staging: brcm80211: removed unneeded call to brcms_b_tx_fifo_suspended
[zen-stable.git] / net / ipv6 / ndisc.c
blob9da6e02eaaebffdb47949ff0c0fbf43dc1b9a208
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,
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 .entry_size = sizeof(struct neighbour) + sizeof(struct in6_addr),
130 .key_len = sizeof(struct in6_addr),
131 .hash = ndisc_hash,
132 .constructor = ndisc_constructor,
133 .pconstructor = pndisc_constructor,
134 .pdestructor = pndisc_destructor,
135 .proxy_redo = pndisc_redo,
136 .id = "ndisc_cache",
137 .parms = {
138 .tbl = &nd_tbl,
139 .base_reachable_time = ND_REACHABLE_TIME,
140 .retrans_time = ND_RETRANS_TIMER,
141 .gc_staletime = 60 * HZ,
142 .reachable_time = ND_REACHABLE_TIME,
143 .delay_probe_time = 5 * HZ,
144 .queue_len = 3,
145 .ucast_probes = 3,
146 .mcast_probes = 3,
147 .anycast_delay = 1 * HZ,
148 .proxy_delay = (8 * HZ) / 10,
149 .proxy_qlen = 64,
151 .gc_interval = 30 * HZ,
152 .gc_thresh1 = 128,
153 .gc_thresh2 = 512,
154 .gc_thresh3 = 1024,
157 /* ND options */
158 struct ndisc_options {
159 struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
160 #ifdef CONFIG_IPV6_ROUTE_INFO
161 struct nd_opt_hdr *nd_opts_ri;
162 struct nd_opt_hdr *nd_opts_ri_end;
163 #endif
164 struct nd_opt_hdr *nd_useropts;
165 struct nd_opt_hdr *nd_useropts_end;
168 #define nd_opts_src_lladdr nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
169 #define nd_opts_tgt_lladdr nd_opt_array[ND_OPT_TARGET_LL_ADDR]
170 #define nd_opts_pi nd_opt_array[ND_OPT_PREFIX_INFO]
171 #define nd_opts_pi_end nd_opt_array[__ND_OPT_PREFIX_INFO_END]
172 #define nd_opts_rh nd_opt_array[ND_OPT_REDIRECT_HDR]
173 #define nd_opts_mtu nd_opt_array[ND_OPT_MTU]
175 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
178 * Return the padding between the option length and the start of the
179 * link addr. Currently only IP-over-InfiniBand needs this, although
180 * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
181 * also need a pad of 2.
183 static int ndisc_addr_option_pad(unsigned short type)
185 switch (type) {
186 case ARPHRD_INFINIBAND: return 2;
187 default: return 0;
191 static inline int ndisc_opt_addr_space(struct net_device *dev)
193 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
196 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
197 unsigned short addr_type)
199 int space = NDISC_OPT_SPACE(data_len);
200 int pad = ndisc_addr_option_pad(addr_type);
202 opt[0] = type;
203 opt[1] = space>>3;
205 memset(opt + 2, 0, pad);
206 opt += pad;
207 space -= pad;
209 memcpy(opt+2, data, data_len);
210 data_len += 2;
211 opt += data_len;
212 if ((space -= data_len) > 0)
213 memset(opt, 0, space);
214 return opt + space;
217 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
218 struct nd_opt_hdr *end)
220 int type;
221 if (!cur || !end || cur >= end)
222 return NULL;
223 type = cur->nd_opt_type;
224 do {
225 cur = ((void *)cur) + (cur->nd_opt_len << 3);
226 } while(cur < end && cur->nd_opt_type != type);
227 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
230 static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
232 return opt->nd_opt_type == ND_OPT_RDNSS;
235 static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
236 struct nd_opt_hdr *end)
238 if (!cur || !end || cur >= end)
239 return NULL;
240 do {
241 cur = ((void *)cur) + (cur->nd_opt_len << 3);
242 } while(cur < end && !ndisc_is_useropt(cur));
243 return cur <= end && ndisc_is_useropt(cur) ? cur : NULL;
246 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
247 struct ndisc_options *ndopts)
249 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
251 if (!nd_opt || opt_len < 0 || !ndopts)
252 return NULL;
253 memset(ndopts, 0, sizeof(*ndopts));
254 while (opt_len) {
255 int l;
256 if (opt_len < sizeof(struct nd_opt_hdr))
257 return NULL;
258 l = nd_opt->nd_opt_len << 3;
259 if (opt_len < l || l == 0)
260 return NULL;
261 switch (nd_opt->nd_opt_type) {
262 case ND_OPT_SOURCE_LL_ADDR:
263 case ND_OPT_TARGET_LL_ADDR:
264 case ND_OPT_MTU:
265 case ND_OPT_REDIRECT_HDR:
266 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
267 ND_PRINTK2(KERN_WARNING
268 "%s(): duplicated ND6 option found: type=%d\n",
269 __func__,
270 nd_opt->nd_opt_type);
271 } else {
272 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
274 break;
275 case ND_OPT_PREFIX_INFO:
276 ndopts->nd_opts_pi_end = nd_opt;
277 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
278 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
279 break;
280 #ifdef CONFIG_IPV6_ROUTE_INFO
281 case ND_OPT_ROUTE_INFO:
282 ndopts->nd_opts_ri_end = nd_opt;
283 if (!ndopts->nd_opts_ri)
284 ndopts->nd_opts_ri = nd_opt;
285 break;
286 #endif
287 default:
288 if (ndisc_is_useropt(nd_opt)) {
289 ndopts->nd_useropts_end = nd_opt;
290 if (!ndopts->nd_useropts)
291 ndopts->nd_useropts = nd_opt;
292 } else {
294 * Unknown options must be silently ignored,
295 * to accommodate future extension to the
296 * protocol.
