xref: /openbmc/linux/net/ipv6/ndisc.c (revision fd5e9fccbd504c5179ab57ff695c610bca8809d6)
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_CAPTIVE_PORTAL ||
204 		opt->nd_opt_type == ND_OPT_PREF64 ||
205 		ndisc_ops_is_useropt(dev, opt->nd_opt_type);
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 ? 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, 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 |= 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 		    (idev->cnf.forwarding &&
911 		     (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
912 		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
913 			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
914 			    skb->pkt_type != PACKET_HOST &&
915 			    inc &&
916 			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
917 				/*
918 				 * for anycast or proxy,
919 				 * sender should delay its response
920 				 * by a random time between 0 and
921 				 * MAX_ANYCAST_DELAY_TIME seconds.
922 				 * (RFC2461) -- yoshfuji
923 				 */
924 				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
925 				if (n)
926 					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
927 				goto out;
928 			}
929 		} else {
930 			SKB_DR_SET(reason, IPV6_NDISC_NS_OTHERHOST);
931 			goto out;
932 		}
933 	}
934 
935 	if (is_router < 0)
936 		is_router = idev->cnf.forwarding;
937 
938 	if (dad) {
939 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
940 			      !!is_router, false, (ifp != NULL), true);
941 		goto out;
942 	}
943 
944 	if (inc)
945 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
946 	else
947 		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
948 
949 	/*
950 	 *	update / create cache entry
951 	 *	for the source address
952 	 */
953 	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
954 			       !inc || lladdr || !dev->addr_len);
955 	if (neigh)
956 		ndisc_update(dev, neigh, lladdr, NUD_STALE,
957 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
958 			     NEIGH_UPDATE_F_OVERRIDE,
959 			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
960 	if (neigh || !dev->header_ops) {
961 		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
962 			      true, (ifp != NULL && inc), inc);
963 		if (neigh)
964 			neigh_release(neigh);
965 		reason = SKB_CONSUMED;
966 	}
967 
968 out:
969 	if (ifp)
970 		in6_ifa_put(ifp);
971 	else
972 		in6_dev_put(idev);
973 	return reason;
974 }
975 
accept_untracked_na(struct net_device * dev,struct in6_addr * saddr)976 static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
977 {
978 	struct inet6_dev *idev = __in6_dev_get(dev);
979 
980 	switch (idev->cnf.accept_untracked_na) {
981 	case 0: /* Don't accept untracked na (absent in neighbor cache) */
982 		return 0;
983 	case 1: /* Create new entries from na if currently untracked */
984 		return 1;
985 	case 2: /* Create new entries from untracked na only if saddr is in the
986 		 * same subnet as an address configured on the interface that
987 		 * received the na
988 		 */
989 		return !!ipv6_chk_prefix(saddr, dev);
990 	default:
991 		return 0;
992 	}
993 }
994 
ndisc_recv_na(struct sk_buff * skb)995 static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)
996 {
997 	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
998 	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
999 	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
1000 	u8 *lladdr = NULL;
1001 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1002 				    offsetof(struct nd_msg, opt));
1003 	struct ndisc_options ndopts;
1004 	struct net_device *dev = skb->dev;
1005 	struct inet6_dev *idev = __in6_dev_get(dev);
1006 	struct inet6_ifaddr *ifp;
1007 	struct neighbour *neigh;
1008 	SKB_DR(reason);
1009 	u8 new_state;
1010 
1011 	if (skb->len < sizeof(struct nd_msg))
1012 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1013 
1014 	if (ipv6_addr_is_multicast(&msg->target)) {
1015 		ND_PRINTK(2, warn, "NA: target address is multicast\n");
1016 		return reason;
1017 	}
1018 
1019 	if (ipv6_addr_is_multicast(daddr) &&
1020 	    msg->icmph.icmp6_solicited) {
1021 		ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
1022 		return reason;
1023 	}
1024 
1025 	/* For some 802.11 wireless deployments (and possibly other networks),
1026 	 * there will be a NA proxy and unsolicitd packets are attacks
1027 	 * and thus should not be accepted.
1028 	 * drop_unsolicited_na takes precedence over accept_untracked_na
1029 	 */
1030 	if (!msg->icmph.icmp6_solicited && idev &&
1031 	    idev->cnf.drop_unsolicited_na)
1032 		return reason;
1033 
1034 	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
1035 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1036 
1037 	if (ndopts.nd_opts_tgt_lladdr) {
1038 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1039 		if (!lladdr) {
1040 			ND_PRINTK(2, warn,
1041 				  "NA: invalid link-layer address length\n");
1042 			return reason;
1043 		}
1044 	}
1045 	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1046 	if (ifp) {
1047 		if (skb->pkt_type != PACKET_LOOPBACK
1048 		    && (ifp->flags & IFA_F_TENTATIVE)) {
1049 				addrconf_dad_failure(skb, ifp);
1050 				return reason;
1051 		}
1052 		/* What should we make now? The advertisement
1053 		   is invalid, but ndisc specs say nothing
1054 		   about it. It could be misconfiguration, or
1055 		   an smart proxy agent tries to help us :-)
1056 
1057 		   We should not print the error if NA has been
1058 		   received from loopback - it is just our own
1059 		   unsolicited advertisement.
