xref: /linux/net/ipv6/ndisc.c (revision 26ba30221c03364d6ed9910be8da4c1fd871b07b)
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_DEFAULT,
134 			[NEIGH_VAR_PROXY_QLEN] = 64,
135 			[NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
136 			[NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
137 		},
138 	},
139 	.gc_interval =	  30 * HZ,
140 	.gc_thresh1 =	 128,
141 	.gc_thresh2 =	 512,
142 	.gc_thresh3 =	1024,
143 };
144 EXPORT_SYMBOL_GPL(nd_tbl);
145 
146 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, const void *data,
147 			      int data_len, int pad)
148 {
149 	int space = __ndisc_opt_addr_space(data_len, pad);
150 	u8 *opt = skb_put(skb, space);
151 
152 	opt[0] = type;
153 	opt[1] = space>>3;
154 
155 	memset(opt + 2, 0, pad);
156 	opt   += pad;
157 	space -= pad;
158 
159 	memcpy(opt+2, data, data_len);
160 	data_len += 2;
161 	opt += data_len;
162 	space -= data_len;
163 	if (space > 0)
164 		memset(opt, 0, space);
165 }
166 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
167 
168 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
169 					  const void *data, u8 icmp6_type)
170 {
171 	__ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
172 				 ndisc_addr_option_pad(skb->dev->type));
173 	ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
174 }
175 
176 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
177 						   void *ha,
178 						   const u8 *ops_data)
179 {
180 	ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
181 	ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
182 }
183 
184 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
185 					    struct nd_opt_hdr *end)
186 {
187 	int type;
188 	if (!cur || !end || cur >= end)
189 		return NULL;
190 	type = cur->nd_opt_type;
191 	do {
192 		cur = ((void *)cur) + (cur->nd_opt_len << 3);
193 	} while (cur < end && cur->nd_opt_type != type);
194 	return cur <= end && cur->nd_opt_type == type ? cur : NULL;
195 }
196 
197 static inline int ndisc_is_useropt(const struct net_device *dev,
198 				   struct nd_opt_hdr *opt)
199 {
200 	return opt->nd_opt_type == ND_OPT_PREFIX_INFO ||
201 		opt->nd_opt_type == ND_OPT_RDNSS ||
202 		opt->nd_opt_type == ND_OPT_DNSSL ||
203 		opt->nd_opt_type == ND_OPT_6CO ||
204 		opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
205 		opt->nd_opt_type == ND_OPT_PREF64;
206 }
207 
208 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
209 					     struct nd_opt_hdr *cur,
210 					     struct nd_opt_hdr *end)
211 {
212 	if (!cur || !end || cur >= end)
213 		return NULL;
214 	do {
215 		cur = ((void *)cur) + (cur->nd_opt_len << 3);
216 	} while (cur < end && !ndisc_is_useropt(dev, cur));
217 	return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
218 }
219 
220 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
221 					  u8 *opt, int opt_len,
222 					  struct ndisc_options *ndopts)
223 {
224 	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
225 
226 	if (!nd_opt || opt_len < 0 || !ndopts)
227 		return NULL;
228 	memset(ndopts, 0, sizeof(*ndopts));
229 	while (opt_len) {
230 		bool unknown = false;
231 		int l;
232 		if (opt_len < sizeof(struct nd_opt_hdr))
233 			return NULL;
234 		l = nd_opt->nd_opt_len << 3;
235 		if (opt_len < l || l == 0)
236 			return NULL;
237 		if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
238 			goto next_opt;
239 		switch (nd_opt->nd_opt_type) {
240 		case ND_OPT_SOURCE_LL_ADDR:
241 		case ND_OPT_TARGET_LL_ADDR:
242 		case ND_OPT_MTU:
243 		case ND_OPT_NONCE:
244 		case ND_OPT_REDIRECT_HDR:
245 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
246 				net_dbg_ratelimited("%s: duplicated ND6 option found: type=%d\n",
247 						    __func__, nd_opt->nd_opt_type);
248 			} else {
249 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
250 			}
251 			break;
252 		case ND_OPT_PREFIX_INFO:
253 			ndopts->nd_opts_pi_end = nd_opt;
254 			if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
255 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
256 			break;
257 #ifdef CONFIG_IPV6_ROUTE_INFO
258 		case ND_OPT_ROUTE_INFO:
259 			ndopts->nd_opts_ri_end = nd_opt;
260 			if (!ndopts->nd_opts_ri)
261 				ndopts->nd_opts_ri = nd_opt;
262 			break;
263 #endif
264 		default:
265 			unknown = true;
266 		}
267 		if (ndisc_is_useropt(dev, nd_opt)) {
268 			ndopts->nd_useropts_end = nd_opt;
269 			if (!ndopts->nd_useropts)
270 				ndopts->nd_useropts = nd_opt;
271 		} else if (unknown) {
272 			/*
273 			 * Unknown options must be silently ignored,
274 			 * to accommodate future extension to the
275 			 * protocol.
276 			 */
277 			net_dbg_ratelimited("%s: ignored unsupported option; type=%d, len=%d\n",
278 					    __func__, nd_opt->nd_opt_type, nd_opt->nd_opt_len);
279 		}
280 next_opt:
281 		opt_len -= l;
282 		nd_opt = ((void *)nd_opt) + l;
283 	}
284 	return ndopts;
285 }
286 EXPORT_SYMBOL_GPL(ndisc_parse_options);
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 net_device *dev = n->dev;
378 	struct in6_addr maddr;
379 
380 	if (!dev)
381 		return -EINVAL;
382 
383 	addrconf_addr_solict_mult(addr, &maddr);
384 	return ipv6_dev_mc_inc(dev, &maddr);
385 }
386 
387 static void pndisc_destructor(struct pneigh_entry *n)
388 {
389 	struct in6_addr *addr = (struct in6_addr *)&n->key;
390 	struct net_device *dev = n->dev;
391 	struct in6_addr maddr;
392 
393 	if (!dev)
394 		return;
395 
396 	addrconf_addr_solict_mult(addr, &maddr);
397 	ipv6_dev_mc_dec(dev, &maddr);
398 }
399 
400 /* called with rtnl held */
401 static bool ndisc_allow_add(const struct net_device *dev,
402 			    struct netlink_ext_ack *extack)
403 {
404 	struct inet6_dev *idev = __in6_dev_get(dev);
405 
406 	if (!idev || idev->cnf.disable_ipv6) {
407 		NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
408 		return false;
409 	}
410 
411 	return true;
412 }
413 
414 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
415 				       int len)
416 {
417 	int hlen = LL_RESERVED_SPACE(dev);
418 	int tlen = dev->needed_tailroom;
419 	struct sk_buff *skb;
420 
421 	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
422 	if (!skb)
423 		return NULL;
424 
425 	skb->protocol = htons(ETH_P_IPV6);
426 	skb->dev = dev;
427 
428 	skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
429 	skb_reset_transport_header(skb);
430 
431 	/* Manually assign socket ownership as we avoid calling
432 	 * sock_alloc_send_pskb() to bypass wmem buffer limits
433 	 */
434 	rcu_read_lock();
435 	skb_set_owner_w(skb, dev_net_rcu(dev)->ipv6.ndisc_sk);
436 	rcu_read_unlock();
437 
438 	return skb;
439 }
440 
441 static void ip6_nd_hdr(struct sk_buff *skb,
442 		       const struct in6_addr *saddr,
443 		       const struct in6_addr *daddr,
444 		       int hop_limit, int len)
445 {
