net: spread __net_init, __net_exit
[linux-3.10.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  *      Pierre Ynard                    :       export userland ND options
19  *                                              through netlink (RDNSS support)
20  *      Lars Fenneberg                  :       fixed MTU setting on receipt
21  *                                              of an RA.
22  *      Janos Farkas                    :       kmalloc failure checks
23  *      Alexey Kuznetsov                :       state machine reworked
24  *                                              and moved to net/core.
25  *      Pekka Savola                    :       RFC2461 validation
26  *      YOSHIFUJI Hideaki @USAGI        :       Verify ND options properly
27  */
28
29 /* Set to 3 to get tracing... */
30 #define ND_DEBUG 1
31
32 #define ND_PRINTK(fmt, args...) do { if (net_ratelimit()) { printk(fmt, ## args); } } while(0)
33 #define ND_NOPRINTK(x...) do { ; } while(0)
34 #define ND_PRINTK0 ND_PRINTK
35 #define ND_PRINTK1 ND_NOPRINTK
36 #define ND_PRINTK2 ND_NOPRINTK
37 #define ND_PRINTK3 ND_NOPRINTK
38 #if ND_DEBUG >= 1
39 #undef ND_PRINTK1
40 #define ND_PRINTK1 ND_PRINTK
41 #endif
42 #if ND_DEBUG >= 2
43 #undef ND_PRINTK2
44 #define ND_PRINTK2 ND_PRINTK
45 #endif
46 #if ND_DEBUG >= 3
47 #undef ND_PRINTK3
48 #define ND_PRINTK3 ND_PRINTK
49 #endif
50
51 #include <linux/module.h>
52 #include <linux/errno.h>
53 #include <linux/types.h>
54 #include <linux/socket.h>
55 #include <linux/sockios.h>
56 #include <linux/sched.h>
57 #include <linux/net.h>
58 #include <linux/in6.h>
59 #include <linux/route.h>
60 #include <linux/init.h>
61 #include <linux/rcupdate.h>
62 #ifdef CONFIG_SYSCTL
63 #include <linux/sysctl.h>
64 #endif
65
66 #include <linux/if_addr.h>
67 #include <linux/if_arp.h>
68 #include <linux/ipv6.h>
69 #include <linux/icmpv6.h>
70 #include <linux/jhash.h>
71
72 #include <net/sock.h>
73 #include <net/snmp.h>
74
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/icmp.h>
81
82 #include <net/netlink.h>
83 #include <linux/rtnetlink.h>
84
85 #include <net/flow.h>
86 #include <net/ip6_checksum.h>
87 #include <net/inet_common.h>
88 #include <linux/proc_fs.h>
89
90 #include <linux/netfilter.h>
91 #include <linux/netfilter_ipv6.h>
92
93 static u32 ndisc_hash(const void *pkey, const struct net_device *dev);
94 static int ndisc_constructor(struct neighbour *neigh);
95 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
96 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
97 static int pndisc_constructor(struct pneigh_entry *n);
98 static void pndisc_destructor(struct pneigh_entry *n);
99 static void pndisc_redo(struct sk_buff *skb);
100
101 static const struct neigh_ops ndisc_generic_ops = {
102         .family =               AF_INET6,
103         .solicit =              ndisc_solicit,
104         .error_report =         ndisc_error_report,
105         .output =               neigh_resolve_output,
106         .connected_output =     neigh_connected_output,
107         .hh_output =            dev_queue_xmit,
108         .queue_xmit =           dev_queue_xmit,
109 };
110
111 static const struct neigh_ops ndisc_hh_ops = {
112         .family =               AF_INET6,
113         .solicit =              ndisc_solicit,
114         .error_report =         ndisc_error_report,
115         .output =               neigh_resolve_output,
116         .connected_output =     neigh_resolve_output,
117         .hh_output =            dev_queue_xmit,
118         .queue_xmit =           dev_queue_xmit,
119 };
120
121
122 static const struct neigh_ops ndisc_direct_ops = {
123         .family =               AF_INET6,
124         .output =               dev_queue_xmit,
125         .connected_output =     dev_queue_xmit,
126         .hh_output =            dev_queue_xmit,
127         .queue_xmit =           dev_queue_xmit,
128 };
129
130 struct neigh_table nd_tbl = {
131         .family =       AF_INET6,
132         .entry_size =   sizeof(struct neighbour) + sizeof(struct in6_addr),
133         .key_len =      sizeof(struct in6_addr),
134         .hash =         ndisc_hash,
135         .constructor =  ndisc_constructor,
136         .pconstructor = pndisc_constructor,
137         .pdestructor =  pndisc_destructor,
138         .proxy_redo =   pndisc_redo,
139         .id =           "ndisc_cache",
140         .parms = {
141                 .tbl =                  &nd_tbl,
142                 .base_reachable_time =  30 * HZ,
143                 .retrans_time =  1 * HZ,
144                 .gc_staletime = 60 * HZ,
145                 .reachable_time =               30 * HZ,
146                 .delay_probe_time =      5 * HZ,
147                 .queue_len =             3,
148                 .ucast_probes =  3,
149                 .mcast_probes =  3,
150                 .anycast_delay =         1 * HZ,
151                 .proxy_delay =          (8 * HZ) / 10,
152                 .proxy_qlen =           64,
153         },
154         .gc_interval =    30 * HZ,
155         .gc_thresh1 =    128,
156         .gc_thresh2 =    512,
157         .gc_thresh3 =   1024,
158 };
159
160 /* ND options */
161 struct ndisc_options {
162         struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
163 #ifdef CONFIG_IPV6_ROUTE_INFO
164         struct nd_opt_hdr *nd_opts_ri;
165         struct nd_opt_hdr *nd_opts_ri_end;
166 #endif
167         struct nd_opt_hdr *nd_useropts;
168         struct nd_opt_hdr *nd_useropts_end;
169 };
170
171 #define nd_opts_src_lladdr      nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
172 #define nd_opts_tgt_lladdr      nd_opt_array[ND_OPT_TARGET_LL_ADDR]
173 #define nd_opts_pi              nd_opt_array[ND_OPT_PREFIX_INFO]
174 #define nd_opts_pi_end          nd_opt_array[__ND_OPT_PREFIX_INFO_END]
175 #define nd_opts_rh              nd_opt_array[ND_OPT_REDIRECT_HDR]
176 #define nd_opts_mtu             nd_opt_array[ND_OPT_MTU]
177
178 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
179
180 /*
181  * Return the padding between the option length and the start of the
182  * link addr.  Currently only IP-over-InfiniBand needs this, although
183  * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
184  * also need a pad of 2.