1060 		 */
1061 		if (skb->pkt_type != PACKET_LOOPBACK)
1062 			ND_PRINTK(1, warn,
1063 				  "NA: %pM advertised our address %pI6c on %s!\n",
1064 				  eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1065 		in6_ifa_put(ifp);
1066 		return reason;
1067 	}
1068 
1069 	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1070 
1071 	/* RFC 9131 updates original Neighbour Discovery RFC 4861.
1072 	 * NAs with Target LL Address option without a corresponding
1073 	 * entry in the neighbour cache can now create a STALE neighbour
1074 	 * cache entry on routers.
1075 	 *
1076 	 *   entry accept  fwding  solicited        behaviour
1077 	 * ------- ------  ------  ---------    ----------------------
1078 	 * present      X       X         0     Set state to STALE
1079 	 * present      X       X         1     Set state to REACHABLE
1080 	 *  absent      0       X         X     Do nothing
1081 	 *  absent      1       0         X     Do nothing
1082 	 *  absent      1       1         X     Add a new STALE entry
1083 	 *
1084 	 * Note that we don't do a (daddr == all-routers-mcast) check.
1085 	 */
1086 	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1087 	if (!neigh && lladdr && idev && idev->cnf.forwarding) {
1088 		if (accept_untracked_na(dev, saddr)) {
1089 			neigh = neigh_create(&nd_tbl, &msg->target, dev);
1090 			new_state = NUD_STALE;
1091 		}
1092 	}
1093 
1094 	if (neigh && !IS_ERR(neigh)) {
1095 		u8 old_flags = neigh->flags;
1096 		struct net *net = dev_net(dev);
1097 
1098 		if (READ_ONCE(neigh->nud_state) & NUD_FAILED)
1099 			goto out;
1100 
1101 		/*
1102 		 * Don't update the neighbor cache entry on a proxy NA from
1103 		 * ourselves because either the proxied node is off link or it
1104 		 * has already sent a NA to us.
1105 		 */
1106 		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1107 		    net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1108 		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1109 			/* XXX: idev->cnf.proxy_ndp */
1110 			goto out;
1111 		}
1112 
1113 		ndisc_update(dev, neigh, lladdr,
1114 			     new_state,
1115 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1116 			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1117 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1118 			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1119 			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1120 
1121 		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1122 			/*
1123 			 * Change: router to host
1124 			 */
1125 			rt6_clean_tohost(dev_net(dev),  saddr);
1126 		}
1127 		reason = SKB_CONSUMED;
1128 out:
1129 		neigh_release(neigh);
1130 	}
1131 	return reason;
1132 }
1133 
ndisc_recv_rs(struct sk_buff * skb)1134 static enum skb_drop_reason ndisc_recv_rs(struct sk_buff *skb)
1135 {
1136 	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1137 	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1138 	struct neighbour *neigh;
1139 	struct inet6_dev *idev;
1140 	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1141 	struct ndisc_options ndopts;
1142 	u8 *lladdr = NULL;
1143 	SKB_DR(reason);
1144 
1145 	if (skb->len < sizeof(*rs_msg))
1146 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1147 
1148 	idev = __in6_dev_get(skb->dev);
1149 	if (!idev) {
1150 		ND_PRINTK(1, err, "RS: can't find in6 device\n");
1151 		return reason;
1152 	}
1153 
1154 	/* Don't accept RS if we're not in router mode */
1155 	if (!idev->cnf.forwarding)
1156 		goto out;
1157 
1158 	/*
1159 	 * Don't update NCE if src = ::;
1160 	 * this implies that the source node has no ip address assigned yet.
