• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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_GC_STALETIME] = 60 * HZ,
132 			[NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
133 			[NEIGH_VAR_PROXY_QLEN] = 64,
134 			[NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
135 			[NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
136 		},
137 	},
138 	.gc_interval =	  30 * HZ,
139 	.gc_thresh1 =	 128,
140 	.gc_thresh2 =	 512,
141 	.gc_thresh3 =	1024,
142 };
143 EXPORT_SYMBOL_GPL(nd_tbl);
144 
__ndisc_fill_addr_option(struct sk_buff * skb,int type,void * data,int data_len,int pad)145 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, void *data,
146 			      int data_len, int pad)
147 {
148 	int space = __ndisc_opt_addr_space(data_len, pad);
149 	u8 *opt = skb_put(skb, space);
150 
151 	opt[0] = type;
152 	opt[1] = space>>3;
153 
154 	memset(opt + 2, 0, pad);
155 	opt   += pad;
156 	space -= pad;
157 
158 	memcpy(opt+2, data, data_len);
159 	data_len += 2;
160 	opt += data_len;
161 	space -= data_len;
162 	if (space > 0)
163 		memset(opt, 0, space);
164 }
165 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
166 
ndisc_fill_addr_option(struct sk_buff * skb,int type,void * data,u8 icmp6_type)167 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
168 					  void *data, u8 icmp6_type)
169 {
170 	__ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
171 				 ndisc_addr_option_pad(skb->dev->type));
172 	ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
173 }
174 
ndisc_fill_redirect_addr_option(struct sk_buff * skb,void * ha,const u8 * ops_data)175 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
176 						   void *ha,
177 						   const u8 *ops_data)
178 {
179 	ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
180 	ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
181 }
182 
ndisc_next_option(struct nd_opt_hdr * cur,struct nd_opt_hdr * end)183 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
184 					    struct nd_opt_hdr *end)
185 {
186 	int type;
187 	if (!cur || !end || cur >= end)
188 		return NULL;
189 	type = cur->nd_opt_type;
190 	do {
191 		cur = ((void *)cur) + (cur->nd_opt_len << 3);
192 	} while (cur < end && cur->nd_opt_type != type);
193 	return cur <= end && cur->nd_opt_type == type ? cur : NULL;
194 }
195 
ndisc_is_useropt(const struct net_device * dev,struct nd_opt_hdr * opt)196 static inline int ndisc_is_useropt(const struct net_device *dev,
197 				   struct nd_opt_hdr *opt)
198 {
199 	return opt->nd_opt_type == ND_OPT_PREFIX_INFO ||
200 		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 
ndisc_next_useropt(const struct net_device * dev,struct nd_opt_hdr * cur,struct nd_opt_hdr * end)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 
ndisc_parse_options(const struct net_device * dev,u8 * opt,int opt_len,struct ndisc_options * ndopts)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 
ndisc_mc_map(const struct in6_addr * addr,char * buf,struct net_device * dev,int dir)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 
ndisc_hash(const void * pkey,const struct net_device * dev,__u32 * hash_rnd)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 
ndisc_key_eq(const struct neighbour * n,const void * pkey)321 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
322 {
323 	return neigh_key_eq128(n, pkey);
324 }
325 
ndisc_constructor(struct neighbour * neigh)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 
pndisc_constructor(struct pneigh_entry * n)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 
pndisc_destructor(struct pneigh_entry * n)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 */
ndisc_allow_add(const struct net_device * dev,struct netlink_ext_ack * extack)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 
ndisc_alloc_skb(struct net_device * dev,int len)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 
ip6_nd_hdr(struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,int hop_limit,int len)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 
ndisc_send_skb(struct sk_buff * skb,const struct in6_addr * daddr,const struct in6_addr * saddr)470 static void ndisc_send_skb(struct sk_buff *skb,
471 			   const struct in6_addr *daddr,
472 			   const struct in6_addr *saddr)
473 {
474 	struct dst_entry *dst = skb_dst(skb);
475 	struct net *net = dev_net(skb->dev);
476 	struct sock *sk = net->ipv6.ndisc_sk;
477 	struct inet6_dev *idev;
478 	int err;
479 	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
480 	u8 type;
481 
482 	type = icmp6h->icmp6_type;
483 
484 	if (!dst) {
485 		struct flowi6 fl6;
486 		int oif = skb->dev->ifindex;
487 
488 		icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
489 		dst = icmp6_dst_alloc(skb->dev, &fl6);
490 		if (IS_ERR(dst)) {
491 			kfree_skb(skb);
492 			return;
493 		}
494 
495 		skb_dst_set(skb, dst);
496 	}
497 
498 	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
499 					      IPPROTO_ICMPV6,
500 					      csum_partial(icmp6h,
501 							   skb->len, 0));
502 
503 	ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
504 
505 	rcu_read_lock();
506 	idev = __in6_dev_get(dst->dev);
507 	IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
508 
509 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
510 		      net, sk, skb, NULL, dst->dev,
511 		      dst_output);
512 	if (!err) {
513 		ICMP6MSGOUT_INC_STATS(net, idev, type);
514 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
515 	}
516 
517 	rcu_read_unlock();
518 }
519 
ndisc_send_na(struct net_device * dev,const struct in6_addr * daddr,const struct in6_addr * solicited_addr,bool router,bool solicited,bool override,bool inc_opt)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 
ndisc_send_unsol_na(struct net_device * dev)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 
ndisc_send_ns(struct net_device * dev,const struct in6_addr * solicit,const struct in6_addr * daddr,const struct in6_addr * saddr,u64 nonce)602 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
603 		   const struct in6_addr *daddr, const struct in6_addr *saddr,
604 		   u64 nonce)
605 {
606 	struct sk_buff *skb;
607 	struct in6_addr addr_buf;
608 	int inc_opt = dev->addr_len;
609 	int optlen = 0;
610 	struct nd_msg *msg;
611 
612 	if (!saddr) {
613 		if (ipv6_get_lladdr(dev, &addr_buf,
614 				   (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
615 			return;
616 		saddr = &addr_buf;
617 	}
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;
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 	ndisc_send_skb(skb, daddr, saddr);
652 }
653 
ndisc_send_rs(struct net_device * dev,const struct in6_addr * saddr,const struct in6_addr * daddr)654 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
655 		   const struct in6_addr *daddr)
656 {
657 	struct sk_buff *skb;
658 	struct rs_msg *msg;
659 	int send_sllao = dev->addr_len;
660 	int optlen = 0;
661 
662 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
663 	/*
664 	 * According to section 2.2 of RFC 4429, we must not
665 	 * send router solicitations with a sllao from
666 	 * optimistic addresses, but we may send the solicitation
667 	 * if we don't include the sllao.  So here we check
668 	 * if our address is optimistic, and if so, we
669 	 * suppress the inclusion of the sllao.