298 ND_PRINTK2(KERN_NOTICE
299 "%s(): ignored unsupported option; type=%d, len=%d\n",
300 __func__,
301 nd_opt->nd_opt_type, nd_opt->nd_opt_len);
304 opt_len -= l;
305 nd_opt = ((void *)nd_opt) + l;
307 return ndopts;
310 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
311 struct net_device *dev)
313 u8 *lladdr = (u8 *)(p + 1);
314 int lladdrlen = p->nd_opt_len << 3;
315 int prepad = ndisc_addr_option_pad(dev->type);
316 if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
317 return NULL;
318 return lladdr + prepad;
321 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
323 switch (dev->type) {
324 case ARPHRD_ETHER:
325 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
326 case ARPHRD_FDDI:
327 ipv6_eth_mc_map(addr, buf);
328 return 0;
329 case ARPHRD_IEEE802_TR:
330 ipv6_tr_mc_map(addr,buf);
331 return 0;
332 case ARPHRD_ARCNET:
333 ipv6_arcnet_mc_map(addr, buf);
334 return 0;
335 case ARPHRD_INFINIBAND:
336 ipv6_ib_mc_map(addr, dev->broadcast, buf);
337 return 0;
338 case ARPHRD_IPGRE:
339 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
340 default:
341 if (dir) {
342 memcpy(buf, dev->broadcast, dev->addr_len);
343 return 0;
346 return -EINVAL;
349 EXPORT_SYMBOL(ndisc_mc_map);
351 static u32 ndisc_hash(const void *pkey,
352 const struct net_device *dev,
353 __u32 hash_rnd)
355 const u32 *p32 = pkey;
356 u32 addr_hash, i;
358 addr_hash = 0;
359 for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
360 addr_hash ^= *p32++;
362 return jhash_2words(addr_hash, dev->ifindex, hash_rnd);
365 static int ndisc_constructor(struct neighbour *neigh)
367 struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
368 struct net_device *dev = neigh->dev;
369 struct inet6_dev *in6_dev;
370 struct neigh_parms *parms;
371 int is_multicast = ipv6_addr_is_multicast(addr);
373 rcu_read_lock();
374 in6_dev = in6_dev_get(dev);
375 if (in6_dev == NULL) {
376 rcu_read_unlock();
377 return -EINVAL;
380 parms = in6_dev->nd_parms;
381 __neigh_parms_put(neigh->parms);
382 neigh->parms = neigh_parms_clone(parms);
383 rcu_read_unlock();
385 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
386 if (!dev->header_ops) {
387 neigh->nud_state = NUD_NOARP;
388 neigh->ops = &ndisc_direct_ops;
389 neigh->output = neigh_direct_output;
390 } else {
391 if (is_multicast) {
392 neigh->nud_state = NUD_NOARP;
393 ndisc_mc_map(addr, neigh->ha, dev, 1);
394 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
395 neigh->nud_state = NUD_NOARP;
396 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
397 if (dev->flags&IFF_LOOPBACK)
398 neigh->type = RTN_LOCAL;
399 } else if (dev->flags&IFF_POINTOPOINT) {
400 neigh->nud_state = NUD_NOARP;
401 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
403 if (dev->header_ops->cache)
404 neigh->ops = &ndisc_hh_ops;
405 else
406 neigh->ops = &ndisc_generic_ops;
407 if (neigh->nud_state&NUD_VALID)
408 neigh->output = neigh->ops->connected_output;
409 else
410 neigh->output = neigh->ops->output;
412 in6_dev_put(in6_dev);
413 return 0;
416 static int pndisc_constructor(struct pneigh_entry *n)
418 struct in6_addr *addr = (struct in6_addr*)&n->key;
419 struct in6_addr maddr;
420 struct net_device *dev = n->dev;
422 if (dev == NULL || __in6_dev_get(dev) == NULL)
423 return -EINVAL;
424 addrconf_addr_solict_mult(addr, &maddr);
425 ipv6_dev_mc_inc(dev, &maddr);
426 return 0;
429 static void pndisc_destructor(struct pneigh_entry *n)
431 struct in6_addr *addr = (struct in6_addr*)&n->key;
432 struct in6_addr maddr;
433 struct net_device *dev = n->dev;
435 if (dev == NULL || __in6_dev_get(dev) == NULL)
436 return;
437 addrconf_addr_solict_mult(addr, &maddr);
438 ipv6_dev_mc_dec(dev, &maddr);
441 struct sk_buff *ndisc_build_skb(struct net_device *dev,
442 const struct in6_addr *daddr,
443 const struct in6_addr *saddr,
444 struct icmp6hdr *icmp6h,
445 const struct in6_addr *target,
446 int llinfo)
448 struct net *net = dev_net(dev);
449 struct sock *sk = net->ipv6.ndisc_sk;
450 struct sk_buff *skb;
451 struct icmp6hdr *hdr;
452 int len;
453 int err;
454 u8 *opt;
456 if (!dev->addr_len)
457 llinfo = 0;
459 len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
460 if (llinfo)
461 len += ndisc_opt_addr_space(dev);
463 skb = sock_alloc_send_skb(sk,
464 (MAX_HEADER + sizeof(struct ipv6hdr) +
465 len + LL_ALLOCATED_SPACE(dev)),
466 1, &err);
467 if (!skb) {
468 ND_PRINTK0(KERN_ERR
469 "ICMPv6 ND: %s() failed to allocate an skb, err=%d.\n",
470 __func__, err);
471 return NULL;
474 skb_reserve(skb, LL_RESERVED_SPACE(dev));
475 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
477 skb->transport_header = skb->tail;
478 skb_put(skb, len);
480 hdr = (struct icmp6hdr *)skb_transport_header(skb);
481 memcpy(hdr, icmp6h, sizeof(*hdr));
483 opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
484 if (target) {
485 ipv6_addr_copy((struct in6_addr *)opt, target);
486 opt += sizeof(*target);
489 if (llinfo)
490 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
491 dev->addr_len, dev->type);
493 hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
494 IPPROTO_ICMPV6,
495 csum_partial(hdr,
496 len, 0));
498 return skb;
501 EXPORT_SYMBOL(ndisc_build_skb);
503 void ndisc_send_skb(struct sk_buff *skb,
504 struct net_device *dev,
505 struct neighbour *neigh,
506 const struct in6_addr *daddr,
507 const struct in6_addr *saddr,
508 struct icmp6hdr *icmp6h)
510 struct flowi6 fl6;
511 struct dst_entry *dst;
512 struct net *net = dev_net(dev);
513 struct sock *sk = net->ipv6.ndisc_sk;
514 struct inet6_dev *idev;
515 int err;
516 u8 type;
518 type = icmp6h->icmp6_type;
520 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, dev->ifindex);
522 dst = icmp6_dst_alloc(dev, neigh, daddr);
523 if (!dst) {
524 kfree_skb(skb);
525 return;
528 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
529 if (IS_ERR(dst)) {
530 kfree_skb(skb);
531 return;
534 skb_dst_set(skb, dst);
536 idev = in6_dev_get(dst->dev);
537 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
539 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
