0304b5fe8d6aa01d77378d4ff939e8f5f04c1d4f
[pandora-kernel.git] / net / ipv6 / ndisc.c
1 /*
2  *      Neighbour Discovery for IPv6
3  *      Linux INET6 implementation 
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>     
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>
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
65 #include <linux/if_addr.h>
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
88 static struct socket *ndisc_socket;
89
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);
97
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,
106 };
107
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,
116 };
117
118
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,
125 };
126
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,
150         },
151         .gc_interval =    30 * HZ,
152         .gc_thresh1 =    128,
153         .gc_thresh2 =    512,
154         .gc_thresh3 =   1024,
155 };
156
157 /* ND options */
158 struct ndisc_options {
159         struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
160 #ifdef CONFIG_IPV6_ROUTE_INFO
161         struct nd_opt_hdr *nd_opts_ri;
162         struct nd_opt_hdr *nd_opts_ri_end;
163 #endif
164 };
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  */
181 static 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
189 static 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
194 static 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
215 static 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
228 static 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;
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
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
285 static 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
296 int 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
322 static u32 ndisc_hash(const void *pkey, const struct net_device *dev)
323 {
324         const u32 *p32 = pkey;
325         u32 addr_hash, i;
326
327         addr_hash = 0;
328         for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
329                 addr_hash ^= *p32++;
330
331         return jhash_2words(addr_hash, dev->ifindex, nd_tbl.hash_rnd);
332 }
333
334 static int ndisc_constructor(struct neighbour *neigh)
335 {
336         struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
337         struct net_device *dev = neigh->dev;
338         struct inet6_dev *in6_dev;
339         struct neigh_parms *parms;
340         int is_multicast = ipv6_addr_is_multicast(addr);
341
342         rcu_read_lock();
343         in6_dev = in6_dev_get(dev);
344         if (in6_dev == NULL) {
345                 rcu_read_unlock();
346                 return -EINVAL;
347         }
348
349         parms = in6_dev->nd_parms;
350         __neigh_parms_put(neigh->parms);
351         neigh->parms = neigh_parms_clone(parms);
352         rcu_read_unlock();
353
354         neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
355         if (dev->hard_header == NULL) {
356                 neigh->nud_state = NUD_NOARP;
357                 neigh->ops = &ndisc_direct_ops;
358                 neigh->output = neigh->ops->queue_xmit;
359         } else {
360                 if (is_multicast) {
361                         neigh->nud_state = NUD_NOARP;
362                         ndisc_mc_map(addr, neigh->ha, dev, 1);
363                 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
364                         neigh->nud_state = NUD_NOARP;
365                         memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
366                         if (dev->flags&IFF_LOOPBACK)
367                                 neigh->type = RTN_LOCAL;
368                 } else if (dev->flags&IFF_POINTOPOINT) {
369                         neigh->nud_state = NUD_NOARP;
370                         memcpy(neigh->ha, dev->broadcast, dev->addr_len);
371                 }
372                 if (dev->hard_header_cache)
373                         neigh->ops = &ndisc_hh_ops;
374                 else
375                         neigh->ops = &ndisc_generic_ops;
376                 if (neigh->nud_state&NUD_VALID)
377                         neigh->output = neigh->ops->connected_output;
378                 else
379                         neigh->output = neigh->ops->output;
380         }
381         in6_dev_put(in6_dev);
382         return 0;
383 }
384
385 static int pndisc_constructor(struct pneigh_entry *n)
386 {
387         struct in6_addr *addr = (struct in6_addr*)&n->key;
388         struct in6_addr maddr;
389         struct net_device *dev = n->dev;
390
391         if (dev == NULL || __in6_dev_get(dev) == NULL)
392                 return -EINVAL;
393         addrconf_addr_solict_mult(addr, &maddr);
394         ipv6_dev_mc_inc(dev, &maddr);
395         return 0;
396 }
397
398 static void pndisc_destructor(struct pneigh_entry *n)
399 {
400         struct in6_addr *addr = (struct in6_addr*)&n->key;
401         struct in6_addr maddr;
402         struct net_device *dev = n->dev;
403
404         if (dev == NULL || __in6_dev_get(dev) == NULL)
405                 return;
406         addrconf_addr_solict_mult(addr, &maddr);
407         ipv6_dev_mc_dec(dev, &maddr);
408 }
409
410 /*
411  *      Send a Neighbour Advertisement
412  */
413
414 static inline void ndisc_flow_init(struct flowi *fl, u8 type,
415                             struct in6_addr *saddr, struct in6_addr *daddr,
416                             int oif)
417 {
418         memset(fl, 0, sizeof(*fl));
419         ipv6_addr_copy(&fl->fl6_src, saddr);
420         ipv6_addr_copy(&fl->fl6_dst, daddr);
421         fl->proto               = IPPROTO_ICMPV6;
422         fl->fl_icmp_type        = type;
423         fl->fl_icmp_code        = 0;
424         fl->oif                 = oif;
425         security_sk_classify_flow(ndisc_socket->sk, fl);
426 }
427
428 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
429                    struct in6_addr *daddr, struct in6_addr *solicited_addr,
430                    int router, int solicited, int override, int inc_opt) 
431 {
432         struct in6_addr tmpaddr;
433         struct inet6_ifaddr *ifp;
434         struct inet6_dev *idev;
435         struct flowi fl;
436         struct dst_entry* dst;
437         struct sock *sk = ndisc_socket->sk;
438         struct in6_addr *src_addr;
439         struct nd_msg *msg;
440         int len;
441         struct sk_buff *skb;
442         int err;
443
444         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
445
446         /* for anycast or proxy, solicited_addr != src_addr */
447         ifp = ipv6_get_ifaddr(solicited_addr, dev, 1);
448         if (ifp) {
449                 src_addr = solicited_addr;
450                 in6_ifa_put(ifp);
451         } else {