670 	 */
671 	if (send_sllao) {
672 		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
673 							   dev, 1);
674 		if (ifp) {
675 			if (ifp->flags & IFA_F_OPTIMISTIC)  {
676 				send_sllao = 0;
677 			}
678 			in6_ifa_put(ifp);
679 		} else {
680 			send_sllao = 0;
681 		}
682 	}
683 #endif
684 	if (send_sllao)
685 		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
686 
687 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
688 	if (!skb)
689 		return;
690 
691 	msg = skb_put(skb, sizeof(*msg));
692 	*msg = (struct rs_msg) {
693 		.icmph = {
694 			.icmp6_type = NDISC_ROUTER_SOLICITATION,
695 		},
696 	};
697 
698 	if (send_sllao)
699 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
700 				       dev->dev_addr,
701 				       NDISC_ROUTER_SOLICITATION);
702 
703 	ndisc_send_skb(skb, daddr, saddr);
704 }
705 
706 
ndisc_error_report(struct neighbour * neigh,struct sk_buff * skb)707 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
708 {
709 	/*
710 	 *	"The sender MUST return an ICMP
711 	 *	 destination unreachable"
712 	 */
713 	dst_link_failure(skb);
714 	kfree_skb(skb);
715 }
716 
717 /* Called with locked neigh: either read or both */
718 
ndisc_solicit(struct neighbour * neigh,struct sk_buff * skb)719 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
720 {
721 	struct in6_addr *saddr = NULL;
722 	struct in6_addr mcaddr;
723 	struct net_device *dev = neigh->dev;
724 	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
725 	int probes = atomic_read(&neigh->probes);
726 
727 	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
728 					   dev, false, 1,
729 					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
730 		saddr = &ipv6_hdr(skb)->saddr;
731 	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
732 	if (probes < 0) {
733 		if (!(neigh->nud_state & NUD_VALID)) {
734 			ND_PRINTK(1, dbg,
735 				  "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
736 				  __func__, target);
737 		}
738 		ndisc_send_ns(dev, target, target, saddr, 0);
739 	} else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
740 		neigh_app_ns(neigh);
741 	} else {
742 		addrconf_addr_solict_mult(target, &mcaddr);
743 		ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
744 	}
745 }
746 
pndisc_is_router(const void * pkey,struct net_device * dev)747 static int pndisc_is_router(const void *pkey,
748 			    struct net_device *dev)
749 {
750 	struct pneigh_entry *n;
751 	int ret = -1;
752 
753 	read_lock_bh(&nd_tbl.lock);
754 	n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
755 	if (n)
756 		ret = !!(n->flags & NTF_ROUTER);
757 	read_unlock_bh(&nd_tbl.lock);
758 
759 	return ret;
760 }
761 
ndisc_update(const struct net_device * dev,struct neighbour * neigh,const u8 * lladdr,u8 new,u32 flags,u8 icmp6_type,struct ndisc_options * ndopts)762 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
763 		  const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
764 		  struct ndisc_options *ndopts)
765 {
766 	neigh_update(neigh, lladdr, new, flags, 0);
767 	/* report ndisc ops about neighbour update */
768 	ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
769 }
770 
ndisc_recv_ns(struct sk_buff * skb)771 static void ndisc_recv_ns(struct sk_buff *skb)
772 {
773 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
774 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
775 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
776 	u8 *lladdr = NULL;
777 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
778 				    offsetof(struct nd_msg, opt));
779 	struct ndisc_options ndopts;
780 	struct net_device *dev = skb->dev;
781 	struct inet6_ifaddr *ifp;
782 	struct inet6_dev *idev = NULL;
783 	struct neighbour *neigh;
784 	int dad = ipv6_addr_any(saddr);
785 	bool inc;
786 	int is_router = -1;
787 	u64 nonce = 0;
788 
789 	if (skb->len < sizeof(struct nd_msg)) {
790 		ND_PRINTK(2, warn, "NS: packet too short\n");
791 		return;
792 	}
793 
794 	if (ipv6_addr_is_multicast(&msg->target)) {
795 		ND_PRINTK(2, warn, "NS: multicast target address\n");
796 		return;
797 	}
798 
799 	/*
800 	 * RFC2461 7.1.1:
801 	 * DAD has to be destined for solicited node multicast address.
802 	 */
803 	if (dad && !ipv6_addr_is_solict_mult(daddr)) {
804 		ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
805 		return;
806 	}
807 
808 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
809 		ND_PRINTK(2, warn, "NS: invalid ND options\n");
810 		return;
811 	}
812 
813 	if (ndopts.nd_opts_src_lladdr) {
814 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
815 		if (!lladdr) {
816 			ND_PRINTK(2, warn,
817 				  "NS: invalid link-layer address length\n");
818 			return;
819 		}
820 
821 		/* RFC2461 7.1.1:
822 		 *	If the IP source address is the unspecified address,
823 		 *	there MUST NOT be source link-layer address option
824 		 *	in the message.
