[IPV6]: ROUTE: Add support for Router Preference (RFC4191).
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / ndisc.c
blob966ab6b3022e0d2b36167f358860a153424f7c1d
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 * Lars Fenneberg : fixed MTU setting on receipt
19 * of an RA.
21 * Janos Farkas : kmalloc failure checks
22 * Alexey Kuznetsov : state machine reworked
23 * and moved to net/core.
24 * Pekka Savola : RFC2461 validation
25 * YOSHIFUJI Hideaki @USAGI : Verify ND options properly
28 /* Set to 3 to get tracing... */
29 #define ND_DEBUG 1
31 #define ND_PRINTK(fmt, args...) do { if (net_ratelimit()) { printk(fmt, ## args); } } while(0)
32 #define ND_NOPRINTK(x...) do { ; } while(0)
33 #define ND_PRINTK0 ND_PRINTK
34 #define ND_PRINTK1 ND_NOPRINTK
35 #define ND_PRINTK2 ND_NOPRINTK
36 #define ND_PRINTK3 ND_NOPRINTK
37 #if ND_DEBUG >= 1
38 #undef ND_PRINTK1
39 #define ND_PRINTK1 ND_PRINTK
40 #endif
41 #if ND_DEBUG >= 2
42 #undef ND_PRINTK2
43 #define ND_PRINTK2 ND_PRINTK
44 #endif
45 #if ND_DEBUG >= 3
46 #undef ND_PRINTK3
47 #define ND_PRINTK3 ND_PRINTK
48 #endif
50 #include <linux/module.h>
51 #include <linux/config.h>
52 #include <linux/errno.h>
53 #include <linux/types.h>
54 #include <linux/socket.h>
55 #include <linux/sockios.h>
56 #include <linux/sched.h>
57 #include <linux/net.h>
58 #include <linux/in6.h>
59 #include <linux/route.h>
60 #include <linux/init.h>
61 #include <linux/rcupdate.h>
62 #ifdef CONFIG_SYSCTL
63 #include <linux/sysctl.h>
64 #endif
66 #include <linux/if_arp.h>
67 #include <linux/ipv6.h>
68 #include <linux/icmpv6.h>
69 #include <linux/jhash.h>
71 #include <net/sock.h>
72 #include <net/snmp.h>
74 #include <net/ipv6.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
79 #include <net/icmp.h>
81 #include <net/flow.h>
82 #include <net/ip6_checksum.h>
83 #include <linux/proc_fs.h>
85 #include <linux/netfilter.h>
86 #include <linux/netfilter_ipv6.h>
88 static struct socket *ndisc_socket;
90 static u32 ndisc_hash(const void *pkey, const struct net_device *dev);
91 static int ndisc_constructor(struct neighbour *neigh);
92 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
93 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
94 static int pndisc_constructor(struct pneigh_entry *n);
95 static void pndisc_destructor(struct pneigh_entry *n);
96 static void pndisc_redo(struct sk_buff *skb);
98 static struct neigh_ops ndisc_generic_ops = {
99 .family = AF_INET6,
100 .solicit = ndisc_solicit,
101 .error_report = ndisc_error_report,
102 .output = neigh_resolve_output,
103 .connected_output = neigh_connected_output,
104 .hh_output = dev_queue_xmit,
105 .queue_xmit = dev_queue_xmit,
108 static struct neigh_ops ndisc_hh_ops = {
109 .family = AF_INET6,
110 .solicit = ndisc_solicit,
111 .error_report = ndisc_error_report,
112 .output = neigh_resolve_output,
113 .connected_output = neigh_resolve_output,
114 .hh_output = dev_queue_xmit,
115 .queue_xmit = dev_queue_xmit,
119 static struct neigh_ops ndisc_direct_ops = {
120 .family = AF_INET6,
121 .output = dev_queue_xmit,
122 .connected_output = dev_queue_xmit,
123 .hh_output = dev_queue_xmit,
124 .queue_xmit = dev_queue_xmit,
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 = 30 * HZ,
140 .retrans_time = 1 * HZ,
141 .gc_staletime = 60 * HZ,
142 .reachable_time = 30 * HZ,
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_MAX];
162 #define nd_opts_src_lladdr nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
163 #define nd_opts_tgt_lladdr nd_opt_array[ND_OPT_TARGET_LL_ADDR]
164 #define nd_opts_pi nd_opt_array[ND_OPT_PREFIX_INFO]
165 #define nd_opts_pi_end nd_opt_array[__ND_OPT_PREFIX_INFO_END]
166 #define nd_opts_rh nd_opt_array[ND_OPT_REDIRECT_HDR]
167 #define nd_opts_mtu nd_opt_array[ND_OPT_MTU]
169 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
172 * Return the padding between the option length and the start of the
173 * link addr. Currently only IP-over-InfiniBand needs this, although
174 * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
175 * also need a pad of 2.
177 static int ndisc_addr_option_pad(unsigned short type)
179 switch (type) {
180 case ARPHRD_INFINIBAND: return 2;
181 default: return 0;
185 static inline int ndisc_opt_addr_space(struct net_device *dev)
187 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
190 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
191 unsigned short addr_type)
193 int space = NDISC_OPT_SPACE(data_len);
194 int pad = ndisc_addr_option_pad(addr_type);
196 opt[0] = type;
197 opt[1] = space>>3;
199 memset(opt + 2, 0, pad);
200 opt += pad;
201 space -= pad;
203 memcpy(opt+2, data, data_len);
204 data_len += 2;
205 opt += data_len;
206 if ((space -= data_len) > 0)
207 memset(opt, 0, space);
208 return opt + space;
211 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
212 struct nd_opt_hdr *end)
214 int type;
215 if (!cur || !end || cur >= end)
216 return NULL;
217 type = cur->nd_opt_type;
218 do {
219 cur = ((void *)cur) + (cur->nd_opt_len << 3);
220 } while(cur < end && cur->nd_opt_type != type);
221 return (cur <= end && cur->nd_opt_type == type ? cur : NULL);
224 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
225 struct ndisc_options *ndopts)
227 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
229 if (!nd_opt || opt_len < 0 || !ndopts)
230 return NULL;
231 memset(ndopts, 0, sizeof(*ndopts));
232 while (opt_len) {
233 int l;
234 if (opt_len < sizeof(struct nd_opt_hdr))
235 return NULL;
236 l = nd_opt->nd_opt_len << 3;
237 if (opt_len < l || l == 0)
238 return NULL;
239 switch (nd_opt->nd_opt_type) {
240 case ND_OPT_SOURCE_LL_ADDR:
241 case ND_OPT_TARGET_LL_ADDR:
242 case ND_OPT_MTU:
243 case ND_OPT_REDIRECT_HDR:
244 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
245 ND_PRINTK2(KERN_WARNING
246 "%s(): duplicated ND6 option found: type=%d\n",
247 __FUNCTION__,
248 nd_opt->nd_opt_type);
249 } else {
250 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
252 break;
253 case ND_OPT_PREFIX_INFO:
254 ndopts->nd_opts_pi_end = nd_opt;
255 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0)
256 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
257 break;
258 default:
260 * Unknown options must be silently ignored,
261 * to accommodate future extension to the protocol.