452                 if (ipv6_dev_get_saddr(dev, daddr, &tmpaddr))
453                         return;
454                 src_addr = &tmpaddr;
455         }
456
457         ndisc_flow_init(&fl, NDISC_NEIGHBOUR_ADVERTISEMENT, src_addr, daddr,
458                         dev->ifindex);
459
460         dst = ndisc_dst_alloc(dev, neigh, daddr, ip6_output);
461         if (!dst)
462                 return;
463
464         err = xfrm_lookup(&dst, &fl, NULL, 0);
465         if (err < 0)
466                 return;
467
468         if (inc_opt) {
469                 if (dev->addr_len)
470                         len += ndisc_opt_addr_space(dev);
471                 else
472                         inc_opt = 0;
473         }
474
475         skb = sock_alloc_send_skb(sk, MAX_HEADER + len + LL_RESERVED_SPACE(dev),
476                                   1, &err);
477
478         if (skb == NULL) {
479                 ND_PRINTK0(KERN_ERR
480                            "ICMPv6 NA: %s() failed to allocate an skb.\n", 
481                            __FUNCTION__);
482                 dst_release(dst);
483                 return;
484         }
485
486         skb_reserve(skb, LL_RESERVED_SPACE(dev));
487         ip6_nd_hdr(sk, skb, dev, src_addr, daddr, IPPROTO_ICMPV6, len);
488
489         msg = (struct nd_msg *)skb_put(skb, len);
490         skb->h.raw = (unsigned char*)msg;
491
492         msg->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
493         msg->icmph.icmp6_code = 0;
494         msg->icmph.icmp6_cksum = 0;
495
496         msg->icmph.icmp6_unused = 0;
497         msg->icmph.icmp6_router    = router;
498         msg->icmph.icmp6_solicited = solicited;
499         msg->icmph.icmp6_override  = override;
500
501         /* Set the target address. */
502         ipv6_addr_copy(&msg->target, solicited_addr);
503
504         if (inc_opt)
505                 ndisc_fill_addr_option(msg->opt, ND_OPT_TARGET_LL_ADDR, dev->dev_addr,
506                                        dev->addr_len, dev->type);
507
508         /* checksum */
509         msg->icmph.icmp6_cksum = csum_ipv6_magic(src_addr, daddr, len, 
510                                                  IPPROTO_ICMPV6,
511                                                  csum_partial((__u8 *) msg, 
512                                                               len, 0));
513
514         skb->dst = dst;
515         idev = in6_dev_get(dst->dev);
516         IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
517         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
518         if (!err) {
519                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTNEIGHBORADVERTISEMENTS);
520                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
521         }
522
523         if (likely(idev != NULL))
524                 in6_dev_put(idev);
525 }        
526
527 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
528                    struct in6_addr *solicit,
529                    struct in6_addr *daddr, struct in6_addr *saddr) 
530 {
531         struct flowi fl;
532         struct dst_entry* dst;
533         struct inet6_dev *idev;
534         struct sock *sk = ndisc_socket->sk;
535         struct sk_buff *skb;
536         struct nd_msg *msg;
537         struct in6_addr addr_buf;
538         int len;
539         int err;
540         int send_llinfo;
541
542         if (saddr == NULL) {
543                 if (ipv6_get_lladdr(dev, &addr_buf))
544                         return;
545                 saddr = &addr_buf;
546         }
547
548         ndisc_flow_init(&fl, NDISC_NEIGHBOUR_SOLICITATION, saddr, daddr,
549                         dev->ifindex);
550
551         dst = ndisc_dst_alloc(dev, neigh, daddr, ip6_output);
552         if (!dst)
553                 return;
554
555         err = xfrm_lookup(&dst, &fl, NULL, 0);
556         if (err < 0)
557                 return;
558
559         len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
560         send_llinfo = dev->addr_len && !ipv6_addr_any(saddr);
561         if (send_llinfo)
562                 len += ndisc_opt_addr_space(dev);
563
564         skb = sock_alloc_send_skb(sk, MAX_HEADER + len + LL_RESERVED_SPACE(dev),
565                                   1, &err);
566         if (skb == NULL) {
567                 ND_PRINTK0(KERN_ERR
568                            "ICMPv6 NA: %s() failed to allocate an skb.\n", 
569                            __FUNCTION__);
570                 dst_release(dst);
571                 return;
572         }
573
574         skb_reserve(skb, LL_RESERVED_SPACE(dev));
575         ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
576
577         msg = (struct nd_msg *)skb_put(skb, len);
578         skb->h.raw = (unsigned char*)msg;
579         msg->icmph.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION;
580         msg->icmph.icmp6_code = 0;
581         msg->icmph.icmp6_cksum = 0;
582         msg->icmph.icmp6_unused = 0;
583
584         /* Set the target address. */
585         ipv6_addr_copy(&msg->target, solicit);
586
587         if (send_llinfo)
588                 ndisc_fill_addr_option(msg->opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr,
589                                        dev->addr_len, dev->type);
590
591         /* checksum */
592         msg->icmph.icmp6_cksum = csum_ipv6_magic(&skb->nh.ipv6h->saddr,
593                                                  daddr, len, 
594                                                  IPPROTO_ICMPV6,
595                                                  csum_partial((__u8 *) msg, 
596                                                               len, 0));
597         /* send it! */
598         skb->dst = dst;
599         idev = in6_dev_get(dst->dev);
600         IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
601         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
602         if (!err) {
603                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTNEIGHBORSOLICITS);
604                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
605         }
606
607         if (likely(idev != NULL))
608                 in6_dev_put(idev);
609 }
610
611 void ndisc_send_rs(struct net_device *dev, struct in6_addr *saddr,
612                    struct in6_addr *daddr)
613 {
614         struct flowi fl;
615         struct dst_entry* dst;
616         struct inet6_dev *idev;
617         struct sock *sk = ndisc_socket->sk;
618         struct sk_buff *skb;
619         struct icmp6hdr *hdr;
620         __u8 * opt;
621         int len;
622         int err;
623
624         ndisc_flow_init(&fl, NDISC_ROUTER_SOLICITATION, saddr, daddr,
625                         dev->ifindex);
626
627         dst = ndisc_dst_alloc(dev, NULL, daddr, ip6_output);
