_nh_dev, NULL,
1373 					  &ipv6_hdr(skb)->saddr);
1374 		if (!neigh) {
1375 			ND_PRINTK(0, err,
1376 				  "RA: %s got default router without neighbour\n",
1377 				  __func__);
1378 			fib6_info_release(rt);
1379 			return;
1380 		}
1381 		neigh->flags |= NTF_ROUTER;
1382 	} else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1383 		struct nl_info nlinfo = {
1384 			.nl_net = net,
1385 		};
1386 		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1387 		inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1388 	}
1389 
1390 	if (rt)
1391 		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1392 	if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1393 	    ra_msg->icmph.icmp6_hop_limit) {
1394 		if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1395 			in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1396 			fib6_metric_set(rt, RTAX_HOPLIMIT,
1397 					ra_msg->icmph.icmp6_hop_limit);
1398 		} else {
1399 			ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1400 		}
1401 	}
1402 
1403 skip_defrtr:
1404 
1405 	/*
1406 	 *	Update Reachable Time and Retrans Timer
1407 	 */
1408 
1409 	if (in6_dev->nd_parms) {
1410 		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1411 
1412 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1413 			rtime = (rtime*HZ)/1000;
1414 			if (rtime < HZ/100)
1415 				rtime = HZ/100;
1416 			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1417 			in6_dev->tstamp = jiffies;
1418 			send_ifinfo_notify = true;
1419 		}
1420 
1421 		rtime = ntohl(ra_msg->reachable_time);
1422 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1423 			rtime = (rtime*HZ)/1000;
1424 
1425 			if (rtime < HZ/10)
1426 				rtime = HZ/10;
1427 
1428 			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1429 				NEIGH_VAR_SET(in6_dev->nd_parms,
1430 					      BASE_REACHABLE_TIME, rtime);
1431 				NEIGH_VAR_SET(in6_dev->nd_parms,
1432 					      GC_STALETIME, 3 * rtime);
1433 				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1434 				in6_dev->tstamp = jiffies;
1435 				send_ifinfo_notify = true;
1436 			}
1437 		}
1438 	}
1439 
1440 skip_linkparms:
1441 
1442 	/*
1443 	 *	Process options.
1444 	 */
1445 
1446 	if (!neigh)
1447 		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1448 				       skb->dev, 1);
1449 	if (neigh) {
1450 		u8 *lladdr = NULL;
1451 		if (ndopts.nd_opts_src_lladdr) {
1452 			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1453 						     skb->dev);
1454 			if (!lladdr) {
1455 				ND_PRINTK(2, warn,
1456 					  "RA: invalid link-layer address length\n");
1457 				goto out;
1458 			}
1459 		}
1460 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1461 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1462 			     NEIGH_UPDATE_F_OVERRIDE|
1463 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1464 			     NEIGH_UPDATE_F_ISROUTER,
1465 			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1466 	}
1467 
1468 	if (!ipv6_accept_ra(in6_dev)) {
1469 		ND_PRINTK(2, info,
1470 			  "RA: %s, accept_ra is false for dev: %s\n",
1471 			  __func__, skb->dev->name);
1472 		goto out;
1473 	}
1474 
1475 #ifdef CONFIG_IPV6_ROUTE_INFO
1476 	if (!in6_dev->cnf.accept_ra_from_local &&
1477 	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1478 			  in6_dev->dev, 0)) {
1479 		ND_PRINTK(2, info,
1480 			  "RA from local address detected on dev: %s: router info ignored.\n",
1481 			  skb->dev->name);
1482 		goto skip_routeinfo;
1483 	}
1484 
1485 	if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1486 		struct nd_opt_hdr *p;
1487 		for (p = ndopts.nd_opts_ri;
1488 		     p;
1489 		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1490 			struct route_info *ri = (struct route_info *)p;
