ARM: 7447/1: rwlocks: remove unused branch labels from trylock routines
[linux-2.6/libata-dev.git] / net / ipv6 / ndisc.c
blob54f62d3b8dd6b9eb24fe791870b81fa8adb97280
1 /*
2 * Neighbour Discovery for IPv6
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Mike Shaver <shaver@ingenia.com>
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.
16 * Changes:
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
30 #define pr_fmt(fmt) "ICMPv6: " fmt
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
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>
54 #include <net/sock.h>
55 #include <net/snmp.h>
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>
64 #include <net/netlink.h>
65 #include <linux/rtnetlink.h>
67 #include <net/flow.h>
68 #include <net/ip6_checksum.h>
69 #include <net/inet_common.h>
70 #include <linux/proc_fs.h>
72 #include <linux/netfilter.h>
73 #include <linux/netfilter_ipv6.h>
75 /* Set to 3 to get tracing... */
76 #define ND_DEBUG 1
78 #define ND_PRINTK(val, level, fmt, ...) \
79 do { \
80 if (val <= ND_DEBUG) \
81 net_##level##_ratelimited(fmt, ##__VA_ARGS__); \
82 } while (0)
84 static u32 ndisc_hash(const void *pkey,
85 const struct net_device *dev,
86 __u32 *hash_rnd);
87 static int ndisc_constructor(struct neighbour *neigh);
88 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
89 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
90 static int pndisc_constructor(struct pneigh_entry *n);
91 static void pndisc_destructor(struct pneigh_entry *n);
92 static void pndisc_redo(struct sk_buff *skb);
94 static 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,
102 static 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,
111 static const struct neigh_ops ndisc_direct_ops = {
112 .family = AF_INET6,
113 .output = neigh_direct_output,
114 .connected_output = neigh_direct_output,
117 struct 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,
140 .gc_interval = 30 * HZ,
141 .gc_thresh1 = 128,
142 .gc_thresh2 = 512,
143 .gc_thresh3 = 1024,
146 /* ND options */
147 struct 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;
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]
164 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
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.
172 static int ndisc_addr_option_pad(unsigned short type)
174 switch (type) {
175 case ARPHRD_INFINIBAND: return 2;
176 default: return 0;
180 static inline int ndisc_opt_addr_space(struct net_device *dev)
182 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
185 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
186 unsigned short addr_type)
188 int space = NDISC_OPT_SPACE(data_len);
189 int pad = ndisc_addr_option_pad(addr_type);
191 opt[0] = type;
192 opt[1] = space>>3;
194 memset(opt + 2, 0, pad);
195 opt += pad;
196 space -= pad;
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;
206 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
207 struct nd_opt_hdr *end)
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;
219 static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
221 return opt->nd_opt_type == ND_OPT_RDNSS ||
222 opt->nd_opt_type == ND_OPT_DNSSL;
225 static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
226 struct nd_opt_hdr *end)
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;
236 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
237 struct ndisc_options *ndopts)
239 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
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;
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 {
283 * Unknown options must be silently ignored,
284 * to accommodate future extension to the
285 * protocol.
