Merge branch 'misc' into test
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / ndisc.c
blob998d6d27e7cf293383b7c4e944899790170ba690
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 #include <linux/slab.h>
63 #ifdef CONFIG_SYSCTL
64 #include <linux/sysctl.h>
65 #endif
67 #include <linux/if_addr.h>
68 #include <linux/if_arp.h>
69 #include <linux/ipv6.h>
70 #include <linux/icmpv6.h>
71 #include <linux/jhash.h>
73 #include <net/sock.h>
74 #include <net/snmp.h>
76 #include <net/ipv6.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
81 #include <net/icmp.h>
83 #include <net/netlink.h>
84 #include <linux/rtnetlink.h>
86 #include <net/flow.h>
87 #include <net/ip6_checksum.h>
88 #include <net/inet_common.h>
89 #include <linux/proc_fs.h>
91 #include <linux/netfilter.h>
92 #include <linux/netfilter_ipv6.h>
94 static u32 ndisc_hash(const void *pkey,
95 const struct net_device *dev,
96 __u32 rnd);
97 static int ndisc_constructor(struct neighbour *neigh);
98 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
99 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
100 static int pndisc_constructor(struct pneigh_entry *n);
101 static void pndisc_destructor(struct pneigh_entry *n);
102 static void pndisc_redo(struct sk_buff *skb);
104 static const struct neigh_ops ndisc_generic_ops = {
105 .family = AF_INET6,
106 .solicit = ndisc_solicit,
107 .error_report = ndisc_error_report,
108 .output = neigh_resolve_output,
109 .connected_output = neigh_connected_output,
110 .hh_output = dev_queue_xmit,
111 .queue_xmit = dev_queue_xmit,
114 static const struct neigh_ops ndisc_hh_ops = {
115 .family = AF_INET6,
116 .solicit = ndisc_solicit,
117 .error_report = ndisc_error_report,
118 .output = neigh_resolve_output,
119 .connected_output = neigh_resolve_output,
120 .hh_output = dev_queue_xmit,
121 .queue_xmit = dev_queue_xmit,
125 static const struct neigh_ops ndisc_direct_ops = {
126 .family = AF_INET6,
127 .output = dev_queue_xmit,
128 .connected_output = dev_queue_xmit,
129 .hh_output = dev_queue_xmit,
130 .queue_xmit = dev_queue_xmit,
133 struct neigh_table nd_tbl = {
134 .family = AF_INET6,
135 .entry_size = sizeof(struct neighbour) + sizeof(struct in6_addr),
136 .key_len = sizeof(struct in6_addr),
137 .hash = ndisc_hash,
138 .constructor = ndisc_constructor,
139 .pconstructor = pndisc_constructor,
140 .pdestructor = pndisc_destructor,
141 .proxy_redo = pndisc_redo,
142 .id = "ndisc_cache",
143 .parms = {
144 .tbl = &nd_tbl,
145 .base_reachable_time = 30 * HZ,
146 .retrans_time = 1 * HZ,
147 .gc_staletime = 60 * HZ,
148 .reachable_time = 30 * HZ,
149 .delay_probe_time = 5 * HZ,
150 .queue_len = 3,
151 .ucast_probes = 3,
152 .mcast_probes = 3,
153 .anycast_delay = 1 * HZ,
154 .proxy_delay = (8 * HZ) / 10,
155 .proxy_qlen = 64,
157 .gc_interval = 30 * HZ,
158 .gc_thresh1 = 128,
159 .gc_thresh2 = 512,
160 .gc_thresh3 = 1024,
163 /* ND options */
164 struct ndisc_options {
165 struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
166 #ifdef CONFIG_IPV6_ROUTE_INFO
167 struct nd_opt_hdr *nd_opts_ri;
168 struct nd_opt_hdr *nd_opts_ri_end;
169 #endif
170 struct nd_opt_hdr *nd_useropts;
171 struct nd_opt_hdr *nd_useropts_end;
174 #define nd_opts_src_lladdr nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
175 #define nd_opts_tgt_lladdr nd_opt_array[ND_OPT_TARGET_LL_ADDR]
176 #define nd_opts_pi nd_opt_array[ND_OPT_PREFIX_INFO]
177 #define nd_opts_pi_end nd_opt_array[__ND_OPT_PREFIX_INFO_END]
178 #define nd_opts_rh nd_opt_array[ND_OPT_REDIRECT_HDR]
179 #define nd_opts_mtu nd_opt_array[ND_OPT_MTU]
181 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
184 * Return the padding between the option length and the start of the
185 * link addr. Currently only IP-over-InfiniBand needs this, although
186 * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
187 * also need a pad of 2.
189 static int ndisc_addr_option_pad(unsigned short type)
191 switch (type) {
192 case ARPHRD_INFINIBAND: return 2;
193 default: return 0;
197 static inline int ndisc_opt_addr_space(struct net_device *dev)
199 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
202 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
203 unsigned short addr_type)
205 int space = NDISC_OPT_SPACE(data_len);
206 int pad = ndisc_addr_option_pad(addr_type);
208 opt[0] = type;
209 opt[1] = space>>3;
211 memset(opt + 2, 0, pad);
212 opt += pad;
213 space -= pad;
215 memcpy(opt+2, data, data_len);
216 data_len += 2;
217 opt += data_len;
218 if ((space -= data_len) > 0)
219 memset(opt, 0, space);
220 return opt + space;
223 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
224 struct nd_opt_hdr *end)
226 int type;
227 if (!cur || !end || cur >= end)
228 return NULL;
229 type = cur->nd_opt_type;
230 do {
231 cur = ((void *)cur) + (cur->nd_opt_len << 3);
232 } while(cur < end && cur->nd_opt_type != type);
233 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
236 static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
238 return opt->nd_opt_type == ND_OPT_RDNSS;
241 static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
242 struct nd_opt_hdr *end)
244 if (!cur || !end || cur >= end)
245 return NULL;
246 do {
247 cur = ((void *)cur) + (cur->nd_opt_len << 3);
248 } while(cur < end && !ndisc_is_useropt(cur));
249 return cur <= end && ndisc_is_useropt(cur) ? cur : NULL;
252 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
253 struct ndisc_options *ndopts)
255 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
257 if (!nd_opt || opt_len < 0 || !ndopts)
258 return NULL;
259 memset(ndopts, 0, sizeof(*ndopts));
260 while (opt_len) {
261 int l;
262 if (opt_len < sizeof(struct nd_opt_hdr))
263 return NULL;
264 l = nd_opt->nd_opt_len << 3;
265 if (opt_len < l || l == 0)
266 return NULL;
267 switch (nd_opt->nd_opt_type) {
268 case ND_OPT_SOURCE_LL_ADDR:
269 case ND_OPT_TARGET_LL_ADDR:
270 case ND_OPT_MTU:
271 case ND_OPT_REDIRECT_HDR:
272 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
273 ND_PRINTK2(KERN_WARNING
274 "%s(): duplicated ND6 option found: type=%d\n",
275 __func__,
276 nd_opt->nd_opt_type);
277 } else {
278 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
280 break;
281 case ND_OPT_PREFIX_INFO:
282 ndopts->nd_opts_pi_end = nd_opt;
283 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
284 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
285 break;
286 #ifdef CONFIG_IPV6_ROUTE_INFO
287 case ND_OPT_ROUTE_INFO:
288 ndopts->nd_opts_ri_end = nd_opt;
289 if (!ndopts->nd_opts_ri)
290 ndopts->nd_opts_ri = nd_opt;
291 break;
292 #endif
293 default:
294 if (ndisc_is_useropt(nd_opt)) {
295 ndopts->nd_useropts_end = nd_opt;
296 if (!ndopts->nd_useropts)
297 ndopts->nd_useropts = nd_opt;
298 } else {
300 * Unknown options must be silently ignored,
301 * to accommodate future extension to the
302 * protocol.
