• 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 sk_buff *skb;
419 
420 	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
421 	if (!skb)
422 		return NULL;
423 
424 	skb->protocol = htons(ETH_P_IPV6);
425 	skb->dev = dev;
426 
427 	skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
428 	skb_reset_transport_header(skb);
429 
430 	/* Manually assign socket ownership as we avoid calling
431 	 * sock_alloc_send_pskb() to bypass wmem buffer limits
432 	 */
433 	rcu_read_lock();
434 	skb_set_owner_w(skb, dev_net_rcu(dev)->ipv6.ndisc_sk);
435 	rcu_read_unlock();
436 
437 	return skb;
438 }
439 
ip6_nd_hdr(struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,int hop_limit,int len)440 static void ip6_nd_hdr(struct sk_buff *skb,
441 		       const struct in6_addr *saddr,
442 		       const struct in6_addr *daddr,
443 		       int hop_limit, int len)
444 {
445 	struct ipv6hdr *hdr;
446 	struct inet6_dev *idev;
447 	unsigned tclass;
448 
449 	rcu_read_lock();
450 	idev = __in6_dev_get(skb->dev);
451 	tclass = idev ? idev->cnf.ndisc_tclass : 0;
452 	rcu_read_unlock();
453 
454 	skb_push(skb, sizeof(*hdr));
455 	skb_reset_network_header(skb);
456 	hdr = ipv6_hdr(skb);
457 
458 	ip6_flow_hdr(hdr, tclass, 0);
459 
460 	hdr->payload_len = htons(len);
461 	hdr->nexthdr = IPPROTO_ICMPV6;
462 	hdr->hop_limit = hop_limit;
463 
464 	hdr->saddr = *saddr;
465 	hdr->daddr = *daddr;
466 }
467 
ndisc_send_skb(struct sk_buff * skb,const struct in6_addr * daddr,const struct in6_addr * saddr)468 static void ndisc_send_skb(struct sk_buff *skb,
469 			   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 inet6_dev *idev;
475 	struct net *net;
476 	struct sock *sk;
477 	int err;
478 	u8 type;
479 
480 	type = icmp6h->icmp6_type;
481 
482 	rcu_read_lock();
483 
484 	net = dev_net_rcu(skb->dev);
485 	sk = net->ipv6.ndisc_sk;
486 	if (!dst) {
487 		struct flowi6 fl6;
488 		int oif = skb->dev->ifindex;
489 
490 		icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
491 		dst = icmp6_dst_alloc(skb->dev, &fl6);
492 		if (IS_ERR(dst)) {
493 			rcu_read_unlock();
494 			kfree_skb(skb);
495 			return;
496 		}
497 
498 		skb_dst_set(skb, dst);
499 	}
500 
501 	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
502 					      IPPROTO_ICMPV6,
503 					      csum_partial(icmp6h,
504 							   skb->len, 0));
505 
506 	ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
507 
508 	idev = __in6_dev_get(dst->dev);
509 	IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
510 
511 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
512 		      net, sk, skb, NULL, dst->dev,
513 		      dst_output);
514 	if (!err) {
515 		ICMP6MSGOUT_INC_STATS(net, idev, type);
516 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
517 	}
518 
519 	rcu_read_unlock();
520 }
521 
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)522 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
523 		   const struct in6_addr *solicited_addr,
524 		   bool router, bool solicited, bool override, bool inc_opt)
525 {
526 	struct sk_buff *skb;
527 	struct in6_addr tmpaddr;
528 	struct inet6_ifaddr *ifp;
529 	const struct in6_addr *src_addr;
530 	struct nd_msg *msg;
531 	int optlen = 0;
532 
533 	/* for anycast or proxy, solicited_addr != src_addr */
534 	ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
535 	if (ifp) {
536 		src_addr = solicited_addr;
537 		if (ifp->flags & IFA_F_OPTIMISTIC)
538 			override = false;
539 		inc_opt |= ifp->idev->cnf.force_tllao;
540 		in6_ifa_put(ifp);
541 	} else {
542 		if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
543 				       inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
544 				       &tmpaddr))
545 			return;
546 		src_addr = &tmpaddr;
547 	}
548 
549 	if (!dev->addr_len)
550 		inc_opt = false;
551 	if (inc_opt)
552 		optlen += ndisc_opt_addr_space(dev,
553 					       NDISC_NEIGHBOUR_ADVERTISEMENT);
554 
555 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
556 	if (!skb)
557 		return;
558 
559 	msg = skb_put(skb, sizeof(*msg));
560 	*msg = (struct nd_msg) {
561 		.icmph = {
562 			.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
563 			.icmp6_router = router,
564 			.icmp6_solicited = solicited,
565 			.icmp6_override = override,
566 		},
567 		.target = *solicited_addr,
568 	};
569 
570 	if (inc_opt)
571 		ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
572 				       dev->dev_addr,
573 				       NDISC_NEIGHBOUR_ADVERTISEMENT);
574 
575 	ndisc_send_skb(skb, daddr, src_addr);
576 }
577 
ndisc_send_unsol_na(struct net_device * dev)578 static void ndisc_send_unsol_na(struct net_device *dev)
579 {
580 	struct inet6_dev *idev;
581 	struct inet6_ifaddr *ifa;
582 
583 	idev = in6_dev_get(dev);
584 	if (!idev)
585 		return;
586 
587 	read_lock_bh(&idev->lock);
588 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
589 		/* skip tentative addresses until dad completes */
590 		if (ifa->flags & IFA_F_TENTATIVE &&
591 		    !(ifa->flags & IFA_F_OPTIMISTIC))
592 			continue;
593 
594 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
595 			      /*router=*/ !!idev->cnf.forwarding,
596 			      /*solicited=*/ false, /*override=*/ true,
597 			      /*inc_opt=*/ true);
598 	}
599 	read_unlock_bh(&idev->lock);
600 
601 	in6_dev_put(idev);
602 }
603 
ndisc_send_ns(struct net_device * dev,const struct in6_addr * solicit,const struct in6_addr * daddr,const struct in6_addr * saddr,u64 nonce)604 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
605 		   const struct in6_addr *daddr, const struct in6_addr *saddr,
606 		   u64 nonce)
607 {
608 	struct sk_buff *skb;
609 	struct in6_addr addr_buf;
610 	int inc_opt = dev->addr_len;
611 	int optlen = 0;
612 	struct nd_msg *msg;
613 
614 	if (!saddr) {
615 		if (ipv6_get_lladdr(dev, &addr_buf,
616 				   (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
