Loading...
1/*
2 * Neighbour Discovery for IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Mike Shaver <shaver@ingenia.com>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
13 */
14
15/*
16 * Changes:
17 *
18 * Alexey I. Froloff : RFC6106 (DNSSL) support
19 * Pierre Ynard : export userland ND options
20 * through netlink (RDNSS support)
21 * Lars Fenneberg : fixed MTU setting on receipt
22 * of an RA.
23 * Janos Farkas : kmalloc failure checks
24 * Alexey Kuznetsov : state machine reworked
25 * and moved to net/core.
26 * Pekka Savola : RFC2461 validation
27 * YOSHIFUJI Hideaki @USAGI : Verify ND options properly
28 */
29
30#define pr_fmt(fmt) "ICMPv6: " fmt
31
32#include <linux/module.h>
33#include <linux/errno.h>
34#include <linux/types.h>
35#include <linux/socket.h>
36#include <linux/sockios.h>
37#include <linux/sched.h>
38#include <linux/net.h>
39#include <linux/in6.h>
40#include <linux/route.h>
41#include <linux/init.h>
42#include <linux/rcupdate.h>
43#include <linux/slab.h>
44#ifdef CONFIG_SYSCTL
45#include <linux/sysctl.h>
46#endif
47
48#include <linux/if_addr.h>
49#include <linux/if_ether.h>
50#include <linux/if_arp.h>
51#include <linux/ipv6.h>
52#include <linux/icmpv6.h>
53#include <linux/jhash.h>
54
55#include <net/sock.h>
56#include <net/snmp.h>
57
58#include <net/ipv6.h>
59#include <net/protocol.h>
60#include <net/ndisc.h>
61#include <net/ip6_route.h>
62#include <net/addrconf.h>
63#include <net/icmp.h>
64
65#include <net/netlink.h>
66#include <linux/rtnetlink.h>
67
68#include <net/flow.h>
69#include <net/ip6_checksum.h>
70#include <net/inet_common.h>
71#include <linux/proc_fs.h>
72
73#include <linux/netfilter.h>
74#include <linux/netfilter_ipv6.h>
75
76static u32 ndisc_hash(const void *pkey,
77 const struct net_device *dev,
78 __u32 *hash_rnd);
79static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
80static int ndisc_constructor(struct neighbour *neigh);
81static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
82static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
83static int pndisc_constructor(struct pneigh_entry *n);
84static void pndisc_destructor(struct pneigh_entry *n);
85static void pndisc_redo(struct sk_buff *skb);
86
87static const struct neigh_ops ndisc_generic_ops = {
88 .family = AF_INET6,
89 .solicit = ndisc_solicit,
90 .error_report = ndisc_error_report,
91 .output = neigh_resolve_output,
92 .connected_output = neigh_connected_output,
93};
94
95static const struct neigh_ops ndisc_hh_ops = {
96 .family = AF_INET6,
97 .solicit = ndisc_solicit,
98 .error_report = ndisc_error_report,
99 .output = neigh_resolve_output,
100 .connected_output = neigh_resolve_output,
101};
102
103
104static const struct neigh_ops ndisc_direct_ops = {
105 .family = AF_INET6,
106 .output = neigh_direct_output,
107 .connected_output = neigh_direct_output,
108};
109
110struct neigh_table nd_tbl = {
111 .family = AF_INET6,
112 .key_len = sizeof(struct in6_addr),
113 .protocol = cpu_to_be16(ETH_P_IPV6),
114 .hash = ndisc_hash,
115 .key_eq = ndisc_key_eq,
116 .constructor = ndisc_constructor,
117 .pconstructor = pndisc_constructor,
118 .pdestructor = pndisc_destructor,
119 .proxy_redo = pndisc_redo,
120 .id = "ndisc_cache",
121 .parms = {
122 .tbl = &nd_tbl,
123 .reachable_time = ND_REACHABLE_TIME,
124 .data = {
125 [NEIGH_VAR_MCAST_PROBES] = 3,
126 [NEIGH_VAR_UCAST_PROBES] = 3,
127 [NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER,
128 [NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME,
129 [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
130 [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
131 [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
132 [NEIGH_VAR_PROXY_QLEN] = 64,
133 [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
134 [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
135 },
136 },
137 .gc_interval = 30 * HZ,
138 .gc_thresh1 = 128,
139 .gc_thresh2 = 512,
140 .gc_thresh3 = 1024,
141};
142EXPORT_SYMBOL_GPL(nd_tbl);
143
144void __ndisc_fill_addr_option(struct sk_buff *skb, int type, void *data,
145 int data_len, int pad)
146{
147 int space = __ndisc_opt_addr_space(data_len, pad);
148 u8 *opt = skb_put(skb, space);
149
150 opt[0] = type;
151 opt[1] = space>>3;
152
153 memset(opt + 2, 0, pad);
154 opt += pad;
155 space -= pad;
156
157 memcpy(opt+2, data, data_len);
158 data_len += 2;
159 opt += data_len;
160 space -= data_len;
161 if (space > 0)
162 memset(opt, 0, space);
163}
164EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
165
166static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
167 void *data, u8 icmp6_type)
168{
169 __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
170 ndisc_addr_option_pad(skb->dev->type));
171 ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
172}
173
174static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
175 void *ha,
176 const u8 *ops_data)
177{
178 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
179 ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
180}
181
182static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
183 struct nd_opt_hdr *end)
184{
185 int type;
186 if (!cur || !end || cur >= end)
187 return NULL;
188 type = cur->nd_opt_type;
189 do {
190 cur = ((void *)cur) + (cur->nd_opt_len << 3);
191 } while (cur < end && cur->nd_opt_type != type);
192 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
193}
194
195static inline int ndisc_is_useropt(const struct net_device *dev,
196 struct nd_opt_hdr *opt)
197{
198 return opt->nd_opt_type == ND_OPT_RDNSS ||
199 opt->nd_opt_type == ND_OPT_DNSSL ||
200 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
201}
202
203static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
204 struct nd_opt_hdr *cur,
205 struct nd_opt_hdr *end)
206{
207 if (!cur || !end || cur >= end)
208 return NULL;
209 do {
210 cur = ((void *)cur) + (cur->nd_opt_len << 3);
211 } while (cur < end && !ndisc_is_useropt(dev, cur));
212 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
213}
214
215struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
216 u8 *opt, int opt_len,
217 struct ndisc_options *ndopts)
218{
219 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
220
221 if (!nd_opt || opt_len < 0 || !ndopts)
222 return NULL;
223 memset(ndopts, 0, sizeof(*ndopts));
224 while (opt_len) {
225 int l;
226 if (opt_len < sizeof(struct nd_opt_hdr))
227 return NULL;
228 l = nd_opt->nd_opt_len << 3;
229 if (opt_len < l || l == 0)
230 return NULL;
231 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
232 goto next_opt;
233 switch (nd_opt->nd_opt_type) {
234 case ND_OPT_SOURCE_LL_ADDR:
235 case ND_OPT_TARGET_LL_ADDR:
236 case ND_OPT_MTU:
237 case ND_OPT_NONCE:
238 case ND_OPT_REDIRECT_HDR:
239 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
240 ND_PRINTK(2, warn,
241 "%s: duplicated ND6 option found: type=%d\n",
242 __func__, nd_opt->nd_opt_type);
243 } else {
244 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
245 }
246 break;
247 case ND_OPT_PREFIX_INFO:
248 ndopts->nd_opts_pi_end = nd_opt;
249 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
250 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
251 break;
252#ifdef CONFIG_IPV6_ROUTE_INFO
253 case ND_OPT_ROUTE_INFO:
254 ndopts->nd_opts_ri_end = nd_opt;
255 if (!ndopts->nd_opts_ri)
256 ndopts->nd_opts_ri = nd_opt;
257 break;
258#endif
259 default:
260 if (ndisc_is_useropt(dev, nd_opt)) {
261 ndopts->nd_useropts_end = nd_opt;
262 if (!ndopts->nd_useropts)
263 ndopts->nd_useropts = nd_opt;
264 } else {
265 /*
266 * Unknown options must be silently ignored,
267 * to accommodate future extension to the
268 * protocol.
269 */
270 ND_PRINTK(2, notice,
271 "%s: ignored unsupported option; type=%d, len=%d\n",
272 __func__,
273 nd_opt->nd_opt_type,
274 nd_opt->nd_opt_len);
275 }
276 }
277next_opt:
278 opt_len -= l;
279 nd_opt = ((void *)nd_opt) + l;
280 }
281 return ndopts;
282}
283
284int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
285{
286 switch (dev->type) {
287 case ARPHRD_ETHER:
288 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
289 case ARPHRD_FDDI:
290 ipv6_eth_mc_map(addr, buf);
291 return 0;
292 case ARPHRD_ARCNET:
293 ipv6_arcnet_mc_map(addr, buf);
294 return 0;
295 case ARPHRD_INFINIBAND:
296 ipv6_ib_mc_map(addr, dev->broadcast, buf);
297 return 0;
298 case ARPHRD_IPGRE:
299 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
300 default:
301 if (dir) {
302 memcpy(buf, dev->broadcast, dev->addr_len);
303 return 0;
304 }
305 }
306 return -EINVAL;
307}
308EXPORT_SYMBOL(ndisc_mc_map);
309
310static u32 ndisc_hash(const void *pkey,
311 const struct net_device *dev,
312 __u32 *hash_rnd)
313{
314 return ndisc_hashfn(pkey, dev, hash_rnd);
315}
316
317static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
318{
319 return neigh_key_eq128(n, pkey);
320}
321
322static int ndisc_constructor(struct neighbour *neigh)
323{
324 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
325 struct net_device *dev = neigh->dev;
326 struct inet6_dev *in6_dev;
327 struct neigh_parms *parms;
328 bool is_multicast = ipv6_addr_is_multicast(addr);
329
330 in6_dev = in6_dev_get(dev);
331 if (!in6_dev) {
332 return -EINVAL;
333 }
334
335 parms = in6_dev->nd_parms;
336 __neigh_parms_put(neigh->parms);
337 neigh->parms = neigh_parms_clone(parms);
338
339 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
340 if (!dev->header_ops) {
341 neigh->nud_state = NUD_NOARP;
342 neigh->ops = &ndisc_direct_ops;
343 neigh->output = neigh_direct_output;
344 } else {
345 if (is_multicast) {
346 neigh->nud_state = NUD_NOARP;
347 ndisc_mc_map(addr, neigh->ha, dev, 1);
348 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
349 neigh->nud_state = NUD_NOARP;
350 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
351 if (dev->flags&IFF_LOOPBACK)
352 neigh->type = RTN_LOCAL;
353 } else if (dev->flags&IFF_POINTOPOINT) {
354 neigh->nud_state = NUD_NOARP;
355 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
356 }
357 if (dev->header_ops->cache)
358 neigh->ops = &ndisc_hh_ops;
359 else
360 neigh->ops = &ndisc_generic_ops;
361 if (neigh->nud_state&NUD_VALID)
362 neigh->output = neigh->ops->connected_output;
363 else
364 neigh->output = neigh->ops->output;
365 }
366 in6_dev_put(in6_dev);
367 return 0;
368}
369
370static int pndisc_constructor(struct pneigh_entry *n)
371{
372 struct in6_addr *addr = (struct in6_addr *)&n->key;
373 struct in6_addr maddr;
374 struct net_device *dev = n->dev;
375
376 if (!dev || !__in6_dev_get(dev))
377 return -EINVAL;
378 addrconf_addr_solict_mult(addr, &maddr);
379 ipv6_dev_mc_inc(dev, &maddr);
380 return 0;
381}
382
383static void pndisc_destructor(struct pneigh_entry *n)
384{
385 struct in6_addr *addr = (struct in6_addr *)&n->key;
386 struct in6_addr maddr;
387 struct net_device *dev = n->dev;
388
389 if (!dev || !__in6_dev_get(dev))
390 return;
391 addrconf_addr_solict_mult(addr, &maddr);
392 ipv6_dev_mc_dec(dev, &maddr);
393}
394
395static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
396 int len)
397{
398 int hlen = LL_RESERVED_SPACE(dev);
399 int tlen = dev->needed_tailroom;
400 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
401 struct sk_buff *skb;
402
403 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
404 if (!skb) {
405 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
406 __func__);
407 return NULL;
408 }
409
410 skb->protocol = htons(ETH_P_IPV6);
411 skb->dev = dev;
412
413 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
414 skb_reset_transport_header(skb);
415
416 /* Manually assign socket ownership as we avoid calling
417 * sock_alloc_send_pskb() to bypass wmem buffer limits
418 */
419 skb_set_owner_w(skb, sk);
420
421 return skb;
422}
423
424static void ip6_nd_hdr(struct sk_buff *skb,
425 const struct in6_addr *saddr,
426 const struct in6_addr *daddr,
427 int hop_limit, int len)
428{
429 struct ipv6hdr *hdr;
430 struct inet6_dev *idev;
431 unsigned tclass;
432
433 rcu_read_lock();
434 idev = __in6_dev_get(skb->dev);
435 tclass = idev ? idev->cnf.ndisc_tclass : 0;
436 rcu_read_unlock();
437
438 skb_push(skb, sizeof(*hdr));
439 skb_reset_network_header(skb);
440 hdr = ipv6_hdr(skb);
441
442 ip6_flow_hdr(hdr, tclass, 0);
443
444 hdr->payload_len = htons(len);
445 hdr->nexthdr = IPPROTO_ICMPV6;
446 hdr->hop_limit = hop_limit;
447
448 hdr->saddr = *saddr;
449 hdr->daddr = *daddr;
450}
451
452static void ndisc_send_skb(struct sk_buff *skb,
453 const struct in6_addr *daddr,
454 const struct in6_addr *saddr)
455{
456 struct dst_entry *dst = skb_dst(skb);
457 struct net *net = dev_net(skb->dev);
458 struct sock *sk = net->ipv6.ndisc_sk;
459 struct inet6_dev *idev;
460 int err;
461 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
462 u8 type;
463
464 type = icmp6h->icmp6_type;
465
466 if (!dst) {
467 struct flowi6 fl6;
468 int oif = skb->dev->ifindex;
469
470 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
471 dst = icmp6_dst_alloc(skb->dev, &fl6);
472 if (IS_ERR(dst)) {
473 kfree_skb(skb);
474 return;
475 }
476
477 skb_dst_set(skb, dst);
478 }
479
480 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
481 IPPROTO_ICMPV6,
