warning removal
[cor_2_6_31.git] / net / ipv6 / ndisc.c
blob9eb68e92cc18108e762f24cc7f12307018f477cc
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 * Pierre Ynard : export userland ND options
19 * through netlink (RDNSS support)
20 * Lars Fenneberg : fixed MTU setting on receipt
21 * of an RA.
22 * Janos Farkas : kmalloc failure checks
23 * Alexey Kuznetsov : state machine reworked
24 * and moved to net/core.
25 * Pekka Savola : RFC2461 validation
26 * YOSHIFUJI Hideaki @USAGI : Verify ND options properly
29 /* Set to 3 to get tracing... */
30 #define ND_DEBUG 1
32 #define ND_PRINTK(fmt, args...) do { if (net_ratelimit()) { printk(fmt, ## args); } } while(0)
33 #define ND_NOPRINTK(x...) do { ; } while(0)
34 #define ND_PRINTK0 ND_PRINTK
35 #define ND_PRINTK1 ND_NOPRINTK
36 #define ND_PRINTK2 ND_NOPRINTK
37 #define ND_PRINTK3 ND_NOPRINTK
38 #if ND_DEBUG >= 1
39 #undef ND_PRINTK1
40 #define ND_PRINTK1 ND_PRINTK
41 #endif
42 #if ND_DEBUG >= 2
43 #undef ND_PRINTK2
44 #define ND_PRINTK2 ND_PRINTK
45 #endif
46 #if ND_DEBUG >= 3
47 #undef ND_PRINTK3
48 #define ND_PRINTK3 ND_PRINTK
49 #endif
51 #include <linux/module.h>
52 #include <linux/errno.h>
53 #include <linux/types.h>
54 #include <linux/socket.h>
55 #include <linux/sockios.h>
56 #include <linux/sched.h>
57 #include <linux/net.h>
58 #include <linux/in6.h>
59 #include <linux/route.h>
60 #include <linux/init.h>
61 #include <linux/rcupdate.h>
62 #ifdef CONFIG_SYSCTL
63 #include <linux/sysctl.h>
64 #endif
66 #include <linux/if_addr.h>
67 #include <linux/if_arp.h>
68 #include <linux/ipv6.h>
69 #include <linux/icmpv6.h>
70 #include <linux/jhash.h>
72 #include <net/sock.h>
73 #include <net/snmp.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/icmp.h>
82 #include <net/netlink.h>
83 #include <linux/rtnetlink.h>
85 #include <net/flow.h>
86 #include <net/ip6_checksum.h>
87 #include <net/inet_common.h>
88 #include <linux/proc_fs.h>
90 #include <linux/netfilter.h>
91 #include <linux/netfilter_ipv6.h>
93 static u32 ndisc_hash(const void *pkey, const struct net_device *dev);
94 static int ndisc_constructor(struct neighbour *neigh);
95 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
96 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
97 static int pndisc_constructor(struct pneigh_entry *n);
98 static void pndisc_destructor(struct pneigh_entry *n);
99 static void pndisc_redo(struct sk_buff *skb);
101 static struct neigh_ops ndisc_generic_ops = {
102 .family = AF_INET6,
103 .solicit = ndisc_solicit,
104 .error_report = ndisc_error_report,
105 .output = neigh_resolve_output,
106 .connected_output = neigh_connected_output,
107 .hh_output = dev_queue_xmit,
108 .queue_xmit = dev_queue_xmit,
111 static struct neigh_ops ndisc_hh_ops = {
112 .family = AF_INET6,
113 .solicit = ndisc_solicit,
114 .error_report = ndisc_error_report,
115 .output = neigh_resolve_output,
116 .connected_output = neigh_resolve_output,
117 .hh_output = dev_queue_xmit,
118 .queue_xmit = dev_queue_xmit,
122 static struct neigh_ops ndisc_direct_ops = {
123 .family = AF_INET6,
124 .output = dev_queue_xmit,
125 .connected_output = dev_queue_xmit,
126 .hh_output = dev_queue_xmit,
127 .queue_xmit = dev_queue_xmit,
130 struct neigh_table nd_tbl = {
131 .family = AF_INET6,
132 .entry_size = sizeof(struct neighbour) + sizeof(struct in6_addr),
133 .key_len = sizeof(struct in6_addr),
134 .hash = ndisc_hash,
135 .constructor = ndisc_constructor,
136 .pconstructor = pndisc_constructor,
137 .pdestructor = pndisc_destructor,
138 .proxy_redo = pndisc_redo,
139 .id = "ndisc_cache",
140 .parms = {
141 .tbl = &nd_tbl,
142 .base_reachable_time = 30 * HZ,
143 .retrans_time = 1 * HZ,
144 .gc_staletime = 60 * HZ,
145 .reachable_time = 30 * HZ,
146 .delay_probe_time = 5 * HZ,
147 .queue_len = 3,
148 .ucast_probes = 3,
149 .mcast_probes = 3,
150 .anycast_delay = 1 * HZ,
151 .proxy_delay = (8 * HZ) / 10,
152 .proxy_qlen = 64,
154 .gc_interval = 30 * HZ,
155 .gc_thresh1 = 128,
156 .gc_thresh2 = 512,
157 .gc_thresh3 = 1024,
160 /* ND options */
161 struct ndisc_options {
162 struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
163 #ifdef CONFIG_IPV6_ROUTE_INFO
164 struct nd_opt_hdr *nd_opts_ri;
165 struct nd_opt_hdr *nd_opts_ri_end;
166 #endif
167 struct nd_opt_hdr *nd_useropts;
168 struct nd_opt_hdr *nd_useropts_end;
171 #define nd_opts_src_lladdr nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
172 #define nd_opts_tgt_lladdr nd_opt_array[ND_OPT_TARGET_LL_ADDR]
173 #define nd_opts_pi nd_opt_array[ND_OPT_PREFIX_INFO]
174 #define nd_opts_pi_end nd_opt_array[__ND_OPT_PREFIX_INFO_END]
175 #define nd_opts_rh nd_opt_array[ND_OPT_REDIRECT_HDR]
176 #define nd_opts_mtu nd_opt_array[ND_OPT_MTU]
178 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
181 * Return the padding between the option length and the start of the
182 * link addr. Currently only IP-over-InfiniBand needs this, although
183 * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
184 * also need a pad of 2.