1161 	 */
1162 	if (ipv6_addr_any(saddr))
1163 		goto out;
1164 
1165 	/* Parse ND options */
1166 	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts))
1167 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1168 
1169 	if (ndopts.nd_opts_src_lladdr) {
1170 		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1171 					     skb->dev);
1172 		if (!lladdr)
1173 			goto out;
1174 	}
1175 
1176 	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1177 	if (neigh) {
1178 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1179 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1180 			     NEIGH_UPDATE_F_OVERRIDE|
1181 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1182 			     NDISC_ROUTER_SOLICITATION, &ndopts);
1183 		neigh_release(neigh);
1184 		reason = SKB_CONSUMED;
1185 	}
1186 out:
1187 	return reason;
1188 }
1189 
ndisc_ra_useropt(struct sk_buff * ra,struct nd_opt_hdr * opt)1190 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1191 {
1192 	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1193 	struct sk_buff *skb;
1194 	struct nlmsghdr *nlh;
1195 	struct nduseroptmsg *ndmsg;
1196 	struct net *net = dev_net(ra->dev);
1197 	int err;
1198 	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1199 				    + (opt->nd_opt_len << 3));
1200 	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1201 
1202 	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1203 	if (!skb) {
1204 		err = -ENOBUFS;
1205 		goto errout;
1206 	}
1207 
1208 	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1209 	if (!nlh) {
1210 		goto nla_put_failure;
1211 	}
1212 
1213 	ndmsg = nlmsg_data(nlh);
1214 	ndmsg->nduseropt_family = AF_INET6;
1215 	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1216 	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1217 	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1218 	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1219 
1220 	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1221 
1222 	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1223 		goto nla_put_failure;
1224 	nlmsg_end(skb, nlh);
1225 
1226 	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1227 	return;
1228 
1229 nla_put_failure:
1230 	nlmsg_free(skb);
1231 	err = -EMSGSIZE;
1232 errout:
1233 	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1234 }
1235 
ndisc_router_discovery(struct sk_buff * skb)1236 static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb)
1237 {
1238 	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1239 	bool send_ifinfo_notify = false;
1240 	struct neighbour *neigh = NULL;
1241 	struct ndisc_options ndopts;
1242 	struct fib6_info *rt = NULL;
1243 	struct inet6_dev *in6_dev;
1244 	u32 defrtr_usr_metric;
1245 	unsigned int pref = 0;
1246 	__u32 old_if_flags;
1247 	struct net *net;
1248 	SKB_DR(reason);
1249 	int lifetime;
1250 	int optlen;
1251 
1252 	__u8 *opt = (__u8 *)(ra_msg + 1);
1253 
1254 	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1255 		sizeof(struct ra_msg);
1256 
1257 	ND_PRINTK(2, info,
1258 		  "RA: %s, dev: %s\n",
1259 		  __func__, skb->dev->name);
1260 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1261 		ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1262 		return reason;
1263 	}
1264 	if (optlen < 0)
1265 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1266 
1267 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1268 	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1269 		ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1270 		return reason;
1271 	}
1272 #endif
1273 
1274 	in6_dev = __in6_dev_get(skb->dev);
1275 	if (!in6_dev) {
1276 		ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1277 			  skb->dev->name);
1278 		return reason;
1279 	}
1280 
1281 	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts))
1282 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1283 
1284 	if (!ipv6_accept_ra(in6_dev)) {
1285 		ND_PRINTK(2, info,
1286 			  "RA: %s, did not accept ra for dev: %s\n",
1287 			  __func__, skb->dev->name);
1288 		goto skip_linkparms;
1289 	}
1290 
1291 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1292 	/* skip link-specific parameters from interior routers */
1293 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1294 		ND_PRINTK(2, info,
1295 			  "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1296 			  __func__, skb->dev->name);
1297 		goto skip_linkparms;
1298 	}
1299 #endif
1300 
1301 	if (in6_dev->if_flags & IF_RS_SENT) {
1302 		/*
1303 		 *	flag that an RA was received after an RS was sent
1304 		 *	out on this interface.
1305 		 */
1306 		in6_dev->if_flags |= IF_RA_RCVD;
1307 	}
1308 
1309 	/*
1310 	 * Remember the managed/otherconf flags from most recently
1311 	 * received RA message (RFC 2462) -- yoshfuji
1312 	 */
1313 	old_if_flags = in6_dev->if_flags;
1314 	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1315 				IF_RA_OTHERCONF)) |
1316 				(ra_msg->icmph.icmp6_addrconf_managed ?
1317 					IF_RA_MANAGED : 0) |
1318 				(ra_msg->icmph.icmp6_addrconf_other ?
1319 					IF_RA_OTHERCONF : 0);
1320 
1321 	if (old_if_flags != in6_dev->if_flags)
1322 		send_ifinfo_notify = true;
1323 
1324 	if (!in6_dev->cnf.accept_ra_defrtr) {
1325 		ND_PRINTK(2, info,
1326 			  "RA: %s, defrtr is false for dev: %s\n",
1327 			  __func__, skb->dev->name);
1328 		goto skip_defrtr;
1329 	}
1330 
1331 	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1332 	if (lifetime != 0 && lifetime < in6_dev->cnf.accept_ra_min_lft) {
1333 		ND_PRINTK(2, info,
1334 			  "RA: router lifetime (%ds) is too short: %s\n",
1335 			  lifetime, skb->dev->name);
1336 		goto skip_defrtr;
1337 	}
1338 
1339 	/* Do not accept RA with source-addr found on local machine unless
1340 	 * accept_ra_from_local is set to true.