825 		 */
826 		if (dad) {
827 			ND_PRINTK(2, warn,
828 				  "NS: bad DAD packet (link-layer address option)\n");
829 			return;
830 		}
831 	}
832 	if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
833 		memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
834 
835 	inc = ipv6_addr_is_multicast(daddr);
836 
837 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
838 	if (ifp) {
839 have_ifp:
840 		if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
841 			if (dad) {
842 				if (nonce != 0 && ifp->dad_nonce == nonce) {
843 					u8 *np = (u8 *)&nonce;
844 					/* Matching nonce if looped back */
845 					ND_PRINTK(2, notice,
846 						  "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
847 						  ifp->idev->dev->name,
848 						  &ifp->addr, np);
849 					goto out;
850 				}
851 				/*
852 				 * We are colliding with another node
853 				 * who is doing DAD
854 				 * so fail our DAD process
855 				 */
856 				addrconf_dad_failure(skb, ifp);
857 				return;
858 			} else {
859 				/*
860 				 * This is not a dad solicitation.
861 				 * If we are an optimistic node,
862 				 * we should respond.
863 				 * Otherwise, we should ignore it.
864 				 */
865 				if (!(ifp->flags & IFA_F_OPTIMISTIC))
866 					goto out;
867 			}
868 		}
869 
870 		idev = ifp->idev;
871 	} else {
872 		struct net *net = dev_net(dev);
873 
874 		/* perhaps an address on the master device */
875 		if (netif_is_l3_slave(dev)) {
876 			struct net_device *mdev;
877 
878 			mdev = netdev_master_upper_dev_get_rcu(dev);
879 			if (mdev) {
880 				ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
881 				if (ifp)
882 					goto have_ifp;
883 			}
884 		}
885 
886 		idev = in6_dev_get(dev);
887 		if (!idev) {
888 			/* XXX: count this drop? */
889 			return;
890 		}
891 
892 		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
893 		    (idev->cnf.forwarding &&
894 		     (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
895 		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
896 			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
897 			    skb->pkt_type != PACKET_HOST &&
898 			    inc &&
899 			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
900 				/*
901 				 * for anycast or proxy,
902 				 * sender should delay its response
903 				 * by a random time between 0 and
904 				 * MAX_ANYCAST_DELAY_TIME seconds.
905 				 * (RFC2461) -- yoshfuji
906 				 */
907 				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
908 				if (n)
909 					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
910 				goto out;
911 			}
912 		} else
913 			goto out;
914 	}
915 
916 	if (is_router < 0)
917 		is_router = idev->cnf.forwarding;
918 
919 	if (dad) {
920 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
921 			      !!is_router, false, (ifp != NULL), true);
922 		goto out;
923 	}
924 
925 	if (inc)
926 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
927 	else
928 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
929 
930 	/*
931 	 *	update / create cache entry
932 	 *	for the source address
933 	 */
934 	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
935 			       !inc || lladdr || !dev->addr_len);
936 	if (neigh)
937 		ndisc_update(dev, neigh, lladdr, NUD_STALE,
938 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
939 			     NEIGH_UPDATE_F_OVERRIDE,
940 			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
941 	if (neigh || !dev->header_ops) {
942 		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
943 			      true, (ifp != NULL && inc), inc);
944 		if (neigh)
945 			neigh_release(neigh);
946 	}
947 
948 out:
949 	if (ifp)
950 		in6_ifa_put(ifp);
951 	else
952 		in6_dev_put(idev);
953 }
954 
ndisc_recv_na(struct sk_buff * skb)955 static void ndisc_recv_na(struct sk_buff *skb)
956 {
957 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
958 	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
959 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
960 	u8 *lladdr = NULL;
961 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
962 				    offsetof(struct nd_msg, opt));
963 	struct ndisc_options ndopts;
964 	struct net_device *dev = skb->dev;
965 	struct inet6_dev *idev = __in6_dev_get(dev);
966 	struct inet6_ifaddr *ifp;
967 	struct neighbour *neigh;
968 
969 	if (skb->len < sizeof(struct nd_msg)) {
970 		ND_PRINTK(2, warn, "NA: packet too short\n");
971 		return;
972 	}
973 
974 	if (ipv6_addr_is_multicast(&msg->target)) {
975 		ND_PRINTK(2, warn, "NA: target address is multicast\n");
976 		return;
977 	}
978 
979 	if (ipv6_addr_is_multicast(daddr) &&
980 	    msg->icmph.icmp6_solicited) {
981 		ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
982 		return;
983 	}
984 
985 	/* For some 802.11 wireless deployments (and possibly other networks),
986 	 * there will be a NA proxy and unsolicitd packets are attacks
987 	 * and thus should not be accepted.
988 	 */
989 	if (!msg->icmph.icmp6_solicited && idev &&
990 	    idev->cnf.drop_unsolicited_na)
991 		return;
992 
993 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
994 		ND_PRINTK(2, warn, "NS: invalid ND option\n");
995 		return;
996 	}
997 	if (ndopts.nd_opts_tgt_lladdr) {
998 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
999 		if (!lladdr) {
1000 			ND_PRINTK(2, warn,
1001 				  "NA: invalid link-layer address length\n");
1002 			return;
1003 		}
1004 	}
1005 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1006 	if (ifp) {
1007 		if (skb->pkt_type != PACKET_LOOPBACK
1008 		    && (ifp->flags & IFA_F_TENTATIVE)) {
1009 				addrconf_dad_failure(skb, ifp);
1010 				return;
1011 		}
1012 		/* What should we make now? The advertisement
1013 		   is invalid, but ndisc specs say nothing
1014 		   about it. It could be misconfiguration, or
1015 		   an smart proxy agent tries to help us :-)
1016 
1017 		   We should not print the error if NA has been
1018 		   received from loopback - it is just our own
1019 		   unsolicited advertisement.