185  */
186 static int ndisc_addr_option_pad(unsigned short type)
187 {
188         switch (type) {
189         case ARPHRD_INFINIBAND: return 2;
190         default:                return 0;
191         }
192 }
193
194 static inline int ndisc_opt_addr_space(struct net_device *dev)
195 {
196         return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
197 }
198
199 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
200                                   unsigned short addr_type)
201 {
202         int space = NDISC_OPT_SPACE(data_len);
203         int pad   = ndisc_addr_option_pad(addr_type);
204
205         opt[0] = type;
206         opt[1] = space>>3;
207
208         memset(opt + 2, 0, pad);
209         opt   += pad;
210         space -= pad;
211
212         memcpy(opt+2, data, data_len);
213         data_len += 2;
214         opt += data_len;
215         if ((space -= data_len) > 0)
216                 memset(opt, 0, space);
217         return opt + space;
218 }
219
220 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
221                                             struct nd_opt_hdr *end)
222 {
223         int type;
224         if (!cur || !end || cur >= end)
225                 return NULL;
226         type = cur->nd_opt_type;
227         do {
228                 cur = ((void *)cur) + (cur->nd_opt_len << 3);
229         } while(cur < end && cur->nd_opt_type != type);
230         return (cur <= end && cur->nd_opt_type == type ? cur : NULL);
231 }
232
233 static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
234 {
235         return (opt->nd_opt_type == ND_OPT_RDNSS);
236 }
237
238 static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
239                                              struct nd_opt_hdr *end)
240 {
241         if (!cur || !end || cur >= end)
242                 return NULL;
243         do {
244                 cur = ((void *)cur) + (cur->nd_opt_len << 3);
245         } while(cur < end && !ndisc_is_useropt(cur));
246         return (cur <= end && ndisc_is_useropt(cur) ? cur : NULL);
247 }
248
249 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
250                                                  struct ndisc_options *ndopts)
251 {
252         struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
253
254         if (!nd_opt || opt_len < 0 || !ndopts)
255                 return NULL;
256         memset(ndopts, 0, sizeof(*ndopts));
257         while (opt_len) {
258                 int l;
259                 if (opt_len < sizeof(struct nd_opt_hdr))
260                         return NULL;
261                 l = nd_opt->nd_opt_len << 3;
262                 if (opt_len < l || l == 0)
263                         return NULL;
264                 switch (nd_opt->nd_opt_type) {
265                 case ND_OPT_SOURCE_LL_ADDR:
266                 case ND_OPT_TARGET_LL_ADDR:
267                 case ND_OPT_MTU:
268                 case ND_OPT_REDIRECT_HDR:
269                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
270                                 ND_PRINTK2(KERN_WARNING
271                                            "%s(): duplicated ND6 option found: type=%d\n",
272                                            __func__,
273                                            nd_opt->nd_opt_type);
274                         } else {
275                                 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
276                         }
277                         break;
278                 case ND_OPT_PREFIX_INFO:
279                         ndopts->nd_opts_pi_end = nd_opt;
280                         if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
281                                 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
282                         break;
283 #ifdef CONFIG_IPV6_ROUTE_INFO
284                 case ND_OPT_ROUTE_INFO:
285                         ndopts->nd_opts_ri_end = nd_opt;
286                         if (!ndopts->nd_opts_ri)
287                                 ndopts->nd_opts_ri = nd_opt;
288                         break;
289 #endif
290                 default:
291                         if (ndisc_is_useropt(nd_opt)) {
292                                 ndopts->nd_useropts_end = nd_opt;
293                                 if (!ndopts->nd_useropts)
294                                         ndopts->nd_useropts = nd_opt;
295                         } else {
296                                 /*
297                                  * Unknown options must be silently ignored,
298                                  * to accommodate future extension to the
299                                  * protocol.
300                                  */
301                                 ND_PRINTK2(KERN_NOTICE
302                                            "%s(): ignored unsupported option; type=%d, len=%d\n",
303                                            __func__,
304                                            nd_opt->nd_opt_type, nd_opt->nd_opt_len);
305                         }
306                 }
307                 opt_len -= l;
308                 nd_opt = ((void *)nd_opt) + l;
309         }
310         return ndopts;
311 }
312
313 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
314                                       struct net_device *dev)
315 {
316         u8 *lladdr = (u8 *)(p + 1);
317         int lladdrlen = p->nd_opt_len << 3;
318         int prepad = ndisc_addr_option_pad(dev->type);
319         if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
320                 return NULL;
321         return (lladdr + prepad);
322 }
323
324 int ndisc_mc_map(struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
325 {
326         switch (dev->type) {
327         case ARPHRD_ETHER:
328         case ARPHRD_IEEE802:    /* Not sure. Check it later. --ANK */
329         case ARPHRD_FDDI:
330                 ipv6_eth_mc_map(addr, buf);
331                 return 0;
332         case ARPHRD_IEEE802_TR:
333                 ipv6_tr_mc_map(addr,buf);
334                 return 0;
335         case ARPHRD_ARCNET:
336                 ipv6_arcnet_mc_map(addr, buf);
337                 return 0;
338         case ARPHRD_INFINIBAND:
339                 ipv6_ib_mc_map(addr, dev->broadcast, buf);
340                 return 0;
341         default:
342                 if (dir) {
343                         memcpy(buf, dev->broadcast, dev->addr_len);
344                         return 0;
345                 }
346         }
347         return -EINVAL;
348 }
349
350 EXPORT_SYMBOL(ndisc_mc_map);
351
352 static u32 ndisc_hash(const void *pkey, const struct net_device *dev)
353 {
354         const u32 *p32 = pkey;
355         u32 addr_hash, i;
356
357         addr_hash = 0;
358         for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
359                 addr_hash ^= *p32++;
360
361         return jhash_2words(addr_hash, dev->ifindex, nd_tbl.hash_rnd);
362 }
363
364 static int ndisc_constructor(struct neighbour *neigh)
365 {
366         struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
367         struct net_device *dev = neigh->dev;
368         struct inet6_dev *in6_dev;
369         struct neigh_parms *parms;
370         int is_multicast = ipv6_addr_is_multicast(addr);
371
372         rcu_read_lock();
373         in6_dev = in6_dev_get(dev);
374         if (in6_dev == NULL) {
375                 rcu_read_unlock();
376                 return -EINVAL;
377         }
378
379         parms = in6_dev->nd_parms;
380         __neigh_parms_put(neigh->parms);
381         neigh->parms = neigh_parms_clone(parms);
382         rcu_read_unlock();
383
384         neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
385         if (!dev->header_ops) {
386                 neigh->nud_state = NUD_NOARP;
387                 neigh->ops = &ndisc_direct_ops;
388                 neigh->output = neigh->ops->queue_xmit;
389         } else {
390                 if (is_multicast) {
391                         neigh->nud_state = NUD_NOARP;
392                         ndisc_mc_map(addr, neigh->ha, dev, 1);
393                 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
394                         neigh->nud_state = NUD_NOARP;
395                         memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
396                         if (dev->flags&IFF_LOOPBACK)
397                                 neigh->type = RTN_LOCAL;
398                 } else if (dev->flags&IFF_POINTOPOINT) {
399                         neigh->nud_state = NUD_NOARP;
400                         memcpy(neigh->ha, dev->broadcast, dev->addr_len);
401                 }
402                 if (dev->header_ops->cache)
403                         neigh->ops = &ndisc_hh_ops;
404                 else
405                         neigh->ops = &ndisc_generic_ops;
