xref: /linux/net/ipv6/ndisc.c (revision 348f968b89bfeec0bb53dd82dba58b94d97fbd34)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Neighbour Discovery for IPv6
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Mike Shaver		<shaver@ingenia.com>
9  */
10 
11 /*
12  *	Changes:
13  *
14  *	Alexey I. Froloff		:	RFC6106 (DNSSL) support
15  *	Pierre Ynard			:	export userland ND options
16  *						through netlink (RDNSS support)
17  *	Lars Fenneberg			:	fixed MTU setting on receipt
18  *						of an RA.
19  *	Janos Farkas			:	kmalloc failure checks
20  *	Alexey Kuznetsov		:	state machine reworked
21  *						and moved to net/core.
22  *	Pekka Savola			:	RFC2461 validation
23  *	YOSHIFUJI Hideaki @USAGI	:	Verify ND options properly
24  */
25 
26 #define pr_fmt(fmt) "ICMPv6: " fmt
27 
28 #include <linux/module.h>
29 #include <linux/errno.h>
30 #include <linux/types.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/sched.h>
34 #include <linux/net.h>
35 #include <linux/in6.h>
36 #include <linux/route.h>
37 #include <linux/init.h>
38 #include <linux/rcupdate.h>
39 #include <linux/slab.h>
40 #ifdef CONFIG_SYSCTL
41 #include <linux/sysctl.h>
42 #endif
43 
44 #include <linux/if_addr.h>
45 #include <linux/if_ether.h>
46 #include <linux/if_arp.h>
47 #include <linux/ipv6.h>
48 #include <linux/icmpv6.h>
49 #include <linux/jhash.h>
50 
51 #include <net/sock.h>
52 #include <net/snmp.h>
53 
54 #include <net/ipv6.h>
55 #include <net/protocol.h>
56 #include <net/ndisc.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #include <net/icmp.h>
60 
61 #include <net/netlink.h>
62 #include <linux/rtnetlink.h>
63 
64 #include <net/flow.h>
65 #include <net/ip6_checksum.h>
66 #include <net/inet_common.h>
67 #include <linux/proc_fs.h>
68 
69 #include <linux/netfilter.h>
70 #include <linux/netfilter_ipv6.h>
71 
72 static u32 ndisc_hash(const void *pkey,
73 		      const struct net_device *dev,
74 		      __u32 *hash_rnd);
75 static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
76 static bool ndisc_allow_add(const struct net_device *dev,
77 			    struct netlink_ext_ack *extack);
78 static int ndisc_constructor(struct neighbour *neigh);
79 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
80 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
81 static int pndisc_constructor(struct pneigh_entry *n);
82 static void pndisc_destructor(struct pneigh_entry *n);
83 static void pndisc_redo(struct sk_buff *skb);
84 static int ndisc_is_multicast(const void *pkey);
85 
86 static const struct neigh_ops ndisc_generic_ops = {
87 	.family =		AF_INET6,
88 	.solicit =		ndisc_solicit,
89 	.error_report =		ndisc_error_report,
90 	.output =		neigh_resolve_output,
91 	.connected_output =	neigh_connected_output,
92 };
93 
94 static const struct neigh_ops ndisc_hh_ops = {
95 	.family =		AF_INET6,
96 	.solicit =		ndisc_solicit,
97 	.error_report =		ndisc_error_report,
98 	.output =		neigh_resolve_output,
99 	.connected_output =	neigh_resolve_output,
100 };
101 
102 
103 static const struct neigh_ops ndisc_direct_ops = {
104 	.family =		AF_INET6,
105 	.output =		neigh_direct_output,
106 	.connected_output =	neigh_direct_output,
107 };
108 
109 struct neigh_table nd_tbl = {
110 	.family =	AF_INET6,
111 	.key_len =	sizeof(struct in6_addr),
112 	.protocol =	cpu_to_be16(ETH_P_IPV6),
113 	.hash =		ndisc_hash,
114 	.key_eq =	ndisc_key_eq,
115 	.constructor =	ndisc_constructor,
116 	.pconstructor =	pndisc_constructor,
117 	.pdestructor =	pndisc_destructor,
118 	.proxy_redo =	pndisc_redo,
119 	.is_multicast =	ndisc_is_multicast,
120 	.allow_add  =   ndisc_allow_add,
121 	.id =		"ndisc_cache",
122 	.parms = {
123 		.tbl			= &nd_tbl,
124 		.reachable_time		= ND_REACHABLE_TIME,
125 		.data = {
126 			[NEIGH_VAR_MCAST_PROBES] = 3,
127 			[NEIGH_VAR_UCAST_PROBES] = 3,
128 			[NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER,
129 			[NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME,
130 			[NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
131 			[NEIGH_VAR_INTERVAL_PROBE_TIME_MS] = 5 * HZ,
132 			[NEIGH_VAR_GC_STALETIME] = 60 * HZ,
133 			[NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
134 			[NEIGH_VAR_PROXY_QLEN] = 64,
135 			[NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
136 			[NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
137 		},
138 	},
139 	.gc_interval =	  30 * HZ,
140 	.gc_thresh1 =	 128,
141 	.gc_thresh2 =	 512,
142 	.gc_thresh3 =	1024,
143 };
144 EXPORT_SYMBOL_GPL(nd_tbl);
145 
__ndisc_fill_addr_option(struct sk_buff * skb,int type,const void * data,int data_len,int pad)146 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, const void *data,
147 			      int data_len, int pad)
148 {
149 	int space = __ndisc_opt_addr_space(data_len, pad);
150 	u8 *opt = skb_put(skb, space);
151 
152 	opt[0] = type;
153 	opt[1] = space>>3;
154 
155 	memset(opt + 2, 0, pad);
156 	opt   += pad;
157 	space -= pad;
158 
159 	memcpy(opt+2, data, data_len);
160 	data_len += 2;
161 	opt += data_len;
162 	space -= data_len;
163 	if (space > 0)
164 		memset(opt, 0, space);
165 }
166 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
167 
ndisc_fill_addr_option(struct sk_buff * skb,int type,const void * data,u8 icmp6_type)168 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
169 					  const void *data, u8 icmp6_type)
170 {
171 	__ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
172 				 ndisc_addr_option_pad(skb->dev->type));
173 	ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
174 }
175 
ndisc_fill_redirect_addr_option(struct sk_buff * skb,void * ha,const u8 * ops_data)176 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
177 						   void *ha,
178 						   const u8 *ops_data)
179 {
180 	ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
181 	ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
182 }
183 
ndisc_next_option(struct nd_opt_hdr * cur,struct nd_opt_hdr * end)184 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
185 					    struct nd_opt_hdr *end)
186 {
187 	int type;
188 	if (!cur || !end || cur >= end)
189 		return NULL;
190 	type = cur->nd_opt_type;
191 	do {
192 		cur = ((void *)cur) + (cur->nd_opt_len << 3);
193 	} while (cur < end && cur->nd_opt_type != type);
194 	return cur <= end && cur->nd_opt_type == type ? cur : NULL;
195 }
196 
ndisc_is_useropt(const struct net_device * dev,struct nd_opt_hdr * opt)197 static inline int ndisc_is_useropt(const struct net_device *dev,
198 				   struct nd_opt_hdr *opt)
199 {
200 	return opt->nd_opt_type == ND_OPT_PREFIX_INFO ||
201 		opt->nd_opt_type == ND_OPT_RDNSS ||
202 		opt->nd_opt_type == ND_OPT_DNSSL ||
203 		opt->nd_opt_type == ND_OPT_6CO ||
204 		opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
205 		opt->nd_opt_type == ND_OPT_PREF64;
206 }
207 
ndisc_next_useropt(const struct net_device * dev,struct nd_opt_hdr * cur,struct nd_opt_hdr * end)208 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
209 					     struct nd_opt_hdr *cur,
210 					     struct nd_opt_hdr *end)
211 {
212 	if (!cur || !end || cur >= end)
213 		return NULL;
214 	do {
215 		cur = ((void *)cur) + (cur->nd_opt_len << 3);
216 	} while (cur < end && !ndisc_is_useropt(dev, cur));
217 	return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
218 }
219 
ndisc_parse_options(const struct net_device * dev,u8 * opt,int opt_len,struct ndisc_options * ndopts)220 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
221 					  u8 *opt, int opt_len,
222 					  struct ndisc_options *ndopts)
223 {
224 	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
225 
226 	if (!nd_opt || opt_len < 0 || !ndopts)
227 		return NULL;
228 	memset(ndopts, 0, sizeof(*ndopts));
229 	while (opt_len) {
230 		bool unknown = false;
231 		int l;
232 		if (opt_len < sizeof(struct nd_opt_hdr))
233 			return NULL;
234 		l = nd_opt->nd_opt_len << 3;
235 		if (opt_len < l || l == 0)
236 			return NULL;
237 		if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
238 			goto next_opt;
239 		switch (nd_opt->nd_opt_type) {
240 		case ND_OPT_SOURCE_LL_ADDR:
241 		case ND_OPT_TARGET_LL_ADDR:
242 		case ND_OPT_MTU:
243 		case ND_OPT_NONCE:
244 		case ND_OPT_REDIRECT_HDR:
245 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
246 				ND_PRINTK(2, warn,
247 					  "%s: duplicated ND6 option found: type=%d\n",
248 					  __func__, nd_opt->nd_opt_type);
249 			} else {
250 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
251 			}
252 			break;
253 		case ND_OPT_PREFIX_INFO:
254 			ndopts->nd_opts_pi_end = nd_opt;
255 			if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
256 				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
257 			break;
258 #ifdef CONFIG_IPV6_ROUTE_INFO
259 		case ND_OPT_ROUTE_INFO:
260 			ndopts->nd_opts_ri_end = nd_opt;
261 			if (!ndopts->nd_opts_ri)
262 				ndopts->nd_opts_ri = nd_opt;
263 			break;
264 #endif
265 		default:
266 			unknown = true;
267 		}
268 		if (ndisc_is_useropt(dev, nd_opt)) {
269 			ndopts->nd_useropts_end = nd_opt;
270 			if (!ndopts->nd_useropts)
271 				ndopts->nd_useropts = nd_opt;
272 		} else if (unknown) {
273 			/*
274 			 * Unknown options must be silently ignored,
275 			 * to accommodate future extension to the
276 			 * protocol.
