[TCP]: TCP Illinois congestion control (rev3)
[linux-block.git] / net / ipv6 / ndisc.c
CommitLineData
1da177e4
LT
1/*
2 * Neighbour Discovery for IPv6
1ab1457c 3 * Linux INET6 implementation
1da177e4
LT
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT
7 * Mike Shaver <shaver@ingenia.com>
8 *
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.
13 */
14
15/*
16 * Changes:
17 *
18 * Lars Fenneberg : fixed MTU setting on receipt
19 * of an RA.
20 *
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
26 */
27
28/* Set to 3 to get tracing... */
29#define ND_DEBUG 1
30
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
49
50#include <linux/module.h>
1da177e4
LT
51#include <linux/errno.h>
52#include <linux/types.h>
53#include <linux/socket.h>
54#include <linux/sockios.h>
55#include <linux/sched.h>
56#include <linux/net.h>
57#include <linux/in6.h>
58#include <linux/route.h>
59#include <linux/init.h>
60#include <linux/rcupdate.h>
61#ifdef CONFIG_SYSCTL
62#include <linux/sysctl.h>
63#endif
64
1823730f 65#include <linux/if_addr.h>
1da177e4
LT
66#include <linux/if_arp.h>
67#include <linux/ipv6.h>
68#include <linux/icmpv6.h>
69#include <linux/jhash.h>
70
71#include <net/sock.h>
72#include <net/snmp.h>
73
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>
80
81#include <net/flow.h>
82#include <net/ip6_checksum.h>
83#include <linux/proc_fs.h>
84
85#include <linux/netfilter.h>
86#include <linux/netfilter_ipv6.h>
87
88static struct socket *ndisc_socket;
89
90static u32 ndisc_hash(const void *pkey, const struct net_device *dev);
91static int ndisc_constructor(struct neighbour *neigh);
92static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
93static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
94static int pndisc_constructor(struct pneigh_entry *n);
95static void pndisc_destructor(struct pneigh_entry *n);
96static void pndisc_redo(struct sk_buff *skb);
97
98static 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,
106};
107
108static 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,
116};
117
118
119static 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,
125};
126
127struct 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,
150 },
151 .gc_interval = 30 * HZ,
152 .gc_thresh1 = 128,
153 .gc_thresh2 = 512,
154 .gc_thresh3 = 1024,
155};
156
157/* ND options */
158struct ndisc_options {
70ceb4f5
YH
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
1da177e4
LT
164};
165
166#define nd_opts_src_lladdr nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
167#define nd_opts_tgt_lladdr nd_opt_array[ND_OPT_TARGET_LL_ADDR]
168#define nd_opts_pi nd_opt_array[ND_OPT_PREFIX_INFO]
169#define nd_opts_pi_end nd_opt_array[__ND_OPT_PREFIX_INFO_END]
170#define nd_opts_rh nd_opt_array[ND_OPT_REDIRECT_HDR]
171#define nd_opts_mtu nd_opt_array[ND_OPT_MTU]
172
173#define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
174
175/*
176 * Return the padding between the option length and the start of the
177 * link addr. Currently only IP-over-InfiniBand needs this, although
178 * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
179 * also need a pad of 2.
180 */
181static int ndisc_addr_option_pad(unsigned short type)
182{
183 switch (type) {
184 case ARPHRD_INFINIBAND: return 2;
185 default: return 0;
186 }
187}
188
189static inline int ndisc_opt_addr_space(struct net_device *dev)
190{
191 return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
192}
193
194static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
195 unsigned short addr_type)
196{
197 int space = NDISC_OPT_SPACE(data_len);
198 int pad = ndisc_addr_option_pad(addr_type);
199
200 opt[0] = type;
201 opt[1] = space>>3;
202
203 memset(opt + 2, 0, pad);
204 opt += pad;
205 space -= pad;
206
207 memcpy(opt+2, data, data_len);
208 data_len += 2;
209 opt += data_len;
210 if ((space -= data_len) > 0)
211 memset(opt, 0, space);
212 return opt + space;
213}
214
215static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
216 struct nd_opt_hdr *end)
217{
218 int type;
219 if (!cur || !end || cur >= end)
220 return NULL;
221 type = cur->nd_opt_type;
222 do {
223 cur = ((void *)cur) + (cur->nd_opt_len << 3);
224 } while(cur < end && cur->nd_opt_type != type);
225 return (cur <= end && cur->nd_opt_type == type ? cur : NULL);
226}
227
228static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
229 struct ndisc_options *ndopts)
230{
231 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
232
233 if (!nd_opt || opt_len < 0 || !ndopts)
234 return NULL;
235 memset(ndopts, 0, sizeof(*ndopts));
236 while (opt_len) {
237 int l;
238 if (opt_len < sizeof(struct nd_opt_hdr))
239 return NULL;
240 l = nd_opt->nd_opt_len << 3;
241 if (opt_len < l || l == 0)
242 return NULL;
243 switch (nd_opt->nd_opt_type) {
244 case ND_OPT_SOURCE_LL_ADDR:
245 case ND_OPT_TARGET_LL_ADDR:
246 case ND_OPT_MTU:
247 case ND_OPT_REDIRECT_HDR:
248 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
249 ND_PRINTK2(KERN_WARNING
250 "%s(): duplicated ND6 option found: type=%d\n",
251 __FUNCTION__,
252 nd_opt->nd_opt_type);
253 } else {
254 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
255 }
256 break;
257 case ND_OPT_PREFIX_INFO:
258 ndopts->nd_opts_pi_end = nd_opt;
259 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0)
260 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
261 break;
70ceb4f5
YH
262#ifdef CONFIG_IPV6_ROUTE_INFO
263 case ND_OPT_ROUTE_INFO:
264 ndopts->nd_opts_ri_end = nd_opt;
265 if (!ndopts->nd_opts_ri)
266 ndopts->nd_opts_ri = nd_opt;
267 break;
268#endif
1da177e4
LT
269 default:
270 /*
271 * Unknown options must be silently ignored,
272 * to accommodate future extension to the protocol.