406                 if (neigh->nud_state&NUD_VALID)
407                         neigh->output = neigh->ops->connected_output;
408                 else
409                         neigh->output = neigh->ops->output;
410         }
411         in6_dev_put(in6_dev);
412         return 0;
413 }
414
415 static int pndisc_constructor(struct pneigh_entry *n)
416 {
417         struct in6_addr *addr = (struct in6_addr*)&n->key;
418         struct in6_addr maddr;
419         struct net_device *dev = n->dev;
420
421         if (dev == NULL || __in6_dev_get(dev) == NULL)
422                 return -EINVAL;
423         addrconf_addr_solict_mult(addr, &maddr);
424         ipv6_dev_mc_inc(dev, &maddr);
425         return 0;
426 }
427
428 static void pndisc_destructor(struct pneigh_entry *n)
429 {
430         struct in6_addr *addr = (struct in6_addr*)&n->key;
431         struct in6_addr maddr;
432         struct net_device *dev = n->dev;
433
434         if (dev == NULL || __in6_dev_get(dev) == NULL)
435                 return;
436         addrconf_addr_solict_mult(addr, &maddr);
437         ipv6_dev_mc_dec(dev, &maddr);
438 }
439
440 struct sk_buff *ndisc_build_skb(struct net_device *dev,
441                                 const struct in6_addr *daddr,
442                                 const struct in6_addr *saddr,
443                                 struct icmp6hdr *icmp6h,
444                                 const struct in6_addr *target,
445                                 int llinfo)
446 {
447         struct net *net = dev_net(dev);
448         struct sock *sk = net->ipv6.ndisc_sk;
449         struct sk_buff *skb;
450         struct icmp6hdr *hdr;
451         int len;
452         int err;
453         u8 *opt;
454
455         if (!dev->addr_len)
456                 llinfo = 0;
457
458         len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
459         if (llinfo)
460                 len += ndisc_opt_addr_space(dev);
461
462         skb = sock_alloc_send_skb(sk,
463                                   (MAX_HEADER + sizeof(struct ipv6hdr) +
464                                    len + LL_ALLOCATED_SPACE(dev)),
465                                   1, &err);
466         if (!skb) {
467                 ND_PRINTK0(KERN_ERR
468                            "ICMPv6 ND: %s() failed to allocate an skb, err=%d.\n",
469                            __func__, err);
470                 return NULL;
471         }
472
473         skb_reserve(skb, LL_RESERVED_SPACE(dev));
474         ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
475
476         skb->transport_header = skb->tail;
477         skb_put(skb, len);
478
479         hdr = (struct icmp6hdr *)skb_transport_header(skb);
480         memcpy(hdr, icmp6h, sizeof(*hdr));
481
482         opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
483         if (target) {
484                 ipv6_addr_copy((struct in6_addr *)opt, target);
485                 opt += sizeof(*target);
486         }
487
488         if (llinfo)
489                 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
490                                        dev->addr_len, dev->type);
491
492         hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
493                                            IPPROTO_ICMPV6,
494                                            csum_partial(hdr,
495                                                         len, 0));
496
497         return skb;
498 }
499
500 EXPORT_SYMBOL(ndisc_build_skb);
501
502 void ndisc_send_skb(struct sk_buff *skb,
503                     struct net_device *dev,
504                     struct neighbour *neigh,
505                     const struct in6_addr *daddr,
506                     const struct in6_addr *saddr,
507                     struct icmp6hdr *icmp6h)
508 {
509         struct flowi fl;
510         struct dst_entry *dst;
511         struct net *net = dev_net(dev);
512         struct sock *sk = net->ipv6.ndisc_sk;
513         struct inet6_dev *idev;
514         int err;
515         u8 type;
516
517         type = icmp6h->icmp6_type;
518
519         icmpv6_flow_init(sk, &fl, type, saddr, daddr, dev->ifindex);
520
521         dst = icmp6_dst_alloc(dev, neigh, daddr);
522         if (!dst) {
523                 kfree_skb(skb);
524                 return;
525         }
526
527         err = xfrm_lookup(net, &dst, &fl, NULL, 0);
528         if (err < 0) {
529                 kfree_skb(skb);
530                 return;
531         }
532
533         skb_dst_set(skb, dst);
534
535         idev = in6_dev_get(dst->dev);
536         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
537
538         err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
539                       dst_output);
540         if (!err) {
541                 ICMP6MSGOUT_INC_STATS(net, idev, type);
542                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
543         }
544
545         if (likely(idev != NULL))
546                 in6_dev_put(idev);
547 }
548
549 EXPORT_SYMBOL(ndisc_send_skb);
550
551 /*
552  *      Send a Neighbour Discover packet
553  */
554 static void __ndisc_send(struct net_device *dev,
555                          struct neighbour *neigh,
556                          const struct in6_addr *daddr,
557                          const struct in6_addr *saddr,
558                          struct icmp6hdr *icmp6h, const struct in6_addr *target,
559                          int llinfo)
560 {
561         struct sk_buff *skb;
562
563         skb = ndisc_build_skb(dev, daddr, saddr, icmp6h, target, llinfo);
564         if (!skb)
565                 return;
566
567         ndisc_send_skb(skb, dev, neigh, daddr, saddr, icmp6h);
568 }
569
570 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
571                           const struct in6_addr *daddr,
572                           const struct in6_addr *solicited_addr,
573                           int router, int solicited, int override, int inc_opt)
574 {
575         struct in6_addr tmpaddr;
576         struct inet6_ifaddr *ifp;
577         const struct in6_addr *src_addr;
578         struct icmp6hdr icmp6h = {
579                 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
580         };
581
582         /* for anycast or proxy, solicited_addr != src_addr */
583         ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
584         if (ifp) {
585                 src_addr = solicited_addr;
586                 if (ifp->flags & IFA_F_OPTIMISTIC)
587                         override = 0;
588                 in6_ifa_put(ifp);
589         } else {
590                 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
591                                        inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
592                                        &tmpaddr))
593                         return;
594                 src_addr = &tmpaddr;
595         }
596
597         icmp6h.icmp6_router = router;
598         icmp6h.icmp6_solicited = solicited;
599         icmp6h.icmp6_override = override;
600
601         inc_opt |= ifp->idev->cnf.force_tllao;
602         __ndisc_send(dev, neigh, daddr, src_addr,
603                      &icmp6h, solicited_addr,
604                      inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
605 }
606
607 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
608                    const struct in6_addr *solicit,
609                    const struct in6_addr *daddr, const struct in6_addr *saddr)
610 {
611         struct in6_addr addr_buf;
612         struct icmp6hdr icmp6h = {
613                 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
614         };
615
616         if (saddr == NULL) {
617                 if (ipv6_get_lladdr(dev, &addr_buf,
618                                    (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
619                         return;
620                 saddr = &addr_buf;
621         }
622
623         __ndisc_send(dev, neigh, daddr, saddr,
624                      &icmp6h, solicit,
625                      !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
626 }
627
628 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
629                    const struct in6_addr *daddr)
630 {
631         struct icmp6hdr icmp6h = {
632                 .icmp6_type = NDISC_ROUTER_SOLICITATION,
633         };
634         int send_sllao = dev->addr_len;
635
636 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
637         /*
638          * According to section 2.2 of RFC 4429, we must not
639          * send router solicitations with a sllao from
640          * optimistic addresses, but we may send the solicitation
641          * if we don't include the sllao.  So here we check
642          * if our address is optimistic, and if so, we
643          * suppress the inclusion of the sllao.