540 dst_output);
541 if (!err) {
542 ICMP6MSGOUT_INC_STATS(net, idev, type);
543 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
546 if (likely(idev != NULL))
547 in6_dev_put(idev);
550 EXPORT_SYMBOL(ndisc_send_skb);
553 * Send a Neighbour Discover packet
555 static void __ndisc_send(struct net_device *dev,
556 struct neighbour *neigh,
557 const struct in6_addr *daddr,
558 const struct in6_addr *saddr,
559 struct icmp6hdr *icmp6h, const struct in6_addr *target,
560 int llinfo)
562 struct sk_buff *skb;
564 skb = ndisc_build_skb(dev, daddr, saddr, icmp6h, target, llinfo);
565 if (!skb)
566 return;
568 ndisc_send_skb(skb, dev, neigh, daddr, saddr, icmp6h);
571 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
572 const struct in6_addr *daddr,
573 const struct in6_addr *solicited_addr,
574 int router, int solicited, int override, int inc_opt)
576 struct in6_addr tmpaddr;
577 struct inet6_ifaddr *ifp;
578 const struct in6_addr *src_addr;
579 struct icmp6hdr icmp6h = {
580 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
583 /* for anycast or proxy, solicited_addr != src_addr */
584 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
585 if (ifp) {
586 src_addr = solicited_addr;
587 if (ifp->flags & IFA_F_OPTIMISTIC)
588 override = 0;
589 inc_opt |= ifp->idev->cnf.force_tllao;
590 in6_ifa_put(ifp);
591 } else {
592 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
593 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
594 &tmpaddr))
595 return;
596 src_addr = &tmpaddr;
599 icmp6h.icmp6_router = router;
600 icmp6h.icmp6_solicited = solicited;
601 icmp6h.icmp6_override = override;
603 __ndisc_send(dev, neigh, daddr, src_addr,
604 &icmp6h, solicited_addr,
605 inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
608 static void ndisc_send_unsol_na(struct net_device *dev)
610 struct inet6_dev *idev;
611 struct inet6_ifaddr *ifa;
612 struct in6_addr mcaddr;
614 idev = in6_dev_get(dev);
615 if (!idev)
616 return;
618 read_lock_bh(&idev->lock);
619 list_for_each_entry(ifa, &idev->addr_list, if_list) {
620 addrconf_addr_solict_mult(&ifa->addr, &mcaddr);
621 ndisc_send_na(dev, NULL, &mcaddr, &ifa->addr,
622 /*router=*/ !!idev->cnf.forwarding,
623 /*solicited=*/ false, /*override=*/ true,
624 /*inc_opt=*/ true);
626 read_unlock_bh(&idev->lock);
628 in6_dev_put(idev);
631 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
632 const struct in6_addr *solicit,
633 const struct in6_addr *daddr, const struct in6_addr *saddr)
635 struct in6_addr addr_buf;
636 struct icmp6hdr icmp6h = {
637 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
640 if (saddr == NULL) {
641 if (ipv6_get_lladdr(dev, &addr_buf,
642 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
643 return;
644 saddr = &addr_buf;
647 __ndisc_send(dev, neigh, daddr, saddr,
648 &icmp6h, solicit,
649 !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
652 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
653 const struct in6_addr *daddr)
655 struct icmp6hdr icmp6h = {
656 .icmp6_type = NDISC_ROUTER_SOLICITATION,
658 int send_sllao = dev->addr_len;
660 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
662 * According to section 2.2 of RFC 4429, we must not
663 * send router solicitations with a sllao from
664 * optimistic addresses, but we may send the solicitation
665 * if we don't include the sllao. So here we check
666 * if our address is optimistic, and if so, we
667 * suppress the inclusion of the sllao.
669 if (send_sllao) {
670 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
671 dev, 1);
672 if (ifp) {
673 if (ifp->flags & IFA_F_OPTIMISTIC) {
674 send_sllao = 0;
676 in6_ifa_put(ifp);
677 } else {
678 send_sllao = 0;
681 #endif
682 __ndisc_send(dev, NULL, daddr, saddr,
683 &icmp6h, NULL,
684 send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
688 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
691 * "The sender MUST return an ICMP
692 * destination unreachable"
694 dst_link_failure(skb);
695 kfree_skb(skb);
698 /* Called with locked neigh: either read or both */
700 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
702 struct in6_addr *saddr = NULL;
703 struct in6_addr mcaddr;
704 struct net_device *dev = neigh->dev;
705 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
706 int probes = atomic_read(&neigh->probes);
708 if (skb && ipv6_chk_addr(dev_net(dev), &ipv6_hdr(skb)->saddr, dev, 1))
709 saddr = &ipv6_hdr(skb)->saddr;
711 if ((probes -= neigh->parms->ucast_probes) < 0) {
712 if (!(neigh->nud_state & NUD_VALID)) {
713 ND_PRINTK1(KERN_DEBUG "%s(): trying to ucast probe in NUD_INVALID: %pI6\n",
714 __func__, target);
716 ndisc_send_ns(dev, neigh, target, target, saddr);
717 } else if ((probes -= neigh->parms->app_probes) < 0) {
718 #ifdef CONFIG_ARPD
719 neigh_app_ns(neigh);
720 #endif
721 } else {
722 addrconf_addr_solict_mult(target, &mcaddr);
723 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
727 static int pndisc_is_router(const void *pkey,
728 struct net_device *dev)
730 struct pneigh_entry *n;
731 int ret = -1;
733 read_lock_bh(&nd_tbl.lock);
734 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
735 if (n)
736 ret = !!(n->flags & NTF_ROUTER);
737 read_unlock_bh(&nd_tbl.lock);
739 return ret;
742 static void ndisc_recv_ns(struct sk_buff *skb)
744 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
745 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
746 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
747 u8 *lladdr = NULL;
748 u32 ndoptlen = skb->tail - (skb->transport_header +
749 offsetof(struct nd_msg, opt));
750 struct ndisc_options ndopts;
751 struct net_device *dev = skb->dev;
752 struct inet6_ifaddr *ifp;
753 struct inet6_dev *idev = NULL;
754 struct neighbour *neigh;
755 int dad = ipv6_addr_any(saddr);
756 int inc;
757 int is_router = -1;
759 if (ipv6_addr_is_multicast(&msg->target)) {
760 ND_PRINTK2(KERN_WARNING
761 "ICMPv6 NS: multicast target address");
762 return;
766 * RFC2461 7.1.1:
767 * DAD has to be destined for solicited node multicast address.