482 csum_partial(icmp6h,
483 skb->len, 0));
484
485 ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
486
487 rcu_read_lock();
488 idev = __in6_dev_get(dst->dev);
489 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
490
491 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
492 net, sk, skb, NULL, dst->dev,
493 dst_output);
494 if (!err) {
495 ICMP6MSGOUT_INC_STATS(net, idev, type);
496 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
497 }
498
499 rcu_read_unlock();
500}
501
502void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
503 const struct in6_addr *solicited_addr,
504 bool router, bool solicited, bool override, bool inc_opt)
505{
506 struct sk_buff *skb;
507 struct in6_addr tmpaddr;
508 struct inet6_ifaddr *ifp;
509 const struct in6_addr *src_addr;
510 struct nd_msg *msg;
511 int optlen = 0;
512
513 /* for anycast or proxy, solicited_addr != src_addr */
514 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
515 if (ifp) {
516 src_addr = solicited_addr;
517 if (ifp->flags & IFA_F_OPTIMISTIC)
518 override = false;
519 inc_opt |= ifp->idev->cnf.force_tllao;
520 in6_ifa_put(ifp);
521 } else {
522 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
523 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
524 &tmpaddr))
525 return;
526 src_addr = &tmpaddr;
527 }
528
529 if (!dev->addr_len)
530 inc_opt = false;
531 if (inc_opt)
532 optlen += ndisc_opt_addr_space(dev,
533 NDISC_NEIGHBOUR_ADVERTISEMENT);
534
535 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
536 if (!skb)
537 return;
538
539 msg = skb_put(skb, sizeof(*msg));
540 *msg = (struct nd_msg) {
541 .icmph = {
542 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
543 .icmp6_router = router,
544 .icmp6_solicited = solicited,
545 .icmp6_override = override,
546 },
547 .target = *solicited_addr,
548 };
549
550 if (inc_opt)
551 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
552 dev->dev_addr,
553 NDISC_NEIGHBOUR_ADVERTISEMENT);
554
555 ndisc_send_skb(skb, daddr, src_addr);
556}
557
558static void ndisc_send_unsol_na(struct net_device *dev)
559{
560 struct inet6_dev *idev;
561 struct inet6_ifaddr *ifa;
562
563 idev = in6_dev_get(dev);
564 if (!idev)
565 return;
566
567 read_lock_bh(&idev->lock);
568 list_for_each_entry(ifa, &idev->addr_list, if_list) {
569 /* skip tentative addresses until dad completes */
570 if (ifa->flags & IFA_F_TENTATIVE &&
571 !(ifa->flags & IFA_F_OPTIMISTIC))
572 continue;
573
574 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
575 /*router=*/ !!idev->cnf.forwarding,
576 /*solicited=*/ false, /*override=*/ true,
577 /*inc_opt=*/ true);
578 }
579 read_unlock_bh(&idev->lock);
580
581 in6_dev_put(idev);
582}
583
584void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
585 const struct in6_addr *daddr, const struct in6_addr *saddr,
586 u64 nonce)
587{
588 struct sk_buff *skb;
589 struct in6_addr addr_buf;
590 int inc_opt = dev->addr_len;
591 int optlen = 0;
592 struct nd_msg *msg;
593
594 if (!saddr) {
595 if (ipv6_get_lladdr(dev, &addr_buf,
596 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
597 return;
598 saddr = &addr_buf;
599 }
600
601 if (ipv6_addr_any(saddr))
602 inc_opt = false;
603 if (inc_opt)
604 optlen += ndisc_opt_addr_space(dev,
605 NDISC_NEIGHBOUR_SOLICITATION);
606 if (nonce != 0)
607 optlen += 8;
608
609 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
610 if (!skb)
611 return;
612
613 msg = skb_put(skb, sizeof(*msg));
614 *msg = (struct nd_msg) {
615 .icmph = {
616 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
617 },
618 .target = *solicit,
619 };
620
621 if (inc_opt)
622 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
623 dev->dev_addr,
624 NDISC_NEIGHBOUR_SOLICITATION);
625 if (nonce != 0) {
626 u8 *opt = skb_put(skb, 8);
627
628 opt[0] = ND_OPT_NONCE;
629 opt[1] = 8 >> 3;
630 memcpy(opt + 2, &nonce, 6);
631 }
632
633 ndisc_send_skb(skb, daddr, saddr);
634}
635
636void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
637 const struct in6_addr *daddr)
638{
639 struct sk_buff *skb;
640 struct rs_msg *msg;
641 int send_sllao = dev->addr_len;
642 int optlen = 0;
643
644#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
645 /*
646 * According to section 2.2 of RFC 4429, we must not
647 * send router solicitations with a sllao from
648 * optimistic addresses, but we may send the solicitation
649 * if we don't include the sllao. So here we check
650 * if our address is optimistic, and if so, we
651 * suppress the inclusion of the sllao.
652 */
653 if (send_sllao) {
654 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
655 dev, 1);
656 if (ifp) {
657 if (ifp->flags & IFA_F_OPTIMISTIC) {
658 send_sllao = 0;
659 }
660 in6_ifa_put(ifp);
661 } else {
662 send_sllao = 0;
663 }
664 }
665#endif
666 if (send_sllao)
667 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
668
669 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
670 if (!skb)
671 return;
672
673 msg = skb_put(skb, sizeof(*msg));
674 *msg = (struct rs_msg) {
675 .icmph = {
676 .icmp6_type = NDISC_ROUTER_SOLICITATION,
677 },
678 };
679
680 if (send_sllao)
681 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
682 dev->dev_addr,
683 NDISC_ROUTER_SOLICITATION);
684
685 ndisc_send_skb(skb, daddr, saddr);
686}
687
688
689static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
690{
691 /*
692 * "The sender MUST return an ICMP
693 * destination unreachable"
694 */
695 dst_link_failure(skb);
696 kfree_skb(skb);
697}
698
699/* Called with locked neigh: either read or both */
700
701static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
702{
703 struct in6_addr *saddr = NULL;
704 struct in6_addr mcaddr;
705 struct net_device *dev = neigh->dev;
706 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
707 int probes = atomic_read(&neigh->probes);
708
709 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
710 dev, false, 1,
711 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
712 saddr = &ipv6_hdr(skb)->saddr;
713 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
714 if (probes < 0) {
715 if (!(neigh->nud_state & NUD_VALID)) {
716 ND_PRINTK(1, dbg,
717 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
718 __func__, target);
719 }
720 ndisc_send_ns(dev, target, target, saddr, 0);
721 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
722 neigh_app_ns(neigh);
723 } else {
724 addrconf_addr_solict_mult(target, &mcaddr);
725 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
726 }
727}
728
729static int pndisc_is_router(const void *pkey,
730 struct net_device *dev)
731{
732 struct pneigh_entry *n;
733 int ret = -1;
734
735 read_lock_bh(&nd_tbl.lock);
736 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
737 if (n)
738 ret = !!(n->flags & NTF_ROUTER);
739 read_unlock_bh(&nd_tbl.lock);
740
741 return ret;
742}
743
744void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
745 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
746 struct ndisc_options *ndopts)
747{
748 neigh_update(neigh, lladdr, new, flags, 0);
749 /* report ndisc ops about neighbour update */
750 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
751}
752
753static void ndisc_recv_ns(struct sk_buff *skb)
754{
755 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
756 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
757 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
758 u8 *lladdr = NULL;
759 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
760 offsetof(struct nd_msg, opt));
761 struct ndisc_options ndopts;
762 struct net_device *dev = skb->dev;
763 struct inet6_ifaddr *ifp;
764 struct inet6_dev *idev = NULL;
765 struct neighbour *neigh;
766 int dad = ipv6_addr_any(saddr);
767 bool inc;
768 int is_router = -1;
769 u64 nonce = 0;
770
771 if (skb->len < sizeof(struct nd_msg)) {
772 ND_PRINTK(2, warn, "NS: packet too short\n");
773 return;
774 }
775
776 if (ipv6_addr_is_multicast(&msg->target)) {
777 ND_PRINTK(2, warn, "NS: multicast target address\n");
778 return;
779 }
780
781 /*
782 * RFC2461 7.1.1:
783 * DAD has to be destined for solicited node multicast address.
784 */
785 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
786 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
787 return;
788 }
789
790 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
791 ND_PRINTK(2, warn, "NS: invalid ND options\n");
792 return;
793 }
794
795 if (ndopts.nd_opts_src_lladdr) {
796 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
797 if (!lladdr) {
798 ND_PRINTK(2, warn,
799 "NS: invalid link-layer address length\n");
800 return;
801 }
802
803 /* RFC2461 7.1.1:
804 * If the IP source address is the unspecified address,
805 * there MUST NOT be source link-layer address option
806 * in the message.
807 */
808 if (dad) {
809 ND_PRINTK(2, warn,
810 "NS: bad DAD packet (link-layer address option)\n");
811 return;
812 }
813 }
814 if (ndopts.nd_opts_nonce)
815 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
816
817 inc = ipv6_addr_is_multicast(daddr);
818
819 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
820 if (ifp) {
821have_ifp:
822 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
823 if (dad) {
824 if (nonce != 0 && ifp->dad_nonce == nonce) {
825 u8 *np = (u8 *)&nonce;
826 /* Matching nonce if looped back */
827 ND_PRINTK(2, notice,
828 "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
829 ifp->idev->dev->name,
830 &ifp->addr, np);
831 goto out;
832 }
833 /*
834 * We are colliding with another node
835 * who is doing DAD
836 * so fail our DAD process
837 */
838 addrconf_dad_failure(skb, ifp);
839 return;
840 } else {
841 /*
842 * This is not a dad solicitation.
843 * If we are an optimistic node,
844 * we should respond.
845 * Otherwise, we should ignore it.
846 */
847 if (!(ifp->flags & IFA_F_OPTIMISTIC))
848 goto out;
849 }
850 }
851
852 idev = ifp->idev;
853 } else {
854 struct net *net = dev_net(dev);
855
856 /* perhaps an address on the master device */
857 if (netif_is_l3_slave(dev)) {
858 struct net_device *mdev;
859
860 mdev = netdev_master_upper_dev_get_rcu(dev);
861 if (mdev) {
862 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
863 if (ifp)
864 goto have_ifp;
865 }
866 }
867
868 idev = in6_dev_get(dev);
869 if (!idev) {
870 /* XXX: count this drop? */
871 return;
872 }
873
874 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
875 (idev->cnf.forwarding &&
876 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
877 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
878 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
879 skb->pkt_type != PACKET_HOST &&
880 inc &&
881 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
882 /*
883 * for anycast or proxy,
884 * sender should delay its response
885 * by a random time between 0 and
886 * MAX_ANYCAST_DELAY_TIME seconds.
887 * (RFC2461) -- yoshfuji
888 */
889 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
890 if (n)
891 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
892 goto out;
893 }
894 } else
895 goto out;
896 }
897
898 if (is_router < 0)
899 is_router = idev->cnf.forwarding;
900
901 if (dad) {
902 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
903 !!is_router, false, (ifp != NULL), true);
904 goto out;
905 }
906
907 if (inc)
908 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
909 else
910 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
911
912 /*
913 * update / create cache entry
914 * for the source address
915 */
916 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
917 !inc || lladdr || !dev->addr_len);
918 if (neigh)
919 ndisc_update(dev, neigh, lladdr, NUD_STALE,
920 NEIGH_UPDATE_F_WEAK_OVERRIDE|
921 NEIGH_UPDATE_F_OVERRIDE,
922 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
923 if (neigh || !dev->header_ops) {
924 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
925 true, (ifp != NULL && inc), inc);
926 if (neigh)
927 neigh_release(neigh);
928 }
929
930out:
931 if (ifp)
932 in6_ifa_put(ifp);
933 else
934 in6_dev_put(idev);
935}
936
937static void ndisc_recv_na(struct sk_buff *skb)
938{
939 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
940 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
941 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
942 u8 *lladdr = NULL;
943 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
944 offsetof(struct nd_msg, opt));
945 struct ndisc_options ndopts;
946 struct net_device *dev = skb->dev;
947 struct inet6_dev *idev = __in6_dev_get(dev);
948 struct inet6_ifaddr *ifp;
949 struct neighbour *neigh;
950
951 if (skb->len < sizeof(struct nd_msg)) {
952 ND_PRINTK(2, warn, "NA: packet too short\n");
953 return;
954 }
955
956 if (ipv6_addr_is_multicast(&msg->target)) {
957 ND_PRINTK(2, warn, "NA: target address is multicast\n");
958 return;
959 }
960
961 if (ipv6_addr_is_multicast(daddr) &&
962 msg->icmph.icmp6_solicited) {
963 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
964 return;
965 }
966
967 /* For some 802.11 wireless deployments (and possibly other networks),
968 * there will be a NA proxy and unsolicitd packets are attacks
969 * and thus should not be accepted.