1020 		 */
1021 		if (skb->pkt_type != PACKET_LOOPBACK)
1022 			ND_PRINTK(1, warn,
1023 				  "NA: %pM advertised our address %pI6c on %s!\n",
1024 				  eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1025 		in6_ifa_put(ifp);
1026 		return;
1027 	}
1028 	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1029 
1030 	if (neigh) {
1031 		u8 old_flags = neigh->flags;
1032 		struct net *net = dev_net(dev);
1033 
1034 		if (neigh->nud_state & NUD_FAILED)
1035 			goto out;
1036 
1037 		/*
1038 		 * Don't update the neighbor cache entry on a proxy NA from
1039 		 * ourselves because either the proxied node is off link or it
1040 		 * has already sent a NA to us.
1041 		 */
1042 		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1043 		    net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1044 		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1045 			/* XXX: idev->cnf.proxy_ndp */
1046 			goto out;
1047 		}
1048 
1049 		ndisc_update(dev, neigh, lladdr,
1050 			     msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1051 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1052 			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1053 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1054 			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1055 			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1056 
1057 		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1058 			/*
1059 			 * Change: router to host
1060 			 */
1061 			rt6_clean_tohost(dev_net(dev),  saddr);
1062 		}
1063 
1064 out:
1065 		neigh_release(neigh);
1066 	}
1067 }
1068 
ndisc_recv_rs(struct sk_buff * skb)1069 static void ndisc_recv_rs(struct sk_buff *skb)
1070 {
1071 	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1072 	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1073 	struct neighbour *neigh;
1074 	struct inet6_dev *idev;
1075 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1076 	struct ndisc_options ndopts;
1077 	u8 *lladdr = NULL;
1078 
1079 	if (skb->len < sizeof(*rs_msg))
1080 		return;
1081 
1082 	idev = __in6_dev_get(skb->dev);
1083 	if (!idev) {
1084 		ND_PRINTK(1, err, "RS: can't find in6 device\n");
1085 		return;
1086 	}
1087 
1088 	/* Don't accept RS if we're not in router mode */
1089 	if (!idev->cnf.forwarding)
1090 		goto out;
1091 
1092 	/*
1093 	 * Don't update NCE if src = ::;
1094 	 * this implies that the source node has no ip address assigned yet.
1095 	 */
1096 	if (ipv6_addr_any(saddr))
1097 		goto out;
1098 
1099 	/* Parse ND options */
1100 	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1101 		ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1102 		goto out;
1103 	}
1104 
1105 	if (ndopts.nd_opts_src_lladdr) {
1106 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1107 					     skb->dev);
1108 		if (!lladdr)
1109 			goto out;
1110 	}
1111 
1112 	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1113 	if (neigh) {
1114 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1115 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1116 			     NEIGH_UPDATE_F_OVERRIDE|
1117 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1118 			     NDISC_ROUTER_SOLICITATION, &ndopts);
1119 		neigh_release(neigh);
1120 	}
1121 out:
1122 	return;
1123 }
1124 
ndisc_ra_useropt(struct sk_buff * ra,struct nd_opt_hdr * opt)1125 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1126 {
1127 	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1128 	struct sk_buff *skb;
1129 	struct nlmsghdr *nlh;
1130 	struct nduseroptmsg *ndmsg;
1131 	struct net *net = dev_net(ra->dev);
1132 	int err;
1133 	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1134 				    + (opt->nd_opt_len << 3));
1135 	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1136 
1137 	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1138 	if (!skb) {
1139 		err = -ENOBUFS;
1140 		goto errout;
1141 	}
1142 
1143 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1144 	if (!nlh) {
1145 		goto nla_put_failure;
1146 	}
1147 
1148 	ndmsg = nlmsg_data(nlh);
1149 	ndmsg->nduseropt_family = AF_INET6;
1150 	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1151 	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1152 	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1153 	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1154 
1155 	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1156 
1157 	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1158 		goto nla_put_failure;
1159 	nlmsg_end(skb, nlh);
1160 
1161 	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1162 	return;
1163 
1164 nla_put_failure:
1165 	nlmsg_free(skb);
1166 	err = -EMSGSIZE;
1167 errout:
1168 	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1169 }
1170 
ndisc_router_discovery(struct sk_buff * skb)1171 static void ndisc_router_discovery(struct sk_buff *skb)
1172 {
1173 	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1174 	struct neighbour *neigh = NULL;
1175 	struct inet6_dev *in6_dev;
1176 	struct fib6_info *rt = NULL;
1177 	struct net *net;
1178 	int lifetime;
1179 	struct ndisc_options ndopts;
1180 	int optlen;
1181 	unsigned int pref = 0;
1182 	__u32 old_if_flags;
1183 	bool send_ifinfo_notify = false;
1184 
1185 	__u8 *opt = (__u8 *)(ra_msg + 1);
1186 
1187 	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1188 		sizeof(struct ra_msg);
1189 
1190 	ND_PRINTK(2, info,
1191 		  "RA: %s, dev: %s\n",
1192 		  __func__, skb->dev->name);
1193 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1194 		ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1195 		return;
1196 	}
1197 	if (optlen < 0) {
1198 		ND_PRINTK(2, warn, "RA: packet too short\n");
1199 		return;
1200 	}
1201 
1202 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1203 	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1204 		ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1205 		return;
1206 	}
1207 #endif
1208 
1209 	/*
1210 	 *	set the RA_RECV flag in the interface
1211 	 */
1212 
1213 	in6_dev = __in6_dev_get(skb->dev);
1214 	if (!in6_dev) {
1215 		ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1216 			  skb->dev->name);
1217 		return;
1218 	}
1219 
1220 	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1221 		ND_PRINTK(2, warn, "RA: invalid ND options\n");
1222 		return;
1223 	}
1224 
1225 	if (!ipv6_accept_ra(in6_dev)) {
1226 		ND_PRINTK(2, info,
1227 			  "RA: %s, did not accept ra for dev: %s\n",
1228 			  __func__, skb->dev->name);
1229 		goto skip_linkparms;
1230 	}
1231 
1232 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1233 	/* skip link-specific parameters from interior routers */
1234 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1235 		ND_PRINTK(2, info,
1236 			  "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1237 			  __func__, skb->dev->name);
1238 		goto skip_linkparms;
1239 	}
1240 #endif
1241 
1242 	if (in6_dev->if_flags & IF_RS_SENT) {
1243 		/*
1244 		 *	flag that an RA was received after an RS was sent
1245 		 *	out on this interface.