273 */
274 ND_PRINTK2(KERN_NOTICE
275 "%s(): ignored unsupported option; type=%d, len=%d\n",
276 __FUNCTION__,
277 nd_opt->nd_opt_type, nd_opt->nd_opt_len);
278 }
279 opt_len -= l;
280 nd_opt = ((void *)nd_opt) + l;
281 }
282 return ndopts;
283}
284
285static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
286 struct net_device *dev)
287{
288 u8 *lladdr = (u8 *)(p + 1);
289 int lladdrlen = p->nd_opt_len << 3;
290 int prepad = ndisc_addr_option_pad(dev->type);
291 if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
292 return NULL;
293 return (lladdr + prepad);
294}
295
296int ndisc_mc_map(struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
297{
298 switch (dev->type) {
299 case ARPHRD_ETHER:
300 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
301 case ARPHRD_FDDI:
302 ipv6_eth_mc_map(addr, buf);
303 return 0;
304 case ARPHRD_IEEE802_TR:
305 ipv6_tr_mc_map(addr,buf);
306 return 0;
307 case ARPHRD_ARCNET:
308 ipv6_arcnet_mc_map(addr, buf);
309 return 0;
310 case ARPHRD_INFINIBAND:
311 ipv6_ib_mc_map(addr, buf);
312 return 0;
313 default:
314 if (dir) {
315 memcpy(buf, dev->broadcast, dev->addr_len);
316 return 0;
317 }
318 }
319 return -EINVAL;
320}
321
7159039a
YH
322EXPORT_SYMBOL(ndisc_mc_map);
323
1da177e4
LT
324static u32 ndisc_hash(const void *pkey, const struct net_device *dev)
325{
326 const u32 *p32 = pkey;
327 u32 addr_hash, i;
328
329 addr_hash = 0;
330 for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
331 addr_hash ^= *p32++;
332
333 return jhash_2words(addr_hash, dev->ifindex, nd_tbl.hash_rnd);
334}
335
336static int ndisc_constructor(struct neighbour *neigh)
337{
338 struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
339 struct net_device *dev = neigh->dev;
340 struct inet6_dev *in6_dev;
341 struct neigh_parms *parms;
342 int is_multicast = ipv6_addr_is_multicast(addr);
343
344 rcu_read_lock();
345 in6_dev = in6_dev_get(dev);
346 if (in6_dev == NULL) {
347 rcu_read_unlock();
348 return -EINVAL;
349 }
350
351 parms = in6_dev->nd_parms;
352 __neigh_parms_put(neigh->parms);
353 neigh->parms = neigh_parms_clone(parms);
354 rcu_read_unlock();
355
356 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
357 if (dev->hard_header == NULL) {
358 neigh->nud_state = NUD_NOARP;
359 neigh->ops = &ndisc_direct_ops;
360 neigh->output = neigh->ops->queue_xmit;
361 } else {
362 if (is_multicast) {
363 neigh->nud_state = NUD_NOARP;
364 ndisc_mc_map(addr, neigh->ha, dev, 1);
365 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
366 neigh->nud_state = NUD_NOARP;
367 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
368 if (dev->flags&IFF_LOOPBACK)
369 neigh->type = RTN_LOCAL;
370 } else if (dev->flags&IFF_POINTOPOINT) {
371 neigh->nud_state = NUD_NOARP;
372 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
373 }
374 if (dev->hard_header_cache)
375 neigh->ops = &ndisc_hh_ops;
376 else
377 neigh->ops = &ndisc_generic_ops;
378 if (neigh->nud_state&NUD_VALID)
379 neigh->output = neigh->ops->connected_output;
380 else
381 neigh->output = neigh->ops->output;
382 }
383 in6_dev_put(in6_dev);
384 return 0;
385}
386
387static int pndisc_constructor(struct pneigh_entry *n)
388{
389 struct in6_addr *addr = (struct in6_addr*)&n->key;
390 struct in6_addr maddr;
391 struct net_device *dev = n->dev;
392
393 if (dev == NULL || __in6_dev_get(dev) == NULL)
394 return -EINVAL;
395 addrconf_addr_solict_mult(addr, &maddr);
396 ipv6_dev_mc_inc(dev, &maddr);
397 return 0;
398}
399
400static void pndisc_destructor(struct pneigh_entry *n)
401{
402 struct in6_addr *addr = (struct in6_addr*)&n->key;
403 struct in6_addr maddr;
404 struct net_device *dev = n->dev;
405
406 if (dev == NULL || __in6_dev_get(dev) == NULL)
407 return;
408 addrconf_addr_solict_mult(addr, &maddr);
409 ipv6_dev_mc_dec(dev, &maddr);
410}
411
412/*
413 * Send a Neighbour Advertisement
414 */
415
416static inline void ndisc_flow_init(struct flowi *fl, u8 type,
af184765
YH
417 struct in6_addr *saddr, struct in6_addr *daddr,
418 int oif)
1da177e4
LT
419{
420 memset(fl, 0, sizeof(*fl));
421 ipv6_addr_copy(&fl->fl6_src, saddr);
422 ipv6_addr_copy(&fl->fl6_dst, daddr);
423 fl->proto = IPPROTO_ICMPV6;
424 fl->fl_icmp_type = type;
425 fl->fl_icmp_code = 0;
af184765 426 fl->oif = oif;
beb8d13b 427 security_sk_classify_flow(ndisc_socket->sk, fl);
1da177e4
LT
428}
429
430static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
431 struct in6_addr *daddr, struct in6_addr *solicited_addr,
1ab1457c 432 int router, int solicited, int override, int inc_opt)
1da177e4
LT
433{
434 struct in6_addr tmpaddr;
435 struct inet6_ifaddr *ifp;
436 struct inet6_dev *idev;
437 struct flowi fl;
438 struct dst_entry* dst;
1ab1457c 439 struct sock *sk = ndisc_socket->sk;
1da177e4 440 struct in6_addr *src_addr;
1ab1457c
YH
441 struct nd_msg *msg;
442 int len;
443 struct sk_buff *skb;
1da177e4
LT
444 int err;
445
446 len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
447
448 /* for anycast or proxy, solicited_addr != src_addr */
449 ifp = ipv6_get_ifaddr(solicited_addr, dev, 1);
1ab1457c 450 if (ifp) {
1da177e4 451 src_addr = solicited_addr;
95c385b4
NH
452 if (ifp->flags & IFA_F_OPTIMISTIC)
453 override = 0;
1da177e4
LT
454 in6_ifa_put(ifp);
455 } else {
456 if (ipv6_dev_get_saddr(dev, daddr, &tmpaddr))
457 return;
458 src_addr = &tmpaddr;
459 }
460
af184765
YH
461 ndisc_flow_init(&fl, NDISC_NEIGHBOUR_ADVERTISEMENT, src_addr, daddr,
462 dev->ifindex);
1da177e4
LT
463
464 dst = ndisc_dst_alloc(dev, neigh, daddr, ip6_output);
465 if (!dst)
466 return;
467
468 err = xfrm_lookup(&dst, &fl, NULL, 0);
e104411b 469 if (err < 0)
1da177e4 470 return;
1da177e4
LT
471
472 if (inc_opt) {
473 if (dev->addr_len)
474 len += ndisc_opt_addr_space(dev);
475 else
476 inc_opt = 0;
477 }
478
d54a81d3
DM
479 skb = sock_alloc_send_skb(sk,
480 (MAX_HEADER + sizeof(struct ipv6hdr) +
481 len + LL_RESERVED_SPACE(dev)),
1da177e4
LT
482 1, &err);
483
484 if (skb == NULL) {
485 ND_PRINTK0(KERN_ERR
1ab1457c 486 "ICMPv6 NA: %s() failed to allocate an skb.\n",
1da177e4
LT
487 __FUNCTION__);
488 dst_release(dst);
489 return;
490 }
491
492 skb_reserve(skb, LL_RESERVED_SPACE(dev));