263 ND_PRINTK2(KERN_NOTICE
264 "%s(): ignored unsupported option; type=%d, len=%d\n",
265 __FUNCTION__,
266 nd_opt->nd_opt_type, nd_opt->nd_opt_len);
268 opt_len -= l;
269 nd_opt = ((void *)nd_opt) + l;
271 return ndopts;
274 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
275 struct net_device *dev)
277 u8 *lladdr = (u8 *)(p + 1);
278 int lladdrlen = p->nd_opt_len << 3;
279 int prepad = ndisc_addr_option_pad(dev->type);
280 if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
281 return NULL;
282 return (lladdr + prepad);
285 int ndisc_mc_map(struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
287 switch (dev->type) {
288 case ARPHRD_ETHER:
289 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
290 case ARPHRD_FDDI:
291 ipv6_eth_mc_map(addr, buf);
292 return 0;
293 case ARPHRD_IEEE802_TR:
294 ipv6_tr_mc_map(addr,buf);
295 return 0;
296 case ARPHRD_ARCNET:
297 ipv6_arcnet_mc_map(addr, buf);
298 return 0;
299 case ARPHRD_INFINIBAND:
300 ipv6_ib_mc_map(addr, buf);
301 return 0;
302 default:
303 if (dir) {
304 memcpy(buf, dev->broadcast, dev->addr_len);
305 return 0;
308 return -EINVAL;
311 static u32 ndisc_hash(const void *pkey, const struct net_device *dev)
313 const u32 *p32 = pkey;
314 u32 addr_hash, i;
316 addr_hash = 0;
317 for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
318 addr_hash ^= *p32++;
320 return jhash_2words(addr_hash, dev->ifindex, nd_tbl.hash_rnd);
323 static int ndisc_constructor(struct neighbour *neigh)
325 struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
326 struct net_device *dev = neigh->dev;
327 struct inet6_dev *in6_dev;
328 struct neigh_parms *parms;
329 int is_multicast = ipv6_addr_is_multicast(addr);
331 rcu_read_lock();
332 in6_dev = in6_dev_get(dev);
333 if (in6_dev == NULL) {
334 rcu_read_unlock();
335 return -EINVAL;
338 parms = in6_dev->nd_parms;
339 __neigh_parms_put(neigh->parms);
340 neigh->parms = neigh_parms_clone(parms);
341 rcu_read_unlock();
343 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
344 if (dev->hard_header == NULL) {
345 neigh->nud_state = NUD_NOARP;
346 neigh->ops = &ndisc_direct_ops;
347 neigh->output = neigh->ops->queue_xmit;
348 } else {
349 if (is_multicast) {
350 neigh->nud_state = NUD_NOARP;
351 ndisc_mc_map(addr, neigh->ha, dev, 1);
352 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
353 neigh->nud_state = NUD_NOARP;
354 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
355 if (dev->flags&IFF_LOOPBACK)
356 neigh->type = RTN_LOCAL;
357 } else if (dev->flags&IFF_POINTOPOINT) {
358 neigh->nud_state = NUD_NOARP;
359 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
361 if (dev->hard_header_cache)
362 neigh->ops = &ndisc_hh_ops;
363 else
364 neigh->ops = &ndisc_generic_ops;
365 if (neigh->nud_state&NUD_VALID)
366 neigh->output = neigh->ops->connected_output;
367 else
368 neigh->output = neigh->ops->output;
370 in6_dev_put(in6_dev);
371 return 0;
374 static int pndisc_constructor(struct pneigh_entry *n)
376 struct in6_addr *addr = (struct in6_addr*)&n->key;
377 struct in6_addr maddr;
378 struct net_device *dev = n->dev;
380 if (dev == NULL || __in6_dev_get(dev) == NULL)
381 return -EINVAL;
382 addrconf_addr_solict_mult(addr, &maddr);
383 ipv6_dev_mc_inc(dev, &maddr);
384 return 0;
387 static void pndisc_destructor(struct pneigh_entry *n)
389 struct in6_addr *addr = (struct in6_addr*)&n->key;
390 struct in6_addr maddr;
391 struct net_device *dev = n->dev;
393 if (dev == NULL || __in6_dev_get(dev) == NULL)
394 return;
395 addrconf_addr_solict_mult(addr, &maddr);
396 ipv6_dev_mc_dec(dev, &maddr);
400 * Send a Neighbour Advertisement
403 static inline void ndisc_flow_init(struct flowi *fl, u8 type,
404 struct in6_addr *saddr, struct in6_addr *daddr)
406 memset(fl, 0, sizeof(*fl));
407 ipv6_addr_copy(&fl->fl6_src, saddr);
408 ipv6_addr_copy(&fl->fl6_dst, daddr);
409 fl->proto = IPPROTO_ICMPV6;
410 fl->fl_icmp_type = type;
411 fl->fl_icmp_code = 0;
414 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
415 struct in6_addr *daddr, struct in6_addr *solicited_addr,
416 int router, int solicited, int override, int inc_opt)
418 struct in6_addr tmpaddr;
419 struct inet6_ifaddr *ifp;
420 struct inet6_dev *idev;
421 struct flowi fl;
422 struct dst_entry* dst;
423 struct sock *sk = ndisc_socket->sk;
424 struct in6_addr *src_addr;
425 struct nd_msg *msg;
426 int len;
427 struct sk_buff *skb;
428 int err;
430 len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
432 /* for anycast or proxy, solicited_addr != src_addr */
433 ifp = ipv6_get_ifaddr(solicited_addr, dev, 1);
434 if (ifp) {
435 src_addr = solicited_addr;
436 in6_ifa_put(ifp);
437 } else {
438 if (ipv6_dev_get_saddr(dev, daddr, &tmpaddr))
439 return;
440 src_addr = &tmpaddr;
443 ndisc_flow_init(&fl, NDISC_NEIGHBOUR_ADVERTISEMENT, src_addr, daddr);
445 dst = ndisc_dst_alloc(dev, neigh, daddr, ip6_output);
446 if (!dst)
447 return;
449 err = xfrm_lookup(&dst, &fl, NULL, 0);
