Merge branch 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband
[linux-2.6] / 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 EXPORT_SYMBOL(ndisc_mc_map);
323
324 static u32 ndisc_hash(const void *pkey, const struct net_device *dev)
325 {
326         const u32 *p32 = pkey;
327         u32 addr_hash, i;
328
329         addr_hash = 0;
330         for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
331                 addr_hash ^= *p32++;
332
333         return jhash_2words(addr_hash, dev->ifindex, nd_tbl.hash_rnd);
334 }
335
336 static int ndisc_constructor(struct neighbour *neigh)
337 {
338         struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
339         struct net_device *dev = neigh->dev;
340         struct inet6_dev *in6_dev;
341         struct neigh_parms *parms;
342         int is_multicast = ipv6_addr_is_multicast(addr);
343
344         rcu_read_lock();
345         in6_dev = in6_dev_get(dev);
346         if (in6_dev == NULL) {
347                 rcu_read_unlock();
348                 return -EINVAL;
349         }
350
351         parms = in6_dev->nd_parms;
352         __neigh_parms_put(neigh->parms);
353         neigh->parms = neigh_parms_clone(parms);
354         rcu_read_unlock();
355
356         neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
357         if (dev->hard_header == NULL) {
358                 neigh->nud_state = NUD_NOARP;
359                 neigh->ops = &ndisc_direct_ops;
360                 neigh->output = neigh->ops->queue_xmit;
361         } else {
362                 if (is_multicast) {
363                         neigh->nud_state = NUD_NOARP;
364                         ndisc_mc_map(addr, neigh->ha, dev, 1);
365                 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
366                         neigh->nud_state = NUD_NOARP;
367                         memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
368                         if (dev->flags&IFF_LOOPBACK)
369                                 neigh->type = RTN_LOCAL;
370                 } else if (dev->flags&IFF_POINTOPOINT) {
371                         neigh->nud_state = NUD_NOARP;
372                         memcpy(neigh->ha, dev->broadcast, dev->addr_len);
373                 }
374                 if (dev->hard_header_cache)
375                         neigh->ops = &ndisc_hh_ops;
376                 else
377                         neigh->ops = &ndisc_generic_ops;
378                 if (neigh->nud_state&NUD_VALID)
379                         neigh->output = neigh->ops->connected_output;
380                 else
381                         neigh->output = neigh->ops->output;
382         }
383         in6_dev_put(in6_dev);
384         return 0;
385 }
386
387 static int pndisc_constructor(struct pneigh_entry *n)
388 {
389         struct in6_addr *addr = (struct in6_addr*)&n->key;
390         struct in6_addr maddr;
391         struct net_device *dev = n->dev;
392
393         if (dev == NULL || __in6_dev_get(dev) == NULL)
394                 return -EINVAL;
395         addrconf_addr_solict_mult(addr, &maddr);
396         ipv6_dev_mc_inc(dev, &maddr);
397         return 0;
398 }
399
400 static void pndisc_destructor(struct pneigh_entry *n)
401 {
402         struct in6_addr *addr = (struct in6_addr*)&n->key;
403         struct in6_addr maddr;
404         struct net_device *dev = n->dev;
405
406         if (dev == NULL || __in6_dev_get(dev) == NULL)
407                 return;
408         addrconf_addr_solict_mult(addr, &maddr);
409         ipv6_dev_mc_dec(dev, &maddr);
410 }
411
412 /*
413  *      Send a Neighbour Advertisement
414  */
415
416 static inline void ndisc_flow_init(struct flowi *fl, u8 type,
417                             struct in6_addr *saddr, struct in6_addr *daddr,
418                             int oif)
419 {
420         memset(fl, 0, sizeof(*fl));
421         ipv6_addr_copy(&fl->fl6_src, saddr);
422         ipv6_addr_copy(&fl->fl6_dst, daddr);
423         fl->proto               = IPPROTO_ICMPV6;
424         fl->fl_icmp_type        = type;
425         fl->fl_icmp_code        = 0;
426         fl->oif                 = oif;
427         security_sk_classify_flow(ndisc_socket->sk, fl);
428 }
429
430 static void __ndisc_send(struct net_device *dev,
431                          struct neighbour *neigh,
432                          struct in6_addr *daddr, struct in6_addr *saddr,
433                          struct icmp6hdr *icmp6h, struct in6_addr *target,
434                          int llinfo, int icmp6_mib_outnd)
435 {
436         struct flowi fl;
437         struct dst_entry *dst;
438         struct sock *sk = ndisc_socket->sk;
439         struct sk_buff *skb;
440         struct icmp6hdr *hdr;
441         struct inet6_dev *idev;
442         int len;
443         int err;
444         u8 *opt;
445
446         ndisc_flow_init(&fl, icmp6h->icmp6_type, saddr, daddr,
447                         dev->ifindex);
448
449         dst = ndisc_dst_alloc(dev, neigh, daddr, ip6_output);
450         if (!dst)
451                 return;
452
453         err = xfrm_lookup(&dst, &fl, NULL, 0);
454         if (err < 0)
455                 return;
456
457         if (!dev->addr_len)
458                 llinfo = 0;
459
460         len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
461         if (llinfo)
462                 len += ndisc_opt_addr_space(dev);
463
464         skb = sock_alloc_send_skb(sk,
465                                   (MAX_HEADER + sizeof(struct ipv6hdr) +
466                                    len + LL_RESERVED_SPACE(dev)),
467                                   1, &err);
468         if (!skb) {
469                 ND_PRINTK0(KERN_ERR
470                            "ICMPv6 ND: %s() failed to allocate an skb.\n",
471                            __FUNCTION__);
472                 dst_release(dst);
473                 return;
474         }
475
476         skb_reserve(skb, LL_RESERVED_SPACE(dev));
477         ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
478
479         skb->transport_header = skb->tail;
480         skb_put(skb, len);
481
482         hdr = (struct icmp6hdr *)skb_transport_header(skb);
483         memcpy(hdr, icmp6h, sizeof(*hdr));
484
485         opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
486         if (target) {
487                 ipv6_addr_copy((struct in6_addr *)opt, target);
488                 opt += sizeof(*target);
489         }
490
491         if (llinfo)
492                 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
493                                        dev->addr_len, dev->type);
494
495         hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
496                                            IPPROTO_ICMPV6,
497                                            csum_partial((__u8 *) hdr,
498                                                         len, 0));