1341 	 */
1342 	net = dev_net(in6_dev->dev);
1343 	if (!in6_dev->cnf.accept_ra_from_local &&
1344 	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1345 		ND_PRINTK(2, info,
1346 			  "RA from local address detected on dev: %s: default router ignored\n",
1347 			  skb->dev->name);
1348 		goto skip_defrtr;
1349 	}
1350 
1351 #ifdef CONFIG_IPV6_ROUTER_PREF
1352 	pref = ra_msg->icmph.icmp6_router_pref;
1353 	/* 10b is handled as if it were 00b (medium) */
1354 	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1355 	    !in6_dev->cnf.accept_ra_rtr_pref)
1356 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1357 #endif
1358 	/* routes added from RAs do not use nexthop objects */
1359 	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1360 	if (rt) {
1361 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1362 					 rt->fib6_nh->fib_nh_dev, NULL,
1363 					  &ipv6_hdr(skb)->saddr);
1364 		if (!neigh) {
1365 			ND_PRINTK(0, err,
1366 				  "RA: %s got default router without neighbour\n",
1367 				  __func__);
1368 			fib6_info_release(rt);
1369 			return reason;
1370 		}
1371 	}
1372 	/* Set default route metric as specified by user */
1373 	defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1374 	/* delete the route if lifetime is 0 or if metric needs change */
1375 	if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1376 		ip6_del_rt(net, rt, false);
1377 		rt = NULL;
1378 	}
1379 
1380 	ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, metric: %d, for dev: %s\n",
1381 		  rt, lifetime, defrtr_usr_metric, skb->dev->name);
1382 	if (!rt && lifetime) {
1383 		ND_PRINTK(3, info, "RA: adding default router\n");
1384 
1385 		if (neigh)
1386 			neigh_release(neigh);
1387 
1388 		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1389 					 skb->dev, pref, defrtr_usr_metric);
1390 		if (!rt) {
1391 			ND_PRINTK(0, err,
1392 				  "RA: %s failed to add default route\n",
1393 				  __func__);
1394 			return reason;
1395 		}
1396 
1397 		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1398 					 rt->fib6_nh->fib_nh_dev, NULL,
1399 					  &ipv6_hdr(skb)->saddr);
1400 		if (!neigh) {
1401 			ND_PRINTK(0, err,
1402 				  "RA: %s got default router without neighbour\n",
1403 				  __func__);
1404 			fib6_info_release(rt);
1405 			return reason;
1406 		}
1407 		neigh->flags |= NTF_ROUTER;
1408 	} else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1409 		struct nl_info nlinfo = {
1410 			.nl_net = net,
1411 		};
1412 		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1413 		inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1414 	}
1415 
1416 	if (rt)
1417 		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1418 	if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1419 	    ra_msg->icmph.icmp6_hop_limit) {
1420 		if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1421 			in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1422 			fib6_metric_set(rt, RTAX_HOPLIMIT,
1423 					ra_msg->icmph.icmp6_hop_limit);
1424 		} else {
1425 			ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1426 		}
1427 	}
1428 
1429 skip_defrtr:
1430 
1431 	/*
1432 	 *	Update Reachable Time and Retrans Timer
1433 	 */
1434 
1435 	if (in6_dev->nd_parms) {
1436 		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1437 
1438 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1439 			rtime = (rtime*HZ)/1000;
1440 			if (rtime < HZ/100)
1441 				rtime = HZ/100;
1442 			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1443 			in6_dev->tstamp = jiffies;
1444 			send_ifinfo_notify = true;
1445 		}
1446 
1447 		rtime = ntohl(ra_msg->reachable_time);
1448 		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1449 			rtime = (rtime*HZ)/1000;
1450 
1451 			if (rtime < HZ/10)
1452 				rtime = HZ/10;
1453 
1454 			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1455 				NEIGH_VAR_SET(in6_dev->nd_parms,
1456 					      BASE_REACHABLE_TIME, rtime);
1457 				NEIGH_VAR_SET(in6_dev->nd_parms,
1458 					      GC_STALETIME, 3 * rtime);
1459 				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1460 				in6_dev->tstamp = jiffies;
1461 				send_ifinfo_notify = true;
1462 			}
1463 		}
1464 	}
1465 
1466 skip_linkparms:
1467 
1468 	/*
1469 	 *	Process options.