769 if (dad &&
770 !(daddr->s6_addr32[0] == htonl(0xff020000) &&
771 daddr->s6_addr32[1] == htonl(0x00000000) &&
772 daddr->s6_addr32[2] == htonl(0x00000001) &&
773 daddr->s6_addr [12] == 0xff )) {
774 ND_PRINTK2(KERN_WARNING
775 "ICMPv6 NS: bad DAD packet (wrong destination)\n");
776 return;
779 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
780 ND_PRINTK2(KERN_WARNING
781 "ICMPv6 NS: invalid ND options\n");
782 return;
785 if (ndopts.nd_opts_src_lladdr) {
786 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
787 if (!lladdr) {
788 ND_PRINTK2(KERN_WARNING
789 "ICMPv6 NS: invalid link-layer address length\n");
790 return;
793 /* RFC2461 7.1.1:
794 * If the IP source address is the unspecified address,
795 * there MUST NOT be source link-layer address option
796 * in the message.
798 if (dad) {
799 ND_PRINTK2(KERN_WARNING
800 "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
801 return;
805 inc = ipv6_addr_is_multicast(daddr);
807 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
808 if (ifp) {
810 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
811 if (dad) {
812 if (dev->type == ARPHRD_IEEE802_TR) {
813 const unsigned char *sadr;
814 sadr = skb_mac_header(skb);
815 if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
816 sadr[9] == dev->dev_addr[1] &&
817 sadr[10] == dev->dev_addr[2] &&
818 sadr[11] == dev->dev_addr[3] &&
819 sadr[12] == dev->dev_addr[4] &&
820 sadr[13] == dev->dev_addr[5]) {
821 /* looped-back to us */
822 goto out;
827 * We are colliding with another node
828 * who is doing DAD
829 * so fail our DAD process
831 addrconf_dad_failure(ifp);
832 return;
833 } else {
835 * This is not a dad solicitation.
836 * If we are an optimistic node,
837 * we should respond.
838 * Otherwise, we should ignore it.
840 if (!(ifp->flags & IFA_F_OPTIMISTIC))
841 goto out;
845 idev = ifp->idev;
846 } else {
847 struct net *net = dev_net(dev);
849 idev = in6_dev_get(dev);
850 if (!idev) {
851 /* XXX: count this drop? */
852 return;
855 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
856 (idev->cnf.forwarding &&
857 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
858 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
859 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
860 skb->pkt_type != PACKET_HOST &&
861 inc != 0 &&
862 idev->nd_parms->proxy_delay != 0) {
864 * for anycast or proxy,
865 * sender should delay its response
866 * by a random time between 0 and
867 * MAX_ANYCAST_DELAY_TIME seconds.
868 * (RFC2461) -- yoshfuji
870 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
871 if (n)
872 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
873 goto out;
875 } else
876 goto out;
879 if (is_router < 0)
880 is_router = !!idev->cnf.forwarding;
882 if (dad) {
883 ndisc_send_na(dev, NULL, &in6addr_linklocal_allnodes, &msg->target,
884 is_router, 0, (ifp != NULL), 1);
885 goto out;
888 if (inc)
889 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
890 else
891 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
894 * update / create cache entry
895 * for the source address
897 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
898 !inc || lladdr || !dev->addr_len);
899 if (neigh)
900 neigh_update(neigh, lladdr, NUD_STALE,
901 NEIGH_UPDATE_F_WEAK_OVERRIDE|
902 NEIGH_UPDATE_F_OVERRIDE);
903 if (neigh || !dev->header_ops) {
904 ndisc_send_na(dev, neigh, saddr, &msg->target,
905 is_router,
906 1, (ifp != NULL && inc), inc);
907 if (neigh)
908 neigh_release(neigh);
911 out:
912 if (ifp)
913 in6_ifa_put(ifp);
914 else
915 in6_dev_put(idev);
918 static void ndisc_recv_na(struct sk_buff *skb)
920 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
921 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
922 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
923 u8 *lladdr = NULL;
924 u32 ndoptlen = skb->tail - (skb->transport_header +
925 offsetof(struct nd_msg, opt));
926 struct ndisc_options ndopts;
927 struct net_device *dev = skb->dev;
928 struct inet6_ifaddr *ifp;
929 struct neighbour *neigh;
931 if (skb->len < sizeof(struct nd_msg)) {
932 ND_PRINTK2(KERN_WARNING
933 "ICMPv6 NA: packet too short\n");
934 return;
937 if (ipv6_addr_is_multicast(&msg->target)) {
938 ND_PRINTK2(KERN_WARNING
939 "ICMPv6 NA: target address is multicast.\n");
940 return;
943 if (ipv6_addr_is_multicast(daddr) &&
944 msg->icmph.icmp6_solicited) {
945 ND_PRINTK2(KERN_WARNING
946 "ICMPv6 NA: solicited NA is multicasted.\n");
947 return;
950 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
951 ND_PRINTK2(KERN_WARNING
952 "ICMPv6 NS: invalid ND option\n");
953 return;
955 if (ndopts.nd_opts_tgt_lladdr) {
956 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
957 if (!lladdr) {
958 ND_PRINTK2(KERN_WARNING
959 "ICMPv6 NA: invalid link-layer address length\n");
960 return;
963 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
964 if (ifp) {
965 if (skb->pkt_type != PACKET_LOOPBACK
966 && (ifp->flags & IFA_F_TENTATIVE)) {
967 addrconf_dad_failure(ifp);
968 return;
970 /* What should we make now? The advertisement
971 is invalid, but ndisc specs say nothing
972 about it. It could be misconfiguration, or
973 an smart proxy agent tries to help us :-)
975 We should not print the error if NA has been
976 received from loopback - it is just our own
977 unsolicited advertisement.