970 */
971 if (!msg->icmph.icmp6_solicited && idev &&
972 idev->cnf.drop_unsolicited_na)
973 return;
974
975 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
976 ND_PRINTK(2, warn, "NS: invalid ND option\n");
977 return;
978 }
979 if (ndopts.nd_opts_tgt_lladdr) {
980 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
981 if (!lladdr) {
982 ND_PRINTK(2, warn,
983 "NA: invalid link-layer address length\n");
984 return;
985 }
986 }
987 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
988 if (ifp) {
989 if (skb->pkt_type != PACKET_LOOPBACK
990 && (ifp->flags & IFA_F_TENTATIVE)) {
991 addrconf_dad_failure(skb, ifp);
992 return;
993 }
994 /* What should we make now? The advertisement
995 is invalid, but ndisc specs say nothing
996 about it. It could be misconfiguration, or
997 an smart proxy agent tries to help us :-)
998
999 We should not print the error if NA has been
1000 received from loopback - it is just our own
1001 unsolicited advertisement.
1002 */
1003 if (skb->pkt_type != PACKET_LOOPBACK)
1004 ND_PRINTK(1, warn,
1005 "NA: %pM advertised our address %pI6c on %s!\n",
1006 eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1007 in6_ifa_put(ifp);
1008 return;
1009 }
1010 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1011
1012 if (neigh) {
1013 u8 old_flags = neigh->flags;
1014 struct net *net = dev_net(dev);
1015
1016 if (neigh->nud_state & NUD_FAILED)
1017 goto out;
1018
1019 /*
1020 * Don't update the neighbor cache entry on a proxy NA from
1021 * ourselves because either the proxied node is off link or it
1022 * has already sent a NA to us.
1023 */
1024 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1025 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1026 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1027 /* XXX: idev->cnf.proxy_ndp */
1028 goto out;
1029 }
1030
1031 ndisc_update(dev, neigh, lladdr,
1032 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1033 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1034 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1035 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1036 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1037 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1038
1039 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1040 /*
1041 * Change: router to host
1042 */
1043 rt6_clean_tohost(dev_net(dev), saddr);
1044 }
1045
1046out:
1047 neigh_release(neigh);
1048 }
1049}
1050
1051static void ndisc_recv_rs(struct sk_buff *skb)
1052{
1053 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1054 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1055 struct neighbour *neigh;
1056 struct inet6_dev *idev;
1057 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1058 struct ndisc_options ndopts;
1059 u8 *lladdr = NULL;
1060
1061 if (skb->len < sizeof(*rs_msg))
1062 return;
1063
1064 idev = __in6_dev_get(skb->dev);
1065 if (!idev) {
1066 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1067 return;
1068 }
1069
1070 /* Don't accept RS if we're not in router mode */
1071 if (!idev->cnf.forwarding)
1072 goto out;
1073
1074 /*
1075 * Don't update NCE if src = ::;
1076 * this implies that the source node has no ip address assigned yet.
1077 */
1078 if (ipv6_addr_any(saddr))
1079 goto out;
1080
1081 /* Parse ND options */
1082 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1083 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1084 goto out;
1085 }
1086
1087 if (ndopts.nd_opts_src_lladdr) {
1088 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1089 skb->dev);
1090 if (!lladdr)
1091 goto out;
1092 }
1093
1094 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1095 if (neigh) {
1096 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1097 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1098 NEIGH_UPDATE_F_OVERRIDE|
1099 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1100 NDISC_ROUTER_SOLICITATION, &ndopts);
1101 neigh_release(neigh);
1102 }
1103out:
1104 return;
1105}
1106
1107static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1108{
1109 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1110 struct sk_buff *skb;
1111 struct nlmsghdr *nlh;
1112 struct nduseroptmsg *ndmsg;
1113 struct net *net = dev_net(ra->dev);
1114 int err;
1115 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1116 + (opt->nd_opt_len << 3));
1117 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1118
1119 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1120 if (!skb) {
1121 err = -ENOBUFS;
1122 goto errout;
1123 }
1124
1125 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1126 if (!nlh) {
1127 goto nla_put_failure;
1128 }
1129
1130 ndmsg = nlmsg_data(nlh);
1131 ndmsg->nduseropt_family = AF_INET6;
1132 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1133 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1134 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1135 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1136
1137 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1138
1139 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1140 goto nla_put_failure;
1141 nlmsg_end(skb, nlh);
1142
1143 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1144 return;
1145
1146nla_put_failure:
1147 nlmsg_free(skb);
1148 err = -EMSGSIZE;
1149errout:
1150 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1151}
1152
1153static void ndisc_router_discovery(struct sk_buff *skb)
1154{
1155 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1156 struct neighbour *neigh = NULL;
1157 struct inet6_dev *in6_dev;
1158 struct rt6_info *rt = NULL;
1159 int lifetime;
1160 struct ndisc_options ndopts;
1161 int optlen;
1162 unsigned int pref = 0;
1163 __u32 old_if_flags;
1164 bool send_ifinfo_notify = false;
1165
1166 __u8 *opt = (__u8 *)(ra_msg + 1);
1167
1168 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1169 sizeof(struct ra_msg);
1170
1171 ND_PRINTK(2, info,
1172 "RA: %s, dev: %s\n",
1173 __func__, skb->dev->name);
1174 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1175 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1176 return;
1177 }
1178 if (optlen < 0) {
1179 ND_PRINTK(2, warn, "RA: packet too short\n");
1180 return;
1181 }
1182
1183#ifdef CONFIG_IPV6_NDISC_NODETYPE
1184 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1185 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1186 return;
1187 }
1188#endif
1189
1190 /*
1191 * set the RA_RECV flag in the interface
1192 */
1193
1194 in6_dev = __in6_dev_get(skb->dev);
1195 if (!in6_dev) {
1196 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1197 skb->dev->name);
1198 return;
1199 }
1200
1201 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1202 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1203 return;
1204 }
1205
1206 if (!ipv6_accept_ra(in6_dev)) {
1207 ND_PRINTK(2, info,
1208 "RA: %s, did not accept ra for dev: %s\n",
1209 __func__, skb->dev->name);
1210 goto skip_linkparms;
1211 }
1212
1213#ifdef CONFIG_IPV6_NDISC_NODETYPE
1214 /* skip link-specific parameters from interior routers */
1215 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1216 ND_PRINTK(2, info,
1217 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1218 __func__, skb->dev->name);
1219 goto skip_linkparms;
1220 }
1221#endif
1222
1223 if (in6_dev->if_flags & IF_RS_SENT) {
1224 /*
1225 * flag that an RA was received after an RS was sent
1226 * out on this interface.
1227 */
1228 in6_dev->if_flags |= IF_RA_RCVD;
1229 }
1230
1231 /*
1232 * Remember the managed/otherconf flags from most recently
1233 * received RA message (RFC 2462) -- yoshfuji
1234 */
1235 old_if_flags = in6_dev->if_flags;
1236 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1237 IF_RA_OTHERCONF)) |
1238 (ra_msg->icmph.icmp6_addrconf_managed ?
1239 IF_RA_MANAGED : 0) |
1240 (ra_msg->icmph.icmp6_addrconf_other ?
1241 IF_RA_OTHERCONF : 0);
1242
1243 if (old_if_flags != in6_dev->if_flags)
1244 send_ifinfo_notify = true;
1245
1246 if (!in6_dev->cnf.accept_ra_defrtr) {
1247 ND_PRINTK(2, info,
1248 "RA: %s, defrtr is false for dev: %s\n",
1249 __func__, skb->dev->name);
1250 goto skip_defrtr;
1251 }
1252
1253 /* Do not accept RA with source-addr found on local machine unless
1254 * accept_ra_from_local is set to true.
1255 */
1256 if (!in6_dev->cnf.accept_ra_from_local &&
1257 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1258 in6_dev->dev, 0)) {
1259 ND_PRINTK(2, info,
1260 "RA from local address detected on dev: %s: default router ignored\n",
1261 skb->dev->name);
1262 goto skip_defrtr;
1263 }
1264
1265 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1266
1267#ifdef CONFIG_IPV6_ROUTER_PREF
1268 pref = ra_msg->icmph.icmp6_router_pref;
1269 /* 10b is handled as if it were 00b (medium) */
1270 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1271 !in6_dev->cnf.accept_ra_rtr_pref)
1272 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1273#endif
1274
1275 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1276
1277 if (rt) {
1278 neigh = dst_neigh_lookup(&rt->dst, &ipv6_hdr(skb)->saddr);
1279 if (!neigh) {
1280 ND_PRINTK(0, err,
1281 "RA: %s got default router without neighbour\n",
1282 __func__);
1283 ip6_rt_put(rt);
1284 return;
1285 }
1286 }
1287 if (rt && lifetime == 0) {
1288 ip6_del_rt(rt);
1289 rt = NULL;
1290 }
1291
1292 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, for dev: %s\n",
1293 rt, lifetime, skb->dev->name);
1294 if (!rt && lifetime) {
1295 ND_PRINTK(3, info, "RA: adding default router\n");
1296
1297 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1298 if (!rt) {
1299 ND_PRINTK(0, err,
1300 "RA: %s failed to add default route\n",
1301 __func__);
1302 return;
1303 }
1304
1305 neigh = dst_neigh_lookup(&rt->dst, &ipv6_hdr(skb)->saddr);
1306 if (!neigh) {
1307 ND_PRINTK(0, err,
1308 "RA: %s got default router without neighbour\n",
1309 __func__);
1310 ip6_rt_put(rt);
1311 return;
1312 }
1313 neigh->flags |= NTF_ROUTER;
1314 } else if (rt) {
1315 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1316 }
1317
1318 if (rt)
1319 rt6_set_expires(rt, jiffies + (HZ * lifetime));
1320 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1321 ra_msg->icmph.icmp6_hop_limit) {
1322 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1323 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1324 if (rt)
1325 dst_metric_set(&rt->dst, RTAX_HOPLIMIT,
1326 ra_msg->icmph.icmp6_hop_limit);
1327 } else {
1328 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1329 }
1330 }
1331
1332skip_defrtr:
1333
1334 /*
1335 * Update Reachable Time and Retrans Timer
1336 */
1337
1338 if (in6_dev->nd_parms) {
1339 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1340
1341 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1342 rtime = (rtime*HZ)/1000;
1343 if (rtime < HZ/10)
1344 rtime = HZ/10;
1345 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1346 in6_dev->tstamp = jiffies;
1347 send_ifinfo_notify = true;
1348 }
1349
1350 rtime = ntohl(ra_msg->reachable_time);
1351 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1352 rtime = (rtime*HZ)/1000;
1353
1354 if (rtime < HZ/10)
1355 rtime = HZ/10;
1356
1357 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1358 NEIGH_VAR_SET(in6_dev->nd_parms,
1359 BASE_REACHABLE_TIME, rtime);
1360 NEIGH_VAR_SET(in6_dev->nd_parms,
1361 GC_STALETIME, 3 * rtime);
1362 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1363 in6_dev->tstamp = jiffies;
1364 send_ifinfo_notify = true;
1365 }
1366 }
1367 }
1368
1369 /*
1370 * Send a notify if RA changed managed/otherconf flags or timer settings
1371 */
1372 if (send_ifinfo_notify)
1373 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1374
1375skip_linkparms:
1376
1377 /*
1378 * Process options.