186 static int ndisc_addr_option_pad(unsigned short type)
188 switch (type) {
189 case ARPHRD_INFINIBAND: return 2;
190 default: return 0;
194 static inline int ndisc_opt_addr_space(struct net_device *dev)
196 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
199 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
200 unsigned short addr_type)
202 int space = NDISC_OPT_SPACE(data_len);
203 int pad = ndisc_addr_option_pad(addr_type);
205 opt[0] = type;
206 opt[1] = space>>3;
208 memset(opt + 2, 0, pad);
209 opt += pad;
210 space -= pad;
212 memcpy(opt+2, data, data_len);
213 data_len += 2;
214 opt += data_len;
215 if ((space -= data_len) > 0)
216 memset(opt, 0, space);
217 return opt + space;
220 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
221 struct nd_opt_hdr *end)
223 int type;
224 if (!cur || !end || cur >= end)
225 return NULL;
226 type = cur->nd_opt_type;
227 do {
228 cur = ((void *)cur) + (cur->nd_opt_len << 3);
229 } while(cur < end && cur->nd_opt_type != type);
230 return (cur <= end && cur->nd_opt_type == type ? cur : NULL);
233 static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
235 return (opt->nd_opt_type == ND_OPT_RDNSS);
238 static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
239 struct nd_opt_hdr *end)
241 if (!cur || !end || cur >= end)
242 return NULL;
243 do {
244 cur = ((void *)cur) + (cur->nd_opt_len << 3);
245 } while(cur < end && !ndisc_is_useropt(cur));
246 return (cur <= end && ndisc_is_useropt(cur) ? cur : NULL);
249 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
250 struct ndisc_options *ndopts)
252 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
254 if (!nd_opt || opt_len < 0 || !ndopts)
255 return NULL;
256 memset(ndopts, 0, sizeof(*ndopts));
257 while (opt_len) {
258 int l;
259 if (opt_len < sizeof(struct nd_opt_hdr))
260 return NULL;
261 l = nd_opt->nd_opt_len << 3;
262 if (opt_len < l || l == 0)
263 return NULL;
264 switch (nd_opt->nd_opt_type) {
265 case ND_OPT_SOURCE_LL_ADDR:
266 case ND_OPT_TARGET_LL_ADDR:
267 case ND_OPT_MTU:
268 case ND_OPT_REDIRECT_HDR:
269 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
270 ND_PRINTK2(KERN_WARNING
271 "%s(): duplicated ND6 option found: type=%d\n",
272 __func__,
273 nd_opt->nd_opt_type);
274 } else {
275 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
277 break;
278 case ND_OPT_PREFIX_INFO:
279 ndopts->nd_opts_pi_end = nd_opt;
280 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
281 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
282 break;
283 #ifdef CONFIG_IPV6_ROUTE_INFO
284 case ND_OPT_ROUTE_INFO:
285 ndopts->nd_opts_ri_end = nd_opt;
286 if (!ndopts->nd_opts_ri)
287 ndopts->nd_opts_ri = nd_opt;
288 break;
289 #endif
290 default:
291 if (ndisc_is_useropt(nd_opt)) {
292 ndopts->nd_useropts_end = nd_opt;
293 if (!ndopts->nd_useropts)
294 ndopts->nd_useropts = nd_opt;
295 } else {
297 * Unknown options must be silently ignored,
298 * to accommodate future extension to the
299 * protocol.
301 ND_PRINTK2(KERN_NOTICE
302 "%s(): ignored unsupported option; type=%d, len=%d\n",
303 __func__,
304 nd_opt->nd_opt_type, nd_opt->nd_opt_len);
307 opt_len -= l;
308 nd_opt = ((void *)nd_opt) + l;
310 return ndopts;
313 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
314 struct net_device *dev)
316 u8 *lladdr = (u8 *)(p + 1);
317 int lladdrlen = p->nd_opt_len << 3;
318 int prepad = ndisc_addr_option_pad(dev->type);
319 if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
320 return NULL;
321 return (lladdr + prepad);
324 int ndisc_mc_map(struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
326 switch (dev->type) {
327 case ARPHRD_ETHER:
328 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
329 case ARPHRD_FDDI:
330 ipv6_eth_mc_map(addr, buf);
331 return 0;
332 case ARPHRD_IEEE802_TR:
333 ipv6_tr_mc_map(addr,buf);
334 return 0;
335 case ARPHRD_ARCNET:
336 ipv6_arcnet_mc_map(addr, buf);
337 return 0;
338 case ARPHRD_INFINIBAND:
339 ipv6_ib_mc_map(addr, dev->broadcast, buf);
340 return 0;
341 default:
342 if (dir) {
343 memcpy(buf, dev->broadcast, dev->addr_len);
344 return 0;
347 return -EINVAL;
350 EXPORT_SYMBOL(ndisc_mc_map);
352 static u32 ndisc_hash(const void *pkey, const struct net_device *dev)
354 const u32 *p32 = pkey;
355 u32 addr_hash, i;
357 addr_hash = 0;
358 for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
359 addr_hash ^= *p32++;
361 return jhash_2words(addr_hash, dev->ifindex, nd_tbl.hash_rnd);
364 static int ndisc_constructor(struct neighbour *neigh)
366 struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
367 struct net_device *dev = neigh->dev;
368 struct inet6_dev *in6_dev;
369 struct neigh_parms *parms;
370 int is_multicast = ipv6_addr_is_multicast(addr);
372 rcu_read_lock();
373 in6_dev = in6_dev_get(dev);
374 if (in6_dev == NULL) {
375 rcu_read_unlock();
376 return -EINVAL;
379 parms = in6_dev->nd_parms;
380 __neigh_parms_put(neigh->parms);
381 neigh->parms = neigh_parms_clone(parms);
382 rcu_read_unlock();
384 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
385 if (!dev->header_ops) {
386 neigh->nud_state = NUD_NOARP;
387 neigh->ops = &ndisc_direct_ops;
388 neigh->output = neigh->ops->queue_xmit;
389 } else {
390 if (is_multicast) {
391 neigh->nud_state = NUD_NOARP;
392 ndisc_mc_map(addr, neigh->ha, dev, 1);
393 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
394 neigh->nud_state = NUD_NOARP;
395 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
396 if (dev->flags&IFF_LOOPBACK)
397 neigh->type = RTN_LOCAL;
398 } else if (dev->flags&IFF_POINTOPOINT) {
399 neigh->nud_state = NUD_NOARP;
400 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
402 if (dev->header_ops->cache)
403 neigh->ops = &ndisc_hh_ops;
404 else
405 neigh->ops = &ndisc_generic_ops;
406 if (neigh->nud_state&NUD_VALID)
407 neigh->output = neigh->ops->connected_output;
408 else
409 neigh->output = neigh->ops->output;