446 	struct ipv6hdr *hdr;
447 	struct inet6_dev *idev;
448 	unsigned tclass;
449 
450 	rcu_read_lock();
451 	idev = __in6_dev_get(skb->dev);
452 	tclass = idev ? READ_ONCE(idev->cnf.ndisc_tclass) : 0;
453 	rcu_read_unlock();
454 
455 	skb_push(skb, sizeof(*hdr));
456 	skb_reset_network_header(skb);
457 	hdr = ipv6_hdr(skb);
458 
459 	ip6_flow_hdr(hdr, tclass, 0);
460 
461 	hdr->payload_len = htons(len);
462 	hdr->nexthdr = IPPROTO_ICMPV6;
463 	hdr->hop_limit = hop_limit;
464 
465 	hdr->saddr = *saddr;
466 	hdr->daddr = *daddr;
467 }
468 
469 void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
470 		    const struct in6_addr *saddr)
471 {
472 	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
473 	struct dst_entry *dst = skb_dst(skb);
474 	struct net_device *dev;
475 	struct inet6_dev *idev;
476 	struct net *net;
477 	struct sock *sk;
478 	int err;
479 	u8 type;
480 
481 	type = icmp6h->icmp6_type;
482 
483 	rcu_read_lock();
484 
485 	net = dev_net_rcu(skb->dev);
486 	sk = net->ipv6.ndisc_sk;
487 	if (!dst) {
488 		struct flowi6 fl6;
489 		int oif = skb->dev->ifindex;
490 
491 		icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
492 		dst = icmp6_dst_alloc(skb->dev, &fl6);
493 		if (IS_ERR(dst)) {
494 			rcu_read_unlock();
495 			kfree_skb(skb);
496 			return;
497 		}
498 
499 		skb_dst_set(skb, dst);
500 	}
501 
502 	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
503 					      IPPROTO_ICMPV6,
504 					      csum_partial(icmp6h,
505 							   skb->len, 0));
506 
507 	ip6_nd_hdr(skb, saddr, daddr, READ_ONCE(inet6_sk(sk)->hop_limit), skb->len);
508 
509 	dev = dst_dev_rcu(dst);
510 	idev = __in6_dev_get(dev);
511 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
512 
513 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
514 		      net, sk, skb, NULL, dev,
515 		      dst_output);
516 	if (!err) {
517 		ICMP6MSGOUT_INC_STATS(net, idev, type);
518 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
519 	}
520 
521 	rcu_read_unlock();
522 }
523 EXPORT_SYMBOL(ndisc_send_skb);
524 
525 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
526 		   const struct in6_addr *solicited_addr,
527 		   bool router, bool solicited, bool override, bool inc_opt)
528 {
529 	struct sk_buff *skb;
530 	struct in6_addr tmpaddr;
531 	struct inet6_ifaddr *ifp;
532 	const struct in6_addr *src_addr;
533 	struct nd_msg *msg;
534 	int optlen = 0;
535 
536 	/* for anycast or proxy, solicited_addr != src_addr */
537 	ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
538 	if (ifp) {
539 		src_addr = solicited_addr;
540 		if (ifp->flags & IFA_F_OPTIMISTIC)
541 			override = false;
542 		inc_opt |= READ_ONCE(ifp->idev->cnf.force_tllao);
543 		in6_ifa_put(ifp);
544 	} else {
545 		if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
546 				       inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
547 				       &tmpaddr))
548 			return;
549 		src_addr = &tmpaddr;
550 	}
551 
552 	if (!dev->addr_len)
553 		inc_opt = false;
554 	if (inc_opt)
555 		optlen += ndisc_opt_addr_space(dev,
556 					       NDISC_NEIGHBOUR_ADVERTISEMENT);
557 
558 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
559 	if (!skb)
560 		return;
561 
562 	msg = skb_put(skb, sizeof(*msg));
563 	*msg = (struct nd_msg) {
564 		.icmph = {
565 			.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
566 			.icmp6_router = router,
567 			.icmp6_solicited = solicited,
568 			.icmp6_override = override,
569 		},
570 		.target = *solicited_addr,
571 	};
572 
573 	if (inc_opt)
574 		ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
575 				       dev->dev_addr,
576 				       NDISC_NEIGHBOUR_ADVERTISEMENT);
577 
578 	ndisc_send_skb(skb, daddr, src_addr);
579 }
580 EXPORT_SYMBOL_GPL(ndisc_send_na);
581 
582 static void ndisc_send_unsol_na(struct net_device *dev)
583 {
584 	struct inet6_dev *idev;
585 	struct inet6_ifaddr *ifa;
586 
587 	idev = in6_dev_get(dev);
588 	if (!idev)
589 		return;
590 
591 	read_lock_bh(&idev->lock);
592 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
593 		/* skip tentative addresses until dad completes */
594 		if (ifa->flags & IFA_F_TENTATIVE &&
595 		    !(ifa->flags & IFA_F_OPTIMISTIC))
596 			continue;
597 
598 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
599 			      /*router=*/ !!idev->cnf.forwarding,
600 			      /*solicited=*/ false, /*override=*/ true,
601 			      /*inc_opt=*/ true);
602 	}
603 	read_unlock_bh(&idev->lock);
604 
605 	in6_dev_put(idev);
606 }
607 
608 struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
609 				const struct in6_addr *saddr, u64 nonce)
610 {
611 	int inc_opt = dev->addr_len;
612 	struct sk_buff *skb;
613 	struct nd_msg *msg;
614 	int optlen = 0;
615 
616 	if (!saddr)
617 		return NULL;
618 
619 	if (ipv6_addr_any(saddr))
620 		inc_opt = false;
621 	if (inc_opt)
622 		optlen += ndisc_opt_addr_space(dev,
623 					       NDISC_NEIGHBOUR_SOLICITATION);
624 	if (nonce != 0)
625 		optlen += 8;
626 
627 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
628 	if (!skb)
629 		return NULL;
630 
631 	msg = skb_put(skb, sizeof(*msg));
632 	*msg = (struct nd_msg) {
633 		.icmph = {
634 			.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
635 		},
636 		.target = *solicit,
637 	};
638 
639 	if (inc_opt)
640 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
641 				       dev->dev_addr,
642 				       NDISC_NEIGHBOUR_SOLICITATION);
643 	if (nonce != 0) {
644 		u8 *opt = skb_put(skb, 8);
645 
646 		opt[0] = ND_OPT_NONCE;
647 		opt[1] = 8 >> 3;
648 		memcpy(opt + 2, &nonce, 6);
649 	}
650 
651 	return skb;
652 }
653 EXPORT_SYMBOL(ndisc_ns_create);
654 
655 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
656 		   const struct in6_addr *daddr, const struct in6_addr *saddr,
657 		   u64 nonce)
658 {
659 	struct in6_addr addr_buf;
660 	struct sk_buff *skb;
661 
662 	if (!saddr) {
663 		if (ipv6_get_lladdr(dev, &addr_buf,
664 				    (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
665 			return;
666 		saddr = &addr_buf;
667 	}
668 
669 	skb = ndisc_ns_create(dev, solicit, saddr, nonce);
670 
671 	if (skb)
672 		ndisc_send_skb(skb, daddr, saddr);
673 }
674 
675 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
676 		   const struct in6_addr *daddr)
677 {
678 	struct sk_buff *skb;
679 	struct rs_msg *msg;
680 	int send_sllao = dev->addr_len;
681 	int optlen = 0;
682 
683 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
684 	/*
685 	 * According to section 2.2 of RFC 4429, we must not
686 	 * send router solicitations with a sllao from
687 	 * optimistic addresses, but we may send the solicitation
688 	 * if we don't include the sllao.  So here we check
689 	 * if our address is optimistic, and if so, we
690 	 * suppress the inclusion of the sllao.