499
500         skb->dst = dst;
501
502         idev = in6_dev_get(dst->dev);
503         IP6_INC_STATS(idev, IPSTATS_MIB_OUTREQUESTS);
504
505         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev, dst_output);
506         if (!err) {
507                 ICMP6_INC_STATS(idev, icmp6_mib_outnd);
508                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
509         }
510
511         if (likely(idev != NULL))
512                 in6_dev_put(idev);
513 }
514
515 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
516                    struct in6_addr *daddr, struct in6_addr *solicited_addr,
517                    int router, int solicited, int override, int inc_opt)
518 {
519         struct in6_addr tmpaddr;
520         struct inet6_ifaddr *ifp;
521         struct in6_addr *src_addr;
522         struct icmp6hdr icmp6h = {
523                 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
524         };
525
526         /* for anycast or proxy, solicited_addr != src_addr */
527         ifp = ipv6_get_ifaddr(solicited_addr, dev, 1);
528         if (ifp) {
529                 src_addr = solicited_addr;
530                 if (ifp->flags & IFA_F_OPTIMISTIC)
531                         override = 0;
532                 in6_ifa_put(ifp);
533         } else {
534                 if (ipv6_dev_get_saddr(dev, daddr, &tmpaddr))
535                         return;
536                 src_addr = &tmpaddr;
537         }
538
539         icmp6h.icmp6_router = router;
540         icmp6h.icmp6_solicited = solicited;
541         icmp6h.icmp6_override = override;
542
543         __ndisc_send(dev, neigh, daddr, src_addr,
544                      &icmp6h, solicited_addr,
545                      inc_opt ? ND_OPT_TARGET_LL_ADDR : 0,
546                      ICMP6_MIB_OUTNEIGHBORADVERTISEMENTS);
547 }
548
549 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
550                    struct in6_addr *solicit,
551                    struct in6_addr *daddr, struct in6_addr *saddr)
552 {
553         struct in6_addr addr_buf;
554         struct icmp6hdr icmp6h = {
555                 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
556         };
557
558         if (saddr == NULL) {
559                 if (ipv6_get_lladdr(dev, &addr_buf,
560                                    (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
561                         return;
562                 saddr = &addr_buf;
563         }
564
565         __ndisc_send(dev, neigh, daddr, saddr,
566                      &icmp6h, solicit,
567                      !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0,
568                      ICMP6_MIB_OUTNEIGHBORSOLICITS);
569 }
570
571 void ndisc_send_rs(struct net_device *dev, struct in6_addr *saddr,
572                    struct in6_addr *daddr)
573 {
574         struct icmp6hdr icmp6h = {
575                 .icmp6_type = NDISC_ROUTER_SOLICITATION,
576         };
577         int send_sllao = dev->addr_len;
578
579 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
580         /*
581          * According to section 2.2 of RFC 4429, we must not
582          * send router solicitations with a sllao from
583          * optimistic addresses, but we may send the solicitation
584          * if we don't include the sllao.  So here we check
585          * if our address is optimistic, and if so, we
586          * supress the inclusion of the sllao.
587          */
588         if (send_sllao) {
589                 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(saddr, dev, 1);
590                 if (ifp) {
591                         if (ifp->flags & IFA_F_OPTIMISTIC)  {
592                                 send_sllao = 0;
593                         }
594                         in6_ifa_put(ifp);
595                 } else {
596                         send_sllao = 0;
597                 }
598         }
599 #endif
600         __ndisc_send(dev, NULL, daddr, saddr,
601                      &icmp6h, NULL,
602                      send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0,
603                      ICMP6_MIB_OUTROUTERSOLICITS);
604 }
605
606
607 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
608 {
609         /*
610          *      "The sender MUST return an ICMP
611          *       destination unreachable"
612          */
613         dst_link_failure(skb);
614         kfree_skb(skb);
615 }
616
617 /* Called with locked neigh: either read or both */
618
619 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
620 {
621         struct in6_addr *saddr = NULL;
622         struct in6_addr mcaddr;
623         struct net_device *dev = neigh->dev;
624         struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
625         int probes = atomic_read(&neigh->probes);
626
627         if (skb && ipv6_chk_addr(&ipv6_hdr(skb)->saddr, dev, 1))
628                 saddr = &ipv6_hdr(skb)->saddr;
629
630         if ((probes -= neigh->parms->ucast_probes) < 0) {
631                 if (!(neigh->nud_state & NUD_VALID)) {
632                         ND_PRINTK1(KERN_DEBUG
633                                    "%s(): trying to ucast probe in NUD_INVALID: "
634                                    NIP6_FMT "\n",
635                                    __FUNCTION__,
636                                    NIP6(*target));
637                 }
638                 ndisc_send_ns(dev, neigh, target, target, saddr);
639         } else if ((probes -= neigh->parms->app_probes) < 0) {
640 #ifdef CONFIG_ARPD
641                 neigh_app_ns(neigh);
642 #endif
643         } else {
644                 addrconf_addr_solict_mult(target, &mcaddr);
645                 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
646         }
647 }
648
649 static void ndisc_recv_ns(struct sk_buff *skb)
650 {
651         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
652         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
653         struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
654         u8 *lladdr = NULL;
655         u32 ndoptlen = skb->tail - (skb->transport_header +
656                                     offsetof(struct nd_msg, opt));
657         struct ndisc_options ndopts;
658         struct net_device *dev = skb->dev;
659         struct inet6_ifaddr *ifp;
660         struct inet6_dev *idev = NULL;
661         struct neighbour *neigh;
662         struct pneigh_entry *pneigh = NULL;
663         int dad = ipv6_addr_any(saddr);
664         int inc;
665         int is_router;
666
667         if (ipv6_addr_is_multicast(&msg->target)) {
668                 ND_PRINTK2(KERN_WARNING
669                            "ICMPv6 NS: multicast target address");
670                 return;
671         }
672
673         /*
674          * RFC2461 7.1.1:
675          * DAD has to be destined for solicited node multicast address.