277 			 */
278 			ND_PRINTK(2, notice,
279 				  "%s: ignored unsupported option; type=%d, len=%d\n",
280 				  __func__,
281 				  nd_opt->nd_opt_type,
282 				  nd_opt->nd_opt_len);
283 		}
284 next_opt:
285 		opt_len -= l;
286 		nd_opt = ((void *)nd_opt) + l;
287 	}
288 	return ndopts;
289 }
290 
ndisc_mc_map(const struct in6_addr * addr,char * buf,struct net_device * dev,int dir)291 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
292 {
293 	switch (dev->type) {
294 	case ARPHRD_ETHER:
295 	case ARPHRD_IEEE802:	/* Not sure. Check it later. --ANK */
296 	case ARPHRD_FDDI:
297 		ipv6_eth_mc_map(addr, buf);
298 		return 0;
299 	case ARPHRD_ARCNET:
300 		ipv6_arcnet_mc_map(addr, buf);
301 		return 0;
302 	case ARPHRD_INFINIBAND:
303 		ipv6_ib_mc_map(addr, dev->broadcast, buf);
304 		return 0;
305 	case ARPHRD_IPGRE:
306 		return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
307 	default:
308 		if (dir) {
309 			memcpy(buf, dev->broadcast, dev->addr_len);
310 			return 0;
311 		}
312 	}
313 	return -EINVAL;
314 }
315 EXPORT_SYMBOL(ndisc_mc_map);
316 
ndisc_hash(const void * pkey,const struct net_device * dev,__u32 * hash_rnd)317 static u32 ndisc_hash(const void *pkey,
318 		      const struct net_device *dev,
319 		      __u32 *hash_rnd)
320 {
321 	return ndisc_hashfn(pkey, dev, hash_rnd);
322 }
323 
ndisc_key_eq(const struct neighbour * n,const void * pkey)324 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
325 {
326 	return neigh_key_eq128(n, pkey);
327 }
328 
ndisc_constructor(struct neighbour * neigh)329 static int ndisc_constructor(struct neighbour *neigh)
330 {
331 	struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
332 	struct net_device *dev = neigh->dev;
333 	struct inet6_dev *in6_dev;
334 	struct neigh_parms *parms;
335 	bool is_multicast = ipv6_addr_is_multicast(addr);
336 
337 	in6_dev = in6_dev_get(dev);
338 	if (!in6_dev) {
339 		return -EINVAL;
340 	}
341 
342 	parms = in6_dev->nd_parms;
343 	__neigh_parms_put(neigh->parms);
344 	neigh->parms = neigh_parms_clone(parms);
345 
346 	neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
347 	if (!dev->header_ops) {
348 		neigh->nud_state = NUD_NOARP;
349 		neigh->ops = &ndisc_direct_ops;
350 		neigh->output = neigh_direct_output;
351 	} else {
352 		if (is_multicast) {
353 			neigh->nud_state = NUD_NOARP;
354 			ndisc_mc_map(addr, neigh->ha, dev, 1);
355 		} else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
356 			neigh->nud_state = NUD_NOARP;
357 			memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
358 			if (dev->flags&IFF_LOOPBACK)
359 				neigh->type = RTN_LOCAL;
360 		} else if (dev->flags&IFF_POINTOPOINT) {
361 			neigh->nud_state = NUD_NOARP;
362 			memcpy(neigh->ha, dev->broadcast, dev->addr_len);
363 		}
364 		if (dev->header_ops->cache)
365 			neigh->ops = &ndisc_hh_ops;
366 		else
367 			neigh->ops = &ndisc_generic_ops;
368 		if (neigh->nud_state&NUD_VALID)
369 			neigh->output = neigh->ops->connected_output;
370 		else
371 			neigh->output = neigh->ops->output;
372 	}
373 	in6_dev_put(in6_dev);
374 	return 0;
375 }
376 
pndisc_constructor(struct pneigh_entry * n)377 static int pndisc_constructor(struct pneigh_entry *n)
378 {
379 	struct in6_addr *addr = (struct in6_addr *)&n->key;
380 	struct in6_addr maddr;
381 	struct net_device *dev = n->dev;
382 
383 	if (!dev || !__in6_dev_get(dev))
384 		return -EINVAL;
385 	addrconf_addr_solict_mult(addr, &maddr);
386 	ipv6_dev_mc_inc(dev, &maddr);
387 	return 0;
388 }
389 
pndisc_destructor(struct pneigh_entry * n)390 static void pndisc_destructor(struct pneigh_entry *n)
391 {
392 	struct in6_addr *addr = (struct in6_addr *)&n->key;
393 	struct in6_addr maddr;
394 	struct net_device *dev = n->dev;
395 
396 	if (!dev || !__in6_dev_get(dev))
397 		return;
398 	addrconf_addr_solict_mult(addr, &maddr);
399 	ipv6_dev_mc_dec(dev, &maddr);
400 }
401 
402 /* called with rtnl held */
ndisc_allow_add(const struct net_device * dev,struct netlink_ext_ack * extack)403 static bool ndisc_allow_add(const struct net_device *dev,
404 			    struct netlink_ext_ack *extack)
405 {
406 	struct inet6_dev *idev = __in6_dev_get(dev);
407 
408 	if (!idev || idev->cnf.disable_ipv6) {
409 		NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
410 		return false;
411 	}
412 
413 	return true;
414 }
415 
ndisc_alloc_skb(struct net_device * dev,int len)416 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
417 				       int len)
418 {
419 	int hlen = LL_RESERVED_SPACE(dev);
420 	int tlen = dev->needed_tailroom;
421 	struct sk_buff *skb;
422 
423 	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
424 	if (!skb)
425 		return NULL;
426 
427 	skb->protocol = htons(ETH_P_IPV6);
428 	skb->dev = dev;
429 
430 	skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
431 	skb_reset_transport_header(skb);
432 
433 	/* Manually assign socket ownership as we avoid calling
434 	 * sock_alloc_send_pskb() to bypass wmem buffer limits
435 	 */
436 	rcu_read_lock();
437 	skb_set_owner_w(skb, dev_net_rcu(dev)->ipv6.ndisc_sk);
438 	rcu_read_unlock();
439 
440 	return skb;
441 }
442 
ip6_nd_hdr(struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr,int hop_limit,int len)443 static void ip6_nd_hdr(struct sk_buff *skb,
444 		       const struct in6_addr *saddr,
445 		       const struct in6_addr *daddr,
446 		       int hop_limit, int len)
447 {
448 	struct ipv6hdr *hdr;
449 	struct inet6_dev *idev;
450 	unsigned tclass;
451 
452 	rcu_read_lock();
453 	idev = __in6_dev_get(skb->dev);
454 	tclass = idev ? READ_ONCE(idev->cnf.ndisc_tclass) : 0;
455 	rcu_read_unlock();
456 
457 	skb_push(skb, sizeof(*hdr));
458 	skb_reset_network_header(skb);
459 	hdr = ipv6_hdr(skb);
460 
461 	ip6_flow_hdr(hdr, tclass, 0);
462 
463 	hdr->payload_len = htons(len);
464 	hdr->nexthdr = IPPROTO_ICMPV6;
465 	hdr->hop_limit = hop_limit;
466 
467 	hdr->saddr = *saddr;
468 	hdr->daddr = *daddr;
469 }
470 
ndisc_send_skb(struct sk_buff * skb,const struct in6_addr * daddr,const struct in6_addr * saddr)471 void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
472 		    const struct in6_addr *saddr)
473 {
474 	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
475 	struct dst_entry *dst = skb_dst(skb);
476 	struct inet6_dev *idev;
477 	struct net *net;
478 	struct sock *sk;