644          */
645         if (send_sllao) {
646                 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
647                                                            dev, 1);
648                 if (ifp) {
649                         if (ifp->flags & IFA_F_OPTIMISTIC)  {
650                                 send_sllao = 0;
651                         }
652                         in6_ifa_put(ifp);
653                 } else {
654                         send_sllao = 0;
655                 }
656         }
657 #endif
658         __ndisc_send(dev, NULL, daddr, saddr,
659                      &icmp6h, NULL,
660                      send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
661 }
662
663
664 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
665 {
666         /*
667          *      "The sender MUST return an ICMP
668          *       destination unreachable"
669          */
670         dst_link_failure(skb);
671         kfree_skb(skb);
672 }
673
674 /* Called with locked neigh: either read or both */
675
676 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
677 {
678         struct in6_addr *saddr = NULL;
679         struct in6_addr mcaddr;
680         struct net_device *dev = neigh->dev;
681         struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
682         int probes = atomic_read(&neigh->probes);
683
684         if (skb && ipv6_chk_addr(dev_net(dev), &ipv6_hdr(skb)->saddr, dev, 1))
685                 saddr = &ipv6_hdr(skb)->saddr;
686
687         if ((probes -= neigh->parms->ucast_probes) < 0) {
688                 if (!(neigh->nud_state & NUD_VALID)) {
689                         ND_PRINTK1(KERN_DEBUG "%s(): trying to ucast probe in NUD_INVALID: %pI6\n",
690                                    __func__, target);
691                 }
692                 ndisc_send_ns(dev, neigh, target, target, saddr);
693         } else if ((probes -= neigh->parms->app_probes) < 0) {
694 #ifdef CONFIG_ARPD
695                 neigh_app_ns(neigh);
696 #endif
697         } else {
698                 addrconf_addr_solict_mult(target, &mcaddr);
699                 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
700         }
701 }
702
703 static int pndisc_is_router(const void *pkey,
704                             struct net_device *dev)
705 {
706         struct pneigh_entry *n;
707         int ret = -1;
708
709         read_lock_bh(&nd_tbl.lock);
710         n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
711         if (n)
712                 ret = !!(n->flags & NTF_ROUTER);
713         read_unlock_bh(&nd_tbl.lock);
714
715         return ret;
716 }
717
718 static void ndisc_recv_ns(struct sk_buff *skb)
719 {
720         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
721         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
722         struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
723         u8 *lladdr = NULL;
724         u32 ndoptlen = skb->tail - (skb->transport_header +
725                                     offsetof(struct nd_msg, opt));
726         struct ndisc_options ndopts;
727         struct net_device *dev = skb->dev;
728         struct inet6_ifaddr *ifp;
729         struct inet6_dev *idev = NULL;
730         struct neighbour *neigh;
731         int dad = ipv6_addr_any(saddr);
732         int inc;
733         int is_router = -1;
734
735         if (ipv6_addr_is_multicast(&msg->target)) {
736                 ND_PRINTK2(KERN_WARNING
737                            "ICMPv6 NS: multicast target address");
738                 return;
739         }
740
741         /*
742          * RFC2461 7.1.1:
743          * DAD has to be destined for solicited node multicast address.
744          */
745         if (dad &&
746             !(daddr->s6_addr32[0] == htonl(0xff020000) &&
747               daddr->s6_addr32[1] == htonl(0x00000000) &&
748               daddr->s6_addr32[2] == htonl(0x00000001) &&
749               daddr->s6_addr [12] == 0xff )) {
750                 ND_PRINTK2(KERN_WARNING
751                            "ICMPv6 NS: bad DAD packet (wrong destination)\n");
752                 return;
753         }
754
755         if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
756                 ND_PRINTK2(KERN_WARNING
757                            "ICMPv6 NS: invalid ND options\n");
758                 return;
759         }
760
761         if (ndopts.nd_opts_src_lladdr) {
762                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
763                 if (!lladdr) {
764                         ND_PRINTK2(KERN_WARNING
765                                    "ICMPv6 NS: invalid link-layer address length\n");
766                         return;
767                 }
768
769                 /* RFC2461 7.1.1:
770                  *      If the IP source address is the unspecified address,
771                  *      there MUST NOT be source link-layer address option
772                  *      in the message.
773                  */
774                 if (dad) {
775                         ND_PRINTK2(KERN_WARNING
776                                    "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
777                         return;
778                 }
779         }
780
781         inc = ipv6_addr_is_multicast(daddr);
782
783         ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
784         if (ifp) {
785
786                 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
787                         if (dad) {
788                                 if (dev->type == ARPHRD_IEEE802_TR) {
789                                         const unsigned char *sadr;
790                                         sadr = skb_mac_header(skb);
791                                         if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
792                                             sadr[9] == dev->dev_addr[1] &&
793                                             sadr[10] == dev->dev_addr[2] &&
794                                             sadr[11] == dev->dev_addr[3] &&
795                                             sadr[12] == dev->dev_addr[4] &&
796                                             sadr[13] == dev->dev_addr[5]) {
797                                                 /* looped-back to us */
798                                                 goto out;
799                                         }
800                                 }
801
802                                 /*
803                                  * We are colliding with another node
804                                  * who is doing DAD
805                                  * so fail our DAD process
806                                  */
807                                 addrconf_dad_failure(ifp);
808                                 return;
809                         } else {
810                                 /*
811                                  * This is not a dad solicitation.
812                                  * If we are an optimistic node,
813                                  * we should respond.
814                                  * Otherwise, we should ignore it.
815                                  */
816                                 if (!(ifp->flags & IFA_F_OPTIMISTIC))
817                                         goto out;
818                         }
819                 }
820
821                 idev = ifp->idev;
822         } else {
823                 struct net *net = dev_net(dev);
824
825                 idev = in6_dev_get(dev);
826                 if (!idev) {
827                         /* XXX: count this drop? */
828                         return;
829                 }
830
831                 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
832                     (idev->cnf.forwarding &&
833                      (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
834                      (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
835                         if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
836                             skb->pkt_type != PACKET_HOST &&
837                             inc != 0 &&
838                             idev->nd_parms->proxy_delay != 0) {
839                                 /*
840                                  * for anycast or proxy,
841                                  * sender should delay its response
842                                  * by a random time between 0 and
843                                  * MAX_ANYCAST_DELAY_TIME seconds.