450 if (err < 0)
451 return;
453 if (inc_opt) {
454 if (dev->addr_len)
455 len += ndisc_opt_addr_space(dev);
456 else
457 inc_opt = 0;
460 skb = sock_alloc_send_skb(sk, MAX_HEADER + len + LL_RESERVED_SPACE(dev),
461 1, &err);
463 if (skb == NULL) {
464 ND_PRINTK0(KERN_ERR
465 "ICMPv6 NA: %s() failed to allocate an skb.\n",
466 __FUNCTION__);
467 dst_release(dst);
468 return;
471 skb_reserve(skb, LL_RESERVED_SPACE(dev));
472 ip6_nd_hdr(sk, skb, dev, src_addr, daddr, IPPROTO_ICMPV6, len);
474 msg = (struct nd_msg *)skb_put(skb, len);
475 skb->h.raw = (unsigned char*)msg;
477 msg->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
478 msg->icmph.icmp6_code = 0;
479 msg->icmph.icmp6_cksum = 0;
481 msg->icmph.icmp6_unused = 0;
482 msg->icmph.icmp6_router = router;
483 msg->icmph.icmp6_solicited = solicited;
484 msg->icmph.icmp6_override = !!override;
486 /* Set the target address. */
487 ipv6_addr_copy(&msg->target, solicited_addr);
489 if (inc_opt)
490 ndisc_fill_addr_option(msg->opt, ND_OPT_TARGET_LL_ADDR, dev->dev_addr,
491 dev->addr_len, dev->type);
493 /* checksum */
494 msg->icmph.icmp6_cksum = csum_ipv6_magic(src_addr, daddr, len,
495 IPPROTO_ICMPV6,
496 csum_partial((__u8 *) msg,
497 len, 0));
499 skb->dst = dst;
500 idev = in6_dev_get(dst->dev);
501 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
502 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
503 if (!err) {
504 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTNEIGHBORADVERTISEMENTS);
505 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
508 if (likely(idev != NULL))
509 in6_dev_put(idev);
512 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
513 struct in6_addr *solicit,
514 struct in6_addr *daddr, struct in6_addr *saddr)
516 struct flowi fl;
517 struct dst_entry* dst;
518 struct inet6_dev *idev;
519 struct sock *sk = ndisc_socket->sk;
520 struct sk_buff *skb;
521 struct nd_msg *msg;
522 struct in6_addr addr_buf;
523 int len;
524 int err;
525 int send_llinfo;
527 if (saddr == NULL) {
528 if (ipv6_get_lladdr(dev, &addr_buf))
529 return;
530 saddr = &addr_buf;
533 ndisc_flow_init(&fl, NDISC_NEIGHBOUR_SOLICITATION, saddr, daddr);
535 dst = ndisc_dst_alloc(dev, neigh, daddr, ip6_output);
536 if (!dst)
537 return;
539 err = xfrm_lookup(&dst, &fl, NULL, 0);
540 if (err < 0)
541 return;
543 len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
544 send_llinfo = dev->addr_len && !ipv6_addr_any(saddr);
545 if (send_llinfo)
546 len += ndisc_opt_addr_space(dev);
548 skb = sock_alloc_send_skb(sk, MAX_HEADER + len + LL_RESERVED_SPACE(dev),
549 1, &err);
550 if (skb == NULL) {
551 ND_PRINTK0(KERN_ERR
552 "ICMPv6 NA: %s() failed to allocate an skb.\n",
553 __FUNCTION__);
554 dst_release(dst);
555 return;
558 skb_reserve(skb, LL_RESERVED_SPACE(dev));
559 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
561 msg = (struct nd_msg *)skb_put(skb, len);
562 skb->h.raw = (unsigned char*)msg;
563 msg->icmph.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION;
564 msg->icmph.icmp6_code = 0;
565 msg->icmph.icmp6_cksum = 0;
566 msg->icmph.icmp6_unused = 0;
568 /* Set the target address. */
569 ipv6_addr_copy(&msg->target, solicit);
571 if (send_llinfo)
572 ndisc_fill_addr_option(msg->opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr,
573 dev->addr_len, dev->type);
575 /* checksum */
576 msg->icmph.icmp6_cksum = csum_ipv6_magic(&skb->nh.ipv6h->saddr,
577 daddr, len,
578 IPPROTO_ICMPV6,
579 csum_partial((__u8 *) msg,
580 len, 0));
581 /* send it! */
582 skb->dst = dst;
583 idev = in6_dev_get(dst->dev);
584 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
585 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
586 if (!err) {
587 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTNEIGHBORSOLICITS);
588 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
591 if (likely(idev != NULL))
592 in6_dev_put(idev);
595 void ndisc_send_rs(struct net_device *dev, struct in6_addr *saddr,
596 struct in6_addr *daddr)
598 struct flowi fl;
599 struct dst_entry* dst;
600 struct inet6_dev *idev;
601 struct sock *sk = ndisc_socket->sk;
602 struct sk_buff *skb;
603 struct icmp6hdr *hdr;
604 __u8 * opt;
605 int len;
606 int err;
608 ndisc_flow_init(&fl, NDISC_ROUTER_SOLICITATION, saddr, daddr);
610 dst = ndisc_dst_alloc(dev, NULL, daddr, ip6_output);
611 if (!dst)
612 return;
614 err = xfrm_lookup(&dst, &fl, NULL, 0);
615 if (err < 0)
616 return;
618 len = sizeof(struct icmp6hdr);
619 if (dev->addr_len)
620 len += ndisc_opt_addr_space(dev);
622 skb = sock_alloc_send_skb(sk, MAX_HEADER + len + LL_RESERVED_SPACE(dev),
623 1, &err);
624 if (skb == NULL) {
625 ND_PRINTK0(KERN_ERR
626 "ICMPv6 RS: %s() failed to allocate an skb.\n",
627 __FUNCTION__);
628 dst_release(dst);
629 return;
632 skb_reserve(skb, LL_RESERVED_SPACE(dev));
633 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
635 hdr = (struct icmp6hdr *)skb_put(skb, len);