1491 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1492 			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1493 			    ri->prefix_len == 0)
1494 				continue;
1495 #endif
1496 			if (ri->prefix_len == 0 &&
1497 			    !in6_dev->cnf.accept_ra_defrtr)
1498 				continue;
1499 			if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1500 				continue;
1501 			if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1502 				continue;
1503 			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1504 				      &ipv6_hdr(skb)->saddr);
1505 		}
1506 	}
1507 
1508 skip_routeinfo:
1509 #endif
1510 
1511 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1512 	/* skip link-specific ndopts from interior routers */
1513 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1514 		ND_PRINTK(2, info,
1515 			  "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1516 			  __func__, skb->dev->name);
1517 		goto out;
1518 	}
1519 #endif
1520 
1521 	if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1522 		struct nd_opt_hdr *p;
1523 		for (p = ndopts.nd_opts_pi;
1524 		     p;
1525 		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1526 			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1527 					    (p->nd_opt_len) << 3,
1528 					    ndopts.nd_opts_src_lladdr != NULL);
1529 		}
1530 	}
1531 
1532 	if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1533 		__be32 n;
1534 		u32 mtu;
1535 
1536 		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1537 		mtu = ntohl(n);
1538 
1539 		if (in6_dev->ra_mtu != mtu) {
1540 			in6_dev->ra_mtu = mtu;
1541 			send_ifinfo_notify = true;
1542 		}
1543 
1544 		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1545 			ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1546 		} else if (in6_dev->cnf.mtu6 != mtu) {
1547 			in6_dev->cnf.mtu6 = mtu;
1548 			fib6_metric_set(rt, RTAX_MTU, mtu);
1549 			rt6_mtu_change(skb->dev, mtu);
1550 		}
1551 	}
1552 
1553 	if (ndopts.nd_useropts) {
1554 		struct nd_opt_hdr *p;
1555 		for (p = ndopts.nd_useropts;
1556 		     p;
1557 		     p = ndisc_next_useropt(skb->dev, p,
1558 					    ndopts.nd_useropts_end)) {
1559 			ndisc_ra_useropt(skb, p);
1560 		}
1561 	}
1562 
1563 	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1564 		ND_PRINTK(2, warn, "RA: invalid RA options\n");
1565 	}
1566 out:
1567 	/* Send a notify if RA changed managed/otherconf flags or
1568 	 * timer settings or ra_mtu value
1569 	 */
1570 	if (send_ifinfo_notify)
1571 		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1572 
1573 	fib6_info_release(rt);
1574 	if (neigh)
1575 		neigh_release(neigh);
1576 }
1577 
1578 static void ndisc_redirect_rcv(struct sk_buff *skb)
1579 {
1580 	u8 *hdr;
1581 	struct ndisc_options ndopts;
1582 	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1583 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1584 				    offsetof(struct rd_msg, opt));
1585 
1586 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1587 	switch (skb->ndisc_nodetype) {
1588 	case NDISC_NODETYPE_HOST:
1589 	case NDISC_NODETYPE_NODEFAULT:
1590 		ND_PRINTK(2, warn,
1591 			  "Redirect: from host or unauthorized router\n");
1592 		return;
1593 	}
1594 #endif
1595 
1596 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1597 		ND_PRINTK(2, warn,
1598 			  "Redirect: source address is not link-local\n");
1599 		return;
1600 	}
1601 
1602 	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1603 		return;
1604 
1605 	if (!ndopts.nd_opts_rh) {
1606 		ip6_redirect_no_header(skb, dev_net(skb->dev),
1607 					skb->dev->ifindex);
1608 		return;
1609 	}
1610 
1611 	hdr = (u8 *)ndopts.nd_opts_rh;
1612 	hdr += 8;
1613 	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1614 		return;