479 	int err;
480 	u8 type;
481 
482 	type = icmp6h->icmp6_type;
483 
484 	rcu_read_lock();
485 
486 	net = dev_net_rcu(skb->dev);
487 	sk = net->ipv6.ndisc_sk;
488 	if (!dst) {
489 		struct flowi6 fl6;
490 		int oif = skb->dev->ifindex;
491 
492 		icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
493 		dst = icmp6_dst_alloc(skb->dev, &fl6);
494 		if (IS_ERR(dst)) {
495 			rcu_read_unlock();
496 			kfree_skb(skb);
497 			return;
498 		}
499 
500 		skb_dst_set(skb, dst);
501 	}
502 
503 	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
504 					      IPPROTO_ICMPV6,
505 					      csum_partial(icmp6h,
506 							   skb->len, 0));
507 
508 	ip6_nd_hdr(skb, saddr, daddr, READ_ONCE(inet6_sk(sk)->hop_limit), skb->len);
509 
510 	idev = __in6_dev_get(dst->dev);
511 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
512 
513 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
514 		      net, sk, skb, NULL, dst->dev,
515 		      dst_output);
516 	if (!err) {
517 		ICMP6MSGOUT_INC_STATS(net, idev, type);
518 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
519 	}
520 
521 	rcu_read_unlock();
522 }
523 EXPORT_SYMBOL(ndisc_send_skb);
524 
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)525 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
526 		   const struct in6_addr *solicited_addr,
527 		   bool router, bool solicited, bool override, bool inc_opt)
528 {
529 	struct sk_buff *skb;
530 	struct in6_addr tmpaddr;
531 	struct inet6_ifaddr *ifp;
532 	const struct in6_addr *src_addr;
533 	struct nd_msg *msg;
534 	int optlen = 0;
535 
536 	/* for anycast or proxy, solicited_addr != src_addr */
537 	ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
538 	if (ifp) {
539 		src_addr = solicited_addr;
540 		if (ifp->flags & IFA_F_OPTIMISTIC)
541 			override = false;
542 		inc_opt |= READ_ONCE(ifp->idev->cnf.force_tllao);
543 		in6_ifa_put(ifp);
544 	} else {
545 		if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
546 				       inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
547 				       &tmpaddr))
548 			return;
549 		src_addr = &tmpaddr;
550 	}
551 
552 	if (!dev->addr_len)
553 		inc_opt = false;
554 	if (inc_opt)
555 		optlen += ndisc_opt_addr_space(dev,
556 					       NDISC_NEIGHBOUR_ADVERTISEMENT);
557 
558 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
559 	if (!skb)
560 		return;
561 
562 	msg = skb_put(skb, sizeof(*msg));
563 	*msg = (struct nd_msg) {
564 		.icmph = {
565 			.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
566 			.icmp6_router = router,
567 			.icmp6_solicited = solicited,
568 			.icmp6_override = override,
569 		},
570 		.target = *solicited_addr,
571 	};
572 
573 	if (inc_opt)
574 		ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
575 				       dev->dev_addr,
576 				       NDISC_NEIGHBOUR_ADVERTISEMENT);
577 
578 	ndisc_send_skb(skb, daddr, src_addr);
579 }
580 
ndisc_send_unsol_na(struct net_device * dev)581 static void ndisc_send_unsol_na(struct net_device *dev)
582 {
583 	struct inet6_dev *idev;
584 	struct inet6_ifaddr *ifa;
585 
586 	idev = in6_dev_get(dev);
587 	if (!idev)
588 		return;
589 
590 	read_lock_bh(&idev->lock);
591 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
592 		/* skip tentative addresses until dad completes */
593 		if (ifa->flags & IFA_F_TENTATIVE &&
594 		    !(ifa->flags & IFA_F_OPTIMISTIC))
595 			continue;
596 
597 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
598 			      /*router=*/ !!idev->cnf.forwarding,
599 			      /*solicited=*/ false, /*override=*/ true,
600 			      /*inc_opt=*/ true);
601 	}
602 	read_unlock_bh(&idev->lock);
603 
604 	in6_dev_put(idev);
605 }
606 
ndisc_ns_create(struct net_device * dev,const struct in6_addr * solicit,const struct in6_addr * saddr,u64 nonce)607 struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
608 				const struct in6_addr *saddr, u64 nonce)
609 {
610 	int inc_opt = dev->addr_len;
611 	struct sk_buff *skb;
612 	struct nd_msg *msg;
613 	int optlen = 0;
614 
615 	if (!saddr)
616 		return NULL;
617 
618 	if (ipv6_addr_any(saddr))
619 		inc_opt = false;
620 	if (inc_opt)
621 		optlen += ndisc_opt_addr_space(dev,
622 					       NDISC_NEIGHBOUR_SOLICITATION);
623 	if (nonce != 0)
624 		optlen += 8;
625 
626 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
627 	if (!skb)
628 		return NULL;
629 
630 	msg = skb_put(skb, sizeof(*msg));
631 	*msg = (struct nd_msg) {
632 		.icmph = {
633 			.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
634 		},
635 		.target = *solicit,
636 	};
637 
638 	if (inc_opt)
639 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
640 				       dev->dev_addr,
641 				       NDISC_NEIGHBOUR_SOLICITATION);
642 	if (nonce != 0) {
643 		u8 *opt = skb_put(skb, 8);
644 
645 		opt[0] = ND_OPT_NONCE;
646 		opt[1] = 8 >> 3;
647 		memcpy(opt + 2, &nonce, 6);
648 	}
649 
650 	return skb;
651 }
652 EXPORT_SYMBOL(ndisc_ns_create);
653 
ndisc_send_ns(struct net_device * dev,const struct in6_addr * solicit,const struct in6_addr * daddr,const struct in6_addr * saddr,u64 nonce)654 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
655 		   const struct in6_addr *daddr, const struct in6_addr *saddr,
656 		   u64 nonce)
657 {
658 	struct in6_addr addr_buf;
659 	struct sk_buff *skb;
660 
661 	if (!saddr) {
662 		if (ipv6_get_lladdr(dev, &addr_buf,
663 				    (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
664 			return;
665 		saddr = &addr_buf;
666 	}
667 
668 	skb = ndisc_ns_create(dev, solicit, saddr, nonce);
669 
670 	if (skb)
671 		ndisc_send_skb(skb, daddr, saddr);
672 }
673 
ndisc_send_rs(struct net_device * dev,const struct in6_addr * saddr,const struct in6_addr * daddr)674 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
675 		   const struct in6_addr *daddr)
676 {
677 	struct sk_buff *skb;
678 	struct rs_msg *msg;
679 	int send_sllao = dev->addr_len;
680 	int optlen = 0;
681 
682 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
683 	/*
684 	 * According to section 2.2 of RFC 4429, we must not
685 	 * send router solicitations with a sllao from
686 	 * optimistic addresses, but we may send the solicitation
687 	 * if we don't include the sllao.  So here we check
688 	 * if our address is optimistic, and if so, we
689 	 * suppress the inclusion of the sllao.