636 skb->h.raw = (unsigned char*)hdr;
637 hdr->icmp6_type = NDISC_ROUTER_SOLICITATION;
638 hdr->icmp6_code = 0;
639 hdr->icmp6_cksum = 0;
640 hdr->icmp6_unused = 0;
642 opt = (u8*) (hdr + 1);
644 if (dev->addr_len)
645 ndisc_fill_addr_option(opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr,
646 dev->addr_len, dev->type);
648 /* checksum */
649 hdr->icmp6_cksum = csum_ipv6_magic(&skb->nh.ipv6h->saddr, daddr, len,
650 IPPROTO_ICMPV6,
651 csum_partial((__u8 *) hdr, len, 0));
653 /* send it! */
654 skb->dst = dst;
655 idev = in6_dev_get(dst->dev);
656 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
657 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
658 if (!err) {
659 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTROUTERSOLICITS);
660 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
663 if (likely(idev != NULL))
664 in6_dev_put(idev);
668 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
671 * "The sender MUST return an ICMP
672 * destination unreachable"
674 dst_link_failure(skb);
675 kfree_skb(skb);
678 /* Called with locked neigh: either read or both */
680 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
682 struct in6_addr *saddr = NULL;
683 struct in6_addr mcaddr;
684 struct net_device *dev = neigh->dev;
685 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
686 int probes = atomic_read(&neigh->probes);
688 if (skb && ipv6_chk_addr(&skb->nh.ipv6h->saddr, dev, 1))
689 saddr = &skb->nh.ipv6h->saddr;
691 if ((probes -= neigh->parms->ucast_probes) < 0) {
692 if (!(neigh->nud_state & NUD_VALID)) {
693 ND_PRINTK1(KERN_DEBUG
694 "%s(): trying to ucast probe in NUD_INVALID: "
695 NIP6_FMT "\n",
696 __FUNCTION__,
697 NIP6(*target));
699 ndisc_send_ns(dev, neigh, target, target, saddr);
700 } else if ((probes -= neigh->parms->app_probes) < 0) {
701 #ifdef CONFIG_ARPD
702 neigh_app_ns(neigh);
703 #endif
704 } else {
705 addrconf_addr_solict_mult(target, &mcaddr);
706 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
710 static void ndisc_recv_ns(struct sk_buff *skb)
712 struct nd_msg *msg = (struct nd_msg *)skb->h.raw;
713 struct in6_addr *saddr = &skb->nh.ipv6h->saddr;
714 struct in6_addr *daddr = &skb->nh.ipv6h->daddr;
715 u8 *lladdr = NULL;
716 u32 ndoptlen = skb->tail - msg->opt;
717 struct ndisc_options ndopts;
718 struct net_device *dev = skb->dev;
719 struct inet6_ifaddr *ifp;
720 struct inet6_dev *idev = NULL;
721 struct neighbour *neigh;
722 int dad = ipv6_addr_any(saddr);
723 int inc;
725 if (ipv6_addr_is_multicast(&msg->target)) {
726 ND_PRINTK2(KERN_WARNING
727 "ICMPv6 NS: multicast target address");
728 return;
732 * RFC2461 7.1.1:
733 * DAD has to be destined for solicited node multicast address.
735 if (dad &&
736 !(daddr->s6_addr32[0] == htonl(0xff020000) &&
737 daddr->s6_addr32[1] == htonl(0x00000000) &&
738 daddr->s6_addr32[2] == htonl(0x00000001) &&
739 daddr->s6_addr [12] == 0xff )) {
740 ND_PRINTK2(KERN_WARNING
741 "ICMPv6 NS: bad DAD packet (wrong destination)\n");
742 return;
745 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
746 ND_PRINTK2(KERN_WARNING
747 "ICMPv6 NS: invalid ND options\n");
748 return;
751 if (ndopts.nd_opts_src_lladdr) {
752 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
753 if (!lladdr) {
754 ND_PRINTK2(KERN_WARNING
755 "ICMPv6 NS: invalid link-layer address length\n");
756 return;
759 /* RFC2461 7.1.1:
760 * If the IP source address is the unspecified address,
761 * there MUST NOT be source link-layer address option
762 * in the message.
764 if (dad) {
765 ND_PRINTK2(KERN_WARNING
766 "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
767 return;
771 inc = ipv6_addr_is_multicast(daddr);
773 if ((ifp = ipv6_get_ifaddr(&msg->target, dev, 1)) != NULL) {
774 if (ifp->flags & IFA_F_TENTATIVE) {
775 /* Address is tentative. If the source
776 is unspecified address, it is someone
777 does DAD, otherwise we ignore solicitations
778 until DAD timer expires.
780 if (!dad)
781 goto out;
782 if (dev->type == ARPHRD_IEEE802_TR) {
783 unsigned char *sadr = skb->mac.raw;
784 if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
785 sadr[9] == dev->dev_addr[1] &&
786 sadr[10] == dev->dev_addr[2] &&
787 sadr[11] == dev->dev_addr[3] &&
788 sadr[12] == dev->dev_addr[4] &&
789 sadr[13] == dev->dev_addr[5]) {
790 /* looped-back to us */
791 goto out;
794 addrconf_dad_failure(ifp);
795 return;
798 idev = ifp->idev;
799 } else {
800 idev = in6_dev_get(dev);
801 if (!idev) {
802 /* XXX: count this drop? */
803 return;
806 if (ipv6_chk_acast_addr(dev, &msg->target) ||
807 (idev->cnf.forwarding &&
808 pneigh_lookup(&nd_tbl, &msg->target, dev, 0))) {
809 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
810 skb->pkt_type != PACKET_HOST &&
811 inc != 0 &&
812 idev->nd_parms->proxy_delay != 0) {
814 * for anycast or proxy,
815 * sender should delay its response
816 * by a random time between 0 and
817 * MAX_ANYCAST_DELAY_TIME seconds.