676          */
677         if (dad &&
678             !(daddr->s6_addr32[0] == htonl(0xff020000) &&
679               daddr->s6_addr32[1] == htonl(0x00000000) &&
680               daddr->s6_addr32[2] == htonl(0x00000001) &&
681               daddr->s6_addr [12] == 0xff )) {
682                 ND_PRINTK2(KERN_WARNING
683                            "ICMPv6 NS: bad DAD packet (wrong destination)\n");
684                 return;
685         }
686
687         if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
688                 ND_PRINTK2(KERN_WARNING
689                            "ICMPv6 NS: invalid ND options\n");
690                 return;
691         }
692
693         if (ndopts.nd_opts_src_lladdr) {
694                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
695                 if (!lladdr) {
696                         ND_PRINTK2(KERN_WARNING
697                                    "ICMPv6 NS: invalid link-layer address length\n");
698                         return;
699                 }
700
701                 /* RFC2461 7.1.1:
702                  *      If the IP source address is the unspecified address,
703                  *      there MUST NOT be source link-layer address option
704                  *      in the message.
705                  */
706                 if (dad) {
707                         ND_PRINTK2(KERN_WARNING
708                                    "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
709                         return;
710                 }
711         }
712
713         inc = ipv6_addr_is_multicast(daddr);
714
715         if ((ifp = ipv6_get_ifaddr(&msg->target, dev, 1)) != NULL) {
716
717                 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
718                         if (dad) {
719                                 if (dev->type == ARPHRD_IEEE802_TR) {
720                                         const unsigned char *sadr;
721                                         sadr = skb_mac_header(skb);
722                                         if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
723                                             sadr[9] == dev->dev_addr[1] &&
724                                             sadr[10] == dev->dev_addr[2] &&
725                                             sadr[11] == dev->dev_addr[3] &&
726                                             sadr[12] == dev->dev_addr[4] &&
727                                             sadr[13] == dev->dev_addr[5]) {
728                                                 /* looped-back to us */
729                                                 goto out;
730                                         }
731                                 }
732
733                                 /*
734                                  * We are colliding with another node
735                                  * who is doing DAD
736                                  * so fail our DAD process
737                                  */
738                                 addrconf_dad_failure(ifp);
739                                 goto out;
740                         } else {
741                                 /*
742                                  * This is not a dad solicitation.
743                                  * If we are an optimistic node,
744                                  * we should respond.
745                                  * Otherwise, we should ignore it.
746                                  */
747                                 if (!(ifp->flags & IFA_F_OPTIMISTIC))
748                                         goto out;
749                         }
750                 }
751
752                 idev = ifp->idev;
753         } else {
754                 idev = in6_dev_get(dev);
755                 if (!idev) {
756                         /* XXX: count this drop? */
757                         return;
758                 }
759
760                 if (ipv6_chk_acast_addr(dev, &msg->target) ||
761                     (idev->cnf.forwarding &&
762                      (ipv6_devconf.proxy_ndp || idev->cnf.proxy_ndp) &&
763                      (pneigh = pneigh_lookup(&nd_tbl,
764                                              &msg->target, dev, 0)) != NULL)) {
765                         if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
766                             skb->pkt_type != PACKET_HOST &&
767                             inc != 0 &&
768                             idev->nd_parms->proxy_delay != 0) {
769                                 /*
770                                  * for anycast or proxy,
771                                  * sender should delay its response
772                                  * by a random time between 0 and
773                                  * MAX_ANYCAST_DELAY_TIME seconds.