304 ND_PRINTK2(KERN_NOTICE
305 "%s(): ignored unsupported option; type=%d, len=%d\n",
306 __func__,
307 nd_opt->nd_opt_type, nd_opt->nd_opt_len);
310 opt_len -= l;
311 nd_opt = ((void *)nd_opt) + l;
313 return ndopts;
316 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
317 struct net_device *dev)
319 u8 *lladdr = (u8 *)(p + 1);
320 int lladdrlen = p->nd_opt_len << 3;
321 int prepad = ndisc_addr_option_pad(dev->type);
322 if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
323 return NULL;
324 return lladdr + prepad;
327 int ndisc_mc_map(struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
329 switch (dev->type) {
330 case ARPHRD_ETHER:
331 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
332 case ARPHRD_FDDI:
333 ipv6_eth_mc_map(addr, buf);
334 return 0;
335 case ARPHRD_IEEE802_TR:
336 ipv6_tr_mc_map(addr,buf);
337 return 0;
338 case ARPHRD_ARCNET:
339 ipv6_arcnet_mc_map(addr, buf);
340 return 0;
341 case ARPHRD_INFINIBAND:
342 ipv6_ib_mc_map(addr, dev->broadcast, buf);
343 return 0;
344 default:
345 if (dir) {
346 memcpy(buf, dev->broadcast, dev->addr_len);
347 return 0;
350 return -EINVAL;
353 EXPORT_SYMBOL(ndisc_mc_map);
355 static u32 ndisc_hash(const void *pkey,
356 const struct net_device *dev,
357 __u32 hash_rnd)
359 const u32 *p32 = pkey;
360 u32 addr_hash, i;
362 addr_hash = 0;
363 for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
364 addr_hash ^= *p32++;
366 return jhash_2words(addr_hash, dev->ifindex, hash_rnd);
369 static int ndisc_constructor(struct neighbour *neigh)
371 struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
372 struct net_device *dev = neigh->dev;
373 struct inet6_dev *in6_dev;
374 struct neigh_parms *parms;
375 int is_multicast = ipv6_addr_is_multicast(addr);
377 rcu_read_lock();
378 in6_dev = in6_dev_get(dev);
379 if (in6_dev == NULL) {
380 rcu_read_unlock();
381 return -EINVAL;
384 parms = in6_dev->nd_parms;
385 __neigh_parms_put(neigh->parms);
386 neigh->parms = neigh_parms_clone(parms);
387 rcu_read_unlock();
389 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
390 if (!dev->header_ops) {
391 neigh->nud_state = NUD_NOARP;
392 neigh->ops = &ndisc_direct_ops;
393 neigh->output = neigh->ops->queue_xmit;
394 } else {
395 if (is_multicast) {
396 neigh->nud_state = NUD_NOARP;
397 ndisc_mc_map(addr, neigh->ha, dev, 1);
398 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
399 neigh->nud_state = NUD_NOARP;
400 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
401 if (dev->flags&IFF_LOOPBACK)
402 neigh->type = RTN_LOCAL;
403 } else if (dev->flags&IFF_POINTOPOINT) {
404 neigh->nud_state = NUD_NOARP;
405 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
407 if (dev->header_ops->cache)
408 neigh->ops = &ndisc_hh_ops;
409 else
410 neigh->ops = &ndisc_generic_ops;
411 if (neigh->nud_state&NUD_VALID)
412 neigh->output = neigh->ops->connected_output;
413 else
414 neigh->output = neigh->ops->output;
416 in6_dev_put(in6_dev);
417 return 0;
420 static int pndisc_constructor(struct pneigh_entry *n)
422 struct in6_addr *addr = (struct in6_addr*)&n->key;
423 struct in6_addr maddr;
424 struct net_device *dev = n->dev;
426 if (dev == NULL || __in6_dev_get(dev) == NULL)
427 return -EINVAL;
428 addrconf_addr_solict_mult(addr, &maddr);
429 ipv6_dev_mc_inc(dev, &maddr);
430 return 0;
433 static void pndisc_destructor(struct pneigh_entry *n)
435 struct in6_addr *addr = (struct in6_addr*)&n->key;
436 struct in6_addr maddr;
437 struct net_device *dev = n->dev;
439 if (dev == NULL || __in6_dev_get(dev) == NULL)
440 return;
441 addrconf_addr_solict_mult(addr, &maddr);
442 ipv6_dev_mc_dec(dev, &maddr);
445 struct sk_buff *ndisc_build_skb(struct net_device *dev,
446 const struct in6_addr *daddr,
447 const struct in6_addr *saddr,
448 struct icmp6hdr *icmp6h,
449 const struct in6_addr *target,
450 int llinfo)
452 struct net *net = dev_net(dev);
453 struct sock *sk = net->ipv6.ndisc_sk;
454 struct sk_buff *skb;
455 struct icmp6hdr *hdr;
456 int len;
457 int err;
458 u8 *opt;
460 if (!dev->addr_len)
461 llinfo = 0;
463 len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
464 if (llinfo)
465 len += ndisc_opt_addr_space(dev);
467 skb = sock_alloc_send_skb(sk,
468 (MAX_HEADER + sizeof(struct ipv6hdr) +
469 len + LL_ALLOCATED_SPACE(dev)),
470 1, &err);
471 if (!skb) {
472 ND_PRINTK0(KERN_ERR
473 "ICMPv6 ND: %s() failed to allocate an skb, err=%d.\n",
474 __func__, err);
475 return NULL;