1470 	 */
1471 
1472 	if (!neigh)
1473 		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1474 				       skb->dev, 1);
1475 	if (neigh) {
1476 		u8 *lladdr = NULL;
1477 		if (ndopts.nd_opts_src_lladdr) {
1478 			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1479 						     skb->dev);
1480 			if (!lladdr) {
1481 				ND_PRINTK(2, warn,
1482 					  "RA: invalid link-layer address length\n");
1483 				goto out;
1484 			}
1485 		}
1486 		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1487 			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1488 			     NEIGH_UPDATE_F_OVERRIDE|
1489 			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1490 			     NEIGH_UPDATE_F_ISROUTER,
1491 			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1492 		reason = SKB_CONSUMED;
1493 	}
1494 
1495 	if (!ipv6_accept_ra(in6_dev)) {
1496 		ND_PRINTK(2, info,
1497 			  "RA: %s, accept_ra is false for dev: %s\n",
1498 			  __func__, skb->dev->name);
1499 		goto out;
1500 	}
1501 
1502 #ifdef CONFIG_IPV6_ROUTE_INFO
1503 	if (!in6_dev->cnf.accept_ra_from_local &&
1504 	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1505 			  in6_dev->dev, 0)) {
1506 		ND_PRINTK(2, info,
1507 			  "RA from local address detected on dev: %s: router info ignored.\n",
1508 			  skb->dev->name);
1509 		goto skip_routeinfo;
1510 	}
1511 
1512 	if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1513 		struct nd_opt_hdr *p;
1514 		for (p = ndopts.nd_opts_ri;
1515 		     p;
1516 		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1517 			struct route_info *ri = (struct route_info *)p;
1518 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1519 			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1520 			    ri->prefix_len == 0)
1521 				continue;
1522 #endif
1523 			if (ri->prefix_len == 0 &&
1524 			    !in6_dev->cnf.accept_ra_defrtr)
1525 				continue;
1526 			if (ri->lifetime != 0 &&
1527 			    ntohl(ri->lifetime) < in6_dev->cnf.accept_ra_min_lft)
1528 				continue;
1529 			if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1530 				continue;
1531 			if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1532 				continue;
1533 			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1534 				      &ipv6_hdr(skb)->saddr);
1535 		}
1536 	}
1537 
1538 skip_routeinfo:
1539 #endif
1540 
1541 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1542 	/* skip link-specific ndopts from interior routers */
1543 	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1544 		ND_PRINTK(2, info,
1545 			  "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1546 			  __func__, skb->dev->name);
1547 		goto out;
1548 	}
1549 #endif
1550 
1551 	if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1552 		struct nd_opt_hdr *p;
1553 		for (p = ndopts.nd_opts_pi;
1554 		     p;
1555 		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1556 			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1557 					    (p->nd_opt_len) << 3,
1558 					    ndopts.nd_opts_src_lladdr != NULL);
1559 		}
1560 	}
1561 
1562 	if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1563 		__be32 n;
1564 		u32 mtu;
1565 
1566 		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1567 		mtu = ntohl(n);
1568 
1569 		if (in6_dev->ra_mtu != mtu) {
1570 			in6_dev->ra_mtu = mtu;
1571 			send_ifinfo_notify = true;
1572 		}
1573 
1574 		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1575 			ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1576 		} else if (in6_dev->cnf.mtu6 != mtu) {
1577 			in6_dev->cnf.mtu6 = mtu;
1578 			fib6_metric_set(rt, RTAX_MTU, mtu);
1579 			rt6_mtu_change(skb->dev, mtu);
1580 		}
1581 	}
1582 
1583 	if (ndopts.nd_useropts) {
1584 		struct nd_opt_hdr *p;
1585 		for (p = ndopts.nd_useropts;
1586 		     p;
1587 		     p = ndisc_next_useropt(skb->dev, p,
1588 					    ndopts.nd_useropts_end)) {
1589 			ndisc_ra_useropt(skb, p);
1590 		}
1591 	}
1592 
1593 	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1594 		ND_PRINTK(2, warn, "RA: invalid RA options\n");
1595 	}
1596 out:
1597 	/* Send a notify if RA changed managed/otherconf flags or
1598 	 * timer settings or ra_mtu value
1599 	 */
1600 	if (send_ifinfo_notify)
1601 		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1602 
1603 	fib6_info_release(rt);
1604 	if (neigh)
1605 		neigh_release(neigh);
1606 	return reason;
1607 }
1608 
ndisc_redirect_rcv(struct sk_buff * skb)1609 static enum skb_drop_reason ndisc_redirect_rcv(struct sk_buff *skb)
1610 {
1611 	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1612 	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1613 				    offsetof(struct rd_msg, opt));
1614 	struct ndisc_options ndopts;
1615 	SKB_DR(reason);
1616 	u8 *hdr;
1617 