818 * (RFC2461) -- yoshfuji
820 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
821 if (n)
822 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
823 goto out;
825 } else
826 goto out;
829 if (dad) {
830 struct in6_addr maddr;
832 ipv6_addr_all_nodes(&maddr);
833 ndisc_send_na(dev, NULL, &maddr, &msg->target,
834 idev->cnf.forwarding, 0, (ifp != NULL), 1);
835 goto out;
838 if (inc)
839 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
840 else
841 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
844 * update / create cache entry
845 * for the source address
847 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
848 !inc || lladdr || !dev->addr_len);
849 if (neigh)
850 neigh_update(neigh, lladdr, NUD_STALE,
851 NEIGH_UPDATE_F_WEAK_OVERRIDE|
852 NEIGH_UPDATE_F_OVERRIDE);
853 if (neigh || !dev->hard_header) {
854 ndisc_send_na(dev, neigh, saddr, &msg->target,
855 idev->cnf.forwarding,
856 1, (ifp != NULL && inc), inc);
857 if (neigh)
858 neigh_release(neigh);
861 out:
862 if (ifp)
863 in6_ifa_put(ifp);
864 else
865 in6_dev_put(idev);
867 return;
870 static void ndisc_recv_na(struct sk_buff *skb)
872 struct nd_msg *msg = (struct nd_msg *)skb->h.raw;
873 struct in6_addr *saddr = &skb->nh.ipv6h->saddr;
874 struct in6_addr *daddr = &skb->nh.ipv6h->daddr;
875 u8 *lladdr = NULL;
876 u32 ndoptlen = skb->tail - msg->opt;
877 struct ndisc_options ndopts;
878 struct net_device *dev = skb->dev;
879 struct inet6_ifaddr *ifp;
880 struct neighbour *neigh;
882 if (skb->len < sizeof(struct nd_msg)) {
883 ND_PRINTK2(KERN_WARNING
884 "ICMPv6 NA: packet too short\n");
885 return;
888 if (ipv6_addr_is_multicast(&msg->target)) {
889 ND_PRINTK2(KERN_WARNING
890 "ICMPv6 NA: target address is multicast.\n");
891 return;
894 if (ipv6_addr_is_multicast(daddr) &&
895 msg->icmph.icmp6_solicited) {
896 ND_PRINTK2(KERN_WARNING
897 "ICMPv6 NA: solicited NA is multicasted.\n");
898 return;
901 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
902 ND_PRINTK2(KERN_WARNING
903 "ICMPv6 NS: invalid ND option\n");
904 return;
906 if (ndopts.nd_opts_tgt_lladdr) {
907 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
908 if (!lladdr) {
909 ND_PRINTK2(KERN_WARNING
910 "ICMPv6 NA: invalid link-layer address length\n");
911 return;
914 if ((ifp = ipv6_get_ifaddr(&msg->target, dev, 1))) {
915 if (ifp->flags & IFA_F_TENTATIVE) {
916 addrconf_dad_failure(ifp);
917 return;
919 /* What should we make now? The advertisement
920 is invalid, but ndisc specs say nothing
921 about it. It could be misconfiguration, or
922 an smart proxy agent tries to help us :-)
924 ND_PRINTK1(KERN_WARNING
925 "ICMPv6 NA: someone advertises our address on %s!\n",
926 ifp->idev->dev->name);
927 in6_ifa_put(ifp);
928 return;
930 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
932 if (neigh) {
933 u8 old_flags = neigh->flags;
935 if (neigh->nud_state & NUD_FAILED)
936 goto out;
938 neigh_update(neigh, lladdr,
939 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
940 NEIGH_UPDATE_F_WEAK_OVERRIDE|
941 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
942 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
943 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
945 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
947 * Change: router to host
949 struct rt6_info *rt;
950 rt = rt6_get_dflt_router(saddr, dev);
951 if (rt)
952 ip6_del_rt(rt, NULL, NULL, NULL);
955 out:
956 neigh_release(neigh);
960 static void ndisc_recv_rs(struct sk_buff *skb)
962 struct rs_msg *rs_msg = (struct rs_msg *) skb->h.raw;
963 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
964 struct neighbour *neigh;
965 struct inet6_dev *idev;
966 struct in6_addr *saddr = &skb->nh.ipv6h->saddr;
967 struct ndisc_options ndopts;
968 u8 *lladdr = NULL;
970 if (skb->len < sizeof(*rs_msg))
971 return;
973 idev = in6_dev_get(skb->dev);
974 if (!idev) {
975 if (net_ratelimit())
976 ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
977 return;
980 /* Don't accept RS if we're not in router mode */
981 if (!idev->cnf.forwarding)
982 goto out;
985 * Don't update NCE if src = ::;
986 * this implies that the source node has no ip address assigned yet.
988 if (ipv6_addr_any(saddr))
989 goto out;
991 /* Parse ND options */
992 if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
993 if (net_ratelimit())
994 ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
995 goto out;
998 if (ndopts.nd_opts_src_lladdr) {
999 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1000 skb->dev);
1001 if (!lladdr)
1002 goto out;
1005 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1006 if (neigh) {
1007 neigh_update(neigh, lladdr, NUD_STALE,
1008 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1009 NEIGH_UPDATE_F_OVERRIDE|
1010 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1011 neigh_release(neigh);
1013 out:
1014 in6_dev_put(idev);
1017 static void ndisc_router_discovery(struct sk_buff *skb)
1019 struct ra_msg *ra_msg = (struct ra_msg *) skb->h.raw;
1020 struct neighbour *neigh = NULL;
1021 struct inet6_dev *in6_dev;
1022 struct rt6_info *rt = NULL;
1023 int lifetime;
1024 struct ndisc_options ndopts;
1025 int optlen;
1026 unsigned int pref = 0;
1028 __u8 * opt = (__u8 *)(ra_msg + 1);
1030 optlen = (skb->tail - skb->h.raw) - sizeof(struct ra_msg);
1032 if (!(ipv6_addr_type(&skb->nh.ipv6h->saddr) & IPV6_ADDR_LINKLOCAL)) {
1033 ND_PRINTK2(KERN_WARNING
1034 "ICMPv6 RA: source address is not link-local.\n");
1035 return;
1037 if (optlen < 0) {
1038 ND_PRINTK2(KERN_WARNING
1039 "ICMPv6 RA: packet too short\n");
1040 return;
1044 * set the RA_RECV flag in the interface
1047 in6_dev = in6_dev_get(skb->dev);
1048 if (in6_dev == NULL) {
1049 ND_PRINTK0(KERN_ERR
1050 "ICMPv6 RA: can't find inet6 device for %s.\n",
1051 skb->dev->name);
1052 return;
1054 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra) {
1055 in6_dev_put(in6_dev);
1056 return;
1059 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1060 in6_dev_put(in6_dev);
1061 ND_PRINTK2(KERN_WARNING
1062 "ICMP6 RA: invalid ND options\n");
1063 return;
1066 if (in6_dev->if_flags & IF_RS_SENT) {
1068 * flag that an RA was received after an RS was sent
1069 * out on this interface.