478 skb_reserve(skb, LL_RESERVED_SPACE(dev));
479 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
481 skb->transport_header = skb->tail;
482 skb_put(skb, len);
484 hdr = (struct icmp6hdr *)skb_transport_header(skb);
485 memcpy(hdr, icmp6h, sizeof(*hdr));
487 opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
488 if (target) {
489 ipv6_addr_copy((struct in6_addr *)opt, target);
490 opt += sizeof(*target);
493 if (llinfo)
494 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
495 dev->addr_len, dev->type);
497 hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
498 IPPROTO_ICMPV6,
499 csum_partial(hdr,
500 len, 0));
502 return skb;
505 EXPORT_SYMBOL(ndisc_build_skb);
507 void ndisc_send_skb(struct sk_buff *skb,
508 struct net_device *dev,
509 struct neighbour *neigh,
510 const struct in6_addr *daddr,
511 const struct in6_addr *saddr,
512 struct icmp6hdr *icmp6h)
514 struct flowi fl;
515 struct dst_entry *dst;
516 struct net *net = dev_net(dev);
517 struct sock *sk = net->ipv6.ndisc_sk;
518 struct inet6_dev *idev;
519 int err;
520 u8 type;
522 type = icmp6h->icmp6_type;
524 icmpv6_flow_init(sk, &fl, type, saddr, daddr, dev->ifindex);
526 dst = icmp6_dst_alloc(dev, neigh, daddr);
527 if (!dst) {
528 kfree_skb(skb);
529 return;
532 err = xfrm_lookup(net, &dst, &fl, NULL, 0);
533 if (err < 0) {
534 kfree_skb(skb);
535 return;
538 skb_dst_set(skb, dst);
540 idev = in6_dev_get(dst->dev);
541 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
543 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
544 dst_output);
545 if (!err) {
546 ICMP6MSGOUT_INC_STATS(net, idev, type);
547 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
550 if (likely(idev != NULL))
551 in6_dev_put(idev);
554 EXPORT_SYMBOL(ndisc_send_skb);
557 * Send a Neighbour Discover packet
559 static void __ndisc_send(struct net_device *dev,
560 struct neighbour *neigh,
561 const struct in6_addr *daddr,
562 const struct in6_addr *saddr,
563 struct icmp6hdr *icmp6h, const struct in6_addr *target,
564 int llinfo)
566 struct sk_buff *skb;
568 skb = ndisc_build_skb(dev, daddr, saddr, icmp6h, target, llinfo);
569 if (!skb)
570 return;
572 ndisc_send_skb(skb, dev, neigh, daddr, saddr, icmp6h);
575 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
576 const struct in6_addr *daddr,
577 const struct in6_addr *solicited_addr,
578 int router, int solicited, int override, int inc_opt)
580 struct in6_addr tmpaddr;
581 struct inet6_ifaddr *ifp;
582 const struct in6_addr *src_addr;
583 struct icmp6hdr icmp6h = {
584 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
587 /* for anycast or proxy, solicited_addr != src_addr */
588 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
589 if (ifp) {
590 src_addr = solicited_addr;
591 if (ifp->flags & IFA_F_OPTIMISTIC)
592 override = 0;
593 inc_opt |= ifp->idev->cnf.force_tllao;
594 in6_ifa_put(ifp);
595 } else {
596 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
597 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
598 &tmpaddr))
599 return;
600 src_addr = &tmpaddr;
603 icmp6h.icmp6_router = router;
604 icmp6h.icmp6_solicited = solicited;
605 icmp6h.icmp6_override = override;
607 __ndisc_send(dev, neigh, daddr, src_addr,
608 &icmp6h, solicited_addr,
609 inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
612 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
613 const struct in6_addr *solicit,
614 const struct in6_addr *daddr, const struct in6_addr *saddr)
616 struct in6_addr addr_buf;
617 struct icmp6hdr icmp6h = {
618 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
621 if (saddr == NULL) {
622 if (ipv6_get_lladdr(dev, &addr_buf,
623 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
624 return;
625 saddr = &addr_buf;
628 __ndisc_send(dev, neigh, daddr, saddr,
629 &icmp6h, solicit,
630 !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
633 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
634 const struct in6_addr *daddr)
636 struct icmp6hdr icmp6h = {
637 .icmp6_type = NDISC_ROUTER_SOLICITATION,
639 int send_sllao = dev->addr_len;
641 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
643 * According to section 2.2 of RFC 4429, we must not
644 * send router solicitations with a sllao from
645 * optimistic addresses, but we may send the solicitation
646 * if we don't include the sllao. So here we check
647 * if our address is optimistic, and if so, we
648 * suppress the inclusion of the sllao.