1379 */
1380
1381 if (!neigh)
1382 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1383 skb->dev, 1);
1384 if (neigh) {
1385 u8 *lladdr = NULL;
1386 if (ndopts.nd_opts_src_lladdr) {
1387 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1388 skb->dev);
1389 if (!lladdr) {
1390 ND_PRINTK(2, warn,
1391 "RA: invalid link-layer address length\n");
1392 goto out;
1393 }
1394 }
1395 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1396 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1397 NEIGH_UPDATE_F_OVERRIDE|
1398 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1399 NEIGH_UPDATE_F_ISROUTER,
1400 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1401 }
1402
1403 if (!ipv6_accept_ra(in6_dev)) {
1404 ND_PRINTK(2, info,
1405 "RA: %s, accept_ra is false for dev: %s\n",
1406 __func__, skb->dev->name);
1407 goto out;
1408 }
1409
1410#ifdef CONFIG_IPV6_ROUTE_INFO
1411 if (!in6_dev->cnf.accept_ra_from_local &&
1412 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1413 in6_dev->dev, 0)) {
1414 ND_PRINTK(2, info,
1415 "RA from local address detected on dev: %s: router info ignored.\n",
1416 skb->dev->name);
1417 goto skip_routeinfo;
1418 }
1419
1420 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1421 struct nd_opt_hdr *p;
1422 for (p = ndopts.nd_opts_ri;
1423 p;
1424 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1425 struct route_info *ri = (struct route_info *)p;
1426#ifdef CONFIG_IPV6_NDISC_NODETYPE
1427 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1428 ri->prefix_len == 0)
1429 continue;
1430#endif
1431 if (ri->prefix_len == 0 &&
1432 !in6_dev->cnf.accept_ra_defrtr)
1433 continue;
1434 if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1435 continue;
1436 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1437 continue;
1438 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1439 &ipv6_hdr(skb)->saddr);
1440 }
1441 }
1442
1443skip_routeinfo:
1444#endif
1445
1446#ifdef CONFIG_IPV6_NDISC_NODETYPE
1447 /* skip link-specific ndopts from interior routers */
1448 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1449 ND_PRINTK(2, info,
1450 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1451 __func__, skb->dev->name);
1452 goto out;
1453 }
1454#endif
1455
1456 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1457 struct nd_opt_hdr *p;
1458 for (p = ndopts.nd_opts_pi;
1459 p;
1460 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1461 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1462 (p->nd_opt_len) << 3,
1463 ndopts.nd_opts_src_lladdr != NULL);
1464 }
1465 }
1466
1467 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1468 __be32 n;
1469 u32 mtu;
1470
1471 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1472 mtu = ntohl(n);
1473
1474 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1475 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1476 } else if (in6_dev->cnf.mtu6 != mtu) {
1477 in6_dev->cnf.mtu6 = mtu;
1478
1479 if (rt)
1480 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
1481
1482 rt6_mtu_change(skb->dev, mtu);
1483 }
1484 }
1485
1486 if (ndopts.nd_useropts) {
1487 struct nd_opt_hdr *p;
1488 for (p = ndopts.nd_useropts;
1489 p;
1490 p = ndisc_next_useropt(skb->dev, p,
1491 ndopts.nd_useropts_end)) {
1492 ndisc_ra_useropt(skb, p);
1493 }
1494 }
1495
1496 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1497 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1498 }
1499out:
1500 ip6_rt_put(rt);
1501 if (neigh)
1502 neigh_release(neigh);
1503}
1504
1505static void ndisc_redirect_rcv(struct sk_buff *skb)
1506{
1507 u8 *hdr;
1508 struct ndisc_options ndopts;
1509 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1510 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1511 offsetof(struct rd_msg, opt));
1512
1513#ifdef CONFIG_IPV6_NDISC_NODETYPE
1514 switch (skb->ndisc_nodetype) {
1515 case NDISC_NODETYPE_HOST:
1516 case NDISC_NODETYPE_NODEFAULT:
1517 ND_PRINTK(2, warn,
1518 "Redirect: from host or unauthorized router\n");
1519 return;
1520 }
1521#endif
1522
1523 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1524 ND_PRINTK(2, warn,
1525 "Redirect: source address is not link-local\n");
1526 return;
1527 }
1528
1529 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1530 return;
1531
1532 if (!ndopts.nd_opts_rh) {
1533 ip6_redirect_no_header(skb, dev_net(skb->dev),
1534 skb->dev->ifindex, 0);
1535 return;
1536 }
1537
1538 hdr = (u8 *)ndopts.nd_opts_rh;
1539 hdr += 8;
1540 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1541 return;
1542
1543 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1544}
1545
1546static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1547 struct sk_buff *orig_skb,
1548 int rd_len)
1549{
1550 u8 *opt = skb_put(skb, rd_len);
1551
1552 memset(opt, 0, 8);
1553 *(opt++) = ND_OPT_REDIRECT_HDR;
1554 *(opt++) = (rd_len >> 3);
1555 opt += 6;
1556
1557 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1558 rd_len - 8);
1559}
1560
1561void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1562{
1563 struct net_device *dev = skb->dev;
1564 struct net *net = dev_net(dev);
1565 struct sock *sk = net->ipv6.ndisc_sk;
1566 int optlen = 0;
1567 struct inet_peer *peer;
1568 struct sk_buff *buff;
1569 struct rd_msg *msg;
1570 struct in6_addr saddr_buf;
1571 struct rt6_info *rt;
1572 struct dst_entry *dst;
1573 struct flowi6 fl6;
1574 int rd_len;
1575 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1576 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1577 bool ret;
1578
1579 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1580 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1581 dev->name);
1582 return;
1583 }
1584
1585 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1586 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1587 ND_PRINTK(2, warn,
1588 "Redirect: target address is not link-local unicast\n");
1589 return;
1590 }
1591
1592 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1593 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1594
1595 dst = ip6_route_output(net, NULL, &fl6);
1596 if (dst->error) {
1597 dst_release(dst);
1598 return;
1599 }
1600 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1601 if (IS_ERR(dst))
1602 return;
1603
1604 rt = (struct rt6_info *) dst;
1605
1606 if (rt->rt6i_flags & RTF_GATEWAY) {
1607 ND_PRINTK(2, warn,
1608 "Redirect: destination is not a neighbour\n");
1609 goto release;
1610 }
1611 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1612 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1613 if (peer)
1614 inet_putpeer(peer);
1615 if (!ret)
1616 goto release;
1617
1618 if (dev->addr_len) {
1619 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1620 if (!neigh) {
1621 ND_PRINTK(2, warn,
1622 "Redirect: no neigh for target address\n");
1623 goto release;
1624 }
1625
1626 read_lock_bh(&neigh->lock);
1627 if (neigh->nud_state & NUD_VALID) {
1628 memcpy(ha_buf, neigh->ha, dev->addr_len);
1629 read_unlock_bh(&neigh->lock);
1630 ha = ha_buf;
1631 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1632 ops_data_buf,
1633 &ops_data);
1634 } else
1635 read_unlock_bh(&neigh->lock);
1636
1637 neigh_release(neigh);
1638 }
1639
1640 rd_len = min_t(unsigned int,
1641 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1642 skb->len + 8);
1643 rd_len &= ~0x7;
1644 optlen += rd_len;
1645
1646 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1647 if (!buff)
1648 goto release;
1649
1650 msg = skb_put(buff, sizeof(*msg));
1651 *msg = (struct rd_msg) {
1652 .icmph = {
1653 .icmp6_type = NDISC_REDIRECT,
1654 },
1655 .target = *target,
1656 .dest = ipv6_hdr(skb)->daddr,
1657 };
1658
1659 /*
1660 * include target_address option
1661 */
1662
1663 if (ha)
1664 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1665
1666 /*
1667 * build redirect option and copy skb over to the new packet.
1668 */
1669
1670 if (rd_len)
1671 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1672
1673 skb_dst_set(buff, dst);
1674 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1675 return;
1676
1677release:
1678 dst_release(dst);
1679}
1680
1681static void pndisc_redo(struct sk_buff *skb)
1682{
1683 ndisc_recv_ns(skb);
1684 kfree_skb(skb);
1685}
1686
1687static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1688{
1689 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1690
1691 if (!idev)
1692 return true;
1693 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1694 idev->cnf.suppress_frag_ndisc) {
1695 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1696 return true;
1697 }
1698 return false;
1699}
1700
1701int ndisc_rcv(struct sk_buff *skb)
1702{
1703 struct nd_msg *msg;
1704
1705 if (ndisc_suppress_frag_ndisc(skb))
1706 return 0;
1707
1708 if (skb_linearize(skb))
1709 return 0;
1710
1711 msg = (struct nd_msg *)skb_transport_header(skb);
1712
1713 __skb_push(skb, skb->data - skb_transport_header(skb));
1714
1715 if (ipv6_hdr(skb)->hop_limit != 255) {
1716 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1717 ipv6_hdr(skb)->hop_limit);
1718 return 0;
1719 }
1720
1721 if (msg->icmph.icmp6_code != 0) {
1722 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1723 msg->icmph.icmp6_code);
1724 return 0;
1725 }
1726
1727 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1728
1729 switch (msg->icmph.icmp6_type) {
1730 case NDISC_NEIGHBOUR_SOLICITATION:
1731 ndisc_recv_ns(skb);
1732 break;
1733
1734 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1735 ndisc_recv_na(skb);
1736 break;
1737
1738 case NDISC_ROUTER_SOLICITATION:
1739 ndisc_recv_rs(skb);
1740 break;
1741
1742 case NDISC_ROUTER_ADVERTISEMENT:
1743 ndisc_router_discovery(skb);
1744 break;
1745
1746 case NDISC_REDIRECT:
1747 ndisc_redirect_rcv(skb);
1748 break;
1749 }
1750
1751 return 0;
1752}
1753
1754static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1755{
1756 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1757 struct netdev_notifier_change_info *change_info;
1758 struct net *net = dev_net(dev);
1759 struct inet6_dev *idev;
1760
1761 switch (event) {
1762 case NETDEV_CHANGEADDR:
1763 neigh_changeaddr(&nd_tbl, dev);
1764 fib6_run_gc(0, net, false);
1765 /* fallthrough */
1766 case NETDEV_UP:
1767 idev = in6_dev_get(dev);
1768 if (!idev)
1769 break;
1770 if (idev->cnf.ndisc_notify ||
1771 net->ipv6.devconf_all->ndisc_notify)
1772 ndisc_send_unsol_na(dev);
1773 in6_dev_put(idev);
1774 break;
1775 case NETDEV_CHANGE:
1776 change_info = ptr;
1777 if (change_info->flags_changed & IFF_NOARP)
1778 neigh_changeaddr(&nd_tbl, dev);
1779 break;
1780 case NETDEV_DOWN:
1781 neigh_ifdown(&nd_tbl, dev);
1782 fib6_run_gc(0, net, false);
1783 break;
1784 case NETDEV_NOTIFY_PEERS:
1785 ndisc_send_unsol_na(dev);
1786 break;
1787 default:
1788 break;
1789 }
1790
1791 return NOTIFY_DONE;
1792}
1793
1794static struct notifier_block ndisc_netdev_notifier = {
1795 .notifier_call = ndisc_netdev_event,
1796 .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1797};
1798
1799#ifdef CONFIG_SYSCTL
1800static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1801 const char *func, const char *dev_name)
1802{
1803 static char warncomm[TASK_COMM_LEN];
1804 static int warned;
1805 if (strcmp(warncomm, current->comm) && warned < 5) {
1806 strcpy(warncomm, current->comm);
1807 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1808 warncomm, func,
1809 dev_name, ctl->procname,
1810 dev_name, ctl->procname);
1811 warned++;
1812 }
1813}
1814
1815int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1816{
1817 struct net_device *dev = ctl->extra1;
1818 struct inet6_dev *idev;
1819 int ret;
1820
1821 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1822 (strcmp(ctl->procname, "base_reachable_time") == 0))
1823 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1824
1825 if (strcmp(ctl->procname, "retrans_time") == 0)
1826 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1827
1828 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1829 ret = neigh_proc_dointvec_jiffies(ctl, write,
1830 buffer, lenp, ppos);
1831
1832 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1833 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1834 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1835 buffer, lenp, ppos);
1836 else
1837 ret = -1;
1838
1839 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1840 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1841 idev->nd_parms->reachable_time =
1842 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1843 idev->tstamp = jiffies;
1844 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1845 in6_dev_put(idev);
1846 }
1847 return ret;
1848}
1849
1850
1851#endif
1852
1853static int __net_init ndisc_net_init(struct net *net)
1854{
1855 struct ipv6_pinfo *np;
1856 struct sock *sk;
1857 int err;
1858
1859 err = inet_ctl_sock_create(&sk, PF_INET6,
1860 SOCK_RAW, IPPROTO_ICMPV6, net);
1861 if (err < 0) {
1862 ND_PRINTK(0, err,
1863 "NDISC: Failed to initialize the control socket (err %d)\n",
1864 err);
1865 return err;
1866 }
1867
1868 net->ipv6.ndisc_sk = sk;
1869
1870 np = inet6_sk(sk);
1871 np->hop_limit = 255;
1872 /* Do not loopback ndisc messages */
1873 np->mc_loop = 0;
1874
1875 return 0;
1876}
1877
1878static void __net_exit ndisc_net_exit(struct net *net)
1879{
1880 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1881}
1882
1883static struct pernet_operations ndisc_net_ops = {
1884 .init = ndisc_net_init,
1885 .exit = ndisc_net_exit,
1886};
1887
1888int __init ndisc_init(void)
1889{
1890 int err;
1891
1892 err = register_pernet_subsys(&ndisc_net_ops);
1893 if (err)
1894 return err;
1895 /*
1896 * Initialize the neighbour table
1897 */
1898 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1899
1900#ifdef CONFIG_SYSCTL
1901 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1902 ndisc_ifinfo_sysctl_change);
1903 if (err)
1904 goto out_unregister_pernet;
1905out:
1906#endif
1907 return err;
1908
1909#ifdef CONFIG_SYSCTL
1910out_unregister_pernet:
1911 unregister_pernet_subsys(&ndisc_net_ops);
1912 goto out;
1913#endif
1914}
1915
1916int __init ndisc_late_init(void)
1917{
1918 return register_netdevice_notifier(&ndisc_netdev_notifier);
1919}
1920
1921void ndisc_late_cleanup(void)
1922{
1923 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1924}
1925
1926void ndisc_cleanup(void)
1927{
1928#ifdef CONFIG_SYSCTL
1929 neigh_sysctl_unregister(&nd_tbl.parms);
1930#endif
1931 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1932 unregister_pernet_subsys(&ndisc_net_ops);
1933}
1/*
2 * Neighbour Discovery for IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Mike Shaver <shaver@ingenia.com>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
13 */
14
15/*
16 * Changes:
17 *
18 * Alexey I. Froloff : RFC6106 (DNSSL) support
19 * Pierre Ynard : export userland ND options
20 * through netlink (RDNSS support)
21 * Lars Fenneberg : fixed MTU setting on receipt
22 * of an RA.