774                                  * (RFC2461) -- yoshfuji
775                                  */
776                                 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
777                                 if (n)
778                                         pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
779                                 goto out;
780                         }
781                 } else
782                         goto out;
783         }
784
785         is_router = !!(pneigh ? pneigh->flags & NTF_ROUTER : idev->cnf.forwarding);
786
787         if (dad) {
788                 struct in6_addr maddr;
789
790                 ipv6_addr_all_nodes(&maddr);
791                 ndisc_send_na(dev, NULL, &maddr, &msg->target,
792                               is_router, 0, (ifp != NULL), 1);
793                 goto out;
794         }
795
796         if (inc)
797                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
798         else
799                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
800
801         /*
802          *      update / create cache entry
803          *      for the source address
804          */
805         neigh = __neigh_lookup(&nd_tbl, saddr, dev,
806                                !inc || lladdr || !dev->addr_len);
807         if (neigh)
808                 neigh_update(neigh, lladdr, NUD_STALE,
809                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
810                              NEIGH_UPDATE_F_OVERRIDE);
811         if (neigh || !dev->hard_header) {
812                 ndisc_send_na(dev, neigh, saddr, &msg->target,
813                               is_router,
814                               1, (ifp != NULL && inc), inc);
815                 if (neigh)
816                         neigh_release(neigh);
817         }
818
819 out:
820         if (ifp)
821                 in6_ifa_put(ifp);
822         else
823                 in6_dev_put(idev);
824
825         return;
826 }
827
828 static void ndisc_recv_na(struct sk_buff *skb)
829 {
830         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
831         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
832         struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
833         u8 *lladdr = NULL;
834         u32 ndoptlen = skb->tail - (skb->transport_header +
835                                     offsetof(struct nd_msg, opt));
836         struct ndisc_options ndopts;
837         struct net_device *dev = skb->dev;
838         struct inet6_ifaddr *ifp;
839         struct neighbour *neigh;
840
841         if (skb->len < sizeof(struct nd_msg)) {
842                 ND_PRINTK2(KERN_WARNING
843                            "ICMPv6 NA: packet too short\n");
844                 return;
845         }
846
847         if (ipv6_addr_is_multicast(&msg->target)) {
848                 ND_PRINTK2(KERN_WARNING
849                            "ICMPv6 NA: target address is multicast.\n");
850                 return;
851         }
852
853         if (ipv6_addr_is_multicast(daddr) &&
854             msg->icmph.icmp6_solicited) {
855                 ND_PRINTK2(KERN_WARNING
856                            "ICMPv6 NA: solicited NA is multicasted.\n");
857                 return;
858         }
859
860         if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
861                 ND_PRINTK2(KERN_WARNING
862                            "ICMPv6 NS: invalid ND option\n");
863                 return;
864         }
865         if (ndopts.nd_opts_tgt_lladdr) {
866                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
867                 if (!lladdr) {
868                         ND_PRINTK2(KERN_WARNING
869                                    "ICMPv6 NA: invalid link-layer address length\n");
870                         return;
871                 }
872         }
873         if ((ifp = ipv6_get_ifaddr(&msg->target, dev, 1))) {
874                 if (ifp->flags & IFA_F_TENTATIVE) {
875                         addrconf_dad_failure(ifp);
876                         return;
877                 }
878                 /* What should we make now? The advertisement
879                    is invalid, but ndisc specs say nothing
880                    about it. It could be misconfiguration, or
881                    an smart proxy agent tries to help us :-)
882                  */
883                 ND_PRINTK1(KERN_WARNING
884                            "ICMPv6 NA: someone advertises our address on %s!\n",
885                            ifp->idev->dev->name);
886                 in6_ifa_put(ifp);
887                 return;
888         }
889         neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
890
891         if (neigh) {
892                 u8 old_flags = neigh->flags;
893
894                 if (neigh->nud_state & NUD_FAILED)
895                         goto out;
896
897                 /*
898                  * Don't update the neighbor cache entry on a proxy NA from
899                  * ourselves because either the proxied node is off link or it
900                  * has already sent a NA to us.
901                  */
902                 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
903                     ipv6_devconf.forwarding && ipv6_devconf.proxy_ndp &&
904                     pneigh_lookup(&nd_tbl, &msg->target, dev, 0)) {
905                         /* XXX: idev->cnf.prixy_ndp */
906                         goto out;
907                 }
908
909                 neigh_update(neigh, lladdr,
910                              msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
911                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
912                              (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
913                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
914                              (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
915
916                 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
917                         /*
918                          * Change: router to host
919                          */
920                         struct rt6_info *rt;
921                         rt = rt6_get_dflt_router(saddr, dev);
922                         if (rt)
923                                 ip6_del_rt(rt);
924                 }
925
926 out:
927                 neigh_release(neigh);
928         }
929 }
930
931 static void ndisc_recv_rs(struct sk_buff *skb)
932 {
933         struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
934         unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
935         struct neighbour *neigh;
936         struct inet6_dev *idev;
937         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
938         struct ndisc_options ndopts;
939         u8 *lladdr = NULL;
940
941         if (skb->len < sizeof(*rs_msg))
942                 return;
943
944         idev = in6_dev_get(skb->dev);
945         if (!idev) {
946                 if (net_ratelimit())
947                         ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
948                 return;
949         }
950
951         /* Don't accept RS if we're not in router mode */
952         if (!idev->cnf.forwarding)
953                 goto out;
954
955         /*
956          * Don't update NCE if src = ::;
957          * this implies that the source node has no ip address assigned yet.
958          */
959         if (ipv6_addr_any(saddr))
960                 goto out;
961
962         /* Parse ND options */
963         if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
964                 if (net_ratelimit())
965                         ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
966                 goto out;
967         }
968
969         if (ndopts.nd_opts_src_lladdr) {
970                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
971                                              skb->dev);
972                 if (!lladdr)
973                         goto out;
974         }
975
976         neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
977         if (neigh) {
978                 neigh_update(neigh, lladdr, NUD_STALE,
979                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
980                              NEIGH_UPDATE_F_OVERRIDE|
981                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
982                 neigh_release(neigh);
983         }
984 out:
985         in6_dev_put(idev);
986 }
987
988 static void ndisc_router_discovery(struct sk_buff *skb)
989 {
990         struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
991         struct neighbour *neigh = NULL;
992         struct inet6_dev *in6_dev;
993         struct rt6_info *rt = NULL;
994         int lifetime;
995         struct ndisc_options ndopts;
996         int optlen;
997         unsigned int pref = 0;
998
999         __u8 * opt = (__u8 *)(ra_msg + 1);
1000
1001         optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1002
1003         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1004                 ND_PRINTK2(KERN_WARNING
1005                            "ICMPv6 RA: source address is not link-local.\n");
1006                 return;
1007         }
1008         if (optlen < 0) {
1009                 ND_PRINTK2(KERN_WARNING
1010                            "ICMPv6 RA: packet too short\n");
1011                 return;
1012         }
1013
1014         /*
1015          *      set the RA_RECV flag in the interface
1016          */
1017
1018         in6_dev = in6_dev_get(skb->dev);
1019         if (in6_dev == NULL) {
1020                 ND_PRINTK0(KERN_ERR
1021                            "ICMPv6 RA: can't find inet6 device for %s.\n",
1022                            skb->dev->name);
1023                 return;
1024         }
1025         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra) {
1026                 in6_dev_put(in6_dev);
1027                 return;
1028         }
1029
1030         if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1031                 in6_dev_put(in6_dev);
1032                 ND_PRINTK2(KERN_WARNING
1033                            "ICMP6 RA: invalid ND options\n");
1034                 return;
1035         }
1036
1037         if (in6_dev->if_flags & IF_RS_SENT) {
1038                 /*
1039                  *      flag that an RA was received after an RS was sent
1040                  *      out on this interface.