1246 		 */
1247 		in6_dev->if_flags |= IF_RA_RCVD;
1248 	}
1249 
1250 	/*
1251 	 * Remember the managed/otherconf flags from most recently
1252 	 * received RA message (RFC 2462) -- yoshfuji
1253 	 */
1254 	old_if_flags = in6_dev->if_flags;
1255 	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1256 				IF_RA_OTHERCONF)) |
1257 				(ra_msg->icmph.icmp6_addrconf_managed ?
1258 					IF_RA_MANAGED : 0) |
1259 				(ra_msg->icmph.icmp6_addrconf_other ?
1260 					IF_RA_OTHERCONF : 0);
1261 
1262 	if (old_if_flags != in6_dev->if_flags)
1263 		send_ifinfo_notify = true;
1264 
1265 	if (!in6_dev->cnf.accept_ra_defrtr) {
1266 		ND_PRINTK(2, info,
1267 			  "RA: %s, defrtr is false for dev: %s\n",
1268 			  __func__, skb->dev->name);
1269 		goto skip_defrtr;
1270 	}
1271 
1272 	/* Do not accept RA with source-addr found on local machine unless
1273 	 * accept_ra_from_local is set to true.
1274 	 */
1275 	net = dev_net(in6_dev->dev);
1276 	if (!in6_dev->cnf.accept_ra_from_local &&
1277 	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1278 		ND_PRINTK(2, info,
1279 			  "RA from local address detected on dev: %s: default router ignored\n",
1280 			  skb->dev->name);
1281 		goto skip_defrtr;
1282 	}
1283 
1284 	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1285 
1286 #ifdef CONFIG_IPV6_ROUTER_PREF
1287 	pref = ra_msg->icmph.icmp6_router_pref;
1288 	/* 10b is handled as if it were 00b (medium) */
1289 	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1290 	    !in6_dev->cnf.accept_ra_rtr_pref)
1291 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1292 #endif
1293 	/* routes added from RAs do not use nexthop objects */
1294 	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1295 	if (rt) {
1296 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1297 					 rt->fib6_nh->fib_nh_dev, NULL,
1298 					  &ipv6_hdr(skb)->saddr);
1299 		if (!neigh) {
1300 			ND_PRINTK(0, err,
1301 				  "RA: %s got default router without neighbour\n",
1302 				  __func__);
1303 			fib6_info_release(rt);
1304 			return;
1305 		}
1306 	}
1307 	if (rt && lifetime == 0) {
1308 		ip6_del_rt(net, rt, false);
1309 		rt = NULL;
1310 	}
1311 
1312 	ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, for dev: %s\n",
1313 		  rt, lifetime, skb->dev->name);
1314 	if (!rt && lifetime) {
1315 		ND_PRINTK(3, info, "RA: adding default router\n");
1316 
1317 		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1318 					 skb->dev, pref);
1319 		if (!rt) {
1320 			ND_PRINTK(0, err,
1321 				  "RA: %s failed to add default route\n",
1322 				  __func__);
1323 			return;
1324 		}
1325 
1326 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1327 					 rt->fib6_nh->fib_nh_dev, NULL,
1328 					  &ipv6_hdr(skb)->saddr);
1329 		if (!neigh) {
1330 			ND_PRINTK(0, err,
1331 				  "RA: %s got default router without neighbour\n",
1332 				  __func__);
1333 			fib6_info_release(rt);
1334 			return;
1335 		}
1336 		neigh->flags |= NTF_ROUTER;
1337 	} else if (rt) {
1338 		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1339 	}
1340 
1341 	if (rt)
1342 		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1343 	if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1344 	    ra_msg->icmph.icmp6_hop_limit) {
1345 		if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1346 			in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1347 			fib6_metric_set(rt, RTAX_HOPLIMIT,
1348 					ra_msg->icmph.icmp6_hop_limit);
1349 		} else {
1350 			ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1351 		}
1352 	}
1353 
1354 skip_defrtr:
1355 
1356 	/*
1357 	 *	Update Reachable Time and Retrans Timer
1358 	 */
1359 
1360 	if (in6_dev->nd_parms) {
1361 		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1362 
1363 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1364 			rtime = (rtime*HZ)/1000;
1365 			if (rtime < HZ/100)
1366 				rtime = HZ/100;
1367 			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1368 			in6_dev->tstamp = jiffies;
1369 			send_ifinfo_notify = true;
1370 		}
1371 
1372 		rtime = ntohl(ra_msg->reachable_time);
1373 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1374 			rtime = (rtime*HZ)/1000;
1375 
1376 			if (rtime < HZ/10)
1377 				rtime = HZ/10;
1378 
1379 			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1380 				NEIGH_VAR_SET(in6_dev->nd_parms,
1381 					      BASE_REACHABLE_TIME, rtime);
1382 				NEIGH_VAR_SET(in6_dev->nd_parms,
1383 					      GC_STALETIME, 3 * rtime);
1384 				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1385 				in6_dev->tstamp = jiffies;
1386 				send_ifinfo_notify = true;
1387 			}
1388 		}
1389 	}
1390 
1391 	/*
1392 	 *	Send a notify if RA changed managed/otherconf flags or timer settings
1393 	 */
1394 	if (send_ifinfo_notify)
1395 		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1396 
1397 skip_linkparms:
1398 
1399 	/*
1400 	 *	Process options.