690 	 */
691 	if (send_sllao) {
692 		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
693 							   dev, 1);
694 		if (ifp) {
695 			if (ifp->flags & IFA_F_OPTIMISTIC)  {
696 				send_sllao = 0;
697 			}
698 			in6_ifa_put(ifp);
699 		} else {
700 			send_sllao = 0;
701 		}
702 	}
703 #endif
704 	if (send_sllao)
705 		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
706 
707 	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
708 	if (!skb)
709 		return;
710 
711 	msg = skb_put(skb, sizeof(*msg));
712 	*msg = (struct rs_msg) {
713 		.icmph = {
714 			.icmp6_type = NDISC_ROUTER_SOLICITATION,
715 		},
716 	};
717 
718 	if (send_sllao)
719 		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
720 				       dev->dev_addr,
721 				       NDISC_ROUTER_SOLICITATION);
722 
723 	ndisc_send_skb(skb, daddr, saddr);
724 }
725 
726 
ndisc_error_report(struct neighbour * neigh,struct sk_buff * skb)727 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
728 {
729 	/*
730 	 *	"The sender MUST return an ICMP
731 	 *	 destination unreachable"
732 	 */
733 	dst_link_failure(skb);
734 	kfree_skb(skb);
735 }
736 
737 /* Called with locked neigh: either read or both */
738 
ndisc_solicit(struct neighbour * neigh,struct sk_buff * skb)739 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
740 {
741 	struct in6_addr *saddr = NULL;
742 	struct in6_addr mcaddr;
743 	struct net_device *dev = neigh->dev;
744 	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
745 	int probes = atomic_read(&neigh->probes);
746 
747 	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
748 					   dev, false, 1,
749 					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
750 		saddr = &ipv6_hdr(skb)->saddr;
751 	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
752 	if (probes < 0) {
753 		if (!(READ_ONCE(neigh->nud_state) & NUD_VALID)) {
754 			ND_PRINTK(1, dbg,
755 				  "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
756 				  __func__, target);
757 		}
758 		ndisc_send_ns(dev, target, target, saddr, 0);
759 	} else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
760 		neigh_app_ns(neigh);
761 	} else {
762 		addrconf_addr_solict_mult(target, &mcaddr);
763 		ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
764 	}
765 }
766 
pndisc_is_router(const void * pkey,struct net_device * dev)767 static int pndisc_is_router(const void *pkey,
768 			    struct net_device *dev)
769 {
770 	struct pneigh_entry *n;
771 	int ret = -1;
772 
773 	read_lock_bh(&nd_tbl.lock);
774 	n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
775 	if (n)
776 		ret = !!(n->flags & NTF_ROUTER);
777 	read_unlock_bh(&nd_tbl.lock);
778 
779 	return ret;
780 }
781 
ndisc_update(const struct net_device * dev,struct neighbour * neigh,const u8 * lladdr,u8 new,u32 flags,u8 icmp6_type,struct ndisc_options * ndopts)782 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
783 		  const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
784 		  struct ndisc_options *ndopts)
785 {
786 	neigh_update(neigh, lladdr, new, flags, 0);
787 	/* report ndisc ops about neighbour update */
788 	ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
789 }
790 
ndisc_recv_ns(struct sk_buff * skb)791 static enum skb_drop_reason ndisc_recv_ns(struct sk_buff *skb)
792 {
793 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
794 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
795 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
796 	u8 *lladdr = NULL;
797 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
798 				    offsetof(struct nd_msg, opt));
799 	struct ndisc_options ndopts;
800 	struct net_device *dev = skb->dev;
801 	struct inet6_ifaddr *ifp;
802 	struct inet6_dev *idev = NULL;
803 	struct neighbour *neigh;
804 	int dad = ipv6_addr_any(saddr);
805 	int is_router = -1;
806 	SKB_DR(reason);
807 	u64 nonce = 0;
808 	bool inc;
809 
810 	if (skb->len < sizeof(struct nd_msg))
811 		return SKB_DROP_REASON_PKT_TOO_SMALL;
812 
813 	if (ipv6_addr_is_multicast(&msg->target)) {
814 		ND_PRINTK(2, warn, "NS: multicast target address\n");
815 		return reason;
816 	}
817 
818 	/*
819 	 * RFC2461 7.1.1:
820 	 * DAD has to be destined for solicited node multicast address.
821 	 */
822 	if (dad && !ipv6_addr_is_solict_mult(daddr)) {
823 		ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
824 		return reason;
825 	}
826 
827 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
828 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
829 
830 	if (ndopts.nd_opts_src_lladdr) {
831 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
832 		if (!lladdr) {
833 			ND_PRINTK(2, warn,
834 				  "NS: invalid link-layer address length\n");
835 			return reason;
836 		}
837 
838 		/* RFC2461 7.1.1:
839 		 *	If the IP source address is the unspecified address,
840 		 *	there MUST NOT be source link-layer address option
841 		 *	in the message.
842 		 */
843 		if (dad) {
844 			ND_PRINTK(2, warn,
845 				  "NS: bad DAD packet (link-layer address option)\n");
846 			return reason;
847 		}
848 	}
849 	if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
850 		memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
851 
852 	inc = ipv6_addr_is_multicast(daddr);
853 
854 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
855 	if (ifp) {
856 have_ifp:
857 		if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
858 			if (dad) {
859 				if (nonce != 0 && ifp->dad_nonce == nonce) {
860 					u8 *np = (u8 *)&nonce;
861 					/* Matching nonce if looped back */
862 					ND_PRINTK(2, notice,
863 						  "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
864 						  ifp->idev->dev->name,
865 						  &ifp->addr, np);
866 					goto out;
867 				}
868 				/*
869 				 * We are colliding with another node
870 				 * who is doing DAD
871 				 * so fail our DAD process
872 				 */
873 				addrconf_dad_failure(skb, ifp);
874 				return reason;
875 			} else {
876 				/*
877 				 * This is not a dad solicitation.
878 				 * If we are an optimistic node,
879 				 * we should respond.
880 				 * Otherwise, we should ignore it.
881 				 */
882 				if (!(ifp->flags & IFA_F_OPTIMISTIC))
883 					goto out;
884 			}
885 		}
886 
887 		idev = ifp->idev;
888 	} else {
889 		struct net *net = dev_net(dev);
890 
891 		/* perhaps an address on the master device */
892 		if (netif_is_l3_slave(dev)) {
893 			struct net_device *mdev;
894 
895 			mdev = netdev_master_upper_dev_get_rcu(dev);
896 			if (mdev) {
897 				ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
898 				if (ifp)
899 					goto have_ifp;
900 			}
901 		}
902 
903 		idev = in6_dev_get(dev);
904 		if (!idev) {
905 			/* XXX: count this drop? */
906 			return reason;
907 		}
908 
909 		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
910 		    (READ_ONCE(idev->cnf.forwarding) &&
911 		     (READ_ONCE(net->ipv6.devconf_all->proxy_ndp) ||
912 		      READ_ONCE(idev->cnf.proxy_ndp)) &&
913 		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
914 			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
915 			    skb->pkt_type != PACKET_HOST &&
916 			    inc &&
917 			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
918 				/*
919 				 * for anycast or proxy,
920 				 * sender should delay its response
921 				 * by a random time between 0 and
922 				 * MAX_ANYCAST_DELAY_TIME seconds.
923 				 * (RFC2461) -- yoshfuji
924 				 */
925 				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
926 				if (n)
927 					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
928 				goto out;
929 			}
930 		} else {
931 			SKB_DR_SET(reason, IPV6_NDISC_NS_OTHERHOST);
932 			goto out;
933 		}
934 	}
935 
936 	if (is_router < 0)
937 		is_router = READ_ONCE(idev->cnf.forwarding);
938 
939 	if (dad) {
940 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
941 			      !!is_router, false, (ifp != NULL), true);
942 		goto out;
943 	}
944 
945 	if (inc)
946 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
947 	else
948 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
949 
950 	/*
951 	 *	update / create cache entry
952 	 *	for the source address
953 	 */
954 	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
955 			       !inc || lladdr || !dev->addr_len);
956 	if (neigh)
957 		ndisc_update(dev, neigh, lladdr, NUD_STALE,
958 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
959 			     NEIGH_UPDATE_F_OVERRIDE,
960 			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
961 	if (neigh || !dev->header_ops) {
962 		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
963 			      true, (ifp != NULL && inc), inc);
964 		if (neigh)
965 			neigh_release(neigh);
966 		reason = SKB_CONSUMED;
967 	}
968 
969 out:
970 	if (ifp)
971 		in6_ifa_put(ifp);
972 	else
973 		in6_dev_put(idev);
974 	return reason;
975 }
976 
accept_untracked_na(struct net_device * dev,struct in6_addr * saddr)977 static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
978 {
979 	struct inet6_dev *idev = __in6_dev_get(dev);
980 
981 	switch (READ_ONCE(idev->cnf.accept_untracked_na)) {
982 	case 0: /* Don't accept untracked na (absent in neighbor cache) */
983 		return 0;
984 	case 1: /* Create new entries from na if currently untracked */
985 		return 1;
986 	case 2: /* Create new entries from untracked na only if saddr is in the
987 		 * same subnet as an address configured on the interface that
988 		 * received the na
989 		 */
990 		return !!ipv6_chk_prefix(saddr, dev);
991 	default:
992 		return 0;
993 	}
994 }
995 
ndisc_recv_na(struct sk_buff * skb)996 static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)
997 {
998 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
999 	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1000 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
1001 	u8 *lladdr = NULL;
1002 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1003 				    offsetof(struct nd_msg, opt));
1004 	struct ndisc_options ndopts;
1005 	struct net_device *dev = skb->dev;
1006 	struct inet6_dev *idev = __in6_dev_get(dev);
1007 	struct inet6_ifaddr *ifp;
1008 	struct neighbour *neigh;
1009 	SKB_DR(reason);
1010 	u8 new_state;
1011 
1012 	if (skb->len < sizeof(struct nd_msg))
1013 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1014 
1015 	if (ipv6_addr_is_multicast(&msg->target)) {
1016 		ND_PRINTK(2, warn, "NA: target address is multicast\n");
1017 		return reason;
1018 	}
1019 
1020 	if (ipv6_addr_is_multicast(daddr) &&
1021 	    msg->icmph.icmp6_solicited) {
1022 		ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
1023 		return reason;
1024 	}
1025 
1026 	/* For some 802.11 wireless deployments (and possibly other networks),
1027 	 * there will be a NA proxy and unsolicitd packets are attacks
1028 	 * and thus should not be accepted.