1615 
1616 	icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1617 }
1618 
1619 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1620 					   struct sk_buff *orig_skb,
1621 					   int rd_len)
1622 {
1623 	u8 *opt = skb_put(skb, rd_len);
1624 
1625 	memset(opt, 0, 8);
1626 	*(opt++) = ND_OPT_REDIRECT_HDR;
1627 	*(opt++) = (rd_len >> 3);
1628 	opt += 6;
1629 
1630 	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1631 		      rd_len - 8);
1632 }
1633 
1634 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1635 {
1636 	struct net_device *dev = skb->dev;
1637 	struct net *net = dev_net(dev);
1638 	struct sock *sk = net->ipv6.ndisc_sk;
1639 	int optlen = 0;
1640 	struct inet_peer *peer;
1641 	struct sk_buff *buff;
1642 	struct rd_msg *msg;
1643 	struct in6_addr saddr_buf;
1644 	struct rt6_info *rt;
1645 	struct dst_entry *dst;
1646 	struct flowi6 fl6;
1647 	int rd_len;
1648 	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1649 	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1650 	bool ret;
1651 
1652 	if (netif_is_l3_master(skb->dev)) {
1653 		dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1654 		if (!dev)
1655 			return;
1656 	}
1657 
1658 	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1659 		ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1660 			  dev->name);
1661 		return;
1662 	}
1663 
1664 	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1665 	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1666 		ND_PRINTK(2, warn,
1667 			  "Redirect: target address is not link-local unicast\n");
1668 		return;
1669 	}
1670 
1671 	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1672 			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1673 
1674 	dst = ip6_route_output(net, NULL, &fl6);
1675 	if (dst->error) {
1676 		dst_release(dst);
1677 		return;
1678 	}
1679 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1680 	if (IS_ERR(dst))
1681 		return;
1682 
1683 	rt = (struct rt6_info *) dst;
1684 
1685 	if (rt->rt6i_flags & RTF_GATEWAY) {
1686 		ND_PRINTK(2, warn,
1687 			  "Redirect: destination is not a neighbour\n");
1688 		goto release;
1689 	}
1690 	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1691 	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1692 	if (peer)
1693 		inet_putpeer(peer);
1694 	if (!ret)
1695 		goto release;
1696 
1697 	if (dev->addr_len) {
1698 		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1699 		if (!neigh) {
1700 			ND_PRINTK(2, warn,
1701 				  "Redirect: no neigh for target address\n");
1702 			goto release;
1703 		}
1704 
1705 		read_lock_bh(&neigh->lock);
1706 		if (neigh->nud_state & NUD_VALID) {
1707 			memcpy(ha_buf, neigh->ha, dev->addr_len);
1708 			read_unlock_bh(&neigh->lock);
1709 			ha = ha_buf;
1710 			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1711 								ops_data_buf,
1712 								&ops_data);
1713 		} else
1714 			read_unlock_bh(&neigh->lock);
1715 
1716 		neigh_release(neigh);
1717 	}
1718 
1719 	rd_len = min_t(unsigned int,
1720 		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1721 		       skb->len + 8);
1722 	rd_len &= ~0x7;
1723 	optlen += rd_len;
1724 
1725 	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1726 	if (!buff)
1727 		goto release;
1728 
1729 	msg = skb_put(buff, sizeof(*msg));
1730 	*msg = (struct rd_msg) {
1731 		.icmph = {
1732 			.icmp6_type = NDISC_REDIRECT,
1733 		},
1734 		.target = *target,
1735 		.dest = ipv6_hdr(skb)->daddr,
1736 	};
1737 
1738 	/*
1739 	 *	include target_address option
1740 	 */
1741 
1742 	if (ha)
1743 		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1744 
1745 	/*
1746 	 *	build redirect option and copy skb over to the new packet.