650 if (send_sllao) {
651 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
652 dev, 1);
653 if (ifp) {
654 if (ifp->flags & IFA_F_OPTIMISTIC) {
655 send_sllao = 0;
657 in6_ifa_put(ifp);
658 } else {
659 send_sllao = 0;
662 #endif
663 __ndisc_send(dev, NULL, daddr, saddr,
664 &icmp6h, NULL,
665 send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
669 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
672 * "The sender MUST return an ICMP
673 * destination unreachable"
675 dst_link_failure(skb);
676 kfree_skb(skb);
679 /* Called with locked neigh: either read or both */
681 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
683 struct in6_addr *saddr = NULL;
684 struct in6_addr mcaddr;
685 struct net_device *dev = neigh->dev;
686 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
687 int probes = atomic_read(&neigh->probes);
689 if (skb && ipv6_chk_addr(dev_net(dev), &ipv6_hdr(skb)->saddr, dev, 1))
690 saddr = &ipv6_hdr(skb)->saddr;
692 if ((probes -= neigh->parms->ucast_probes) < 0) {
693 if (!(neigh->nud_state & NUD_VALID)) {
694 ND_PRINTK1(KERN_DEBUG "%s(): trying to ucast probe in NUD_INVALID: %pI6\n",
695 __func__, target);
697 ndisc_send_ns(dev, neigh, target, target, saddr);
698 } else if ((probes -= neigh->parms->app_probes) < 0) {
699 #ifdef CONFIG_ARPD
700 neigh_app_ns(neigh);
701 #endif
702 } else {
703 addrconf_addr_solict_mult(target, &mcaddr);
704 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
708 static int pndisc_is_router(const void *pkey,
709 struct net_device *dev)
711 struct pneigh_entry *n;
712 int ret = -1;
714 read_lock_bh(&nd_tbl.lock);
715 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
716 if (n)
717 ret = !!(n->flags & NTF_ROUTER);
718 read_unlock_bh(&nd_tbl.lock);
720 return ret;
723 static void ndisc_recv_ns(struct sk_buff *skb)
725 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
726 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
727 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
728 u8 *lladdr = NULL;
729 u32 ndoptlen = skb->tail - (skb->transport_header +
730 offsetof(struct nd_msg, opt));
731 struct ndisc_options ndopts;
732 struct net_device *dev = skb->dev;
733 struct inet6_ifaddr *ifp;
734 struct inet6_dev *idev = NULL;
735 struct neighbour *neigh;
736 int dad = ipv6_addr_any(saddr);
737 int inc;
738 int is_router = -1;
740 if (ipv6_addr_is_multicast(&msg->target)) {
741 ND_PRINTK2(KERN_WARNING
742 "ICMPv6 NS: multicast target address");
743 return;
747 * RFC2461 7.1.1:
748 * DAD has to be destined for solicited node multicast address.
750 if (dad &&
751 !(daddr->s6_addr32[0] == htonl(0xff020000) &&
752 daddr->s6_addr32[1] == htonl(0x00000000) &&
753 daddr->s6_addr32[2] == htonl(0x00000001) &&
754 daddr->s6_addr [12] == 0xff )) {
755 ND_PRINTK2(KERN_WARNING
756 "ICMPv6 NS: bad DAD packet (wrong destination)\n");
757 return;
760 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
761 ND_PRINTK2(KERN_WARNING
762 "ICMPv6 NS: invalid ND options\n");
763 return;
766 if (ndopts.nd_opts_src_lladdr) {
767 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
768 if (!lladdr) {
769 ND_PRINTK2(KERN_WARNING
770 "ICMPv6 NS: invalid link-layer address length\n");
771 return;
774 /* RFC2461 7.1.1:
775 * If the IP source address is the unspecified address,
776 * there MUST NOT be source link-layer address option
777 * in the message.
779 if (dad) {
780 ND_PRINTK2(KERN_WARNING
781 "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
782 return;
786 inc = ipv6_addr_is_multicast(daddr);
788 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
789 if (ifp) {
791 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
792 if (dad) {
793 if (dev->type == ARPHRD_IEEE802_TR) {
794 const unsigned char *sadr;
795 sadr = skb_mac_header(skb);
796 if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
797 sadr[9] == dev->dev_addr[1] &&
798 sadr[10] == dev->dev_addr[2] &&
799 sadr[11] == dev->dev_addr[3] &&
800 sadr[12] == dev->dev_addr[4] &&
801 sadr[13] == dev->dev_addr[5]) {
802 /* looped-back to us */
803 goto out;
808 * We are colliding with another node
809 * who is doing DAD
810 * so fail our DAD process
812 addrconf_dad_failure(ifp);
813 return;
814 } else {
816 * This is not a dad solicitation.
817 * If we are an optimistic node,
818 * we should respond.
819 * Otherwise, we should ignore it.
821 if (!(ifp->flags & IFA_F_OPTIMISTIC))
822 goto out;
826 idev = ifp->idev;
827 } else {
828 struct net *net = dev_net(dev);
830 idev = in6_dev_get(dev);
831 if (!idev) {
832 /* XXX: count this drop? */
833 return;
836 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
837 (idev->cnf.forwarding &&
838 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
839 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
840 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
841 skb->pkt_type != PACKET_HOST &&
842 inc != 0 &&
843 idev->nd_parms->proxy_delay != 0) {
845 * for anycast or proxy,
846 * sender should delay its response
847 * by a random time between 0 and
848 * MAX_ANYCAST_DELAY_TIME seconds.
849 * (RFC2461) -- yoshfuji
851 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
852 if (n)
853 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
854 goto out;
856 } else
857 goto out;
860 if (is_router < 0)
861 is_router = !!idev->cnf.forwarding;
863 if (dad) {
864 ndisc_send_na(dev, NULL, &in6addr_linklocal_allnodes, &msg->target,
865 is_router, 0, (ifp != NULL), 1);
866 goto out;
869 if (inc)
870 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
871 else
872 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
875 * update / create cache entry
876 * for the source address
878 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
879 !inc || lladdr || !dev->addr_len);
880 if (neigh)
881 neigh_update(neigh, lladdr, NUD_STALE,
882 NEIGH_UPDATE_F_WEAK_OVERRIDE|
883 NEIGH_UPDATE_F_OVERRIDE);
884 if (neigh || !dev->header_ops) {
885 ndisc_send_na(dev, neigh, saddr, &msg->target,
886 is_router,
887 1, (ifp != NULL && inc), inc);
888 if (neigh)
889 neigh_release(neigh);
892 out:
893 if (ifp)
894 in6_ifa_put(ifp);
895 else
896 in6_dev_put(idev);
899 static void ndisc_recv_na(struct sk_buff *skb)
901 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
902 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
903 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
904 u8 *lladdr = NULL;
905 u32 ndoptlen = skb->tail - (skb->transport_header +
906 offsetof(struct nd_msg, opt));
907 struct ndisc_options ndopts;
908 struct net_device *dev = skb->dev;
909 struct inet6_ifaddr *ifp;
910 struct neighbour *neigh;
912 if (skb->len < sizeof(struct nd_msg)) {
913 ND_PRINTK2(KERN_WARNING
914 "ICMPv6 NA: packet too short\n");
915 return;
918 if (ipv6_addr_is_multicast(&msg->target)) {
919 ND_PRINTK2(KERN_WARNING
920 "ICMPv6 NA: target address is multicast.\n");
921 return;
924 if (ipv6_addr_is_multicast(daddr) &&
925 msg->icmph.icmp6_solicited) {
926 ND_PRINTK2(KERN_WARNING
927 "ICMPv6 NA: solicited NA is multicasted.\n");
928 return;
931 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
932 ND_PRINTK2(KERN_WARNING
933 "ICMPv6 NS: invalid ND option\n");
934 return;
936 if (ndopts.nd_opts_tgt_lladdr) {
937 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
938 if (!lladdr) {
939 ND_PRINTK2(KERN_WARNING
940 "ICMPv6 NA: invalid link-layer address length\n");
941 return;
944 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
945 if (ifp) {
946 if (ifp->flags & IFA_F_TENTATIVE) {
947 addrconf_dad_failure(ifp);
948 return;
950 /* What should we make now? The advertisement
951 is invalid, but ndisc specs say nothing
952 about it. It could be misconfiguration, or
953 an smart proxy agent tries to help us :-)
955 We should not print the error if NA has been
956 received from loopback - it is just our own
957 unsolicited advertisement.