617 			return;
618 		saddr = &addr_buf;
619 	}
620 
621 	if (ipv6_addr_any(saddr))
622 		inc_opt = false;
623 	if (inc_opt)
624 		optlen += ndisc_opt_addr_space(dev,
625 					       NDISC_NEIGHBOUR_SOLICITATION);
626 	if (nonce != 0)
627 		optlen += 8;
628 
629 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
630 	if (!skb)
631 		return;
632 
633 	msg = skb_put(skb, sizeof(*msg));
634 	*msg = (struct nd_msg) {
635 		.icmph = {
636 			.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
637 		},
638 		.target = *solicit,
639 	};
640 
641 	if (inc_opt)
642 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
643 				       dev->dev_addr,
644 				       NDISC_NEIGHBOUR_SOLICITATION);
645 	if (nonce != 0) {
646 		u8 *opt = skb_put(skb, 8);
647 
648 		opt[0] = ND_OPT_NONCE;
649 		opt[1] = 8 >> 3;
650 		memcpy(opt + 2, &nonce, 6);
651 	}
652 
653 	ndisc_send_skb(skb, daddr, saddr);
654 }
655 
ndisc_send_rs(struct net_device * dev,const struct in6_addr * saddr,const struct in6_addr * daddr)656 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
657 		   const struct in6_addr *daddr)
658 {
659 	struct sk_buff *skb;
660 	struct rs_msg *msg;
661 	int send_sllao = dev->addr_len;
662 	int optlen = 0;
663 
664 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
665 	/*
666 	 * According to section 2.2 of RFC 4429, we must not
667 	 * send router solicitations with a sllao from
668 	 * optimistic addresses, but we may send the solicitation
669 	 * if we don't include the sllao.  So here we check
670 	 * if our address is optimistic, and if so, we
671 	 * suppress the inclusion of the sllao.
672 	 */
673 	if (send_sllao) {
674 		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
675 							   dev, 1);
676 		if (ifp) {
677 			if (ifp->flags & IFA_F_OPTIMISTIC)  {
678 				send_sllao = 0;
679 			}
680 			in6_ifa_put(ifp);
681 		} else {
682 			send_sllao = 0;
683 		}
684 	}
685 #endif
686 	if (send_sllao)
687 		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
688 
689 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
690 	if (!skb)
691 		return;
692 
693 	msg = skb_put(skb, sizeof(*msg));
694 	*msg = (struct rs_msg) {
695 		.icmph = {
696 			.icmp6_type = NDISC_ROUTER_SOLICITATION,
697 		},
698 	};
699 
700 	if (send_sllao)
701 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
702 				       dev->dev_addr,
703 				       NDISC_ROUTER_SOLICITATION);
704 
705 	ndisc_send_skb(skb, daddr, saddr);
706 }
707 
708 
ndisc_error_report(struct neighbour * neigh,struct sk_buff * skb)709 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
710 {
711 	/*
712 	 *	"The sender MUST return an ICMP
713 	 *	 destination unreachable"
714 	 */
715 	dst_link_failure(skb);
716 	kfree_skb(skb);
717 }
718 
719 /* Called with locked neigh: either read or both */
720 
ndisc_solicit(struct neighbour * neigh,struct sk_buff * skb)721 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
722 {
723 	struct in6_addr *saddr = NULL;
724 	struct in6_addr mcaddr;
725 	struct net_device *dev = neigh->dev;
726 	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
727 	int probes = atomic_read(&neigh->probes);
728 
729 	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
730 					   dev, false, 1,
731 					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
732 		saddr = &ipv6_hdr(skb)->saddr;
733 	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
734 	if (probes < 0) {
735 		if (!(neigh->nud_state & NUD_VALID)) {
736 			ND_PRINTK(1, dbg,
737 				  "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
738 				  __func__, target);
739 		}
740 		ndisc_send_ns(dev, target, target, saddr, 0);
741 	} else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
742 		neigh_app_ns(neigh);
743 	} else {
744 		addrconf_addr_solict_mult(target, &mcaddr);
745 		ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
746 	}
747 }
748 
pndisc_is_router(const void * pkey,struct net_device * dev)749 static int pndisc_is_router(const void *pkey,
750 			    struct net_device *dev)
751 {
752 	struct pneigh_entry *n;
753 	int ret = -1;
754 
755 	read_lock_bh(&nd_tbl.lock);
756 	n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
757 	if (n)
758 		ret = !!(n->flags & NTF_ROUTER);
759 	read_unlock_bh(&nd_tbl.lock);
760 
761 	return ret;
762 }
763 
ndisc_update(const struct net_device * dev,struct neighbour * neigh,const u8 * lladdr,u8 new,u32 flags,u8 icmp6_type,struct ndisc_options * ndopts)764 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
765 		  const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
766 		  struct ndisc_options *ndopts)
767 {
768 	neigh_update(neigh, lladdr, new, flags, 0);
769 	/* report ndisc ops about neighbour update */
770 	ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
771 }
772 
ndisc_recv_ns(struct sk_buff * skb)773 static void ndisc_recv_ns(struct sk_buff *skb)
774 {
775 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
776 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
777 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
778 	u8 *lladdr = NULL;
779 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
780 				    offsetof(struct nd_msg, opt));
781 	struct ndisc_options ndopts;
782 	struct net_device *dev = skb->dev;
783 	struct inet6_ifaddr *ifp;
784 	struct inet6_dev *idev = NULL;
785 	struct neighbour *neigh;
786 	int dad = ipv6_addr_any(saddr);
787 	bool inc;
788 	int is_router = -1;
789 	u64 nonce = 0;
790 
791 	if (skb->len < sizeof(struct nd_msg)) {
792 		ND_PRINTK(2, warn, "NS: packet too short\n");
793 		return;
794 	}
795 
796 	if (ipv6_addr_is_multicast(&msg->target)) {
797 		ND_PRINTK(2, warn, "NS: multicast target address\n");
798 		return;
799 	}
800 
801 	/*
802 	 * RFC2461 7.1.1:
803 	 * DAD has to be destined for solicited node multicast address.