1747 	 */
1748 
1749 	if (rd_len)
1750 		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1751 
1752 	skb_dst_set(buff, dst);
1753 	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1754 	return;
1755 
1756 release:
1757 	dst_release(dst);
1758 }
1759 
1760 static void pndisc_redo(struct sk_buff *skb)
1761 {
1762 	ndisc_recv_ns(skb);
1763 	kfree_skb(skb);
1764 }
1765 
1766 static int ndisc_is_multicast(const void *pkey)
1767 {
1768 	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1769 }
1770 
1771 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1772 {
1773 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1774 
1775 	if (!idev)
1776 		return true;
1777 	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1778 	    idev->cnf.suppress_frag_ndisc) {
1779 		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1780 		return true;
1781 	}
1782 	return false;
1783 }
1784 
1785 int ndisc_rcv(struct sk_buff *skb)
1786 {
1787 	struct nd_msg *msg;
1788 
1789 	if (ndisc_suppress_frag_ndisc(skb))
1790 		return 0;
1791 
1792 	if (skb_linearize(skb))
1793 		return 0;
1794 
1795 	msg = (struct nd_msg *)skb_transport_header(skb);
1796 
1797 	__skb_push(skb, skb->data - skb_transport_header(skb));
1798 
1799 	if (ipv6_hdr(skb)->hop_limit != 255) {
1800 		ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1801 			  ipv6_hdr(skb)->hop_limit);
1802 		return 0;
1803 	}
1804 
1805 	if (msg->icmph.icmp6_code != 0) {
1806 		ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1807 			  msg->icmph.icmp6_code);
1808 		return 0;
1809 	}
1810 
1811 	switch (msg->icmph.icmp6_type) {
1812 	case NDISC_NEIGHBOUR_SOLICITATION:
1813 		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1814 		ndisc_recv_ns(skb);
1815 		break;
1816 
1817 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1818 		ndisc_recv_na(skb);
1819 		break;
1820 
1821 	case NDISC_ROUTER_SOLICITATION:
1822 		ndisc_recv_rs(skb);
1823 		break;
1824 
1825 	case NDISC_ROUTER_ADVERTISEMENT:
1826 		ndisc_router_discovery(skb);
1827 		break;
1828 
1829 	case NDISC_REDIRECT:
1830 		ndisc_redirect_rcv(skb);
1831 		break;
1832 	}
1833 
1834 	return 0;
1835 }
1836 
1837 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1838 {
1839 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1840 	struct netdev_notifier_change_info *change_info;
1841 	struct net *net = dev_net(dev);
1842 	struct inet6_dev *idev;
1843 	bool evict_nocarrier;
1844 
1845 	switch (event) {
1846 	case NETDEV_CHANGEADDR:
1847 		neigh_changeaddr(&nd_tbl, dev);
1848 		fib6_run_gc(0, net, false);
1849 		fallthrough;
1850 	case NETDEV_UP:
1851 		idev = in6_dev_get(dev);
1852 		if (!idev)
1853 			break;
1854 		if (idev->cnf.ndisc_notify ||
1855 		    net->ipv6.devconf_all->ndisc_notify)
1856 			ndisc_send_unsol_na(dev);
1857 		in6_dev_put(idev);
1858 		break;
1859 	case NETDEV_CHANGE:
1860 		idev = in6_dev_get(dev);
1861 		if (!idev)
1862 			evict_nocarrier = true;
1863 		else {
1864 			evict_nocarrier = idev->cnf.ndisc_evict_nocarrier &&
1865 					  net->ipv6.devconf_all->ndisc_evict_nocarrier;
1866 			in6_dev_put(idev);
1867 		}
1868 
1869 		change_info = ptr;
1870 		if (change_info->flags_changed & IFF_NOARP)
1871 			neigh_changeaddr(&nd_tbl, dev);
1872 		if (evict_nocarrier && !netif_carrier_ok(dev))
1873 			neigh_carrier_down(&nd_tbl, dev);
1874 		break;
1875 	case NETDEV_DOWN:
1876 		neigh_ifdown(&nd_tbl, dev);
1877 		fib6_run_gc(0, net, false);
1878 		break;
1879 	case NETDEV_NOTIFY_PEERS:
1880 		ndisc_send_unsol_na(dev);
1881 		break;
1882 	default:
1883 		break;
1884 	}
1885 
1886 	return NOTIFY_DONE;
1887 }
1888 
1889 static struct notifier_block ndisc_netdev_notifier = {
1890 	.notifier_call = ndisc_netdev_event,
1891 	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1892 };