979 if (skb->pkt_type != PACKET_LOOPBACK)
980 ND_PRINTK1(KERN_WARNING
981 "ICMPv6 NA: someone advertises our address %pI6 on %s!\n",
982 &ifp->addr, ifp->idev->dev->name);
983 in6_ifa_put(ifp);
984 return;
986 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
988 if (neigh) {
989 u8 old_flags = neigh->flags;
990 struct net *net = dev_net(dev);
992 if (neigh->nud_state & NUD_FAILED)
993 goto out;
996 * Don't update the neighbor cache entry on a proxy NA from
997 * ourselves because either the proxied node is off link or it
998 * has already sent a NA to us.
1000 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1001 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1002 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1003 /* XXX: idev->cnf.prixy_ndp */
1004 goto out;
1007 neigh_update(neigh, lladdr,
1008 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1009 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1010 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1011 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1012 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
1014 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1016 * Change: router to host
1018 struct rt6_info *rt;
1019 rt = rt6_get_dflt_router(saddr, dev);
1020 if (rt)
1021 ip6_del_rt(rt);
1024 out:
1025 neigh_release(neigh);
1029 static void ndisc_recv_rs(struct sk_buff *skb)
1031 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1032 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1033 struct neighbour *neigh;
1034 struct inet6_dev *idev;
1035 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1036 struct ndisc_options ndopts;
1037 u8 *lladdr = NULL;
1039 if (skb->len < sizeof(*rs_msg))
1040 return;
1042 idev = in6_dev_get(skb->dev);
1043 if (!idev) {
1044 if (net_ratelimit())
1045 ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
1046 return;
1049 /* Don't accept RS if we're not in router mode */
1050 if (!idev->cnf.forwarding)
1051 goto out;
1054 * Don't update NCE if src = ::;
1055 * this implies that the source node has no ip address assigned yet.
1057 if (ipv6_addr_any(saddr))
1058 goto out;
1060 /* Parse ND options */
1061 if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1062 if (net_ratelimit())
1063 ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1064 goto out;
1067 if (ndopts.nd_opts_src_lladdr) {
1068 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1069 skb->dev);
1070 if (!lladdr)
1071 goto out;
1074 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1075 if (neigh) {
1076 neigh_update(neigh, lladdr, NUD_STALE,
1077 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1078 NEIGH_UPDATE_F_OVERRIDE|
1079 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1080 neigh_release(neigh);
1082 out:
1083 in6_dev_put(idev);
1086 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1088 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1089 struct sk_buff *skb;
1090 struct nlmsghdr *nlh;
1091 struct nduseroptmsg *ndmsg;
1092 struct net *net = dev_net(ra->dev);
1093 int err;
1094 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1095 + (opt->nd_opt_len << 3));
1096 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1098 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1099 if (skb == NULL) {
1100 err = -ENOBUFS;
1101 goto errout;
1104 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1105 if (nlh == NULL) {
1106 goto nla_put_failure;
1109 ndmsg = nlmsg_data(nlh);
1110 ndmsg->nduseropt_family = AF_INET6;
1111 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1112 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1113 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1114 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1116 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1118 NLA_PUT(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
1119 &ipv6_hdr(ra)->saddr);
1120 nlmsg_end(skb, nlh);
1122 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1123 return;
1125 nla_put_failure:
1126 nlmsg_free(skb);
1127 err = -EMSGSIZE;
1128 errout:
1129 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1132 static inline int accept_ra(struct inet6_dev *in6_dev)
1135 * If forwarding is enabled, RA are not accepted unless the special
1136 * hybrid mode (accept_ra=2) is enabled.
1138 if (in6_dev->cnf.forwarding && in6_dev->cnf.accept_ra < 2)
1139 return 0;
1141 return in6_dev->cnf.accept_ra;
1144 static void ndisc_router_discovery(struct sk_buff *skb)
1146 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1147 struct neighbour *neigh = NULL;
1148 struct inet6_dev *in6_dev;
1149 struct rt6_info *rt = NULL;
1150 int lifetime;
1151 struct ndisc_options ndopts;
1152 int optlen;
1153 unsigned int pref = 0;
1155 __u8 * opt = (__u8 *)(ra_msg + 1);
1157 optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1159 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1160 ND_PRINTK2(KERN_WARNING
1161 "ICMPv6 RA: source address is not link-local.\n");
1162 return;
1164 if (optlen < 0) {
1165 ND_PRINTK2(KERN_WARNING
1166 "ICMPv6 RA: packet too short\n");
1167 return;
1170 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1171 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1172 ND_PRINTK2(KERN_WARNING
1173 "ICMPv6 RA: from host or unauthorized router\n");
1174 return;
1176 #endif
1179 * set the RA_RECV flag in the interface
1182 in6_dev = in6_dev_get(skb->dev);
1183 if (in6_dev == NULL) {
1184 ND_PRINTK0(KERN_ERR
1185 "ICMPv6 RA: can't find inet6 device for %s.\n",
1186 skb->dev->name);
1187 return;
1190 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1191 in6_dev_put(in6_dev);
1192 ND_PRINTK2(KERN_WARNING
1193 "ICMP6 RA: invalid ND options\n");
1194 return;
1197 if (!accept_ra(in6_dev))
1198 goto skip_linkparms;
1200 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1201 /* skip link-specific parameters from interior routers */
1202 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1203 goto skip_linkparms;
1204 #endif
1206 if (in6_dev->if_flags & IF_RS_SENT) {
1208 * flag that an RA was received after an RS was sent
1209 * out on this interface.
1211 in6_dev->if_flags |= IF_RA_RCVD;
1215 * Remember the managed/otherconf flags from most recently
1216 * received RA message (RFC 2462) -- yoshfuji
1218 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1219 IF_RA_OTHERCONF)) |
1220 (ra_msg->icmph.icmp6_addrconf_managed ?