1618 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1619 	switch (skb->ndisc_nodetype) {
1620 	case NDISC_NODETYPE_HOST:
1621 	case NDISC_NODETYPE_NODEFAULT:
1622 		ND_PRINTK(2, warn,
1623 			  "Redirect: from host or unauthorized router\n");
1624 		return reason;
1625 	}
1626 #endif
1627 
1628 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1629 		ND_PRINTK(2, warn,
1630 			  "Redirect: source address is not link-local\n");
1631 		return reason;
1632 	}
1633 
1634 	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1635 		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1636 
1637 	if (!ndopts.nd_opts_rh) {
1638 		ip6_redirect_no_header(skb, dev_net(skb->dev),
1639 					skb->dev->ifindex);
1640 		return reason;
1641 	}
1642 
1643 	hdr = (u8 *)ndopts.nd_opts_rh;
1644 	hdr += 8;
1645 	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1646 		return SKB_DROP_REASON_PKT_TOO_SMALL;
1647 
1648 	return icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1649 }
1650 
ndisc_fill_redirect_hdr_option(struct sk_buff * skb,struct sk_buff * orig_skb,int rd_len)1651 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1652 					   struct sk_buff *orig_skb,
1653 					   int rd_len)
1654 {
1655 	u8 *opt = skb_put(skb, rd_len);
1656 
1657 	memset(opt, 0, 8);
1658 	*(opt++) = ND_OPT_REDIRECT_HDR;
1659 	*(opt++) = (rd_len >> 3);
1660 	opt += 6;
1661 
1662 	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1663 		      rd_len - 8);
1664 }
1665 
ndisc_send_redirect(struct sk_buff * skb,const struct in6_addr * target)1666 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1667 {
1668 	struct net_device *dev = skb->dev;
1669 	struct net *net = dev_net(dev);
1670 	struct sock *sk = net->ipv6.ndisc_sk;
1671 	int optlen = 0;
1672 	struct inet_peer *peer;
1673 	struct sk_buff *buff;
1674 	struct rd_msg *msg;
1675 	struct in6_addr saddr_buf;
1676 	struct rt6_info *rt;
1677 	struct dst_entry *dst;
1678 	struct flowi6 fl6;
1679 	int rd_len;
1680 	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1681 	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1682 	bool ret;
1683 
1684 	if (netif_is_l3_master(skb->dev)) {
1685 		dev = dev_get_by_index_rcu(dev_net(skb->dev), IPCB(skb)->iif);
1686 		if (!dev)
1687 			return;
1688 	}
1689 
1690 	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1691 		ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1692 			  dev->name);
1693 		return;
1694 	}
1695 
1696 	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1697 	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1698 		ND_PRINTK(2, warn,
1699 			  "Redirect: target address is not link-local unicast\n");
1700 		return;
1701 	}
1702 
1703 	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1704 			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1705 
1706 	dst = ip6_route_output(net, NULL, &fl6);
1707 	if (dst->error) {
1708 		dst_release(dst);
1709 		return;
1710 	}
1711 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1712 	if (IS_ERR(dst))
1713 		return;
1714 
1715 	rt = dst_rt6_info(dst);
1716 
1717 	if (rt->rt6i_flags & RTF_GATEWAY) {
1718 		ND_PRINTK(2, warn,
1719 			  "Redirect: destination is not a neighbour\n");
1720 		goto release;
1721 	}
1722 
1723 	rcu_read_lock();
1724 	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr);
1725 	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1726 	rcu_read_unlock();
1727 
1728 	if (!ret)
1729 		goto release;
1730 
1731 	if (dev->addr_len) {
1732 		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1733 		if (!neigh) {
1734 			ND_PRINTK(2, warn,
1735 				  "Redirect: no neigh for target address\n");
1736 			goto release;
1737 		}
1738 
1739 		read_lock_bh(&neigh->lock);
1740 		if (neigh->nud_state & NUD_VALID) {
1741 			memcpy(ha_buf, neigh->ha, dev->addr_len);
1742 			read_unlock_bh(&neigh->lock);
1743 			ha = ha_buf;
1744 			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1745 								ops_data_buf,
1746 								&ops_data);
1747 		} else
1748 			read_unlock_bh(&neigh->lock);
1749 
1750 		neigh_release(neigh);
1751 	}
1752 
1753 	rd_len = min_t(unsigned int,
1754 		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1755 		       skb->len + 8);
1756 	rd_len &= ~0x7;
1757 	optlen += rd_len;
1758 
1759 	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1760 	if (!buff)
1761 		goto release;
1762 
1763 	msg = skb_put(buff, sizeof(*msg));
1764 	*msg = (struct rd_msg) {
1765 		.icmph = {
1766 			.icmp6_type = NDISC_REDIRECT,
1767 		},
1768 		.target = *target,
1769 		.dest = ipv6_hdr(skb)->daddr,
1770 	};
1771 
1772 	/*
1773 	 *	include target_address option
1774 	 */
1775 
1776 	if (ha)
1777 		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1778 
1779 	/*
1780 	 *	build redirect option and copy skb over to the new packet.