1029 	 * drop_unsolicited_na takes precedence over accept_untracked_na
1030 	 */
1031 	if (!msg->icmph.icmp6_solicited && idev &&
1032 	    READ_ONCE(idev->cnf.drop_unsolicited_na))
1033 		return reason;
1034 
1035 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
1036 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1037 
1038 	if (ndopts.nd_opts_tgt_lladdr) {
1039 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1040 		if (!lladdr) {
1041 			ND_PRINTK(2, warn,
1042 				  "NA: invalid link-layer address length\n");
1043 			return reason;
1044 		}
1045 	}
1046 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1047 	if (ifp) {
1048 		if (skb->pkt_type != PACKET_LOOPBACK
1049 		    && (ifp->flags & IFA_F_TENTATIVE)) {
1050 				addrconf_dad_failure(skb, ifp);
1051 				return reason;
1052 		}
1053 		/* What should we make now? The advertisement
1054 		   is invalid, but ndisc specs say nothing
1055 		   about it. It could be misconfiguration, or
1056 		   an smart proxy agent tries to help us :-)
1057 
1058 		   We should not print the error if NA has been
1059 		   received from loopback - it is just our own
1060 		   unsolicited advertisement.
1061 		 */
1062 		if (skb->pkt_type != PACKET_LOOPBACK)
1063 			ND_PRINTK(1, warn,
1064 				  "NA: %pM advertised our address %pI6c on %s!\n",
1065 				  eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1066 		in6_ifa_put(ifp);
1067 		return reason;
1068 	}
1069 
1070 	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1071 
1072 	/* RFC 9131 updates original Neighbour Discovery RFC 4861.
1073 	 * NAs with Target LL Address option without a corresponding
1074 	 * entry in the neighbour cache can now create a STALE neighbour
1075 	 * cache entry on routers.
1076 	 *
1077 	 *   entry accept  fwding  solicited        behaviour
1078 	 * ------- ------  ------  ---------    ----------------------
1079 	 * present      X       X         0     Set state to STALE
1080 	 * present      X       X         1     Set state to REACHABLE
1081 	 *  absent      0       X         X     Do nothing
1082 	 *  absent      1       0         X     Do nothing
1083 	 *  absent      1       1         X     Add a new STALE entry
1084 	 *
1085 	 * Note that we don't do a (daddr == all-routers-mcast) check.
1086 	 */
1087 	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1088 	if (!neigh && lladdr && idev && READ_ONCE(idev->cnf.forwarding)) {
1089 		if (accept_untracked_na(dev, saddr)) {
1090 			neigh = neigh_create(&nd_tbl, &msg->target, dev);
1091 			new_state = NUD_STALE;
1092 		}
1093 	}
1094 
1095 	if (neigh && !IS_ERR(neigh)) {
1096 		u8 old_flags = neigh->flags;
1097 		struct net *net = dev_net(dev);
1098 
1099 		if (READ_ONCE(neigh->nud_state) & NUD_FAILED)
1100 			goto out;
1101 
1102 		/*
1103 		 * Don't update the neighbor cache entry on a proxy NA from
1104 		 * ourselves because either the proxied node is off link or it
1105 		 * has already sent a NA to us.
1106 		 */
1107 		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1108 		    READ_ONCE(net->ipv6.devconf_all->forwarding) &&
1109 		    READ_ONCE(net->ipv6.devconf_all->proxy_ndp) &&
1110 		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1111 			/* XXX: idev->cnf.proxy_ndp */
1112 			goto out;
1113 		}
1114 
1115 		ndisc_update(dev, neigh, lladdr,
1116 			     new_state,
1117 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1118 			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1119 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1120 			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1121 			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1122 
1123 		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1124 			/*
1125 			 * Change: router to host
1126 			 */
1127 			rt6_clean_tohost(dev_net(dev),  saddr);
1128 		}
1129 		reason = SKB_CONSUMED;
1130 out:
1131 		neigh_release(neigh);
1132 	}
1133 	return reason;
1134 }
1135 
ndisc_recv_rs(struct sk_buff * skb)1136 static enum skb_drop_reason ndisc_recv_rs(struct sk_buff *skb)
1137 {
1138 	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1139 	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1140 	struct neighbour *neigh;
1141 	struct inet6_dev *idev;
1142 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1143 	struct ndisc_options ndopts;
1144 	u8 *lladdr = NULL;
1145 	SKB_DR(reason);
1146 
1147 	if (skb->len < sizeof(*rs_msg))
1148 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1149 
1150 	idev = __in6_dev_get(skb->dev);
1151 	if (!idev) {
1152 		ND_PRINTK(1, err, "RS: can't find in6 device\n");
1153 		return reason;
1154 	}
1155 
1156 	/* Don't accept RS if we're not in router mode */
1157 	if (!READ_ONCE(idev->cnf.forwarding))
1158 		goto out;
1159 
1160 	/*
1161 	 * Don't update NCE if src = ::;
1162 	 * this implies that the source node has no ip address assigned yet.
1163 	 */
1164 	if (ipv6_addr_any(saddr))
1165 		goto out;
1166 
1167 	/* Parse ND options */
1168 	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts))
1169 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1170 
1171 	if (ndopts.nd_opts_src_lladdr) {
1172 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1173 					     skb->dev);
1174 		if (!lladdr)
1175 			goto out;
1176 	}
1177 
1178 	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1179 	if (neigh) {
1180 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1181 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1182 			     NEIGH_UPDATE_F_OVERRIDE|
1183 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1184 			     NDISC_ROUTER_SOLICITATION, &ndopts);
1185 		neigh_release(neigh);
1186 		reason = SKB_CONSUMED;
1187 	}
1188 out:
1189 	return reason;
1190 }
1191 
ndisc_ra_useropt(struct sk_buff * ra,struct nd_opt_hdr * opt)1192 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1193 {
1194 	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1195 	struct sk_buff *skb;
1196 	struct nlmsghdr *nlh;
1197 	struct nduseroptmsg *ndmsg;
1198 	struct net *net = dev_net(ra->dev);
1199 	int err;
1200 	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1201 				    + (opt->nd_opt_len << 3));
1202 	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1203 
1204 	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1205 	if (!skb) {
1206 		err = -ENOBUFS;
1207 		goto errout;
1208 	}
1209 
1210 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1211 	if (!nlh) {
1212 		goto nla_put_failure;
1213 	}
1214 
1215 	ndmsg = nlmsg_data(nlh);
1216 	ndmsg->nduseropt_family = AF_INET6;
1217 	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1218 	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1219 	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1220 	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1221 
1222 	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1223 
1224 	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1225 		goto nla_put_failure;
1226 	nlmsg_end(skb, nlh);
1227 
1228 	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1229 	return;
1230 
1231 nla_put_failure:
1232 	nlmsg_free(skb);
1233 	err = -EMSGSIZE;
1234 errout:
1235 	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1236 }
1237 
ndisc_router_discovery(struct sk_buff * skb)1238 static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb)
1239 {
1240 	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1241 	bool send_ifinfo_notify = false;
1242 	struct neighbour *neigh = NULL;
1243 	struct ndisc_options ndopts;
1244 	struct fib6_info *rt = NULL;
1245 	struct inet6_dev *in6_dev;
1246 	struct fib6_table *table;
1247 	u32 defrtr_usr_metric;
1248 	unsigned int pref = 0;
1249 	__u32 old_if_flags;
1250 	struct net *net;
1251 	SKB_DR(reason);
1252 	int lifetime;
1253 	int optlen;
1254 
1255 	__u8 *opt = (__u8 *)(ra_msg + 1);
1256 
1257 	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1258 		sizeof(struct ra_msg);
1259 
1260 	ND_PRINTK(2, info,
1261 		  "RA: %s, dev: %s\n",
1262 		  __func__, skb->dev->name);
1263 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1264 		ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1265 		return reason;
1266 	}
1267 	if (optlen < 0)
1268 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1269 
1270 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1271 	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1272 		ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1273 		return reason;
1274 	}
1275 #endif
1276 
1277 	in6_dev = __in6_dev_get(skb->dev);
1278 	if (!in6_dev) {
1279 		ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1280 			  skb->dev->name);
1281 		return reason;
1282 	}
1283 
1284 	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts))
1285 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1286 
1287 	if (!ipv6_accept_ra(in6_dev)) {
1288 		ND_PRINTK(2, info,
1289 			  "RA: %s, did not accept ra for dev: %s\n",
1290 			  __func__, skb->dev->name);
1291 		goto skip_linkparms;
1292 	}
1293 
1294 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1295 	/* skip link-specific parameters from interior routers */
1296 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1297 		ND_PRINTK(2, info,
1298 			  "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1299 			  __func__, skb->dev->name);
1300 		goto skip_linkparms;
1301 	}
1302 #endif
1303 
1304 	if (in6_dev->if_flags & IF_RS_SENT) {
1305 		/*
1306 		 *	flag that an RA was received after an RS was sent
1307 		 *	out on this interface.