1221 IF_RA_MANAGED : 0) |
1222 (ra_msg->icmph.icmp6_addrconf_other ?
1223 IF_RA_OTHERCONF : 0);
1225 if (!in6_dev->cnf.accept_ra_defrtr)
1226 goto skip_defrtr;
1228 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1230 #ifdef CONFIG_IPV6_ROUTER_PREF
1231 pref = ra_msg->icmph.icmp6_router_pref;
1232 /* 10b is handled as if it were 00b (medium) */
1233 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1234 !in6_dev->cnf.accept_ra_rtr_pref)
1235 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1236 #endif
1238 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1240 if (rt)
1241 neigh = dst_get_neighbour(&rt->dst);
1243 if (rt && lifetime == 0) {
1244 neigh_clone(neigh);
1245 ip6_del_rt(rt);
1246 rt = NULL;
1249 if (rt == NULL && lifetime) {
1250 ND_PRINTK3(KERN_DEBUG
1251 "ICMPv6 RA: adding default router.\n");
1253 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1254 if (rt == NULL) {
1255 ND_PRINTK0(KERN_ERR
1256 "ICMPv6 RA: %s() failed to add default route.\n",
1257 __func__);
1258 in6_dev_put(in6_dev);
1259 return;
1262 neigh = dst_get_neighbour(&rt->dst);
1263 if (neigh == NULL) {
1264 ND_PRINTK0(KERN_ERR
1265 "ICMPv6 RA: %s() got default router without neighbour.\n",
1266 __func__);
1267 dst_release(&rt->dst);
1268 in6_dev_put(in6_dev);
1269 return;
1271 neigh->flags |= NTF_ROUTER;
1272 } else if (rt) {
1273 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1276 if (rt)
1277 rt->rt6i_expires = jiffies + (HZ * lifetime);
1279 if (ra_msg->icmph.icmp6_hop_limit) {
1280 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1281 if (rt)
1282 dst_metric_set(&rt->dst, RTAX_HOPLIMIT,
1283 ra_msg->icmph.icmp6_hop_limit);
1286 skip_defrtr:
1289 * Update Reachable Time and Retrans Timer
1292 if (in6_dev->nd_parms) {
1293 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1295 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1296 rtime = (rtime*HZ)/1000;
1297 if (rtime < HZ/10)
1298 rtime = HZ/10;
1299 in6_dev->nd_parms->retrans_time = rtime;
1300 in6_dev->tstamp = jiffies;
1301 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1304 rtime = ntohl(ra_msg->reachable_time);
1305 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1306 rtime = (rtime*HZ)/1000;
1308 if (rtime < HZ/10)
1309 rtime = HZ/10;
1311 if (rtime != in6_dev->nd_parms->base_reachable_time) {
1312 in6_dev->nd_parms->base_reachable_time = rtime;
1313 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1314 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1315 in6_dev->tstamp = jiffies;
1316 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1321 skip_linkparms:
1324 * Process options.
1327 if (!neigh)
1328 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1329 skb->dev, 1);
1330 if (neigh) {
1331 u8 *lladdr = NULL;
1332 if (ndopts.nd_opts_src_lladdr) {
1333 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1334 skb->dev);
1335 if (!lladdr) {
1336 ND_PRINTK2(KERN_WARNING
1337 "ICMPv6 RA: invalid link-layer address length\n");
1338 goto out;
1341 neigh_update(neigh, lladdr, NUD_STALE,
1342 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1343 NEIGH_UPDATE_F_OVERRIDE|
1344 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1345 NEIGH_UPDATE_F_ISROUTER);
1348 if (!accept_ra(in6_dev))
1349 goto out;
1351 #ifdef CONFIG_IPV6_ROUTE_INFO
1352 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1353 struct nd_opt_hdr *p;
1354 for (p = ndopts.nd_opts_ri;
1356 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1357 struct route_info *ri = (struct route_info *)p;
1358 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1359 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1360 ri->prefix_len == 0)
1361 continue;
1362 #endif
1363 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1364 continue;
1365 rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1366 &ipv6_hdr(skb)->saddr);
1369 #endif
1371 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1372 /* skip link-specific ndopts from interior routers */
1373 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1374 goto out;
1375 #endif
1377 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1378 struct nd_opt_hdr *p;
1379 for (p = ndopts.nd_opts_pi;
1381 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1382 addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1386 if (ndopts.nd_opts_mtu) {
1387 __be32 n;
1388 u32 mtu;
1390 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1391 mtu = ntohl(n);
1393 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1394 ND_PRINTK2(KERN_WARNING
1395 "ICMPv6 RA: invalid mtu: %d\n",
1396 mtu);
1397 } else if (in6_dev->cnf.mtu6 != mtu) {
1398 in6_dev->cnf.mtu6 = mtu;
1400 if (rt)
1401 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
1403 rt6_mtu_change(skb->dev, mtu);
1407 if (ndopts.nd_useropts) {
1408 struct nd_opt_hdr *p;
1409 for (p = ndopts.nd_useropts;
1411 p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1412 ndisc_ra_useropt(skb, p);
1416 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1417 ND_PRINTK2(KERN_WARNING
1418 "ICMPv6 RA: invalid RA options");
1420 out:
1421 if (rt)
1422 dst_release(&rt->dst);
1423 else if (neigh)
1424 neigh_release(neigh);
1425 in6_dev_put(in6_dev);
1428 static void ndisc_redirect_rcv(struct sk_buff *skb)
1430 struct inet6_dev *in6_dev;
1431 struct icmp6hdr *icmph;
1432 const struct in6_addr *dest;
1433 const struct in6_addr *target; /* new first hop to destination */
1434 struct neighbour *neigh;
1435 int on_link = 0;
1436 struct ndisc_options ndopts;
1437 int optlen;
1438 u8 *lladdr = NULL;
1440 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1441 switch (skb->ndisc_nodetype) {
1442 case NDISC_NODETYPE_HOST:
1443 case NDISC_NODETYPE_NODEFAULT:
1444 ND_PRINTK2(KERN_WARNING
1445 "ICMPv6 Redirect: from host or unauthorized router\n");
1446 return;
1448 #endif
1450 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1451 ND_PRINTK2(KERN_WARNING
1452 "ICMPv6 Redirect: source address is not link-local.\n");
1453 return;
1456 optlen = skb->tail - skb->transport_header;
1457 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1459 if (optlen < 0) {
1460 ND_PRINTK2(KERN_WARNING
1461 "ICMPv6 Redirect: packet too short\n");
1462 return;
1465 icmph = icmp6_hdr(skb);
1466 target = (const struct in6_addr *) (icmph + 1);
1467 dest = target + 1;
1469 if (ipv6_addr_is_multicast(dest)) {
1470 ND_PRINTK2(KERN_WARNING
1471 "ICMPv6 Redirect: destination address is multicast.\n");
1472 return;
1475 if (ipv6_addr_equal(dest, target)) {
1476 on_link = 1;
1477 } else if (ipv6_addr_type(target) !=
1478 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1479 ND_PRINTK2(KERN_WARNING
1480 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1481 return;
1484 in6_dev = in6_dev_get(skb->dev);
1485 if (!in6_dev)
1486 return;
1487 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1488 in6_dev_put(in6_dev);
1489 return;
1492 /* RFC2461 8.1:
1493 * The IP source address of the Redirect MUST be the same as the current
1494 * first-hop router for the specified ICMP Destination Address.