493 ip6_nd_hdr(sk, skb, dev, src_addr, daddr, IPPROTO_ICMPV6, len);
494
27a884dc 495 skb->transport_header = skb->tail;
d10ba34b
ACM
496 skb_put(skb, len);
497 msg = (struct nd_msg *)skb_transport_header(skb);
1da177e4 498
1ab1457c
YH
499 msg->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
500 msg->icmph.icmp6_code = 0;
501 msg->icmph.icmp6_cksum = 0;
1da177e4 502
1ab1457c
YH
503 msg->icmph.icmp6_unused = 0;
504 msg->icmph.icmp6_router = router;
505 msg->icmph.icmp6_solicited = solicited;
506 msg->icmph.icmp6_override = override;
1da177e4 507
1ab1457c 508 /* Set the target address. */
1da177e4
LT
509 ipv6_addr_copy(&msg->target, solicited_addr);
510
511 if (inc_opt)
512 ndisc_fill_addr_option(msg->opt, ND_OPT_TARGET_LL_ADDR, dev->dev_addr,
513 dev->addr_len, dev->type);
514
515 /* checksum */
1ab1457c 516 msg->icmph.icmp6_cksum = csum_ipv6_magic(src_addr, daddr, len,
1da177e4 517 IPPROTO_ICMPV6,
1ab1457c 518 csum_partial((__u8 *) msg,
1da177e4
LT
519 len, 0));
520
521 skb->dst = dst;
522 idev = in6_dev_get(dst->dev);
a11d206d 523 IP6_INC_STATS(idev, IPSTATS_MIB_OUTREQUESTS);
1da177e4
LT
524 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
525 if (!err) {
526 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTNEIGHBORADVERTISEMENTS);
527 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
528 }
529
530 if (likely(idev != NULL))
531 in6_dev_put(idev);
1ab1457c 532}
1da177e4
LT
533
534void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
535 struct in6_addr *solicit,
1ab1457c 536 struct in6_addr *daddr, struct in6_addr *saddr)
1da177e4
LT
537{
538 struct flowi fl;
539 struct dst_entry* dst;
540 struct inet6_dev *idev;
1ab1457c
YH
541 struct sock *sk = ndisc_socket->sk;
542 struct sk_buff *skb;
543 struct nd_msg *msg;
1da177e4 544 struct in6_addr addr_buf;
1ab1457c 545 int len;
1da177e4
LT
546 int err;
547 int send_llinfo;
548
549 if (saddr == NULL) {
95c385b4
NH
550 if (ipv6_get_lladdr(dev, &addr_buf,
551 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
1da177e4
LT
552 return;
553 saddr = &addr_buf;
554 }
555
af184765
YH
556 ndisc_flow_init(&fl, NDISC_NEIGHBOUR_SOLICITATION, saddr, daddr,
557 dev->ifindex);
1da177e4
LT
558
559 dst = ndisc_dst_alloc(dev, neigh, daddr, ip6_output);
560 if (!dst)
561 return;
562
563 err = xfrm_lookup(&dst, &fl, NULL, 0);
e104411b 564 if (err < 0)
1da177e4 565 return;
1da177e4
LT
566
567 len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
568 send_llinfo = dev->addr_len && !ipv6_addr_any(saddr);
569 if (send_llinfo)
570 len += ndisc_opt_addr_space(dev);
571
d54a81d3
DM
572 skb = sock_alloc_send_skb(sk,
573 (MAX_HEADER + sizeof(struct ipv6hdr) +
574 len + LL_RESERVED_SPACE(dev)),
1da177e4
LT
575 1, &err);
576 if (skb == NULL) {
577 ND_PRINTK0(KERN_ERR
1ab1457c 578 "ICMPv6 NA: %s() failed to allocate an skb.\n",
1da177e4
LT
579 __FUNCTION__);
580 dst_release(dst);
581 return;
582 }
583
584 skb_reserve(skb, LL_RESERVED_SPACE(dev));
585 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
586
27a884dc 587 skb->transport_header = skb->tail;
d10ba34b
ACM
588 skb_put(skb, len);
589 msg = (struct nd_msg *)skb_transport_header(skb);
1da177e4
LT
590 msg->icmph.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION;
591 msg->icmph.icmp6_code = 0;
592 msg->icmph.icmp6_cksum = 0;
593 msg->icmph.icmp6_unused = 0;
594
595 /* Set the target address. */
596 ipv6_addr_copy(&msg->target, solicit);
597
598 if (send_llinfo)
599 ndisc_fill_addr_option(msg->opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr,
600 dev->addr_len, dev->type);
601
602 /* checksum */
0660e03f 603 msg->icmph.icmp6_cksum = csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
1ab1457c 604 daddr, len,
1da177e4 605 IPPROTO_ICMPV6,
1ab1457c 606 csum_partial((__u8 *) msg,
1da177e4
LT
607 len, 0));
608 /* send it! */
609 skb->dst = dst;
610 idev = in6_dev_get(dst->dev);
a11d206d 611 IP6_INC_STATS(idev, IPSTATS_MIB_OUTREQUESTS);
1da177e4
LT
612 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
613 if (!err) {
614 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTNEIGHBORSOLICITS);
615 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
616 }
617
618 if (likely(idev != NULL))
619 in6_dev_put(idev);
620}
621
622void ndisc_send_rs(struct net_device *dev, struct in6_addr *saddr,
623 struct in6_addr *daddr)
624{
625 struct flowi fl;
626 struct dst_entry* dst;
627 struct inet6_dev *idev;
628 struct sock *sk = ndisc_socket->sk;
1ab1457c
YH
629 struct sk_buff *skb;
630 struct icmp6hdr *hdr;
1da177e4 631 __u8 * opt;
95c385b4 632 int send_sllao = dev->addr_len;
1ab1457c 633 int len;
1da177e4
LT
634 int err;
635
95c385b4
NH
636
637#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
638 /*
639 * According to section 2.2 of RFC 4429, we must not
640 * send router solicitations with a sllao from
641 * optimistic addresses, but we may send the solicitation
642 * if we don't include the sllao. So here we check
643 * if our address is optimistic, and if so, we
644 * supress the inclusion of the sllao.
645 */
646 if (send_sllao) {
ca043569 647 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(saddr, dev, 1);
95c385b4
NH
648 if (ifp) {
649 if (ifp->flags & IFA_F_OPTIMISTIC) {
ca043569 650 send_sllao = 0;
95c385b4 651 }
ca043569 652 in6_ifa_put(ifp);
95c385b4
NH
653 } else {
654 send_sllao = 0;
655 }
656 }
657#endif
af184765
YH
658 ndisc_flow_init(&fl, NDISC_ROUTER_SOLICITATION, saddr, daddr,
659 dev->ifindex);
1da177e4
LT
660
661 dst = ndisc_dst_alloc(dev, NULL, daddr, ip6_output);
662 if (!dst)
663 return;
664
665 err = xfrm_lookup(&dst, &fl, NULL, 0);
e104411b 666 if (err < 0)
1da177e4 667 return;
1da177e4
LT
668
669 len = sizeof(struct icmp6hdr);
95c385b4 670 if (send_sllao)
1da177e4
LT
671 len += ndisc_opt_addr_space(dev);
672
1ab1457c 673 skb = sock_alloc_send_skb(sk,
d54a81d3
DM
674 (MAX_HEADER + sizeof(struct ipv6hdr) +
675 len + LL_RESERVED_SPACE(dev)),
1da177e4
LT
676 1, &err);
677 if (skb == NULL) {
678 ND_PRINTK0(KERN_ERR
1ab1457c 679 "ICMPv6 RS: %s() failed to allocate an skb.\n",
1da177e4
LT
680 __FUNCTION__);
681 dst_release(dst);
682 return;
683 }
684
685 skb_reserve(skb, LL_RESERVED_SPACE(dev));
686 ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