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);
294 opt_len -= l;
295 nd_opt = ((void *)nd_opt) + l;
297 return ndopts;
300 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
301 struct net_device *dev)
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;
311 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
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;
333 return -EINVAL;
336 EXPORT_SYMBOL(ndisc_mc_map);
338 static u32 ndisc_hash(const void *pkey,
339 const struct net_device *dev,
340 __u32 *hash_rnd)
342 return ndisc_hashfn(pkey, dev, hash_rnd);
345 static int ndisc_constructor(struct neighbour *neigh)
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);
353 in6_dev = in6_dev_get(dev);
354 if (in6_dev == NULL) {
355 return -EINVAL;
358 parms = in6_dev->nd_parms;
359 __neigh_parms_put(neigh->parms);
360 neigh->parms = neigh_parms_clone(parms);
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);
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;
389 in6_dev_put(in6_dev);
390 return 0;
393 static int pndisc_constructor(struct pneigh_entry *n)
395 struct in6_addr *addr = (struct in6_addr*)&n->key;
396 struct in6_addr maddr;
397 struct net_device *dev = n->dev;
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;
406 static void pndisc_destructor(struct pneigh_entry *n)
408 struct in6_addr *addr = (struct in6_addr*)&n->key;
409 struct in6_addr maddr;
410 struct net_device *dev = n->dev;
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);
418 struct 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)
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;
435 if (!dev->addr_len)
436 llinfo = 0;
438 len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
439 if (llinfo)
440 len += ndisc_opt_addr_space(dev);
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;
452 skb_reserve(skb, hlen);
453 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
455 skb->transport_header = skb->tail;
456 skb_put(skb, len);
458 hdr = (struct icmp6hdr *)skb_transport_header(skb);
459 memcpy(hdr, icmp6h, sizeof(*hdr));
461 opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
462 if (target) {
463 *(struct in6_addr *)opt = *target;
464 opt += sizeof(*target);
467 if (llinfo)
468 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
469 dev->addr_len, dev->type);
471 hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
472 IPPROTO_ICMPV6,
473 csum_partial(hdr,
474 len, 0));
476 return skb;
479 EXPORT_SYMBOL(ndisc_build_skb);
481 void 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)
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;
496 type = icmp6h->icmp6_type;
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;
505 skb_dst_set(skb, dst);
507 rcu_read_lock();
508 idev = __in6_dev_get(dst->dev);
509 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
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);
518 rcu_read_unlock();
521 EXPORT_SYMBOL(ndisc_send_skb);
524 * Send a Neighbour Discover packet
526 static 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)
533 struct sk_buff *skb;
535 skb = ndisc_build_skb(dev, daddr, saddr, icmp6h, target, llinfo);
536 if (!skb)
537 return;
539 ndisc_send_skb(skb, dev, neigh, daddr, saddr, icmp6h);
542 static 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)
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,
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;
570 icmp6h.icmp6_router = router;
571 icmp6h.icmp6_solicited = solicited;
572 icmp6h.icmp6_override = override;
574 __ndisc_send(dev, neigh, daddr, src_addr,
575 &icmp6h, solicited_addr,
576 inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
579 static void ndisc_send_unsol_na(struct net_device *dev)
581 struct inet6_dev *idev;
582 struct inet6_ifaddr *ifa;
583 struct in6_addr mcaddr;
585 idev = in6_dev_get(dev);
586 if (!idev)
587 return;
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);
597 read_unlock_bh(&idev->lock);
599 in6_dev_put(idev);
602 void 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)
606 struct in6_addr addr_buf;
607 struct icmp6hdr icmp6h = {
608 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
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;
618 __ndisc_send(dev, neigh, daddr, saddr,
619 &icmp6h, solicit,
620 !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
623 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
624 const struct in6_addr *daddr)
626 struct icmp6hdr icmp6h = {
627 .icmp6_type = NDISC_ROUTER_SOLICITATION,
629 int send_sllao = dev->addr_len;
631 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
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.
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;
647 in6_ifa_put(ifp);
648 } else {
649 send_sllao = 0;
652 #endif
653 __ndisc_send(dev, NULL, daddr, saddr,
654 &icmp6h, NULL,
655 send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
659 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
662 * "The sender MUST return an ICMP
663 * destination unreachable"
665 dst_link_failure(skb);
666 kfree_skb(skb);
669 /* Called with locked neigh: either read or both */
671 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
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);
679 if (skb && ipv6_chk_addr(dev_net(dev), &ipv6_hdr(skb)->saddr, dev, 1))
680 saddr = &ipv6_hdr(skb)->saddr;
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);
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);
699 static int pndisc_is_router(const void *pkey,
700 struct net_device *dev)
702 struct pneigh_entry *n;
703 int ret = -1;
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);
711 return ret;
714 static void ndisc_recv_ns(struct sk_buff *skb)
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;
731 if (ipv6_addr_is_multicast(&msg->target)) {
732 ND_PRINTK(2, warn, "NS: multicast target address\n");
733 return;
737 * RFC2461 7.1.1:
738 * DAD has to be destined for solicited node multicast address.
