driver core: remove drivers/base/sys.c and include/linux/sysdev.h
[zen-stable.git] / net / ipv6 / ndisc.c
blobd8f02ef88e59b265322890025aa22e51815abb3a
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 *hash_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,
112 static const struct neigh_ops ndisc_hh_ops = {
113 .family = AF_INET6,
114 .solicit = ndisc_solicit,
115 .error_report = ndisc_error_report,
116 .output = neigh_resolve_output,
117 .connected_output = neigh_resolve_output,
121 static const struct neigh_ops ndisc_direct_ops = {
122 .family = AF_INET6,
123 .output = neigh_direct_output,
124 .connected_output = neigh_direct_output,
127 struct neigh_table nd_tbl = {
128 .family = AF_INET6,
129 .key_len = sizeof(struct in6_addr),
130 .hash = ndisc_hash,
131 .constructor = ndisc_constructor,
132 .pconstructor = pndisc_constructor,
133 .pdestructor = pndisc_destructor,
134 .proxy_redo = pndisc_redo,
135 .id = "ndisc_cache",
136 .parms = {
137 .tbl = &nd_tbl,
138 .base_reachable_time = ND_REACHABLE_TIME,
139 .retrans_time = ND_RETRANS_TIMER,
140 .gc_staletime = 60 * HZ,
141 .reachable_time = ND_REACHABLE_TIME,
142 .delay_probe_time = 5 * HZ,
143 .queue_len_bytes = 64*1024,
144 .ucast_probes = 3,
145 .mcast_probes = 3,
146 .anycast_delay = 1 * HZ,
147 .proxy_delay = (8 * HZ) / 10,
148 .proxy_qlen = 64,
150 .gc_interval = 30 * HZ,
151 .gc_thresh1 = 128,
152 .gc_thresh2 = 512,
153 .gc_thresh3 = 1024,
156 /* ND options */
157 struct ndisc_options {
158 struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
159 #ifdef CONFIG_IPV6_ROUTE_INFO
160 struct nd_opt_hdr *nd_opts_ri;
161 struct nd_opt_hdr *nd_opts_ri_end;
162 #endif
163 struct nd_opt_hdr *nd_useropts;
164 struct nd_opt_hdr *nd_useropts_end;
167 #define nd_opts_src_lladdr nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
168 #define nd_opts_tgt_lladdr nd_opt_array[ND_OPT_TARGET_LL_ADDR]
169 #define nd_opts_pi nd_opt_array[ND_OPT_PREFIX_INFO]
170 #define nd_opts_pi_end nd_opt_array[__ND_OPT_PREFIX_INFO_END]
171 #define nd_opts_rh nd_opt_array[ND_OPT_REDIRECT_HDR]
172 #define nd_opts_mtu nd_opt_array[ND_OPT_MTU]
174 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
177 * Return the padding between the option length and the start of the
178 * link addr. Currently only IP-over-InfiniBand needs this, although
179 * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
180 * also need a pad of 2.
182 static int ndisc_addr_option_pad(unsigned short type)
184 switch (type) {
185 case ARPHRD_INFINIBAND: return 2;
186 default: return 0;
190 static inline int ndisc_opt_addr_space(struct net_device *dev)
192 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
195 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
196 unsigned short addr_type)
198 int space = NDISC_OPT_SPACE(data_len);
199 int pad = ndisc_addr_option_pad(addr_type);
201 opt[0] = type;
202 opt[1] = space>>3;
204 memset(opt + 2, 0, pad);
205 opt += pad;
206 space -= pad;
208 memcpy(opt+2, data, data_len);
209 data_len += 2;
210 opt += data_len;
211 if ((space -= data_len) > 0)
212 memset(opt, 0, space);
213 return opt + space;
216 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
217 struct nd_opt_hdr *end)
219 int type;
220 if (!cur || !end || cur >= end)
221 return NULL;
222 type = cur->nd_opt_type;
223 do {
224 cur = ((void *)cur) + (cur->nd_opt_len << 3);
225 } while(cur < end && cur->nd_opt_type != type);
226 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
229 static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
231 return opt->nd_opt_type == ND_OPT_RDNSS;
234 static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
235 struct nd_opt_hdr *end)
237 if (!cur || !end || cur >= end)
238 return NULL;
239 do {
240 cur = ((void *)cur) + (cur->nd_opt_len << 3);
241 } while(cur < end && !ndisc_is_useropt(cur));
242 return cur <= end && ndisc_is_useropt(cur) ? cur : NULL;
245 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
246 struct ndisc_options *ndopts)
248 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
250 if (!nd_opt || opt_len < 0 || !ndopts)
251 return NULL;
252 memset(ndopts, 0, sizeof(*ndopts));
253 while (opt_len) {
254 int l;
255 if (opt_len < sizeof(struct nd_opt_hdr))
256 return NULL;
257 l = nd_opt->nd_opt_len << 3;
258 if (opt_len < l || l == 0)
259 return NULL;
260 switch (nd_opt->nd_opt_type) {
261 case ND_OPT_SOURCE_LL_ADDR:
262 case ND_OPT_TARGET_LL_ADDR:
263 case ND_OPT_MTU:
264 case ND_OPT_REDIRECT_HDR:
265 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
266 ND_PRINTK2(KERN_WARNING
267 "%s(): duplicated ND6 option found: type=%d\n",
268 __func__,
269 nd_opt->nd_opt_type);
270 } else {
271 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
273 break;
274 case ND_OPT_PREFIX_INFO:
275 ndopts->nd_opts_pi_end = nd_opt;
276 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
277 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
278 break;
279 #ifdef CONFIG_IPV6_ROUTE_INFO
280 case ND_OPT_ROUTE_INFO:
281 ndopts->nd_opts_ri_end = nd_opt;
282 if (!ndopts->nd_opts_ri)
283 ndopts->nd_opts_ri = nd_opt;
284 break;
285 #endif
286 default:
287 if (ndisc_is_useropt(nd_opt)) {
288 ndopts->nd_useropts_end = nd_opt;
289 if (!ndopts->nd_useropts)
290 ndopts->nd_useropts = nd_opt;
291 } else {
293 * Unknown options must be silently ignored,
294 * to accommodate future extension to the
295 * protocol.