687
27a884dc 688 skb->transport_header = skb->tail;
d10ba34b
ACM
689 skb_put(skb, len);
690 hdr = icmp6_hdr(skb);
1ab1457c
YH
691 hdr->icmp6_type = NDISC_ROUTER_SOLICITATION;
692 hdr->icmp6_code = 0;
693 hdr->icmp6_cksum = 0;
694 hdr->icmp6_unused = 0;
1da177e4
LT
695
696 opt = (u8*) (hdr + 1);
697
95c385b4 698 if (send_sllao)
1da177e4
LT
699 ndisc_fill_addr_option(opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr,
700 dev->addr_len, dev->type);
701
702 /* checksum */
0660e03f 703 hdr->icmp6_cksum = csum_ipv6_magic(&ipv6_hdr(skb)->saddr, daddr, len,
1da177e4
LT
704 IPPROTO_ICMPV6,
705 csum_partial((__u8 *) hdr, len, 0));
706
707 /* send it! */
708 skb->dst = dst;
709 idev = in6_dev_get(dst->dev);
a11d206d 710 IP6_INC_STATS(idev, IPSTATS_MIB_OUTREQUESTS);
1da177e4
LT
711 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
712 if (!err) {
713 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTROUTERSOLICITS);
714 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
715 }
716
717 if (likely(idev != NULL))
718 in6_dev_put(idev);
719}
1ab1457c 720
1da177e4
LT
721
722static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
723{
724 /*
725 * "The sender MUST return an ICMP
726 * destination unreachable"
727 */
728 dst_link_failure(skb);
729 kfree_skb(skb);
730}
731
732/* Called with locked neigh: either read or both */
733
734static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
735{
736 struct in6_addr *saddr = NULL;
737 struct in6_addr mcaddr;
738 struct net_device *dev = neigh->dev;
739 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
740 int probes = atomic_read(&neigh->probes);
741
0660e03f
ACM
742 if (skb && ipv6_chk_addr(&ipv6_hdr(skb)->saddr, dev, 1))
743 saddr = &ipv6_hdr(skb)->saddr;
1da177e4
LT
744
745 if ((probes -= neigh->parms->ucast_probes) < 0) {
746 if (!(neigh->nud_state & NUD_VALID)) {
747 ND_PRINTK1(KERN_DEBUG
748 "%s(): trying to ucast probe in NUD_INVALID: "
46b86a2d 749 NIP6_FMT "\n",
1da177e4
LT
750 __FUNCTION__,
751 NIP6(*target));
752 }
753 ndisc_send_ns(dev, neigh, target, target, saddr);
754 } else if ((probes -= neigh->parms->app_probes) < 0) {
755#ifdef CONFIG_ARPD
756 neigh_app_ns(neigh);
757#endif
758 } else {
759 addrconf_addr_solict_mult(target, &mcaddr);
760 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
761 }
762}
763
764static void ndisc_recv_ns(struct sk_buff *skb)
765{
9c70220b 766 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
0660e03f
ACM
767 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
768 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
1da177e4 769 u8 *lladdr = NULL;
27a884dc
ACM
770 u32 ndoptlen = skb->tail - (skb->transport_header +
771 offsetof(struct nd_msg, opt));
1da177e4
LT
772 struct ndisc_options ndopts;
773 struct net_device *dev = skb->dev;
774 struct inet6_ifaddr *ifp;
775 struct inet6_dev *idev = NULL;
776 struct neighbour *neigh;
62dd9318 777 struct pneigh_entry *pneigh = NULL;
1da177e4
LT
778 int dad = ipv6_addr_any(saddr);
779 int inc;
62dd9318 780 int is_router;
1da177e4
LT
781
782 if (ipv6_addr_is_multicast(&msg->target)) {
1ab1457c 783 ND_PRINTK2(KERN_WARNING
1da177e4
LT
784 "ICMPv6 NS: multicast target address");
785 return;
786 }
787
788 /*
789 * RFC2461 7.1.1:
790 * DAD has to be destined for solicited node multicast address.
791 */
792 if (dad &&
793 !(daddr->s6_addr32[0] == htonl(0xff020000) &&
794 daddr->s6_addr32[1] == htonl(0x00000000) &&
795 daddr->s6_addr32[2] == htonl(0x00000001) &&
796 daddr->s6_addr [12] == 0xff )) {
797 ND_PRINTK2(KERN_WARNING
798 "ICMPv6 NS: bad DAD packet (wrong destination)\n");
799 return;
800 }
801
802 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
1ab1457c 803 ND_PRINTK2(KERN_WARNING
1da177e4
LT
804 "ICMPv6 NS: invalid ND options\n");
805 return;
806 }
807
808 if (ndopts.nd_opts_src_lladdr) {
809 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
810 if (!lladdr) {
811 ND_PRINTK2(KERN_WARNING
812 "ICMPv6 NS: invalid link-layer address length\n");
813 return;
814 }
815
816 /* RFC2461 7.1.1:
1ab1457c
YH
817 * If the IP source address is the unspecified address,
818 * there MUST NOT be source link-layer address option
1da177e4
LT
819 * in the message.
820 */
821 if (dad) {
1ab1457c 822 ND_PRINTK2(KERN_WARNING
1da177e4
LT
823 "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
824 return;
825 }
826 }
827
828 inc = ipv6_addr_is_multicast(daddr);
829
830 if ((ifp = ipv6_get_ifaddr(&msg->target, dev, 1)) != NULL) {
95c385b4
NH
831
832 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
833 if (dad) {
834 if (dev->type == ARPHRD_IEEE802_TR) {
98e399f8
ACM
835 const unsigned char *sadr;
836 sadr = skb_mac_header(skb);
95c385b4
NH
837 if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
838 sadr[9] == dev->dev_addr[1] &&
839 sadr[10] == dev->dev_addr[2] &&
840 sadr[11] == dev->dev_addr[3] &&
841 sadr[12] == dev->dev_addr[4] &&
842 sadr[13] == dev->dev_addr[5]) {
843 /* looped-back to us */
844 goto out;
845 }
846 }
847
848 /*
849 * We are colliding with another node
850 * who is doing DAD
851 * so fail our DAD process
852 */
853 addrconf_dad_failure(ifp);
1da177e4 854 goto out;
95c385b4
NH
855 } else {
856 /*
857 * This is not a dad solicitation.
858 * If we are an optimistic node,
859 * we should respond.
860 * Otherwise, we should ignore it.
861 */
862 if (!(ifp->flags & IFA_F_OPTIMISTIC))
1da177e4 863 goto out;
1da177e4 864 }
1da177e4
LT
865 }
866
867 idev = ifp->idev;
868 } else {
869 idev = in6_dev_get(dev);
870 if (!idev) {
871 /* XXX: count this drop? */
872 return;
873 }
874
875 if (ipv6_chk_acast_addr(dev, &msg->target) ||
1ab1457c 876 (idev->cnf.forwarding &&
fbea49e1 877 (ipv6_devconf.proxy_ndp || idev->cnf.proxy_ndp) &&
62dd9318
VN
878 (pneigh = pneigh_lookup(&nd_tbl,
879 &msg->target, dev, 0)) != NULL)) {
a61bbcf2 880 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
1da177e4
LT
881 skb->pkt_type != PACKET_HOST &&
882 inc != 0 &&
883 idev->nd_parms->proxy_delay != 0) {
884 /*
885 * for anycast or proxy,
1ab1457c
YH
886 * sender should delay its response
887 * by a random time between 0 and
1da177e4
LT
888 * MAX_ANYCAST_DELAY_TIME seconds.