411 in6_dev_put(in6_dev);
412 return 0;
415 static int pndisc_constructor(struct pneigh_entry *n)
417 struct in6_addr *addr = (struct in6_addr*)&n->key;
418 struct in6_addr maddr;
419 struct net_device *dev = n->dev;
421 if (dev == NULL || __in6_dev_get(dev) == NULL)
422 return -EINVAL;
423 addrconf_addr_solict_mult(addr, &maddr);
424 ipv6_dev_mc_inc(dev, &maddr);
425 return 0;
428 static void pndisc_destructor(struct pneigh_entry *n)
430 struct in6_addr *addr = (struct in6_addr*)&n->key;
431 struct in6_addr maddr;
432 struct net_device *dev = n->dev;
434 if (dev == NULL || __in6_dev_get(dev) == NULL)
435 return;
436 addrconf_addr_solict_mult(addr, &maddr);
437 ipv6_dev_mc_dec(dev, &maddr);
440 struct sk_buff *ndisc_build_skb(struct net_device *dev,
441 const struct in6_addr *daddr,
442 const struct in6_addr *saddr,
443 struct icmp6hdr *icmp6h,
444 const struct in6_addr *target,
445 int llinfo)
447 struct net *net = dev_net(dev);
448 struct sock *sk = net->ipv6.ndisc_sk;
449 struct sk_buff *skb;
450 struct icmp6hdr *hdr;
451 int len;
452 int err;
453 u8 *opt;
455 if (!dev->addr_len)
456 llinfo = 0;
458 len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
459 if (llinfo)
460 len += ndisc_opt_addr_space(dev);
462 skb = sock_alloc_send_skb(sk,
463 (MAX_HEADER + sizeof(struct ipv6hdr) +
464 len + LL_ALLOCATED_SPACE(dev)),
465 1, &err);
466 if (!skb) {
467 ND_PRINTK0(KERN_ERR
468 "ICMPv6 ND: %s() failed to allocate an skb, err=%d.\n",
469 __func__, err);
470 return NULL;
473 skb_reserve(skb, LL_RESERVED_SPACE(dev));
474 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
476 skb->transport_header = skb->tail;
477 skb_put(skb, len);
479 hdr = (struct icmp6hdr *)skb_transport_header(skb);
480 memcpy(hdr, icmp6h, sizeof(*hdr));
482 opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
483 if (target) {
484 ipv6_addr_copy((struct in6_addr *)opt, target);
485 opt += sizeof(*target);
488 if (llinfo)
489 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
490 dev->addr_len, dev->type);
492 hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
493 IPPROTO_ICMPV6,
494 csum_partial(hdr,
495 len, 0));
497 return skb;
500 EXPORT_SYMBOL(ndisc_build_skb);
502 void ndisc_send_skb(struct sk_buff *skb,
503 struct net_device *dev,
504 struct neighbour *neigh,
505 const struct in6_addr *daddr,
506 const struct in6_addr *saddr,
507 struct icmp6hdr *icmp6h)
509 struct flowi fl;
510 struct dst_entry *dst;
511 struct net *net = dev_net(dev);
512 struct sock *sk = net->ipv6.ndisc_sk;
513 struct inet6_dev *idev;
514 int err;
515 u8 type;
517 type = icmp6h->icmp6_type;
519 icmpv6_flow_init(sk, &fl, type, saddr, daddr, dev->ifindex);
521 dst = icmp6_dst_alloc(dev, neigh, daddr);
522 if (!dst) {
523 kfree_skb(skb);
524 return;
527 err = xfrm_lookup(net, &dst, &fl, NULL, 0);
528 if (err < 0) {
529 kfree_skb(skb);
530 return;
533 skb_dst_set(skb, dst);
535 idev = in6_dev_get(dst->dev);
536 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
538 err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
539 dst_output);
540 if (!err) {
541 ICMP6MSGOUT_INC_STATS(net, idev, type);
542 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
545 if (likely(idev != NULL))
546 in6_dev_put(idev);
549 EXPORT_SYMBOL(ndisc_send_skb);
552 * Send a Neighbour Discover packet
554 static void __ndisc_send(struct net_device *dev,
555 struct neighbour *neigh,
556 const struct in6_addr *daddr,
557 const struct in6_addr *saddr,
558 struct icmp6hdr *icmp6h, const struct in6_addr *target,
559 int llinfo)
561 struct sk_buff *skb;
563 skb = ndisc_build_skb(dev, daddr, saddr, icmp6h, target, llinfo);
564 if (!skb)
565 return;
567 ndisc_send_skb(skb, dev, neigh, daddr, saddr, icmp6h);
570 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
571 const struct in6_addr *daddr,
572 const struct in6_addr *solicited_addr,
573 int router, int solicited, int override, int inc_opt)
575 struct in6_addr tmpaddr;
576 struct inet6_ifaddr *ifp;
577 const struct in6_addr *src_addr;
578 struct icmp6hdr icmp6h = {
579 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
582 /* for anycast or proxy, solicited_addr != src_addr */
583 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
584 if (ifp) {
585 src_addr = solicited_addr;
586 if (ifp->flags & IFA_F_OPTIMISTIC)
587 override = 0;
588 in6_ifa_put(ifp);
589 } else {
590 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
591 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
592 &tmpaddr))
593 return;
594 src_addr = &tmpaddr;
597 icmp6h.icmp6_router = router;
598 icmp6h.icmp6_solicited = solicited;
599 icmp6h.icmp6_override = override;
601 __ndisc_send(dev, neigh, daddr, src_addr,
602 &icmp6h, solicited_addr,
603 inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
606 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
607 const struct in6_addr *solicit,
608 const struct in6_addr *daddr, const struct in6_addr *saddr)
610 struct in6_addr addr_buf;
611 struct icmp6hdr icmp6h = {
612 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
615 if (saddr == NULL) {
616 if (ipv6_get_lladdr(dev, &addr_buf,
617 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
618 return;
619 saddr = &addr_buf;
622 __ndisc_send(dev, neigh, daddr, saddr,
623 &icmp6h, solicit,
624 !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
627 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
628 const struct in6_addr *daddr)
630 struct icmp6hdr icmp6h = {
631 .icmp6_type = NDISC_ROUTER_SOLICITATION,
633 int send_sllao = dev->addr_len;
635 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
637 * According to section 2.2 of RFC 4429, we must not
638 * send router solicitations with a sllao from
639 * optimistic addresses, but we may send the solicitation
640 * if we don't include the sllao. So here we check
641 * if our address is optimistic, and if so, we
642 * suppress the inclusion of the sllao.