628         if (!dst)
629                 return;
630
631         err = xfrm_lookup(&dst, &fl, NULL, 0);
632         if (err < 0)
633                 return;
634
635         len = sizeof(struct icmp6hdr);
636         if (dev->addr_len)
637                 len += ndisc_opt_addr_space(dev);
638
639         skb = sock_alloc_send_skb(sk, MAX_HEADER + len + LL_RESERVED_SPACE(dev),
640                                   1, &err);
641         if (skb == NULL) {
642                 ND_PRINTK0(KERN_ERR
643                            "ICMPv6 RS: %s() failed to allocate an skb.\n", 
644                            __FUNCTION__);
645                 dst_release(dst);
646                 return;
647         }
648
649         skb_reserve(skb, LL_RESERVED_SPACE(dev));
650         ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
651
652         hdr = (struct icmp6hdr *)skb_put(skb, len);
653         skb->h.raw = (unsigned char*)hdr;
654         hdr->icmp6_type = NDISC_ROUTER_SOLICITATION;
655         hdr->icmp6_code = 0;
656         hdr->icmp6_cksum = 0;
657         hdr->icmp6_unused = 0;
658
659         opt = (u8*) (hdr + 1);
660
661         if (dev->addr_len)
662                 ndisc_fill_addr_option(opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr,
663                                        dev->addr_len, dev->type);
664
665         /* checksum */
666         hdr->icmp6_cksum = csum_ipv6_magic(&skb->nh.ipv6h->saddr, daddr, len,
667                                            IPPROTO_ICMPV6,
668                                            csum_partial((__u8 *) hdr, len, 0));
669
670         /* send it! */
671         skb->dst = dst;
672         idev = in6_dev_get(dst->dev);
673         IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS); 
674         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
675         if (!err) {
676                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTROUTERSOLICITS);
677                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
678         }
679
680         if (likely(idev != NULL))
681                 in6_dev_put(idev);
682 }
683                    
684
685 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
686 {
687         /*
688          *      "The sender MUST return an ICMP
689          *       destination unreachable"
690          */
691         dst_link_failure(skb);
692         kfree_skb(skb);
693 }
694
695 /* Called with locked neigh: either read or both */
696
697 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
698 {
699         struct in6_addr *saddr = NULL;
700         struct in6_addr mcaddr;
701         struct net_device *dev = neigh->dev;
702         struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
703         int probes = atomic_read(&neigh->probes);
704
705         if (skb && ipv6_chk_addr(&skb->nh.ipv6h->saddr, dev, 1))
706                 saddr = &skb->nh.ipv6h->saddr;
707
708         if ((probes -= neigh->parms->ucast_probes) < 0) {
709                 if (!(neigh->nud_state & NUD_VALID)) {
710                         ND_PRINTK1(KERN_DEBUG
711                                    "%s(): trying to ucast probe in NUD_INVALID: "
712                                    NIP6_FMT "\n",
713                                    __FUNCTION__,
714                                    NIP6(*target));
715                 }
716                 ndisc_send_ns(dev, neigh, target, target, saddr);
717         } else if ((probes -= neigh->parms->app_probes) < 0) {
718 #ifdef CONFIG_ARPD
719                 neigh_app_ns(neigh);
720 #endif
721         } else {
722                 addrconf_addr_solict_mult(target, &mcaddr);
723                 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
724         }
725 }
726
727 static void ndisc_recv_ns(struct sk_buff *skb)
728 {
729         struct nd_msg *msg = (struct nd_msg *)skb->h.raw;
730         struct in6_addr *saddr = &skb->nh.ipv6h->saddr;
731         struct in6_addr *daddr = &skb->nh.ipv6h->daddr;
732         u8 *lladdr = NULL;
733         u32 ndoptlen = skb->tail - msg->opt;
734         struct ndisc_options ndopts;
735         struct net_device *dev = skb->dev;
736         struct inet6_ifaddr *ifp;
737         struct inet6_dev *idev = NULL;
738         struct neighbour *neigh;
739         struct pneigh_entry *pneigh = NULL;
740         int dad = ipv6_addr_any(saddr);
741         int inc;
742         int is_router;
743
744         if (ipv6_addr_is_multicast(&msg->target)) {
745                 ND_PRINTK2(KERN_WARNING 
746                            "ICMPv6 NS: multicast target address");
747                 return;
748         }
749
750         /*
751          * RFC2461 7.1.1:
752          * DAD has to be destined for solicited node multicast address.
753          */
754         if (dad &&
755             !(daddr->s6_addr32[0] == htonl(0xff020000) &&
756               daddr->s6_addr32[1] == htonl(0x00000000) &&
757               daddr->s6_addr32[2] == htonl(0x00000001) &&
758               daddr->s6_addr [12] == 0xff )) {
759                 ND_PRINTK2(KERN_WARNING
760                            "ICMPv6 NS: bad DAD packet (wrong destination)\n");
761                 return;
762         }
763
764         if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
765                 ND_PRINTK2(KERN_WARNING 
766                            "ICMPv6 NS: invalid ND options\n");
767                 return;
768         }
769
770         if (ndopts.nd_opts_src_lladdr) {
771                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
772                 if (!lladdr) {
773                         ND_PRINTK2(KERN_WARNING
774                                    "ICMPv6 NS: invalid link-layer address length\n");
775                         return;
776                 }
777
778                 /* RFC2461 7.1.1:
779                  *      If the IP source address is the unspecified address, 
780                  *      there MUST NOT be source link-layer address option 
781                  *      in the message.
782                  */
783                 if (dad) {
784                         ND_PRINTK2(KERN_WARNING 
785                                    "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
786                         return;
787                 }
788         }
789
790         inc = ipv6_addr_is_multicast(daddr);
791
792         if ((ifp = ipv6_get_ifaddr(&msg->target, dev, 1)) != NULL) {
793                 if (ifp->flags & IFA_F_TENTATIVE) {
794                         /* Address is tentative. If the source
795                            is unspecified address, it is someone
796                            does DAD, otherwise we ignore solicitations
797                            until DAD timer expires.