889 * (RFC2461) -- yoshfuji
890 */
891 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
892 if (n)
893 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
894 goto out;
895 }
896 } else
897 goto out;
898 }
899
fc26d0ab 900 is_router = !!(pneigh ? pneigh->flags & NTF_ROUTER : idev->cnf.forwarding);
62dd9318 901
1da177e4
LT
902 if (dad) {
903 struct in6_addr maddr;
904
905 ipv6_addr_all_nodes(&maddr);
906 ndisc_send_na(dev, NULL, &maddr, &msg->target,
62dd9318 907 is_router, 0, (ifp != NULL), 1);
1da177e4
LT
908 goto out;
909 }
910
911 if (inc)
912 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
913 else
914 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
915
1ab1457c 916 /*
1da177e4
LT
917 * update / create cache entry
918 * for the source address
919 */
920 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
921 !inc || lladdr || !dev->addr_len);
922 if (neigh)
1ab1457c 923 neigh_update(neigh, lladdr, NUD_STALE,
1da177e4
LT
924 NEIGH_UPDATE_F_WEAK_OVERRIDE|
925 NEIGH_UPDATE_F_OVERRIDE);
926 if (neigh || !dev->hard_header) {
927 ndisc_send_na(dev, neigh, saddr, &msg->target,
62dd9318 928 is_router,
1da177e4
LT
929 1, (ifp != NULL && inc), inc);
930 if (neigh)
931 neigh_release(neigh);
932 }
933
934out:
935 if (ifp)
936 in6_ifa_put(ifp);
937 else
938 in6_dev_put(idev);
939
940 return;
941}
942
943static void ndisc_recv_na(struct sk_buff *skb)
944{
9c70220b 945 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
0660e03f
ACM
946 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
947 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
1da177e4 948 u8 *lladdr = NULL;
27a884dc
ACM
949 u32 ndoptlen = skb->tail - (skb->transport_header +
950 offsetof(struct nd_msg, opt));
1da177e4
LT
951 struct ndisc_options ndopts;
952 struct net_device *dev = skb->dev;
953 struct inet6_ifaddr *ifp;
954 struct neighbour *neigh;
955
956 if (skb->len < sizeof(struct nd_msg)) {
957 ND_PRINTK2(KERN_WARNING
958 "ICMPv6 NA: packet too short\n");
959 return;
960 }
961
962 if (ipv6_addr_is_multicast(&msg->target)) {
963 ND_PRINTK2(KERN_WARNING
964 "ICMPv6 NA: target address is multicast.\n");
965 return;
966 }
967
968 if (ipv6_addr_is_multicast(daddr) &&
969 msg->icmph.icmp6_solicited) {
970 ND_PRINTK2(KERN_WARNING
971 "ICMPv6 NA: solicited NA is multicasted.\n");
972 return;
973 }
1ab1457c 974
1da177e4
LT
975 if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
976 ND_PRINTK2(KERN_WARNING
977 "ICMPv6 NS: invalid ND option\n");
978 return;
979 }
980 if (ndopts.nd_opts_tgt_lladdr) {
981 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
982 if (!lladdr) {
983 ND_PRINTK2(KERN_WARNING
984 "ICMPv6 NA: invalid link-layer address length\n");
985 return;
986 }
987 }
988 if ((ifp = ipv6_get_ifaddr(&msg->target, dev, 1))) {
989 if (ifp->flags & IFA_F_TENTATIVE) {
990 addrconf_dad_failure(ifp);
991 return;
992 }
993 /* What should we make now? The advertisement
994 is invalid, but ndisc specs say nothing
995 about it. It could be misconfiguration, or
996 an smart proxy agent tries to help us :-)
997 */
998 ND_PRINTK1(KERN_WARNING
999 "ICMPv6 NA: someone advertises our address on %s!\n",
1000 ifp->idev->dev->name);
1001 in6_ifa_put(ifp);
1002 return;
1003 }
1004 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1005
1006 if (neigh) {
1007 u8 old_flags = neigh->flags;
1008
1009 if (neigh->nud_state & NUD_FAILED)
1010 goto out;
1011
5f3e6e9e
VN
1012 /*
1013 * Don't update the neighbor cache entry on a proxy NA from
1014 * ourselves because either the proxied node is off link or it
1015 * has already sent a NA to us.
1016 */
1017 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
fbea49e1
YH
1018 ipv6_devconf.forwarding && ipv6_devconf.proxy_ndp &&
1019 pneigh_lookup(&nd_tbl, &msg->target, dev, 0)) {
1020 /* XXX: idev->cnf.prixy_ndp */
5f3e6e9e 1021 goto out;
fbea49e1 1022 }
5f3e6e9e 1023
1da177e4
LT
1024 neigh_update(neigh, lladdr,
1025 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1026 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1027 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1028 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1029 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
1030
1031 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1032 /*
1033 * Change: router to host
1034 */
1035 struct rt6_info *rt;
1036 rt = rt6_get_dflt_router(saddr, dev);
1037 if (rt)
e0a1ad73 1038 ip6_del_rt(rt);
1da177e4
LT
1039 }
1040
1041out:
1042 neigh_release(neigh);
1043 }
1044}
1045
1046static void ndisc_recv_rs(struct sk_buff *skb)
1047{
9c70220b 1048 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1da177e4
LT
1049 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1050 struct neighbour *neigh;
1051 struct inet6_dev *idev;
0660e03f 1052 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1da177e4
LT
1053 struct ndisc_options ndopts;
1054 u8 *lladdr = NULL;
1055
1056 if (skb->len < sizeof(*rs_msg))
1057 return;
1058
1059 idev = in6_dev_get(skb->dev);
1060 if (!idev) {
1061 if (net_ratelimit())
1062 ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
1063 return;
1064 }
1065
1066 /* Don't accept RS if we're not in router mode */
1067 if (!idev->cnf.forwarding)
1068 goto out;
1069
1070 /*
1071 * Don't update NCE if src = ::;
1072 * this implies that the source node has no ip address assigned yet.
1073 */
1074 if (ipv6_addr_any(saddr))
1075 goto out;
1076
1077 /* Parse ND options */
1078 if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1079 if (net_ratelimit())
1080 ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1081 goto out;
1082 }
1083
1084 if (ndopts.nd_opts_src_lladdr) {
1085 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1086 skb->dev);
1087 if (!lladdr)
1088 goto out;
1089 }
1090
1091 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1092 if (neigh) {
1093 neigh_update(neigh, lladdr, NUD_STALE,
1094 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1095 NEIGH_UPDATE_F_OVERRIDE|
1096 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1097 neigh_release(neigh);
1098 }
1099out:
1100 in6_dev_put(idev);
1101}
1102
1103static void ndisc_router_discovery(struct sk_buff *skb)
1104{
9c70220b 1105 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1da177e4
LT
1106 struct neighbour *neigh = NULL;
1107 struct inet6_dev *in6_dev;
65f5c7c1 1108 struct rt6_info *rt = NULL;
1da177e4
LT
1109 int lifetime;
1110 struct ndisc_options ndopts;
1111 int optlen;
ebacaaa0 1112 unsigned int pref = 0;
1da177e4
LT
1113
1114 __u8 * opt = (__u8 *)(ra_msg + 1);
1115
27a884dc 1116 optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1da177e4 1117
0660e03f 1118 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1da177e4
LT
1119 ND_PRINTK2(KERN_WARNING
1120 "ICMPv6 RA: source address is not link-local.\n");
1121 return;
1122 }
1123 if (optlen < 0) {
1ab1457c 1124 ND_PRINTK2(KERN_WARNING
1da177e4
LT
1125 "ICMPv6 RA: packet too short\n");
1126 return;
1127 }
1128
1129 /*
1130 * set the RA_RECV flag in the interface
1131 */
1132
1133 in6_dev = in6_dev_get(skb->dev);
1134 if (in6_dev == NULL) {
1135 ND_PRINTK0(KERN_ERR
1136 "ICMPv6 RA: can't find inet6 device for %s.\n",
1137 skb->dev->name);
1138 return;
1139 }
1140 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra) {
1141 in6_dev_put(in6_dev);
1142 return;
1143 }
1144
1145 if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1146 in6_dev_put(in6_dev);
1147 ND_PRINTK2(KERN_WARNING
1148 "ICMP6 RA: invalid ND options\n");
1149 return;
1150 }
1151
1152 if (in6_dev->if_flags & IF_RS_SENT) {
1153 /*
1154 * flag that an RA was received after an RS was sent
1155 * out on this interface.