844                                  * (RFC2461) -- yoshfuji
845                                  */
846                                 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
847                                 if (n)
848                                         pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
849                                 goto out;
850                         }
851                 } else
852                         goto out;
853         }
854
855         if (is_router < 0)
856                 is_router = !!idev->cnf.forwarding;
857
858         if (dad) {
859                 ndisc_send_na(dev, NULL, &in6addr_linklocal_allnodes, &msg->target,
860                               is_router, 0, (ifp != NULL), 1);
861                 goto out;
862         }
863
864         if (inc)
865                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
866         else
867                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
868
869         /*
870          *      update / create cache entry
871          *      for the source address
872          */
873         neigh = __neigh_lookup(&nd_tbl, saddr, dev,
874                                !inc || lladdr || !dev->addr_len);
875         if (neigh)
876                 neigh_update(neigh, lladdr, NUD_STALE,
877                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
878                              NEIGH_UPDATE_F_OVERRIDE);
879         if (neigh || !dev->header_ops) {
880                 ndisc_send_na(dev, neigh, saddr, &msg->target,
881                               is_router,
882                               1, (ifp != NULL && inc), inc);
883                 if (neigh)
884                         neigh_release(neigh);
885         }
886
887 out:
888         if (ifp)
889                 in6_ifa_put(ifp);
890         else
891                 in6_dev_put(idev);
892
893         return;
894 }
895
896 static void ndisc_recv_na(struct sk_buff *skb)
897 {
898         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
899         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
900         struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
901         u8 *lladdr = NULL;
902         u32 ndoptlen = skb->tail - (skb->transport_header +
903                                     offsetof(struct nd_msg, opt));
904         struct ndisc_options ndopts;
905         struct net_device *dev = skb->dev;
906         struct inet6_ifaddr *ifp;
907         struct neighbour *neigh;
908
909         if (skb->len < sizeof(struct nd_msg)) {
910                 ND_PRINTK2(KERN_WARNING
911                            "ICMPv6 NA: packet too short\n");
912                 return;
913         }
914
915         if (ipv6_addr_is_multicast(&msg->target)) {
916                 ND_PRINTK2(KERN_WARNING
917                            "ICMPv6 NA: target address is multicast.\n");
918                 return;
919         }
920
921         if (ipv6_addr_is_multicast(daddr) &&
922             msg->icmph.icmp6_solicited) {
923                 ND_PRINTK2(KERN_WARNING
924                            "ICMPv6 NA: solicited NA is multicasted.\n");
925                 return;
926         }
927
928         if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
929                 ND_PRINTK2(KERN_WARNING
930                            "ICMPv6 NS: invalid ND option\n");
931                 return;
932         }
933         if (ndopts.nd_opts_tgt_lladdr) {
934                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
935                 if (!lladdr) {
936                         ND_PRINTK2(KERN_WARNING
937                                    "ICMPv6 NA: invalid link-layer address length\n");
938                         return;
939                 }
940         }
941         ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
942         if (ifp) {
943                 if (ifp->flags & IFA_F_TENTATIVE) {
944                         addrconf_dad_failure(ifp);
945                         return;
946                 }
947                 /* What should we make now? The advertisement
948                    is invalid, but ndisc specs say nothing
949                    about it. It could be misconfiguration, or
950                    an smart proxy agent tries to help us :-)
951
952                    We should not print the error if NA has been
953                    received from loopback - it is just our own
954                    unsolicited advertisement.
955                  */
956                 if (skb->pkt_type != PACKET_LOOPBACK)
957                         ND_PRINTK1(KERN_WARNING
958                            "ICMPv6 NA: someone advertises our address %pI6 on %s!\n",
959                            &ifp->addr, ifp->idev->dev->name);
960                 in6_ifa_put(ifp);
961                 return;
962         }
963         neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
964
965         if (neigh) {
966                 u8 old_flags = neigh->flags;
967                 struct net *net = dev_net(dev);
968
969                 if (neigh->nud_state & NUD_FAILED)
970                         goto out;
971
972                 /*
973                  * Don't update the neighbor cache entry on a proxy NA from
974                  * ourselves because either the proxied node is off link or it
975                  * has already sent a NA to us.
976                  */
977                 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
978                     net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
979                     pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
980                         /* XXX: idev->cnf.prixy_ndp */
981                         goto out;
982                 }
983
984                 neigh_update(neigh, lladdr,
985                              msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
986                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
987                              (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
988                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
989                              (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
990
991                 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
992                         /*
993                          * Change: router to host
994                          */
995                         struct rt6_info *rt;
996                         rt = rt6_get_dflt_router(saddr, dev);
997                         if (rt)
998                                 ip6_del_rt(rt);
999                 }
1000
1001 out:
1002                 neigh_release(neigh);
1003         }
1004 }
1005
1006 static void ndisc_recv_rs(struct sk_buff *skb)
1007 {
1008         struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1009         unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1010         struct neighbour *neigh;
1011         struct inet6_dev *idev;
1012         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1013         struct ndisc_options ndopts;
1014         u8 *lladdr = NULL;
1015
1016         if (skb->len < sizeof(*rs_msg))
1017                 return;
1018
1019         idev = in6_dev_get(skb->dev);
1020         if (!idev) {
1021                 if (net_ratelimit())
1022                         ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
1023                 return;
1024         }
1025
1026         /* Don't accept RS if we're not in router mode */
1027         if (!idev->cnf.forwarding)
1028                 goto out;
1029
1030         /*
1031          * Don't update NCE if src = ::;
1032          * this implies that the source node has no ip address assigned yet.
1033          */
1034         if (ipv6_addr_any(saddr))
1035                 goto out;
1036
1037         /* Parse ND options */
1038         if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1039                 if (net_ratelimit())
1040                         ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1041                 goto out;
1042         }
1043
1044         if (ndopts.nd_opts_src_lladdr) {
1045                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1046                                              skb->dev);
1047                 if (!lladdr)
1048                         goto out;
1049         }
1050
1051         neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1052         if (neigh) {
1053                 neigh_update(neigh, lladdr, NUD_STALE,
1054                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1055                              NEIGH_UPDATE_F_OVERRIDE|
1056                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1057                 neigh_release(neigh);
1058         }
1059 out:
1060         in6_dev_put(idev);
1061 }
1062
1063 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1064 {
1065         struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1066         struct sk_buff *skb;
1067         struct nlmsghdr *nlh;
1068         struct nduseroptmsg *ndmsg;
1069         struct net *net = dev_net(ra->dev);
1070         int err;
1071         int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1072                                     + (opt->nd_opt_len << 3));
1073         size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1074
1075         skb = nlmsg_new(msg_size, GFP_ATOMIC);
1076         if (skb == NULL) {
1077                 err = -ENOBUFS;
1078                 goto errout;
1079         }
1080
1081         nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1082         if (nlh == NULL) {
1083                 goto nla_put_failure;
1084         }
1085
1086         ndmsg = nlmsg_data(nlh);
1087         ndmsg->nduseropt_family = AF_INET6;
1088         ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1089         ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1090         ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1091         ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1092
1093         memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1094
1095         NLA_PUT(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
1096                 &ipv6_hdr(ra)->saddr);
1097         nlmsg_end(skb, nlh);
1098
1099         rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1100         return;
1101
1102 nla_put_failure:
1103         nlmsg_free(skb);
1104         err = -EMSGSIZE;
1105 errout:
1106         rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1107 }
1108
1109 static void ndisc_router_discovery(struct sk_buff *skb)
1110 {
1111         struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1112         struct neighbour *neigh = NULL;
1113         struct inet6_dev *in6_dev;
1114         struct rt6_info *rt = NULL;
1115         int lifetime;
1116         struct ndisc_options ndopts;
1117         int optlen;
1118         unsigned int pref = 0;
1119
1120         __u8 * opt = (__u8 *)(ra_msg + 1);
1121
1122         optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1123
1124         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1125                 ND_PRINTK2(KERN_WARNING
1126                            "ICMPv6 RA: source address is not link-local.\n");
1127                 return;
1128         }
1129         if (optlen < 0) {
1130                 ND_PRINTK2(KERN_WARNING
1131                            "ICMPv6 RA: packet too short\n");
1132                 return;
1133         }
1134
1135 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1136         if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1137                 ND_PRINTK2(KERN_WARNING
1138                            "ICMPv6 RA: from host or unauthorized router\n");
1139                 return;
1140         }
1141 #endif
1142
1143         /*
1144          *      set the RA_RECV flag in the interface
1145          */
1146
1147         in6_dev = in6_dev_get(skb->dev);
1148         if (in6_dev == NULL) {
1149                 ND_PRINTK0(KERN_ERR
1150                            "ICMPv6 RA: can't find inet6 device for %s.\n",
1151                            skb->dev->name);
1152                 return;
1153         }
1154
1155         if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1156                 in6_dev_put(in6_dev);
1157                 ND_PRINTK2(KERN_WARNING
1158                            "ICMP6 RA: invalid ND options\n");
1159                 return;
1160         }
1161
1162         /* skip route and link configuration on routers */
1163         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra)
1164                 goto skip_linkparms;
1165
1166 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1167         /* skip link-specific parameters from interior routers */
1168         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1169                 goto skip_linkparms;
1170 #endif
1171
1172         if (in6_dev->if_flags & IF_RS_SENT) {
1173                 /*
1174                  *      flag that an RA was received after an RS was sent
1175                  *      out on this interface.