644 if (send_sllao) {
645 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
646 dev, 1);
647 if (ifp) {
648 if (ifp->flags & IFA_F_OPTIMISTIC) {
649 send_sllao = 0;
651 in6_ifa_put(ifp);
652 } else {
653 send_sllao = 0;
656 #endif
657 __ndisc_send(dev, NULL, daddr, saddr,
658 &icmp6h, NULL,
659 send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
661 EXPORT_SYMBOL(ndisc_send_rs);
664 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
667 * "The sender MUST return an ICMP
668 * destination unreachable"
670 dst_link_failure(skb);
671 kfree_skb(skb);
674 /* Called with locked neigh: either read or both */
676 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
678 struct in6_addr *saddr = NULL;
679 struct in6_addr mcaddr;
680 struct net_device *dev = neigh->dev;
681 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
682 int probes = atomic_read(&neigh->probes);
684 if (skb && ipv6_chk_addr(dev_net(dev), &ipv6_hdr(skb)->saddr, dev, 1))
685 saddr = &ipv6_hdr(skb)->saddr;
687 if ((probes -= neigh->parms->ucast_probes) < 0) {
688 if (!(neigh->nud_state & NUD_VALID)) {
689 ND_PRINTK1(KERN_DEBUG "%s(): trying to ucast probe in NUD_INVALID: %pI6\n",
690 __func__, target);
692 ndisc_send_ns(dev, neigh, target, target, saddr);
693 } else if ((probes -= neigh->parms->app_probes) < 0) {
694 #ifdef CONFIG_ARPD
695 neigh_app_ns(neigh);
696 #endif
697 } else {
698 addrconf_addr_solict_mult(target, &mcaddr);
699 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
703 static int pndisc_is_router(const void *pkey,
704 struct net_device *dev)
706 struct pneigh_entry *n;
707 int ret = -1;
709 read_lock_bh(&nd_tbl.lock);
710 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
711 if (n)
712 ret = !!(n->flags & NTF_ROUTER);
713 read_unlock_bh(&nd_tbl.lock);
715 return ret;
718 static void ndisc_recv_ns(struct sk_buff *skb)
720 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
721 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
722 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
723 u8 *lladdr = NULL;
724 u32 ndoptlen = skb->tail - (skb->transport_header +
725 offsetof(struct nd_msg, opt));
726 struct ndisc_options ndopts;
727 struct net_device *dev = skb->dev;
728 struct inet6_ifaddr *ifp;
729 struct inet6_dev *idev = NULL;
730 struct neighbour *neigh;
731 int dad = ipv6_addr_any(saddr);
732 int inc;
733 int is_router = -1;
735 if (ipv6_addr_is_multicast(&msg->target)) {
736 ND_PRINTK2(KERN_WARNING
737 "ICMPv6 NS: multicast target address");
738 return;
742 * RFC2461 7.1.1:
743 * DAD has to be destined for solicited node multicast address.
745 if (dad &&
746 !(daddr->s6_addr32[0] == htonl(0xff020000) &&
747 daddr->s6_addr32[1] == htonl(0x00000000) &&
748 daddr->s6_addr32[2] == htonl(0x00000001) &&
749 daddr->s6_addr [12] == 0xff )) {
750 ND_PRINTK2(KERN_WARNING
751 "ICMPv6 NS: bad DAD packet (wrong destination)\n");
752 return;
755 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
756 ND_PRINTK2(KERN_WARNING
757 "ICMPv6 NS: invalid ND options\n");
758 return;
761 if (ndopts.nd_opts_src_lladdr) {
762 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
763 if (!lladdr) {
764 ND_PRINTK2(KERN_WARNING
765 "ICMPv6 NS: invalid link-layer address length\n");
766 return;
769 /* RFC2461 7.1.1:
770 * If the IP source address is the unspecified address,
771 * there MUST NOT be source link-layer address option
772 * in the message.
774 if (dad) {
775 ND_PRINTK2(KERN_WARNING
776 "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
777 return;
781 inc = ipv6_addr_is_multicast(daddr);
783 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
784 if (ifp) {
786 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
787 if (dad) {
788 if (dev->type == ARPHRD_IEEE802_TR) {
789 const unsigned char *sadr;
790 sadr = skb_mac_header(skb);
791 if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
792 sadr[9] == dev->dev_addr[1] &&
793 sadr[10] == dev->dev_addr[2] &&
794 sadr[11] == dev->dev_addr[3] &&
795 sadr[12] == dev->dev_addr[4] &&
796 sadr[13] == dev->dev_addr[5]) {
797 /* looped-back to us */
798 goto out;
803 * We are colliding with another node
804 * who is doing DAD
805 * so fail our DAD process
807 addrconf_dad_failure(ifp);
808 return;
809 } else {
811 * This is not a dad solicitation.
812 * If we are an optimistic node,
813 * we should respond.
814 * Otherwise, we should ignore it.
816 if (!(ifp->flags & IFA_F_OPTIMISTIC))
817 goto out;
821 idev = ifp->idev;
822 } else {
823 struct net *net = dev_net(dev);
825 idev = in6_dev_get(dev);
826 if (!idev) {
827 /* XXX: count this drop? */
828 return;
831 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
832 (idev->cnf.forwarding &&
833 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
834 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
835 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
836 skb->pkt_type != PACKET_HOST &&
837 inc != 0 &&
838 idev->nd_parms->proxy_delay != 0) {
840 * for anycast or proxy,
841 * sender should delay its response
842 * by a random time between 0 and
843 * MAX_ANYCAST_DELAY_TIME seconds.