1401 	 */
1402 
1403 	if (!neigh)
1404 		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1405 				       skb->dev, 1);
1406 	if (neigh) {
1407 		u8 *lladdr = NULL;
1408 		if (ndopts.nd_opts_src_lladdr) {
1409 			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1410 						     skb->dev);
1411 			if (!lladdr) {
1412 				ND_PRINTK(2, warn,
1413 					  "RA: invalid link-layer address length\n");
1414 				goto out;
1415 			}
1416 		}
1417 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1418 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1419 			     NEIGH_UPDATE_F_OVERRIDE|
1420 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1421 			     NEIGH_UPDATE_F_ISROUTER,
1422 			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1423 	}
1424 
1425 	if (!ipv6_accept_ra(in6_dev)) {
1426 		ND_PRINTK(2, info,
1427 			  "RA: %s, accept_ra is false for dev: %s\n",
1428 			  __func__, skb->dev->name);
1429 		goto out;
1430 	}
1431 
1432 #ifdef CONFIG_IPV6_ROUTE_INFO
1433 	if (!in6_dev->cnf.accept_ra_from_local &&
1434 	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1435 			  in6_dev->dev, 0)) {
1436 		ND_PRINTK(2, info,
1437 			  "RA from local address detected on dev: %s: router info ignored.\n",
1438 			  skb->dev->name);
1439 		goto skip_routeinfo;
1440 	}
1441 
1442 	if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1443 		struct nd_opt_hdr *p;
1444 		for (p = ndopts.nd_opts_ri;
1445 		     p;
1446 		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1447 			struct route_info *ri = (struct route_info *)p;
1448 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1449 			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1450 			    ri->prefix_len == 0)
1451 				continue;
1452 #endif
1453 			if (ri->prefix_len == 0 &&
1454 			    !in6_dev->cnf.accept_ra_defrtr)
1455 				continue;
1456 			if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1457 				continue;
1458 			if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1459 				continue;
1460 			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1461 				      &ipv6_hdr(skb)->saddr);
1462 		}
1463 	}
1464 
1465 skip_routeinfo:
1466 #endif
1467 
1468 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1469 	/* skip link-specific ndopts from interior routers */
1470 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1471 		ND_PRINTK(2, info,
1472 			  "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1473 			  __func__, skb->dev->name);
1474 		goto out;
1475 	}
1476 #endif
1477 
1478 	if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1479 		struct nd_opt_hdr *p;
1480 		for (p = ndopts.nd_opts_pi;
1481 		     p;
1482 		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1483 			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1484 					    (p->nd_opt_len) << 3,
1485 					    ndopts.nd_opts_src_lladdr != NULL);
1486 		}
1487 	}
1488 
1489 	if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1490 		__be32 n;
1491 		u32 mtu;
1492 
1493 		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1494 		mtu = ntohl(n);
1495 
1496 		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1497 			ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1498 		} else if (in6_dev->cnf.mtu6 != mtu) {
1499 			in6_dev->cnf.mtu6 = mtu;
1500 			fib6_metric_set(rt, RTAX_MTU, mtu);
1501 			rt6_mtu_change(skb->dev, mtu);
1502 		}
1503 	}
1504 
1505 	if (ndopts.nd_useropts) {
1506 		struct nd_opt_hdr *p;
1507 		for (p = ndopts.nd_useropts;
1508 		     p;
1509 		     p = ndisc_next_useropt(skb->dev, p,
1510 					    ndopts.nd_useropts_end)) {
1511 			ndisc_ra_useropt(skb, p);
1512 		}
1513 	}
1514 
1515 	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1516 		ND_PRINTK(2, warn, "RA: invalid RA options\n");
1517 	}
1518 out:
1519 	fib6_info_release(rt);
1520 	if (neigh)
1521 		neigh_release(neigh);
1522 }
1523 
ndisc_redirect_rcv(struct sk_buff * skb)1524 static void ndisc_redirect_rcv(struct sk_buff *skb)
1525 {
1526 	u8 *hdr;
1527 	struct ndisc_options ndopts;
1528 	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1529 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1530 				    offsetof(struct rd_msg, opt));
1531 
1532 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1533 	switch (skb->ndisc_nodetype) {
1534 	case NDISC_NODETYPE_HOST:
1535 	case NDISC_NODETYPE_NODEFAULT:
1536 		ND_PRINTK(2, warn,
1537 			  "Redirect: from host or unauthorized router\n");
1538 		return;
1539 	}
1540 #endif
1541 
1542 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1543 		ND_PRINTK(2, warn,
1544 			  "Redirect: source address is not link-local\n");
1545 		return;
1546 	}
1547 
1548 	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1549 		return;
1550 
1551 	if (!ndopts.nd_opts_rh) {
1552 		ip6_redirect_no_header(skb, dev_net(skb->dev),
1553 					skb->dev->ifindex);
1554 		return;
1555 	}
1556 
1557 	hdr = (u8 *)ndopts.nd_opts_rh;
1558 	hdr += 8;
1559 	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1560 		return;
1561 
1562 	icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1563 }
1564 
ndisc_fill_redirect_hdr_option(struct sk_buff * skb,struct sk_buff * orig_skb,int rd_len)1565 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1566 					   struct sk_buff *orig_skb,
1567 					   int rd_len)
1568 {
1569 	u8 *opt = skb_put(skb, rd_len);
1570 
1571 	memset(opt, 0, 8);
1572 	*(opt++) = ND_OPT_REDIRECT_HDR;
1573 	*(opt++) = (rd_len >> 3);
1574 	opt += 6;
1575 
1576 	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1577 		      rd_len - 8);
1578 }
1579 