297 ND_PRINTK2(KERN_NOTICE
298 "%s(): ignored unsupported option; type=%d, len=%d\n",
299 __func__,
300 nd_opt->nd_opt_type, nd_opt->nd_opt_len);
303 opt_len -= l;
304 nd_opt = ((void *)nd_opt) + l;
306 return ndopts;
309 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
310 struct net_device *dev)
312 u8 *lladdr = (u8 *)(p + 1);
313 int lladdrlen = p->nd_opt_len << 3;
314 int prepad = ndisc_addr_option_pad(dev->type);
315 if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
316 return NULL;
317 return lladdr + prepad;
320 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
322 switch (dev->type) {
323 case ARPHRD_ETHER:
324 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
325 case ARPHRD_FDDI:
326 ipv6_eth_mc_map(addr, buf);
327 return 0;
328 case ARPHRD_IEEE802_TR:
329 ipv6_tr_mc_map(addr,buf);
330 return 0;
331 case ARPHRD_ARCNET:
332 ipv6_arcnet_mc_map(addr, buf);
333 return 0;
334 case ARPHRD_INFINIBAND:
335 ipv6_ib_mc_map(addr, dev->broadcast, buf);
336 return 0;
337 case ARPHRD_IPGRE:
338 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
339 default:
340 if (dir) {
341 memcpy(buf, dev->broadcast, dev->addr_len);
342 return 0;
345 return -EINVAL;
348 EXPORT_SYMBOL(ndisc_mc_map);
350 static u32 ndisc_hash(const void *pkey,
351 const struct net_device *dev,
352 __u32 *hash_rnd)
354 return ndisc_hashfn(pkey, dev, hash_rnd);
357 static int ndisc_constructor(struct neighbour *neigh)
359 struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
360 struct net_device *dev = neigh->dev;
361 struct inet6_dev *in6_dev;
362 struct neigh_parms *parms;
363 int is_multicast = ipv6_addr_is_multicast(addr);
365 in6_dev = in6_dev_get(dev);
366 if (in6_dev == NULL) {
367 return -EINVAL;
370 parms = in6_dev->nd_parms;
371 __neigh_parms_put(neigh->parms);
372 neigh->parms = neigh_parms_clone(parms);
374 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
375 if (!dev->header_ops) {
376 neigh->nud_state = NUD_NOARP;
377 neigh->ops = &ndisc_direct_ops;
378 neigh->output = neigh_direct_output;
379 } else {
380 if (is_multicast) {
381 neigh->nud_state = NUD_NOARP;
382 ndisc_mc_map(addr, neigh->ha, dev, 1);
383 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
384 neigh->nud_state = NUD_NOARP;
385 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
386 if (dev->flags&IFF_LOOPBACK)
387 neigh->type = RTN_LOCAL;
388 } else if (dev->flags&IFF_POINTOPOINT) {
389 neigh->nud_state = NUD_NOARP;
390 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
392 if (dev->header_ops->cache)
393 neigh->ops = &ndisc_hh_ops;
394 else
395 neigh->ops = &ndisc_generic_ops;
396 if (neigh->nud_state&NUD_VALID)
397 neigh->output = neigh->ops->connected_output;
398 else
399 neigh->output = neigh->ops->output;
401 in6_dev_put(in6_dev);
402 return 0;
405 static int pndisc_constructor(struct pneigh_entry *n)
407 struct in6_addr *addr = (struct in6_addr*)&n->key;
408 struct in6_addr maddr;
409 struct net_device *dev = n->dev;
411 if (dev == NULL || __in6_dev_get(dev) == NULL)
412 return -EINVAL;
413 addrconf_addr_solict_mult(addr, &maddr);
414 ipv6_dev_mc_inc(dev, &maddr);
415 return 0;
418 static void pndisc_destructor(struct pneigh_entry *n)
420 struct in6_addr *addr = (struct in6_addr*)&n->key;
421 struct in6_addr maddr;
422 struct net_device *dev = n->dev;
424 if (dev == NULL || __in6_dev_get(dev) == NULL)
425 return;
426 addrconf_addr_solict_mult(addr, &maddr);
427 ipv6_dev_mc_dec(dev, &maddr);
430 struct sk_buff *ndisc_build_skb(struct net_device *dev,
431 const struct in6_addr *daddr,
432 const struct in6_addr *saddr,
433 struct icmp6hdr *icmp6h,
434 const struct in6_addr *target,
435 int llinfo)
437 struct net *net = dev_net(dev);
438 struct sock *sk = net->ipv6.ndisc_sk;
439 struct sk_buff *skb;
440 struct icmp6hdr *hdr;
441 int hlen = LL_RESERVED_SPACE(dev);
442 int tlen = dev->needed_tailroom;
443 int len;
444 int err;
445 u8 *opt;
447 if (!dev->addr_len)
448 llinfo = 0;
450 len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
451 if (llinfo)
452 len += ndisc_opt_addr_space(dev);
454 skb = sock_alloc_send_skb(sk,
455 (MAX_HEADER + sizeof(struct ipv6hdr) +
456 len + hlen + tlen),
457 1, &err);
458 if (!skb) {
459 ND_PRINTK0(KERN_ERR
460 "ICMPv6 ND: %s() failed to allocate an skb, err=%d.\n",
461 __func__, err);
462 return NULL;
465 skb_reserve(skb, hlen);
466 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
468 skb->transport_header = skb->tail;
469 skb_put(skb, len);
471 hdr = (struct icmp6hdr *)skb_transport_header(skb);
472 memcpy(hdr, icmp6h, sizeof(*hdr));
474 opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
475 if (target) {
476 *(struct in6_addr *)opt = *target;
477 opt += sizeof(*target);
480 if (llinfo)
481 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
482 dev->addr_len, dev->type);
484 hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
485 IPPROTO_ICMPV6,
486 csum_partial(hdr,
487 len, 0));
489 return skb;
492 EXPORT_SYMBOL(ndisc_build_skb);
494 void ndisc_send_skb(struct sk_buff *skb,
495 struct net_device *dev,
496 struct neighbour *neigh,
497 const struct in6_addr *daddr,
498 const struct in6_addr *saddr,
499 struct icmp6hdr *icmp6h)
501 struct flowi6 fl6;
502 struct dst_entry *dst;
503 struct net *net = dev_net(dev);
504 struct sock *sk = net->ipv6.ndisc_sk;
505 struct inet6_dev *idev;
506 int err;
507 u8 type;
509 type = icmp6h->icmp6_type;
511 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, dev->ifindex);
512 dst = icmp6_dst_alloc(dev, neigh, &fl6);
513 if (IS_ERR(dst)) {
514 kfree_skb(skb);
515 return;
518 skb_dst_set(skb, dst);
520 rcu_read_lock();
521 idev = __in6_dev_get(dst->dev);
522 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
524 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
525 dst_output);
526 if (!err) {
527 ICMP6MSGOUT_INC_STATS(net, idev, type);
528 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