844 * (RFC2461) -- yoshfuji
846 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
847 if (n)
848 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
849 goto out;
851 } else
852 goto out;
855 if (is_router < 0)
856 is_router = !!idev->cnf.forwarding;
858 if (dad) {
859 ndisc_send_na(dev, NULL, &in6addr_linklocal_allnodes, &msg->target,
860 is_router, 0, (ifp != NULL), 1);
861 goto out;
864 if (inc)
865 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
866 else
867 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
870 * update / create cache entry
871 * for the source address
873 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
874 !inc || lladdr || !dev->addr_len);
875 if (neigh)
876 neigh_update(neigh, lladdr, NUD_STALE,
877 NEIGH_UPDATE_F_WEAK_OVERRIDE|
878 NEIGH_UPDATE_F_OVERRIDE);
879 if (neigh || !dev->header_ops) {
880 ndisc_send_na(dev, neigh, saddr, &msg->target,
881 is_router,
882 1, (ifp != NULL && inc), inc);
883 if (neigh)
884 neigh_release(neigh);
887 out:
888 if (ifp)
889 in6_ifa_put(ifp);
890 else
891 in6_dev_put(idev);
893 return;
896 static void ndisc_recv_na(struct sk_buff *skb)
898 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
899 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
900 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
901 u8 *lladdr = NULL;
902 u32 ndoptlen = skb->tail - (skb->transport_header +
903 offsetof(struct nd_msg, opt));
904 struct ndisc_options ndopts;
905 struct net_device *dev = skb->dev;
906 struct inet6_ifaddr *ifp;
907 struct neighbour *neigh;
909 if (skb->len < sizeof(struct nd_msg)) {
910 ND_PRINTK2(KERN_WARNING
911 "ICMPv6 NA: packet too short\n");
912 return;
915 if (ipv6_addr_is_multicast(&msg->target)) {
916 ND_PRINTK2(KERN_WARNING
917 "ICMPv6 NA: target address is multicast.\n");
918 return;
921 if (ipv6_addr_is_multicast(daddr) &&
922 msg->icmph.icmp6_solicited) {
923 ND_PRINTK2(KERN_WARNING
924 "ICMPv6 NA: solicited NA is multicasted.\n");
925 return;
928 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
929 ND_PRINTK2(KERN_WARNING
930 "ICMPv6 NS: invalid ND option\n");
931 return;
933 if (ndopts.nd_opts_tgt_lladdr) {
934 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
935 if (!lladdr) {
936 ND_PRINTK2(KERN_WARNING
937 "ICMPv6 NA: invalid link-layer address length\n");
938 return;
941 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
942 if (ifp) {
943 if (ifp->flags & IFA_F_TENTATIVE) {
944 addrconf_dad_failure(ifp);
945 return;
947 /* What should we make now? The advertisement
948 is invalid, but ndisc specs say nothing
949 about it. It could be misconfiguration, or
950 an smart proxy agent tries to help us :-)
952 We should not print the error if NA has been
953 received from loopback - it is just our own
954 unsolicited advertisement.
956 if (skb->pkt_type != PACKET_LOOPBACK)
957 ND_PRINTK1(KERN_WARNING
958 "ICMPv6 NA: someone advertises our address on %s!\n",
959 ifp->idev->dev->name);
960 in6_ifa_put(ifp);
961 return;
963 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
965 if (neigh) {
966 u8 old_flags = neigh->flags;
967 struct net *net = dev_net(dev);
969 if (neigh->nud_state & NUD_FAILED)
970 goto out;
973 * Don't update the neighbor cache entry on a proxy NA from
974 * ourselves because either the proxied node is off link or it
975 * has already sent a NA to us.
977 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
978 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
979 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
980 /* XXX: idev->cnf.prixy_ndp */
981 goto out;
984 neigh_update(neigh, lladdr,
985 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
986 NEIGH_UPDATE_F_WEAK_OVERRIDE|
987 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
988 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
989 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
991 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
993 * Change: router to host
995 struct rt6_info *rt;
996 rt = rt6_get_dflt_router(saddr, dev);
997 if (rt)
998 ip6_del_rt(rt);
1001 out:
1002 neigh_release(neigh);
1006 static void ndisc_recv_rs(struct sk_buff *skb)
1008 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1009 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1010 struct neighbour *neigh;
1011 struct inet6_dev *idev;
1012 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1013 struct ndisc_options ndopts;
1014 u8 *lladdr = NULL;
1016 if (skb->len < sizeof(*rs_msg))
1017 return;
1019 idev = in6_dev_get(skb->dev);
1020 if (!idev) {
1021 if (net_ratelimit())
1022 ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
1023 return;
1026 /* Don't accept RS if we're not in router mode */
1027 if (!idev->cnf.forwarding)
1028 goto out;
1031 * Don't update NCE if src = ::;
1032 * this implies that the source node has no ip address assigned yet.
1034 if (ipv6_addr_any(saddr))
1035 goto out;
1037 /* Parse ND options */
1038 if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1039 if (net_ratelimit())
1040 ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1041 goto out;
1044 if (ndopts.nd_opts_src_lladdr) {
1045 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1046 skb->dev);
1047 if (!lladdr)
1048 goto out;
1051 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1052 if (neigh) {
1053 neigh_update(neigh, lladdr, NUD_STALE,
1054 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1055 NEIGH_UPDATE_F_OVERRIDE|
1056 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1057 neigh_release(neigh);
1059 out:
1060 in6_dev_put(idev);
1063 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1065 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1066 struct sk_buff *skb;
1067 struct nlmsghdr *nlh;
1068 struct nduseroptmsg *ndmsg;
1069 struct net *net = dev_net(ra->dev);
1070 int err;
1071 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1072 + (opt->nd_opt_len << 3));
1073 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1075 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1076 if (skb == NULL) {
1077 err = -ENOBUFS;
1078 goto errout;
1081 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1082 if (nlh == NULL) {
1083 goto nla_put_failure;
1086 ndmsg = nlmsg_data(nlh);
1087 ndmsg->nduseropt_family = AF_INET6;
1088 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1089 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1090 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1091 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1093 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1095 NLA_PUT(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
1096 &ipv6_hdr(ra)->saddr);
1097 nlmsg_end(skb, nlh);
1099 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1100 return;
1102 nla_put_failure:
1103 nlmsg_free(skb);
1104 err = -EMSGSIZE;
1105 errout:
1106 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1109 static void ndisc_router_discovery(struct sk_buff *skb)
1111 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1112 struct neighbour *neigh = NULL;
1113 struct inet6_dev *in6_dev;
1114 struct rt6_info *rt = NULL;
1115 int lifetime;
1116 struct ndisc_options ndopts;
1117 int optlen;
1118 unsigned int pref = 0;
1120 __u8 * opt = (__u8 *)(ra_msg + 1);
1122 optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1124 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1125 ND_PRINTK2(KERN_WARNING
1126 "ICMPv6 RA: source address is not link-local.\n");
1127 return;
1129 if (optlen < 0) {
1130 ND_PRINTK2(KERN_WARNING
1131 "ICMPv6 RA: packet too short\n");
1132 return;
1135 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1136 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1137 ND_PRINTK2(KERN_WARNING
1138 "ICMPv6 RA: from host or unauthorized router\n");
1139 return;
1141 #endif
1144 * set the RA_RECV flag in the interface
1147 in6_dev = in6_dev_get(skb->dev);
1148 if (in6_dev == NULL) {
1149 ND_PRINTK0(KERN_ERR
1150 "ICMPv6 RA: can't find inet6 device for %s.\n",
1151 skb->dev->name);
1152 return;
1154 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra) {
1155 in6_dev_put(in6_dev);
1156 return;
1159 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1160 in6_dev_put(in6_dev);
1161 ND_PRINTK2(KERN_WARNING
1162 "ICMP6 RA: invalid ND options\n");
1163 return;
1166 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1167 /* skip link-specific parameters from interior routers */
1168 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1169 goto skip_linkparms;
1170 #endif
1172 if (in6_dev->if_flags & IF_RS_SENT) {
1174 * flag that an RA was received after an RS was sent
1175 * out on this interface.