959 if (skb->pkt_type != PACKET_LOOPBACK)
960 ND_PRINTK1(KERN_WARNING
961 "ICMPv6 NA: someone advertises our address %pI6 on %s!\n",
962 &ifp->addr, ifp->idev->dev->name);
963 in6_ifa_put(ifp);
964 return;
966 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
968 if (neigh) {
969 u8 old_flags = neigh->flags;
970 struct net *net = dev_net(dev);
972 if (neigh->nud_state & NUD_FAILED)
973 goto out;
976 * Don't update the neighbor cache entry on a proxy NA from
977 * ourselves because either the proxied node is off link or it
978 * has already sent a NA to us.
980 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
981 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
982 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
983 /* XXX: idev->cnf.prixy_ndp */
984 goto out;
987 neigh_update(neigh, lladdr,
988 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
989 NEIGH_UPDATE_F_WEAK_OVERRIDE|
990 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
991 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
992 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
994 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
996 * Change: router to host
998 struct rt6_info *rt;
999 rt = rt6_get_dflt_router(saddr, dev);
1000 if (rt)
1001 ip6_del_rt(rt);
1004 out:
1005 neigh_release(neigh);
1009 static void ndisc_recv_rs(struct sk_buff *skb)
1011 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1012 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1013 struct neighbour *neigh;
1014 struct inet6_dev *idev;
1015 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1016 struct ndisc_options ndopts;
1017 u8 *lladdr = NULL;
1019 if (skb->len < sizeof(*rs_msg))
1020 return;
1022 idev = in6_dev_get(skb->dev);
1023 if (!idev) {
1024 if (net_ratelimit())
1025 ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
1026 return;
1029 /* Don't accept RS if we're not in router mode */
1030 if (!idev->cnf.forwarding)
1031 goto out;
1034 * Don't update NCE if src = ::;
1035 * this implies that the source node has no ip address assigned yet.
1037 if (ipv6_addr_any(saddr))
1038 goto out;
1040 /* Parse ND options */
1041 if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1042 if (net_ratelimit())
1043 ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1044 goto out;
1047 if (ndopts.nd_opts_src_lladdr) {
1048 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1049 skb->dev);
1050 if (!lladdr)
1051 goto out;
1054 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1055 if (neigh) {
1056 neigh_update(neigh, lladdr, NUD_STALE,
1057 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1058 NEIGH_UPDATE_F_OVERRIDE|
1059 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1060 neigh_release(neigh);
1062 out:
1063 in6_dev_put(idev);
1066 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1068 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1069 struct sk_buff *skb;
1070 struct nlmsghdr *nlh;
1071 struct nduseroptmsg *ndmsg;
1072 struct net *net = dev_net(ra->dev);
1073 int err;
1074 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1075 + (opt->nd_opt_len << 3));
1076 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1078 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1079 if (skb == NULL) {
1080 err = -ENOBUFS;
1081 goto errout;
1084 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1085 if (nlh == NULL) {
1086 goto nla_put_failure;
1089 ndmsg = nlmsg_data(nlh);
1090 ndmsg->nduseropt_family = AF_INET6;
1091 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1092 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1093 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1094 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1096 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1098 NLA_PUT(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
1099 &ipv6_hdr(ra)->saddr);
1100 nlmsg_end(skb, nlh);
1102 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1103 return;
1105 nla_put_failure:
1106 nlmsg_free(skb);
1107 err = -EMSGSIZE;
1108 errout:
1109 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1112 static inline int accept_ra(struct inet6_dev *in6_dev)
1115 * If forwarding is enabled, RA are not accepted unless the special
1116 * hybrid mode (accept_ra=2) is enabled.
1118 if (in6_dev->cnf.forwarding && in6_dev->cnf.accept_ra < 2)
1119 return 0;
1121 return in6_dev->cnf.accept_ra;
1124 static void ndisc_router_discovery(struct sk_buff *skb)
1126 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1127 struct neighbour *neigh = NULL;
1128 struct inet6_dev *in6_dev;
1129 struct rt6_info *rt = NULL;
1130 int lifetime;
1131 struct ndisc_options ndopts;
1132 int optlen;
1133 unsigned int pref = 0;
1135 __u8 * opt = (__u8 *)(ra_msg + 1);
1137 optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1139 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1140 ND_PRINTK2(KERN_WARNING
1141 "ICMPv6 RA: source address is not link-local.\n");
1142 return;
1144 if (optlen < 0) {
1145 ND_PRINTK2(KERN_WARNING
1146 "ICMPv6 RA: packet too short\n");
1147 return;
1150 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1151 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1152 ND_PRINTK2(KERN_WARNING
1153 "ICMPv6 RA: from host or unauthorized router\n");
1154 return;
1156 #endif
1159 * set the RA_RECV flag in the interface
1162 in6_dev = in6_dev_get(skb->dev);
1163 if (in6_dev == NULL) {
1164 ND_PRINTK0(KERN_ERR
1165 "ICMPv6 RA: can't find inet6 device for %s.\n",
1166 skb->dev->name);
1167 return;
1170 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1171 in6_dev_put(in6_dev);
1172 ND_PRINTK2(KERN_WARNING
1173 "ICMP6 RA: invalid ND options\n");
1174 return;
1177 if (!accept_ra(in6_dev))
1178 goto skip_linkparms;
1180 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1181 /* skip link-specific parameters from interior routers */
1182 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1183 goto skip_linkparms;
1184 #endif
1186 if (in6_dev->if_flags & IF_RS_SENT) {
1188 * flag that an RA was received after an RS was sent
1189 * out on this interface.