531 rcu_read_unlock();
534 EXPORT_SYMBOL(ndisc_send_skb);
537 * Send a Neighbour Discover packet
539 static void __ndisc_send(struct net_device *dev,
540 struct neighbour *neigh,
541 const struct in6_addr *daddr,
542 const struct in6_addr *saddr,
543 struct icmp6hdr *icmp6h, const struct in6_addr *target,
544 int llinfo)
546 struct sk_buff *skb;
548 skb = ndisc_build_skb(dev, daddr, saddr, icmp6h, target, llinfo);
549 if (!skb)
550 return;
552 ndisc_send_skb(skb, dev, neigh, daddr, saddr, icmp6h);
555 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
556 const struct in6_addr *daddr,
557 const struct in6_addr *solicited_addr,
558 int router, int solicited, int override, int inc_opt)
560 struct in6_addr tmpaddr;
561 struct inet6_ifaddr *ifp;
562 const struct in6_addr *src_addr;
563 struct icmp6hdr icmp6h = {
564 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
567 /* for anycast or proxy, solicited_addr != src_addr */
568 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
569 if (ifp) {
570 src_addr = solicited_addr;
571 if (ifp->flags & IFA_F_OPTIMISTIC)
572 override = 0;
573 inc_opt |= ifp->idev->cnf.force_tllao;
574 in6_ifa_put(ifp);
575 } else {
576 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
577 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
578 &tmpaddr))
579 return;
580 src_addr = &tmpaddr;
583 icmp6h.icmp6_router = router;
584 icmp6h.icmp6_solicited = solicited;
585 icmp6h.icmp6_override = override;
587 __ndisc_send(dev, neigh, daddr, src_addr,
588 &icmp6h, solicited_addr,
589 inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
592 static void ndisc_send_unsol_na(struct net_device *dev)
594 struct inet6_dev *idev;
595 struct inet6_ifaddr *ifa;
596 struct in6_addr mcaddr;
598 idev = in6_dev_get(dev);
599 if (!idev)
600 return;
602 read_lock_bh(&idev->lock);
603 list_for_each_entry(ifa, &idev->addr_list, if_list) {
604 addrconf_addr_solict_mult(&ifa->addr, &mcaddr);
605 ndisc_send_na(dev, NULL, &mcaddr, &ifa->addr,
606 /*router=*/ !!idev->cnf.forwarding,
607 /*solicited=*/ false, /*override=*/ true,
608 /*inc_opt=*/ true);
610 read_unlock_bh(&idev->lock);
612 in6_dev_put(idev);
615 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
616 const struct in6_addr *solicit,
617 const struct in6_addr *daddr, const struct in6_addr *saddr)
619 struct in6_addr addr_buf;
620 struct icmp6hdr icmp6h = {
621 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
624 if (saddr == NULL) {
625 if (ipv6_get_lladdr(dev, &addr_buf,
626 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
627 return;
628 saddr = &addr_buf;
631 __ndisc_send(dev, neigh, daddr, saddr,
632 &icmp6h, solicit,
633 !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
636 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
637 const struct in6_addr *daddr)
639 struct icmp6hdr icmp6h = {
640 .icmp6_type = NDISC_ROUTER_SOLICITATION,
642 int send_sllao = dev->addr_len;
644 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
646 * According to section 2.2 of RFC 4429, we must not
647 * send router solicitations with a sllao from
648 * optimistic addresses, but we may send the solicitation
649 * if we don't include the sllao. So here we check
650 * if our address is optimistic, and if so, we
651 * suppress the inclusion of the sllao.
653 if (send_sllao) {
654 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
655 dev, 1);
656 if (ifp) {
657 if (ifp->flags & IFA_F_OPTIMISTIC) {
658 send_sllao = 0;
660 in6_ifa_put(ifp);
661 } else {
662 send_sllao = 0;
665 #endif
666 __ndisc_send(dev, NULL, daddr, saddr,
667 &icmp6h, NULL,
668 send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
672 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
675 * "The sender MUST return an ICMP
676 * destination unreachable"
678 dst_link_failure(skb);
679 kfree_skb(skb);
682 /* Called with locked neigh: either read or both */
684 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
686 struct in6_addr *saddr = NULL;
687 struct in6_addr mcaddr;
688 struct net_device *dev = neigh->dev;
689 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
690 int probes = atomic_read(&neigh->probes);
692 if (skb && ipv6_chk_addr(dev_net(dev), &ipv6_hdr(skb)->saddr, dev, 1))
693 saddr = &ipv6_hdr(skb)->saddr;
695 if ((probes -= neigh->parms->ucast_probes) < 0) {
696 if (!(neigh->nud_state & NUD_VALID)) {
697 ND_PRINTK1(KERN_DEBUG "%s(): trying to ucast probe in NUD_INVALID: %pI6\n",
698 __func__, target);
700 ndisc_send_ns(dev, neigh, target, target, saddr);
701 } else if ((probes -= neigh->parms->app_probes) < 0) {
702 #ifdef CONFIG_ARPD
703 neigh_app_ns(neigh);
704 #endif
705 } else {
706 addrconf_addr_solict_mult(target, &mcaddr);
707 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
711 static int pndisc_is_router(const void *pkey,
712 struct net_device *dev)
714 struct pneigh_entry *n;
715 int ret = -1;
717 read_lock_bh(&nd_tbl.lock);
718 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
719 if (n)
720 ret = !!(n->flags & NTF_ROUTER);
721 read_unlock_bh(&nd_tbl.lock);
723 return ret;
726 static void ndisc_recv_ns(struct sk_buff *skb)
728 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
729 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
730 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
731 u8 *lladdr = NULL;
732 u32 ndoptlen = skb->tail - (skb->transport_header +
733 offsetof(struct nd_msg, opt));
734 struct ndisc_options ndopts;
735 struct net_device *dev = skb->dev;
736 struct inet6_ifaddr *ifp;
737 struct inet6_dev *idev = NULL;
738 struct neighbour *neigh;
739 int dad = ipv6_addr_any(saddr);
740 int inc;
741 int is_router = -1;
743 if (ipv6_addr_is_multicast(&msg->target)) {
744 ND_PRINTK2(KERN_WARNING
745 "ICMPv6 NS: multicast target address");
746 return;
750 * RFC2461 7.1.1:
751 * DAD has to be destined for solicited node multicast address.