1177 in6_dev->if_flags |= IF_RA_RCVD;
1181 * Remember the managed/otherconf flags from most recently
1182 * received RA message (RFC 2462) -- yoshfuji
1184 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1185 IF_RA_OTHERCONF)) |
1186 (ra_msg->icmph.icmp6_addrconf_managed ?
1187 IF_RA_MANAGED : 0) |
1188 (ra_msg->icmph.icmp6_addrconf_other ?
1189 IF_RA_OTHERCONF : 0);
1191 if (!in6_dev->cnf.accept_ra_defrtr)
1192 goto skip_defrtr;
1194 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1196 #ifdef CONFIG_IPV6_ROUTER_PREF
1197 pref = ra_msg->icmph.icmp6_router_pref;
1198 /* 10b is handled as if it were 00b (medium) */
1199 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1200 !in6_dev->cnf.accept_ra_rtr_pref)
1201 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1202 #endif
1204 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1206 if (rt)
1207 neigh = rt->rt6i_nexthop;
1209 if (rt && lifetime == 0) {
1210 neigh_clone(neigh);
1211 ip6_del_rt(rt);
1212 rt = NULL;
1215 if (rt == NULL && lifetime) {
1216 ND_PRINTK3(KERN_DEBUG
1217 "ICMPv6 RA: adding default router.\n");
1219 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1220 if (rt == NULL) {
1221 ND_PRINTK0(KERN_ERR
1222 "ICMPv6 RA: %s() failed to add default route.\n",
1223 __func__);
1224 in6_dev_put(in6_dev);
1225 return;
1228 neigh = rt->rt6i_nexthop;
1229 if (neigh == NULL) {
1230 ND_PRINTK0(KERN_ERR
1231 "ICMPv6 RA: %s() got default router without neighbour.\n",
1232 __func__);
1233 dst_release(&rt->u.dst);
1234 in6_dev_put(in6_dev);
1235 return;
1237 neigh->flags |= NTF_ROUTER;
1238 } else if (rt) {
1239 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1242 if (rt)
1243 rt->rt6i_expires = jiffies + (HZ * lifetime);
1245 if (ra_msg->icmph.icmp6_hop_limit) {
1246 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1247 if (rt)
1248 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1251 skip_defrtr:
1254 * Update Reachable Time and Retrans Timer
1257 if (in6_dev->nd_parms) {
1258 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1260 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1261 rtime = (rtime*HZ)/1000;
1262 if (rtime < HZ/10)
1263 rtime = HZ/10;
1264 in6_dev->nd_parms->retrans_time = rtime;
1265 in6_dev->tstamp = jiffies;
1266 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1269 rtime = ntohl(ra_msg->reachable_time);
1270 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1271 rtime = (rtime*HZ)/1000;
1273 if (rtime < HZ/10)
1274 rtime = HZ/10;
1276 if (rtime != in6_dev->nd_parms->base_reachable_time) {
1277 in6_dev->nd_parms->base_reachable_time = rtime;
1278 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1279 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1280 in6_dev->tstamp = jiffies;
1281 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1286 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1287 skip_linkparms:
1288 #endif
1291 * Process options.
1294 if (!neigh)
1295 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1296 skb->dev, 1);
1297 if (neigh) {
1298 u8 *lladdr = NULL;
1299 if (ndopts.nd_opts_src_lladdr) {
1300 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1301 skb->dev);
1302 if (!lladdr) {
1303 ND_PRINTK2(KERN_WARNING
1304 "ICMPv6 RA: invalid link-layer address length\n");
1305 goto out;
1308 neigh_update(neigh, lladdr, NUD_STALE,
1309 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1310 NEIGH_UPDATE_F_OVERRIDE|
1311 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1312 NEIGH_UPDATE_F_ISROUTER);
1315 #ifdef CONFIG_IPV6_ROUTE_INFO
1316 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1317 struct nd_opt_hdr *p;
1318 for (p = ndopts.nd_opts_ri;
1320 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1321 struct route_info *ri = (struct route_info *)p;
1322 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1323 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1324 ri->prefix_len == 0)
1325 continue;
1326 #endif
1327 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1328 continue;
1329 rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1330 &ipv6_hdr(skb)->saddr);
1333 #endif
1335 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1336 /* skip link-specific ndopts from interior routers */
1337 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1338 goto out;
1339 #endif
1341 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1342 struct nd_opt_hdr *p;
1343 for (p = ndopts.nd_opts_pi;
1345 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1346 addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1350 if (ndopts.nd_opts_mtu) {
1351 __be32 n;
1352 u32 mtu;
1354 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1355 mtu = ntohl(n);
1357 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1358 ND_PRINTK2(KERN_WARNING
1359 "ICMPv6 RA: invalid mtu: %d\n",
1360 mtu);
1361 } else if (in6_dev->cnf.mtu6 != mtu) {
1362 in6_dev->cnf.mtu6 = mtu;
1364 if (rt)
1365 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
1367 rt6_mtu_change(skb->dev, mtu);
1371 if (ndopts.nd_useropts) {
1372 struct nd_opt_hdr *p;
1373 for (p = ndopts.nd_useropts;
1375 p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1376 ndisc_ra_useropt(skb, p);
1380 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1381 ND_PRINTK2(KERN_WARNING
1382 "ICMPv6 RA: invalid RA options");
1384 out:
1385 if (rt)
1386 dst_release(&rt->u.dst);
1387 else if (neigh)
1388 neigh_release(neigh);
1389 in6_dev_put(in6_dev);
1392 static void ndisc_redirect_rcv(struct sk_buff *skb)
1394 struct inet6_dev *in6_dev;
1395 struct icmp6hdr *icmph;
1396 struct in6_addr *dest;
1397 struct in6_addr *target; /* new first hop to destination */
1398 struct neighbour *neigh;
1399 int on_link = 0;
1400 struct ndisc_options ndopts;
1401 int optlen;
1402 u8 *lladdr = NULL;
1404 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1405 switch (skb->ndisc_nodetype) {
1406 case NDISC_NODETYPE_HOST:
1407 case NDISC_NODETYPE_NODEFAULT:
1408 ND_PRINTK2(KERN_WARNING
1409 "ICMPv6 Redirect: from host or unauthorized router\n");
1410 return;
1412 #endif
1414 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1415 ND_PRINTK2(KERN_WARNING
1416 "ICMPv6 Redirect: source address is not link-local.\n");
1417 return;
1420 optlen = skb->tail - skb->transport_header;
1421 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1423 if (optlen < 0) {
1424 ND_PRINTK2(KERN_WARNING
1425 "ICMPv6 Redirect: packet too short\n");
1426 return;
1429 icmph = icmp6_hdr(skb);
1430 target = (struct in6_addr *) (icmph + 1);
1431 dest = target + 1;
1433 if (ipv6_addr_is_multicast(dest)) {
1434 ND_PRINTK2(KERN_WARNING
1435 "ICMPv6 Redirect: destination address is multicast.\n");
1436 return;
1439 if (ipv6_addr_equal(dest, target)) {
1440 on_link = 1;
1441 } else if (ipv6_addr_type(target) !=
1442 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1443 ND_PRINTK2(KERN_WARNING
1444 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1445 return;
1448 in6_dev = in6_dev_get(skb->dev);
1449 if (!in6_dev)
1450 return;
1451 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1452 in6_dev_put(in6_dev);
1453 return;
1456 /* RFC2461 8.1:
1457 * The IP source address of the Redirect MUST be the same as the current
1458 * first-hop router for the specified ICMP Destination Address.