691 	 */
692 	if (send_sllao) {
693 		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
694 							   dev, 1);
695 		if (ifp) {
696 			if (ifp->flags & IFA_F_OPTIMISTIC)  {
697 				send_sllao = 0;
698 			}
699 			in6_ifa_put(ifp);
700 		} else {
701 			send_sllao = 0;
702 		}
703 	}
704 #endif
705 	if (send_sllao)
706 		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
707 
708 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
709 	if (!skb)
710 		return;
711 
712 	msg = skb_put(skb, sizeof(*msg));
713 	*msg = (struct rs_msg) {
714 		.icmph = {
715 			.icmp6_type = NDISC_ROUTER_SOLICITATION,
716 		},
717 	};
718 
719 	if (send_sllao)
720 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
721 				       dev->dev_addr,
722 				       NDISC_ROUTER_SOLICITATION);
723 
724 	ndisc_send_skb(skb, daddr, saddr);
725 }
726 
727 
728 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
729 {
730 	/*
731 	 *	"The sender MUST return an ICMP
732 	 *	 destination unreachable"
733 	 */
734 	dst_link_failure(skb);
735 	kfree_skb(skb);
736 }
737 
738 /* Called with locked neigh: either read or both */
739 
740 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
741 {
742 	struct in6_addr *saddr = NULL;
743 	struct in6_addr mcaddr;
744 	struct net_device *dev = neigh->dev;
745 	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
746 	int probes = atomic_read(&neigh->probes);
747 
748 	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
749 					   dev, false, 1,
750 					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
751 		saddr = &ipv6_hdr(skb)->saddr;
752 	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
753 	if (probes < 0) {
754 		if (!(READ_ONCE(neigh->nud_state) & NUD_VALID)) {
755 			net_dbg_ratelimited("%s: trying to ucast probe in NUD_INVALID: %pI6\n",
756 					    __func__, target);
757 		}
758 		ndisc_send_ns(dev, target, target, saddr, 0);
759 	} else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
760 		neigh_app_ns(neigh);
761 	} else {
762 		addrconf_addr_solict_mult(target, &mcaddr);
763 		ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
764 	}
765 }
766 
767 static int pndisc_is_router(const void *pkey,
768 			    struct net_device *dev)
769 {
770 	struct pneigh_entry *n;
771 	int ret = -1;
772 
773 	n = pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
774 	if (n)
775 		ret = !!(READ_ONCE(n->flags) & NTF_ROUTER);
776 
777 	return ret;
778 }
779 
780 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
781 		  const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
782 		  struct ndisc_options *ndopts)
783 {
784 	neigh_update(neigh, lladdr, new, flags, 0);
785 	/* report ndisc ops about neighbour update */
786 	ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
787 }
788 
789 static enum skb_drop_reason ndisc_recv_ns(struct sk_buff *skb)
790 {
791 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
792 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
793 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
794 	u8 *lladdr = NULL;
795 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
796 				    offsetof(struct nd_msg, opt));
797 	struct ndisc_options ndopts;
798 	struct net_device *dev = skb->dev;
799 	struct inet6_ifaddr *ifp;
800 	struct inet6_dev *idev = NULL;
801 	struct neighbour *neigh;
802 	int dad = ipv6_addr_any(saddr);
803 	int is_router = -1;
804 	SKB_DR(reason);
805 	u64 nonce = 0;
806 	bool inc;
807 
808 	if (skb->len < sizeof(struct nd_msg))
809 		return SKB_DROP_REASON_PKT_TOO_SMALL;
810 
811 	if (ipv6_addr_is_multicast(&msg->target)) {
812 		net_dbg_ratelimited("NS: multicast target address\n");
813 		return reason;
814 	}
815 
816 	/*
817 	 * RFC2461 7.1.1:
818 	 * DAD has to be destined for solicited node multicast address.
819 	 */
820 	if (dad && !ipv6_addr_is_solict_mult(daddr)) {
821 		net_dbg_ratelimited("NS: bad DAD packet (wrong destination)\n");
822 		return reason;
823 	}
824 
825 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
826 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
827 
828 	if (ndopts.nd_opts_src_lladdr) {
829 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
830 		if (!lladdr) {
831 			net_dbg_ratelimited("NS: invalid link-layer address length\n");
832 			return reason;
833 		}
834 
835 		/* RFC2461 7.1.1:
836 		 *	If the IP source address is the unspecified address,
837 		 *	there MUST NOT be source link-layer address option
838 		 *	in the message.
839 		 */
840 		if (dad) {
841 			net_dbg_ratelimited("NS: bad DAD packet (link-layer address option)\n");
842 			return reason;
843 		}
844 	}
845 	if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
846 		memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
847 
848 	inc = ipv6_addr_is_multicast(daddr);
849 
850 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
851 	if (ifp) {
852 have_ifp:
853 		if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
854 			if (dad) {
855 				if (nonce != 0 && ifp->dad_nonce == nonce) {
856 					u8 *np = (u8 *)&nonce;
857 					/* Matching nonce if looped back */
858 					net_dbg_ratelimited("%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
859 							    ifp->idev->dev->name, &ifp->addr, np);
860 					goto out;
861 				}
862 				/*
863 				 * We are colliding with another node
864 				 * who is doing DAD
865 				 * so fail our DAD process
866 				 */
867 				addrconf_dad_failure(skb, ifp);
868 				return reason;
869 			} else {
870 				/*
871 				 * This is not a dad solicitation.
872 				 * If we are an optimistic node,
873 				 * we should respond.
874 				 * Otherwise, we should ignore it.
875 				 */
876 				if (!(ifp->flags & IFA_F_OPTIMISTIC))
877 					goto out;
878 			}
879 		}
880 
881 		idev = ifp->idev;
882 	} else {
883 		struct net *net = dev_net(dev);
884 
885 		/* perhaps an address on the master device */
886 		if (netif_is_l3_slave(dev)) {
887 			struct net_device *mdev;
888 
889 			mdev = netdev_master_upper_dev_get_rcu(dev);
890 			if (mdev) {
891 				ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
892 				if (ifp)
893 					goto have_ifp;
894 			}
895 		}
896 
897 		idev = in6_dev_get(dev);
898 		if (!idev) {
899 			/* XXX: count this drop? */
900 			return reason;
901 		}
902 
903 		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
904 		    (READ_ONCE(idev->cnf.forwarding) &&
905 		     (READ_ONCE(net->ipv6.devconf_all->proxy_ndp) ||
906 		      READ_ONCE(idev->cnf.proxy_ndp)) &&
907 		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
908 			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
909 			    skb->pkt_type != PACKET_HOST &&
910 			    inc &&
911 			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
912 				/*
913 				 * for anycast or proxy,
914 				 * sender should delay its response
915 				 * by a random time between 0 and
916 				 * MAX_ANYCAST_DELAY_TIME seconds.