1308 		 */
1309 		in6_dev->if_flags |= IF_RA_RCVD;
1310 	}
1311 
1312 	/*
1313 	 * Remember the managed/otherconf flags from most recently
1314 	 * received RA message (RFC 2462) -- yoshfuji
1315 	 */
1316 	old_if_flags = in6_dev->if_flags;
1317 	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1318 				IF_RA_OTHERCONF)) |
1319 				(ra_msg->icmph.icmp6_addrconf_managed ?
1320 					IF_RA_MANAGED : 0) |
1321 				(ra_msg->icmph.icmp6_addrconf_other ?
1322 					IF_RA_OTHERCONF : 0);
1323 
1324 	if (old_if_flags != in6_dev->if_flags)
1325 		send_ifinfo_notify = true;
1326 
1327 	if (!READ_ONCE(in6_dev->cnf.accept_ra_defrtr)) {
1328 		ND_PRINTK(2, info,
1329 			  "RA: %s, defrtr is false for dev: %s\n",
1330 			  __func__, skb->dev->name);
1331 		goto skip_defrtr;
1332 	}
1333 
1334 	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1335 	if (lifetime != 0 &&
1336 	    lifetime < READ_ONCE(in6_dev->cnf.accept_ra_min_lft)) {
1337 		ND_PRINTK(2, info,
1338 			  "RA: router lifetime (%ds) is too short: %s\n",
1339 			  lifetime, skb->dev->name);
1340 		goto skip_defrtr;
1341 	}
1342 
1343 	/* Do not accept RA with source-addr found on local machine unless
1344 	 * accept_ra_from_local is set to true.
1345 	 */
1346 	net = dev_net(in6_dev->dev);
1347 	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1348 	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1349 		ND_PRINTK(2, info,
1350 			  "RA from local address detected on dev: %s: default router ignored\n",
1351 			  skb->dev->name);
1352 		goto skip_defrtr;
1353 	}
1354 
1355 #ifdef CONFIG_IPV6_ROUTER_PREF
1356 	pref = ra_msg->icmph.icmp6_router_pref;
1357 	/* 10b is handled as if it were 00b (medium) */
1358 	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1359 	    !READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref))
1360 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1361 #endif
1362 	/* routes added from RAs do not use nexthop objects */
1363 	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1364 	if (rt) {
1365 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1366 					 rt->fib6_nh->fib_nh_dev, NULL,
1367 					  &ipv6_hdr(skb)->saddr);
1368 		if (!neigh) {
1369 			ND_PRINTK(0, err,
1370 				  "RA: %s got default router without neighbour\n",
1371 				  __func__);
1372 			fib6_info_release(rt);
1373 			return reason;
1374 		}
1375 	}
1376 	/* Set default route metric as specified by user */
1377 	defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1378 	/* delete the route if lifetime is 0 or if metric needs change */
1379 	if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1380 		ip6_del_rt(net, rt, false);
1381 		rt = NULL;
1382 	}
1383 
1384 	ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, metric: %d, for dev: %s\n",
1385 		  rt, lifetime, defrtr_usr_metric, skb->dev->name);
1386 	if (!rt && lifetime) {
1387 		ND_PRINTK(3, info, "RA: adding default router\n");
1388 
1389 		if (neigh)
1390 			neigh_release(neigh);
1391 
1392 		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1393 					 skb->dev, pref, defrtr_usr_metric,
1394 					 lifetime);
1395 		if (!rt) {
1396 			ND_PRINTK(0, err,
1397 				  "RA: %s failed to add default route\n",
1398 				  __func__);
1399 			return reason;
1400 		}
1401 
1402 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1403 					 rt->fib6_nh->fib_nh_dev, NULL,
1404 					  &ipv6_hdr(skb)->saddr);
1405 		if (!neigh) {
1406 			ND_PRINTK(0, err,
1407 				  "RA: %s got default router without neighbour\n",
1408 				  __func__);
1409 			fib6_info_release(rt);
1410 			return reason;
1411 		}
1412 		neigh->flags |= NTF_ROUTER;
1413 	} else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1414 		struct nl_info nlinfo = {
1415 			.nl_net = net,
1416 		};
1417 		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1418 		inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1419 	}
1420 
1421 	if (rt) {
1422 		table = rt->fib6_table;
1423 		spin_lock_bh(&table->tb6_lock);
1424 
1425 		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1426 		fib6_add_gc_list(rt);
1427 
1428 		spin_unlock_bh(&table->tb6_lock);
1429 	}
1430 	if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) < 256 &&
1431 	    ra_msg->icmph.icmp6_hop_limit) {
1432 		if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) <=
1433 		    ra_msg->icmph.icmp6_hop_limit) {
1434 			WRITE_ONCE(in6_dev->cnf.hop_limit,
1435 				   ra_msg->icmph.icmp6_hop_limit);
1436 			fib6_metric_set(rt, RTAX_HOPLIMIT,
1437 					ra_msg->icmph.icmp6_hop_limit);
1438 		} else {
1439 			ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1440 		}
1441 	}
1442 
1443 skip_defrtr:
1444 
1445 	/*
1446 	 *	Update Reachable Time and Retrans Timer
1447 	 */
1448 
1449 	if (in6_dev->nd_parms) {
1450 		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1451 
1452 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1453 			rtime = (rtime*HZ)/1000;
1454 			if (rtime < HZ/100)
1455 				rtime = HZ/100;
1456 			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1457 			in6_dev->tstamp = jiffies;
1458 			send_ifinfo_notify = true;
1459 		}
1460 
1461 		rtime = ntohl(ra_msg->reachable_time);
1462 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1463 			rtime = (rtime*HZ)/1000;
1464 
1465 			if (rtime < HZ/10)
1466 				rtime = HZ/10;
1467 
1468 			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1469 				NEIGH_VAR_SET(in6_dev->nd_parms,
1470 					      BASE_REACHABLE_TIME, rtime);
1471 				NEIGH_VAR_SET(in6_dev->nd_parms,
1472 					      GC_STALETIME, 3 * rtime);
1473 				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1474 				in6_dev->tstamp = jiffies;
1475 				send_ifinfo_notify = true;
1476 			}
1477 		}
1478 	}
1479 
1480 skip_linkparms:
1481 
1482 	/*
1483 	 *	Process options.