1781 	 */
1782 
1783 	if (rd_len)
1784 		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1785 
1786 	skb_dst_set(buff, dst);
1787 	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1788 	return;
1789 
1790 release:
1791 	dst_release(dst);
1792 }
1793 
pndisc_redo(struct sk_buff * skb)1794 static void pndisc_redo(struct sk_buff *skb)
1795 {
1796 	enum skb_drop_reason reason = ndisc_recv_ns(skb);
1797 
1798 	kfree_skb_reason(skb, reason);
1799 }
1800 
ndisc_is_multicast(const void * pkey)1801 static int ndisc_is_multicast(const void *pkey)
1802 {
1803 	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1804 }
1805 
ndisc_suppress_frag_ndisc(struct sk_buff * skb)1806 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1807 {
1808 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1809 
1810 	if (!idev)
1811 		return true;
1812 	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1813 	    idev->cnf.suppress_frag_ndisc) {
1814 		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1815 		return true;
1816 	}
1817 	return false;
1818 }
1819 
ndisc_rcv(struct sk_buff * skb)1820 enum skb_drop_reason ndisc_rcv(struct sk_buff *skb)
1821 {
1822 	struct nd_msg *msg;
1823 	SKB_DR(reason);
1824 
1825 	if (ndisc_suppress_frag_ndisc(skb))
1826 		return SKB_DROP_REASON_IPV6_NDISC_FRAG;
1827 
1828 	if (skb_linearize(skb))
1829 		return SKB_DROP_REASON_NOMEM;
1830 
1831 	msg = (struct nd_msg *)skb_transport_header(skb);
1832 
1833 	__skb_push(skb, skb->data - skb_transport_header(skb));
1834 
1835 	if (ipv6_hdr(skb)->hop_limit != 255) {
1836 		ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1837 			  ipv6_hdr(skb)->hop_limit);
1838 		return SKB_DROP_REASON_IPV6_NDISC_HOP_LIMIT;
1839 	}
1840 
1841 	if (msg->icmph.icmp6_code != 0) {
1842 		ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1843 			  msg->icmph.icmp6_code);
1844 		return SKB_DROP_REASON_IPV6_NDISC_BAD_CODE;
1845 	}
1846 
1847 	switch (msg->icmph.icmp6_type) {
1848 	case NDISC_NEIGHBOUR_SOLICITATION:
1849 		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1850 		reason = ndisc_recv_ns(skb);
1851 		break;
1852 
1853 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1854 		reason = ndisc_recv_na(skb);
1855 		break;
1856 
1857 	case NDISC_ROUTER_SOLICITATION:
1858 		reason = ndisc_recv_rs(skb);
1859 		break;
1860 
1861 	case NDISC_ROUTER_ADVERTISEMENT:
1862 		reason = ndisc_router_discovery(skb);
1863 		break;
1864 
1865 	case NDISC_REDIRECT:
1866 		reason = ndisc_redirect_rcv(skb);
1867 		break;
1868 	}
1869 
1870 	return reason;
1871 }
1872 
ndisc_netdev_event(struct notifier_block * this,unsigned long event,void * ptr)1873 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1874 {
1875 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1876 	struct netdev_notifier_change_info *change_info;
1877 	struct net *net = dev_net(dev);
1878 	struct inet6_dev *idev;
1879 	bool evict_nocarrier;
1880 
1881 	switch (event) {
1882 	case NETDEV_CHANGEADDR:
1883 		neigh_changeaddr(&nd_tbl, dev);
1884 		fib6_run_gc(0, net, false);
1885 		fallthrough;
1886 	case NETDEV_UP:
1887 		idev = in6_dev_get(dev);
1888 		if (!idev)
1889 			break;
1890 		if (idev->cnf.ndisc_notify ||
1891 		    net->ipv6.devconf_all->ndisc_notify)
1892 			ndisc_send_unsol_na(dev);
1893 		in6_dev_put(idev);
1894 		break;
1895 	case NETDEV_CHANGE:
1896 		idev = in6_dev_get(dev);
1897 		if (!idev)
1898 			evict_nocarrier = true;
1899 		else {
1900 			evict_nocarrier = idev->cnf.ndisc_evict_nocarrier &&
1901 					  net->ipv6.devconf_all->ndisc_evict_nocarrier;
1902 			in6_dev_put(idev);
1903 		}
1904 
1905 		change_info = ptr;
1906 		if (change_info->flags_changed & IFF_NOARP)
1907 			neigh_changeaddr(&nd_tbl, dev);
1908 		if (evict_nocarrier && !netif_carrier_ok(dev))
1909 			neigh_carrier_down(&nd_tbl, dev);
1910 		break;
1911 	case NETDEV_DOWN:
1912 		neigh_ifdown(&nd_tbl, dev);
1913 		fib6_run_gc(0, net, false);
1914 		break;
1915 	case NETDEV_NOTIFY_PEERS:
1916 		ndisc_send_unsol_na(dev);
1917 		break;
1918 	default:
1919 		break;
1920 	}
1921 
1922 	return NOTIFY_DONE;
1923 }
1924 
1925 static struct notifier_block ndisc_netdev_notifier = {
1926 	.notifier_call = ndisc_netdev_event,