917 				 * (RFC2461) -- yoshfuji
918 				 */
919 				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
920 				if (n)
921 					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
922 				goto out;
923 			}
924 		} else {
925 			SKB_DR_SET(reason, IPV6_NDISC_NS_OTHERHOST);
926 			goto out;
927 		}
928 	}
929 
930 	if (is_router < 0)
931 		is_router = READ_ONCE(idev->cnf.forwarding);
932 
933 	if (dad) {
934 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
935 			      !!is_router, false, (ifp != NULL), true);
936 		goto out;
937 	}
938 
939 	if (inc)
940 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
941 	else
942 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
943 
944 	/*
945 	 *	update / create cache entry
946 	 *	for the source address
947 	 */
948 	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
949 			       !inc || lladdr || !dev->addr_len);
950 	if (neigh)
951 		ndisc_update(dev, neigh, lladdr, NUD_STALE,
952 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
953 			     NEIGH_UPDATE_F_OVERRIDE,
954 			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
955 	if (neigh || !dev->header_ops) {
956 		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
957 			      true, (ifp != NULL && inc), inc);
958 		if (neigh)
959 			neigh_release(neigh);
960 		reason = SKB_CONSUMED;
961 	}
962 
963 out:
964 	if (ifp)
965 		in6_ifa_put(ifp);
966 	else
967 		in6_dev_put(idev);
968 	return reason;
969 }
970 
971 static int accept_untracked_na(struct inet6_dev *idev, struct in6_addr *saddr)
972 {
973 	switch (READ_ONCE(idev->cnf.accept_untracked_na)) {
974 	case 0: /* Don't accept untracked na (absent in neighbor cache) */
975 		return 0;
976 	case 1: /* Create new entries from na if currently untracked */
977 		return 1;
978 	case 2: /* Create new entries from untracked na only if saddr is in the
979 		 * same subnet as an address configured on the interface that
980 		 * received the na
981 		 */
982 		return !!ipv6_chk_prefix(saddr, idev->dev);
983 	default:
984 		return 0;
985 	}
986 }
987 
988 static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)
989 {
990 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
991 	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
992 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
993 	u8 *lladdr = NULL;
994 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
995 				    offsetof(struct nd_msg, opt));
996 	struct ndisc_options ndopts;
997 	struct net_device *dev = skb->dev;
998 	struct inet6_dev *idev = __in6_dev_get(dev);
999 	struct inet6_ifaddr *ifp;
1000 	struct neighbour *neigh;
1001 	SKB_DR(reason);
1002 	u8 new_state;
1003 
1004 	if (skb->len < sizeof(struct nd_msg))
1005 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1006 
1007 	if (ipv6_addr_is_multicast(&msg->target)) {
1008 		net_dbg_ratelimited("NA: target address is multicast\n");
1009 		return reason;
1010 	}
1011 
1012 	if (ipv6_addr_is_multicast(daddr) &&
1013 	    msg->icmph.icmp6_solicited) {
1014 		net_dbg_ratelimited("NA: solicited NA is multicasted\n");
1015 		return reason;
1016 	}
1017 
1018 	/* For some 802.11 wireless deployments (and possibly other networks),
1019 	 * there will be a NA proxy and unsolicitd packets are attacks
1020 	 * and thus should not be accepted.
1021 	 * drop_unsolicited_na takes precedence over accept_untracked_na
1022 	 */
1023 	if (!msg->icmph.icmp6_solicited && idev &&
1024 	    READ_ONCE(idev->cnf.drop_unsolicited_na))
1025 		return reason;
1026 
1027 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
1028 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1029 
1030 	if (ndopts.nd_opts_tgt_lladdr) {
1031 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1032 		if (!lladdr) {
1033 			net_dbg_ratelimited("NA: invalid link-layer address length\n");
1034 			return reason;
1035 		}
1036 	}
1037 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1038 	if (ifp) {
1039 		if (skb->pkt_type != PACKET_LOOPBACK
1040 		    && (ifp->flags & IFA_F_TENTATIVE)) {
1041 				addrconf_dad_failure(skb, ifp);
1042 				return reason;
1043 		}
1044 		/* What should we make now? The advertisement
1045 		   is invalid, but ndisc specs say nothing
1046 		   about it. It could be misconfiguration, or
1047 		   an smart proxy agent tries to help us :-)
1048 
1049 		   We should not print the error if NA has been
1050 		   received from loopback - it is just our own
1051 		   unsolicited advertisement.
1052 		 */
1053 		if (skb->pkt_type != PACKET_LOOPBACK)
1054 			net_warn_ratelimited("NA: %pM advertised our address %pI6c on %s!\n",
1055 					     eth_hdr(skb)->h_source, &ifp->addr,
1056 					     ifp->idev->dev->name);
1057 		in6_ifa_put(ifp);
1058 		return reason;
1059 	}
1060 
1061 	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1062 
1063 	/* RFC 9131 updates original Neighbour Discovery RFC 4861.
1064 	 * NAs with Target LL Address option without a corresponding
1065 	 * entry in the neighbour cache can now create a STALE neighbour
1066 	 * cache entry on routers.
1067 	 *
1068 	 *   entry accept  fwding  solicited        behaviour
1069 	 * ------- ------  ------  ---------    ----------------------
1070 	 * present      X       X         0     Set state to STALE
1071 	 * present      X       X         1     Set state to REACHABLE
1072 	 *  absent      0       X         X     Do nothing
1073 	 *  absent      1       0         X     Do nothing
1074 	 *  absent      1       1         X     Add a new STALE entry
1075 	 *
1076 	 * Note that we don't do a (daddr == all-routers-mcast) check.
1077 	 */
1078 	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1079 	if (!neigh && lladdr && idev && READ_ONCE(idev->cnf.forwarding)) {
1080 		if (accept_untracked_na(idev, saddr)) {
1081 			neigh = neigh_create(&nd_tbl, &msg->target, dev);
1082 			new_state = NUD_STALE;
1083 		}
1084 	}
1085 
1086 	if (neigh && !IS_ERR(neigh)) {
1087 		u8 old_flags = neigh->flags;
1088 		struct net *net = dev_net(dev);
1089 
1090 		if (READ_ONCE(neigh->nud_state) & NUD_FAILED)
1091 			goto out;
1092 
1093 		/*
1094 		 * Don't update the neighbor cache entry on a proxy NA from
1095 		 * ourselves because either the proxied node is off link or it
1096 		 * has already sent a NA to us.
1097 		 */
1098 		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1099 		    READ_ONCE(net->ipv6.devconf_all->forwarding) &&
1100 		    READ_ONCE(net->ipv6.devconf_all->proxy_ndp) &&
1101 		    pneigh_lookup(&nd_tbl, net, &msg->target, dev)) {
1102 			/* XXX: idev->cnf.proxy_ndp */
1103 			goto out;
1104 		}
1105 
1106 		ndisc_update(dev, neigh, lladdr,
1107 			     new_state,
1108 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1109 			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1110 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1111 			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1112 			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1113 
1114 		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1115 			/*
1116 			 * Change: router to host
1117 			 */
1118 			rt6_clean_tohost(dev_net(dev),  saddr);
1119 		}
1120 		reason = SKB_CONSUMED;
1121 out:
1122 		neigh_release(neigh);
1123 	}
1124 	return reason;
1125 }
1126 
1127 static enum skb_drop_reason ndisc_recv_rs(struct sk_buff *skb)
1128 {
1129 	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1130 	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1131 	struct neighbour *neigh;
1132 	struct inet6_dev *idev;
1133 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1134 	struct ndisc_options ndopts;
1135 	u8 *lladdr = NULL;
1136 	SKB_DR(reason);
1137 
1138 	if (skb->len < sizeof(*rs_msg))
1139 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1140 
1141 	idev = __in6_dev_get(skb->dev);
1142 	if (!idev) {
1143 		net_err_ratelimited("RS: can't find in6 device\n");
1144 		return reason;
1145 	}
1146 
1147 	/* Don't accept RS if we're not in router mode */
1148 	if (!READ_ONCE(idev->cnf.forwarding))
1149 		goto out;
1150 
1151 	/*
1152 	 * Don't update NCE if src = ::;
1153 	 * this implies that the source node has no ip address assigned yet.