1893 
1894 #ifdef CONFIG_SYSCTL
1895 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1896 					 const char *func, const char *dev_name)
1897 {
1898 	static char warncomm[TASK_COMM_LEN];
1899 	static int warned;
1900 	if (strcmp(warncomm, current->comm) && warned < 5) {
1901 		strcpy(warncomm, current->comm);
1902 		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1903 			warncomm, func,
1904 			dev_name, ctl->procname,
1905 			dev_name, ctl->procname);
1906 		warned++;
1907 	}
1908 }
1909 
1910 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1911 		size_t *lenp, loff_t *ppos)
1912 {
1913 	struct net_device *dev = ctl->extra1;
1914 	struct inet6_dev *idev;
1915 	int ret;
1916 
1917 	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1918 	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1919 		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1920 
1921 	if (strcmp(ctl->procname, "retrans_time") == 0)
1922 		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1923 
1924 	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1925 		ret = neigh_proc_dointvec_jiffies(ctl, write,
1926 						  buffer, lenp, ppos);
1927 
1928 	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1929 		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1930 		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1931 						     buffer, lenp, ppos);
1932 	else
1933 		ret = -1;
1934 
1935 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1936 		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1937 			idev->nd_parms->reachable_time =
1938 					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1939 		idev->tstamp = jiffies;
1940 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1941 		in6_dev_put(idev);
1942 	}
1943 	return ret;
1944 }
1945 
1946 
1947 #endif
1948 
1949 static int __net_init ndisc_net_init(struct net *net)
1950 {
1951 	struct ipv6_pinfo *np;
1952 	struct sock *sk;
1953 	int err;
1954 
1955 	err = inet_ctl_sock_create(&sk, PF_INET6,
1956 				   SOCK_RAW, IPPROTO_ICMPV6, net);
1957 	if (err < 0) {
1958 		ND_PRINTK(0, err,
1959 			  "NDISC: Failed to initialize the control socket (err %d)\n",
1960 			  err);
1961 		return err;
1962 	}
1963 
1964 	net->ipv6.ndisc_sk = sk;
1965 
1966 	np = inet6_sk(sk);
1967 	np->hop_limit = 255;
1968 	/* Do not loopback ndisc messages */
1969 	np->mc_loop = 0;
1970 
1971 	return 0;
1972 }
1973 
1974 static void __net_exit ndisc_net_exit(struct net *net)
1975 {
1976 	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1977 }
1978 
1979 static struct pernet_operations ndisc_net_ops = {
1980 	.init = ndisc_net_init,
1981 	.exit = ndisc_net_exit,
1982 };
1983 
1984 int __init ndisc_init(void)
1985 {
1986 	int err;
1987 
1988 	err = register_pernet_subsys(&ndisc_net_ops);
1989 	if (err)
1990 		return err;
1991 	/*
1992 	 * Initialize the neighbour table
1993 	 */
1994 	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1995 
1996 #ifdef CONFIG_SYSCTL
1997 	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1998 				    ndisc_ifinfo_sysctl_change);
1999 	if (err)
2000 		goto out_unregister_pernet;
2001 out:
2002 #endif
2003 	return err;
2004 
2005 #ifdef CONFIG_SYSCTL
2006 out_unregister_pernet:
2007 	unregister_pernet_subsys(&ndisc_net_ops);
2008 	goto out;
2009 #endif
2010 }
2011 
2012 int __init ndisc_late_init(void)
2013 {
2014 	return register_netdevice_notifier(&ndisc_netdev_notifier);
2015 }
2016 
2017 void ndisc_late_cleanup(void)
2018 {
2019 	unregister_netdevice_notifier(&ndisc_netdev_notifier);
2020 }
2021 
2022 void ndisc_cleanup(void)
2023 {
2024 #ifdef CONFIG_SYSCTL
2025 	neigh_sysctl_unregister(&nd_tbl.parms);
2026 #endif
2027 	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2028 	unregister_pernet_subsys(&ndisc_net_ops);
2029 }
2030