798                          */
799                         if (!dad)
800                                 goto out;
801                         if (dev->type == ARPHRD_IEEE802_TR) {
802                                 unsigned char *sadr = skb->mac.raw;
803                                 if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
804                                     sadr[9] == dev->dev_addr[1] &&
805                                     sadr[10] == dev->dev_addr[2] &&
806                                     sadr[11] == dev->dev_addr[3] &&
807                                     sadr[12] == dev->dev_addr[4] &&
808                                     sadr[13] == dev->dev_addr[5]) {
809                                         /* looped-back to us */
810                                         goto out;
811                                 }
812                         }
813                         addrconf_dad_failure(ifp); 
814                         return;
815                 }
816
817                 idev = ifp->idev;
818         } else {
819                 idev = in6_dev_get(dev);
820                 if (!idev) {
821                         /* XXX: count this drop? */
822                         return;
823                 }
824
825                 if (ipv6_chk_acast_addr(dev, &msg->target) ||
826                     (idev->cnf.forwarding && 
827                      (ipv6_devconf.proxy_ndp || idev->cnf.proxy_ndp) &&
828                      (pneigh = pneigh_lookup(&nd_tbl,
829                                              &msg->target, dev, 0)) != NULL)) {
830                         if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
831                             skb->pkt_type != PACKET_HOST &&
832                             inc != 0 &&
833                             idev->nd_parms->proxy_delay != 0) {
834                                 /*
835                                  * for anycast or proxy,
836                                  * sender should delay its response 
837                                  * by a random time between 0 and 
838                                  * MAX_ANYCAST_DELAY_TIME seconds.
839                                  * (RFC2461) -- yoshfuji
840                                  */
841                                 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
842                                 if (n)
843                                         pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
844                                 goto out;
845                         }
846                 } else
847                         goto out;
848         }
849
850         is_router = !!(pneigh ? pneigh->flags & NTF_ROUTER : idev->cnf.forwarding);
851
852         if (dad) {
853                 struct in6_addr maddr;
854
855                 ipv6_addr_all_nodes(&maddr);
856                 ndisc_send_na(dev, NULL, &maddr, &msg->target,
857                               is_router, 0, (ifp != NULL), 1);
858                 goto out;
859         }
860
861         if (inc)
862                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
863         else
864                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
865
866         /* 
867          *      update / create cache entry
868          *      for the source address
869          */
870         neigh = __neigh_lookup(&nd_tbl, saddr, dev,
871                                !inc || lladdr || !dev->addr_len);
872         if (neigh)
873                 neigh_update(neigh, lladdr, NUD_STALE, 
874                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
875                              NEIGH_UPDATE_F_OVERRIDE);
876         if (neigh || !dev->hard_header) {
877                 ndisc_send_na(dev, neigh, saddr, &msg->target,
878                               is_router,
879                               1, (ifp != NULL && inc), inc);
880                 if (neigh)
881                         neigh_release(neigh);
882         }
883
884 out:
885         if (ifp)
886                 in6_ifa_put(ifp);
887         else
888                 in6_dev_put(idev);
889
890         return;
891 }
892
893 static void ndisc_recv_na(struct sk_buff *skb)
894 {
895         struct nd_msg *msg = (struct nd_msg *)skb->h.raw;
896         struct in6_addr *saddr = &skb->nh.ipv6h->saddr;
897         struct in6_addr *daddr = &skb->nh.ipv6h->daddr;
898         u8 *lladdr = NULL;
899         u32 ndoptlen = skb->tail - msg->opt;
900         struct ndisc_options ndopts;
901         struct net_device *dev = skb->dev;
902         struct inet6_ifaddr *ifp;
903         struct neighbour *neigh;
904
905         if (skb->len < sizeof(struct nd_msg)) {
906                 ND_PRINTK2(KERN_WARNING
907                            "ICMPv6 NA: packet too short\n");
908                 return;
909         }
910
911         if (ipv6_addr_is_multicast(&msg->target)) {
912                 ND_PRINTK2(KERN_WARNING
913                            "ICMPv6 NA: target address is multicast.\n");
914                 return;
915         }
916
917         if (ipv6_addr_is_multicast(daddr) &&
918             msg->icmph.icmp6_solicited) {
919                 ND_PRINTK2(KERN_WARNING
920                            "ICMPv6 NA: solicited NA is multicasted.\n");
921                 return;
922         }
923                 
924         if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
925                 ND_PRINTK2(KERN_WARNING
926                            "ICMPv6 NS: invalid ND option\n");
927                 return;
928         }
929         if (ndopts.nd_opts_tgt_lladdr) {
930                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
931                 if (!lladdr) {
932                         ND_PRINTK2(KERN_WARNING
933                                    "ICMPv6 NA: invalid link-layer address length\n");
934                         return;
935                 }
936         }
937         if ((ifp = ipv6_get_ifaddr(&msg->target, dev, 1))) {
938                 if (ifp->flags & IFA_F_TENTATIVE) {
939                         addrconf_dad_failure(ifp);
940                         return;
941                 }
942                 /* What should we make now? The advertisement
943                    is invalid, but ndisc specs say nothing
944                    about it. It could be misconfiguration, or
945                    an smart proxy agent tries to help us :-)
946                  */
947                 ND_PRINTK1(KERN_WARNING
948                            "ICMPv6 NA: someone advertises our address on %s!\n",
949                            ifp->idev->dev->name);
950                 in6_ifa_put(ifp);
951                 return;
952         }
953         neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
954
955         if (neigh) {
956                 u8 old_flags = neigh->flags;
957
958                 if (neigh->nud_state & NUD_FAILED)
959                         goto out;
960
961                 /*
962                  * Don't update the neighbor cache entry on a proxy NA from
963                  * ourselves because either the proxied node is off link or it
964                  * has already sent a NA to us.
965                  */
966                 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
967                     ipv6_devconf.forwarding && ipv6_devconf.proxy_ndp &&
968                     pneigh_lookup(&nd_tbl, &msg->target, dev, 0)) {
969                         /* XXX: idev->cnf.prixy_ndp */
970                         WARN_ON(skb->dst != NULL &&
971                                 ((struct rt6_info *)skb->dst)->rt6i_idev);
972                         goto out;
973                 }
974
975                 neigh_update(neigh, lladdr,
976                              msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
977                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
978                              (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
979                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
980                              (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
981
982                 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
983                         /*
984                          * Change: router to host
985                          */
986                         struct rt6_info *rt;
987                         rt = rt6_get_dflt_router(saddr, dev);
988                         if (rt)
989                                 ip6_del_rt(rt);
990                 }
991
992 out:
993                 neigh_release(neigh);
994         }
995 }
996
997 static void ndisc_recv_rs(struct sk_buff *skb)
998 {
999         struct rs_msg *rs_msg = (struct rs_msg *) skb->h.raw;
1000         unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1001         struct neighbour *neigh;
1002         struct inet6_dev *idev;
1003         struct in6_addr *saddr = &skb->nh.ipv6h->saddr;
1004         struct ndisc_options ndopts;
1005         u8 *lladdr = NULL;
1006
1007         if (skb->len < sizeof(*rs_msg))
1008                 return;
1009
1010         idev = in6_dev_get(skb->dev);
1011         if (!idev) {
1012                 if (net_ratelimit())
1013                         ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
1014                 return;
1015         }
1016
1017         /* Don't accept RS if we're not in router mode */
1018         if (!idev->cnf.forwarding)
1019                 goto out;
1020
1021         /*
1022          * Don't update NCE if src = ::;
1023          * this implies that the source node has no ip address assigned yet.