753 if (dad &&
754 !(daddr->s6_addr32[0] == htonl(0xff020000) &&
755 daddr->s6_addr32[1] == htonl(0x00000000) &&
756 daddr->s6_addr32[2] == htonl(0x00000001) &&
757 daddr->s6_addr [12] == 0xff )) {
758 ND_PRINTK2(KERN_WARNING
759 "ICMPv6 NS: bad DAD packet (wrong destination)\n");
760 return;
763 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
764 ND_PRINTK2(KERN_WARNING
765 "ICMPv6 NS: invalid ND options\n");
766 return;
769 if (ndopts.nd_opts_src_lladdr) {
770 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
771 if (!lladdr) {
772 ND_PRINTK2(KERN_WARNING
773 "ICMPv6 NS: invalid link-layer address length\n");
774 return;
777 /* RFC2461 7.1.1:
778 * If the IP source address is the unspecified address,
779 * there MUST NOT be source link-layer address option
780 * in the message.
782 if (dad) {
783 ND_PRINTK2(KERN_WARNING
784 "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
785 return;
789 inc = ipv6_addr_is_multicast(daddr);
791 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
792 if (ifp) {
794 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
795 if (dad) {
796 if (dev->type == ARPHRD_IEEE802_TR) {
797 const unsigned char *sadr;
798 sadr = skb_mac_header(skb);
799 if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
800 sadr[9] == dev->dev_addr[1] &&
801 sadr[10] == dev->dev_addr[2] &&
802 sadr[11] == dev->dev_addr[3] &&
803 sadr[12] == dev->dev_addr[4] &&
804 sadr[13] == dev->dev_addr[5]) {
805 /* looped-back to us */
806 goto out;
811 * We are colliding with another node
812 * who is doing DAD
813 * so fail our DAD process
815 addrconf_dad_failure(ifp);
816 return;
817 } else {
819 * This is not a dad solicitation.
820 * If we are an optimistic node,
821 * we should respond.
822 * Otherwise, we should ignore it.
824 if (!(ifp->flags & IFA_F_OPTIMISTIC))
825 goto out;
829 idev = ifp->idev;
830 } else {
831 struct net *net = dev_net(dev);
833 idev = in6_dev_get(dev);
834 if (!idev) {
835 /* XXX: count this drop? */
836 return;
839 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
840 (idev->cnf.forwarding &&
841 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
842 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
843 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
844 skb->pkt_type != PACKET_HOST &&
845 inc != 0 &&
846 idev->nd_parms->proxy_delay != 0) {
848 * for anycast or proxy,
849 * sender should delay its response
850 * by a random time between 0 and
851 * MAX_ANYCAST_DELAY_TIME seconds.
852 * (RFC2461) -- yoshfuji
854 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
855 if (n)
856 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
857 goto out;
859 } else
860 goto out;
863 if (is_router < 0)
864 is_router = !!idev->cnf.forwarding;
866 if (dad) {
867 ndisc_send_na(dev, NULL, &in6addr_linklocal_allnodes, &msg->target,
868 is_router, 0, (ifp != NULL), 1);
869 goto out;
872 if (inc)
873 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
874 else
875 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
878 * update / create cache entry
879 * for the source address
881 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
882 !inc || lladdr || !dev->addr_len);
883 if (neigh)
884 neigh_update(neigh, lladdr, NUD_STALE,
885 NEIGH_UPDATE_F_WEAK_OVERRIDE|
886 NEIGH_UPDATE_F_OVERRIDE);
887 if (neigh || !dev->header_ops) {
888 ndisc_send_na(dev, neigh, saddr, &msg->target,
889 is_router,
890 1, (ifp != NULL && inc), inc);
891 if (neigh)
892 neigh_release(neigh);
895 out:
896 if (ifp)
897 in6_ifa_put(ifp);
898 else
899 in6_dev_put(idev);
902 static void ndisc_recv_na(struct sk_buff *skb)
904 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
905 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
906 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
907 u8 *lladdr = NULL;
908 u32 ndoptlen = skb->tail - (skb->transport_header +
909 offsetof(struct nd_msg, opt));
910 struct ndisc_options ndopts;
911 struct net_device *dev = skb->dev;
912 struct inet6_ifaddr *ifp;
913 struct neighbour *neigh;
915 if (skb->len < sizeof(struct nd_msg)) {
916 ND_PRINTK2(KERN_WARNING
917 "ICMPv6 NA: packet too short\n");
918 return;
921 if (ipv6_addr_is_multicast(&msg->target)) {
922 ND_PRINTK2(KERN_WARNING
923 "ICMPv6 NA: target address is multicast.\n");
924 return;
927 if (ipv6_addr_is_multicast(daddr) &&
928 msg->icmph.icmp6_solicited) {
929 ND_PRINTK2(KERN_WARNING
930 "ICMPv6 NA: solicited NA is multicasted.\n");
931 return;
934 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
935 ND_PRINTK2(KERN_WARNING
936 "ICMPv6 NS: invalid ND option\n");
937 return;
939 if (ndopts.nd_opts_tgt_lladdr) {
940 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
941 if (!lladdr) {
942 ND_PRINTK2(KERN_WARNING
943 "ICMPv6 NA: invalid link-layer address length\n");
944 return;
947 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
948 if (ifp) {
949 if (skb->pkt_type != PACKET_LOOPBACK
950 && (ifp->flags & IFA_F_TENTATIVE)) {
951 addrconf_dad_failure(ifp);
952 return;
954 /* What should we make now? The advertisement
955 is invalid, but ndisc specs say nothing
956 about it. It could be misconfiguration, or
957 an smart proxy agent tries to help us :-)
959 We should not print the error if NA has been
960 received from loopback - it is just our own
961 unsolicited advertisement.