1484 	 */
1485 
1486 	if (!neigh)
1487 		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1488 				       skb->dev, 1);
1489 	if (neigh) {
1490 		u8 *lladdr = NULL;
1491 		if (ndopts.nd_opts_src_lladdr) {
1492 			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1493 						     skb->dev);
1494 			if (!lladdr) {
1495 				ND_PRINTK(2, warn,
1496 					  "RA: invalid link-layer address length\n");
1497 				goto out;
1498 			}
1499 		}
1500 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1501 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1502 			     NEIGH_UPDATE_F_OVERRIDE|
1503 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1504 			     NEIGH_UPDATE_F_ISROUTER,
1505 			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1506 		reason = SKB_CONSUMED;
1507 	}
1508 
1509 	if (!ipv6_accept_ra(in6_dev)) {
1510 		ND_PRINTK(2, info,
1511 			  "RA: %s, accept_ra is false for dev: %s\n",
1512 			  __func__, skb->dev->name);
1513 		goto out;
1514 	}
1515 
1516 #ifdef CONFIG_IPV6_ROUTE_INFO
1517 	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1518 	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1519 			  in6_dev->dev, 0)) {
1520 		ND_PRINTK(2, info,
1521 			  "RA from local address detected on dev: %s: router info ignored.\n",
1522 			  skb->dev->name);
1523 		goto skip_routeinfo;
1524 	}
1525 
1526 	if (READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref) && ndopts.nd_opts_ri) {
1527 		struct nd_opt_hdr *p;
1528 		for (p = ndopts.nd_opts_ri;
1529 		     p;
1530 		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1531 			struct route_info *ri = (struct route_info *)p;
1532 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1533 			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1534 			    ri->prefix_len == 0)
1535 				continue;
1536 #endif
1537 			if (ri->prefix_len == 0 &&
1538 			    !READ_ONCE(in6_dev->cnf.accept_ra_defrtr))
1539 				continue;
1540 			if (ri->lifetime != 0 &&
1541 			    ntohl(ri->lifetime) < READ_ONCE(in6_dev->cnf.accept_ra_min_lft))
1542 				continue;
1543 			if (ri->prefix_len < READ_ONCE(in6_dev->cnf.accept_ra_rt_info_min_plen))
1544 				continue;
1545 			if (ri->prefix_len > READ_ONCE(in6_dev->cnf.accept_ra_rt_info_max_plen))
1546 				continue;
1547 			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1548 				      &ipv6_hdr(skb)->saddr);
1549 		}
1550 	}
1551 
1552 skip_routeinfo:
1553 #endif
1554 
1555 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1556 	/* skip link-specific ndopts from interior routers */
1557 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1558 		ND_PRINTK(2, info,
1559 			  "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1560 			  __func__, skb->dev->name);
1561 		goto out;
1562 	}
1563 #endif
1564 
1565 	if (READ_ONCE(in6_dev->cnf.accept_ra_pinfo) && ndopts.nd_opts_pi) {
1566 		struct nd_opt_hdr *p;
1567 		for (p = ndopts.nd_opts_pi;
1568 		     p;
1569 		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1570 			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1571 					    (p->nd_opt_len) << 3,
1572 					    ndopts.nd_opts_src_lladdr != NULL);
1573 		}
1574 	}
1575 
1576 	if (ndopts.nd_opts_mtu && READ_ONCE(in6_dev->cnf.accept_ra_mtu)) {
1577 		__be32 n;
1578 		u32 mtu;
1579 
1580 		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1581 		mtu = ntohl(n);
1582 
1583 		if (in6_dev->ra_mtu != mtu) {
1584 			in6_dev->ra_mtu = mtu;
1585 			send_ifinfo_notify = true;
1586 		}
1587 
1588 		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1589 			ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1590 		} else if (READ_ONCE(in6_dev->cnf.mtu6) != mtu) {
1591 			WRITE_ONCE(in6_dev->cnf.mtu6, mtu);
1592 			fib6_metric_set(rt, RTAX_MTU, mtu);
1593 			rt6_mtu_change(skb->dev, mtu);
1594 		}
1595 	}
1596 
1597 	if (ndopts.nd_useropts) {
1598 		struct nd_opt_hdr *p;
1599 		for (p = ndopts.nd_useropts;
1600 		     p;
1601 		     p = ndisc_next_useropt(skb->dev, p,
1602 					    ndopts.nd_useropts_end)) {
1603 			ndisc_ra_useropt(skb, p);
1604 		}
1605 	}
1606 
1607 	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1608 		ND_PRINTK(2, warn, "RA: invalid RA options\n");
1609 	}
1610 out:
1611 	/* Send a notify if RA changed managed/otherconf flags or
1612 	 * timer settings or ra_mtu value
1613 	 */
1614 	if (send_ifinfo_notify)
1615 		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1616 
1617 	fib6_info_release(rt);
1618 	if (neigh)
1619 		neigh_release(neigh);
1620 	return reason;
1621 }
1622 
ndisc_redirect_rcv(struct sk_buff * skb)1623 static enum skb_drop_reason ndisc_redirect_rcv(struct sk_buff *skb)
1624 {
1625 	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1626 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1627 				    offsetof(struct rd_msg, opt));
1628 	struct ndisc_options ndopts;
1629 	SKB_DR(reason);
1630 	u8 *hdr;
1631 
1632 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1633 	switch (skb->ndisc_nodetype) {
1634 	case NDISC_NODETYPE_HOST:
1635 	case NDISC_NODETYPE_NODEFAULT:
1636 		ND_PRINTK(2, warn,
1637 			  "Redirect: from host or unauthorized router\n");
1638 		return reason;
1639 	}
1640 #endif
1641 
1642 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1643 		ND_PRINTK(2, warn,
1644 			  "Redirect: source address is not link-local\n");
1645 		return reason;
1646 	}
1647 
1648 	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1649 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1650 
1651 	if (!ndopts.nd_opts_rh) {
1652 		ip6_redirect_no_header(skb, dev_net(skb->dev),
1653 					skb->dev->ifindex);
1654 		return reason;
1655 	}
1656 
1657 	hdr = (u8 *)ndopts.nd_opts_rh;
1658 	hdr += 8;
1659 	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1660 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1661 
1662 	return icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1663 }
1664 
ndisc_fill_redirect_hdr_option(struct sk_buff * skb,struct sk_buff * orig_skb,int rd_len)1665 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1666 					   struct sk_buff *orig_skb,
1667 					   int rd_len)
1668 {
1669 	u8 *opt = skb_put(skb, rd_len);
1670 
1671 	memset(opt, 0, 8);
1672 	*(opt++) = ND_OPT_REDIRECT_HDR;
1673 	*(opt++) = (rd_len >> 3);
1674 	opt += 6;
1675 
1676 	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1677 		      rd_len - 8);
1678 }
1679 
ndisc_send_redirect(struct sk_buff * skb,const struct in6_addr * target)1680 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1681 {
1682 	struct net_device *dev = skb->dev;
1683 	struct net *net = dev_net(dev);
1684 	struct sock *sk = net->ipv6.ndisc_sk;
1685 	int optlen = 0;
1686 	struct inet_peer *peer;
1687 	struct sk_buff *buff;
1688 	struct rd_msg *msg;
1689 	struct in6_addr saddr_buf;
1690 	struct rt6_info *rt;
1691 	struct dst_entry *dst;
1692 	struct flowi6 fl6;
1693 	int rd_len;
1694 	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1695 	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1696 	bool ret;
1697 
1698 	if (netif_is_l3_master(skb->dev)) {
1699 		dev = dev_get_by_index_rcu(dev_net(skb->dev), IPCB(skb)->iif);
1700 		if (!dev)
1701 			return;
1702 	}
1703 
1704 	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1705 		ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1706 			  dev->name);
1707 		return;
1708 	}
1709 
1710 	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1711 	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1712 		ND_PRINTK(2, warn,
1713 			  "Redirect: target address is not link-local unicast\n");
1714 		return;
1715 	}
1716 
1717 	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1718 			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1719 
1720 	dst = ip6_route_output(net, NULL, &fl6);
1721 	if (dst->error) {
1722 		dst_release(dst);
1723 		return;
1724 	}
1725 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1726 	if (IS_ERR(dst))
1727 		return;
1728 
1729 	rt = dst_rt6_info(dst);
1730 
1731 	if (rt->rt6i_flags & RTF_GATEWAY) {
1732 		ND_PRINTK(2, warn,
1733 			  "Redirect: destination is not a neighbour\n");
1734 		goto release;
1735 	}
1736 
1737 	rcu_read_lock();
1738 	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr);
1739 	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1740 	rcu_read_unlock();
1741 
1742 	if (!ret)
1743 		goto release;
1744 
1745 	if (dev->addr_len) {
1746 		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1747 		if (!neigh) {
1748 			ND_PRINTK(2, warn,
1749 				  "Redirect: no neigh for target address\n");
1750 			goto release;
1751 		}
1752 
1753 		read_lock_bh(&neigh->lock);
1754 		if (neigh->nud_state & NUD_VALID) {
1755 			memcpy(ha_buf, neigh->ha, dev->addr_len);
1756 			read_unlock_bh(&neigh->lock);
1757 			ha = ha_buf;
1758 			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1759 								ops_data_buf,
1760 								&ops_data);
1761 		} else
1762 			read_unlock_bh(&neigh->lock);
1763 
1764 		neigh_release(neigh);
1765 	}
1766 
1767 	rd_len = min_t(unsigned int,
1768 		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1769 		       skb->len + 8);
1770 	rd_len &= ~0x7;
1771 	optlen += rd_len;
1772 
1773 	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1774 	if (!buff)
1775 		goto release;
1776 
1777 	msg = skb_put(buff, sizeof(*msg));
1778 	*msg = (struct rd_msg) {
1779 		.icmph = {
1780 			.icmp6_type = NDISC_REDIRECT,
1781 		},
1782 		.target = *target,
1783 		.dest = ipv6_hdr(skb)->daddr,
1784 	};
1785 
1786 	/*
1787 	 *	include target_address option
1788 	 */
1789 
1790 	if (ha)
1791 		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1792 
1793 	/*
1794 	 *	build redirect option and copy skb over to the new packet.