1156 */
1157 in6_dev->if_flags |= IF_RA_RCVD;
1158 }
1159
1160 /*
1161 * Remember the managed/otherconf flags from most recently
1162 * received RA message (RFC 2462) -- yoshfuji
1163 */
1164 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1165 IF_RA_OTHERCONF)) |
1166 (ra_msg->icmph.icmp6_addrconf_managed ?
1167 IF_RA_MANAGED : 0) |
1168 (ra_msg->icmph.icmp6_addrconf_other ?
1169 IF_RA_OTHERCONF : 0);
1170
65f5c7c1
YH
1171 if (!in6_dev->cnf.accept_ra_defrtr)
1172 goto skip_defrtr;
1173
1da177e4
LT
1174 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1175
ebacaaa0
YH
1176#ifdef CONFIG_IPV6_ROUTER_PREF
1177 pref = ra_msg->icmph.icmp6_router_pref;
1178 /* 10b is handled as if it were 00b (medium) */
930d6ff2
YH
1179 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1180 in6_dev->cnf.accept_ra_rtr_pref)
ebacaaa0
YH
1181 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1182#endif
1183
0660e03f 1184 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1da177e4
LT
1185
1186 if (rt)
1187 neigh = rt->rt6i_nexthop;
1188
1189 if (rt && lifetime == 0) {
1190 neigh_clone(neigh);
e0a1ad73 1191 ip6_del_rt(rt);
1da177e4
LT
1192 rt = NULL;
1193 }
1194
1195 if (rt == NULL && lifetime) {
1196 ND_PRINTK3(KERN_DEBUG
1197 "ICMPv6 RA: adding default router.\n");
1198
0660e03f 1199 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1da177e4
LT
1200 if (rt == NULL) {
1201 ND_PRINTK0(KERN_ERR
1202 "ICMPv6 RA: %s() failed to add default route.\n",
1203 __FUNCTION__);
1204 in6_dev_put(in6_dev);
1205 return;
1206 }
1207
1208 neigh = rt->rt6i_nexthop;
1209 if (neigh == NULL) {
1210 ND_PRINTK0(KERN_ERR
1211 "ICMPv6 RA: %s() got default router without neighbour.\n",
1212 __FUNCTION__);
1213 dst_release(&rt->u.dst);
1214 in6_dev_put(in6_dev);
1215 return;
1216 }
1217 neigh->flags |= NTF_ROUTER;
ebacaaa0
YH
1218 } else if (rt) {
1219 rt->rt6i_flags |= (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1da177e4
LT
1220 }
1221
1222 if (rt)
1223 rt->rt6i_expires = jiffies + (HZ * lifetime);
1224
1225 if (ra_msg->icmph.icmp6_hop_limit) {
1226 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1227 if (rt)
1228 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1229 }
1230
65f5c7c1
YH
1231skip_defrtr:
1232
1da177e4
LT
1233 /*
1234 * Update Reachable Time and Retrans Timer
1235 */
1236
1237 if (in6_dev->nd_parms) {
1238 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1239
1240 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1241 rtime = (rtime*HZ)/1000;
1242 if (rtime < HZ/10)
1243 rtime = HZ/10;
1244 in6_dev->nd_parms->retrans_time = rtime;
1245 in6_dev->tstamp = jiffies;
1246 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1247 }
1248
1249 rtime = ntohl(ra_msg->reachable_time);
1250 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1251 rtime = (rtime*HZ)/1000;
1252
1253 if (rtime < HZ/10)
1254 rtime = HZ/10;
1255
1256 if (rtime != in6_dev->nd_parms->base_reachable_time) {
1257 in6_dev->nd_parms->base_reachable_time = rtime;
1258 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1259 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1260 in6_dev->tstamp = jiffies;
1261 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1262 }
1263 }
1264 }
1265
1266 /*
1267 * Process options.
1268 */
1269
1270 if (!neigh)
0660e03f 1271 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1da177e4
LT
1272 skb->dev, 1);
1273 if (neigh) {
1274 u8 *lladdr = NULL;
1275 if (ndopts.nd_opts_src_lladdr) {
1276 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1277 skb->dev);
1278 if (!lladdr) {
1279 ND_PRINTK2(KERN_WARNING
1280 "ICMPv6 RA: invalid link-layer address length\n");
1281 goto out;
1282 }
1283 }
1284 neigh_update(neigh, lladdr, NUD_STALE,
1285 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1286 NEIGH_UPDATE_F_OVERRIDE|
1287 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1288 NEIGH_UPDATE_F_ISROUTER);
1289 }
1290
70ceb4f5 1291#ifdef CONFIG_IPV6_ROUTE_INFO
09c884d4 1292 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
70ceb4f5
YH
1293 struct nd_opt_hdr *p;
1294 for (p = ndopts.nd_opts_ri;
1295 p;
1296 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
09c884d4
YH
1297 if (((struct route_info *)p)->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1298 continue;
70ceb4f5 1299 rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
0660e03f 1300 &ipv6_hdr(skb)->saddr);
70ceb4f5
YH
1301 }
1302 }
1303#endif
1304
c4fd30eb 1305 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1da177e4
LT
1306 struct nd_opt_hdr *p;
1307 for (p = ndopts.nd_opts_pi;
1308 p;
1309 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1310 addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1311 }
1312 }
1313
1314 if (ndopts.nd_opts_mtu) {
e69a4adc 1315 __be32 n;
1da177e4
LT
1316 u32 mtu;
1317
e69a4adc
AV
1318 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1319 mtu = ntohl(n);
1da177e4
LT
1320
1321 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1322 ND_PRINTK2(KERN_WARNING
1323 "ICMPv6 RA: invalid mtu: %d\n",
1324 mtu);
1325 } else if (in6_dev->cnf.mtu6 != mtu) {
1326 in6_dev->cnf.mtu6 = mtu;
1327
1328 if (rt)
1329 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
1330
1331 rt6_mtu_change(skb->dev, mtu);
1332 }
1333 }
1ab1457c 1334
1da177e4
LT
1335 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1336 ND_PRINTK2(KERN_WARNING
1337 "ICMPv6 RA: invalid RA options");
1338 }
1339out:
1340 if (rt)
1341 dst_release(&rt->u.dst);
1342 else if (neigh)
1343 neigh_release(neigh);
1344 in6_dev_put(in6_dev);
1345}
1346
1347static void ndisc_redirect_rcv(struct sk_buff *skb)
1348{
1349 struct inet6_dev *in6_dev;
1350 struct icmp6hdr *icmph;
1351 struct in6_addr *dest;
1352 struct in6_addr *target; /* new first hop to destination */
1353 struct neighbour *neigh;
1354 int on_link = 0;
1355 struct ndisc_options ndopts;
1356 int optlen;
1357 u8 *lladdr = NULL;
1358
0660e03f 1359 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1da177e4
LT
1360 ND_PRINTK2(KERN_WARNING
1361 "ICMPv6 Redirect: source address is not link-local.\n");
1362 return;
1363 }
1364
27a884dc 1365 optlen = skb->tail - skb->transport_header;
1da177e4
LT
1366 optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1367
1368 if (optlen < 0) {
1369 ND_PRINTK2(KERN_WARNING
1370 "ICMPv6 Redirect: packet too short\n");
1371 return;
1372 }
1373
cc70ab26 1374 icmph = icmp6_hdr(skb);
1da177e4
LT
1375 target = (struct in6_addr *) (icmph + 1);
1376 dest = target + 1;
1377
1378 if (ipv6_addr_is_multicast(dest)) {
1379 ND_PRINTK2(KERN_WARNING
1380 "ICMPv6 Redirect: destination address is multicast.\n");
1381 return;
1382 }
1383
1384 if (ipv6_addr_equal(dest, target)) {
1385 on_link = 1;
1386 } else if (!(ipv6_addr_type(target) & IPV6_ADDR_LINKLOCAL)) {
1ab1457c 1387 ND_PRINTK2(KERN_WARNING
1da177e4
LT
1388 "ICMPv6 Redirect: target address is not link-local.\n");
1389 return;
1390 }
1391
1392 in6_dev = in6_dev_get(skb->dev);
1393 if (!in6_dev)
1394 return;
1395 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1396 in6_dev_put(in6_dev);
1397 return;
1398 }
1399
1ab1457c 1400 /* RFC2461 8.1:
1da177e4
LT
1401 * The IP source address of the Redirect MUST be the same as the current
1402 * first-hop router for the specified ICMP Destination Address.