963 if (skb->pkt_type != PACKET_LOOPBACK)
964 ND_PRINTK1(KERN_WARNING
965 "ICMPv6 NA: someone advertises our address %pI6 on %s!\n",
966 &ifp->addr, ifp->idev->dev->name);
967 in6_ifa_put(ifp);
968 return;
970 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
972 if (neigh) {
973 u8 old_flags = neigh->flags;
974 struct net *net = dev_net(dev);
976 if (neigh->nud_state & NUD_FAILED)
977 goto out;
980 * Don't update the neighbor cache entry on a proxy NA from
981 * ourselves because either the proxied node is off link or it
982 * has already sent a NA to us.
984 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
985 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
986 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
987 /* XXX: idev->cnf.prixy_ndp */
988 goto out;
991 neigh_update(neigh, lladdr,
992 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
993 NEIGH_UPDATE_F_WEAK_OVERRIDE|
994 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
995 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
996 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
998 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1000 * Change: router to host
1002 struct rt6_info *rt;
1003 rt = rt6_get_dflt_router(saddr, dev);
1004 if (rt)
1005 ip6_del_rt(rt);
1008 out:
1009 neigh_release(neigh);
1013 static void ndisc_recv_rs(struct sk_buff *skb)
1015 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1016 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1017 struct neighbour *neigh;
1018 struct inet6_dev *idev;
1019 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1020 struct ndisc_options ndopts;
1021 u8 *lladdr = NULL;
1023 if (skb->len < sizeof(*rs_msg))
1024 return;
1026 idev = __in6_dev_get(skb->dev);
1027 if (!idev) {
1028 if (net_ratelimit())
1029 ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
1030 return;
1033 /* Don't accept RS if we're not in router mode */
1034 if (!idev->cnf.forwarding)
1035 goto out;
1038 * Don't update NCE if src = ::;
1039 * this implies that the source node has no ip address assigned yet.
1041 if (ipv6_addr_any(saddr))
1042 goto out;
1044 /* Parse ND options */
1045 if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1046 if (net_ratelimit())
1047 ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1048 goto out;
1051 if (ndopts.nd_opts_src_lladdr) {
1052 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1053 skb->dev);
1054 if (!lladdr)
1055 goto out;
1058 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1059 if (neigh) {
1060 neigh_update(neigh, lladdr, NUD_STALE,
1061 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1062 NEIGH_UPDATE_F_OVERRIDE|
1063 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1064 neigh_release(neigh);
1066 out:
1067 return;
1070 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1072 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1073 struct sk_buff *skb;
1074 struct nlmsghdr *nlh;
1075 struct nduseroptmsg *ndmsg;
1076 struct net *net = dev_net(ra->dev);
1077 int err;
1078 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1079 + (opt->nd_opt_len << 3));
1080 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1082 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1083 if (skb == NULL) {
1084 err = -ENOBUFS;
1085 goto errout;
1088 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1089 if (nlh == NULL) {
1090 goto nla_put_failure;
1093 ndmsg = nlmsg_data(nlh);
1094 ndmsg->nduseropt_family = AF_INET6;
1095 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1096 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1097 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1098 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1100 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1102 NLA_PUT(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
1103 &ipv6_hdr(ra)->saddr);
1104 nlmsg_end(skb, nlh);
1106 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1107 return;
1109 nla_put_failure:
1110 nlmsg_free(skb);
1111 err = -EMSGSIZE;
1112 errout:
1113 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1116 static inline int accept_ra(struct inet6_dev *in6_dev)
1119 * If forwarding is enabled, RA are not accepted unless the special
1120 * hybrid mode (accept_ra=2) is enabled.
1122 if (in6_dev->cnf.forwarding && in6_dev->cnf.accept_ra < 2)
1123 return 0;
1125 return in6_dev->cnf.accept_ra;
1128 static void ndisc_router_discovery(struct sk_buff *skb)
1130 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1131 struct neighbour *neigh = NULL;
1132 struct inet6_dev *in6_dev;
1133 struct rt6_info *rt = NULL;
1134 int lifetime;
1135 struct ndisc_options ndopts;
1136 int optlen;
1137 unsigned int pref = 0;
1139 __u8 * opt = (__u8 *)(ra_msg + 1);
1141 optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1143 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1144 ND_PRINTK2(KERN_WARNING
1145 "ICMPv6 RA: source address is not link-local.\n");
1146 return;
1148 if (optlen < 0) {
1149 ND_PRINTK2(KERN_WARNING
1150 "ICMPv6 RA: packet too short\n");
1151 return;
1154 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1155 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1156 ND_PRINTK2(KERN_WARNING
1157 "ICMPv6 RA: from host or unauthorized router\n");
1158 return;
1160 #endif
1163 * set the RA_RECV flag in the interface
1166 in6_dev = __in6_dev_get(skb->dev);
1167 if (in6_dev == NULL) {
1168 ND_PRINTK0(KERN_ERR
1169 "ICMPv6 RA: can't find inet6 device for %s.\n",
1170 skb->dev->name);
1171 return;
1174 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1175 ND_PRINTK2(KERN_WARNING
1176 "ICMP6 RA: invalid ND options\n");
1177 return;
1180 if (!accept_ra(in6_dev))
1181 goto skip_linkparms;
1183 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1184 /* skip link-specific parameters from interior routers */
1185 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1186 goto skip_linkparms;
1187 #endif
1189 if (in6_dev->if_flags & IF_RS_SENT) {
1191 * flag that an RA was received after an RS was sent
1192 * out on this interface.
1194 in6_dev->if_flags |= IF_RA_RCVD;
1198 * Remember the managed/otherconf flags from most recently
1199 * received RA message (RFC 2462) -- yoshfuji
1201 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1202 IF_RA_OTHERCONF)) |
1203 (ra_msg->icmph.icmp6_addrconf_managed ?