23 * Janos Farkas : kmalloc failure checks
24 * Alexey Kuznetsov : state machine reworked
25 * and moved to net/core.
26 * Pekka Savola : RFC2461 validation
27 * YOSHIFUJI Hideaki @USAGI : Verify ND options properly
28 */
29
30#define pr_fmt(fmt) "ICMPv6: " fmt
31
32#include <linux/module.h>
33#include <linux/errno.h>
34#include <linux/types.h>
35#include <linux/socket.h>
36#include <linux/sockios.h>
37#include <linux/sched.h>
38#include <linux/net.h>
39#include <linux/in6.h>
40#include <linux/route.h>
41#include <linux/init.h>
42#include <linux/rcupdate.h>
43#include <linux/slab.h>
44#ifdef CONFIG_SYSCTL
45#include <linux/sysctl.h>
46#endif
47
48#include <linux/if_addr.h>
49#include <linux/if_arp.h>
50#include <linux/ipv6.h>
51#include <linux/icmpv6.h>
52#include <linux/jhash.h>
53
54#include <net/sock.h>
55#include <net/snmp.h>
56
57#include <net/ipv6.h>
58#include <net/protocol.h>
59#include <net/ndisc.h>
60#include <net/ip6_route.h>
61#include <net/addrconf.h>
62#include <net/icmp.h>
63
64#include <net/netlink.h>
65#include <linux/rtnetlink.h>
66
67#include <net/flow.h>
68#include <net/ip6_checksum.h>
69#include <net/inet_common.h>
70#include <linux/proc_fs.h>
71
72#include <linux/netfilter.h>
73#include <linux/netfilter_ipv6.h>
74
75/* Set to 3 to get tracing... */
76#define ND_DEBUG 1
77
78#define ND_PRINTK(val, level, fmt, ...) \
79do { \
80 if (val <= ND_DEBUG) \
81 net_##level##_ratelimited(fmt, ##__VA_ARGS__); \
82} while (0)
83
84static u32 ndisc_hash(const void *pkey,
85 const struct net_device *dev,
86 __u32 *hash_rnd);
87static int ndisc_constructor(struct neighbour *neigh);
88static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
89static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
90static int pndisc_constructor(struct pneigh_entry *n);
91static void pndisc_destructor(struct pneigh_entry *n);
92static void pndisc_redo(struct sk_buff *skb);
93
94static const struct neigh_ops ndisc_generic_ops = {
95 .family = AF_INET6,
96 .solicit = ndisc_solicit,
97 .error_report = ndisc_error_report,
98 .output = neigh_resolve_output,
99 .connected_output = neigh_connected_output,
100};
101
102static const struct neigh_ops ndisc_hh_ops = {
103 .family = AF_INET6,
104 .solicit = ndisc_solicit,
105 .error_report = ndisc_error_report,
106 .output = neigh_resolve_output,
107 .connected_output = neigh_resolve_output,
108};
109
110
111static const struct neigh_ops ndisc_direct_ops = {
112 .family = AF_INET6,
113 .output = neigh_direct_output,
114 .connected_output = neigh_direct_output,
115};
116
117struct neigh_table nd_tbl = {
118 .family = AF_INET6,
119 .key_len = sizeof(struct in6_addr),
120 .hash = ndisc_hash,
121 .constructor = ndisc_constructor,
122 .pconstructor = pndisc_constructor,
123 .pdestructor = pndisc_destructor,
124 .proxy_redo = pndisc_redo,
125 .id = "ndisc_cache",
126 .parms = {
127 .tbl = &nd_tbl,
128 .base_reachable_time = ND_REACHABLE_TIME,
129 .retrans_time = ND_RETRANS_TIMER,
130 .gc_staletime = 60 * HZ,
131 .reachable_time = ND_REACHABLE_TIME,
132 .delay_probe_time = 5 * HZ,
133 .queue_len_bytes = 64*1024,
134 .ucast_probes = 3,
135 .mcast_probes = 3,
136 .anycast_delay = 1 * HZ,
137 .proxy_delay = (8 * HZ) / 10,
138 .proxy_qlen = 64,
139 },
140 .gc_interval = 30 * HZ,
141 .gc_thresh1 = 128,
142 .gc_thresh2 = 512,
143 .gc_thresh3 = 1024,
144};
145
146/* ND options */
147struct ndisc_options {
148 struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
149#ifdef CONFIG_IPV6_ROUTE_INFO
150 struct nd_opt_hdr *nd_opts_ri;
151 struct nd_opt_hdr *nd_opts_ri_end;
152#endif
153 struct nd_opt_hdr *nd_useropts;
154 struct nd_opt_hdr *nd_useropts_end;
155};
156
157#define nd_opts_src_lladdr nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
158#define nd_opts_tgt_lladdr nd_opt_array[ND_OPT_TARGET_LL_ADDR]
159#define nd_opts_pi nd_opt_array[ND_OPT_PREFIX_INFO]
160#define nd_opts_pi_end nd_opt_array[__ND_OPT_PREFIX_INFO_END]
161#define nd_opts_rh nd_opt_array[ND_OPT_REDIRECT_HDR]
162#define nd_opts_mtu nd_opt_array[ND_OPT_MTU]
163
164#define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
165
166/*
167 * Return the padding between the option length and the start of the
168 * link addr. Currently only IP-over-InfiniBand needs this, although
169 * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
170 * also need a pad of 2.
171 */
172static int ndisc_addr_option_pad(unsigned short type)
173{
174 switch (type) {
175 case ARPHRD_INFINIBAND: return 2;
176 default: return 0;
177 }
178}
179
180static inline int ndisc_opt_addr_space(struct net_device *dev)
181{
182 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
183}
184
185static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
186 unsigned short addr_type)
187{
188 int space = NDISC_OPT_SPACE(data_len);
189 int pad = ndisc_addr_option_pad(addr_type);
190
191 opt[0] = type;
192 opt[1] = space>>3;
193
194 memset(opt + 2, 0, pad);
195 opt += pad;
196 space -= pad;
197
198 memcpy(opt+2, data, data_len);
199 data_len += 2;
200 opt += data_len;
201 if ((space -= data_len) > 0)
202 memset(opt, 0, space);
203 return opt + space;
204}
205
206static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
207 struct nd_opt_hdr *end)
208{
209 int type;
210 if (!cur || !end || cur >= end)
211 return NULL;
212 type = cur->nd_opt_type;
213 do {
214 cur = ((void *)cur) + (cur->nd_opt_len << 3);
215 } while(cur < end && cur->nd_opt_type != type);
216 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
217}
218
219static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
220{
221 return opt->nd_opt_type == ND_OPT_RDNSS ||
222 opt->nd_opt_type == ND_OPT_DNSSL;
223}
224
225static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
226 struct nd_opt_hdr *end)
227{
228 if (!cur || !end || cur >= end)
229 return NULL;
230 do {
231 cur = ((void *)cur) + (cur->nd_opt_len << 3);
232 } while(cur < end && !ndisc_is_useropt(cur));
233 return cur <= end && ndisc_is_useropt(cur) ? cur : NULL;
234}
235
236static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
237 struct ndisc_options *ndopts)
238{
239 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
240
241 if (!nd_opt || opt_len < 0 || !ndopts)
242 return NULL;
243 memset(ndopts, 0, sizeof(*ndopts));
244 while (opt_len) {
245 int l;
246 if (opt_len < sizeof(struct nd_opt_hdr))
247 return NULL;
248 l = nd_opt->nd_opt_len << 3;
249 if (opt_len < l || l == 0)
250 return NULL;
251 switch (nd_opt->nd_opt_type) {
252 case ND_OPT_SOURCE_LL_ADDR:
253 case ND_OPT_TARGET_LL_ADDR:
254 case ND_OPT_MTU:
255 case ND_OPT_REDIRECT_HDR:
256 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
257 ND_PRINTK(2, warn,
258 "%s: duplicated ND6 option found: type=%d\n",
259 __func__, nd_opt->nd_opt_type);
260 } else {
261 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
262 }
263 break;
264 case ND_OPT_PREFIX_INFO:
265 ndopts->nd_opts_pi_end = nd_opt;
266 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
267 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
268 break;
269#ifdef CONFIG_IPV6_ROUTE_INFO
270 case ND_OPT_ROUTE_INFO:
271 ndopts->nd_opts_ri_end = nd_opt;
272 if (!ndopts->nd_opts_ri)
273 ndopts->nd_opts_ri = nd_opt;
274 break;
275#endif
276 default:
277 if (ndisc_is_useropt(nd_opt)) {
278 ndopts->nd_useropts_end = nd_opt;
279 if (!ndopts->nd_useropts)
280 ndopts->nd_useropts = nd_opt;
281 } else {
282 /*
283 * Unknown options must be silently ignored,
284 * to accommodate future extension to the
285 * protocol.
286 */
287 ND_PRINTK(2, notice,
288 "%s: ignored unsupported option; type=%d, len=%d\n",
289 __func__,
290 nd_opt->nd_opt_type,
291 nd_opt->nd_opt_len);
292 }
293 }
294 opt_len -= l;
295 nd_opt = ((void *)nd_opt) + l;
296 }
297 return ndopts;
298}
299
300static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
301 struct net_device *dev)
302{
303 u8 *lladdr = (u8 *)(p + 1);
304 int lladdrlen = p->nd_opt_len << 3;
305 int prepad = ndisc_addr_option_pad(dev->type);
306 if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
307 return NULL;
308 return lladdr + prepad;
309}
310
311int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
312{
313 switch (dev->type) {
314 case ARPHRD_ETHER:
315 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
316 case ARPHRD_FDDI:
317 ipv6_eth_mc_map(addr, buf);
318 return 0;
319 case ARPHRD_ARCNET:
320 ipv6_arcnet_mc_map(addr, buf);
321 return 0;
322 case ARPHRD_INFINIBAND:
323 ipv6_ib_mc_map(addr, dev->broadcast, buf);
324 return 0;
325 case ARPHRD_IPGRE:
326 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
327 default:
328 if (dir) {
329 memcpy(buf, dev->broadcast, dev->addr_len);
330 return 0;
331 }
332 }
333 return -EINVAL;
334}
335
336EXPORT_SYMBOL(ndisc_mc_map);
337
338static u32 ndisc_hash(const void *pkey,
339 const struct net_device *dev,
340 __u32 *hash_rnd)
341{
342 return ndisc_hashfn(pkey, dev, hash_rnd);
343}
344
345static int ndisc_constructor(struct neighbour *neigh)
346{
347 struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
348 struct net_device *dev = neigh->dev;
349 struct inet6_dev *in6_dev;
350 struct neigh_parms *parms;
351 bool is_multicast = ipv6_addr_is_multicast(addr);
352
353 in6_dev = in6_dev_get(dev);
354 if (in6_dev == NULL) {
355 return -EINVAL;
356 }
357
358 parms = in6_dev->nd_parms;
359 __neigh_parms_put(neigh->parms);
360 neigh->parms = neigh_parms_clone(parms);
361
362 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
363 if (!dev->header_ops) {
364 neigh->nud_state = NUD_NOARP;
365 neigh->ops = &ndisc_direct_ops;
366 neigh->output = neigh_direct_output;
367 } else {
368 if (is_multicast) {
369 neigh->nud_state = NUD_NOARP;
370 ndisc_mc_map(addr, neigh->ha, dev, 1);
371 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
372 neigh->nud_state = NUD_NOARP;
373 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
374 if (dev->flags&IFF_LOOPBACK)
375 neigh->type = RTN_LOCAL;
376 } else if (dev->flags&IFF_POINTOPOINT) {
377 neigh->nud_state = NUD_NOARP;
378 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
379 }
380 if (dev->header_ops->cache)
381 neigh->ops = &ndisc_hh_ops;
382 else
383 neigh->ops = &ndisc_generic_ops;
384 if (neigh->nud_state&NUD_VALID)
385 neigh->output = neigh->ops->connected_output;
386 else
387 neigh->output = neigh->ops->output;
388 }
389 in6_dev_put(in6_dev);
390 return 0;
391}
392
393static int pndisc_constructor(struct pneigh_entry *n)
394{
395 struct in6_addr *addr = (struct in6_addr*)&n->key;
396 struct in6_addr maddr;
397 struct net_device *dev = n->dev;
398
399 if (dev == NULL || __in6_dev_get(dev) == NULL)
400 return -EINVAL;
401 addrconf_addr_solict_mult(addr, &maddr);
402 ipv6_dev_mc_inc(dev, &maddr);
403 return 0;
404}
405
406static void pndisc_destructor(struct pneigh_entry *n)
407{
408 struct in6_addr *addr = (struct in6_addr*)&n->key;
409 struct in6_addr maddr;
410 struct net_device *dev = n->dev;
411
412 if (dev == NULL || __in6_dev_get(dev) == NULL)
413 return;
414 addrconf_addr_solict_mult(addr, &maddr);
415 ipv6_dev_mc_dec(dev, &maddr);
416}
417
418struct sk_buff *ndisc_build_skb(struct net_device *dev,
419 const struct in6_addr *daddr,
420 const struct in6_addr *saddr,
421 struct icmp6hdr *icmp6h,
422 const struct in6_addr *target,
423 int llinfo)
424{
425 struct net *net = dev_net(dev);
426 struct sock *sk = net->ipv6.ndisc_sk;
427 struct sk_buff *skb;
428 struct icmp6hdr *hdr;
429 int hlen = LL_RESERVED_SPACE(dev);
430 int tlen = dev->needed_tailroom;
431 int len;
432 int err;
433 u8 *opt;
434
435 if (!dev->addr_len)
436 llinfo = 0;
437
438 len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
439 if (llinfo)
440 len += ndisc_opt_addr_space(dev);
441
442 skb = sock_alloc_send_skb(sk,
443 (MAX_HEADER + sizeof(struct ipv6hdr) +
444 len + hlen + tlen),
445 1, &err);
446 if (!skb) {
447 ND_PRINTK(0, err, "ND: %s failed to allocate an skb, err=%d\n",
448 __func__, err);
449 return NULL;
450 }
451
452 skb_reserve(skb, hlen);
453 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
454
455 skb->transport_header = skb->tail;
456 skb_put(skb, len);
457
458 hdr = (struct icmp6hdr *)skb_transport_header(skb);
459 memcpy(hdr, icmp6h, sizeof(*hdr));
460
461 opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
462 if (target) {
463 *(struct in6_addr *)opt = *target;
464 opt += sizeof(*target);
465 }
466
467 if (llinfo)
468 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
469 dev->addr_len, dev->type);