1154 	 */
1155 	if (ipv6_addr_any(saddr))
1156 		goto out;
1157 
1158 	/* Parse ND options */
1159 	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts))
1160 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1161 
1162 	if (ndopts.nd_opts_src_lladdr) {
1163 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1164 					     skb->dev);
1165 		if (!lladdr)
1166 			goto out;
1167 	}
1168 
1169 	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1170 	if (neigh) {
1171 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1172 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1173 			     NEIGH_UPDATE_F_OVERRIDE|
1174 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1175 			     NDISC_ROUTER_SOLICITATION, &ndopts);
1176 		neigh_release(neigh);
1177 		reason = SKB_CONSUMED;
1178 	}
1179 out:
1180 	return reason;
1181 }
1182 
1183 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1184 {
1185 	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1186 	struct sk_buff *skb;
1187 	struct nlmsghdr *nlh;
1188 	struct nduseroptmsg *ndmsg;
1189 	struct net *net = dev_net(ra->dev);
1190 	int err;
1191 	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1192 				    + (opt->nd_opt_len << 3));
1193 	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1194 
1195 	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1196 	if (!skb) {
1197 		err = -ENOBUFS;
1198 		goto errout;
1199 	}
1200 
1201 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1202 	if (!nlh) {
1203 		goto nla_put_failure;
1204 	}
1205 
1206 	ndmsg = nlmsg_data(nlh);
1207 	ndmsg->nduseropt_family = AF_INET6;
1208 	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1209 	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1210 	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1211 	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1212 	ndmsg->nduseropt_pad1 = 0;
1213 	ndmsg->nduseropt_pad2 = 0;
1214 	ndmsg->nduseropt_pad3 = 0;
1215 
1216 	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1217 
1218 	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1219 		goto nla_put_failure;
1220 	nlmsg_end(skb, nlh);
1221 
1222 	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1223 	return;
1224 
1225 nla_put_failure:
1226 	nlmsg_free(skb);
1227 	err = -EMSGSIZE;
1228 errout:
1229 	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1230 }
1231 
1232 static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb)
1233 {
1234 	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1235 	bool send_ifinfo_notify = false;
1236 	struct neighbour *neigh = NULL;
1237 	struct ndisc_options ndopts;
1238 	struct fib6_info *rt = NULL;
1239 	struct inet6_dev *in6_dev;
1240 	struct fib6_table *table;
1241 	u32 defrtr_usr_metric;
1242 	unsigned int pref = 0;
1243 	__u32 old_if_flags;
1244 	struct net *net;
1245 	SKB_DR(reason);
1246 	int lifetime;
1247 	int optlen;
1248 
1249 	__u8 *opt = (__u8 *)(ra_msg + 1);
1250 
1251 	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1252 		sizeof(struct ra_msg);
1253 
1254 	net_dbg_ratelimited("RA: %s, dev: %s\n", __func__, skb->dev->name);
1255 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1256 		net_dbg_ratelimited("RA: source address is not link-local\n");
1257 		return reason;
1258 	}
1259 	if (optlen < 0)
1260 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1261 
1262 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1263 	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1264 		net_dbg_ratelimited("RA: from host or unauthorized router\n");
1265 		return reason;
1266 	}
1267 #endif
1268 
1269 	in6_dev = __in6_dev_get(skb->dev);
1270 	if (!in6_dev) {
1271 		net_err_ratelimited("RA: can't find inet6 device for %s\n", skb->dev->name);
1272 		return reason;
1273 	}
1274 
1275 	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts))
1276 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1277 
1278 	if (!ipv6_accept_ra(in6_dev)) {
1279 		net_dbg_ratelimited("RA: %s, did not accept ra for dev: %s\n", __func__,
1280 				    skb->dev->name);
1281 		goto skip_linkparms;
1282 	}
1283 
1284 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1285 	/* skip link-specific parameters from interior routers */
1286 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1287 		net_dbg_ratelimited("RA: %s, nodetype is NODEFAULT, dev: %s\n", __func__,
1288 				    skb->dev->name);
1289 		goto skip_linkparms;
1290 	}
1291 #endif
1292 
1293 	if (in6_dev->if_flags & IF_RS_SENT) {
1294 		/*
1295 		 *	flag that an RA was received after an RS was sent
1296 		 *	out on this interface.
1297 		 */
1298 		in6_dev->if_flags |= IF_RA_RCVD;
1299 	}
1300 
1301 	/*
1302 	 * Remember the managed/otherconf flags from most recently
1303 	 * received RA message (RFC 2462) -- yoshfuji
1304 	 */
1305 	old_if_flags = in6_dev->if_flags;
1306 	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1307 				IF_RA_OTHERCONF)) |
1308 				(ra_msg->icmph.icmp6_addrconf_managed ?
1309 					IF_RA_MANAGED : 0) |
1310 				(ra_msg->icmph.icmp6_addrconf_other ?
1311 					IF_RA_OTHERCONF : 0);
1312 
1313 	if (old_if_flags != in6_dev->if_flags)
1314 		send_ifinfo_notify = true;
1315 
1316 	if (!READ_ONCE(in6_dev->cnf.accept_ra_defrtr)) {
1317 		net_dbg_ratelimited("RA: %s, defrtr is false for dev: %s\n", __func__,
1318 				    skb->dev->name);
1319 		goto skip_defrtr;
1320 	}
1321 
1322 	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1323 	if (lifetime != 0 &&
1324 	    lifetime < READ_ONCE(in6_dev->cnf.accept_ra_min_lft)) {
1325 		net_dbg_ratelimited("RA: router lifetime (%ds) is too short: %s\n", lifetime,
1326 				    skb->dev->name);
1327 		goto skip_defrtr;
1328 	}
1329 
1330 	/* Do not accept RA with source-addr found on local machine unless
1331 	 * accept_ra_from_local is set to true.