1191 in6_dev->if_flags |= IF_RA_RCVD;
1195 * Remember the managed/otherconf flags from most recently
1196 * received RA message (RFC 2462) -- yoshfuji
1198 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1199 IF_RA_OTHERCONF)) |
1200 (ra_msg->icmph.icmp6_addrconf_managed ?
1201 IF_RA_MANAGED : 0) |
1202 (ra_msg->icmph.icmp6_addrconf_other ?
1203 IF_RA_OTHERCONF : 0);
1205 if (!in6_dev->cnf.accept_ra_defrtr)
1206 goto skip_defrtr;
1208 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1210 #ifdef CONFIG_IPV6_ROUTER_PREF
1211 pref = ra_msg->icmph.icmp6_router_pref;
1212 /* 10b is handled as if it were 00b (medium) */
1213 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1214 !in6_dev->cnf.accept_ra_rtr_pref)
1215 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1216 #endif
1218 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1220 if (rt)
1221 neigh = rt->rt6i_nexthop;
1223 if (rt && lifetime == 0) {
1224 neigh_clone(neigh);
1225 ip6_del_rt(rt);
1226 rt = NULL;
1229 if (rt == NULL && lifetime) {
1230 ND_PRINTK3(KERN_DEBUG
1231 "ICMPv6 RA: adding default router.\n");
1233 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1234 if (rt == NULL) {
1235 ND_PRINTK0(KERN_ERR
1236 "ICMPv6 RA: %s() failed to add default route.\n",
1237 __func__);
1238 in6_dev_put(in6_dev);
1239 return;
1242 neigh = rt->rt6i_nexthop;
1243 if (neigh == NULL) {
1244 ND_PRINTK0(KERN_ERR
1245 "ICMPv6 RA: %s() got default router without neighbour.\n",
1246 __func__);
1247 dst_release(&rt->dst);
1248 in6_dev_put(in6_dev);
1249 return;
1251 neigh->flags |= NTF_ROUTER;
1252 } else if (rt) {
1253 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1256 if (rt)
1257 rt->rt6i_expires = jiffies + (HZ * lifetime);
1259 if (ra_msg->icmph.icmp6_hop_limit) {
1260 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1261 if (rt)
1262 rt->dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1265 skip_defrtr:
1268 * Update Reachable Time and Retrans Timer
1271 if (in6_dev->nd_parms) {
1272 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1274 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1275 rtime = (rtime*HZ)/1000;
1276 if (rtime < HZ/10)
1277 rtime = HZ/10;
1278 in6_dev->nd_parms->retrans_time = rtime;
1279 in6_dev->tstamp = jiffies;
1280 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1283 rtime = ntohl(ra_msg->reachable_time);
1284 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1285 rtime = (rtime*HZ)/1000;
1287 if (rtime < HZ/10)
1288 rtime = HZ/10;
1290 if (rtime != in6_dev->nd_parms->base_reachable_time) {
1291 in6_dev->nd_parms->base_reachable_time = rtime;
1292 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1293 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1294 in6_dev->tstamp = jiffies;
1295 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1300 skip_linkparms:
1303 * Process options.
1306 if (!neigh)
1307 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1308 skb->dev, 1);
1309 if (neigh) {
1310 u8 *lladdr = NULL;
1311 if (ndopts.nd_opts_src_lladdr) {
1312 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1313 skb->dev);
1314 if (!lladdr) {
1315 ND_PRINTK2(KERN_WARNING
1316 "ICMPv6 RA: invalid link-layer address length\n");
1317 goto out;
1320 neigh_update(neigh, lladdr, NUD_STALE,
1321 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1322 NEIGH_UPDATE_F_OVERRIDE|
1323 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1324 NEIGH_UPDATE_F_ISROUTER);
1327 if (!accept_ra(in6_dev))
1328 goto out;
1330 #ifdef CONFIG_IPV6_ROUTE_INFO
1331 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1332 struct nd_opt_hdr *p;
1333 for (p = ndopts.nd_opts_ri;
1335 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1336 struct route_info *ri = (struct route_info *)p;
1337 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1338 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1339 ri->prefix_len == 0)
1340 continue;
1341 #endif
1342 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1343 continue;
1344 rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1345 &ipv6_hdr(skb)->saddr);
1348 #endif
1350 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1351 /* skip link-specific ndopts from interior routers */
1352 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1353 goto out;
1354 #endif
1356 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1357 struct nd_opt_hdr *p;
1358 for (p = ndopts.nd_opts_pi;
1360 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1361 addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1365 if (ndopts.nd_opts_mtu) {
1366 __be32 n;
1367 u32 mtu;
1369 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1370 mtu = ntohl(n);
1372 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1373 ND_PRINTK2(KERN_WARNING
1374 "ICMPv6 RA: invalid mtu: %d\n",
1375 mtu);
1376 } else if (in6_dev->cnf.mtu6 != mtu) {
1377 in6_dev->cnf.mtu6 = mtu;
1379 if (rt)
1380 rt->dst.metrics[RTAX_MTU-1] = mtu;
1382 rt6_mtu_change(skb->dev, mtu);
1386 if (ndopts.nd_useropts) {
1387 struct nd_opt_hdr *p;
1388 for (p = ndopts.nd_useropts;
1390 p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1391 ndisc_ra_useropt(skb, p);
1395 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1396 ND_PRINTK2(KERN_WARNING
1397 "ICMPv6 RA: invalid RA options");
1399 out:
1400 if (rt)
1401 dst_release(&rt->dst);
1402 else if (neigh)
1403 neigh_release(neigh);
1404 in6_dev_put(in6_dev);
1407 static void ndisc_redirect_rcv(struct sk_buff *skb)
1409 struct inet6_dev *in6_dev;
1410 struct icmp6hdr *icmph;
1411 struct in6_addr *dest;
1412 struct in6_addr *target; /* new first hop to destination */
1413 struct neighbour *neigh;
1414 int on_link = 0;
1415 struct ndisc_options ndopts;
1416 int optlen;
1417 u8 *lladdr = NULL;
1419 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1420 switch (skb->ndisc_nodetype) {
1421 case NDISC_NODETYPE_HOST:
1422 case NDISC_NODETYPE_NODEFAULT:
1423 ND_PRINTK2(KERN_WARNING
1424 "ICMPv6 Redirect: from host or unauthorized router\n");
1425 return;
1427 #endif
1429 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1430 ND_PRINTK2(KERN_WARNING
1431 "ICMPv6 Redirect: source address is not link-local.\n");
1432 return;
1435 optlen = skb->tail - skb->transport_header;
1436 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1438 if (optlen < 0) {
1439 ND_PRINTK2(KERN_WARNING
1440 "ICMPv6 Redirect: packet too short\n");
1441 return;
1444 icmph = icmp6_hdr(skb);
1445 target = (struct in6_addr *) (icmph + 1);
1446 dest = target + 1;
1448 if (ipv6_addr_is_multicast(dest)) {
1449 ND_PRINTK2(KERN_WARNING
1450 "ICMPv6 Redirect: destination address is multicast.\n");
1451 return;
1454 if (ipv6_addr_equal(dest, target)) {
1455 on_link = 1;
1456 } else if (ipv6_addr_type(target) !=
1457 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1458 ND_PRINTK2(KERN_WARNING
1459 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1460 return;
1463 in6_dev = in6_dev_get(skb->dev);
1464 if (!in6_dev)
1465 return;
1466 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1467 in6_dev_put(in6_dev);
1468 return;
1471 /* RFC2461 8.1:
1472 * The IP source address of the Redirect MUST be the same as the current
1473 * first-hop router for the specified ICMP Destination Address.