804 	 */
805 	if (dad && !ipv6_addr_is_solict_mult(daddr)) {
806 		ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
807 		return;
808 	}
809 
810 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
811 		ND_PRINTK(2, warn, "NS: invalid ND options\n");
812 		return;
813 	}
814 
815 	if (ndopts.nd_opts_src_lladdr) {
816 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
817 		if (!lladdr) {
818 			ND_PRINTK(2, warn,
819 				  "NS: invalid link-layer address length\n");
820 			return;
821 		}
822 
823 		/* RFC2461 7.1.1:
824 		 *	If the IP source address is the unspecified address,
825 		 *	there MUST NOT be source link-layer address option
826 		 *	in the message.
827 		 */
828 		if (dad) {
829 			ND_PRINTK(2, warn,
830 				  "NS: bad DAD packet (link-layer address option)\n");
831 			return;
832 		}
833 	}
834 	if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
835 		memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
836 
837 	inc = ipv6_addr_is_multicast(daddr);
838 
839 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
840 	if (ifp) {
841 have_ifp:
842 		if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
843 			if (dad) {
844 				if (nonce != 0 && ifp->dad_nonce == nonce) {
845 					u8 *np = (u8 *)&nonce;
846 					/* Matching nonce if looped back */
847 					ND_PRINTK(2, notice,
848 						  "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
849 						  ifp->idev->dev->name,
850 						  &ifp->addr, np);
851 					goto out;
852 				}
853 				/*
854 				 * We are colliding with another node
855 				 * who is doing DAD
856 				 * so fail our DAD process
857 				 */
858 				addrconf_dad_failure(skb, ifp);
859 				return;
860 			} else {
861 				/*
862 				 * This is not a dad solicitation.
863 				 * If we are an optimistic node,
864 				 * we should respond.
865 				 * Otherwise, we should ignore it.
866 				 */
867 				if (!(ifp->flags & IFA_F_OPTIMISTIC))
868 					goto out;
869 			}
870 		}
871 
872 		idev = ifp->idev;
873 	} else {
874 		struct net *net = dev_net(dev);
875 
876 		/* perhaps an address on the master device */
877 		if (netif_is_l3_slave(dev)) {
878 			struct net_device *mdev;
879 
880 			mdev = netdev_master_upper_dev_get_rcu(dev);
881 			if (mdev) {
882 				ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
883 				if (ifp)
884 					goto have_ifp;
885 			}
886 		}
887 
888 		idev = in6_dev_get(dev);
889 		if (!idev) {
890 			/* XXX: count this drop? */
891 			return;
892 		}
893 
894 		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
895 		    (idev->cnf.forwarding &&
896 		     (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
897 		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
898 			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
899 			    skb->pkt_type != PACKET_HOST &&
900 			    inc &&
901 			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
902 				/*
903 				 * for anycast or proxy,
904 				 * sender should delay its response
905 				 * by a random time between 0 and
906 				 * MAX_ANYCAST_DELAY_TIME seconds.
907 				 * (RFC2461) -- yoshfuji
908 				 */
909 				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
910 				if (n)
911 					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
912 				goto out;
913 			}
914 		} else
915 			goto out;
916 	}
917 
918 	if (is_router < 0)
919 		is_router = idev->cnf.forwarding;
920 
921 	if (dad) {
922 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
923 			      !!is_router, false, (ifp != NULL), true);
924 		goto out;
925 	}
926 
927 	if (inc)
928 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
929 	else
930 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
931 
932 	/*
933 	 *	update / create cache entry
934 	 *	for the source address
935 	 */
936 	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
937 			       !inc || lladdr || !dev->addr_len);
938 	if (neigh)
939 		ndisc_update(dev, neigh, lladdr, NUD_STALE,
940 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
941 			     NEIGH_UPDATE_F_OVERRIDE,
942 			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
943 	if (neigh || !dev->header_ops) {
944 		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
945 			      true, (ifp != NULL && inc), inc);
946 		if (neigh)
947 			neigh_release(neigh);
948 	}
949 
950 out:
951 	if (ifp)
952 		in6_ifa_put(ifp);
953 	else
954 		in6_dev_put(idev);
955 }
956 
ndisc_recv_na(struct sk_buff * skb)957 static void ndisc_recv_na(struct sk_buff *skb)
958 {
959 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
960 	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
961 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
962 	u8 *lladdr = NULL;
963 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
964 				    offsetof(struct nd_msg, opt));
965 	struct ndisc_options ndopts;
966 	struct net_device *dev = skb->dev;
967 	struct inet6_dev *idev = __in6_dev_get(dev);
968 	struct inet6_ifaddr *ifp;
969 	struct neighbour *neigh;
970 
971 	if (skb->len < sizeof(struct nd_msg)) {
972 		ND_PRINTK(2, warn, "NA: packet too short\n");
973 		return;
974 	}
975 
976 	if (ipv6_addr_is_multicast(&msg->target)) {
977 		ND_PRINTK(2, warn, "NA: target address is multicast\n");
978 		return;
979 	}
980 
981 	if (ipv6_addr_is_multicast(daddr) &&
982 	    msg->icmph.icmp6_solicited) {
983 		ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
984 		return;
985 	}
986 
987 	/* For some 802.11 wireless deployments (and possibly other networks),
988 	 * there will be a NA proxy and unsolicitd packets are attacks
989 	 * and thus should not be accepted.