1497 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1498 ND_PRINTK2(KERN_WARNING
1499 "ICMPv6 Redirect: invalid ND options\n");
1500 in6_dev_put(in6_dev);
1501 return;
1503 if (ndopts.nd_opts_tgt_lladdr) {
1504 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1505 skb->dev);
1506 if (!lladdr) {
1507 ND_PRINTK2(KERN_WARNING
1508 "ICMPv6 Redirect: invalid link-layer address length\n");
1509 in6_dev_put(in6_dev);
1510 return;
1514 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1515 if (neigh) {
1516 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1517 &ipv6_hdr(skb)->saddr, neigh, lladdr,
1518 on_link);
1519 neigh_release(neigh);
1521 in6_dev_put(in6_dev);
1524 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1525 const struct in6_addr *target)
1527 struct net_device *dev = skb->dev;
1528 struct net *net = dev_net(dev);
1529 struct sock *sk = net->ipv6.ndisc_sk;
1530 int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1531 struct sk_buff *buff;
1532 struct icmp6hdr *icmph;
1533 struct in6_addr saddr_buf;
1534 struct in6_addr *addrp;
1535 struct rt6_info *rt;
1536 struct dst_entry *dst;
1537 struct inet6_dev *idev;
1538 struct flowi6 fl6;
1539 u8 *opt;
1540 int rd_len;
1541 int err;
1542 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1544 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1545 ND_PRINTK2(KERN_WARNING
1546 "ICMPv6 Redirect: no link-local address on %s\n",
1547 dev->name);
1548 return;
1551 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1552 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1553 ND_PRINTK2(KERN_WARNING
1554 "ICMPv6 Redirect: target address is not link-local unicast.\n");
1555 return;
1558 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1559 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1561 dst = ip6_route_output(net, NULL, &fl6);
1562 if (dst == NULL)
1563 return;
1565 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1566 if (IS_ERR(dst))
1567 return;
1569 rt = (struct rt6_info *) dst;
1571 if (rt->rt6i_flags & RTF_GATEWAY) {
1572 ND_PRINTK2(KERN_WARNING
1573 "ICMPv6 Redirect: destination is not a neighbour.\n");
1574 goto release;
1576 if (!rt->rt6i_peer)
1577 rt6_bind_peer(rt, 1);
1578 if (inet_peer_xrlim_allow(rt->rt6i_peer, 1*HZ))
1579 goto release;
1581 if (dev->addr_len) {
1582 read_lock_bh(&neigh->lock);
1583 if (neigh->nud_state & NUD_VALID) {
1584 memcpy(ha_buf, neigh->ha, dev->addr_len);
1585 read_unlock_bh(&neigh->lock);
1586 ha = ha_buf;
1587 len += ndisc_opt_addr_space(dev);
1588 } else
1589 read_unlock_bh(&neigh->lock);
1592 rd_len = min_t(unsigned int,
1593 IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1594 rd_len &= ~0x7;
1595 len += rd_len;
1597 buff = sock_alloc_send_skb(sk,
1598 (MAX_HEADER + sizeof(struct ipv6hdr) +
1599 len + LL_ALLOCATED_SPACE(dev)),
1600 1, &err);
1601 if (buff == NULL) {
1602 ND_PRINTK0(KERN_ERR
1603 "ICMPv6 Redirect: %s() failed to allocate an skb, err=%d.\n",
1604 __func__, err);
1605 goto release;
1608 skb_reserve(buff, LL_RESERVED_SPACE(dev));
1609 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1610 IPPROTO_ICMPV6, len);
1612 skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1613 skb_put(buff, len);
1614 icmph = icmp6_hdr(buff);
1616 memset(icmph, 0, sizeof(struct icmp6hdr));
1617 icmph->icmp6_type = NDISC_REDIRECT;
1620 * copy target and destination addresses
1623 addrp = (struct in6_addr *)(icmph + 1);
1624 ipv6_addr_copy(addrp, target);
1625 addrp++;
1626 ipv6_addr_copy(addrp, &ipv6_hdr(skb)->daddr);
1628 opt = (u8*) (addrp + 1);
1631 * include target_address option
1634 if (ha)
1635 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1636 dev->addr_len, dev->type);
1639 * build redirect option and copy skb over to the new packet.