1476 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1477 ND_PRINTK2(KERN_WARNING
1478 "ICMPv6 Redirect: invalid ND options\n");
1479 in6_dev_put(in6_dev);
1480 return;
1482 if (ndopts.nd_opts_tgt_lladdr) {
1483 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1484 skb->dev);
1485 if (!lladdr) {
1486 ND_PRINTK2(KERN_WARNING
1487 "ICMPv6 Redirect: invalid link-layer address length\n");
1488 in6_dev_put(in6_dev);
1489 return;
1493 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1494 if (neigh) {
1495 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1496 &ipv6_hdr(skb)->saddr, neigh, lladdr,
1497 on_link);
1498 neigh_release(neigh);
1500 in6_dev_put(in6_dev);
1503 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1504 const struct in6_addr *target)
1506 struct net_device *dev = skb->dev;
1507 struct net *net = dev_net(dev);
1508 struct sock *sk = net->ipv6.ndisc_sk;
1509 int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1510 struct sk_buff *buff;
1511 struct icmp6hdr *icmph;
1512 struct in6_addr saddr_buf;
1513 struct in6_addr *addrp;
1514 struct rt6_info *rt;
1515 struct dst_entry *dst;
1516 struct inet6_dev *idev;
1517 struct flowi fl;
1518 u8 *opt;
1519 int rd_len;
1520 int err;
1521 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1523 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1524 ND_PRINTK2(KERN_WARNING
1525 "ICMPv6 Redirect: no link-local address on %s\n",
1526 dev->name);
1527 return;
1530 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1531 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1532 ND_PRINTK2(KERN_WARNING
1533 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1534 return;
1537 icmpv6_flow_init(sk, &fl, NDISC_REDIRECT,
1538 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1540 dst = ip6_route_output(net, NULL, &fl);
1541 if (dst == NULL)
1542 return;
1544 err = xfrm_lookup(net, &dst, &fl, NULL, 0);
1545 if (err)
1546 return;
1548 rt = (struct rt6_info *) dst;
1550 if (rt->rt6i_flags & RTF_GATEWAY) {
1551 ND_PRINTK2(KERN_WARNING
1552 "ICMPv6 Redirect: destination is not a neighbour.\n");
1553 goto release;
1555 if (!xrlim_allow(dst, 1*HZ))
1556 goto release;
1558 if (dev->addr_len) {
1559 read_lock_bh(&neigh->lock);
1560 if (neigh->nud_state & NUD_VALID) {
1561 memcpy(ha_buf, neigh->ha, dev->addr_len);
1562 read_unlock_bh(&neigh->lock);
1563 ha = ha_buf;
1564 len += ndisc_opt_addr_space(dev);
1565 } else
1566 read_unlock_bh(&neigh->lock);
1569 rd_len = min_t(unsigned int,
1570 IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1571 rd_len &= ~0x7;
1572 len += rd_len;
1574 buff = sock_alloc_send_skb(sk,
1575 (MAX_HEADER + sizeof(struct ipv6hdr) +
1576 len + LL_ALLOCATED_SPACE(dev)),
1577 1, &err);
1578 if (buff == NULL) {
1579 ND_PRINTK0(KERN_ERR
1580 "ICMPv6 Redirect: %s() failed to allocate an skb, err=%d.\n",
1581 __func__, err);
1582 goto release;
1585 skb_reserve(buff, LL_RESERVED_SPACE(dev));
1586 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1587 IPPROTO_ICMPV6, len);
1589 skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1590 skb_put(buff, len);
1591 icmph = icmp6_hdr(buff);
1593 memset(icmph, 0, sizeof(struct icmp6hdr));
1594 icmph->icmp6_type = NDISC_REDIRECT;
1597 * copy target and destination addresses
1600 addrp = (struct in6_addr *)(icmph + 1);
1601 ipv6_addr_copy(addrp, target);
1602 addrp++;
1603 ipv6_addr_copy(addrp, &ipv6_hdr(skb)->daddr);
1605 opt = (u8*) (addrp + 1);
1608 * include target_address option
1611 if (ha)
1612 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1613 dev->addr_len, dev->type);
1616 * build redirect option and copy skb over to the new packet.