990 	 */
991 	if (!msg->icmph.icmp6_solicited && idev &&
992 	    idev->cnf.drop_unsolicited_na)
993 		return;
994 
995 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
996 		ND_PRINTK(2, warn, "NS: invalid ND option\n");
997 		return;
998 	}
999 	if (ndopts.nd_opts_tgt_lladdr) {
1000 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1001 		if (!lladdr) {
1002 			ND_PRINTK(2, warn,
1003 				  "NA: invalid link-layer address length\n");
1004 			return;
1005 		}
1006 	}
1007 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1008 	if (ifp) {
1009 		if (skb->pkt_type != PACKET_LOOPBACK
1010 		    && (ifp->flags & IFA_F_TENTATIVE)) {
1011 				addrconf_dad_failure(skb, ifp);
1012 				return;
1013 		}
1014 		/* What should we make now? The advertisement
1015 		   is invalid, but ndisc specs say nothing
1016 		   about it. It could be misconfiguration, or
1017 		   an smart proxy agent tries to help us :-)
1018 
1019 		   We should not print the error if NA has been
1020 		   received from loopback - it is just our own
1021 		   unsolicited advertisement.
1022 		 */
1023 		if (skb->pkt_type != PACKET_LOOPBACK)
1024 			ND_PRINTK(1, warn,
1025 				  "NA: %pM advertised our address %pI6c on %s!\n",
1026 				  eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1027 		in6_ifa_put(ifp);
1028 		return;
1029 	}
1030 	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1031 
1032 	if (neigh) {
1033 		u8 old_flags = neigh->flags;
1034 		struct net *net = dev_net(dev);
1035 
1036 		if (neigh->nud_state & NUD_FAILED)
1037 			goto out;
1038 
1039 		/*
1040 		 * Don't update the neighbor cache entry on a proxy NA from
1041 		 * ourselves because either the proxied node is off link or it
1042 		 * has already sent a NA to us.
1043 		 */
1044 		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1045 		    net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1046 		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1047 			/* XXX: idev->cnf.proxy_ndp */
1048 			goto out;
1049 		}
1050 
1051 		ndisc_update(dev, neigh, lladdr,
1052 			     msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1053 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1054 			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1055 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1056 			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1057 			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1058 
1059 		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1060 			/*
1061 			 * Change: router to host
1062 			 */
1063 			rt6_clean_tohost(dev_net(dev),  saddr);
1064 		}
1065 
1066 out:
1067 		neigh_release(neigh);
1068 	}
1069 }
1070 
ndisc_recv_rs(struct sk_buff * skb)1071 static void ndisc_recv_rs(struct sk_buff *skb)
1072 {
1073 	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1074 	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1075 	struct neighbour *neigh;
1076 	struct inet6_dev *idev;
1077 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1078 	struct ndisc_options ndopts;
1079 	u8 *lladdr = NULL;
1080 
1081 	if (skb->len < sizeof(*rs_msg))
1082 		return;
1083 
1084 	idev = __in6_dev_get(skb->dev);
1085 	if (!idev) {
1086 		ND_PRINTK(1, err, "RS: can't find in6 device\n");
1087 		return;
1088 	}
1089 
1090 	/* Don't accept RS if we're not in router mode */
1091 	if (!idev->cnf.forwarding)
1092 		goto out;
1093 
1094 	/*
1095 	 * Don't update NCE if src = ::;
1096 	 * this implies that the source node has no ip address assigned yet.
1097 	 */
1098 	if (ipv6_addr_any(saddr))
1099 		goto out;
1100 
1101 	/* Parse ND options */
1102 	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1103 		ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1104 		goto out;
1105 	}
1106 
1107 	if (ndopts.nd_opts_src_lladdr) {
1108 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1109 					     skb->dev);
1110 		if (!lladdr)
1111 			goto out;
1112 	}
1113 
1114 	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1115 	if (neigh) {
1116 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1117 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1118 			     NEIGH_UPDATE_F_OVERRIDE|
1119 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1120 			     NDISC_ROUTER_SOLICITATION, &ndopts);
1121 		neigh_release(neigh);
1122 	}
1123 out:
1124 	return;
1125 }
1126 
ndisc_ra_useropt(struct sk_buff * ra,struct nd_opt_hdr * opt)1127 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1128 {
1129 	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1130 	struct sk_buff *skb;
1131 	struct nlmsghdr *nlh;
1132 	struct nduseroptmsg *ndmsg;
1133 	struct net *net = dev_net(ra->dev);
1134 	int err;
1135 	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1136 				    + (opt->nd_opt_len << 3));
1137 	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1138 
1139 	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1140 	if (!skb) {
1141 		err = -ENOBUFS;
1142 		goto errout;
1143 	}
1144 
1145 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1146 	if (!nlh) {
1147 		goto nla_put_failure;
1148 	}
1149 
1150 	ndmsg = nlmsg_data(nlh);
1151 	ndmsg->nduseropt_family = AF_INET6;
1152 	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1153 	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1154 	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1155 	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1156 
1157 	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1158 
1159 	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1160 		goto nla_put_failure;
1161 	nlmsg_end(skb, nlh);
1162 
1163 	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1164 	return;
1165 
1166 nla_put_failure:
1167 	nlmsg_free(skb);
1168 	err = -EMSGSIZE;
1169 errout:
1170 	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1171 }
1172 
ndisc_router_discovery(struct sk_buff * skb)1173 static void ndisc_router_discovery(struct sk_buff *skb)
1174 {
1175 	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1176 	struct neighbour *neigh = NULL;
1177 	struct inet6_dev *in6_dev;
1178 	struct fib6_info *rt = NULL;
1179 	struct net *net;
1180 	int lifetime;
1181 	struct ndisc_options ndopts;
1182 	int optlen;
1183 	unsigned int pref = 0;
1184 	__u32 old_if_flags;
1185 	bool send_ifinfo_notify = false;
1186 
1187 	__u8 *opt = (__u8 *)(ra_msg + 1);
1188 
1189 	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1190 		sizeof(struct ra_msg);
1191 
1192 	ND_PRINTK(2, info,
1193 		  "RA: %s, dev: %s\n",
1194 		  __func__, skb->dev->name);
1195 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1196 		ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1197 		return;
1198 	}
1199 	if (optlen < 0) {
1200 		ND_PRINTK(2, warn, "RA: packet too short\n");
1201 		return;
1202 	}
1203 
1204 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1205 	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1206 		ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1207 		return;
1208 	}
1209 #endif
1210 
1211 	/*
1212 	 *	set the RA_RECV flag in the interface
1213 	 */
1214 
1215 	in6_dev = __in6_dev_get(skb->dev);
1216 	if (!in6_dev) {
1217 		ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1218 			  skb->dev->name);
1219 		return;
1220 	}
1221 
1222 	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1223 		ND_PRINTK(2, warn, "RA: invalid ND options\n");
1224 		return;
1225 	}
1226 
1227 	if (!ipv6_accept_ra(in6_dev)) {
1228 		ND_PRINTK(2, info,
1229 			  "RA: %s, did not accept ra for dev: %s\n",
1230 			  __func__, skb->dev->name);
1231 		goto skip_linkparms;
1232 	}
1233 
1234 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1235 	/* skip link-specific parameters from interior routers */
1236 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1237 		ND_PRINTK(2, info,
1238 			  "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1239 			  __func__, skb->dev->name);
1240 		goto skip_linkparms;
1241 	}
1242 #endif
1243 
1244 	if (in6_dev->if_flags & IF_RS_SENT) {
1245 		/*
1246 		 *	flag that an RA was received after an RS was sent
1247 		 *	out on this interface.