1071 in6_dev->if_flags |= IF_RA_RCVD;
1075 * Remember the managed/otherconf flags from most recently
1076 * received RA message (RFC 2462) -- yoshfuji
1078 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1079 IF_RA_OTHERCONF)) |
1080 (ra_msg->icmph.icmp6_addrconf_managed ?
1081 IF_RA_MANAGED : 0) |
1082 (ra_msg->icmph.icmp6_addrconf_other ?
1083 IF_RA_OTHERCONF : 0);
1085 if (!in6_dev->cnf.accept_ra_defrtr)
1086 goto skip_defrtr;
1088 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1090 #ifdef CONFIG_IPV6_ROUTER_PREF
1091 pref = ra_msg->icmph.icmp6_router_pref;
1092 /* 10b is handled as if it were 00b (medium) */
1093 if (pref == ICMPV6_ROUTER_PREF_INVALID)
1094 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1095 #endif
1097 rt = rt6_get_dflt_router(&skb->nh.ipv6h->saddr, skb->dev);
1099 if (rt)
1100 neigh = rt->rt6i_nexthop;
1102 if (rt && lifetime == 0) {
1103 neigh_clone(neigh);
1104 ip6_del_rt(rt, NULL, NULL, NULL);
1105 rt = NULL;
1108 if (rt == NULL && lifetime) {
1109 ND_PRINTK3(KERN_DEBUG
1110 "ICMPv6 RA: adding default router.\n");
1112 rt = rt6_add_dflt_router(&skb->nh.ipv6h->saddr, skb->dev, pref);
1113 if (rt == NULL) {
1114 ND_PRINTK0(KERN_ERR
1115 "ICMPv6 RA: %s() failed to add default route.\n",
1116 __FUNCTION__);
1117 in6_dev_put(in6_dev);
1118 return;
1121 neigh = rt->rt6i_nexthop;
1122 if (neigh == NULL) {
1123 ND_PRINTK0(KERN_ERR
1124 "ICMPv6 RA: %s() got default router without neighbour.\n",
1125 __FUNCTION__);
1126 dst_release(&rt->u.dst);
1127 in6_dev_put(in6_dev);
1128 return;
1130 neigh->flags |= NTF_ROUTER;
1131 } else if (rt) {
1132 rt->rt6i_flags |= (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1135 if (rt)
1136 rt->rt6i_expires = jiffies + (HZ * lifetime);
1138 if (ra_msg->icmph.icmp6_hop_limit) {
1139 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1140 if (rt)
1141 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1144 skip_defrtr:
1147 * Update Reachable Time and Retrans Timer
1150 if (in6_dev->nd_parms) {
1151 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1153 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1154 rtime = (rtime*HZ)/1000;
1155 if (rtime < HZ/10)
1156 rtime = HZ/10;
1157 in6_dev->nd_parms->retrans_time = rtime;
1158 in6_dev->tstamp = jiffies;
1159 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1162 rtime = ntohl(ra_msg->reachable_time);
1163 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1164 rtime = (rtime*HZ)/1000;
1166 if (rtime < HZ/10)
1167 rtime = HZ/10;
1169 if (rtime != in6_dev->nd_parms->base_reachable_time) {
1170 in6_dev->nd_parms->base_reachable_time = rtime;
1171 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1172 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1173 in6_dev->tstamp = jiffies;
1174 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1180 * Process options.
1183 if (!neigh)
1184 neigh = __neigh_lookup(&nd_tbl, &skb->nh.ipv6h->saddr,
1185 skb->dev, 1);
1186 if (neigh) {
1187 u8 *lladdr = NULL;
1188 if (ndopts.nd_opts_src_lladdr) {
1189 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1190 skb->dev);
1191 if (!lladdr) {
1192 ND_PRINTK2(KERN_WARNING
1193 "ICMPv6 RA: invalid link-layer address length\n");
1194 goto out;
1197 neigh_update(neigh, lladdr, NUD_STALE,
1198 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1199 NEIGH_UPDATE_F_OVERRIDE|
1200 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1201 NEIGH_UPDATE_F_ISROUTER);
1204 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1205 struct nd_opt_hdr *p;
1206 for (p = ndopts.nd_opts_pi;
1208 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1209 addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1213 if (ndopts.nd_opts_mtu) {
1214 u32 mtu;
1216 memcpy(&mtu, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1217 mtu = ntohl(mtu);
1219 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1220 ND_PRINTK2(KERN_WARNING
1221 "ICMPv6 RA: invalid mtu: %d\n",
1222 mtu);
1223 } else if (in6_dev->cnf.mtu6 != mtu) {
1224 in6_dev->cnf.mtu6 = mtu;
1226 if (rt)
1227 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
1229 rt6_mtu_change(skb->dev, mtu);
1233 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1234 ND_PRINTK2(KERN_WARNING
1235 "ICMPv6 RA: invalid RA options");
1237 out:
1238 if (rt)
1239 dst_release(&rt->u.dst);
1240 else if (neigh)
1241 neigh_release(neigh);
1242 in6_dev_put(in6_dev);
1245 static void ndisc_redirect_rcv(struct sk_buff *skb)
1247 struct inet6_dev *in6_dev;
1248 struct icmp6hdr *icmph;
1249 struct in6_addr *dest;
1250 struct in6_addr *target; /* new first hop to destination */
1251 struct neighbour *neigh;
1252 int on_link = 0;
1253 struct ndisc_options ndopts;
1254 int optlen;
1255 u8 *lladdr = NULL;
1257 if (!(ipv6_addr_type(&skb->nh.ipv6h->saddr) & IPV6_ADDR_LINKLOCAL)) {
1258 ND_PRINTK2(KERN_WARNING
1259 "ICMPv6 Redirect: source address is not link-local.\n");
1260 return;
1263 optlen = skb->tail - skb->h.raw;
1264 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1266 if (optlen < 0) {
1267 ND_PRINTK2(KERN_WARNING
1268 "ICMPv6 Redirect: packet too short\n");
1269 return;
1272 icmph = (struct icmp6hdr *) skb->h.raw;
1273 target = (struct in6_addr *) (icmph + 1);
1274 dest = target + 1;
1276 if (ipv6_addr_is_multicast(dest)) {
1277 ND_PRINTK2(KERN_WARNING
1278 "ICMPv6 Redirect: destination address is multicast.\n");
1279 return;
1282 if (ipv6_addr_equal(dest, target)) {
1283 on_link = 1;
1284 } else if (!(ipv6_addr_type(target) & IPV6_ADDR_LINKLOCAL)) {
1285 ND_PRINTK2(KERN_WARNING
1286 "ICMPv6 Redirect: target address is not link-local.\n");
1287 return;
1290 in6_dev = in6_dev_get(skb->dev);
1291 if (!in6_dev)
1292 return;
1293 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1294 in6_dev_put(in6_dev);
1295 return;
1298 /* RFC2461 8.1:
1299 * The IP source address of the Redirect MUST be the same as the current
1300 * first-hop router for the specified ICMP Destination Address.