ndisc_send_redirect(struct sk_buff * skb,const struct in6_addr * target)1580 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1581 {
1582 	struct net_device *dev = skb->dev;
1583 	struct net *net = dev_net(dev);
1584 	struct sock *sk = net->ipv6.ndisc_sk;
1585 	int optlen = 0;
1586 	struct inet_peer *peer;
1587 	struct sk_buff *buff;
1588 	struct rd_msg *msg;
1589 	struct in6_addr saddr_buf;
1590 	struct rt6_info *rt;
1591 	struct dst_entry *dst;
1592 	struct flowi6 fl6;
1593 	int rd_len;
1594 	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1595 	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1596 	bool ret;
1597 
1598 	if (netif_is_l3_master(skb->dev)) {
1599 		dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1600 		if (!dev)
1601 			return;
1602 	}
1603 
1604 	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1605 		ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1606 			  dev->name);
1607 		return;
1608 	}
1609 
1610 	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1611 	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1612 		ND_PRINTK(2, warn,
1613 			  "Redirect: target address is not link-local unicast\n");
1614 		return;
1615 	}
1616 
1617 	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1618 			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1619 
1620 	dst = ip6_route_output(net, NULL, &fl6);
1621 	if (dst->error) {
1622 		dst_release(dst);
1623 		return;
1624 	}
1625 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1626 	if (IS_ERR(dst))
1627 		return;
1628 
1629 	rt = (struct rt6_info *) dst;
1630 
1631 	if (rt->rt6i_flags & RTF_GATEWAY) {
1632 		ND_PRINTK(2, warn,
1633 			  "Redirect: destination is not a neighbour\n");
1634 		goto release;
1635 	}
1636 	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1637 	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1638 	if (peer)
1639 		inet_putpeer(peer);
1640 	if (!ret)
1641 		goto release;
1642 
1643 	if (dev->addr_len) {
1644 		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1645 		if (!neigh) {
1646 			ND_PRINTK(2, warn,
1647 				  "Redirect: no neigh for target address\n");
1648 			goto release;
1649 		}
1650 
1651 		read_lock_bh(&neigh->lock);
1652 		if (neigh->nud_state & NUD_VALID) {
1653 			memcpy(ha_buf, neigh->ha, dev->addr_len);
1654 			read_unlock_bh(&neigh->lock);
1655 			ha = ha_buf;
1656 			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1657 								ops_data_buf,
1658 								&ops_data);
1659 		} else
1660 			read_unlock_bh(&neigh->lock);
1661 
1662 		neigh_release(neigh);
1663 	}
1664 
1665 	rd_len = min_t(unsigned int,
1666 		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1667 		       skb->len + 8);
1668 	rd_len &= ~0x7;
1669 	optlen += rd_len;
1670 
1671 	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1672 	if (!buff)
1673 		goto release;
1674 
1675 	msg = skb_put(buff, sizeof(*msg));
1676 	*msg = (struct rd_msg) {
1677 		.icmph = {
1678 			.icmp6_type = NDISC_REDIRECT,
1679 		},
1680 		.target = *target,
1681 		.dest = ipv6_hdr(skb)->daddr,
1682 	};
1683 
1684 	/*
1685 	 *	include target_address option
1686 	 */
1687 
1688 	if (ha)
1689 		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1690 
1691 	/*
1692 	 *	build redirect option and copy skb over to the new packet.
1693 	 */
1694 
1695 	if (rd_len)
1696 		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1697 
1698 	skb_dst_set(buff, dst);
1699 	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1700 	return;
1701 
1702 release:
1703 	dst_release(dst);
1704 }
1705 
pndisc_redo(struct sk_buff * skb)1706 static void pndisc_redo(struct sk_buff *skb)
1707 {
1708 	ndisc_recv_ns(skb);
1709 	kfree_skb(skb);
1710 }
1711 
ndisc_is_multicast(const void * pkey)1712 static int ndisc_is_multicast(const void *pkey)
1713 {
1714 	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1715 }
1716 
ndisc_suppress_frag_ndisc(struct sk_buff * skb)1717 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1718 {
1719 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1720 
1721 	if (!idev)
1722 		return true;
1723 	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1724 	    idev->cnf.suppress_frag_ndisc) {
1725 		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1726 		return true;
1727 	}
1728 	return false;
1729 }
1730 
ndisc_rcv(struct sk_buff * skb)1731 int ndisc_rcv(struct sk_buff *skb)
1732 {
1733 	struct nd_msg *msg;
1734 
1735 	if (ndisc_suppress_frag_ndisc(skb))
1736 		return 0;
1737 
1738 	if (skb_linearize(skb))
1739 		return 0;
1740 
1741 	msg = (struct nd_msg *)skb_transport_header(skb);
1742 
1743 	__skb_push(skb, skb->data - skb_transport_header(skb));
1744 
1745 	if (ipv6_hdr(skb)->hop_limit != 255) {
1746 		ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1747 			  ipv6_hdr(skb)->hop_limit);
1748 		return 0;
1749 	}
1750 
1751 	if (msg->icmph.icmp6_code != 0) {
1752 		ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1753 			  msg->icmph.icmp6_code);
1754 		return 0;
1755 	}
1756 
1757 	switch (msg->icmph.icmp6_type) {
1758 	case NDISC_NEIGHBOUR_SOLICITATION:
1759 		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1760 		ndisc_recv_ns(skb);
1761 		break;
1762 
1763 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1764 		ndisc_recv_na(skb);
1765 		break;
1766 
1767 	case NDISC_ROUTER_SOLICITATION:
1768 		ndisc_recv_rs(skb);
1769 		break;
1770 
1771 	case NDISC_ROUTER_ADVERTISEMENT:
1772 		ndisc_router_discovery(skb);
1773 		break;
1774 
1775 	case NDISC_REDIRECT:
1776 		ndisc_redirect_rcv(skb);