1332 	 */
1333 	net = dev_net(in6_dev->dev);
1334 	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1335 	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1336 		net_dbg_ratelimited("RA from local address detected on dev: %s: default router ignored\n",
1337 				    skb->dev->name);
1338 		goto skip_defrtr;
1339 	}
1340 
1341 #ifdef CONFIG_IPV6_ROUTER_PREF
1342 	pref = ra_msg->icmph.icmp6_router_pref;
1343 	/* 10b is handled as if it were 00b (medium) */
1344 	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1345 	    !READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref))
1346 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1347 #endif
1348 	/* routes added from RAs do not use nexthop objects */
1349 	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1350 	if (rt) {
1351 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1352 					 rt->fib6_nh->fib_nh_dev, NULL,
1353 					  &ipv6_hdr(skb)->saddr);
1354 		if (!neigh) {
1355 			net_err_ratelimited("RA: %s got default router without neighbour\n",
1356 					    __func__);
1357 			fib6_info_release(rt);
1358 			return reason;
1359 		}
1360 	}
1361 	/* Set default route metric as specified by user */
1362 	defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1363 	/* delete the route if lifetime is 0 or if metric needs change */
1364 	if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1365 		ip6_del_rt(net, rt, false);
1366 		rt = NULL;
1367 	}
1368 
1369 	net_dbg_ratelimited("RA: rt: %p  lifetime: %d, metric: %d, for dev: %s\n", rt, lifetime,
1370 			    defrtr_usr_metric, skb->dev->name);
1371 	if (!rt && lifetime) {
1372 		net_dbg_ratelimited("RA: adding default router\n");
1373 
1374 		if (neigh)
1375 			neigh_release(neigh);
1376 
1377 		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1378 					 skb->dev, pref, defrtr_usr_metric,
1379 					 lifetime);
1380 		if (!rt) {
1381 			net_err_ratelimited("RA: %s failed to add default route\n", __func__);
1382 			return reason;
1383 		}
1384 
1385 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1386 					 rt->fib6_nh->fib_nh_dev, NULL,
1387 					  &ipv6_hdr(skb)->saddr);
1388 		if (!neigh) {
1389 			net_err_ratelimited("RA: %s got default router without neighbour\n",
1390 					    __func__);
1391 			fib6_info_release(rt);
1392 			return reason;
1393 		}
1394 		neigh->flags |= NTF_ROUTER;
1395 	} else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1396 		struct nl_info nlinfo = {
1397 			.nl_net = net,
1398 		};
1399 		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1400 		inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1401 	}
1402 
1403 	if (rt) {
1404 		table = rt->fib6_table;
1405 		spin_lock_bh(&table->tb6_lock);
1406 
1407 		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1408 		fib6_add_gc_list(rt);
1409 
1410 		spin_unlock_bh(&table->tb6_lock);
1411 	}
1412 	if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) < 256 &&
1413 	    ra_msg->icmph.icmp6_hop_limit) {
1414 		if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) <=
1415 		    ra_msg->icmph.icmp6_hop_limit) {
1416 			WRITE_ONCE(in6_dev->cnf.hop_limit,
1417 				   ra_msg->icmph.icmp6_hop_limit);
1418 			fib6_metric_set(rt, RTAX_HOPLIMIT,
1419 					ra_msg->icmph.icmp6_hop_limit);
1420 		} else {
1421 			net_dbg_ratelimited("RA: Got route advertisement with lower hop_limit than minimum\n");
1422 		}
1423 	}
1424 
1425 skip_defrtr:
1426 
1427 	/*
1428 	 *	Update Reachable Time and Retrans Timer
1429 	 */
1430 
1431 	if (in6_dev->nd_parms) {
1432 		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1433 
1434 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1435 			rtime = (rtime*HZ)/1000;
1436 			if (rtime < HZ/100)
1437 				rtime = HZ/100;
1438 			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1439 			in6_dev->tstamp = jiffies;
1440 			send_ifinfo_notify = true;
1441 		}
1442 
1443 		rtime = ntohl(ra_msg->reachable_time);
1444 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1445 			rtime = (rtime*HZ)/1000;
1446 
1447 			if (rtime < HZ/10)
1448 				rtime = HZ/10;
1449 
1450 			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1451 				NEIGH_VAR_SET(in6_dev->nd_parms,
1452 					      BASE_REACHABLE_TIME, rtime);
1453 				NEIGH_VAR_SET(in6_dev->nd_parms,
1454 					      GC_STALETIME, 3 * rtime);
1455 				neigh_set_reach_time(in6_dev->nd_parms);
1456 				in6_dev->tstamp = jiffies;
1457 				send_ifinfo_notify = true;
1458 			}
1459 		}
1460 	}
1461 
1462 skip_linkparms:
1463 
1464 	/*
1465 	 *	Process options.
1466 	 */
1467 
1468 	if (!neigh)
1469 		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1470 				       skb->dev, 1);
1471 	if (neigh) {
1472 		u8 *lladdr = NULL;
1473 		if (ndopts.nd_opts_src_lladdr) {
1474 			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1475 						     skb->dev);
1476 			if (!lladdr) {
1477 				net_dbg_ratelimited("RA: invalid link-layer address length\n");
1478 				goto out;
1479 			}
1480 		}
1481 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1482 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1483 			     NEIGH_UPDATE_F_OVERRIDE|
1484 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1485 			     NEIGH_UPDATE_F_ISROUTER,
1486 			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1487 		reason = SKB_CONSUMED;
1488 	}
1489 
1490 	if (!ipv6_accept_ra(in6_dev)) {
1491 		net_dbg_ratelimited("RA: %s, accept_ra is false for dev: %s\n", __func__,
1492 				    skb->dev->name);
1493 		goto out;
1494 	}
1495 
1496 #ifdef CONFIG_IPV6_ROUTE_INFO
1497 	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1498 	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1499 			  in6_dev->dev, 0)) {
1500 		net_dbg_ratelimited("RA from local address detected on dev: %s: router info ignored.\n",
1501 				    skb->dev->name);
1502 		goto skip_routeinfo;
1503 	}
1504 
1505 	if (READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref) && ndopts.nd_opts_ri) {
1506 		struct nd_opt_hdr *p;
1507 		for (p = ndopts.nd_opts_ri;
1508 		     p;
1509 		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1510 			struct route_info *ri = (struct route_info *)p;
1511 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1512 			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1513 			    ri->prefix_len == 0)
1514 				continue;
1515 #endif
1516 			if (ri->prefix_len == 0 &&
1517 			    !READ_ONCE(in6_dev->cnf.accept_ra_defrtr))
1518 				continue;
1519 			if (ri->lifetime != 0 &&
1520 			    ntohl(ri->lifetime) < READ_ONCE(in6_dev->cnf.accept_ra_min_lft))
1521 				continue;
1522 			if (ri->prefix_len < READ_ONCE(in6_dev->cnf.accept_ra_rt_info_min_plen))
1523 				continue;
1524 			if (ri->prefix_len > READ_ONCE(in6_dev->cnf.accept_ra_rt_info_max_plen))
1525 				continue;
1526 			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1527 				      &ipv6_hdr(skb)->saddr);
1528 		}
1529 	}
1530 
1531 skip_routeinfo:
1532 #endif
1533 