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;
749 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
750 ND_PRINTK(2, warn, "NS: invalid ND options\n");
751 return;
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;
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.
767 if (dad) {
768 ND_PRINTK(2, warn,
769 "NS: bad DAD packet (link-layer address option)\n");
770 return;
774 inc = ipv6_addr_is_multicast(daddr);
776 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
777 if (ifp) {
779 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
780 if (dad) {
782 * We are colliding with another node
783 * who is doing DAD
784 * so fail our DAD process
786 addrconf_dad_failure(ifp);
787 return;
788 } else {
790 * This is not a dad solicitation.
791 * If we are an optimistic node,
792 * we should respond.
793 * Otherwise, we should ignore it.
795 if (!(ifp->flags & IFA_F_OPTIMISTIC))
796 goto out;
800 idev = ifp->idev;
801 } else {
802 struct net *net = dev_net(dev);
804 idev = in6_dev_get(dev);
805 if (!idev) {
806 /* XXX: count this drop? */
807 return;
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) {
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
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;
830 } else
831 goto out;
834 if (is_router < 0)
835 is_router = !!idev->cnf.forwarding;
837 if (dad) {
838 ndisc_send_na(dev, NULL, &in6addr_linklocal_allnodes, &msg->target,
839 is_router, 0, (ifp != NULL), 1);
840 goto out;
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);
849 * update / create cache entry
850 * for the source address
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);
866 out:
867 if (ifp)
868 in6_ifa_put(ifp);
869 else
870 in6_dev_put(idev);
873 static void ndisc_recv_na(struct sk_buff *skb)
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;
886 if (skb->len < sizeof(struct nd_msg)) {
887 ND_PRINTK(2, warn, "NA: packet too short\n");
888 return;
891 if (ipv6_addr_is_multicast(&msg->target)) {
892 ND_PRINTK(2, warn, "NA: target address is multicast\n");
893 return;
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;
902 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
903 ND_PRINTK(2, warn, "NS: invalid ND option\n");
904 return;
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;
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;
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 :-)
926 We should not print the error if NA has been
927 received from loopback - it is just our own
928 unsolicited advertisement.
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;
937 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
939 if (neigh) {
940 u8 old_flags = neigh->flags;
941 struct net *net = dev_net(dev);
943 if (neigh->nud_state & NUD_FAILED)
944 goto out;
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.
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;
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));
965 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
967 * Change: router to host
969 struct rt6_info *rt;
970 rt = rt6_get_dflt_router(saddr, dev);
971 if (rt)
972 ip6_del_rt(rt);
975 out:
976 neigh_release(neigh);
980 static void ndisc_recv_rs(struct sk_buff *skb)
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;
990 if (skb->len < sizeof(*rs_msg))
991 return;
993 idev = __in6_dev_get(skb->dev);
994 if (!idev) {
995 ND_PRINTK(1, err, "RS: can't find in6 device\n");
996 return;
999 /* Don't accept RS if we're not in router mode */
1000 if (!idev->cnf.forwarding)
1001 goto out;
1004 * Don't update NCE if src = ::;
1005 * this implies that the source node has no ip address assigned yet.
1007 if (ipv6_addr_any(saddr))
1008 goto out;
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;
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;
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);
1031 out:
1032 return;
1035 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
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));
1047 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1048 if (skb == NULL) {
1049 err = -ENOBUFS;
1050 goto errout;
1053 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1054 if (nlh == NULL) {
1055 goto nla_put_failure;
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;
1065 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
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);
1072 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1073 return;
1075 nla_put_failure:
1076 nlmsg_free(skb);
1077 err = -EMSGSIZE;
1078 errout:
1079 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1082 static inline int accept_ra(struct inet6_dev *in6_dev)
1085 * If forwarding is enabled, RA are not accepted unless the special
1086 * hybrid mode (accept_ra=2) is enabled.