1041                  */
1042                 in6_dev->if_flags |= IF_RA_RCVD;
1043         }
1044
1045         /*
1046          * Remember the managed/otherconf flags from most recently
1047          * received RA message (RFC 2462) -- yoshfuji
1048          */
1049         in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1050                                 IF_RA_OTHERCONF)) |
1051                                 (ra_msg->icmph.icmp6_addrconf_managed ?
1052                                         IF_RA_MANAGED : 0) |
1053                                 (ra_msg->icmph.icmp6_addrconf_other ?
1054                                         IF_RA_OTHERCONF : 0);
1055
1056         if (!in6_dev->cnf.accept_ra_defrtr)
1057                 goto skip_defrtr;
1058
1059         lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1060
1061 #ifdef CONFIG_IPV6_ROUTER_PREF
1062         pref = ra_msg->icmph.icmp6_router_pref;
1063         /* 10b is handled as if it were 00b (medium) */
1064         if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1065             in6_dev->cnf.accept_ra_rtr_pref)
1066                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1067 #endif
1068
1069         rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1070
1071         if (rt)
1072                 neigh = rt->rt6i_nexthop;
1073
1074         if (rt && lifetime == 0) {
1075                 neigh_clone(neigh);
1076                 ip6_del_rt(rt);
1077                 rt = NULL;
1078         }
1079
1080         if (rt == NULL && lifetime) {
1081                 ND_PRINTK3(KERN_DEBUG
1082                            "ICMPv6 RA: adding default router.\n");
1083
1084                 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1085                 if (rt == NULL) {
1086                         ND_PRINTK0(KERN_ERR
1087                                    "ICMPv6 RA: %s() failed to add default route.\n",
1088                                    __FUNCTION__);
1089                         in6_dev_put(in6_dev);
1090                         return;
1091                 }
1092
1093                 neigh = rt->rt6i_nexthop;
1094                 if (neigh == NULL) {
1095                         ND_PRINTK0(KERN_ERR
1096                                    "ICMPv6 RA: %s() got default router without neighbour.\n",
1097                                    __FUNCTION__);
1098                         dst_release(&rt->u.dst);
1099                         in6_dev_put(in6_dev);
1100                         return;
1101                 }
1102                 neigh->flags |= NTF_ROUTER;
1103         } else if (rt) {
1104                 rt->rt6i_flags |= (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1105         }
1106
1107         if (rt)
1108                 rt->rt6i_expires = jiffies + (HZ * lifetime);
1109
1110         if (ra_msg->icmph.icmp6_hop_limit) {
1111                 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1112                 if (rt)
1113                         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1114         }
1115
1116 skip_defrtr:
1117
1118         /*
1119          *      Update Reachable Time and Retrans Timer
1120          */
1121
1122         if (in6_dev->nd_parms) {
1123                 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1124
1125                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1126                         rtime = (rtime*HZ)/1000;
1127                         if (rtime < HZ/10)
1128                                 rtime = HZ/10;
1129                         in6_dev->nd_parms->retrans_time = rtime;
1130                         in6_dev->tstamp = jiffies;
1131                         inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1132                 }
1133
1134                 rtime = ntohl(ra_msg->reachable_time);
1135                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1136                         rtime = (rtime*HZ)/1000;
1137
1138                         if (rtime < HZ/10)
1139                                 rtime = HZ/10;
1140
1141                         if (rtime != in6_dev->nd_parms->base_reachable_time) {
1142                                 in6_dev->nd_parms->base_reachable_time = rtime;
1143                                 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1144                                 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1145                                 in6_dev->tstamp = jiffies;
1146                                 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1147                         }
1148                 }
1149         }
1150
1151         /*
1152          *      Process options.