1204 IF_RA_MANAGED : 0) |
1205 (ra_msg->icmph.icmp6_addrconf_other ?
1206 IF_RA_OTHERCONF : 0);
1208 if (!in6_dev->cnf.accept_ra_defrtr)
1209 goto skip_defrtr;
1211 if (ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr, NULL, 0))
1212 goto skip_defrtr;
1214 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1216 #ifdef CONFIG_IPV6_ROUTER_PREF
1217 pref = ra_msg->icmph.icmp6_router_pref;
1218 /* 10b is handled as if it were 00b (medium) */
1219 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1220 !in6_dev->cnf.accept_ra_rtr_pref)
1221 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1222 #endif
1224 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1226 if (rt)
1227 neigh = dst_get_neighbour_noref(&rt->dst);
1229 if (rt && lifetime == 0) {
1230 neigh_clone(neigh);
1231 ip6_del_rt(rt);
1232 rt = NULL;
1235 if (rt == NULL && lifetime) {
1236 ND_PRINTK3(KERN_DEBUG
1237 "ICMPv6 RA: adding default router.\n");
1239 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1240 if (rt == NULL) {
1241 ND_PRINTK0(KERN_ERR
1242 "ICMPv6 RA: %s() failed to add default route.\n",
1243 __func__);
1244 return;
1247 neigh = dst_get_neighbour_noref(&rt->dst);
1248 if (neigh == NULL) {
1249 ND_PRINTK0(KERN_ERR
1250 "ICMPv6 RA: %s() got default router without neighbour.\n",
1251 __func__);
1252 dst_release(&rt->dst);
1253 return;
1255 neigh->flags |= NTF_ROUTER;
1256 } else if (rt) {
1257 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1260 if (rt)
1261 rt->dst.expires = jiffies + (HZ * lifetime);
1263 if (ra_msg->icmph.icmp6_hop_limit) {
1264 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1265 if (rt)
1266 dst_metric_set(&rt->dst, RTAX_HOPLIMIT,
1267 ra_msg->icmph.icmp6_hop_limit);
1270 skip_defrtr:
1273 * Update Reachable Time and Retrans Timer
1276 if (in6_dev->nd_parms) {
1277 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1279 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1280 rtime = (rtime*HZ)/1000;
1281 if (rtime < HZ/10)
1282 rtime = HZ/10;
1283 in6_dev->nd_parms->retrans_time = rtime;
1284 in6_dev->tstamp = jiffies;
1285 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1288 rtime = ntohl(ra_msg->reachable_time);
1289 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1290 rtime = (rtime*HZ)/1000;
1292 if (rtime < HZ/10)
1293 rtime = HZ/10;
1295 if (rtime != in6_dev->nd_parms->base_reachable_time) {
1296 in6_dev->nd_parms->base_reachable_time = rtime;
1297 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1298 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1299 in6_dev->tstamp = jiffies;
1300 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1305 skip_linkparms:
1308 * Process options.
1311 if (!neigh)
1312 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1313 skb->dev, 1);
1314 if (neigh) {
1315 u8 *lladdr = NULL;
1316 if (ndopts.nd_opts_src_lladdr) {
1317 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1318 skb->dev);
1319 if (!lladdr) {
1320 ND_PRINTK2(KERN_WARNING
1321 "ICMPv6 RA: invalid link-layer address length\n");
1322 goto out;
1325 neigh_update(neigh, lladdr, NUD_STALE,
1326 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1327 NEIGH_UPDATE_F_OVERRIDE|
1328 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1329 NEIGH_UPDATE_F_ISROUTER);
1332 if (!accept_ra(in6_dev))
1333 goto out;
1335 #ifdef CONFIG_IPV6_ROUTE_INFO
1336 if (ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr, NULL, 0))
1337 goto skip_routeinfo;
1339 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1340 struct nd_opt_hdr *p;
1341 for (p = ndopts.nd_opts_ri;
1343 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1344 struct route_info *ri = (struct route_info *)p;
1345 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1346 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1347 ri->prefix_len == 0)
1348 continue;
1349 #endif
1350 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1351 continue;
1352 rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1353 &ipv6_hdr(skb)->saddr);
1357 skip_routeinfo:
1358 #endif
1360 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1361 /* skip link-specific ndopts from interior routers */
1362 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1363 goto out;
1364 #endif
1366 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1367 struct nd_opt_hdr *p;
1368 for (p = ndopts.nd_opts_pi;
1370 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1371 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1372 (p->nd_opt_len) << 3,
1373 ndopts.nd_opts_src_lladdr != NULL);
1377 if (ndopts.nd_opts_mtu) {
1378 __be32 n;
1379 u32 mtu;
1381 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1382 mtu = ntohl(n);
1384 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1385 ND_PRINTK2(KERN_WARNING
1386 "ICMPv6 RA: invalid mtu: %d\n",
1387 mtu);
1388 } else if (in6_dev->cnf.mtu6 != mtu) {
1389 in6_dev->cnf.mtu6 = mtu;
1391 if (rt)
1392 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
1394 rt6_mtu_change(skb->dev, mtu);
1398 if (ndopts.nd_useropts) {
1399 struct nd_opt_hdr *p;
1400 for (p = ndopts.nd_useropts;
1402 p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1403 ndisc_ra_useropt(skb, p);
1407 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1408 ND_PRINTK2(KERN_WARNING
1409 "ICMPv6 RA: invalid RA options");
1411 out:
1412 if (rt)
1413 dst_release(&rt->dst);
1414 else if (neigh)
1415 neigh_release(neigh);
1418 static void ndisc_redirect_rcv(struct sk_buff *skb)
1420 struct inet6_dev *in6_dev;
1421 struct icmp6hdr *icmph;
1422 const struct in6_addr *dest;
1423 const struct in6_addr *target; /* new first hop to destination */
1424 struct neighbour *neigh;
1425 int on_link = 0;
1426 struct ndisc_options ndopts;
1427 int optlen;
1428 u8 *lladdr = NULL;
1430 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1431 switch (skb->ndisc_nodetype) {
1432 case NDISC_NODETYPE_HOST:
1433 case NDISC_NODETYPE_NODEFAULT:
1434 ND_PRINTK2(KERN_WARNING
1435 "ICMPv6 Redirect: from host or unauthorized router\n");
1436 return;
1438 #endif
1440 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1441 ND_PRINTK2(KERN_WARNING
1442 "ICMPv6 Redirect: source address is not link-local.\n");
1443 return;
1446 optlen = skb->tail - skb->transport_header;
1447 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1449 if (optlen < 0) {
1450 ND_PRINTK2(KERN_WARNING
1451 "ICMPv6 Redirect: packet too short\n");
1452 return;
1455 icmph = icmp6_hdr(skb);
1456 target = (const struct in6_addr *) (icmph + 1);
1457 dest = target + 1;
1459 if (ipv6_addr_is_multicast(dest)) {
1460 ND_PRINTK2(KERN_WARNING
1461 "ICMPv6 Redirect: destination address is multicast.\n");
1462 return;
1465 if (ipv6_addr_equal(dest, target)) {
1466 on_link = 1;
1467 } else if (ipv6_addr_type(target) !=
1468 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1469 ND_PRINTK2(KERN_WARNING
1470 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1471 return;
1474 in6_dev = __in6_dev_get(skb->dev);
1475 if (!in6_dev)
1476 return;
1477 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1478 return;
1480 /* RFC2461 8.1:
1481 * The IP source address of the Redirect MUST be the same as the current
1482 * first-hop router for the specified ICMP Destination Address.