1088 if (in6_dev->cnf.forwarding && in6_dev->cnf.accept_ra < 2)
1089 return 0;
1091 return in6_dev->cnf.accept_ra;
1094 static void ndisc_router_discovery(struct sk_buff *skb)
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;
1105 __u8 * opt = (__u8 *)(ra_msg + 1);
1107 optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
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;
1113 if (optlen < 0) {
1114 ND_PRINTK(2, warn, "RA: packet too short\n");
1115 return;
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;
1123 #endif
1126 * set the RA_RECV flag in the interface
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;
1136 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1137 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1138 return;
1141 if (!accept_ra(in6_dev))
1142 goto skip_linkparms;
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
1150 if (in6_dev->if_flags & IF_RS_SENT) {
1152 * flag that an RA was received after an RS was sent
1153 * out on this interface.
1155 in6_dev->if_flags |= IF_RA_RCVD;
1159 * Remember the managed/otherconf flags from most recently
1160 * received RA message (RFC 2462) -- yoshfuji
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);
1169 if (!in6_dev->cnf.accept_ra_defrtr)
1170 goto skip_defrtr;
1172 if (ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr, NULL, 0))
1173 goto skip_defrtr;
1175 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
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
1185 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
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;
1197 if (rt && lifetime == 0) {
1198 ip6_del_rt(rt);
1199 rt = NULL;
1202 if (rt == NULL && lifetime) {
1203 ND_PRINTK(3, dbg, "RA: adding default router\n");
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;
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;
1221 neigh->flags |= NTF_ROUTER;
1222 } else if (rt) {
1223 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
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);
1235 skip_defrtr:
1238 * Update Reachable Time and Retrans Timer
1241 if (in6_dev->nd_parms) {
1242 unsigned long rtime = ntohl(ra_msg->retrans_timer);
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);
1253 rtime = ntohl(ra_msg->reachable_time);
1254 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1255 rtime = (rtime*HZ)/1000;
1257 if (rtime < HZ/10)
1258 rtime = HZ/10;
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);
1270 skip_linkparms:
1273 * Process options.
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;
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);
1297 if (!accept_ra(in6_dev))
1298 goto out;
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;
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;
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);
1322 skip_routeinfo:
1323 #endif
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
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;
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);
1342 if (ndopts.nd_opts_mtu) {
1343 __be32 n;
1344 u32 mtu;
1346 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1347 mtu = ntohl(n);
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;
1354 if (rt)
1355 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
1357 rt6_mtu_change(skb->dev, mtu);
1361 if (ndopts.nd_useropts) {
1362 struct nd_opt_hdr *p;
1363 for (p = ndopts.nd_useropts;
1365 p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1366 ndisc_ra_useropt(skb, p);
1370 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1371 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1373 out:
1374 if (rt)
1375 dst_release(&rt->dst);
1376 if (neigh)
1377 neigh_release(neigh);
1380 static void ndisc_redirect_rcv(struct sk_buff *skb)
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;
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;
1400 #endif
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;
1408 optlen = skb->tail - skb->transport_header;
1409 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1411 if (optlen < 0) {
1412 ND_PRINTK(2, warn, "Redirect: packet too short\n");
1413 return;
1416 icmph = icmp6_hdr(skb);
1417 target = (const struct in6_addr *) (icmph + 1);
1418 dest = target + 1;
1420 if (ipv6_addr_is_multicast(dest)) {
1421 ND_PRINTK(2, warn,
1422 "Redirect: destination address is multicast\n");
1423 return;
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;
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;
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.
1446 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1447 ND_PRINTK(2, warn, "Redirect: invalid ND options\n");
1448 return;
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;
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);
1469 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
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;
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;
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;
1502 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1503 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1505 dst = ip6_route_output(net, NULL, &fl6);
1506 if (dst->error) {
1507 dst_release(dst);
1508 return;
1510 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1511 if (IS_ERR(dst))
1512 return;
1514 rt = (struct rt6_info *) dst;
1516 if (rt->rt6i_flags & RTF_GATEWAY) {
1517 ND_PRINTK(2, warn,
1518 "Redirect: destination is not a neighbour\n");
1519 goto release;
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;
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;
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);
1543 neigh_release(neigh);
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;
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;
1564 skb_reserve(buff, hlen);
1565 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1566 IPPROTO_ICMPV6, len);
1568 skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1569 skb_put(buff, len);
1570 icmph = icmp6_hdr(buff);
1572 memset(icmph, 0, sizeof(struct icmp6hdr));
1573 icmph->icmp6_type = NDISC_REDIRECT;
1576 * copy target and destination addresses
1579 addrp = (struct in6_addr *)(icmph + 1);
1580 *addrp = *target;
1581 addrp++;
1582 *addrp = ipv6_hdr(skb)->daddr;
1584 opt = (u8*) (addrp + 1);
1587 * include target_address option
1590 if (ha)
1591 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1592 dev->addr_len, dev->type);
1595 * build redirect option and copy skb over to the new packet.