1024          */
1025         if (ipv6_addr_any(saddr))
1026                 goto out;
1027
1028         /* Parse ND options */
1029         if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1030                 if (net_ratelimit())
1031                         ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1032                 goto out;
1033         }
1034
1035         if (ndopts.nd_opts_src_lladdr) {
1036                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1037                                              skb->dev);
1038                 if (!lladdr)
1039                         goto out;
1040         }
1041
1042         neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1043         if (neigh) {
1044                 neigh_update(neigh, lladdr, NUD_STALE,
1045                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1046                              NEIGH_UPDATE_F_OVERRIDE|
1047                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1048                 neigh_release(neigh);
1049         }
1050 out:
1051         in6_dev_put(idev);
1052 }
1053
1054 static void ndisc_router_discovery(struct sk_buff *skb)
1055 {
1056         struct ra_msg *ra_msg = (struct ra_msg *) skb->h.raw;
1057         struct neighbour *neigh = NULL;
1058         struct inet6_dev *in6_dev;
1059         struct rt6_info *rt = NULL;
1060         int lifetime;
1061         struct ndisc_options ndopts;
1062         int optlen;
1063         unsigned int pref = 0;
1064
1065         __u8 * opt = (__u8 *)(ra_msg + 1);
1066
1067         optlen = (skb->tail - skb->h.raw) - sizeof(struct ra_msg);
1068
1069         if (!(ipv6_addr_type(&skb->nh.ipv6h->saddr) & IPV6_ADDR_LINKLOCAL)) {
1070                 ND_PRINTK2(KERN_WARNING
1071                            "ICMPv6 RA: source address is not link-local.\n");
1072                 return;
1073         }
1074         if (optlen < 0) {
1075                 ND_PRINTK2(KERN_WARNING 
1076                            "ICMPv6 RA: packet too short\n");
1077                 return;
1078         }
1079
1080         /*
1081          *      set the RA_RECV flag in the interface
1082          */
1083
1084         in6_dev = in6_dev_get(skb->dev);
1085         if (in6_dev == NULL) {
1086                 ND_PRINTK0(KERN_ERR
1087                            "ICMPv6 RA: can't find inet6 device for %s.\n",
1088                            skb->dev->name);
1089                 return;
1090         }
1091         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra) {
1092                 in6_dev_put(in6_dev);
1093                 return;
1094         }
1095
1096         if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1097                 in6_dev_put(in6_dev);
1098                 ND_PRINTK2(KERN_WARNING
1099                            "ICMP6 RA: invalid ND options\n");
1100                 return;
1101         }
1102
1103         if (in6_dev->if_flags & IF_RS_SENT) {
1104                 /*
1105                  *      flag that an RA was received after an RS was sent
1106                  *      out on this interface.
1107                  */
1108                 in6_dev->if_flags |= IF_RA_RCVD;
1109         }
1110
1111         /*
1112          * Remember the managed/otherconf flags from most recently
1113          * received RA message (RFC 2462) -- yoshfuji
1114          */
1115         in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1116                                 IF_RA_OTHERCONF)) |
1117                                 (ra_msg->icmph.icmp6_addrconf_managed ?
1118                                         IF_RA_MANAGED : 0) |
1119                                 (ra_msg->icmph.icmp6_addrconf_other ?
1120                                         IF_RA_OTHERCONF : 0);
1121
1122         if (!in6_dev->cnf.accept_ra_defrtr)
1123                 goto skip_defrtr;
1124
1125         lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1126
1127 #ifdef CONFIG_IPV6_ROUTER_PREF
1128         pref = ra_msg->icmph.icmp6_router_pref;
1129         /* 10b is handled as if it were 00b (medium) */
1130         if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1131             in6_dev->cnf.accept_ra_rtr_pref)
1132                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1133 #endif
1134
1135         rt = rt6_get_dflt_router(&skb->nh.ipv6h->saddr, skb->dev);
1136
1137         if (rt)
1138                 neigh = rt->rt6i_nexthop;
1139
1140         if (rt && lifetime == 0) {
1141                 neigh_clone(neigh);
1142                 ip6_del_rt(rt);
1143                 rt = NULL;
1144         }
1145
1146         if (rt == NULL && lifetime) {
1147                 ND_PRINTK3(KERN_DEBUG
1148                            "ICMPv6 RA: adding default router.\n");
1149
1150                 rt = rt6_add_dflt_router(&skb->nh.ipv6h->saddr, skb->dev, pref);
1151                 if (rt == NULL) {
1152                         ND_PRINTK0(KERN_ERR
1153                                    "ICMPv6 RA: %s() failed to add default route.\n",
1154                                    __FUNCTION__);
1155                         in6_dev_put(in6_dev);
1156                         return;
1157                 }
1158
1159                 neigh = rt->rt6i_nexthop;
1160                 if (neigh == NULL) {
1161                         ND_PRINTK0(KERN_ERR
1162                                    "ICMPv6 RA: %s() got default router without neighbour.\n",
1163                                    __FUNCTION__);
1164                         dst_release(&rt->u.dst);
1165                         in6_dev_put(in6_dev);
1166                         return;
1167                 }
1168                 neigh->flags |= NTF_ROUTER;
1169         } else if (rt) {
1170                 rt->rt6i_flags |= (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1171         }
1172
1173         if (rt)
1174                 rt->rt6i_expires = jiffies + (HZ * lifetime);
1175
1176         if (ra_msg->icmph.icmp6_hop_limit) {
1177                 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1178                 if (rt)
1179                         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1180         }
1181
1182 skip_defrtr:
1183
1184         /*
1185          *      Update Reachable Time and Retrans Timer
1186          */
1187
1188         if (in6_dev->nd_parms) {
1189                 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1190
1191                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1192                         rtime = (rtime*HZ)/1000;
1193                         if (rtime < HZ/10)
1194                                 rtime = HZ/10;
1195                         in6_dev->nd_parms->retrans_time = rtime;
1196                         in6_dev->tstamp = jiffies;
1197                         inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1198                 }
1199
1200                 rtime = ntohl(ra_msg->reachable_time);
1201                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1202                         rtime = (rtime*HZ)/1000;
1203
1204                         if (rtime < HZ/10)
1205                                 rtime = HZ/10;
1206
1207                         if (rtime != in6_dev->nd_parms->base_reachable_time) {
1208                                 in6_dev->nd_parms->base_reachable_time = rtime;
1209                                 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1210                                 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1211                                 in6_dev->tstamp = jiffies;
1212                                 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1213                         }
1214                 }
1215         }
1216
1217         /*
1218          *      Process options.