1303 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1304 ND_PRINTK2(KERN_WARNING
1305 "ICMPv6 Redirect: invalid ND options\n");
1306 in6_dev_put(in6_dev);
1307 return;
1309 if (ndopts.nd_opts_tgt_lladdr) {
1310 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1311 skb->dev);
1312 if (!lladdr) {
1313 ND_PRINTK2(KERN_WARNING
1314 "ICMPv6 Redirect: invalid link-layer address length\n");
1315 in6_dev_put(in6_dev);
1316 return;
1320 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1321 if (neigh) {
1322 rt6_redirect(dest, &skb->nh.ipv6h->saddr, neigh, lladdr,
1323 on_link);
1324 neigh_release(neigh);
1326 in6_dev_put(in6_dev);
1329 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1330 struct in6_addr *target)
1332 struct sock *sk = ndisc_socket->sk;
1333 int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1334 struct sk_buff *buff;
1335 struct icmp6hdr *icmph;
1336 struct in6_addr saddr_buf;
1337 struct in6_addr *addrp;
1338 struct net_device *dev;
1339 struct rt6_info *rt;
1340 struct dst_entry *dst;
1341 struct inet6_dev *idev;
1342 struct flowi fl;
1343 u8 *opt;
1344 int rd_len;
1345 int err;
1346 int hlen;
1347 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1349 dev = skb->dev;
1351 if (ipv6_get_lladdr(dev, &saddr_buf)) {
1352 ND_PRINTK2(KERN_WARNING
1353 "ICMPv6 Redirect: no link-local address on %s\n",
1354 dev->name);
1355 return;
1358 ndisc_flow_init(&fl, NDISC_REDIRECT, &saddr_buf, &skb->nh.ipv6h->saddr);
1360 dst = ip6_route_output(NULL, &fl);
1361 if (dst == NULL)
1362 return;
1364 err = xfrm_lookup(&dst, &fl, NULL, 0);
1365 if (err)
1366 return;
1368 rt = (struct rt6_info *) dst;
1370 if (rt->rt6i_flags & RTF_GATEWAY) {
1371 ND_PRINTK2(KERN_WARNING
1372 "ICMPv6 Redirect: destination is not a neighbour.\n");
1373 dst_release(dst);
1374 return;
1376 if (!xrlim_allow(dst, 1*HZ)) {
1377 dst_release(dst);
1378 return;
1381 if (dev->addr_len) {
1382 read_lock_bh(&neigh->lock);
1383 if (neigh->nud_state & NUD_VALID) {
1384 memcpy(ha_buf, neigh->ha, dev->addr_len);
1385 read_unlock_bh(&neigh->lock);
1386 ha = ha_buf;
1387 len += ndisc_opt_addr_space(dev);
1388 } else
1389 read_unlock_bh(&neigh->lock);
1392 rd_len = min_t(unsigned int,
1393 IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1394 rd_len &= ~0x7;
1395 len += rd_len;
1397 buff = sock_alloc_send_skb(sk, MAX_HEADER + len + LL_RESERVED_SPACE(dev),
1398 1, &err);
1399 if (buff == NULL) {
1400 ND_PRINTK0(KERN_ERR
1401 "ICMPv6 Redirect: %s() failed to allocate an skb.\n",
1402 __FUNCTION__);
1403 dst_release(dst);
1404 return;
1407 hlen = 0;
1409 skb_reserve(buff, LL_RESERVED_SPACE(dev));
1410 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &skb->nh.ipv6h->saddr,
1411 IPPROTO_ICMPV6, len);
1413 icmph = (struct icmp6hdr *)skb_put(buff, len);
1414 buff->h.raw = (unsigned char*)icmph;
1416 memset(icmph, 0, sizeof(struct icmp6hdr));
1417 icmph->icmp6_type = NDISC_REDIRECT;
1420 * copy target and destination addresses
1423 addrp = (struct in6_addr *)(icmph + 1);
1424 ipv6_addr_copy(addrp, target);
1425 addrp++;
1426 ipv6_addr_copy(addrp, &skb->nh.ipv6h->daddr);
1428 opt = (u8*) (addrp + 1);
1431 * include target_address option
1434 if (ha)
1435 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1436 dev->addr_len, dev->type);
1439 * build redirect option and copy skb over to the new packet.