1403 */
1ab1457c 1404
1da177e4
LT
1405 if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1406 ND_PRINTK2(KERN_WARNING
1407 "ICMPv6 Redirect: invalid ND options\n");
1408 in6_dev_put(in6_dev);
1409 return;
1410 }
1411 if (ndopts.nd_opts_tgt_lladdr) {
1412 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1413 skb->dev);
1414 if (!lladdr) {
1415 ND_PRINTK2(KERN_WARNING
1416 "ICMPv6 Redirect: invalid link-layer address length\n");
1417 in6_dev_put(in6_dev);
1418 return;
1419 }
1420 }
1421
1422 neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1423 if (neigh) {
0660e03f
ACM
1424 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1425 &ipv6_hdr(skb)->saddr, neigh, lladdr,
1da177e4
LT
1426 on_link);
1427 neigh_release(neigh);
1428 }
1429 in6_dev_put(in6_dev);
1430}
1431
1432void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1433 struct in6_addr *target)
1434{
1435 struct sock *sk = ndisc_socket->sk;
1436 int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1437 struct sk_buff *buff;
1438 struct icmp6hdr *icmph;
1439 struct in6_addr saddr_buf;
1440 struct in6_addr *addrp;
1441 struct net_device *dev;
1442 struct rt6_info *rt;
1443 struct dst_entry *dst;
1444 struct inet6_dev *idev;
1445 struct flowi fl;
1446 u8 *opt;
1447 int rd_len;
1448 int err;
1449 int hlen;
1450 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1451
1452 dev = skb->dev;
1453
95c385b4 1454 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1da177e4
LT
1455 ND_PRINTK2(KERN_WARNING
1456 "ICMPv6 Redirect: no link-local address on %s\n",
1457 dev->name);
1ab1457c
YH
1458 return;
1459 }
1da177e4 1460
0660e03f 1461 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
29556526
LY
1462 !(ipv6_addr_type(target) & IPV6_ADDR_LINKLOCAL)) {
1463 ND_PRINTK2(KERN_WARNING
1464 "ICMPv6 Redirect: target address is not link-local.\n");
1465 return;
1466 }
1467
0660e03f 1468 ndisc_flow_init(&fl, NDISC_REDIRECT, &saddr_buf, &ipv6_hdr(skb)->saddr,
af184765 1469 dev->ifindex);
1da177e4
LT
1470
1471 dst = ip6_route_output(NULL, &fl);
1472 if (dst == NULL)
1473 return;
1474
1475 err = xfrm_lookup(&dst, &fl, NULL, 0);
e104411b 1476 if (err)
1da177e4 1477 return;
1da177e4
LT
1478
1479 rt = (struct rt6_info *) dst;
1480
1481 if (rt->rt6i_flags & RTF_GATEWAY) {
1482 ND_PRINTK2(KERN_WARNING
1483 "ICMPv6 Redirect: destination is not a neighbour.\n");
1484 dst_release(dst);
1485 return;
1486 }
1487 if (!xrlim_allow(dst, 1*HZ)) {
1488 dst_release(dst);
1489 return;
1490 }
1491
1492 if (dev->addr_len) {
1493 read_lock_bh(&neigh->lock);
1494 if (neigh->nud_state & NUD_VALID) {
1495 memcpy(ha_buf, neigh->ha, dev->addr_len);
1496 read_unlock_bh(&neigh->lock);
1497 ha = ha_buf;
1498 len += ndisc_opt_addr_space(dev);
1499 } else
1500 read_unlock_bh(&neigh->lock);
1501 }
1502
1503 rd_len = min_t(unsigned int,
1504 IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1505 rd_len &= ~0x7;
1506 len += rd_len;
1507
d54a81d3
DM
1508 buff = sock_alloc_send_skb(sk,
1509 (MAX_HEADER + sizeof(struct ipv6hdr) +
1510 len + LL_RESERVED_SPACE(dev)),
1da177e4
LT
1511 1, &err);
1512 if (buff == NULL) {
1513 ND_PRINTK0(KERN_ERR
1514 "ICMPv6 Redirect: %s() failed to allocate an skb.\n",
1515 __FUNCTION__);
1516 dst_release(dst);
1517 return;
1518 }
1519
1520 hlen = 0;
1521
1522 skb_reserve(buff, LL_RESERVED_SPACE(dev));
0660e03f 1523 ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1da177e4
LT
1524 IPPROTO_ICMPV6, len);
1525
27a884dc 1526 skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
d10ba34b
ACM
1527 skb_put(buff, len);
1528 icmph = icmp6_hdr(buff);
1da177e4
LT
1529
1530 memset(icmph, 0, sizeof(struct icmp6hdr));
1531 icmph->icmp6_type = NDISC_REDIRECT;
1532
1533 /*
1534 * copy target and destination addresses
1535 */
1536
1537 addrp = (struct in6_addr *)(icmph + 1);
1538 ipv6_addr_copy(addrp, target);
1539 addrp++;
0660e03f 1540 ipv6_addr_copy(addrp, &ipv6_hdr(skb)->daddr);
1da177e4
LT
1541
1542 opt = (u8*) (addrp + 1);
1543
1544 /*
1545 * include target_address option
1546 */
1547
1548 if (ha)
1549 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1550 dev->addr_len, dev->type);
1551
1552 /*
1553 * build redirect option and copy skb over to the new packet.