1248 		 */
1249 		in6_dev->if_flags |= IF_RA_RCVD;
1250 	}
1251 
1252 	/*
1253 	 * Remember the managed/otherconf flags from most recently
1254 	 * received RA message (RFC 2462) -- yoshfuji
1255 	 */
1256 	old_if_flags = in6_dev->if_flags;
1257 	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1258 				IF_RA_OTHERCONF)) |
1259 				(ra_msg->icmph.icmp6_addrconf_managed ?
1260 					IF_RA_MANAGED : 0) |
1261 				(ra_msg->icmph.icmp6_addrconf_other ?
1262 					IF_RA_OTHERCONF : 0);
1263 
1264 	if (old_if_flags != in6_dev->if_flags)
1265 		send_ifinfo_notify = true;
1266 
1267 	if (!in6_dev->cnf.accept_ra_defrtr) {
1268 		ND_PRINTK(2, info,
1269 			  "RA: %s, defrtr is false for dev: %s\n",
1270 			  __func__, skb->dev->name);
1271 		goto skip_defrtr;
1272 	}
1273 
1274 	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1275 	if (lifetime != 0 && lifetime < in6_dev->cnf.accept_ra_min_lft) {
1276 		ND_PRINTK(2, info,
1277 			  "RA: router lifetime (%ds) is too short: %s\n",
1278 			  lifetime, skb->dev->name);
1279 		goto skip_defrtr;
1280 	}
1281 
1282 	/* Do not accept RA with source-addr found on local machine unless
1283 	 * accept_ra_from_local is set to true.
1284 	 */
1285 	net = dev_net(in6_dev->dev);
1286 	if (!in6_dev->cnf.accept_ra_from_local &&
1287 	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1288 		ND_PRINTK(2, info,
1289 			  "RA from local address detected on dev: %s: default router ignored\n",
1290 			  skb->dev->name);
1291 		goto skip_defrtr;
1292 	}
1293 
1294 #ifdef CONFIG_IPV6_ROUTER_PREF
1295 	pref = ra_msg->icmph.icmp6_router_pref;
1296 	/* 10b is handled as if it were 00b (medium) */
1297 	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1298 	    !in6_dev->cnf.accept_ra_rtr_pref)
1299 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1300 #endif
1301 	/* routes added from RAs do not use nexthop objects */
1302 	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1303 	if (rt) {
1304 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1305 					 rt->fib6_nh->fib_nh_dev, NULL,
1306 					  &ipv6_hdr(skb)->saddr);
1307 		if (!neigh) {
1308 			ND_PRINTK(0, err,
1309 				  "RA: %s got default router without neighbour\n",
1310 				  __func__);
1311 			fib6_info_release(rt);
1312 			return;
1313 		}
1314 	}
1315 	if (rt && lifetime == 0) {
1316 		ip6_del_rt(net, rt, false);
1317 		rt = NULL;
1318 	}
1319 
1320 	ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, for dev: %s\n",
1321 		  rt, lifetime, skb->dev->name);
1322 	if (!rt && lifetime) {
1323 		ND_PRINTK(3, info, "RA: adding default router\n");
1324 
1325 		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1326 					 skb->dev, pref);
1327 		if (!rt) {
1328 			ND_PRINTK(0, err,
1329 				  "RA: %s failed to add default route\n",
1330 				  __func__);
1331 			return;
1332 		}
1333 
1334 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1335 					 rt->fib6_nh->fib_nh_dev, NULL,
1336 					  &ipv6_hdr(skb)->saddr);
1337 		if (!neigh) {
1338 			ND_PRINTK(0, err,
1339 				  "RA: %s got default router without neighbour\n",
1340 				  __func__);
1341 			fib6_info_release(rt);
1342 			return;
1343 		}
1344 		neigh->flags |= NTF_ROUTER;
1345 	} else if (rt) {
1346 		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1347 	}
1348 
1349 	if (rt)
1350 		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1351 	if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1352 	    ra_msg->icmph.icmp6_hop_limit) {
1353 		if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1354 			in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1355 			fib6_metric_set(rt, RTAX_HOPLIMIT,
1356 					ra_msg->icmph.icmp6_hop_limit);
1357 		} else {
1358 			ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1359 		}
1360 	}
1361 
1362 skip_defrtr:
1363 
1364 	/*
1365 	 *	Update Reachable Time and Retrans Timer
1366 	 */
1367 
1368 	if (in6_dev->nd_parms) {
1369 		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1370 
1371 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1372 			rtime = (rtime*HZ)/1000;
1373 			if (rtime < HZ/100)
1374 				rtime = HZ/100;
1375 			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1376 			in6_dev->tstamp = jiffies;
1377 			send_ifinfo_notify = true;
1378 		}
1379 
1380 		rtime = ntohl(ra_msg->reachable_time);
1381 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1382 			rtime = (rtime*HZ)/1000;
1383 
1384 			if (rtime < HZ/10)
1385 				rtime = HZ/10;
1386 
1387 			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1388 				NEIGH_VAR_SET(in6_dev->nd_parms,
1389 					      BASE_REACHABLE_TIME, rtime);
1390 				NEIGH_VAR_SET(in6_dev->nd_parms,
1391 					      GC_STALETIME, 3 * rtime);
1392 				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1393 				in6_dev->tstamp = jiffies;
1394 				send_ifinfo_notify = true;
1395 			}
1396 		}
1397 	}
1398 
1399 	/*
1400 	 *	Send a notify if RA changed managed/otherconf flags or timer settings
1401 	 */
1402 	if (send_ifinfo_notify)
1403 		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1404 
1405 skip_linkparms:
1406 
1407 	/*
1408 	 *	Process options.