1219          */
1220
1221         if (!neigh)
1222                 neigh = __neigh_lookup(&nd_tbl, &skb->nh.ipv6h->saddr,
1223                                        skb->dev, 1);
1224         if (neigh) {
1225                 u8 *lladdr = NULL;
1226                 if (ndopts.nd_opts_src_lladdr) {
1227                         lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1228                                                      skb->dev);
1229                         if (!lladdr) {
1230                                 ND_PRINTK2(KERN_WARNING
1231                                            "ICMPv6 RA: invalid link-layer address length\n");
1232                                 goto out;
1233                         }
1234                 }
1235                 neigh_update(neigh, lladdr, NUD_STALE,
1236                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1237                              NEIGH_UPDATE_F_OVERRIDE|
1238                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1239                              NEIGH_UPDATE_F_ISROUTER);
1240         }
1241
1242 #ifdef CONFIG_IPV6_ROUTE_INFO
1243         if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1244                 struct nd_opt_hdr *p;
1245                 for (p = ndopts.nd_opts_ri;
1246                      p;
1247                      p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1248                         if (((struct route_info *)p)->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1249                                 continue;
1250                         rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1251                                       &skb->nh.ipv6h->saddr);
1252                 }
1253         }
1254 #endif
1255
1256         if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1257                 struct nd_opt_hdr *p;
1258                 for (p = ndopts.nd_opts_pi;
1259                      p;
1260                      p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1261                         addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1262                 }
1263         }
1264
1265         if (ndopts.nd_opts_mtu) {
1266                 u32 mtu;
1267
1268                 memcpy(&mtu, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1269                 mtu = ntohl(mtu);
1270
1271                 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1272                         ND_PRINTK2(KERN_WARNING
1273                                    "ICMPv6 RA: invalid mtu: %d\n",
1274                                    mtu);
1275                 } else if (in6_dev->cnf.mtu6 != mtu) {
1276                         in6_dev->cnf.mtu6 = mtu;
1277
1278                         if (rt)
1279                                 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
1280
1281                         rt6_mtu_change(skb->dev, mtu);
1282                 }
1283         }
1284                         
1285         if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1286                 ND_PRINTK2(KERN_WARNING
1287                            "ICMPv6 RA: invalid RA options");
1288         }
1289 out:
1290         if (rt)
1291                 dst_release(&rt->u.dst);
1292         else if (neigh)
1293                 neigh_release(neigh);
1294         in6_dev_put(in6_dev);
1295 }
1296
1297 static void ndisc_redirect_rcv(struct sk_buff *skb)
1298 {
1299         struct inet6_dev *in6_dev;
1300         struct icmp6hdr *icmph;
1301         struct in6_addr *dest;
1302         struct in6_addr *target;        /* new first hop to destination */
1303         struct neighbour *neigh;
1304         int on_link = 0;
1305         struct ndisc_options ndopts;
1306         int optlen;
1307         u8 *lladdr = NULL;
1308
1309         if (!(ipv6_addr_type(&skb->nh.ipv6h->saddr) & IPV6_ADDR_LINKLOCAL)) {
1310                 ND_PRINTK2(KERN_WARNING
1311                            "ICMPv6 Redirect: source address is not link-local.\n");
1312                 return;
1313         }
1314
1315         optlen = skb->tail - skb->h.raw;
1316         optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1317
1318         if (optlen < 0) {
1319                 ND_PRINTK2(KERN_WARNING
1320                            "ICMPv6 Redirect: packet too short\n");
1321                 return;
1322         }
1323
1324         icmph = (struct icmp6hdr *) skb->h.raw;
1325         target = (struct in6_addr *) (icmph + 1);
1326         dest = target + 1;
1327
1328         if (ipv6_addr_is_multicast(dest)) {
1329                 ND_PRINTK2(KERN_WARNING
1330                            "ICMPv6 Redirect: destination address is multicast.\n");
1331                 return;
1332         }
1333
1334         if (ipv6_addr_equal(dest, target)) {
1335                 on_link = 1;
1336         } else if (!(ipv6_addr_type(target) & IPV6_ADDR_LINKLOCAL)) {
1337                 ND_PRINTK2(KERN_WARNING 
1338                            "ICMPv6 Redirect: target address is not link-local.\n");
1339                 return;
1340         }
1341
1342         in6_dev = in6_dev_get(skb->dev);
1343         if (!in6_dev)
1344                 return;
1345         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1346                 in6_dev_put(in6_dev);
1347                 return;
1348         }
1349
1350         /* RFC2461 8.1: 
1351          *      The IP source address of the Redirect MUST be the same as the current
1352          *      first-hop router for the specified ICMP Destination Address.