1485 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1486 ND_PRINTK2(KERN_WARNING
1487 "ICMPv6 Redirect: invalid ND options\n");
1488 return;
1490 if (ndopts.nd_opts_tgt_lladdr) {
1491 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1492 skb->dev);
1493 if (!lladdr) {
1494 ND_PRINTK2(KERN_WARNING
1495 "ICMPv6 Redirect: invalid link-layer address length\n");
1496 return;
1500 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1501 if (neigh) {
1502 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1503 &ipv6_hdr(skb)->saddr, neigh, lladdr,
1504 on_link);
1505 neigh_release(neigh);
1509 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1510 const struct in6_addr *target)
1512 struct net_device *dev = skb->dev;
1513 struct net *net = dev_net(dev);
1514 struct sock *sk = net->ipv6.ndisc_sk;
1515 int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1516 struct sk_buff *buff;
1517 struct icmp6hdr *icmph;
1518 struct in6_addr saddr_buf;
1519 struct in6_addr *addrp;
1520 struct rt6_info *rt;
1521 struct dst_entry *dst;
1522 struct inet6_dev *idev;
1523 struct flowi6 fl6;
1524 u8 *opt;
1525 int hlen, tlen;
1526 int rd_len;
1527 int err;
1528 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1530 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1531 ND_PRINTK2(KERN_WARNING
1532 "ICMPv6 Redirect: no link-local address on %s\n",
1533 dev->name);
1534 return;
1537 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1538 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1539 ND_PRINTK2(KERN_WARNING
1540 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1541 return;
1544 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1545 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1547 dst = ip6_route_output(net, NULL, &fl6);
1548 if (dst == NULL)
1549 return;
1551 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1552 if (IS_ERR(dst))
1553 return;
1555 rt = (struct rt6_info *) dst;
1557 if (rt->rt6i_flags & RTF_GATEWAY) {
1558 ND_PRINTK2(KERN_WARNING
1559 "ICMPv6 Redirect: destination is not a neighbour.\n");
1560 goto release;
1562 if (!rt->rt6i_peer)
1563 rt6_bind_peer(rt, 1);
1564 if (!inet_peer_xrlim_allow(rt->rt6i_peer, 1*HZ))
1565 goto release;
1567 if (dev->addr_len) {
1568 read_lock_bh(&neigh->lock);
1569 if (neigh->nud_state & NUD_VALID) {
1570 memcpy(ha_buf, neigh->ha, dev->addr_len);
1571 read_unlock_bh(&neigh->lock);
1572 ha = ha_buf;
1573 len += ndisc_opt_addr_space(dev);
1574 } else
1575 read_unlock_bh(&neigh->lock);
1578 rd_len = min_t(unsigned int,
1579 IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1580 rd_len &= ~0x7;
1581 len += rd_len;
1583 hlen = LL_RESERVED_SPACE(dev);
1584 tlen = dev->needed_tailroom;
1585 buff = sock_alloc_send_skb(sk,
1586 (MAX_HEADER + sizeof(struct ipv6hdr) +
1587 len + hlen + tlen),
1588 1, &err);
1589 if (buff == NULL) {
1590 ND_PRINTK0(KERN_ERR
1591 "ICMPv6 Redirect: %s() failed to allocate an skb, err=%d.\n",
1592 __func__, err);
1593 goto release;
1596 skb_reserve(buff, hlen);
1597 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1598 IPPROTO_ICMPV6, len);
1600 skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1601 skb_put(buff, len);
1602 icmph = icmp6_hdr(buff);
1604 memset(icmph, 0, sizeof(struct icmp6hdr));
1605 icmph->icmp6_type = NDISC_REDIRECT;
1608 * copy target and destination addresses
1611 addrp = (struct in6_addr *)(icmph + 1);
1612 *addrp = *target;
1613 addrp++;
1614 *addrp = ipv6_hdr(skb)->daddr;
1616 opt = (u8*) (addrp + 1);
1619 * include target_address option
1622 if (ha)
1623 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1624 dev->addr_len, dev->type);
1627 * build redirect option and copy skb over to the new packet.