1176                  */
1177                 in6_dev->if_flags |= IF_RA_RCVD;
1178         }
1179
1180         /*
1181          * Remember the managed/otherconf flags from most recently
1182          * received RA message (RFC 2462) -- yoshfuji
1183          */
1184         in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1185                                 IF_RA_OTHERCONF)) |
1186                                 (ra_msg->icmph.icmp6_addrconf_managed ?
1187                                         IF_RA_MANAGED : 0) |
1188                                 (ra_msg->icmph.icmp6_addrconf_other ?
1189                                         IF_RA_OTHERCONF : 0);
1190
1191         if (!in6_dev->cnf.accept_ra_defrtr)
1192                 goto skip_defrtr;
1193
1194         lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1195
1196 #ifdef CONFIG_IPV6_ROUTER_PREF
1197         pref = ra_msg->icmph.icmp6_router_pref;
1198         /* 10b is handled as if it were 00b (medium) */
1199         if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1200             !in6_dev->cnf.accept_ra_rtr_pref)
1201                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1202 #endif
1203
1204         rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1205
1206         if (rt)
1207                 neigh = rt->rt6i_nexthop;
1208
1209         if (rt && lifetime == 0) {
1210                 neigh_clone(neigh);
1211                 ip6_del_rt(rt);
1212                 rt = NULL;
1213         }
1214
1215         if (rt == NULL && lifetime) {
1216                 ND_PRINTK3(KERN_DEBUG
1217                            "ICMPv6 RA: adding default router.\n");
1218
1219                 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1220                 if (rt == NULL) {
1221                         ND_PRINTK0(KERN_ERR
1222                                    "ICMPv6 RA: %s() failed to add default route.\n",
1223                                    __func__);
1224                         in6_dev_put(in6_dev);
1225                         return;
1226                 }
1227
1228                 neigh = rt->rt6i_nexthop;
1229                 if (neigh == NULL) {
1230                         ND_PRINTK0(KERN_ERR
1231                                    "ICMPv6 RA: %s() got default router without neighbour.\n",
1232                                    __func__);
1233                         dst_release(&rt->u.dst);
1234                         in6_dev_put(in6_dev);
1235                         return;
1236                 }
1237                 neigh->flags |= NTF_ROUTER;
1238         } else if (rt) {
1239                 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1240         }
1241
1242         if (rt)
1243                 rt->rt6i_expires = jiffies + (HZ * lifetime);
1244
1245         if (ra_msg->icmph.icmp6_hop_limit) {
1246                 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1247                 if (rt)
1248                         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1249         }
1250
1251 skip_defrtr:
1252
1253         /*
1254          *      Update Reachable Time and Retrans Timer
1255          */
1256
1257         if (in6_dev->nd_parms) {
1258                 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1259
1260                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1261                         rtime = (rtime*HZ)/1000;
1262                         if (rtime < HZ/10)
1263                                 rtime = HZ/10;
1264                         in6_dev->nd_parms->retrans_time = rtime;
1265                         in6_dev->tstamp = jiffies;
1266                         inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1267                 }
1268
1269                 rtime = ntohl(ra_msg->reachable_time);
1270                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1271                         rtime = (rtime*HZ)/1000;
1272
1273                         if (rtime < HZ/10)
1274                                 rtime = HZ/10;
1275
1276                         if (rtime != in6_dev->nd_parms->base_reachable_time) {
1277                                 in6_dev->nd_parms->base_reachable_time = rtime;
1278                                 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1279                                 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1280                                 in6_dev->tstamp = jiffies;
1281                                 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1282                         }
1283                 }
1284         }
1285
1286 skip_linkparms:
1287
1288         /*
1289          *      Process options.
1290          */
1291
1292         if (!neigh)
1293                 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1294                                        skb->dev, 1);
1295         if (neigh) {
1296                 u8 *lladdr = NULL;
1297                 if (ndopts.nd_opts_src_lladdr) {
1298                         lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1299                                                      skb->dev);
1300                         if (!lladdr) {
1301                                 ND_PRINTK2(KERN_WARNING
1302                                            "ICMPv6 RA: invalid link-layer address length\n");
1303                                 goto out;
1304                         }
1305                 }
1306                 neigh_update(neigh, lladdr, NUD_STALE,
1307                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1308                              NEIGH_UPDATE_F_OVERRIDE|
1309                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1310                              NEIGH_UPDATE_F_ISROUTER);
1311         }
1312
1313         /* skip route and link configuration on routers */
1314         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra)
1315                 goto out;
1316
1317 #ifdef CONFIG_IPV6_ROUTE_INFO
1318         if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1319                 struct nd_opt_hdr *p;
1320                 for (p = ndopts.nd_opts_ri;
1321                      p;
1322                      p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1323                         struct route_info *ri = (struct route_info *)p;
1324 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1325                         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1326                             ri->prefix_len == 0)
1327                                 continue;
1328 #endif
1329                         if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1330                                 continue;
1331                         rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1332                                       &ipv6_hdr(skb)->saddr);
1333                 }
1334         }
1335 #endif
1336
1337 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1338         /* skip link-specific ndopts from interior routers */
1339         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT)
1340                 goto out;
1341 #endif
1342
1343         if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1344                 struct nd_opt_hdr *p;
1345                 for (p = ndopts.nd_opts_pi;
1346                      p;
1347                      p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1348                         addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1349                 }
1350         }
1351
1352         if (ndopts.nd_opts_mtu) {
1353                 __be32 n;
1354                 u32 mtu;
1355
1356                 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1357                 mtu = ntohl(n);
1358
1359                 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1360                         ND_PRINTK2(KERN_WARNING
1361                                    "ICMPv6 RA: invalid mtu: %d\n",
1362                                    mtu);
1363                 } else if (in6_dev->cnf.mtu6 != mtu) {
1364                         in6_dev->cnf.mtu6 = mtu;
1365
1366                         if (rt)
1367                                 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
1368
1369                         rt6_mtu_change(skb->dev, mtu);
1370                 }
1371         }
1372
1373         if (ndopts.nd_useropts) {
1374                 struct nd_opt_hdr *p;
1375                 for (p = ndopts.nd_useropts;
1376                      p;
1377                      p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1378                         ndisc_ra_useropt(skb, p);
1379                 }
1380         }
1381
1382         if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1383                 ND_PRINTK2(KERN_WARNING
1384                            "ICMPv6 RA: invalid RA options");
1385         }
1386 out:
1387         if (rt)
1388                 dst_release(&rt->u.dst);
1389         else if (neigh)
1390                 neigh_release(neigh);
1391         in6_dev_put(in6_dev);
1392 }
1393
1394 static void ndisc_redirect_rcv(struct sk_buff *skb)
1395 {
1396         struct inet6_dev *in6_dev;
1397         struct icmp6hdr *icmph;
1398         struct in6_addr *dest;
1399         struct in6_addr *target;        /* new first hop to destination */
1400         struct neighbour *neigh;
1401         int on_link = 0;
1402         struct ndisc_options ndopts;
1403         int optlen;
1404         u8 *lladdr = NULL;
1405
1406 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1407         switch (skb->ndisc_nodetype) {
1408         case NDISC_NODETYPE_HOST:
1409         case NDISC_NODETYPE_NODEFAULT:
1410                 ND_PRINTK2(KERN_WARNING
1411                            "ICMPv6 Redirect: from host or unauthorized router\n");
1412                 return;
1413         }
1414 #endif
1415
1416         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1417                 ND_PRINTK2(KERN_WARNING
1418                            "ICMPv6 Redirect: source address is not link-local.\n");
1419                 return;
1420         }
1421
1422         optlen = skb->tail - skb->transport_header;
1423         optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1424
1425         if (optlen < 0) {
1426                 ND_PRINTK2(KERN_WARNING
1427                            "ICMPv6 Redirect: packet too short\n");
1428                 return;
1429         }
1430
1431         icmph = icmp6_hdr(skb);
1432         target = (struct in6_addr *) (icmph + 1);
1433         dest = target + 1;
1434
1435         if (ipv6_addr_is_multicast(dest)) {
1436                 ND_PRINTK2(KERN_WARNING
1437                            "ICMPv6 Redirect: destination address is multicast.\n");
1438                 return;
1439         }
1440
1441         if (ipv6_addr_equal(dest, target)) {
1442                 on_link = 1;
1443         } else if (ipv6_addr_type(target) !=
1444                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1445                 ND_PRINTK2(KERN_WARNING
1446                            "ICMPv6 Redirect: target address is not link-local unicast.\n");
1447                 return;
1448         }
1449
1450         in6_dev = in6_dev_get(skb->dev);
1451         if (!in6_dev)
1452                 return;
1453         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1454                 in6_dev_put(in6_dev);
1455                 return;
1456         }
1457
1458         /* RFC2461 8.1:
1459          *      The IP source address of the Redirect MUST be the same as the current
1460          *      first-hop router for the specified ICMP Destination Address.