1795 	 */
1796 
1797 	if (rd_len)
1798 		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1799 
1800 	skb_dst_set(buff, dst);
1801 	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1802 	return;
1803 
1804 release:
1805 	dst_release(dst);
1806 }
1807 
pndisc_redo(struct sk_buff * skb)1808 static void pndisc_redo(struct sk_buff *skb)
1809 {
1810 	enum skb_drop_reason reason = ndisc_recv_ns(skb);
1811 
1812 	kfree_skb_reason(skb, reason);
1813 }
1814 
ndisc_is_multicast(const void * pkey)1815 static int ndisc_is_multicast(const void *pkey)
1816 {
1817 	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1818 }
1819 
ndisc_suppress_frag_ndisc(struct sk_buff * skb)1820 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1821 {
1822 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1823 
1824 	if (!idev)
1825 		return true;
1826 	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1827 	    READ_ONCE(idev->cnf.suppress_frag_ndisc)) {
1828 		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1829 		return true;
1830 	}
1831 	return false;
1832 }
1833 
ndisc_rcv(struct sk_buff * skb)1834 enum skb_drop_reason ndisc_rcv(struct sk_buff *skb)
1835 {
1836 	struct nd_msg *msg;
1837 	SKB_DR(reason);
1838 
1839 	if (ndisc_suppress_frag_ndisc(skb))
1840 		return SKB_DROP_REASON_IPV6_NDISC_FRAG;
1841 
1842 	if (skb_linearize(skb))
1843 		return SKB_DROP_REASON_NOMEM;
1844 
1845 	msg = (struct nd_msg *)skb_transport_header(skb);
1846 
1847 	__skb_push(skb, skb->data - skb_transport_header(skb));
1848 
1849 	if (ipv6_hdr(skb)->hop_limit != 255) {
1850 		ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1851 			  ipv6_hdr(skb)->hop_limit);
1852 		return SKB_DROP_REASON_IPV6_NDISC_HOP_LIMIT;
1853 	}
1854 
1855 	if (msg->icmph.icmp6_code != 0) {
1856 		ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1857 			  msg->icmph.icmp6_code);
1858 		return SKB_DROP_REASON_IPV6_NDISC_BAD_CODE;
1859 	}
1860 
1861 	switch (msg->icmph.icmp6_type) {
1862 	case NDISC_NEIGHBOUR_SOLICITATION:
1863 		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1864 		reason = ndisc_recv_ns(skb);
1865 		break;
1866 
1867 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1868 		reason = ndisc_recv_na(skb);
1869 		break;
1870 
1871 	case NDISC_ROUTER_SOLICITATION:
1872 		reason = ndisc_recv_rs(skb);
1873 		break;
1874 
1875 	case NDISC_ROUTER_ADVERTISEMENT:
1876 		reason = ndisc_router_discovery(skb);
1877 		break;
1878 
1879 	case NDISC_REDIRECT:
1880 		reason = ndisc_redirect_rcv(skb);
1881 		break;
1882 	}
1883 
1884 	return reason;
1885 }
1886 
ndisc_netdev_event(struct notifier_block * this,unsigned long event,void * ptr)1887 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1888 {
1889 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1890 	struct netdev_notifier_change_info *change_info;
1891 	struct net *net = dev_net(dev);
1892 	struct inet6_dev *idev;
1893 	bool evict_nocarrier;
1894 
1895 	switch (event) {
1896 	case NETDEV_CHANGEADDR:
1897 		neigh_changeaddr(&nd_tbl, dev);
1898 		fib6_run_gc(0, net, false);
1899 		fallthrough;
1900 	case NETDEV_UP:
1901 		idev = in6_dev_get(dev);
1902 		if (!idev)
1903 			break;
1904 		if (READ_ONCE(idev->cnf.ndisc_notify) ||
1905 		    READ_ONCE(net->ipv6.devconf_all->ndisc_notify))
1906 			ndisc_send_unsol_na(dev);
1907 		in6_dev_put(idev);
1908 		break;
1909 	case NETDEV_CHANGE:
1910 		idev = in6_dev_get(dev);
1911 		if (!idev)
1912 			evict_nocarrier = true;
1913 		else {
1914 			evict_nocarrier = READ_ONCE(idev->cnf.ndisc_evict_nocarrier) &&
1915 					  READ_ONCE(net->ipv6.devconf_all->ndisc_evict_nocarrier);
1916 			in6_dev_put(idev);
1917 		}
1918 
1919 		change_info = ptr;
1920 		if (change_info->flags_changed & IFF_NOARP)
1921 			neigh_changeaddr(&nd_tbl, dev);
1922 		if (evict_nocarrier && !netif_carrier_ok(dev))
1923 			neigh_carrier_down(&nd_tbl, dev);
1924 		break;
1925 	case NETDEV_DOWN:
1926 		neigh_ifdown(&nd_tbl, dev);
1927 		fib6_run_gc(0, net, false);
1928 		break;
1929 	case NETDEV_NOTIFY_PEERS:
1930 		ndisc_send_unsol_na(dev);
1931 		break;
1932 	default:
1933 		break;
1934 	}
1935 
1936 	return NOTIFY_DONE;
1937 }
1938 
1939 static struct notifier_block ndisc_netdev_notifier = {
1940 	.notifier_call = ndisc_netdev_event,
1941 	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1942 };
1943 
1944 #ifdef CONFIG_SYSCTL
ndisc_warn_deprecated_sysctl(const struct ctl_table * ctl,const char * func,const char * dev_name)1945 static void ndisc_warn_deprecated_sysctl(const struct ctl_table *ctl,
1946 					 const char *func, const char *dev_name)
1947 {
1948 	static char warncomm[TASK_COMM_LEN];
1949 	static int warned;
1950 	if (strcmp(warncomm, current->comm) && warned < 5) {
1951 		strscpy(warncomm, current->comm);
1952 		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1953 			warncomm, func,
1954 			dev_name, ctl->procname,
1955 			dev_name, ctl->procname);
1956 		warned++;
1957 	}
1958 }
1959 
ndisc_ifinfo_sysctl_change(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)1960 int ndisc_ifinfo_sysctl_change(const struct ctl_table *ctl, int write, void *buffer,
1961 		size_t *lenp, loff_t *ppos)
1962 {
1963 	struct net_device *dev = ctl->extra1;
1964 	struct inet6_dev *idev;
1965 	int ret;
1966 
1967 	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1968 	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1969 		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1970 
1971 	if (strcmp(ctl->procname, "retrans_time") == 0)
1972 		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1973 
1974 	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1975 		ret = neigh_proc_dointvec_jiffies(ctl, write,
1976 						  buffer, lenp, ppos);
1977 
1978 	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1979 		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1980 		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1981 						     buffer, lenp, ppos);
1982 	else
1983 		ret = -1;
1984 
1985 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1986 		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1987 			idev->nd_parms->reachable_time =
1988 					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1989 		WRITE_ONCE(idev->tstamp, jiffies);
1990 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1991 		in6_dev_put(idev);
1992 	}
1993 	return ret;
1994 }
1995 
1996 
1997 #endif
1998 
ndisc_net_init(struct net * net)1999 static int __net_init ndisc_net_init(struct net *net)
2000 {
2001 	struct ipv6_pinfo *np;
2002 	struct sock *sk;
2003 	int err;
2004 
2005 	err = inet_ctl_sock_create(&sk, PF_INET6,
2006 				   SOCK_RAW, IPPROTO_ICMPV6, net);
2007 	if (err < 0) {
2008 		ND_PRINTK(0, err,
2009 			  "NDISC: Failed to initialize the control socket (err %d)\n",
2010 			  err);
2011 		return err;
2012 	}
2013 
2014 	net->ipv6.ndisc_sk = sk;
2015 
2016 	np = inet6_sk(sk);
2017 	np->hop_limit = 255;
2018 	/* Do not loopback ndisc messages */
2019 	inet6_clear_bit(MC6_LOOP, sk);
2020 
2021 	return 0;
2022 }
2023 
ndisc_net_exit(struct net * net)2024 static void __net_exit ndisc_net_exit(struct net *net)
2025 {
2026 	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
2027 }
2028 
2029 static struct pernet_operations ndisc_net_ops = {
2030 	.init = ndisc_net_init,
2031 	.exit = ndisc_net_exit,
2032 };
2033 
ndisc_init(void)2034 int __init ndisc_init(void)
2035 {
2036 	int err;
2037 
2038 	err = register_pernet_subsys(&ndisc_net_ops);
2039 	if (err)
2040 		return err;
2041 	/*
2042 	 * Initialize the neighbour table
2043 	 */
2044 	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2045 
2046 #ifdef CONFIG_SYSCTL
2047 	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2048 				    ndisc_ifinfo_sysctl_change);
2049 	if (err)
2050 		goto out_unregister_pernet;
2051 out:
2052 #endif
2053 	return err;
2054 
2055 #ifdef CONFIG_SYSCTL
2056 out_unregister_pernet:
2057 	unregister_pernet_subsys(&ndisc_net_ops);
2058 	goto out;
2059 #endif
2060 }
2061 
ndisc_late_init(void)2062 int __init ndisc_late_init(void)
2063 {
2064 	return register_netdevice_notifier(&ndisc_netdev_notifier);
2065 }
2066 
ndisc_late_cleanup(void)2067 void ndisc_late_cleanup(void)
2068 {
2069 	unregister_netdevice_notifier(&ndisc_netdev_notifier);
2070 }
2071 
ndisc_cleanup(void)2072 void ndisc_cleanup(void)
2073 {
2074 #ifdef CONFIG_SYSCTL
2075 	neigh_sysctl_unregister(&nd_tbl.parms);
2076 #endif
2077 	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2078 	unregister_pernet_subsys(&ndisc_net_ops);
2079 }
2080