470
471 hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
472 IPPROTO_ICMPV6,
473 csum_partial(hdr,
474 len, 0));
475
476 return skb;
477}
478
479EXPORT_SYMBOL(ndisc_build_skb);
480
481void ndisc_send_skb(struct sk_buff *skb,
482 struct net_device *dev,
483 struct neighbour *neigh,
484 const struct in6_addr *daddr,
485 const struct in6_addr *saddr,
486 struct icmp6hdr *icmp6h)
487{
488 struct flowi6 fl6;
489 struct dst_entry *dst;
490 struct net *net = dev_net(dev);
491 struct sock *sk = net->ipv6.ndisc_sk;
492 struct inet6_dev *idev;
493 int err;
494 u8 type;
495
496 type = icmp6h->icmp6_type;
497
498 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, dev->ifindex);
499 dst = icmp6_dst_alloc(dev, neigh, &fl6);
500 if (IS_ERR(dst)) {
501 kfree_skb(skb);
502 return;
503 }
504
505 skb_dst_set(skb, dst);
506
507 rcu_read_lock();
508 idev = __in6_dev_get(dst->dev);
509 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
510
511 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
512 dst_output);
513 if (!err) {
514 ICMP6MSGOUT_INC_STATS(net, idev, type);
515 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
516 }
517
518 rcu_read_unlock();
519}
520
521EXPORT_SYMBOL(ndisc_send_skb);
522
523/*
524 * Send a Neighbour Discover packet
525 */
526static void __ndisc_send(struct net_device *dev,
527 struct neighbour *neigh,
528 const struct in6_addr *daddr,
529 const struct in6_addr *saddr,
530 struct icmp6hdr *icmp6h, const struct in6_addr *target,
531 int llinfo)
532{
533 struct sk_buff *skb;
534
535 skb = ndisc_build_skb(dev, daddr, saddr, icmp6h, target, llinfo);
536 if (!skb)
537 return;
538
539 ndisc_send_skb(skb, dev, neigh, daddr, saddr, icmp6h);
540}
541
542static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
543 const struct in6_addr *daddr,
544 const struct in6_addr *solicited_addr,
545 int router, int solicited, int override, int inc_opt)
546{
547 struct in6_addr tmpaddr;
548 struct inet6_ifaddr *ifp;
549 const struct in6_addr *src_addr;
550 struct icmp6hdr icmp6h = {
551 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
552 };
553
554 /* for anycast or proxy, solicited_addr != src_addr */
555 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
556 if (ifp) {
557 src_addr = solicited_addr;
558 if (ifp->flags & IFA_F_OPTIMISTIC)
559 override = 0;
560 inc_opt |= ifp->idev->cnf.force_tllao;
561 in6_ifa_put(ifp);
562 } else {
563 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
564 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
565 &tmpaddr))
566 return;
567 src_addr = &tmpaddr;
568 }
569
570 icmp6h.icmp6_router = router;
571 icmp6h.icmp6_solicited = solicited;
572 icmp6h.icmp6_override = override;
573
574 __ndisc_send(dev, neigh, daddr, src_addr,
575 &icmp6h, solicited_addr,
576 inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
577}
578
579static void ndisc_send_unsol_na(struct net_device *dev)
580{
581 struct inet6_dev *idev;
582 struct inet6_ifaddr *ifa;
583 struct in6_addr mcaddr;
584
585 idev = in6_dev_get(dev);
586 if (!idev)
587 return;
588
589 read_lock_bh(&idev->lock);
590 list_for_each_entry(ifa, &idev->addr_list, if_list) {
591 addrconf_addr_solict_mult(&ifa->addr, &mcaddr);
592 ndisc_send_na(dev, NULL, &mcaddr, &ifa->addr,
593 /*router=*/ !!idev->cnf.forwarding,
594 /*solicited=*/ false, /*override=*/ true,
595 /*inc_opt=*/ true);
596 }
597 read_unlock_bh(&idev->lock);
598
599 in6_dev_put(idev);
600}
601
602void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
603 const struct in6_addr *solicit,
604 const struct in6_addr *daddr, const struct in6_addr *saddr)
605{
606 struct in6_addr addr_buf;
607 struct icmp6hdr icmp6h = {
608 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
609 };
610
611 if (saddr == NULL) {
612 if (ipv6_get_lladdr(dev, &addr_buf,
613 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
614 return;
615 saddr = &addr_buf;
616 }
617
618 __ndisc_send(dev, neigh, daddr, saddr,
619 &icmp6h, solicit,
620 !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
621}
622
623void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
624 const struct in6_addr *daddr)
625{
626 struct icmp6hdr icmp6h = {
627 .icmp6_type = NDISC_ROUTER_SOLICITATION,
628 };
629 int send_sllao = dev->addr_len;
630
631#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
632 /*
633 * According to section 2.2 of RFC 4429, we must not
634 * send router solicitations with a sllao from
635 * optimistic addresses, but we may send the solicitation
636 * if we don't include the sllao. So here we check
637 * if our address is optimistic, and if so, we
638 * suppress the inclusion of the sllao.
639 */
640 if (send_sllao) {
641 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
642 dev, 1);
643 if (ifp) {
644 if (ifp->flags & IFA_F_OPTIMISTIC) {
645 send_sllao = 0;
646 }
647 in6_ifa_put(ifp);
648 } else {
649 send_sllao = 0;
650 }
651 }
652#endif
653 __ndisc_send(dev, NULL, daddr, saddr,
654 &icmp6h, NULL,
655 send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
656}
657
658
659static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
660{
661 /*
662 * "The sender MUST return an ICMP
663 * destination unreachable"
664 */
665 dst_link_failure(skb);
666 kfree_skb(skb);
667}
668
669/* Called with locked neigh: either read or both */
670
671static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
672{
673 struct in6_addr *saddr = NULL;
674 struct in6_addr mcaddr;
675 struct net_device *dev = neigh->dev;
676 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
677 int probes = atomic_read(&neigh->probes);
678
679 if (skb && ipv6_chk_addr(dev_net(dev), &ipv6_hdr(skb)->saddr, dev, 1))
680 saddr = &ipv6_hdr(skb)->saddr;
681
682 if ((probes -= neigh->parms->ucast_probes) < 0) {
683 if (!(neigh->nud_state & NUD_VALID)) {
684 ND_PRINTK(1, dbg,
685 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
686 __func__, target);
687 }
688 ndisc_send_ns(dev, neigh, target, target, saddr);
689 } else if ((probes -= neigh->parms->app_probes) < 0) {
690#ifdef CONFIG_ARPD
691 neigh_app_ns(neigh);
692#endif
693 } else {
694 addrconf_addr_solict_mult(target, &mcaddr);
695 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
696 }
697}
698
699static int pndisc_is_router(const void *pkey,
700 struct net_device *dev)
701{
702 struct pneigh_entry *n;
703 int ret = -1;
704
705 read_lock_bh(&nd_tbl.lock);
706 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
707 if (n)
708 ret = !!(n->flags & NTF_ROUTER);
709 read_unlock_bh(&nd_tbl.lock);
710
711 return ret;
712}
713
714static void ndisc_recv_ns(struct sk_buff *skb)
715{
716 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
717 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
718 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
719 u8 *lladdr = NULL;
720 u32 ndoptlen = skb->tail - (skb->transport_header +
721 offsetof(struct nd_msg, opt));
722 struct ndisc_options ndopts;
723 struct net_device *dev = skb->dev;
724 struct inet6_ifaddr *ifp;
725 struct inet6_dev *idev = NULL;
726 struct neighbour *neigh;
727 int dad = ipv6_addr_any(saddr);
728 bool inc;
729 int is_router = -1;
730
731 if (ipv6_addr_is_multicast(&msg->target)) {
732 ND_PRINTK(2, warn, "NS: multicast target address\n");
733 return;
734 }
735
736 /*
737 * RFC2461 7.1.1:
738 * DAD has to be destined for solicited node multicast address.
739 */
740 if (dad &&
741 !(daddr->s6_addr32[0] == htonl(0xff020000) &&
742 daddr->s6_addr32[1] == htonl(0x00000000) &&
743 daddr->s6_addr32[2] == htonl(0x00000001) &&
744 daddr->s6_addr [12] == 0xff )) {
745 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
746 return;
747 }
748
749 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
750 ND_PRINTK(2, warn, "NS: invalid ND options\n");
751 return;
752 }
753
754 if (ndopts.nd_opts_src_lladdr) {
755 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
756 if (!lladdr) {
757 ND_PRINTK(2, warn,
758 "NS: invalid link-layer address length\n");
759 return;
760 }
761
762 /* RFC2461 7.1.1:
763 * If the IP source address is the unspecified address,
764 * there MUST NOT be source link-layer address option
765 * in the message.
766 */
767 if (dad) {
768 ND_PRINTK(2, warn,
769 "NS: bad DAD packet (link-layer address option)\n");
770 return;
771 }
772 }
773
774 inc = ipv6_addr_is_multicast(daddr);
775
776 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
777 if (ifp) {
778
779 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
780 if (dad) {
781 /*
782 * We are colliding with another node
783 * who is doing DAD
784 * so fail our DAD process
785 */
786 addrconf_dad_failure(ifp);
787 return;
788 } else {
789 /*
790 * This is not a dad solicitation.
791 * If we are an optimistic node,
792 * we should respond.
793 * Otherwise, we should ignore it.
794 */
795 if (!(ifp->flags & IFA_F_OPTIMISTIC))
796 goto out;
797 }
798 }
799
800 idev = ifp->idev;
801 } else {
802 struct net *net = dev_net(dev);
803
804 idev = in6_dev_get(dev);
805 if (!idev) {
806 /* XXX: count this drop? */
807 return;
808 }
809
810 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
811 (idev->cnf.forwarding &&
812 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
813 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
814 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
815 skb->pkt_type != PACKET_HOST &&
816 inc != 0 &&
817 idev->nd_parms->proxy_delay != 0) {
818 /*
819 * for anycast or proxy,
820 * sender should delay its response
821 * by a random time between 0 and
822 * MAX_ANYCAST_DELAY_TIME seconds.
823 * (RFC2461) -- yoshfuji
824 */
825 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
826 if (n)
827 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
828 goto out;
829 }
830 } else
831 goto out;
832 }
833
834 if (is_router < 0)
835 is_router = !!idev->cnf.forwarding;
836
837 if (dad) {
838 ndisc_send_na(dev, NULL, &in6addr_linklocal_allnodes, &msg->target,
839 is_router, 0, (ifp != NULL), 1);
840 goto out;
841 }
842
843 if (inc)
844 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
845 else
846 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
847
848 /*
849 * update / create cache entry
850 * for the source address
851 */
852 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
853 !inc || lladdr || !dev->addr_len);
854 if (neigh)
855 neigh_update(neigh, lladdr, NUD_STALE,
856 NEIGH_UPDATE_F_WEAK_OVERRIDE|
857 NEIGH_UPDATE_F_OVERRIDE);
858 if (neigh || !dev->header_ops) {
859 ndisc_send_na(dev, neigh, saddr, &msg->target,
860 is_router,
861 1, (ifp != NULL && inc), inc);
862 if (neigh)
863 neigh_release(neigh);
864 }
865
866out:
867 if (ifp)
868 in6_ifa_put(ifp);
869 else
870 in6_dev_put(idev);
871}
872
873static void ndisc_recv_na(struct sk_buff *skb)
874{
875 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
876 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
877 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
878 u8 *lladdr = NULL;
879 u32 ndoptlen = skb->tail - (skb->transport_header +
880 offsetof(struct nd_msg, opt));
881 struct ndisc_options ndopts;
882 struct net_device *dev = skb->dev;
883 struct inet6_ifaddr *ifp;
884 struct neighbour *neigh;
885
886 if (skb->len < sizeof(struct nd_msg)) {
887 ND_PRINTK(2, warn, "NA: packet too short\n");
888 return;
889 }
890
891 if (ipv6_addr_is_multicast(&msg->target)) {
892 ND_PRINTK(2, warn, "NA: target address is multicast\n");
893 return;
894 }
895
896 if (ipv6_addr_is_multicast(daddr) &&
897 msg->icmph.icmp6_solicited) {
898 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
899 return;
900 }
901
902 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
903 ND_PRINTK(2, warn, "NS: invalid ND option\n");
904 return;
905 }
906 if (ndopts.nd_opts_tgt_lladdr) {
907 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
908 if (!lladdr) {
909 ND_PRINTK(2, warn,
910 "NA: invalid link-layer address length\n");
911 return;
912 }
913 }
914 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
915 if (ifp) {
916 if (skb->pkt_type != PACKET_LOOPBACK
917 && (ifp->flags & IFA_F_TENTATIVE)) {
918 addrconf_dad_failure(ifp);
919 return;
920 }
921 /* What should we make now? The advertisement
922 is invalid, but ndisc specs say nothing
923 about it. It could be misconfiguration, or
924 an smart proxy agent tries to help us :-)
925
926 We should not print the error if NA has been
927 received from loopback - it is just our own
928 unsolicited advertisement.