1461          */
1462
1463         if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1464                 ND_PRINTK2(KERN_WARNING
1465                            "ICMPv6 Redirect: invalid ND options\n");
1466                 in6_dev_put(in6_dev);
1467                 return;
1468         }
1469         if (ndopts.nd_opts_tgt_lladdr) {
1470                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1471                                              skb->dev);
1472                 if (!lladdr) {
1473                         ND_PRINTK2(KERN_WARNING
1474                                    "ICMPv6 Redirect: invalid link-layer address length\n");
1475                         in6_dev_put(in6_dev);
1476                         return;
1477                 }
1478         }
1479
1480         neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1481         if (neigh) {
1482                 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1483                              &ipv6_hdr(skb)->saddr, neigh, lladdr,
1484                              on_link);
1485                 neigh_release(neigh);
1486         }
1487         in6_dev_put(in6_dev);
1488 }
1489
1490 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1491                          const struct in6_addr *target)
1492 {
1493         struct net_device *dev = skb->dev;
1494         struct net *net = dev_net(dev);
1495         struct sock *sk = net->ipv6.ndisc_sk;
1496         int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1497         struct sk_buff *buff;
1498         struct icmp6hdr *icmph;
1499         struct in6_addr saddr_buf;
1500         struct in6_addr *addrp;
1501         struct rt6_info *rt;
1502         struct dst_entry *dst;
1503         struct inet6_dev *idev;
1504         struct flowi fl;
1505         u8 *opt;
1506         int rd_len;
1507         int err;
1508         u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1509
1510         if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1511                 ND_PRINTK2(KERN_WARNING
1512                            "ICMPv6 Redirect: no link-local address on %s\n",
1513                            dev->name);
1514                 return;
1515         }
1516
1517         if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1518             ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1519                 ND_PRINTK2(KERN_WARNING
1520                         "ICMPv6 Redirect: target address is not link-local unicast.\n");
1521                 return;
1522         }
1523
1524         icmpv6_flow_init(sk, &fl, NDISC_REDIRECT,
1525                          &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1526
1527         dst = ip6_route_output(net, NULL, &fl);
1528         if (dst == NULL)
1529                 return;
1530
1531         err = xfrm_lookup(net, &dst, &fl, NULL, 0);
1532         if (err)
1533                 return;
1534
1535         rt = (struct rt6_info *) dst;
1536
1537         if (rt->rt6i_flags & RTF_GATEWAY) {
1538                 ND_PRINTK2(KERN_WARNING
1539                            "ICMPv6 Redirect: destination is not a neighbour.\n");
1540                 goto release;
1541         }
1542         if (!xrlim_allow(dst, 1*HZ))
1543                 goto release;
1544
1545         if (dev->addr_len) {
1546                 read_lock_bh(&neigh->lock);
1547                 if (neigh->nud_state & NUD_VALID) {
1548                         memcpy(ha_buf, neigh->ha, dev->addr_len);
1549                         read_unlock_bh(&neigh->lock);
1550                         ha = ha_buf;
1551                         len += ndisc_opt_addr_space(dev);
1552                 } else
1553                         read_unlock_bh(&neigh->lock);
1554         }
1555
1556         rd_len = min_t(unsigned int,
1557                      IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1558         rd_len &= ~0x7;
1559         len += rd_len;
1560
1561         buff = sock_alloc_send_skb(sk,
1562                                    (MAX_HEADER + sizeof(struct ipv6hdr) +
1563                                     len + LL_ALLOCATED_SPACE(dev)),
1564                                    1, &err);
1565         if (buff == NULL) {
1566                 ND_PRINTK0(KERN_ERR
1567                            "ICMPv6 Redirect: %s() failed to allocate an skb, err=%d.\n",
1568                            __func__, err);
1569                 goto release;
1570         }
1571
1572         skb_reserve(buff, LL_RESERVED_SPACE(dev));
1573         ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1574                    IPPROTO_ICMPV6, len);
1575
1576         skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1577         skb_put(buff, len);
1578         icmph = icmp6_hdr(buff);
1579
1580         memset(icmph, 0, sizeof(struct icmp6hdr));
1581         icmph->icmp6_type = NDISC_REDIRECT;
1582
1583         /*
1584          *      copy target and destination addresses
1585          */
1586
1587         addrp = (struct in6_addr *)(icmph + 1);
1588         ipv6_addr_copy(addrp, target);
1589         addrp++;
1590         ipv6_addr_copy(addrp, &ipv6_hdr(skb)->daddr);
1591
1592         opt = (u8*) (addrp + 1);
1593
1594         /*
1595          *      include target_address option
1596          */
1597
1598         if (ha)
1599                 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1600                                              dev->addr_len, dev->type);
1601
1602         /*
1603          *      build redirect option and copy skb over to the new packet.