1927 	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1928 };
1929 
1930 #ifdef CONFIG_SYSCTL
ndisc_warn_deprecated_sysctl(struct ctl_table * ctl,const char * func,const char * dev_name)1931 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1932 					 const char *func, const char *dev_name)
1933 {
1934 	static char warncomm[TASK_COMM_LEN];
1935 	static int warned;
1936 	if (strcmp(warncomm, current->comm) && warned < 5) {
1937 		strcpy(warncomm, current->comm);
1938 		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1939 			warncomm, func,
1940 			dev_name, ctl->procname,
1941 			dev_name, ctl->procname);
1942 		warned++;
1943 	}
1944 }
1945 
ndisc_ifinfo_sysctl_change(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)1946 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1947 		size_t *lenp, loff_t *ppos)
1948 {
1949 	struct net_device *dev = ctl->extra1;
1950 	struct inet6_dev *idev;
1951 	int ret;
1952 
1953 	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1954 	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1955 		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1956 
1957 	if (strcmp(ctl->procname, "retrans_time") == 0)
1958 		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1959 
1960 	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1961 		ret = neigh_proc_dointvec_jiffies(ctl, write,
1962 						  buffer, lenp, ppos);
1963 
1964 	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1965 		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1966 		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1967 						     buffer, lenp, ppos);
1968 	else
1969 		ret = -1;
1970 
1971 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1972 		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1973 			idev->nd_parms->reachable_time =
1974 					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1975 		idev->tstamp = jiffies;
1976 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1977 		in6_dev_put(idev);
1978 	}
1979 	return ret;
1980 }
1981 
1982 
1983 #endif
1984 
ndisc_net_init(struct net * net)1985 static int __net_init ndisc_net_init(struct net *net)
1986 {
1987 	struct ipv6_pinfo *np;
1988 	struct sock *sk;
1989 	int err;
1990 
1991 	err = inet_ctl_sock_create(&sk, PF_INET6,
1992 				   SOCK_RAW, IPPROTO_ICMPV6, net);
1993 	if (err < 0) {
1994 		ND_PRINTK(0, err,
1995 			  "NDISC: Failed to initialize the control socket (err %d)\n",
1996 			  err);
1997 		return err;
1998 	}
1999 
2000 	net->ipv6.ndisc_sk = sk;
2001 
2002 	np = inet6_sk(sk);
2003 	np->hop_limit = 255;
2004 	/* Do not loopback ndisc messages */
2005 	np->mc_loop = 0;
2006 
2007 	return 0;
2008 }
2009 
ndisc_net_exit(struct net * net)2010 static void __net_exit ndisc_net_exit(struct net *net)
2011 {
2012 	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
2013 }
2014 
2015 static struct pernet_operations ndisc_net_ops = {
2016 	.init = ndisc_net_init,
2017 	.exit = ndisc_net_exit,
2018 };
2019 
ndisc_init(void)2020 int __init ndisc_init(void)
2021 {
2022 	int err;
2023 
2024 	err = register_pernet_subsys(&ndisc_net_ops);
2025 	if (err)
2026 		return err;
2027 	/*
2028 	 * Initialize the neighbour table
2029 	 */
2030 	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2031 
2032 #ifdef CONFIG_SYSCTL
2033 	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2034 				    ndisc_ifinfo_sysctl_change);
2035 	if (err)
2036 		goto out_unregister_pernet;
2037 out:
2038 #endif
2039 	return err;
2040 
2041 #ifdef CONFIG_SYSCTL
2042 out_unregister_pernet:
2043 	unregister_pernet_subsys(&ndisc_net_ops);
2044 	goto out;
2045 #endif
2046 }
2047 
ndisc_late_init(void)2048 int __init ndisc_late_init(void)
2049 {
2050 	return register_netdevice_notifier(&ndisc_netdev_notifier);
2051 }
2052 
ndisc_late_cleanup(void)2053 void ndisc_late_cleanup(void)
2054 {
2055 	unregister_netdevice_notifier(&ndisc_netdev_notifier);
2056 }
2057 
ndisc_cleanup(void)2058 void ndisc_cleanup(void)
2059 {
2060 #ifdef CONFIG_SYSCTL
2061 	neigh_sysctl_unregister(&nd_tbl.parms);
2062 #endif
2063 	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2064 	unregister_pernet_subsys(&ndisc_net_ops);
2065 }
2066