1777 		break;
1778 	}
1779 
1780 	return 0;
1781 }
1782 
ndisc_netdev_event(struct notifier_block * this,unsigned long event,void * ptr)1783 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1784 {
1785 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1786 	struct netdev_notifier_change_info *change_info;
1787 	struct net *net = dev_net(dev);
1788 	struct inet6_dev *idev;
1789 
1790 	switch (event) {
1791 	case NETDEV_CHANGEADDR:
1792 		neigh_changeaddr(&nd_tbl, dev);
1793 		fib6_run_gc(0, net, false);
1794 		fallthrough;
1795 	case NETDEV_UP:
1796 		idev = in6_dev_get(dev);
1797 		if (!idev)
1798 			break;
1799 		if (idev->cnf.ndisc_notify ||
1800 		    net->ipv6.devconf_all->ndisc_notify)
1801 			ndisc_send_unsol_na(dev);
1802 		in6_dev_put(idev);
1803 		break;
1804 	case NETDEV_CHANGE:
1805 		change_info = ptr;
1806 		if (change_info->flags_changed & IFF_NOARP)
1807 			neigh_changeaddr(&nd_tbl, dev);
1808 		if (!netif_carrier_ok(dev))
1809 			neigh_carrier_down(&nd_tbl, dev);
1810 		break;
1811 	case NETDEV_DOWN:
1812 		neigh_ifdown(&nd_tbl, dev);
1813 		fib6_run_gc(0, net, false);
1814 		break;
1815 	case NETDEV_NOTIFY_PEERS:
1816 		ndisc_send_unsol_na(dev);
1817 		break;
1818 	default:
1819 		break;
1820 	}
1821 
1822 	return NOTIFY_DONE;
1823 }
1824 
1825 static struct notifier_block ndisc_netdev_notifier = {
1826 	.notifier_call = ndisc_netdev_event,
1827 	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1828 };
1829 
1830 #ifdef CONFIG_SYSCTL
ndisc_warn_deprecated_sysctl(struct ctl_table * ctl,const char * func,const char * dev_name)1831 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1832 					 const char *func, const char *dev_name)
1833 {
1834 	static char warncomm[TASK_COMM_LEN];
1835 	static int warned;
1836 	if (strcmp(warncomm, current->comm) && warned < 5) {
1837 		strcpy(warncomm, current->comm);
1838 		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1839 			warncomm, func,
1840 			dev_name, ctl->procname,
1841 			dev_name, ctl->procname);
1842 		warned++;
1843 	}
1844 }
1845 
ndisc_ifinfo_sysctl_change(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)1846 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1847 		size_t *lenp, loff_t *ppos)
1848 {
1849 	struct net_device *dev = ctl->extra1;
1850 	struct inet6_dev *idev;
1851 	int ret;
1852 
1853 	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1854 	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1855 		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1856 
1857 	if (strcmp(ctl->procname, "retrans_time") == 0)
1858 		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1859 
1860 	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1861 		ret = neigh_proc_dointvec_jiffies(ctl, write,
1862 						  buffer, lenp, ppos);
1863 
1864 	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1865 		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1866 		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1867 						     buffer, lenp, ppos);
1868 	else
1869 		ret = -1;
1870 
1871 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1872 		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1873 			idev->nd_parms->reachable_time =
1874 					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1875 		idev->tstamp = jiffies;
1876 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1877 		in6_dev_put(idev);
1878 	}
1879 	return ret;
1880 }
1881 
1882 
1883 #endif
1884 
ndisc_net_init(struct net * net)1885 static int __net_init ndisc_net_init(struct net *net)
1886 {
1887 	struct ipv6_pinfo *np;
1888 	struct sock *sk;
1889 	int err;
1890 
1891 	err = inet_ctl_sock_create(&sk, PF_INET6,
1892 				   SOCK_RAW, IPPROTO_ICMPV6, net);
1893 	if (err < 0) {
1894 		ND_PRINTK(0, err,
1895 			  "NDISC: Failed to initialize the control socket (err %d)\n",
1896 			  err);
1897 		return err;
1898 	}
1899 
1900 	net->ipv6.ndisc_sk = sk;
1901 
1902 	np = inet6_sk(sk);
1903 	np->hop_limit = 255;
1904 	/* Do not loopback ndisc messages */
1905 	np->mc_loop = 0;
1906 
1907 	return 0;
1908 }
1909 
ndisc_net_exit(struct net * net)1910 static void __net_exit ndisc_net_exit(struct net *net)
1911 {
1912 	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1913 }
1914 
1915 static struct pernet_operations ndisc_net_ops = {
1916 	.init = ndisc_net_init,
1917 	.exit = ndisc_net_exit,
1918 };
1919 
ndisc_init(void)1920 int __init ndisc_init(void)
1921 {
1922 	int err;
1923 
1924 	err = register_pernet_subsys(&ndisc_net_ops);
1925 	if (err)
1926 		return err;
1927 	/*
1928 	 * Initialize the neighbour table
1929 	 */
1930 	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1931 
1932 #ifdef CONFIG_SYSCTL
1933 	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1934 				    ndisc_ifinfo_sysctl_change);
1935 	if (err)
1936 		goto out_unregister_pernet;
1937 out:
1938 #endif
1939 	return err;
1940 
1941 #ifdef CONFIG_SYSCTL
1942 out_unregister_pernet:
1943 	unregister_pernet_subsys(&ndisc_net_ops);
1944 	goto out;
1945 #endif
1946 }
1947 
ndisc_late_init(void)1948 int __init ndisc_late_init(void)
1949 {
1950 	return register_netdevice_notifier(&ndisc_netdev_notifier);
1951 }
1952 
ndisc_late_cleanup(void)1953 void ndisc_late_cleanup(void)
1954 {
1955 	unregister_netdevice_notifier(&ndisc_netdev_notifier);
1956 }
1957 
ndisc_cleanup(void)1958 void ndisc_cleanup(void)
1959 {
1960 #ifdef CONFIG_SYSCTL
1961 	neigh_sysctl_unregister(&nd_tbl.parms);
1962 #endif
1963 	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1964 	unregister_pernet_subsys(&ndisc_net_ops);
1965 }
1966