1461 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1462 ND_PRINTK2(KERN_WARNING
1463 "ICMPv6 Redirect: invalid ND options\n");
1464 in6_dev_put(in6_dev);
1465 return;
1467 if (ndopts.nd_opts_tgt_lladdr) {
1468 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1469 skb->dev);
1470 if (!lladdr) {
1471 ND_PRINTK2(KERN_WARNING
1472 "ICMPv6 Redirect: invalid link-layer address length\n");
1473 in6_dev_put(in6_dev);
1474 return;
1478 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1479 if (neigh) {
1480 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1481 &ipv6_hdr(skb)->saddr, neigh, lladdr,
1482 on_link);
1483 neigh_release(neigh);
1485 in6_dev_put(in6_dev);
1488 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1489 const struct in6_addr *target)
1491 struct net_device *dev = skb->dev;
1492 struct net *net = dev_net(dev);
1493 struct sock *sk = net->ipv6.ndisc_sk;
1494 int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1495 struct sk_buff *buff;
1496 struct icmp6hdr *icmph;
1497 struct in6_addr saddr_buf;
1498 struct in6_addr *addrp;
1499 struct rt6_info *rt;
1500 struct dst_entry *dst;
1501 struct inet6_dev *idev;
1502 struct flowi fl;
1503 u8 *opt;
1504 int rd_len;
1505 int err;
1506 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1508 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1509 ND_PRINTK2(KERN_WARNING
1510 "ICMPv6 Redirect: no link-local address on %s\n",
1511 dev->name);
1512 return;
1515 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1516 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1517 ND_PRINTK2(KERN_WARNING
1518 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1519 return;
1522 icmpv6_flow_init(sk, &fl, NDISC_REDIRECT,
1523 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1525 dst = ip6_route_output(net, NULL, &fl);
1526 if (dst == NULL)
1527 return;
1529 err = xfrm_lookup(net, &dst, &fl, NULL, 0);
1530 if (err)
1531 return;
1533 rt = (struct rt6_info *) dst;
1535 if (rt->rt6i_flags & RTF_GATEWAY) {
1536 ND_PRINTK2(KERN_WARNING
1537 "ICMPv6 Redirect: destination is not a neighbour.\n");
1538 goto release;
1540 if (!xrlim_allow(dst, 1*HZ))
1541 goto release;
1543 if (dev->addr_len) {
1544 read_lock_bh(&neigh->lock);
1545 if (neigh->nud_state & NUD_VALID) {
1546 memcpy(ha_buf, neigh->ha, dev->addr_len);
1547 read_unlock_bh(&neigh->lock);
1548 ha = ha_buf;
1549 len += ndisc_opt_addr_space(dev);
1550 } else
1551 read_unlock_bh(&neigh->lock);
1554 rd_len = min_t(unsigned int,
1555 IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1556 rd_len &= ~0x7;
1557 len += rd_len;
1559 buff = sock_alloc_send_skb(sk,
1560 (MAX_HEADER + sizeof(struct ipv6hdr) +
1561 len + LL_ALLOCATED_SPACE(dev)),
1562 1, &err);
1563 if (buff == NULL) {
1564 ND_PRINTK0(KERN_ERR
1565 "ICMPv6 Redirect: %s() failed to allocate an skb, err=%d.\n",
1566 __func__, err);
1567 goto release;
1570 skb_reserve(buff, LL_RESERVED_SPACE(dev));
1571 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1572 IPPROTO_ICMPV6, len);
1574 skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1575 skb_put(buff, len);
1576 icmph = icmp6_hdr(buff);
1578 memset(icmph, 0, sizeof(struct icmp6hdr));
1579 icmph->icmp6_type = NDISC_REDIRECT;
1582 * copy target and destination addresses
1585 addrp = (struct in6_addr *)(icmph + 1);
1586 ipv6_addr_copy(addrp, target);
1587 addrp++;
1588 ipv6_addr_copy(addrp, &ipv6_hdr(skb)->daddr);
1590 opt = (u8*) (addrp + 1);
1593 * include target_address option
1596 if (ha)
1597 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1598 dev->addr_len, dev->type);
1601 * build redirect option and copy skb over to the new packet.