1630 memset(opt, 0, 8);
1631 *(opt++) = ND_OPT_REDIRECT_HDR;
1632 *(opt++) = (rd_len >> 3);
1633 opt += 6;
1635 memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1637 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1638 len, IPPROTO_ICMPV6,
1639 csum_partial(icmph, len, 0));
1641 skb_dst_set(buff, dst);
1642 rcu_read_lock();
1643 idev = __in6_dev_get(dst->dev);
1644 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1645 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, buff, NULL, dst->dev,
1646 dst_output);
1647 if (!err) {
1648 ICMP6MSGOUT_INC_STATS(net, idev, NDISC_REDIRECT);
1649 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1652 rcu_read_unlock();
1653 return;
1655 release:
1656 dst_release(dst);
1659 static void pndisc_redo(struct sk_buff *skb)
1661 ndisc_recv_ns(skb);
1662 kfree_skb(skb);
1665 int ndisc_rcv(struct sk_buff *skb)
1667 struct nd_msg *msg;
1669 if (!pskb_may_pull(skb, skb->len))
1670 return 0;
1672 msg = (struct nd_msg *)skb_transport_header(skb);
1674 __skb_push(skb, skb->data - skb_transport_header(skb));
1676 if (ipv6_hdr(skb)->hop_limit != 255) {
1677 ND_PRINTK2(KERN_WARNING
1678 "ICMPv6 NDISC: invalid hop-limit: %d\n",
1679 ipv6_hdr(skb)->hop_limit);
1680 return 0;
1683 if (msg->icmph.icmp6_code != 0) {
1684 ND_PRINTK2(KERN_WARNING
1685 "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1686 msg->icmph.icmp6_code);
1687 return 0;
1690 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1692 switch (msg->icmph.icmp6_type) {
1693 case NDISC_NEIGHBOUR_SOLICITATION:
1694 ndisc_recv_ns(skb);
1695 break;
1697 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1698 ndisc_recv_na(skb);
1699 break;
1701 case NDISC_ROUTER_SOLICITATION:
1702 ndisc_recv_rs(skb);
1703 break;
1705 case NDISC_ROUTER_ADVERTISEMENT:
1706 ndisc_router_discovery(skb);
1707 break;
1709 case NDISC_REDIRECT:
1710 ndisc_redirect_rcv(skb);
1711 break;
1714 return 0;
1717 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1719 struct net_device *dev = ptr;
1720 struct net *net = dev_net(dev);
1722 switch (event) {
1723 case NETDEV_CHANGEADDR:
1724 neigh_changeaddr(&nd_tbl, dev);
1725 fib6_run_gc(~0UL, net);
1726 break;
1727 case NETDEV_DOWN:
1728 neigh_ifdown(&nd_tbl, dev);
1729 fib6_run_gc(~0UL, net);
1730 break;
1731 case NETDEV_NOTIFY_PEERS:
1732 ndisc_send_unsol_na(dev);
1733 break;
1734 default:
1735 break;
1738 return NOTIFY_DONE;
1741 static struct notifier_block ndisc_netdev_notifier = {
1742 .notifier_call = ndisc_netdev_event,
1745 #ifdef CONFIG_SYSCTL
1746 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1747 const char *func, const char *dev_name)
1749 static char warncomm[TASK_COMM_LEN];
1750 static int warned;
1751 if (strcmp(warncomm, current->comm) && warned < 5) {
1752 strcpy(warncomm, current->comm);
1753 printk(KERN_WARNING
1754 "process `%s' is using deprecated sysctl (%s) "
1755 "net.ipv6.neigh.%s.%s; "
1756 "Use net.ipv6.neigh.%s.%s_ms "
1757 "instead.\n",
1758 warncomm, func,
1759 dev_name, ctl->procname,
1760 dev_name, ctl->procname);
1761 warned++;
1765 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1767 struct net_device *dev = ctl->extra1;
1768 struct inet6_dev *idev;
1769 int ret;
1771 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1772 (strcmp(ctl->procname, "base_reachable_time") == 0))
1773 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1775 if (strcmp(ctl->procname, "retrans_time") == 0)
1776 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1778 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1779 ret = proc_dointvec_jiffies(ctl, write,
1780 buffer, lenp, ppos);
1782 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1783 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1784 ret = proc_dointvec_ms_jiffies(ctl, write,
1785 buffer, lenp, ppos);
1786 else
1787 ret = -1;
1789 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1790 if (ctl->data == &idev->nd_parms->base_reachable_time)
1791 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1792 idev->tstamp = jiffies;
1793 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1794 in6_dev_put(idev);
1796 return ret;
1800 #endif
1802 static int __net_init ndisc_net_init(struct net *net)
1804 struct ipv6_pinfo *np;
1805 struct sock *sk;
1806 int err;
1808 err = inet_ctl_sock_create(&sk, PF_INET6,
1809 SOCK_RAW, IPPROTO_ICMPV6, net);
1810 if (err < 0) {
1811 ND_PRINTK0(KERN_ERR
1812 "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1813 err);
1814 return err;
1817 net->ipv6.ndisc_sk = sk;
1819 np = inet6_sk(sk);
1820 np->hop_limit = 255;
1821 /* Do not loopback ndisc messages */
1822 np->mc_loop = 0;
1824 return 0;
1827 static void __net_exit ndisc_net_exit(struct net *net)
1829 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1832 static struct pernet_operations ndisc_net_ops = {
1833 .init = ndisc_net_init,
1834 .exit = ndisc_net_exit,
1837 int __init ndisc_init(void)
1839 int err;
1841 err = register_pernet_subsys(&ndisc_net_ops);
1842 if (err)
1843 return err;
1845 * Initialize the neighbour table
1847 neigh_table_init(&nd_tbl);
1849 #ifdef CONFIG_SYSCTL
1850 err = neigh_sysctl_register(NULL, &nd_tbl.parms, "ipv6",
1851 &ndisc_ifinfo_sysctl_change);
1852 if (err)
1853 goto out_unregister_pernet;
1854 #endif
1855 err = register_netdevice_notifier(&ndisc_netdev_notifier);
1856 if (err)
1857 goto out_unregister_sysctl;
1858 out:
1859 return err;
1861 out_unregister_sysctl:
1862 #ifdef CONFIG_SYSCTL
1863 neigh_sysctl_unregister(&nd_tbl.parms);
1864 out_unregister_pernet:
1865 #endif
1866 unregister_pernet_subsys(&ndisc_net_ops);
1867 goto out;
1870 void ndisc_cleanup(void)
1872 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1873 #ifdef CONFIG_SYSCTL
1874 neigh_sysctl_unregister(&nd_tbl.parms);
1875 #endif
1876 neigh_table_clear(&nd_tbl);
1877 unregister_pernet_subsys(&ndisc_net_ops);