929 */
930 if (skb->pkt_type != PACKET_LOOPBACK)
931 ND_PRINTK(1, warn,
932 "NA: someone advertises our address %pI6 on %s!\n",
933 &ifp->addr, ifp->idev->dev->name);
934 in6_ifa_put(ifp);
935 return;
936 }
937 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
938
939 if (neigh) {
940 u8 old_flags = neigh->flags;
941 struct net *net = dev_net(dev);
942
943 if (neigh->nud_state & NUD_FAILED)
944 goto out;
945
946 /*
947 * Don't update the neighbor cache entry on a proxy NA from
948 * ourselves because either the proxied node is off link or it
949 * has already sent a NA to us.
950 */
951 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
952 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
953 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
954 /* XXX: idev->cnf.prixy_ndp */
955 goto out;
956 }
957
958 neigh_update(neigh, lladdr,
959 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
960 NEIGH_UPDATE_F_WEAK_OVERRIDE|
961 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
962 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
963 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
964
965 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
966 /*
967 * Change: router to host
968 */
969 struct rt6_info *rt;
970 rt = rt6_get_dflt_router(saddr, dev);
971 if (rt)
972 ip6_del_rt(rt);
973 }
974
975out:
976 neigh_release(neigh);
977 }
978}
979
980static void ndisc_recv_rs(struct sk_buff *skb)
981{
982 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
983 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
984 struct neighbour *neigh;
985 struct inet6_dev *idev;
986 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
987 struct ndisc_options ndopts;
988 u8 *lladdr = NULL;
989
990 if (skb->len < sizeof(*rs_msg))
991 return;
992
993 idev = __in6_dev_get(skb->dev);
994 if (!idev) {
995 ND_PRINTK(1, err, "RS: can't find in6 device\n");
996 return;
997 }
998
999 /* Don't accept RS if we're not in router mode */
1000 if (!idev->cnf.forwarding)
1001 goto out;
1002
1003 /*
1004 * Don't update NCE if src = ::;
1005 * this implies that the source node has no ip address assigned yet.
1006 */
1007 if (ipv6_addr_any(saddr))
1008 goto out;
1009
1010 /* Parse ND options */
1011 if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1012 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1013 goto out;
1014 }
1015
1016 if (ndopts.nd_opts_src_lladdr) {
1017 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1018 skb->dev);
1019 if (!lladdr)
1020 goto out;
1021 }
1022
1023 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1024 if (neigh) {
1025 neigh_update(neigh, lladdr, NUD_STALE,
1026 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1027 NEIGH_UPDATE_F_OVERRIDE|
1028 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1029 neigh_release(neigh);
1030 }
1031out:
1032 return;
1033}
1034
1035static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1036{
1037 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1038 struct sk_buff *skb;
1039 struct nlmsghdr *nlh;
1040 struct nduseroptmsg *ndmsg;
1041 struct net *net = dev_net(ra->dev);
1042 int err;
1043 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1044 + (opt->nd_opt_len << 3));
1045 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1046
1047 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1048 if (skb == NULL) {
1049 err = -ENOBUFS;
1050 goto errout;
1051 }
1052
1053 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1054 if (nlh == NULL) {
1055 goto nla_put_failure;
1056 }
1057
1058 ndmsg = nlmsg_data(nlh);
1059 ndmsg->nduseropt_family = AF_INET6;
1060 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1061 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1062 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1063 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1064
1065 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1066
1067 if (nla_put(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
1068 &ipv6_hdr(ra)->saddr))
1069 goto nla_put_failure;
1070 nlmsg_end(skb, nlh);
1071
1072 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1073 return;
1074
1075nla_put_failure:
1076 nlmsg_free(skb);
1077 err = -EMSGSIZE;
1078errout:
1079 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1080}
1081
1082static inline int accept_ra(struct inet6_dev *in6_dev)
1083{
1084 /*
1085 * If forwarding is enabled, RA are not accepted unless the special
1086 * hybrid mode (accept_ra=2) is enabled.
1087 */
1088 if (in6_dev->cnf.forwarding && in6_dev->cnf.accept_ra < 2)
1089 return 0;
1090
1091 return in6_dev->cnf.accept_ra;
1092}
1093
1094static void ndisc_router_discovery(struct sk_buff *skb)
1095{
1096 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1097 struct neighbour *neigh = NULL;
1098 struct inet6_dev *in6_dev;
1099 struct rt6_info *rt = NULL;
1100 int lifetime;
1101 struct ndisc_options ndopts;
1102 int optlen;
1103 unsigned int pref = 0;
1104
1105 __u8 * opt = (__u8 *)(ra_msg + 1);
1106
1107 optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1108
1109 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1110 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1111 return;
1112 }
1113 if (optlen < 0) {
1114 ND_PRINTK(2, warn, "RA: packet too short\n");
1115 return;
1116 }
1117
1118#ifdef CONFIG_IPV6_NDISC_NODETYPE
1119 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1120 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1121 return;
1122 }
1123#endif
1124
1125 /*
1126 * set the RA_RECV flag in the interface
1127 */
1128
1129 in6_dev = __in6_dev_get(skb->dev);
1130 if (in6_dev == NULL) {
1131 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1132 skb->dev->name);
1133 return;
1134 }
1135
1136 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1137 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1138 return;
1139 }
1140
1141 if (!accept_ra(in6_dev))
1142 goto skip_linkparms;
1143
1144#ifdef CONFIG_IPV6_NDISC_NODETYPE
1145 /* skip link-specific parameters from interior routers */
1146 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1147 goto skip_linkparms;
1148#endif
1149
1150 if (in6_dev->if_flags & IF_RS_SENT) {
1151 /*
1152 * flag that an RA was received after an RS was sent
1153 * out on this interface.
1154 */
1155 in6_dev->if_flags |= IF_RA_RCVD;
1156 }
1157
1158 /*
1159 * Remember the managed/otherconf flags from most recently
1160 * received RA message (RFC 2462) -- yoshfuji
1161 */
1162 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1163 IF_RA_OTHERCONF)) |
1164 (ra_msg->icmph.icmp6_addrconf_managed ?
1165 IF_RA_MANAGED : 0) |
1166 (ra_msg->icmph.icmp6_addrconf_other ?
1167 IF_RA_OTHERCONF : 0);
1168
1169 if (!in6_dev->cnf.accept_ra_defrtr)
1170 goto skip_defrtr;
1171
1172 if (ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr, NULL, 0))
1173 goto skip_defrtr;
1174
1175 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1176
1177#ifdef CONFIG_IPV6_ROUTER_PREF
1178 pref = ra_msg->icmph.icmp6_router_pref;
1179 /* 10b is handled as if it were 00b (medium) */
1180 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1181 !in6_dev->cnf.accept_ra_rtr_pref)
1182 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1183#endif
1184
1185 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1186
1187 if (rt) {
1188 neigh = dst_neigh_lookup(&rt->dst, &ipv6_hdr(skb)->saddr);
1189 if (!neigh) {
1190 ND_PRINTK(0, err,
1191 "RA: %s got default router without neighbour\n",
1192 __func__);
1193 dst_release(&rt->dst);
1194 return;
1195 }
1196 }
1197 if (rt && lifetime == 0) {
1198 ip6_del_rt(rt);
1199 rt = NULL;
1200 }
1201
1202 if (rt == NULL && lifetime) {
1203 ND_PRINTK(3, dbg, "RA: adding default router\n");
1204
1205 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1206 if (rt == NULL) {
1207 ND_PRINTK(0, err,
1208 "RA: %s failed to add default route\n",
1209 __func__);
1210 return;
1211 }
1212
1213 neigh = dst_neigh_lookup(&rt->dst, &ipv6_hdr(skb)->saddr);
1214 if (neigh == NULL) {
1215 ND_PRINTK(0, err,
1216 "RA: %s got default router without neighbour\n",
1217 __func__);
1218 dst_release(&rt->dst);
1219 return;
1220 }
1221 neigh->flags |= NTF_ROUTER;
1222 } else if (rt) {
1223 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1224 }
1225
1226 if (rt)
1227 rt6_set_expires(rt, jiffies + (HZ * lifetime));
1228 if (ra_msg->icmph.icmp6_hop_limit) {
1229 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1230 if (rt)
1231 dst_metric_set(&rt->dst, RTAX_HOPLIMIT,
1232 ra_msg->icmph.icmp6_hop_limit);
1233 }
1234
1235skip_defrtr:
1236
1237 /*
1238 * Update Reachable Time and Retrans Timer
1239 */
1240
1241 if (in6_dev->nd_parms) {
1242 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1243
1244 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1245 rtime = (rtime*HZ)/1000;
1246 if (rtime < HZ/10)
1247 rtime = HZ/10;
1248 in6_dev->nd_parms->retrans_time = rtime;
1249 in6_dev->tstamp = jiffies;
1250 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1251 }
1252
1253 rtime = ntohl(ra_msg->reachable_time);
1254 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1255 rtime = (rtime*HZ)/1000;
1256
1257 if (rtime < HZ/10)
1258 rtime = HZ/10;
1259
1260 if (rtime != in6_dev->nd_parms->base_reachable_time) {
1261 in6_dev->nd_parms->base_reachable_time = rtime;
1262 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1263 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1264 in6_dev->tstamp = jiffies;
1265 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1266 }
1267 }
1268 }
1269
1270skip_linkparms:
1271
1272 /*
1273 * Process options.
1274 */
1275
1276 if (!neigh)
1277 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1278 skb->dev, 1);
1279 if (neigh) {
1280 u8 *lladdr = NULL;
1281 if (ndopts.nd_opts_src_lladdr) {
1282 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1283 skb->dev);
1284 if (!lladdr) {
1285 ND_PRINTK(2, warn,
1286 "RA: invalid link-layer address length\n");
1287 goto out;
1288 }
1289 }
1290 neigh_update(neigh, lladdr, NUD_STALE,
1291 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1292 NEIGH_UPDATE_F_OVERRIDE|
1293 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1294 NEIGH_UPDATE_F_ISROUTER);
1295 }
1296
1297 if (!accept_ra(in6_dev))
1298 goto out;
1299
1300#ifdef CONFIG_IPV6_ROUTE_INFO
1301 if (ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr, NULL, 0))
1302 goto skip_routeinfo;
1303
1304 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1305 struct nd_opt_hdr *p;
1306 for (p = ndopts.nd_opts_ri;
1307 p;
1308 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1309 struct route_info *ri = (struct route_info *)p;
1310#ifdef CONFIG_IPV6_NDISC_NODETYPE
1311 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1312 ri->prefix_len == 0)
1313 continue;
1314#endif
1315 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1316 continue;
1317 rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1318 &ipv6_hdr(skb)->saddr);
1319 }
1320 }
1321
1322skip_routeinfo:
1323#endif
1324
1325#ifdef CONFIG_IPV6_NDISC_NODETYPE
1326 /* skip link-specific ndopts from interior routers */
1327 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1328 goto out;
1329#endif
1330
1331 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1332 struct nd_opt_hdr *p;
1333 for (p = ndopts.nd_opts_pi;
1334 p;
1335 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1336 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1337 (p->nd_opt_len) << 3,
1338 ndopts.nd_opts_src_lladdr != NULL);
1339 }
1340 }
1341
1342 if (ndopts.nd_opts_mtu) {
1343 __be32 n;
1344 u32 mtu;
1345
1346 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1347 mtu = ntohl(n);
1348
1349 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1350 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1351 } else if (in6_dev->cnf.mtu6 != mtu) {
1352 in6_dev->cnf.mtu6 = mtu;
1353
1354 if (rt)
1355 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
1356
1357 rt6_mtu_change(skb->dev, mtu);
1358 }
1359 }
1360
1361 if (ndopts.nd_useropts) {
1362 struct nd_opt_hdr *p;
1363 for (p = ndopts.nd_useropts;
1364 p;
1365 p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1366 ndisc_ra_useropt(skb, p);
1367 }
1368 }
1369
1370 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1371 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1372 }
1373out:
1374 if (rt)
1375 dst_release(&rt->dst);
1376 if (neigh)
1377 neigh_release(neigh);
1378}
1379
1380static void ndisc_redirect_rcv(struct sk_buff *skb)
1381{
1382 struct inet6_dev *in6_dev;
1383 struct icmp6hdr *icmph;
1384 const struct in6_addr *dest;
1385 const struct in6_addr *target; /* new first hop to destination */
1386 struct neighbour *neigh;
1387 int on_link = 0;
1388 struct ndisc_options ndopts;
1389 int optlen;
1390 u8 *lladdr = NULL;
1391
1392#ifdef CONFIG_IPV6_NDISC_NODETYPE
1393 switch (skb->ndisc_nodetype) {
1394 case NDISC_NODETYPE_HOST:
1395 case NDISC_NODETYPE_NODEFAULT:
1396 ND_PRINTK(2, warn,
1397 "Redirect: from host or unauthorized router\n");
1398 return;
1399 }
1400#endif
1401
1402 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1403 ND_PRINTK(2, warn,
1404 "Redirect: source address is not link-local\n");
1405 return;
1406 }
1407
1408 optlen = skb->tail - skb->transport_header;
1409 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1410
1411 if (optlen < 0) {
1412 ND_PRINTK(2, warn, "Redirect: packet too short\n");
1413 return;
1414 }
1415
1416 icmph = icmp6_hdr(skb);
1417 target = (const struct in6_addr *) (icmph + 1);
1418 dest = target + 1;
1419
1420 if (ipv6_addr_is_multicast(dest)) {
1421 ND_PRINTK(2, warn,
1422 "Redirect: destination address is multicast\n");
1423 return;
1424 }
1425
1426 if (ipv6_addr_equal(dest, target)) {
1427 on_link = 1;
1428 } else if (ipv6_addr_type(target) !=
1429 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1430 ND_PRINTK(2, warn,
1431 "Redirect: target address is not link-local unicast\n");
1432 return;
1433 }
1434
1435 in6_dev = __in6_dev_get(skb->dev);
1436 if (!in6_dev)
1437 return;
1438 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1439 return;
1440
1441 /* RFC2461 8.1:
1442 * The IP source address of the Redirect MUST be the same as the current
1443 * first-hop router for the specified ICMP Destination Address.