1409 	 */
1410 
1411 	if (!neigh)
1412 		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1413 				       skb->dev, 1);
1414 	if (neigh) {
1415 		u8 *lladdr = NULL;
1416 		if (ndopts.nd_opts_src_lladdr) {
1417 			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1418 						     skb->dev);
1419 			if (!lladdr) {
1420 				ND_PRINTK(2, warn,
1421 					  "RA: invalid link-layer address length\n");
1422 				goto out;
1423 			}
1424 		}
1425 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1426 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1427 			     NEIGH_UPDATE_F_OVERRIDE|
1428 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1429 			     NEIGH_UPDATE_F_ISROUTER,
1430 			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1431 	}
1432 
1433 	if (!ipv6_accept_ra(in6_dev)) {
1434 		ND_PRINTK(2, info,
1435 			  "RA: %s, accept_ra is false for dev: %s\n",
1436 			  __func__, skb->dev->name);
1437 		goto out;
1438 	}
1439 
1440 #ifdef CONFIG_IPV6_ROUTE_INFO
1441 	if (!in6_dev->cnf.accept_ra_from_local &&
1442 	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1443 			  in6_dev->dev, 0)) {
1444 		ND_PRINTK(2, info,
1445 			  "RA from local address detected on dev: %s: router info ignored.\n",
1446 			  skb->dev->name);
1447 		goto skip_routeinfo;
1448 	}
1449 
1450 	if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1451 		struct nd_opt_hdr *p;
1452 		for (p = ndopts.nd_opts_ri;
1453 		     p;
1454 		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1455 			struct route_info *ri = (struct route_info *)p;
1456 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1457 			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1458 			    ri->prefix_len == 0)
1459 				continue;
1460 #endif
1461 			if (ri->prefix_len == 0 &&
1462 			    !in6_dev->cnf.accept_ra_defrtr)
1463 				continue;
1464 			if (ri->lifetime != 0 &&
1465 			    ntohl(ri->lifetime) < in6_dev->cnf.accept_ra_min_lft)
1466 				continue;
1467 			if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1468 				continue;
1469 			if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1470 				continue;
1471 			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1472 				      &ipv6_hdr(skb)->saddr);
1473 		}
1474 	}
1475 
1476 skip_routeinfo:
1477 #endif
1478 
1479 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1480 	/* skip link-specific ndopts from interior routers */
1481 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1482 		ND_PRINTK(2, info,
1483 			  "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1484 			  __func__, skb->dev->name);
1485 		goto out;
1486 	}
1487 #endif
1488 
1489 	if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1490 		struct nd_opt_hdr *p;
1491 		for (p = ndopts.nd_opts_pi;
1492 		     p;
1493 		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1494 			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1495 					    (p->nd_opt_len) << 3,
1496 					    ndopts.nd_opts_src_lladdr != NULL);
1497 		}
1498 	}
1499 
1500 	if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1501 		__be32 n;
1502 		u32 mtu;
1503 
1504 		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1505 		mtu = ntohl(n);
1506 
1507 		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1508 			ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1509 		} else if (in6_dev->cnf.mtu6 != mtu) {
1510 			in6_dev->cnf.mtu6 = mtu;
1511 			fib6_metric_set(rt, RTAX_MTU, mtu);
1512 			rt6_mtu_change(skb->dev, mtu);
1513 		}
1514 	}
1515 
1516 	if (ndopts.nd_useropts) {
1517 		struct nd_opt_hdr *p;
1518 		for (p = ndopts.nd_useropts;
1519 		     p;
1520 		     p = ndisc_next_useropt(skb->dev, p,
1521 					    ndopts.nd_useropts_end)) {
1522 			ndisc_ra_useropt(skb, p);
1523 		}
1524 	}
1525 
1526 	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1527 		ND_PRINTK(2, warn, "RA: invalid RA options\n");
1528 	}
1529 out:
1530 	fib6_info_release(rt);
1531 	if (neigh)
1532 		neigh_release(neigh);
1533 }
1534 
ndisc_redirect_rcv(struct sk_buff * skb)1535 static void ndisc_redirect_rcv(struct sk_buff *skb)
1536 {
1537 	u8 *hdr;
1538 	struct ndisc_options ndopts;
1539 	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1540 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1541 				    offsetof(struct rd_msg, opt));
1542 
1543 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1544 	switch (skb->ndisc_nodetype) {
1545 	case NDISC_NODETYPE_HOST:
1546 	case NDISC_NODETYPE_NODEFAULT:
1547 		ND_PRINTK(2, warn,
1548 			  "Redirect: from host or unauthorized router\n");
1549 		return;
1550 	}
1551 #endif
1552 
1553 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1554 		ND_PRINTK(2, warn,
1555 			  "Redirect: source address is not link-local\n");
1556 		return;
1557 	}
1558 
1559 	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1560 		return;
1561 
1562 	if (!ndopts.nd_opts_rh) {
1563 		ip6_redirect_no_header(skb, dev_net(skb->dev),
1564 					skb->dev->ifindex);
1565 		return;
1566 	}
1567 
1568 	hdr = (u8 *)ndopts.nd_opts_rh;
1569 	hdr += 8;
1570 	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1571 		return;
1572 
1573 	icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1574 }
1575 
ndisc_fill_redirect_hdr_option(struct sk_buff * skb,struct sk_buff * orig_skb,int rd_len)1576 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1577 					   struct sk_buff *orig_skb,
1578 					   int rd_len)
1579 {
1580 	u8 *opt = skb_put(skb, rd_len);
1581 
1582 	memset(opt, 0, 8);
1583 	*(opt++) = ND_OPT_REDIRECT_HDR;
1584 	*(opt++) = (rd_len >> 3);
1585 	opt += 6;
1586 