1534 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1535 	/* skip link-specific ndopts from interior routers */
1536 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1537 		net_dbg_ratelimited("RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1538 				    __func__, skb->dev->name);
1539 		goto out;
1540 	}
1541 #endif
1542 
1543 	if (READ_ONCE(in6_dev->cnf.accept_ra_pinfo) && ndopts.nd_opts_pi) {
1544 		struct nd_opt_hdr *p;
1545 		for (p = ndopts.nd_opts_pi;
1546 		     p;
1547 		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1548 			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1549 					    (p->nd_opt_len) << 3,
1550 					    ndopts.nd_opts_src_lladdr != NULL);
1551 		}
1552 	}
1553 
1554 	if (ndopts.nd_opts_mtu && READ_ONCE(in6_dev->cnf.accept_ra_mtu)) {
1555 		__be32 n;
1556 		u32 mtu;
1557 
1558 		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1559 		mtu = ntohl(n);
1560 
1561 		if (READ_ONCE(in6_dev->ra_mtu) != mtu) {
1562 			WRITE_ONCE(in6_dev->ra_mtu, mtu);
1563 			send_ifinfo_notify = true;
1564 		}
1565 
1566 		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1567 			net_dbg_ratelimited("RA: invalid mtu: %d\n", mtu);
1568 		} else if (READ_ONCE(in6_dev->cnf.mtu6) != mtu) {
1569 			WRITE_ONCE(in6_dev->cnf.mtu6, mtu);
1570 			fib6_metric_set(rt, RTAX_MTU, mtu);
1571 			rt6_mtu_change(skb->dev, mtu);
1572 		}
1573 	}
1574 
1575 	if (ndopts.nd_useropts) {
1576 		struct nd_opt_hdr *p;
1577 		for (p = ndopts.nd_useropts;
1578 		     p;
1579 		     p = ndisc_next_useropt(skb->dev, p,
1580 					    ndopts.nd_useropts_end)) {
1581 			ndisc_ra_useropt(skb, p);
1582 		}
1583 	}
1584 
1585 	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1586 		net_dbg_ratelimited("RA: invalid RA options\n");
1587 	}
1588 out:
1589 	/* Send a notify if RA changed managed/otherconf flags or
1590 	 * timer settings or ra_mtu value
1591 	 */
1592 	if (send_ifinfo_notify)
1593 		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1594 
1595 	fib6_info_release(rt);
1596 	if (neigh)
1597 		neigh_release(neigh);
1598 	return reason;
1599 }
1600 
1601 static enum skb_drop_reason ndisc_redirect_rcv(struct sk_buff *skb)
1602 {
1603 	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1604 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1605 				    offsetof(struct rd_msg, opt));
1606 	struct ndisc_options ndopts;
1607 	SKB_DR(reason);
1608 	u8 *hdr;
1609 
1610 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1611 	switch (skb->ndisc_nodetype) {
1612 	case NDISC_NODETYPE_HOST:
1613 	case NDISC_NODETYPE_NODEFAULT:
1614 		net_dbg_ratelimited("Redirect: from host or unauthorized router\n");
1615 		return reason;
1616 	}
1617 #endif
1618 
1619 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1620 		net_dbg_ratelimited("Redirect: source address is not link-local\n");
1621 		return reason;
1622 	}
1623 
1624 	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1625 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1626 
1627 	if (!ndopts.nd_opts_rh) {
1628 		ip6_redirect_no_header(skb, dev_net(skb->dev),
1629 					skb->dev->ifindex);
1630 		return reason;
1631 	}
1632 
1633 	hdr = (u8 *)ndopts.nd_opts_rh;
1634 	hdr += 8;
1635 	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1636 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1637 
1638 	return icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1639 }
1640 
1641 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1642 					   struct sk_buff *orig_skb,
1643 					   int rd_len)
1644 {
1645 	u8 *opt = skb_put(skb, rd_len);
1646 
1647 	memset(opt, 0, 8);
1648 	*(opt++) = ND_OPT_REDIRECT_HDR;
1649 	*(opt++) = (rd_len >> 3);
1650 	opt += 6;
1651 
1652 	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1653 		      rd_len - 8);
1654 }
1655 
1656 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1657 {
1658 	struct net_device *dev = skb->dev;
1659 	struct net *net = dev_net_rcu(dev);
1660 	struct sock *sk = net->ipv6.ndisc_sk;
1661 	int optlen = 0;
1662 	struct inet_peer *peer;
1663 	struct sk_buff *buff;
1664 	struct rd_msg *msg;
1665 	struct in6_addr saddr_buf;
1666 	struct rt6_info *rt;
1667 	struct dst_entry *dst;
1668 	struct flowi6 fl6;
1669 	int rd_len;
1670 	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1671 	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1672 	bool ret;
1673 
1674 	if (netif_is_l3_master(dev)) {
1675 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
1676 		if (!dev)
1677 			return;
1678 	}
1679 
1680 	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1681 		net_dbg_ratelimited("Redirect: no link-local address on %s\n", dev->name);
1682 		return;
1683 	}
1684 
1685 	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1686 	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1687 		net_dbg_ratelimited("Redirect: target address is not link-local unicast\n");
1688 		return;
1689 	}
1690 
1691 	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1692 			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1693 
1694 	dst = ip6_route_output(net, NULL, &fl6);
1695 	if (dst->error) {
1696 		dst_release(dst);
1697 		return;
1698 	}
1699 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1700 	if (IS_ERR(dst))
1701 		return;
1702 
1703 	rt = dst_rt6_info(dst);
1704 
1705 	if (rt->rt6i_flags & RTF_GATEWAY) {
1706 		net_dbg_ratelimited("Redirect: destination is not a neighbour\n");
1707 		goto release;
1708 	}
1709 
1710 	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr);
1711 	if (!peer)
1712 		goto release;
1713 	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1714 
1715 	if (!ret)
1716 		goto release;
1717 
1718 	if (dev->addr_len) {
1719 		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1720 		if (!neigh) {
1721 			net_dbg_ratelimited("Redirect: no neigh for target address\n");
1722 			goto release;
1723 		}
1724 
1725 		read_lock_bh(&neigh->lock);
1726 		if (neigh->nud_state & NUD_VALID) {
1727 			memcpy(ha_buf, neigh->ha, dev->addr_len);
1728 			read_unlock_bh(&neigh->lock);
1729 			ha = ha_buf;
1730 			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1731 								ops_data_buf,
1732 								&ops_data);
1733 		} else
1734 			read_unlock_bh(&neigh->lock);
1735 
1736 		neigh_release(neigh);
1737 	}
1738 
1739 	rd_len = min_t(unsigned int,
1740 		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1741 		       skb->len + 8);
1742 	rd_len &= ~0x7;
1743 	optlen += rd_len;
1744 
1745 	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1746 	if (!buff)
1747 		goto release;
1748 
1749 	msg = skb_put(buff, sizeof(*msg));
1750 	*msg = (struct rd_msg) {
1751 		.icmph = {
1752 			.icmp6_type = NDISC_REDIRECT,
1753 		},
1754 		.target = *target,
1755 		.dest = ipv6_hdr(skb)->daddr,
1756 	};
1757 
1758 	/*
1759 	 *	include target_address option
1760 	 */
1761 
1762 	if (ha)
1763 		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1764 
1765 	/*
1766 	 *	build redirect option and copy skb over to the new packet.
