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