1444 */
1445
1446 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1447 ND_PRINTK(2, warn, "Redirect: invalid ND options\n");
1448 return;
1449 }
1450 if (ndopts.nd_opts_tgt_lladdr) {
1451 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1452 skb->dev);
1453 if (!lladdr) {
1454 ND_PRINTK(2, warn,
1455 "Redirect: invalid link-layer address length\n");
1456 return;
1457 }
1458 }
1459
1460 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1461 if (neigh) {
1462 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1463 &ipv6_hdr(skb)->saddr, neigh, lladdr,
1464 on_link);
1465 neigh_release(neigh);
1466 }
1467}
1468
1469void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1470{
1471 struct net_device *dev = skb->dev;
1472 struct net *net = dev_net(dev);
1473 struct sock *sk = net->ipv6.ndisc_sk;
1474 int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1475 struct sk_buff *buff;
1476 struct icmp6hdr *icmph;
1477 struct in6_addr saddr_buf;
1478 struct in6_addr *addrp;
1479 struct rt6_info *rt;
1480 struct dst_entry *dst;
1481 struct inet6_dev *idev;
1482 struct flowi6 fl6;
1483 u8 *opt;
1484 int hlen, tlen;
1485 int rd_len;
1486 int err;
1487 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1488
1489 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1490 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1491 dev->name);
1492 return;
1493 }
1494
1495 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1496 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1497 ND_PRINTK(2, warn,
1498 "Redirect: target address is not link-local unicast\n");
1499 return;
1500 }
1501
1502 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1503 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1504
1505 dst = ip6_route_output(net, NULL, &fl6);
1506 if (dst->error) {
1507 dst_release(dst);
1508 return;
1509 }
1510 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1511 if (IS_ERR(dst))
1512 return;
1513
1514 rt = (struct rt6_info *) dst;
1515
1516 if (rt->rt6i_flags & RTF_GATEWAY) {
1517 ND_PRINTK(2, warn,
1518 "Redirect: destination is not a neighbour\n");
1519 goto release;
1520 }
1521 if (!rt->rt6i_peer)
1522 rt6_bind_peer(rt, 1);
1523 if (!inet_peer_xrlim_allow(rt->rt6i_peer, 1*HZ))
1524 goto release;
1525
1526 if (dev->addr_len) {
1527 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1528 if (!neigh) {
1529 ND_PRINTK(2, warn,
1530 "Redirect: no neigh for target address\n");
1531 goto release;
1532 }
1533
1534 read_lock_bh(&neigh->lock);
1535 if (neigh->nud_state & NUD_VALID) {
1536 memcpy(ha_buf, neigh->ha, dev->addr_len);
1537 read_unlock_bh(&neigh->lock);
1538 ha = ha_buf;
1539 len += ndisc_opt_addr_space(dev);
1540 } else
1541 read_unlock_bh(&neigh->lock);
1542
1543 neigh_release(neigh);
1544 }
1545
1546 rd_len = min_t(unsigned int,
1547 IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1548 rd_len &= ~0x7;
1549 len += rd_len;
1550
1551 hlen = LL_RESERVED_SPACE(dev);
1552 tlen = dev->needed_tailroom;
1553 buff = sock_alloc_send_skb(sk,
1554 (MAX_HEADER + sizeof(struct ipv6hdr) +
1555 len + hlen + tlen),
1556 1, &err);
1557 if (buff == NULL) {
1558 ND_PRINTK(0, err,
1559 "Redirect: %s failed to allocate an skb, err=%d\n",
1560 __func__, err);
1561 goto release;
1562 }
1563
1564 skb_reserve(buff, hlen);
1565 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1566 IPPROTO_ICMPV6, len);
1567
1568 skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1569 skb_put(buff, len);
1570 icmph = icmp6_hdr(buff);
1571
1572 memset(icmph, 0, sizeof(struct icmp6hdr));
1573 icmph->icmp6_type = NDISC_REDIRECT;
1574
1575 /*
1576 * copy target and destination addresses
1577 */
1578
1579 addrp = (struct in6_addr *)(icmph + 1);
1580 *addrp = *target;
1581 addrp++;
1582 *addrp = ipv6_hdr(skb)->daddr;
1583
1584 opt = (u8*) (addrp + 1);
1585
1586 /*
1587 * include target_address option
1588 */
1589
1590 if (ha)
1591 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1592 dev->addr_len, dev->type);
1593
1594 /*
1595 * build redirect option and copy skb over to the new packet.
1596 */
1597
1598 memset(opt, 0, 8);
1599 *(opt++) = ND_OPT_REDIRECT_HDR;
1600 *(opt++) = (rd_len >> 3);
1601 opt += 6;
1602
1603 memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1604
1605 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1606 len, IPPROTO_ICMPV6,
1607 csum_partial(icmph, len, 0));
1608
1609 skb_dst_set(buff, dst);
1610 rcu_read_lock();
1611 idev = __in6_dev_get(dst->dev);
1612 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1613 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, buff, NULL, dst->dev,
1614 dst_output);
1615 if (!err) {
1616 ICMP6MSGOUT_INC_STATS(net, idev, NDISC_REDIRECT);
1617 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1618 }
1619
1620 rcu_read_unlock();
1621 return;
1622
1623release:
1624 dst_release(dst);
1625}
1626
1627static void pndisc_redo(struct sk_buff *skb)
1628{
1629 ndisc_recv_ns(skb);
1630 kfree_skb(skb);
1631}
1632
1633int ndisc_rcv(struct sk_buff *skb)
1634{
1635 struct nd_msg *msg;
1636
1637 if (!pskb_may_pull(skb, skb->len))
1638 return 0;
1639
1640 msg = (struct nd_msg *)skb_transport_header(skb);
1641
1642 __skb_push(skb, skb->data - skb_transport_header(skb));
1643
1644 if (ipv6_hdr(skb)->hop_limit != 255) {
1645 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1646 ipv6_hdr(skb)->hop_limit);
1647 return 0;
1648 }
1649
1650 if (msg->icmph.icmp6_code != 0) {
1651 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1652 msg->icmph.icmp6_code);
1653 return 0;
1654 }
1655
1656 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1657
1658 switch (msg->icmph.icmp6_type) {
1659 case NDISC_NEIGHBOUR_SOLICITATION:
1660 ndisc_recv_ns(skb);
1661 break;
1662
1663 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1664 ndisc_recv_na(skb);
1665 break;
1666
1667 case NDISC_ROUTER_SOLICITATION:
1668 ndisc_recv_rs(skb);
1669 break;
1670
1671 case NDISC_ROUTER_ADVERTISEMENT:
1672 ndisc_router_discovery(skb);
1673 break;
1674
1675 case NDISC_REDIRECT:
1676 ndisc_redirect_rcv(skb);
1677 break;
1678 }
1679
1680 return 0;
1681}
1682
1683static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1684{
1685 struct net_device *dev = ptr;
1686 struct net *net = dev_net(dev);
1687
1688 switch (event) {
1689 case NETDEV_CHANGEADDR:
1690 neigh_changeaddr(&nd_tbl, dev);
1691 fib6_run_gc(~0UL, net);
1692 break;
1693 case NETDEV_DOWN:
1694 neigh_ifdown(&nd_tbl, dev);
1695 fib6_run_gc(~0UL, net);
1696 break;
1697 case NETDEV_NOTIFY_PEERS:
1698 ndisc_send_unsol_na(dev);
1699 break;
1700 default:
1701 break;
1702 }
1703
1704 return NOTIFY_DONE;
1705}
1706
1707static struct notifier_block ndisc_netdev_notifier = {
1708 .notifier_call = ndisc_netdev_event,
1709};
1710
1711#ifdef CONFIG_SYSCTL
1712static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1713 const char *func, const char *dev_name)
1714{
1715 static char warncomm[TASK_COMM_LEN];
1716 static int warned;
1717 if (strcmp(warncomm, current->comm) && warned < 5) {
1718 strcpy(warncomm, current->comm);
1719 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1720 warncomm, func,
1721 dev_name, ctl->procname,
1722 dev_name, ctl->procname);
1723 warned++;
1724 }
1725}
1726
1727int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1728{
1729 struct net_device *dev = ctl->extra1;
1730 struct inet6_dev *idev;
1731 int ret;
1732
1733 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1734 (strcmp(ctl->procname, "base_reachable_time") == 0))
1735 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1736
1737 if (strcmp(ctl->procname, "retrans_time") == 0)
1738 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1739
1740 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1741 ret = proc_dointvec_jiffies(ctl, write,
1742 buffer, lenp, ppos);
1743
1744 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1745 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1746 ret = proc_dointvec_ms_jiffies(ctl, write,
1747 buffer, lenp, ppos);
1748 else
1749 ret = -1;
1750
1751 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1752 if (ctl->data == &idev->nd_parms->base_reachable_time)
1753 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1754 idev->tstamp = jiffies;
1755 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1756 in6_dev_put(idev);
1757 }
1758 return ret;
1759}
1760
1761
1762#endif
1763
1764static int __net_init ndisc_net_init(struct net *net)
1765{
1766 struct ipv6_pinfo *np;
1767 struct sock *sk;
1768 int err;
1769
1770 err = inet_ctl_sock_create(&sk, PF_INET6,
1771 SOCK_RAW, IPPROTO_ICMPV6, net);
1772 if (err < 0) {
1773 ND_PRINTK(0, err,
1774 "NDISC: Failed to initialize the control socket (err %d)\n",
1775 err);
1776 return err;
1777 }
1778
1779 net->ipv6.ndisc_sk = sk;
1780
1781 np = inet6_sk(sk);
1782 np->hop_limit = 255;
1783 /* Do not loopback ndisc messages */
1784 np->mc_loop = 0;
1785
1786 return 0;
1787}
1788
1789static void __net_exit ndisc_net_exit(struct net *net)
1790{
1791 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1792}
1793
1794static struct pernet_operations ndisc_net_ops = {
1795 .init = ndisc_net_init,
1796 .exit = ndisc_net_exit,
1797};
1798
1799int __init ndisc_init(void)
1800{
1801 int err;
1802
1803 err = register_pernet_subsys(&ndisc_net_ops);
1804 if (err)
1805 return err;
1806 /*
1807 * Initialize the neighbour table
1808 */
1809 neigh_table_init(&nd_tbl);
1810
1811#ifdef CONFIG_SYSCTL
1812 err = neigh_sysctl_register(NULL, &nd_tbl.parms, "ipv6",
1813 &ndisc_ifinfo_sysctl_change);
1814 if (err)
1815 goto out_unregister_pernet;
1816#endif
1817 err = register_netdevice_notifier(&ndisc_netdev_notifier);
1818 if (err)
1819 goto out_unregister_sysctl;
1820out:
1821 return err;
1822
1823out_unregister_sysctl:
1824#ifdef CONFIG_SYSCTL
1825 neigh_sysctl_unregister(&nd_tbl.parms);
1826out_unregister_pernet:
1827#endif
1828 unregister_pernet_subsys(&ndisc_net_ops);
1829 goto out;
1830}
1831
1832void ndisc_cleanup(void)
1833{
1834 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1835#ifdef CONFIG_SYSCTL
1836 neigh_sysctl_unregister(&nd_tbl.parms);
1837#endif
1838 neigh_table_clear(&nd_tbl);
1839 unregister_pernet_subsys(&ndisc_net_ops);
1840}