1604          */
1605
1606         memset(opt, 0, 8);
1607         *(opt++) = ND_OPT_REDIRECT_HDR;
1608         *(opt++) = (rd_len >> 3);
1609         opt += 6;
1610
1611         memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1612
1613         icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1614                                              len, IPPROTO_ICMPV6,
1615                                              csum_partial(icmph, len, 0));
1616
1617         skb_dst_set(buff, dst);
1618         idev = in6_dev_get(dst->dev);
1619         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1620         err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, buff, NULL, dst->dev,
1621                       dst_output);
1622         if (!err) {
1623                 ICMP6MSGOUT_INC_STATS(net, idev, NDISC_REDIRECT);
1624                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1625         }
1626
1627         if (likely(idev != NULL))
1628                 in6_dev_put(idev);
1629         return;
1630
1631 release:
1632         dst_release(dst);
1633 }
1634
1635 static void pndisc_redo(struct sk_buff *skb)
1636 {
1637         ndisc_recv_ns(skb);
1638         kfree_skb(skb);
1639 }
1640
1641 int ndisc_rcv(struct sk_buff *skb)
1642 {
1643         struct nd_msg *msg;
1644
1645         if (!pskb_may_pull(skb, skb->len))
1646                 return 0;
1647
1648         msg = (struct nd_msg *)skb_transport_header(skb);
1649
1650         __skb_push(skb, skb->data - skb_transport_header(skb));
1651
1652         if (ipv6_hdr(skb)->hop_limit != 255) {
1653                 ND_PRINTK2(KERN_WARNING
1654                            "ICMPv6 NDISC: invalid hop-limit: %d\n",
1655                            ipv6_hdr(skb)->hop_limit);
1656                 return 0;
1657         }
1658
1659         if (msg->icmph.icmp6_code != 0) {
1660                 ND_PRINTK2(KERN_WARNING
1661                            "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1662                            msg->icmph.icmp6_code);
1663                 return 0;
1664         }
1665
1666         memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1667
1668         switch (msg->icmph.icmp6_type) {
1669         case NDISC_NEIGHBOUR_SOLICITATION:
1670                 ndisc_recv_ns(skb);
1671                 break;
1672
1673         case NDISC_NEIGHBOUR_ADVERTISEMENT:
1674                 ndisc_recv_na(skb);
1675                 break;
1676
1677         case NDISC_ROUTER_SOLICITATION:
1678                 ndisc_recv_rs(skb);
1679                 break;
1680
1681         case NDISC_ROUTER_ADVERTISEMENT:
1682                 ndisc_router_discovery(skb);
1683                 break;
1684
1685         case NDISC_REDIRECT:
1686                 ndisc_redirect_rcv(skb);
1687                 break;
1688         }
1689
1690         return 0;
1691 }
1692
1693 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1694 {
1695         struct net_device *dev = ptr;
1696         struct net *net = dev_net(dev);
1697
1698         switch (event) {
1699         case NETDEV_CHANGEADDR:
1700                 neigh_changeaddr(&nd_tbl, dev);
1701                 fib6_run_gc(~0UL, net);
1702                 break;
1703         case NETDEV_DOWN:
1704                 neigh_ifdown(&nd_tbl, dev);
1705                 fib6_run_gc(~0UL, net);
1706                 break;
1707         default:
1708                 break;
1709         }
1710
1711         return NOTIFY_DONE;
1712 }
1713
1714 static struct notifier_block ndisc_netdev_notifier = {
1715         .notifier_call = ndisc_netdev_event,
1716 };
1717
1718 #ifdef CONFIG_SYSCTL
1719 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1720                                          const char *func, const char *dev_name)
1721 {
1722         static char warncomm[TASK_COMM_LEN];
1723         static int warned;
1724         if (strcmp(warncomm, current->comm) && warned < 5) {
1725                 strcpy(warncomm, current->comm);
1726                 printk(KERN_WARNING
1727                         "process `%s' is using deprecated sysctl (%s) "
1728                         "net.ipv6.neigh.%s.%s; "
1729                         "Use net.ipv6.neigh.%s.%s_ms "
1730                         "instead.\n",
1731                         warncomm, func,
1732                         dev_name, ctl->procname,
1733                         dev_name, ctl->procname);
1734                 warned++;
1735         }
1736 }
1737
1738 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1739 {
1740         struct net_device *dev = ctl->extra1;
1741         struct inet6_dev *idev;
1742         int ret;
1743
1744         if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1745             (strcmp(ctl->procname, "base_reachable_time") == 0))
1746                 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1747
1748         if (strcmp(ctl->procname, "retrans_time") == 0)
1749                 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1750
1751         else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1752                 ret = proc_dointvec_jiffies(ctl, write,
1753                                             buffer, lenp, ppos);
1754
1755         else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1756                  (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1757                 ret = proc_dointvec_ms_jiffies(ctl, write,
1758                                                buffer, lenp, ppos);
1759         else
1760                 ret = -1;
1761
1762         if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1763                 if (ctl->data == &idev->nd_parms->base_reachable_time)
1764                         idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1765                 idev->tstamp = jiffies;
1766                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1767                 in6_dev_put(idev);
1768         }
1769         return ret;
1770 }
1771
1772
1773 #endif
1774
1775 static int __net_init ndisc_net_init(struct net *net)
1776 {
1777         struct ipv6_pinfo *np;
1778         struct sock *sk;
1779         int err;
1780
1781         err = inet_ctl_sock_create(&sk, PF_INET6,
1782                                    SOCK_RAW, IPPROTO_ICMPV6, net);
1783         if (err < 0) {
1784                 ND_PRINTK0(KERN_ERR
1785                            "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1786                            err);
1787                 return err;
1788         }
1789
1790         net->ipv6.ndisc_sk = sk;
1791
1792         np = inet6_sk(sk);
1793         np->hop_limit = 255;
1794         /* Do not loopback ndisc messages */
1795         np->mc_loop = 0;
1796
1797         return 0;
1798 }
1799
1800 static void __net_exit ndisc_net_exit(struct net *net)
1801 {
1802         inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1803 }
1804
1805 static struct pernet_operations ndisc_net_ops = {
1806         .init = ndisc_net_init,
1807         .exit = ndisc_net_exit,
1808 };
1809
1810 int __init ndisc_init(void)
1811 {
1812         int err;
1813
1814         err = register_pernet_subsys(&ndisc_net_ops);
1815         if (err)
1816                 return err;
1817         /*
1818          * Initialize the neighbour table
1819          */
1820         neigh_table_init(&nd_tbl);
1821
1822 #ifdef CONFIG_SYSCTL
1823         err = neigh_sysctl_register(NULL, &nd_tbl.parms, NET_IPV6,
1824                                     NET_IPV6_NEIGH, "ipv6",
1825                                     &ndisc_ifinfo_sysctl_change);
1826         if (err)
1827                 goto out_unregister_pernet;
1828 #endif
1829         err = register_netdevice_notifier(&ndisc_netdev_notifier);
1830         if (err)
1831                 goto out_unregister_sysctl;
1832 out:
1833         return err;
1834
1835 out_unregister_sysctl:
1836 #ifdef CONFIG_SYSCTL
1837         neigh_sysctl_unregister(&nd_tbl.parms);
1838 out_unregister_pernet:
1839 #endif
1840         unregister_pernet_subsys(&ndisc_net_ops);
1841         goto out;
1842 }
1843
1844 void ndisc_cleanup(void)
1845 {
1846         unregister_netdevice_notifier(&ndisc_netdev_notifier);
1847 #ifdef CONFIG_SYSCTL
1848         neigh_sysctl_unregister(&nd_tbl.parms);
1849 #endif
1850         neigh_table_clear(&nd_tbl);
1851         unregister_pernet_subsys(&ndisc_net_ops);
1852 }