1153          */
1154
1155         if (!neigh)
1156                 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1157                                        skb->dev, 1);
1158         if (neigh) {
1159                 u8 *lladdr = NULL;
1160                 if (ndopts.nd_opts_src_lladdr) {
1161                         lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1162                                                      skb->dev);
1163                         if (!lladdr) {
1164                                 ND_PRINTK2(KERN_WARNING
1165                                            "ICMPv6 RA: invalid link-layer address length\n");
1166                                 goto out;
1167                         }
1168                 }
1169                 neigh_update(neigh, lladdr, NUD_STALE,
1170                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1171                              NEIGH_UPDATE_F_OVERRIDE|
1172                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1173                              NEIGH_UPDATE_F_ISROUTER);
1174         }
1175
1176 #ifdef CONFIG_IPV6_ROUTE_INFO
1177         if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1178                 struct nd_opt_hdr *p;
1179                 for (p = ndopts.nd_opts_ri;
1180                      p;
1181                      p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1182                         if (((struct route_info *)p)->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1183                                 continue;
1184                         rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1185                                       &ipv6_hdr(skb)->saddr);
1186                 }
1187         }
1188 #endif
1189
1190         if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1191                 struct nd_opt_hdr *p;
1192                 for (p = ndopts.nd_opts_pi;
1193                      p;
1194                      p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1195                         addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1196                 }
1197         }
1198
1199         if (ndopts.nd_opts_mtu) {
1200                 __be32 n;
1201                 u32 mtu;
1202
1203                 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1204                 mtu = ntohl(n);
1205
1206                 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1207                         ND_PRINTK2(KERN_WARNING
1208                                    "ICMPv6 RA: invalid mtu: %d\n",
1209                                    mtu);
1210                 } else if (in6_dev->cnf.mtu6 != mtu) {
1211                         in6_dev->cnf.mtu6 = mtu;
1212
1213                         if (rt)
1214                                 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
1215
1216                         rt6_mtu_change(skb->dev, mtu);
1217                 }
1218         }
1219
1220         if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1221                 ND_PRINTK2(KERN_WARNING
1222                            "ICMPv6 RA: invalid RA options");
1223         }
1224 out:
1225         if (rt)
1226                 dst_release(&rt->u.dst);
1227         else if (neigh)
1228                 neigh_release(neigh);
1229         in6_dev_put(in6_dev);
1230 }
1231
1232 static void ndisc_redirect_rcv(struct sk_buff *skb)
1233 {
1234         struct inet6_dev *in6_dev;
1235         struct icmp6hdr *icmph;
1236         struct in6_addr *dest;
1237         struct in6_addr *target;        /* new first hop to destination */
1238         struct neighbour *neigh;
1239         int on_link = 0;
1240         struct ndisc_options ndopts;
1241         int optlen;
1242         u8 *lladdr = NULL;
1243
1244         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1245                 ND_PRINTK2(KERN_WARNING
1246                            "ICMPv6 Redirect: source address is not link-local.\n");
1247                 return;
1248         }
1249
1250         optlen = skb->tail - skb->transport_header;
1251         optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1252
1253         if (optlen < 0) {
1254                 ND_PRINTK2(KERN_WARNING
1255                            "ICMPv6 Redirect: packet too short\n");
1256                 return;
1257         }
1258
1259         icmph = icmp6_hdr(skb);
1260         target = (struct in6_addr *) (icmph + 1);
1261         dest = target + 1;
1262
1263         if (ipv6_addr_is_multicast(dest)) {
1264                 ND_PRINTK2(KERN_WARNING
1265                            "ICMPv6 Redirect: destination address is multicast.\n");
1266                 return;
1267         }
1268
1269         if (ipv6_addr_equal(dest, target)) {
1270                 on_link = 1;
1271         } else if (!(ipv6_addr_type(target) & IPV6_ADDR_LINKLOCAL)) {
1272                 ND_PRINTK2(KERN_WARNING
1273                            "ICMPv6 Redirect: target address is not link-local.\n");
1274                 return;
1275         }
1276
1277         in6_dev = in6_dev_get(skb->dev);
1278         if (!in6_dev)
1279                 return;
1280         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1281                 in6_dev_put(in6_dev);
1282                 return;
1283         }
1284
1285         /* RFC2461 8.1:
1286          *      The IP source address of the Redirect MUST be the same as the current
1287          *      first-hop router for the specified ICMP Destination Address.
1288          */
1289
1290         if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1291                 ND_PRINTK2(KERN_WARNING
1292                            "ICMPv6 Redirect: invalid ND options\n");
1293                 in6_dev_put(in6_dev);
1294                 return;
1295         }
1296         if (ndopts.nd_opts_tgt_lladdr) {
1297                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1298                                              skb->dev);
1299                 if (!lladdr) {
1300                         ND_PRINTK2(KERN_WARNING
1301                                    "ICMPv6 Redirect: invalid link-layer address length\n");
1302                         in6_dev_put(in6_dev);
1303                         return;
1304                 }
1305         }
1306
1307         neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1308         if (neigh) {
1309                 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1310                              &ipv6_hdr(skb)->saddr, neigh, lladdr,
1311                              on_link);
1312                 neigh_release(neigh);
1313         }
1314         in6_dev_put(in6_dev);
1315 }
1316
1317 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1318                          struct in6_addr *target)
1319 {
1320         struct sock *sk = ndisc_socket->sk;