1767 	 */
1768 
1769 	if (rd_len)
1770 		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1771 
1772 	skb_dst_set(buff, dst);
1773 	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1774 	return;
1775 
1776 release:
1777 	dst_release(dst);
1778 }
1779 
1780 static void pndisc_redo(struct sk_buff *skb)
1781 {
1782 	enum skb_drop_reason reason = ndisc_recv_ns(skb);
1783 
1784 	kfree_skb_reason(skb, reason);
1785 }
1786 
1787 static int ndisc_is_multicast(const void *pkey)
1788 {
1789 	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1790 }
1791 
1792 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1793 {
1794 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1795 
1796 	if (!idev)
1797 		return true;
1798 	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1799 	    READ_ONCE(idev->cnf.suppress_frag_ndisc)) {
1800 		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1801 		return true;
1802 	}
1803 	return false;
1804 }
1805 
1806 enum skb_drop_reason ndisc_rcv(struct sk_buff *skb)
1807 {
1808 	struct nd_msg *msg;
1809 	SKB_DR(reason);
1810 
1811 	if (ndisc_suppress_frag_ndisc(skb))
1812 		return SKB_DROP_REASON_IPV6_NDISC_FRAG;
1813 
1814 	if (skb_linearize(skb))
1815 		return SKB_DROP_REASON_NOMEM;
1816 
1817 	msg = (struct nd_msg *)skb_transport_header(skb);
1818 
1819 	__skb_push(skb, skb->data - skb_transport_header(skb));
1820 
1821 	if (ipv6_hdr(skb)->hop_limit != 255) {
1822 		net_dbg_ratelimited("NDISC: invalid hop-limit: %d\n", ipv6_hdr(skb)->hop_limit);
1823 		return SKB_DROP_REASON_IPV6_NDISC_HOP_LIMIT;
1824 	}
1825 
1826 	if (msg->icmph.icmp6_code != 0) {
1827 		net_dbg_ratelimited("NDISC: invalid ICMPv6 code: %d\n", msg->icmph.icmp6_code);
1828 		return SKB_DROP_REASON_IPV6_NDISC_BAD_CODE;
1829 	}
1830 
1831 	switch (msg->icmph.icmp6_type) {
1832 	case NDISC_NEIGHBOUR_SOLICITATION:
1833 		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1834 		reason = ndisc_recv_ns(skb);
1835 		break;
1836 
1837 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1838 		reason = ndisc_recv_na(skb);
1839 		break;
1840 
1841 	case NDISC_ROUTER_SOLICITATION:
1842 		reason = ndisc_recv_rs(skb);
1843 		break;
1844 
1845 	case NDISC_ROUTER_ADVERTISEMENT:
1846 		reason = ndisc_router_discovery(skb);
1847 		break;
1848 
1849 	case NDISC_REDIRECT:
1850 		reason = ndisc_redirect_rcv(skb);
1851 		break;
1852 	}
1853 
1854 	return reason;
1855 }
1856 
1857 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1858 {
1859 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1860 	struct netdev_notifier_change_info *change_info;
1861 	struct net *net = dev_net(dev);
1862 	struct inet6_dev *idev;
1863 	bool evict_nocarrier;
1864 
1865 	switch (event) {
1866 	case NETDEV_CHANGEADDR:
1867 		neigh_changeaddr(&nd_tbl, dev);
1868 		fib6_run_gc(0, net, false);
1869 		fallthrough;
1870 	case NETDEV_UP:
1871 		idev = in6_dev_get(dev);
1872 		if (!idev)
1873 			break;
1874 		if (READ_ONCE(idev->cnf.ndisc_notify) ||
1875 		    READ_ONCE(net->ipv6.devconf_all->ndisc_notify))
1876 			ndisc_send_unsol_na(dev);
1877 		in6_dev_put(idev);
1878 		break;
1879 	case NETDEV_CHANGE:
1880 		idev = in6_dev_get(dev);
1881 		if (!idev)
1882 			evict_nocarrier = true;
1883 		else {
1884 			evict_nocarrier = READ_ONCE(idev->cnf.ndisc_evict_nocarrier) &&
1885 					  READ_ONCE(net->ipv6.devconf_all->ndisc_evict_nocarrier);
1886 			in6_dev_put(idev);
1887 		}
1888 
1889 		change_info = ptr;
1890 		if (change_info->flags_changed & IFF_NOARP)
1891 			neigh_changeaddr(&nd_tbl, dev);
1892 		if (evict_nocarrier && !netif_carrier_ok(dev))
1893 			neigh_carrier_down(&nd_tbl, dev);
1894 		break;
1895 	case NETDEV_DOWN:
1896 		neigh_ifdown(&nd_tbl, dev);
1897 		fib6_run_gc(0, net, false);
1898 		break;
1899 	case NETDEV_NOTIFY_PEERS:
1900 		ndisc_send_unsol_na(dev);
1901 		break;
1902 	default:
1903 		break;
1904 	}
1905 
1906 	return NOTIFY_DONE;
1907 }
1908 
1909 static struct notifier_block ndisc_netdev_notifier = {
1910 	.notifier_call = ndisc_netdev_event,
1911 	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1912 };
1913 
1914 #ifdef CONFIG_SYSCTL
1915 static void ndisc_warn_deprecated_sysctl(const struct ctl_table *ctl,
1916 					 const char *func, const char *dev_name)
1917 {
1918 	static char warncomm[TASK_COMM_LEN];
1919 	static int warned;
1920 	if (strcmp(warncomm, current->comm) && warned < 5) {
1921 		strscpy(warncomm, current->comm);
1922 		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1923 			warncomm, func,
1924 			dev_name, ctl->procname,
1925 			dev_name, ctl->procname);
1926 		warned++;
1927 	}
1928 }
1929 
1930 int ndisc_ifinfo_sysctl_change(const struct ctl_table *ctl, int write, void *buffer,
1931 		size_t *lenp, loff_t *ppos)
1932 {
1933 	struct net_device *dev = ctl->extra1;
1934 	struct inet6_dev *idev;
1935 	int ret;
1936 
1937 	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1938 	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1939 		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1940 
1941 	if (strcmp(ctl->procname, "retrans_time") == 0)
1942 		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1943 
1944 	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1945 		ret = neigh_proc_dointvec_jiffies(ctl, write,
1946 						  buffer, lenp, ppos);
1947 
1948 	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1949 		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1950 		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1951 						     buffer, lenp, ppos);
1952 	else
1953 		ret = -1;
1954 
1955 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1956 		if (ctl->data == NEIGH_VAR_PTR(idev->nd_parms, BASE_REACHABLE_TIME))
1957 			neigh_set_reach_time(idev->nd_parms);
1958 
1959 		WRITE_ONCE(idev->tstamp, jiffies);
1960 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1961 		in6_dev_put(idev);
1962 	}
1963 	return ret;
1964 }
1965 
1966 
1967 #endif
1968 
1969 static int __net_init ndisc_net_init(struct net *net)
1970 {
1971 	struct ipv6_pinfo *np;
1972 	struct sock *sk;
1973 	int err;
1974 
1975 	err = inet_ctl_sock_create(&sk, PF_INET6,
1976 				   SOCK_RAW, IPPROTO_ICMPV6, net);
1977 	if (err < 0) {
1978 		net_err_ratelimited("NDISC: Failed to initialize the control socket (err %d)\n",
1979 				    err);
1980 		return err;
1981 	}
1982 
1983 	net->ipv6.ndisc_sk = sk;
1984 
1985 	np = inet6_sk(sk);
1986 	np->hop_limit = 255;
1987 	/* Do not loopback ndisc messages */
1988 	inet6_clear_bit(MC6_LOOP, sk);
1989 
1990 	return 0;
1991 }
1992 
1993 static void __net_exit ndisc_net_exit(struct net *net)
1994 {
1995 	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1996 }
1997 
1998 static struct pernet_operations ndisc_net_ops = {
1999 	.init = ndisc_net_init,
2000 	.exit = ndisc_net_exit,
2001 };
2002 
2003 int __init ndisc_init(void)
2004 {
2005 	int err;
2006 
2007 	err = register_pernet_subsys(&ndisc_net_ops);
2008 	if (err)
2009 		return err;
2010 	/*
2011 	 * Initialize the neighbour table
2012 	 */
2013 	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2014 
2015 #ifdef CONFIG_SYSCTL
2016 	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2017 				    ndisc_ifinfo_sysctl_change);
2018 	if (err)
2019 		goto out_unregister_pernet;
2020 out:
2021 #endif
2022 	return err;
2023 
2024 #ifdef CONFIG_SYSCTL
2025 out_unregister_pernet:
2026 	unregister_pernet_subsys(&ndisc_net_ops);
2027 	goto out;
2028 #endif
2029 }
2030 
2031 int __init ndisc_late_init(void)
2032 {
2033 	return register_netdevice_notifier(&ndisc_netdev_notifier);
2034 }
2035 
2036 void ndisc_late_cleanup(void)
2037 {
2038 	unregister_netdevice_notifier(&ndisc_netdev_notifier);
2039 }
2040 
2041 void ndisc_cleanup(void)
2042 {
2043 #ifdef CONFIG_SYSCTL
2044 	neigh_sysctl_unregister(&nd_tbl.parms);
2045 #endif
2046 	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2047 	unregister_pernet_subsys(&ndisc_net_ops);
2048 }
2049