1619 memset(opt, 0, 8);
1620 *(opt++) = ND_OPT_REDIRECT_HDR;
1621 *(opt++) = (rd_len >> 3);
1622 opt += 6;
1624 memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1626 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1627 len, IPPROTO_ICMPV6,
1628 csum_partial(icmph, len, 0));
1630 skb_dst_set(buff, dst);
1631 idev = in6_dev_get(dst->dev);
1632 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1633 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, buff, NULL, dst->dev,
1634 dst_output);
1635 if (!err) {
1636 ICMP6MSGOUT_INC_STATS(net, idev, NDISC_REDIRECT);
1637 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1640 if (likely(idev != NULL))
1641 in6_dev_put(idev);
1642 return;
1644 release:
1645 dst_release(dst);
1648 static void pndisc_redo(struct sk_buff *skb)
1650 ndisc_recv_ns(skb);
1651 kfree_skb(skb);
1654 int ndisc_rcv(struct sk_buff *skb)
1656 struct nd_msg *msg;
1658 if (!pskb_may_pull(skb, skb->len))
1659 return 0;
1661 msg = (struct nd_msg *)skb_transport_header(skb);
1663 __skb_push(skb, skb->data - skb_transport_header(skb));
1665 if (ipv6_hdr(skb)->hop_limit != 255) {
1666 ND_PRINTK2(KERN_WARNING
1667 "ICMPv6 NDISC: invalid hop-limit: %d\n",
1668 ipv6_hdr(skb)->hop_limit);
1669 return 0;
1672 if (msg->icmph.icmp6_code != 0) {
1673 ND_PRINTK2(KERN_WARNING
1674 "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1675 msg->icmph.icmp6_code);
1676 return 0;
1679 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1681 switch (msg->icmph.icmp6_type) {
1682 case NDISC_NEIGHBOUR_SOLICITATION:
1683 ndisc_recv_ns(skb);
1684 break;
1686 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1687 ndisc_recv_na(skb);
1688 break;
1690 case NDISC_ROUTER_SOLICITATION:
1691 ndisc_recv_rs(skb);
1692 break;
1694 case NDISC_ROUTER_ADVERTISEMENT:
1695 ndisc_router_discovery(skb);
1696 break;
1698 case NDISC_REDIRECT:
1699 ndisc_redirect_rcv(skb);
1700 break;
1703 return 0;
1706 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1708 struct net_device *dev = ptr;
1709 struct net *net = dev_net(dev);
1711 switch (event) {
1712 case NETDEV_CHANGEADDR:
1713 neigh_changeaddr(&nd_tbl, dev);
1714 fib6_run_gc(~0UL, net);
1715 break;
1716 case NETDEV_DOWN:
1717 neigh_ifdown(&nd_tbl, dev);
1718 fib6_run_gc(~0UL, net);
1719 break;
1720 default:
1721 break;
1724 return NOTIFY_DONE;
1727 static struct notifier_block ndisc_netdev_notifier = {
1728 .notifier_call = ndisc_netdev_event,
1731 #ifdef CONFIG_SYSCTL
1732 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1733 const char *func, const char *dev_name)
1735 static char warncomm[TASK_COMM_LEN];
1736 static int warned;
1737 if (strcmp(warncomm, current->comm) && warned < 5) {
1738 strcpy(warncomm, current->comm);
1739 printk(KERN_WARNING
1740 "process `%s' is using deprecated sysctl (%s) "
1741 "net.ipv6.neigh.%s.%s; "
1742 "Use net.ipv6.neigh.%s.%s_ms "
1743 "instead.\n",
1744 warncomm, func,
1745 dev_name, ctl->procname,
1746 dev_name, ctl->procname);
1747 warned++;
1751 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1753 struct net_device *dev = ctl->extra1;
1754 struct inet6_dev *idev;
1755 int ret;
1757 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1758 (strcmp(ctl->procname, "base_reachable_time") == 0))
1759 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1761 if (strcmp(ctl->procname, "retrans_time") == 0)
1762 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1764 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1765 ret = proc_dointvec_jiffies(ctl, write,
1766 buffer, lenp, ppos);
1768 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1769 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1770 ret = proc_dointvec_ms_jiffies(ctl, write,
1771 buffer, lenp, ppos);
1772 else
1773 ret = -1;
1775 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1776 if (ctl->data == &idev->nd_parms->base_reachable_time)
1777 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1778 idev->tstamp = jiffies;
1779 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1780 in6_dev_put(idev);
1782 return ret;
1786 #endif
1788 static int __net_init ndisc_net_init(struct net *net)
1790 struct ipv6_pinfo *np;
1791 struct sock *sk;
1792 int err;
1794 err = inet_ctl_sock_create(&sk, PF_INET6,
1795 SOCK_RAW, IPPROTO_ICMPV6, net);
1796 if (err < 0) {
1797 ND_PRINTK0(KERN_ERR
1798 "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1799 err);
1800 return err;
1803 net->ipv6.ndisc_sk = sk;
1805 np = inet6_sk(sk);
1806 np->hop_limit = 255;
1807 /* Do not loopback ndisc messages */
1808 np->mc_loop = 0;
1810 return 0;
1813 static void __net_exit ndisc_net_exit(struct net *net)
1815 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1818 static struct pernet_operations ndisc_net_ops = {
1819 .init = ndisc_net_init,
1820 .exit = ndisc_net_exit,
1823 int __init ndisc_init(void)
1825 int err;
1827 err = register_pernet_subsys(&ndisc_net_ops);
1828 if (err)
1829 return err;
1831 * Initialize the neighbour table
1833 neigh_table_init(&nd_tbl);
1835 #ifdef CONFIG_SYSCTL
1836 err = neigh_sysctl_register(NULL, &nd_tbl.parms, "ipv6",
1837 &ndisc_ifinfo_sysctl_change);
1838 if (err)
1839 goto out_unregister_pernet;
1840 #endif
1841 err = register_netdevice_notifier(&ndisc_netdev_notifier);
1842 if (err)
1843 goto out_unregister_sysctl;
1844 out:
1845 return err;
1847 out_unregister_sysctl:
1848 #ifdef CONFIG_SYSCTL
1849 neigh_sysctl_unregister(&nd_tbl.parms);
1850 out_unregister_pernet:
1851 #endif
1852 unregister_pernet_subsys(&ndisc_net_ops);
1853 goto out;
1856 void ndisc_cleanup(void)
1858 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1859 #ifdef CONFIG_SYSCTL
1860 neigh_sysctl_unregister(&nd_tbl.parms);
1861 #endif
1862 neigh_table_clear(&nd_tbl);
1863 unregister_pernet_subsys(&ndisc_net_ops);