1642 memset(opt, 0, 8);
1643 *(opt++) = ND_OPT_REDIRECT_HDR;
1644 *(opt++) = (rd_len >> 3);
1645 opt += 6;
1647 memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1649 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1650 len, IPPROTO_ICMPV6,
1651 csum_partial(icmph, len, 0));
1653 skb_dst_set(buff, dst);
1654 idev = in6_dev_get(dst->dev);
1655 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1656 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, buff, NULL, dst->dev,
1657 dst_output);
1658 if (!err) {
1659 ICMP6MSGOUT_INC_STATS(net, idev, NDISC_REDIRECT);
1660 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1663 if (likely(idev != NULL))
1664 in6_dev_put(idev);
1665 return;
1667 release:
1668 dst_release(dst);
1671 static void pndisc_redo(struct sk_buff *skb)
1673 ndisc_recv_ns(skb);
1674 kfree_skb(skb);
1677 int ndisc_rcv(struct sk_buff *skb)
1679 struct nd_msg *msg;
1681 if (!pskb_may_pull(skb, skb->len))
1682 return 0;
1684 msg = (struct nd_msg *)skb_transport_header(skb);
1686 __skb_push(skb, skb->data - skb_transport_header(skb));
1688 if (ipv6_hdr(skb)->hop_limit != 255) {
1689 ND_PRINTK2(KERN_WARNING
1690 "ICMPv6 NDISC: invalid hop-limit: %d\n",
1691 ipv6_hdr(skb)->hop_limit);
1692 return 0;
1695 if (msg->icmph.icmp6_code != 0) {
1696 ND_PRINTK2(KERN_WARNING
1697 "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1698 msg->icmph.icmp6_code);
1699 return 0;
1702 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1704 switch (msg->icmph.icmp6_type) {
1705 case NDISC_NEIGHBOUR_SOLICITATION:
1706 ndisc_recv_ns(skb);
1707 break;
1709 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1710 ndisc_recv_na(skb);
1711 break;
1713 case NDISC_ROUTER_SOLICITATION:
1714 ndisc_recv_rs(skb);
1715 break;
1717 case NDISC_ROUTER_ADVERTISEMENT:
1718 ndisc_router_discovery(skb);
1719 break;
1721 case NDISC_REDIRECT:
1722 ndisc_redirect_rcv(skb);
1723 break;
1726 return 0;
1729 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1731 struct net_device *dev = ptr;
1732 struct net *net = dev_net(dev);
1734 switch (event) {
1735 case NETDEV_CHANGEADDR:
1736 neigh_changeaddr(&nd_tbl, dev);
1737 fib6_run_gc(~0UL, net);
1738 break;
1739 case NETDEV_DOWN:
1740 neigh_ifdown(&nd_tbl, dev);
1741 fib6_run_gc(~0UL, net);
1742 break;
1743 case NETDEV_NOTIFY_PEERS:
1744 ndisc_send_unsol_na(dev);
1745 break;
1746 default:
1747 break;
1750 return NOTIFY_DONE;
1753 static struct notifier_block ndisc_netdev_notifier = {
1754 .notifier_call = ndisc_netdev_event,
1757 #ifdef CONFIG_SYSCTL
1758 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1759 const char *func, const char *dev_name)
1761 static char warncomm[TASK_COMM_LEN];
1762 static int warned;
1763 if (strcmp(warncomm, current->comm) && warned < 5) {
1764 strcpy(warncomm, current->comm);
1765 printk(KERN_WARNING
1766 "process `%s' is using deprecated sysctl (%s) "
1767 "net.ipv6.neigh.%s.%s; "
1768 "Use net.ipv6.neigh.%s.%s_ms "
1769 "instead.\n",
1770 warncomm, func,
1771 dev_name, ctl->procname,
1772 dev_name, ctl->procname);
1773 warned++;
1777 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1779 struct net_device *dev = ctl->extra1;
1780 struct inet6_dev *idev;
1781 int ret;
1783 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1784 (strcmp(ctl->procname, "base_reachable_time") == 0))
1785 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1787 if (strcmp(ctl->procname, "retrans_time") == 0)
1788 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1790 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1791 ret = proc_dointvec_jiffies(ctl, write,
1792 buffer, lenp, ppos);
1794 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1795 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1796 ret = proc_dointvec_ms_jiffies(ctl, write,
1797 buffer, lenp, ppos);
1798 else
1799 ret = -1;
1801 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1802 if (ctl->data == &idev->nd_parms->base_reachable_time)
1803 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1804 idev->tstamp = jiffies;
1805 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1806 in6_dev_put(idev);
1808 return ret;
1812 #endif
1814 static int __net_init ndisc_net_init(struct net *net)
1816 struct ipv6_pinfo *np;
1817 struct sock *sk;
1818 int err;
1820 err = inet_ctl_sock_create(&sk, PF_INET6,
1821 SOCK_RAW, IPPROTO_ICMPV6, net);
1822 if (err < 0) {
1823 ND_PRINTK0(KERN_ERR
1824 "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1825 err);
1826 return err;
1829 net->ipv6.ndisc_sk = sk;
1831 np = inet6_sk(sk);
1832 np->hop_limit = 255;
1833 /* Do not loopback ndisc messages */
1834 np->mc_loop = 0;
1836 return 0;
1839 static void __net_exit ndisc_net_exit(struct net *net)
1841 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1844 static struct pernet_operations ndisc_net_ops = {
1845 .init = ndisc_net_init,
1846 .exit = ndisc_net_exit,
1849 int __init ndisc_init(void)
1851 int err;
1853 err = register_pernet_subsys(&ndisc_net_ops);
1854 if (err)
1855 return err;
1857 * Initialize the neighbour table
1859 neigh_table_init(&nd_tbl);
1861 #ifdef CONFIG_SYSCTL
1862 err = neigh_sysctl_register(NULL, &nd_tbl.parms, "ipv6",
1863 &ndisc_ifinfo_sysctl_change);
1864 if (err)
1865 goto out_unregister_pernet;
1866 #endif
1867 err = register_netdevice_notifier(&ndisc_netdev_notifier);
1868 if (err)
1869 goto out_unregister_sysctl;
1870 out:
1871 return err;
1873 out_unregister_sysctl:
1874 #ifdef CONFIG_SYSCTL
1875 neigh_sysctl_unregister(&nd_tbl.parms);
1876 out_unregister_pernet:
1877 #endif
1878 unregister_pernet_subsys(&ndisc_net_ops);
1879 goto out;
1882 void ndisc_cleanup(void)
1884 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1885 #ifdef CONFIG_SYSCTL
1886 neigh_sysctl_unregister(&nd_tbl.parms);
1887 #endif
1888 neigh_table_clear(&nd_tbl);
1889 unregister_pernet_subsys(&ndisc_net_ops);