1587 	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1588 		      rd_len - 8);
1589 }
1590 
ndisc_send_redirect(struct sk_buff * skb,const struct in6_addr * target)1591 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1592 {
1593 	struct net_device *dev = skb->dev;
1594 	struct net *net = dev_net(dev);
1595 	struct sock *sk = net->ipv6.ndisc_sk;
1596 	int optlen = 0;
1597 	struct inet_peer *peer;
1598 	struct sk_buff *buff;
1599 	struct rd_msg *msg;
1600 	struct in6_addr saddr_buf;
1601 	struct rt6_info *rt;
1602 	struct dst_entry *dst;
1603 	struct flowi6 fl6;
1604 	int rd_len;
1605 	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1606 	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1607 	bool ret;
1608 
1609 	if (netif_is_l3_master(skb->dev)) {
1610 		dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1611 		if (!dev)
1612 			return;
1613 	}
1614 
1615 	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1616 		ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1617 			  dev->name);
1618 		return;
1619 	}
1620 
1621 	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1622 	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1623 		ND_PRINTK(2, warn,
1624 			  "Redirect: target address is not link-local unicast\n");
1625 		return;
1626 	}
1627 
1628 	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1629 			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1630 
1631 	dst = ip6_route_output(net, NULL, &fl6);
1632 	if (dst->error) {
1633 		dst_release(dst);
1634 		return;
1635 	}
1636 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1637 	if (IS_ERR(dst))
1638 		return;
1639 
1640 	rt = (struct rt6_info *) dst;
1641 
1642 	if (rt->rt6i_flags & RTF_GATEWAY) {
1643 		ND_PRINTK(2, warn,
1644 			  "Redirect: destination is not a neighbour\n");
1645 		goto release;
1646 	}
1647 	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1648 	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1649 	if (peer)
1650 		inet_putpeer(peer);
1651 	if (!ret)
1652 		goto release;
1653 
1654 	if (dev->addr_len) {
1655 		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1656 		if (!neigh) {
1657 			ND_PRINTK(2, warn,
1658 				  "Redirect: no neigh for target address\n");
1659 			goto release;
1660 		}
1661 
1662 		read_lock_bh(&neigh->lock);
1663 		if (neigh->nud_state & NUD_VALID) {
1664 			memcpy(ha_buf, neigh->ha, dev->addr_len);
1665 			read_unlock_bh(&neigh->lock);
1666 			ha = ha_buf;
1667 			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1668 								ops_data_buf,
1669 								&ops_data);
1670 		} else
1671 			read_unlock_bh(&neigh->lock);
1672 
1673 		neigh_release(neigh);
1674 	}
1675 
1676 	rd_len = min_t(unsigned int,
1677 		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1678 		       skb->len + 8);
1679 	rd_len &= ~0x7;
1680 	optlen += rd_len;
1681 
1682 	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1683 	if (!buff)
1684 		goto release;
1685 
1686 	msg = skb_put(buff, sizeof(*msg));
1687 	*msg = (struct rd_msg) {
1688 		.icmph = {
1689 			.icmp6_type = NDISC_REDIRECT,
1690 		},
1691 		.target = *target,
1692 		.dest = ipv6_hdr(skb)->daddr,
1693 	};
1694 
1695 	/*
1696 	 *	include target_address option
1697 	 */
1698 
1699 	if (ha)
1700 		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1701 
1702 	/*
1703 	 *	build redirect option and copy skb over to the new packet.
1704 	 */
1705 
1706 	if (rd_len)
1707 		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1708 
1709 	skb_dst_set(buff, dst);
1710 	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1711 	return;
1712 
1713 release:
1714 	dst_release(dst);
1715 }
1716 
pndisc_redo(struct sk_buff * skb)1717 static void pndisc_redo(struct sk_buff *skb)
1718 {
1719 	ndisc_recv_ns(skb);
1720 	kfree_skb(skb);
1721 }
1722 
ndisc_is_multicast(const void * pkey)1723 static int ndisc_is_multicast(const void *pkey)
1724 {
1725 	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1726 }
1727 
ndisc_suppress_frag_ndisc(struct sk_buff * skb)1728 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1729 {
1730 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1731 
1732 	if (!idev)
1733 		return true;
1734 	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1735 	    idev->cnf.suppress_frag_ndisc) {
1736 		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1737 		return true;
1738 	}
1739 	return false;
1740 }
1741 
ndisc_rcv(struct sk_buff * skb)1742 int ndisc_rcv(struct sk_buff *skb)
1743 {
1744 	struct nd_msg *msg;
1745 
1746 	if (ndisc_suppress_frag_ndisc(skb))
1747 		return 0;
1748 
1749 	if (skb_linearize(skb))
1750 		return 0;
1751 
1752 	msg = (struct nd_msg *)skb_transport_header(skb);
1753 
1754 	__skb_push(skb, skb->data - skb_transport_header(skb));
1755 
1756 	if (ipv6_hdr(skb)->hop_limit != 255) {
1757 		ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1758 			  ipv6_hdr(skb)->hop_limit);
1759 		return 0;
1760 	}
1761 
1762 	if (msg->icmph.icmp6_code != 0) {
1763 		ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1764 			  msg->icmph.icmp6_code);
1765 		return 0;
1766 	}
1767 
1768 	switch (msg->icmph.icmp6_type) {
1769 	case NDISC_NEIGHBOUR_SOLICITATION:
1770 		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1771 		ndisc_recv_ns(skb);
1772 		break;
1773 
1774 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1775 		ndisc_recv_na(skb);
1776 		break;
1777 
1778 	case NDISC_ROUTER_SOLICITATION:
1779 		ndisc_recv_rs(skb);
1780 		break;
1781 
1782 	case NDISC_ROUTER_ADVERTISEMENT:
1783 		ndisc_router_discovery(skb);
1784 		break;
1785 
1786 	case NDISC_REDIRECT:
1787 		ndisc_redirect_rcv(skb);
1788 		break;
1789 	}
1790 
1791 	return 0;
1792 }
1793 
ndisc_netdev_event(struct notifier_block * this,unsigned long event,void * ptr)1794 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1795 {
1796 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1797 	struct netdev_notifier_change_info *change_info;
1798 	struct net *net = dev_net(dev);
1799 	struct inet6_dev *idev;
1800 
1801 	switch (event) {
1802 	case NETDEV_CHANGEADDR:
1803 		neigh_changeaddr(&nd_tbl, dev);
1804 		fib6_run_gc(0, net, false);
1805 		fallthrough;
1806 	case NETDEV_UP:
1807 		idev = in6_dev_get(dev);
1808 		if (!idev)
1809 			break;
1810 		if (idev->cnf.ndisc_notify ||
1811 		    net->ipv6.devconf_all->ndisc_notify)
1812 			ndisc_send_unsol_na(dev);
1813 		in6_dev_put(idev);
1814 		break;
1815 	case NETDEV_CHANGE:
1816 		change_info = ptr;
1817 		if (change_info->flags_changed & IFF_NOARP)
1818 			neigh_changeaddr(&nd_tbl, dev);
1819 		if (!netif_carrier_ok(dev))
1820 			neigh_carrier_down(&nd_tbl, dev);
1821 		break;
1822 	case NETDEV_DOWN:
1823 		neigh_ifdown(&nd_tbl, dev);
1824 		fib6_run_gc(0, net, false);
1825 		break;
1826 	case NETDEV_NOTIFY_PEERS:
1827 		ndisc_send_unsol_na(dev);
1828 		break;
1829 	default:
1830 		break;
1831 	}
1832 
1833 	return NOTIFY_DONE;
1834 }
1835 
1836 static struct notifier_block ndisc_netdev_notifier = {
1837 	.notifier_call = ndisc_netdev_event,
1838 	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1839 };
1840 
1841 #ifdef CONFIG_SYSCTL
ndisc_warn_deprecated_sysctl(struct ctl_table * ctl,const char * func,const char * dev_name)1842 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1843 					 const char *func, const char *dev_name)
1844 {
1845 	static char warncomm[TASK_COMM_LEN];
1846 	static int warned;
1847 	if (strcmp(warncomm, current->comm) && warned < 5) {
1848 		strcpy(warncomm, current->comm);
1849 		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1850 			warncomm, func,
1851 			dev_name, ctl->procname,
1852 			dev_name, ctl->procname);
1853 		warned++;
1854 	}
1855 }
1856 
ndisc_ifinfo_sysctl_change(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)1857 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1858 		size_t *lenp, loff_t *ppos)
1859 {
1860 	struct net_device *dev = ctl->extra1;
1861 	struct inet6_dev *idev;
1862 	int ret;
1863 
1864 	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1865 	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1866 		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1867 
1868 	if (strcmp(ctl->procname, "retrans_time") == 0)
1869 		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1870 
1871 	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1872 		ret = neigh_proc_dointvec_jiffies(ctl, write,
1873 						  buffer, lenp, ppos);
1874 
1875 	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1876 		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1877 		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1878 						     buffer, lenp, ppos);
1879 	else
1880 		ret = -1;
1881 
1882 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1883 		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1884 			idev->nd_parms->reachable_time =
1885 					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1886 		idev->tstamp = jiffies;
1887 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1888 		in6_dev_put(idev);
1889 	}
1890 	return ret;
1891 }
1892 
1893 
1894 #endif
1895 
ndisc_net_init(struct net * net)1896 static int __net_init ndisc_net_init(struct net *net)
1897 {
1898 	struct ipv6_pinfo *np;
1899 	struct sock *sk;
1900 	int err;
1901 
1902 	err = inet_ctl_sock_create(&sk, PF_INET6,
1903 				   SOCK_RAW, IPPROTO_ICMPV6, net);
1904 	if (err < 0) {
1905 		ND_PRINTK(0, err,
1906 			  "NDISC: Failed to initialize the control socket (err %d)\n",
1907 			  err);
1908 		return err;
1909 	}
1910 
1911 	net->ipv6.ndisc_sk = sk;
1912 
1913 	np = inet6_sk(sk);
1914 	np->hop_limit = 255;
1915 	/* Do not loopback ndisc messages */
1916 	np->mc_loop = 0;
1917 
1918 	return 0;
1919 }
1920 
ndisc_net_exit(struct net * net)1921 static void __net_exit ndisc_net_exit(struct net *net)
1922 {
1923 	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1924 }
1925 
1926 static struct pernet_operations ndisc_net_ops = {
1927 	.init = ndisc_net_init,
1928 	.exit = ndisc_net_exit,
1929 };
1930 
ndisc_init(void)1931 int __init ndisc_init(void)
1932 {
1933 	int err;
1934 
1935 	err = register_pernet_subsys(&ndisc_net_ops);
1936 	if (err)
1937 		return err;
1938 	/*
1939 	 * Initialize the neighbour table
1940 	 */
1941 	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1942 
1943 #ifdef CONFIG_SYSCTL
1944 	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1945 				    ndisc_ifinfo_sysctl_change);
1946 	if (err)
1947 		goto out_unregister_pernet;
1948 out:
1949 #endif
1950 	return err;
1951 
1952 #ifdef CONFIG_SYSCTL
1953 out_unregister_pernet:
1954 	unregister_pernet_subsys(&ndisc_net_ops);
1955 	goto out;
1956 #endif
1957 }
1958 
ndisc_late_init(void)1959 int __init ndisc_late_init(void)
1960 {
1961 	return register_netdevice_notifier(&ndisc_netdev_notifier);
1962 }
1963 
ndisc_late_cleanup(void)1964 void ndisc_late_cleanup(void)
1965 {
1966 	unregister_netdevice_notifier(&ndisc_netdev_notifier);
1967 }
1968 
ndisc_cleanup(void)1969 void ndisc_cleanup(void)
1970 {
1971 #ifdef CONFIG_SYSCTL
1972 	neigh_sysctl_unregister(&nd_tbl.parms);
1973 #endif
1974 	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1975 	unregister_pernet_subsys(&ndisc_net_ops);
1976 }
1977