1598 memset(opt, 0, 8);
1599 *(opt++) = ND_OPT_REDIRECT_HDR;
1600 *(opt++) = (rd_len >> 3);
1601 opt += 6;
1603 memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1605 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1606 len, IPPROTO_ICMPV6,
1607 csum_partial(icmph, len, 0));
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);
1620 rcu_read_unlock();
1621 return;
1623 release:
1624 dst_release(dst);
1627 static void pndisc_redo(struct sk_buff *skb)
1629 ndisc_recv_ns(skb);
1630 kfree_skb(skb);
1633 int ndisc_rcv(struct sk_buff *skb)
1635 struct nd_msg *msg;
1637 if (!pskb_may_pull(skb, skb->len))
1638 return 0;
1640 msg = (struct nd_msg *)skb_transport_header(skb);
1642 __skb_push(skb, skb->data - skb_transport_header(skb));
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;
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;
1656 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1658 switch (msg->icmph.icmp6_type) {
1659 case NDISC_NEIGHBOUR_SOLICITATION:
1660 ndisc_recv_ns(skb);
1661 break;
1663 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1664 ndisc_recv_na(skb);
1665 break;
1667 case NDISC_ROUTER_SOLICITATION:
1668 ndisc_recv_rs(skb);
1669 break;
1671 case NDISC_ROUTER_ADVERTISEMENT:
1672 ndisc_router_discovery(skb);
1673 break;
1675 case NDISC_REDIRECT:
1676 ndisc_redirect_rcv(skb);
1677 break;
1680 return 0;
1683 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1685 struct net_device *dev = ptr;
1686 struct net *net = dev_net(dev);
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;
1704 return NOTIFY_DONE;
1707 static struct notifier_block ndisc_netdev_notifier = {
1708 .notifier_call = ndisc_netdev_event,
1711 #ifdef CONFIG_SYSCTL
1712 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1713 const char *func, const char *dev_name)
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++;
1727 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1729 struct net_device *dev = ctl->extra1;
1730 struct inet6_dev *idev;
1731 int ret;
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");
1737 if (strcmp(ctl->procname, "retrans_time") == 0)
1738 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1740 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1741 ret = proc_dointvec_jiffies(ctl, write,
1742 buffer, lenp, ppos);
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;
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);
1758 return ret;
1762 #endif
1764 static int __net_init ndisc_net_init(struct net *net)
1766 struct ipv6_pinfo *np;
1767 struct sock *sk;
1768 int err;
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;
1779 net->ipv6.ndisc_sk = sk;
1781 np = inet6_sk(sk);
1782 np->hop_limit = 255;
1783 /* Do not loopback ndisc messages */
1784 np->mc_loop = 0;
1786 return 0;
1789 static void __net_exit ndisc_net_exit(struct net *net)
1791 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1794 static struct pernet_operations ndisc_net_ops = {
1795 .init = ndisc_net_init,
1796 .exit = ndisc_net_exit,
1799 int __init ndisc_init(void)
1801 int err;
1803 err = register_pernet_subsys(&ndisc_net_ops);
1804 if (err)
1805 return err;
1807 * Initialize the neighbour table
1809 neigh_table_init(&nd_tbl);
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;
1820 out:
1821 return err;
1823 out_unregister_sysctl:
1824 #ifdef CONFIG_SYSCTL
1825 neigh_sysctl_unregister(&nd_tbl.parms);
1826 out_unregister_pernet:
1827 #endif
1828 unregister_pernet_subsys(&ndisc_net_ops);
1829 goto out;
1832 void ndisc_cleanup(void)
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);