1321         int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1322         struct sk_buff *buff;
1323         struct icmp6hdr *icmph;
1324         struct in6_addr saddr_buf;
1325         struct in6_addr *addrp;
1326         struct net_device *dev;
1327         struct rt6_info *rt;
1328         struct dst_entry *dst;
1329         struct inet6_dev *idev;
1330         struct flowi fl;
1331         u8 *opt;
1332         int rd_len;
1333         int err;
1334         int hlen;
1335         u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1336
1337         dev = skb->dev;
1338
1339         if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1340                 ND_PRINTK2(KERN_WARNING
1341                            "ICMPv6 Redirect: no link-local address on %s\n",
1342                            dev->name);
1343                 return;
1344         }
1345
1346         if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1347             !(ipv6_addr_type(target) & IPV6_ADDR_LINKLOCAL)) {
1348                 ND_PRINTK2(KERN_WARNING
1349                         "ICMPv6 Redirect: target address is not link-local.\n");
1350                 return;
1351         }
1352
1353         ndisc_flow_init(&fl, NDISC_REDIRECT, &saddr_buf, &ipv6_hdr(skb)->saddr,
1354                         dev->ifindex);
1355
1356         dst = ip6_route_output(NULL, &fl);
1357         if (dst == NULL)
1358                 return;
1359
1360         err = xfrm_lookup(&dst, &fl, NULL, 0);
1361         if (err)
1362                 return;
1363
1364         rt = (struct rt6_info *) dst;
1365
1366         if (rt->rt6i_flags & RTF_GATEWAY) {
1367                 ND_PRINTK2(KERN_WARNING
1368                            "ICMPv6 Redirect: destination is not a neighbour.\n");
1369                 dst_release(dst);
1370                 return;
1371         }
1372         if (!xrlim_allow(dst, 1*HZ)) {
1373                 dst_release(dst);
1374                 return;
1375         }
1376
1377         if (dev->addr_len) {
1378                 read_lock_bh(&neigh->lock);
1379                 if (neigh->nud_state & NUD_VALID) {
1380                         memcpy(ha_buf, neigh->ha, dev->addr_len);
1381                         read_unlock_bh(&neigh->lock);
1382                         ha = ha_buf;
1383                         len += ndisc_opt_addr_space(dev);
1384                 } else
1385                         read_unlock_bh(&neigh->lock);
1386         }
1387
1388         rd_len = min_t(unsigned int,
1389                      IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1390         rd_len &= ~0x7;
1391         len += rd_len;
1392
1393         buff = sock_alloc_send_skb(sk,
1394                                    (MAX_HEADER + sizeof(struct ipv6hdr) +
1395                                     len + LL_RESERVED_SPACE(dev)),
1396                                    1, &err);
1397         if (buff == NULL) {
1398                 ND_PRINTK0(KERN_ERR
1399                            "ICMPv6 Redirect: %s() failed to allocate an skb.\n",
1400                            __FUNCTION__);
1401                 dst_release(dst);
1402                 return;
1403         }
1404
1405         hlen = 0;
1406
1407         skb_reserve(buff, LL_RESERVED_SPACE(dev));
1408         ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1409                    IPPROTO_ICMPV6, len);
1410
1411         skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1412         skb_put(buff, len);
1413         icmph = icmp6_hdr(buff);
1414
1415         memset(icmph, 0, sizeof(struct icmp6hdr));
1416         icmph->icmp6_type = NDISC_REDIRECT;
1417
1418         /*
1419          *      copy target and destination addresses
1420          */
1421
1422         addrp = (struct in6_addr *)(icmph + 1);
1423         ipv6_addr_copy(addrp, target);
1424         addrp++;
1425         ipv6_addr_copy(addrp, &ipv6_hdr(skb)->daddr);
1426
1427         opt = (u8*) (addrp + 1);
1428
1429         /*
1430          *      include target_address option
1431          */
1432
1433         if (ha)
1434                 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1435                                              dev->addr_len, dev->type);
1436
1437         /*
1438          *      build redirect option and copy skb over to the new packet.
1439          */
1440
1441         memset(opt, 0, 8);
1442         *(opt++) = ND_OPT_REDIRECT_HDR;
1443         *(opt++) = (rd_len >> 3);
1444         opt += 6;
1445
1446         memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1447
1448         icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1449                                              len, IPPROTO_ICMPV6,
1450                                              csum_partial((u8 *) icmph, len, 0));
1451
1452         buff->dst = dst;
1453         idev = in6_dev_get(dst->dev);
1454         IP6_INC_STATS(idev, IPSTATS_MIB_OUTREQUESTS);
1455         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, buff, NULL, dst->dev, dst_output);
1456         if (!err) {
1457                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTREDIRECTS);
1458                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
1459         }
1460
1461         if (likely(idev != NULL))
1462                 in6_dev_put(idev);
1463 }
1464
1465 static void pndisc_redo(struct sk_buff *skb)
1466 {
1467         ndisc_recv_ns(skb);
1468         kfree_skb(skb);
1469 }
1470
1471 int ndisc_rcv(struct sk_buff *skb)
1472 {
1473         struct nd_msg *msg;
1474
1475         if (!pskb_may_pull(skb, skb->len))
1476                 return 0;
1477
1478         msg = (struct nd_msg *)skb_transport_header(skb);
1479
1480         __skb_push(skb, skb->data - skb_transport_header(skb));
1481
1482         if (ipv6_hdr(skb)->hop_limit != 255) {
1483                 ND_PRINTK2(KERN_WARNING
1484                            "ICMPv6 NDISC: invalid hop-limit: %d\n",
1485                            ipv6_hdr(skb)->hop_limit);
1486                 return 0;
1487         }
1488
1489         if (msg->icmph.icmp6_code != 0) {
1490                 ND_PRINTK2(KERN_WARNING
1491                            "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1492                            msg->icmph.icmp6_code);
1493                 return 0;
1494         }
1495
1496         memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1497
1498         switch (msg->icmph.icmp6_type) {
1499         case NDISC_NEIGHBOUR_SOLICITATION:
1500                 ndisc_recv_ns(skb);
1501                 break;
1502
1503         case NDISC_NEIGHBOUR_ADVERTISEMENT:
1504                 ndisc_recv_na(skb);
1505                 break;
1506
1507         case NDISC_ROUTER_SOLICITATION:
1508                 ndisc_recv_rs(skb);
1509                 break;
1510
1511         case NDISC_ROUTER_ADVERTISEMENT:
1512                 ndisc_router_discovery(skb);
1513                 break;
1514