1604 memset(opt, 0, 8);
1605 *(opt++) = ND_OPT_REDIRECT_HDR;
1606 *(opt++) = (rd_len >> 3);
1607 opt += 6;
1609 memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1611 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1612 len, IPPROTO_ICMPV6,
1613 csum_partial(icmph, len, 0));
1615 skb_dst_set(buff, dst);
1616 idev = in6_dev_get(dst->dev);
1617 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1618 err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, buff, NULL, dst->dev,
1619 dst_output);
1620 if (!err) {
1621 ICMP6MSGOUT_INC_STATS(net, idev, NDISC_REDIRECT);
1622 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1625 if (likely(idev != NULL))
1626 in6_dev_put(idev);
1627 return;
1629 release:
1630 dst_release(dst);
1633 static void pndisc_redo(struct sk_buff *skb)
1635 ndisc_recv_ns(skb);
1636 kfree_skb(skb);
1639 int ndisc_rcv(struct sk_buff *skb)
1641 struct nd_msg *msg;
1643 if (!pskb_may_pull(skb, skb->len))
1644 return 0;
1646 msg = (struct nd_msg *)skb_transport_header(skb);
1648 __skb_push(skb, skb->data - skb_transport_header(skb));
1650 if (ipv6_hdr(skb)->hop_limit != 255) {
1651 ND_PRINTK2(KERN_WARNING
1652 "ICMPv6 NDISC: invalid hop-limit: %d\n",
1653 ipv6_hdr(skb)->hop_limit);
1654 return 0;
1657 if (msg->icmph.icmp6_code != 0) {
1658 ND_PRINTK2(KERN_WARNING
1659 "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1660 msg->icmph.icmp6_code);
1661 return 0;
1664 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1666 switch (msg->icmph.icmp6_type) {
1667 case NDISC_NEIGHBOUR_SOLICITATION:
1668 ndisc_recv_ns(skb);
1669 break;
1671 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1672 ndisc_recv_na(skb);
1673 break;
1675 case NDISC_ROUTER_SOLICITATION:
1676 ndisc_recv_rs(skb);
1677 break;
1679 case NDISC_ROUTER_ADVERTISEMENT:
1680 ndisc_router_discovery(skb);
1681 break;
1683 case NDISC_REDIRECT:
1684 ndisc_redirect_rcv(skb);
1685 break;
1688 return 0;
1691 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1693 struct net_device *dev = ptr;
1694 struct net *net = dev_net(dev);
1696 switch (event) {
1697 case NETDEV_CHANGEADDR:
1698 neigh_changeaddr(&nd_tbl, dev);
1699 fib6_run_gc(~0UL, net);
1700 break;
1701 case NETDEV_DOWN:
1702 neigh_ifdown(&nd_tbl, dev);
1703 fib6_run_gc(~0UL, net);
1704 break;
1705 default:
1706 break;
1709 return NOTIFY_DONE;
1712 static struct notifier_block ndisc_netdev_notifier = {
1713 .notifier_call = ndisc_netdev_event,
1716 #ifdef CONFIG_SYSCTL
1717 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1718 const char *func, const char *dev_name)
1720 static char warncomm[TASK_COMM_LEN];
1721 static int warned;
1722 if (strcmp(warncomm, current->comm) && warned < 5) {
1723 strcpy(warncomm, current->comm);
1724 printk(KERN_WARNING
1725 "process `%s' is using deprecated sysctl (%s) "
1726 "net.ipv6.neigh.%s.%s; "
1727 "Use net.ipv6.neigh.%s.%s_ms "
1728 "instead.\n",
1729 warncomm, func,
1730 dev_name, ctl->procname,
1731 dev_name, ctl->procname);
1732 warned++;
1736 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, struct file * filp, void __user *buffer, size_t *lenp, loff_t *ppos)
1738 struct net_device *dev = ctl->extra1;
1739 struct inet6_dev *idev;
1740 int ret;
1742 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1743 (strcmp(ctl->procname, "base_reachable_time") == 0))
1744 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1746 if (strcmp(ctl->procname, "retrans_time") == 0)
1747 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1749 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1750 ret = proc_dointvec_jiffies(ctl, write,
1751 filp, buffer, lenp, ppos);
1753 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1754 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1755 ret = proc_dointvec_ms_jiffies(ctl, write,
1756 filp, buffer, lenp, ppos);
1757 else
1758 ret = -1;
1760 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1761 if (ctl->data == &idev->nd_parms->base_reachable_time)
1762 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1763 idev->tstamp = jiffies;
1764 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1765 in6_dev_put(idev);
1767 return ret;
1770 int ndisc_ifinfo_sysctl_strategy(ctl_table *ctl,
1771 void __user *oldval, size_t __user *oldlenp,
1772 void __user *newval, size_t newlen)
1774 struct net_device *dev = ctl->extra1;
1775 struct inet6_dev *idev;
1776 int ret;
1778 if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1779 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1780 ndisc_warn_deprecated_sysctl(ctl, "procfs", dev ? dev->name : "default");
1782 switch (ctl->ctl_name) {
1783 case NET_NEIGH_REACHABLE_TIME:
1784 ret = sysctl_jiffies(ctl, oldval, oldlenp, newval, newlen);
1785 break;
1786 case NET_NEIGH_RETRANS_TIME_MS:
1787 case NET_NEIGH_REACHABLE_TIME_MS:
1788 ret = sysctl_ms_jiffies(ctl, oldval, oldlenp, newval, newlen);
1789 break;
1790 default:
1791 ret = 0;
1794 if (newval && newlen && ret > 0 &&
1795 dev && (idev = in6_dev_get(dev)) != NULL) {
1796 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1797 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1798 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1799 idev->tstamp = jiffies;
1800 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1801 in6_dev_put(idev);
1804 return ret;
1807 #endif
1809 static int ndisc_net_init(struct net *net)
1811 struct ipv6_pinfo *np;
1812 struct sock *sk;
1813 int err;
1815 err = inet_ctl_sock_create(&sk, PF_INET6,
1816 SOCK_RAW, IPPROTO_ICMPV6, net);
1817 if (err < 0) {
1818 ND_PRINTK0(KERN_ERR
1819 "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1820 err);
1821 return err;
1824 net->ipv6.ndisc_sk = sk;
1826 np = inet6_sk(sk);
1827 np->hop_limit = 255;
1828 /* Do not loopback ndisc messages */
1829 np->mc_loop = 0;
1831 return 0;
1834 static void ndisc_net_exit(struct net *net)
1836 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1839 static struct pernet_operations ndisc_net_ops = {
1840 .init = ndisc_net_init,
1841 .exit = ndisc_net_exit,
1844 int __init ndisc_init(void)
1846 int err;
1848 err = register_pernet_subsys(&ndisc_net_ops);
1849 if (err)
1850 return err;
1852 * Initialize the neighbour table
1854 neigh_table_init(&nd_tbl);
1856 #ifdef CONFIG_SYSCTL
1857 err = neigh_sysctl_register(NULL, &nd_tbl.parms, NET_IPV6,
1858 NET_IPV6_NEIGH, "ipv6",
1859 &ndisc_ifinfo_sysctl_change,
1860 &ndisc_ifinfo_sysctl_strategy);
1861 if (err)
1862 goto out_unregister_pernet;
1863 #endif
1864 err = register_netdevice_notifier(&ndisc_netdev_notifier);
1865 if (err)
1866 goto out_unregister_sysctl;
1867 out:
1868 return err;
1870 out_unregister_sysctl:
1871 #ifdef CONFIG_SYSCTL
1872 neigh_sysctl_unregister(&nd_tbl.parms);
1873 out_unregister_pernet:
1874 #endif
1875 unregister_pernet_subsys(&ndisc_net_ops);
1876 goto out;
1879 void ndisc_cleanup(void)
1881 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1882 #ifdef CONFIG_SYSCTL
1883 neigh_sysctl_unregister(&nd_tbl.parms);
1884 #endif
1885 neigh_table_clear(&nd_tbl);
1886 unregister_pernet_subsys(&ndisc_net_ops);