1554 */
1555
1ab1457c 1556 memset(opt, 0, 8);
1da177e4
LT
1557 *(opt++) = ND_OPT_REDIRECT_HDR;
1558 *(opt++) = (rd_len >> 3);
1559 opt += 6;
1560
0660e03f 1561 memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1da177e4 1562
0660e03f 1563 icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1da177e4
LT
1564 len, IPPROTO_ICMPV6,
1565 csum_partial((u8 *) icmph, len, 0));
1566
1567 buff->dst = dst;
1568 idev = in6_dev_get(dst->dev);
a11d206d 1569 IP6_INC_STATS(idev, IPSTATS_MIB_OUTREQUESTS);
1da177e4
LT
1570 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, buff, NULL, dst->dev, dst_output);
1571 if (!err) {
1572 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTREDIRECTS);
1573 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
1574 }
1575
1576 if (likely(idev != NULL))
1577 in6_dev_put(idev);
1578}
1579
1580static void pndisc_redo(struct sk_buff *skb)
1581{
140e26fc 1582 ndisc_recv_ns(skb);
1da177e4
LT
1583 kfree_skb(skb);
1584}
1585
1586int ndisc_rcv(struct sk_buff *skb)
1587{
1588 struct nd_msg *msg;
1589
1590 if (!pskb_may_pull(skb, skb->len))
1591 return 0;
1592
9c70220b 1593 msg = (struct nd_msg *)skb_transport_header(skb);
1da177e4 1594
9c70220b 1595 __skb_push(skb, skb->data - skb_transport_header(skb));
1da177e4 1596
0660e03f 1597 if (ipv6_hdr(skb)->hop_limit != 255) {
1da177e4
LT
1598 ND_PRINTK2(KERN_WARNING
1599 "ICMPv6 NDISC: invalid hop-limit: %d\n",
0660e03f 1600 ipv6_hdr(skb)->hop_limit);
1da177e4
LT
1601 return 0;
1602 }
1603
1604 if (msg->icmph.icmp6_code != 0) {
1ab1457c 1605 ND_PRINTK2(KERN_WARNING
1da177e4
LT
1606 "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1607 msg->icmph.icmp6_code);
1608 return 0;
1609 }
1610
a61bbcf2
PM
1611 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1612
1da177e4
LT
1613 switch (msg->icmph.icmp6_type) {
1614 case NDISC_NEIGHBOUR_SOLICITATION:
1615 ndisc_recv_ns(skb);
1616 break;
1617
1618 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1619 ndisc_recv_na(skb);
1620 break;
1621
1622 case NDISC_ROUTER_SOLICITATION:
1623 ndisc_recv_rs(skb);
1624 break;
1625
1626 case NDISC_ROUTER_ADVERTISEMENT:
1627 ndisc_router_discovery(skb);
1628 break;
1629
1630 case NDISC_REDIRECT:
1631 ndisc_redirect_rcv(skb);
1632 break;
1633 };
1634
1635 return 0;
1636}
1637
1638static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1639{
1640 struct net_device *dev = ptr;
1641
1642 switch (event) {
1643 case NETDEV_CHANGEADDR:
1644 neigh_changeaddr(&nd_tbl, dev);
1645 fib6_run_gc(~0UL);
1646 break;
1647 case NETDEV_DOWN:
1648 neigh_ifdown(&nd_tbl, dev);
1649 fib6_run_gc(~0UL);
1650 break;
1651 default:
1652 break;
1653 }
1654
1655 return NOTIFY_DONE;
1656}
1657
1658static struct notifier_block ndisc_netdev_notifier = {
1659 .notifier_call = ndisc_netdev_event,
1660};
1661
1662#ifdef CONFIG_SYSCTL
1663static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1664 const char *func, const char *dev_name)
1665{
1666 static char warncomm[TASK_COMM_LEN];
1667 static int warned;
1668 if (strcmp(warncomm, current->comm) && warned < 5) {
1669 strcpy(warncomm, current->comm);
1670 printk(KERN_WARNING
1671 "process `%s' is using deprecated sysctl (%s) "
1672 "net.ipv6.neigh.%s.%s; "
1673 "Use net.ipv6.neigh.%s.%s_ms "
1674 "instead.\n",
1675 warncomm, func,
1676 dev_name, ctl->procname,
1677 dev_name, ctl->procname);
1678 warned++;
1679 }
1680}
1681
1682int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, struct file * filp, void __user *buffer, size_t *lenp, loff_t *ppos)
1683{
1684 struct net_device *dev = ctl->extra1;
1685 struct inet6_dev *idev;
1686 int ret;
1687
1688 if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1689 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1690 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1691
1692 switch (ctl->ctl_name) {
1693 case NET_NEIGH_RETRANS_TIME:
1694 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1695 break;
1696 case NET_NEIGH_REACHABLE_TIME:
1697 ret = proc_dointvec_jiffies(ctl, write,
1698 filp, buffer, lenp, ppos);
1699 break;
1700 case NET_NEIGH_RETRANS_TIME_MS:
1701 case NET_NEIGH_REACHABLE_TIME_MS:
1702 ret = proc_dointvec_ms_jiffies(ctl, write,
1703 filp, buffer, lenp, ppos);
1704 break;
1705 default:
1706 ret = -1;
1707 }
1708
1709 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1710 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1711 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1712 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1713 idev->tstamp = jiffies;
1714 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1715 in6_dev_put(idev);
1716 }
1717 return ret;
1718}
1719
1720static int ndisc_ifinfo_sysctl_strategy(ctl_table *ctl, int __user *name,
1721 int nlen, void __user *oldval,
1722 size_t __user *oldlenp,
1f29bcd7 1723 void __user *newval, size_t newlen)
1da177e4
LT
1724{
1725 struct net_device *dev = ctl->extra1;
1726 struct inet6_dev *idev;
1727 int ret;
1728
1729 if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1730 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1731 ndisc_warn_deprecated_sysctl(ctl, "procfs", dev ? dev->name : "default");
1732
1733 switch (ctl->ctl_name) {
1734 case NET_NEIGH_REACHABLE_TIME:
1735 ret = sysctl_jiffies(ctl, name, nlen,
1f29bcd7 1736 oldval, oldlenp, newval, newlen);
1da177e4
LT
1737 break;
1738 case NET_NEIGH_RETRANS_TIME_MS:
1739 case NET_NEIGH_REACHABLE_TIME_MS:
1740 ret = sysctl_ms_jiffies(ctl, name, nlen,
1f29bcd7 1741 oldval, oldlenp, newval, newlen);
1da177e4
LT
1742 break;
1743 default:
1744 ret = 0;
1745 }
1746
1747 if (newval && newlen && ret > 0 &&
1748 dev && (idev = in6_dev_get(dev)) != NULL) {
1749 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1750 ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1751 idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1752 idev->tstamp = jiffies;
1753 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1754 in6_dev_put(idev);
1755 }
1756
1757 return ret;
1758}
1759
1760#endif
1761
1762int __init ndisc_init(struct net_proto_family *ops)
1763{
1764 struct ipv6_pinfo *np;
1765 struct sock *sk;
1ab1457c 1766 int err;
1da177e4
LT
1767
1768 err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, &ndisc_socket);
1769 if (err < 0) {
1770 ND_PRINTK0(KERN_ERR
1ab1457c 1771 "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1da177e4
LT
1772 err);
1773 ndisc_socket = NULL; /* For safety. */
1774 return err;
1775 }
1776
1777 sk = ndisc_socket->sk;
1778 np = inet6_sk(sk);
1779 sk->sk_allocation = GFP_ATOMIC;
1780 np->hop_limit = 255;
1781 /* Do not loopback ndisc messages */
1782 np->mc_loop = 0;
1783 sk->sk_prot->unhash(sk);
1784
1ab1457c
YH
1785 /*
1786 * Initialize the neighbour table
1787 */
1788
1da177e4
LT
1789 neigh_table_init(&nd_tbl);
1790
1791#ifdef CONFIG_SYSCTL
1ab1457c 1792 neigh_sysctl_register(NULL, &nd_tbl.parms, NET_IPV6, NET_IPV6_NEIGH,
1da177e4
LT
1793 "ipv6",
1794 &ndisc_ifinfo_sysctl_change,
1795 &ndisc_ifinfo_sysctl_strategy);
1796#endif
1797
1798 register_netdevice_notifier(&ndisc_netdev_notifier);
1799 return 0;
1800}
1801
1802void ndisc_cleanup(void)
1803{
36f73d0c 1804 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1da177e4
LT
1805#ifdef CONFIG_SYSCTL
1806 neigh_sysctl_unregister(&nd_tbl.parms);
1807#endif
1808 neigh_table_clear(&nd_tbl);
1809 sock_release(ndisc_socket);
1810 ndisc_socket = NULL; /* For safety. */
1811}