1353          */
1354                 
1355         if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1356                 ND_PRINTK2(KERN_WARNING
1357                            "ICMPv6 Redirect: invalid ND options\n");
1358                 in6_dev_put(in6_dev);
1359                 return;
1360         }
1361         if (ndopts.nd_opts_tgt_lladdr) {
1362                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1363                                              skb->dev);
1364                 if (!lladdr) {
1365                         ND_PRINTK2(KERN_WARNING
1366                                    "ICMPv6 Redirect: invalid link-layer address length\n");
1367                         in6_dev_put(in6_dev);
1368                         return;
1369                 }
1370         }
1371
1372         neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1373         if (neigh) {
1374                 rt6_redirect(dest, &skb->nh.ipv6h->daddr,
1375                              &skb->nh.ipv6h->saddr, neigh, lladdr,
1376                              on_link);
1377                 neigh_release(neigh);
1378         }
1379         in6_dev_put(in6_dev);
1380 }
1381
1382 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1383                          struct in6_addr *target)
1384 {
1385         struct sock *sk = ndisc_socket->sk;
1386         int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1387         struct sk_buff *buff;
1388         struct icmp6hdr *icmph;
1389         struct in6_addr saddr_buf;
1390         struct in6_addr *addrp;
1391         struct net_device *dev;
1392         struct rt6_info *rt;
1393         struct dst_entry *dst;
1394         struct inet6_dev *idev;
1395         struct flowi fl;
1396         u8 *opt;
1397         int rd_len;
1398         int err;
1399         int hlen;
1400         u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1401
1402         dev = skb->dev;
1403
1404         if (ipv6_get_lladdr(dev, &saddr_buf)) {
1405                 ND_PRINTK2(KERN_WARNING
1406                            "ICMPv6 Redirect: no link-local address on %s\n",
1407                            dev->name);
1408                 return;
1409         }
1410
1411         ndisc_flow_init(&fl, NDISC_REDIRECT, &saddr_buf, &skb->nh.ipv6h->saddr,
1412                         dev->ifindex);
1413
1414         dst = ip6_route_output(NULL, &fl);
1415         if (dst == NULL)
1416                 return;
1417
1418         err = xfrm_lookup(&dst, &fl, NULL, 0);
1419         if (err)
1420                 return;
1421
1422         rt = (struct rt6_info *) dst;
1423
1424         if (rt->rt6i_flags & RTF_GATEWAY) {
1425                 ND_PRINTK2(KERN_WARNING
1426                            "ICMPv6 Redirect: destination is not a neighbour.\n");
1427                 dst_release(dst);
1428                 return;
1429         }
1430         if (!xrlim_allow(dst, 1*HZ)) {
1431                 dst_release(dst);
1432                 return;
1433         }
1434
1435         if (dev->addr_len) {
1436                 read_lock_bh(&neigh->lock);
1437                 if (neigh->nud_state & NUD_VALID) {
1438                         memcpy(ha_buf, neigh->ha, dev->addr_len);
1439                         read_unlock_bh(&neigh->lock);
1440                         ha = ha_buf;
1441                         len += ndisc_opt_addr_space(dev);
1442                 } else
1443                         read_unlock_bh(&neigh->lock);
1444         }
1445
1446         rd_len = min_t(unsigned int,
1447                      IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1448         rd_len &= ~0x7;
1449         len += rd_len;
1450
1451         buff = sock_alloc_send_skb(sk, MAX_HEADER + len + LL_RESERVED_SPACE(dev),
1452                                    1, &err);
1453         if (buff == NULL) {
1454                 ND_PRINTK0(KERN_ERR
1455                            "ICMPv6 Redirect: %s() failed to allocate an skb.\n",
1456                            __FUNCTION__);
1457                 dst_release(dst);
1458                 return;
1459         }
1460
1461         hlen = 0;
1462
1463         skb_reserve(buff, LL_RESERVED_SPACE(dev));
1464         ip6_nd_hdr(sk, buff, dev, &saddr_buf, &skb->nh.ipv6h->saddr,
1465                    IPPROTO_ICMPV6, len);
1466
1467         icmph = (struct icmp6hdr *)skb_put(buff, len);
1468         buff->h.raw = (unsigned char*)icmph;
1469
1470         memset(icmph, 0, sizeof(struct icmp6hdr));
1471         icmph->icmp6_type = NDISC_REDIRECT;
1472
1473         /*
1474          *      copy target and destination addresses
1475          */
1476
1477         addrp = (struct in6_addr *)(icmph + 1);
1478         ipv6_addr_copy(addrp, target);
1479         addrp++;
1480         ipv6_addr_copy(addrp, &skb->nh.ipv6h->daddr);
1481
1482         opt = (u8*) (addrp + 1);
1483
1484         /*
1485          *      include target_address option
1486          */
1487
1488         if (ha)
1489                 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1490                                              dev->addr_len, dev->type);
1491
1492         /*
1493          *      build redirect option and copy skb over to the new packet.
1494          */
1495
1496         memset(opt, 0, 8);      
1497         *(opt++) = ND_OPT_REDIRECT_HDR;
1498         *(opt++) = (rd_len >> 3);
1499         opt += 6;
1500
1501         memcpy(opt, skb->nh.ipv6h, rd_len - 8);
1502
1503         icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &skb->nh.ipv6h->saddr,
1504                                              len, IPPROTO_ICMPV6,
1505                                              csum_partial((u8 *) icmph, len, 0));
1506
1507         buff->dst = dst;
1508         idev = in6_dev_get(dst->dev);
1509         IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
1510         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, buff, NULL, dst->dev, dst_output);
1511         if (!err) {
1512                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTREDIRECTS);
1513                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
1514         }
1515
1516         if (likely(idev != NULL))
1517                 in6_dev_put(idev);
1518 }
1519
1520 static void pndisc_redo(struct sk_buff *skb)
1521 {
1522         ndisc_recv_ns(skb);
1523         kfree_skb(skb);
1524 }
1525
1526 int ndisc_rcv(struct sk_buff *skb)
1527 {
1528         struct nd_msg *msg;
1529
1530         if (!pskb_may_pull(skb, skb->len))
1531                 return 0;
1532
1533         msg = (struct nd_msg *) skb->h.raw;
1534
1535         __skb_push(skb, skb->data-skb->h.raw);
1536
1537         if (skb->nh.ipv6h->hop_limit != 255) {
1538                 ND_PRINTK2(KERN_WARNING
1539                            "ICMPv6 NDISC: invalid hop-limit: %d\n",
1540                            skb->nh.ipv6h->hop_limit);
1541                 return 0;
1542         }
1543
1544         if (msg->icmph.icmp6_code != 0) {
1545                 ND_PRINTK2(KERN_WARNING 
1546                            "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1547                            msg->icmph.icmp6_code);
1548                 return 0;
1549         }
1550
1551         memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1552
1553         switch (msg->icmph.icmp6_type) {
1554         case NDISC_NEIGHBOUR_SOLICITATION:
1555                 ndisc_recv_ns(skb);