1515         case NDISC_REDIRECT:
1516                 ndisc_redirect_rcv(skb);
1517                 break;
1518         }
1519
1520         return 0;
1521 }
1522
1523 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1524 {
1525         struct net_device *dev = ptr;
1526
1527         switch (event) {
1528         case NETDEV_CHANGEADDR:
1529                 neigh_changeaddr(&nd_tbl, dev);
1530                 fib6_run_gc(~0UL);
1531                 break;
1532         case NETDEV_DOWN:
1533                 neigh_ifdown(&nd_tbl, dev);
1534                 fib6_run_gc(~0UL);
1535                 break;
1536         default:
1537                 break;
1538         }
1539
1540         return NOTIFY_DONE;
1541 }
1542
1543 static struct notifier_block ndisc_netdev_notifier = {
1544         .notifier_call = ndisc_netdev_event,
1545 };
1546
1547 #ifdef CONFIG_SYSCTL
1548 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1549                                          const char *func, const char *dev_name)
1550 {
1551         static char warncomm[TASK_COMM_LEN];
1552         static int warned;
1553         if (strcmp(warncomm, current->comm) && warned < 5) {
1554                 strcpy(warncomm, current->comm);
1555                 printk(KERN_WARNING
1556                         "process `%s' is using deprecated sysctl (%s) "
1557                         "net.ipv6.neigh.%s.%s; "
1558                         "Use net.ipv6.neigh.%s.%s_ms "
1559                         "instead.\n",
1560                         warncomm, func,
1561                         dev_name, ctl->procname,
1562                         dev_name, ctl->procname);
1563                 warned++;
1564         }
1565 }
1566
1567 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, struct file * filp, void __user *buffer, size_t *lenp, loff_t *ppos)
1568 {
1569         struct net_device *dev = ctl->extra1;
1570         struct inet6_dev *idev;
1571         int ret;
1572
1573         if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1574             ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1575                 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1576
1577         switch (ctl->ctl_name) {
1578         case NET_NEIGH_RETRANS_TIME:
1579                 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1580                 break;
1581         case NET_NEIGH_REACHABLE_TIME:
1582                 ret = proc_dointvec_jiffies(ctl, write,
1583                                             filp, buffer, lenp, ppos);
1584                 break;
1585         case NET_NEIGH_RETRANS_TIME_MS:
1586         case NET_NEIGH_REACHABLE_TIME_MS:
1587                 ret = proc_dointvec_ms_jiffies(ctl, write,
1588                                                filp, buffer, lenp, ppos);
1589                 break;
1590         default:
1591                 ret = -1;
1592         }
1593
1594         if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1595                 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1596                     ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1597                         idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1598                 idev->tstamp = jiffies;
1599                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1600                 in6_dev_put(idev);
1601         }
1602         return ret;
1603 }
1604
1605 static int ndisc_ifinfo_sysctl_strategy(ctl_table *ctl, int __user *name,
1606                                         int nlen, void __user *oldval,
1607                                         size_t __user *oldlenp,
1608                                         void __user *newval, size_t newlen)
1609 {
1610         struct net_device *dev = ctl->extra1;
1611         struct inet6_dev *idev;
1612         int ret;
1613
1614         if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1615             ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1616                 ndisc_warn_deprecated_sysctl(ctl, "procfs", dev ? dev->name : "default");
1617
1618         switch (ctl->ctl_name) {
1619         case NET_NEIGH_REACHABLE_TIME:
1620                 ret = sysctl_jiffies(ctl, name, nlen,
1621                                      oldval, oldlenp, newval, newlen);
1622                 break;
1623         case NET_NEIGH_RETRANS_TIME_MS:
1624         case NET_NEIGH_REACHABLE_TIME_MS:
1625                  ret = sysctl_ms_jiffies(ctl, name, nlen,
1626                                          oldval, oldlenp, newval, newlen);
1627                  break;
1628         default:
1629                 ret = 0;
1630         }
1631
1632         if (newval && newlen && ret > 0 &&
1633             dev && (idev = in6_dev_get(dev)) != NULL) {
1634                 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1635                     ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1636                         idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1637                 idev->tstamp = jiffies;
1638                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1639                 in6_dev_put(idev);
1640         }
1641
1642         return ret;
1643 }
1644
1645 #endif
1646
1647 int __init ndisc_init(struct net_proto_family *ops)
1648 {
1649         struct ipv6_pinfo *np;
1650         struct sock *sk;
1651         int err;
1652
1653         err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, &ndisc_socket);
1654         if (err < 0) {
1655                 ND_PRINTK0(KERN_ERR
1656                            "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1657                            err);
1658                 ndisc_socket = NULL; /* For safety. */
1659                 return err;
1660         }
1661
1662         sk = ndisc_socket->sk;
1663         np = inet6_sk(sk);
1664         sk->sk_allocation = GFP_ATOMIC;
1665         np->hop_limit = 255;
1666         /* Do not loopback ndisc messages */
1667         np->mc_loop = 0;
1668         sk->sk_prot->unhash(sk);
1669
1670         /*
1671          * Initialize the neighbour table
1672          */
1673
1674         neigh_table_init(&nd_tbl);
1675
1676 #ifdef CONFIG_SYSCTL
1677         neigh_sysctl_register(NULL, &nd_tbl.parms, NET_IPV6, NET_IPV6_NEIGH,
1678                               "ipv6",
1679                               &ndisc_ifinfo_sysctl_change,
1680                               &ndisc_ifinfo_sysctl_strategy);
1681 #endif
1682
1683         register_netdevice_notifier(&ndisc_netdev_notifier);
1684         return 0;
1685 }
1686
1687 void ndisc_cleanup(void)
1688 {
1689         unregister_netdevice_notifier(&ndisc_netdev_notifier);
1690 #ifdef CONFIG_SYSCTL
1691         neigh_sysctl_unregister(&nd_tbl.parms);
1692 #endif
1693         neigh_table_clear(&nd_tbl);
1694         sock_release(ndisc_socket);
1695         ndisc_socket = NULL; /* For safety. */
1696 }