1442 memset(opt, 0, 8);
1443 *(opt++) = ND_OPT_REDIRECT_HDR;
1444 *(opt++) = (rd_len >> 3);
1445 opt += 6;
1447 memcpy(opt, skb->nh.ipv6h, rd_len - 8);
1449 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &skb->nh.ipv6h->saddr,
1450 len, IPPROTO_ICMPV6,
1451 csum_partial((u8 *) icmph, len, 0));
1453 buff->dst = dst;
1454 idev = in6_dev_get(dst->dev);
1455 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
1456 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, buff, NULL, dst->dev, dst_output);
1457 if (!err) {
1458 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTREDIRECTS);
1459 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
1462 if (likely(idev != NULL))
1463 in6_dev_put(idev);
1466 static void pndisc_redo(struct sk_buff *skb)
1468 ndisc_recv_ns(skb);
1469 kfree_skb(skb);
1472 int ndisc_rcv(struct sk_buff *skb)
1474 struct nd_msg *msg;
1476 if (!pskb_may_pull(skb, skb->len))
1477 return 0;
1479 msg = (struct nd_msg *) skb->h.raw;
1481 __skb_push(skb, skb->data-skb->h.raw);
1483 if (skb->nh.ipv6h->hop_limit != 255) {
1484 ND_PRINTK2(KERN_WARNING
1485 "ICMPv6 NDISC: invalid hop-limit: %d\n",
1486 skb->nh.ipv6h->hop_limit);
1487 return 0;
1490 if (msg->icmph.icmp6_code != 0) {
1491 ND_PRINTK2(KERN_WARNING
1492 "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1493 msg->icmph.icmp6_code);
1494 return 0;
1497 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1499 switch (msg->icmph.icmp6_type) {
1500 case NDISC_NEIGHBOUR_SOLICITATION:
1501 ndisc_recv_ns(skb);
1502 break;
1504 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1505 ndisc_recv_na(skb);
1506 break;
1508 case NDISC_ROUTER_SOLICITATION:
1509 ndisc_recv_rs(skb);
1510 break;
1512 case NDISC_ROUTER_ADVERTISEMENT:
1513 ndisc_router_discovery(skb);
1514 break;
1516 case NDISC_REDIRECT:
1517 ndisc_redirect_rcv(skb);
1518 break;
1521 return 0;
1524 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1526 struct net_device *dev = ptr;
1528 switch (event) {
1529 case NETDEV_CHANGEADDR:
1530 neigh_changeaddr(&nd_tbl, dev);
1531 fib6_run_gc(~0UL);
1532 break;
1533 case NETDEV_DOWN:
1534 neigh_ifdown(&nd_tbl, dev);
1535 fib6_run_gc(~0UL);
1536 break;
1537 default:
1538 break;
1541 return NOTIFY_DONE;
1544 static struct notifier_block ndisc_netdev_notifier = {
1545 .notifier_call = ndisc_netdev_event,
1548 #ifdef CONFIG_SYSCTL
1549 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1550 const char *func, const char *dev_name)
1552 static char warncomm[TASK_COMM_LEN];
1553 static int warned;
1554 if (strcmp(warncomm, current->comm) && warned < 5) {
1555 strcpy(warncomm, current->comm);
1556 printk(KERN_WARNING
1557 "process `%s' is using deprecated sysctl (%s) "
1558 "net.ipv6.neigh.%s.%s; "
1559 "Use net.ipv6.neigh.%s.%s_ms "
1560 "instead.\n",
1561 warncomm, func,
1562 dev_name, ctl->procname,
1563 dev_name, ctl->procname);
1564 warned++;
1568 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, struct file * filp, void __user *buffer, size_t *lenp, loff_t *ppos)
1570 struct net_device *dev = ctl->extra1;
1571 struct inet6_dev *idev;
1572 int ret;
1574 if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1575 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1576 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1578 switch (ctl->ctl_name) {
1579 case NET_NEIGH_RETRANS_TIME:
1580 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1581 break;
1582 case NET_NEIGH_REACHABLE_TIME:
1583 ret = proc_dointvec_jiffies(ctl, write,
1584 filp, buffer, lenp, ppos);
1585 break;
1586 case NET_NEIGH_RETRANS_TIME_MS:
1587 case NET_NEIGH_REACHABLE_TIME_MS:
1588 ret = proc_dointvec_ms_jiffies(ctl, write,
1589 filp, buffer, lenp, ppos);
1590 break;
1591 default:
1592 ret = -1;
1595 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1596 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1597 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1598 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1599 idev->tstamp = jiffies;
1600 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1601 in6_dev_put(idev);
1603 return ret;
1606 static int ndisc_ifinfo_sysctl_strategy(ctl_table *ctl, int __user *name,
1607 int nlen, void __user *oldval,
1608 size_t __user *oldlenp,
1609 void __user *newval, size_t newlen,
1610 void **context)
1612 struct net_device *dev = ctl->extra1;
1613 struct inet6_dev *idev;
1614 int ret;
1616 if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1617 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1618 ndisc_warn_deprecated_sysctl(ctl, "procfs", dev ? dev->name : "default");
1620 switch (ctl->ctl_name) {
1621 case NET_NEIGH_REACHABLE_TIME:
1622 ret = sysctl_jiffies(ctl, name, nlen,
1623 oldval, oldlenp, newval, newlen,
1624 context);
1625 break;
1626 case NET_NEIGH_RETRANS_TIME_MS:
1627 case NET_NEIGH_REACHABLE_TIME_MS:
1628 ret = sysctl_ms_jiffies(ctl, name, nlen,
1629 oldval, oldlenp, newval, newlen,
1630 context);
1631 break;
1632 default:
1633 ret = 0;
1636 if (newval && newlen && ret > 0 &&
1637 dev && (idev = in6_dev_get(dev)) != NULL) {
1638 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1639 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1640 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1641 idev->tstamp = jiffies;
1642 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1643 in6_dev_put(idev);
1646 return ret;
1649 #endif
1651 int __init ndisc_init(struct net_proto_family *ops)
1653 struct ipv6_pinfo *np;
1654 struct sock *sk;
1655 int err;
1657 err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, &ndisc_socket);
1658 if (err < 0) {
1659 ND_PRINTK0(KERN_ERR
1660 "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1661 err);
1662 ndisc_socket = NULL; /* For safety. */
1663 return err;
1666 sk = ndisc_socket->sk;
1667 np = inet6_sk(sk);
1668 sk->sk_allocation = GFP_ATOMIC;
1669 np->hop_limit = 255;
1670 /* Do not loopback ndisc messages */
1671 np->mc_loop = 0;
1672 sk->sk_prot->unhash(sk);
1675 * Initialize the neighbour table
1678 neigh_table_init(&nd_tbl);
1680 #ifdef CONFIG_SYSCTL
1681 neigh_sysctl_register(NULL, &nd_tbl.parms, NET_IPV6, NET_IPV6_NEIGH,
1682 "ipv6",
1683 &ndisc_ifinfo_sysctl_change,
1684 &ndisc_ifinfo_sysctl_strategy);
1685 #endif
1687 register_netdevice_notifier(&ndisc_netdev_notifier);
1688 return 0;
1691 void ndisc_cleanup(void)
1693 #ifdef CONFIG_SYSCTL
1694 neigh_sysctl_unregister(&nd_tbl.parms);
1695 #endif
1696 neigh_table_clear(&nd_tbl);
1697 sock_release(ndisc_socket);
1698 ndisc_socket = NULL; /* For safety. */