1556                 break;
1557
1558         case NDISC_NEIGHBOUR_ADVERTISEMENT:
1559                 ndisc_recv_na(skb);
1560                 break;
1561
1562         case NDISC_ROUTER_SOLICITATION:
1563                 ndisc_recv_rs(skb);
1564                 break;
1565
1566         case NDISC_ROUTER_ADVERTISEMENT:
1567                 ndisc_router_discovery(skb);
1568                 break;
1569
1570         case NDISC_REDIRECT:
1571                 ndisc_redirect_rcv(skb);
1572                 break;
1573         };
1574
1575         return 0;
1576 }
1577
1578 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1579 {
1580         struct net_device *dev = ptr;
1581
1582         switch (event) {
1583         case NETDEV_CHANGEADDR:
1584                 neigh_changeaddr(&nd_tbl, dev);
1585                 fib6_run_gc(~0UL);
1586                 break;
1587         case NETDEV_DOWN:
1588                 neigh_ifdown(&nd_tbl, dev);
1589                 fib6_run_gc(~0UL);
1590                 break;
1591         default:
1592                 break;
1593         }
1594
1595         return NOTIFY_DONE;
1596 }
1597
1598 static struct notifier_block ndisc_netdev_notifier = {
1599         .notifier_call = ndisc_netdev_event,
1600 };
1601
1602 #ifdef CONFIG_SYSCTL
1603 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1604                                          const char *func, const char *dev_name)
1605 {
1606         static char warncomm[TASK_COMM_LEN];
1607         static int warned;
1608         if (strcmp(warncomm, current->comm) && warned < 5) {
1609                 strcpy(warncomm, current->comm);
1610                 printk(KERN_WARNING
1611                         "process `%s' is using deprecated sysctl (%s) "
1612                         "net.ipv6.neigh.%s.%s; "
1613                         "Use net.ipv6.neigh.%s.%s_ms "
1614                         "instead.\n",
1615                         warncomm, func,
1616                         dev_name, ctl->procname,
1617                         dev_name, ctl->procname);
1618                 warned++;
1619         }
1620 }
1621
1622 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, struct file * filp, void __user *buffer, size_t *lenp, loff_t *ppos)
1623 {
1624         struct net_device *dev = ctl->extra1;
1625         struct inet6_dev *idev;
1626         int ret;
1627
1628         if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1629             ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1630                 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1631
1632         switch (ctl->ctl_name) {
1633         case NET_NEIGH_RETRANS_TIME:
1634                 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1635                 break;
1636         case NET_NEIGH_REACHABLE_TIME:
1637                 ret = proc_dointvec_jiffies(ctl, write,
1638                                             filp, buffer, lenp, ppos);
1639                 break;
1640         case NET_NEIGH_RETRANS_TIME_MS:
1641         case NET_NEIGH_REACHABLE_TIME_MS:
1642                 ret = proc_dointvec_ms_jiffies(ctl, write,
1643                                                filp, buffer, lenp, ppos);
1644                 break;
1645         default:
1646                 ret = -1;
1647         }
1648
1649         if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1650                 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1651                     ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1652                         idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1653                 idev->tstamp = jiffies;
1654                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1655                 in6_dev_put(idev);
1656         }
1657         return ret;
1658 }
1659
1660 static int ndisc_ifinfo_sysctl_strategy(ctl_table *ctl, int __user *name,
1661                                         int nlen, void __user *oldval,
1662                                         size_t __user *oldlenp,
1663                                         void __user *newval, size_t newlen,
1664                                         void **context)
1665 {
1666         struct net_device *dev = ctl->extra1;
1667         struct inet6_dev *idev;
1668         int ret;
1669
1670         if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1671             ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1672                 ndisc_warn_deprecated_sysctl(ctl, "procfs", dev ? dev->name : "default");
1673
1674         switch (ctl->ctl_name) {
1675         case NET_NEIGH_REACHABLE_TIME:
1676                 ret = sysctl_jiffies(ctl, name, nlen,
1677                                      oldval, oldlenp, newval, newlen,
1678                                      context);
1679                 break;
1680         case NET_NEIGH_RETRANS_TIME_MS:
1681         case NET_NEIGH_REACHABLE_TIME_MS:
1682                  ret = sysctl_ms_jiffies(ctl, name, nlen,
1683                                          oldval, oldlenp, newval, newlen,
1684                                          context);
1685                  break;
1686         default:
1687                 ret = 0;
1688         }
1689
1690         if (newval && newlen && ret > 0 &&
1691             dev && (idev = in6_dev_get(dev)) != NULL) {
1692                 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1693                     ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1694                         idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1695                 idev->tstamp = jiffies;
1696                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1697                 in6_dev_put(idev);
1698         }
1699
1700         return ret;
1701 }
1702
1703 #endif
1704
1705 int __init ndisc_init(struct net_proto_family *ops)
1706 {
1707         struct ipv6_pinfo *np;
1708         struct sock *sk;
1709         int err;
1710
1711         err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, &ndisc_socket);
1712         if (err < 0) {
1713                 ND_PRINTK0(KERN_ERR
1714                            "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n", 
1715                            err);
1716                 ndisc_socket = NULL; /* For safety. */
1717                 return err;
1718         }
1719
1720         sk = ndisc_socket->sk;
1721         np = inet6_sk(sk);
1722         sk->sk_allocation = GFP_ATOMIC;
1723         np->hop_limit = 255;
1724         /* Do not loopback ndisc messages */
1725         np->mc_loop = 0;
1726         sk->sk_prot->unhash(sk);
1727
1728         /*
1729          * Initialize the neighbour table
1730          */
1731         
1732         neigh_table_init(&nd_tbl);
1733
1734 #ifdef CONFIG_SYSCTL
1735         neigh_sysctl_register(NULL, &nd_tbl.parms, NET_IPV6, NET_IPV6_NEIGH, 
1736                               "ipv6",
1737                               &ndisc_ifinfo_sysctl_change,
1738                               &ndisc_ifinfo_sysctl_strategy);
1739 #endif
1740
1741         register_netdevice_notifier(&ndisc_netdev_notifier);
1742         return 0;
1743 }
1744
1745 void ndisc_cleanup(void)
1746 {
1747 #ifdef CONFIG_SYSCTL
1748         neigh_sysctl_unregister(&nd_tbl.parms);
1749 #endif
1750         neigh_table_clear(&nd_tbl);
1751         sock_release(ndisc_socket);
1752         ndisc_socket = NULL; /* For safety. */
1753 }