[NET]: Use local_irq_{save,restore}() in napi_complete().
[linux-2.6.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 <linux/proc_fs.h>
88
89 #include <linux/netfilter.h>
90 #include <linux/netfilter_ipv6.h>
91
92 static struct socket *ndisc_socket;
93
94 static u32 ndisc_hash(const void *pkey, const struct net_device *dev);
95 static int ndisc_constructor(struct neighbour *neigh);
96 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
97 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
98 static int pndisc_constructor(struct pneigh_entry *n);
99 static void pndisc_destructor(struct pneigh_entry *n);
100 static void pndisc_redo(struct sk_buff *skb);
101
102 static struct neigh_ops ndisc_generic_ops = {
103         .family =               AF_INET6,
104         .solicit =              ndisc_solicit,
105         .error_report =         ndisc_error_report,
106         .output =               neigh_resolve_output,
107         .connected_output =     neigh_connected_output,
108         .hh_output =            dev_queue_xmit,
109         .queue_xmit =           dev_queue_xmit,
110 };
111
112 static struct neigh_ops ndisc_hh_ops = {
113         .family =               AF_INET6,
114         .solicit =              ndisc_solicit,
115         .error_report =         ndisc_error_report,
116         .output =               neigh_resolve_output,
117         .connected_output =     neigh_resolve_output,
118         .hh_output =            dev_queue_xmit,
119         .queue_xmit =           dev_queue_xmit,
120 };
121
122
123 static struct neigh_ops ndisc_direct_ops = {
124         .family =               AF_INET6,
125         .output =               dev_queue_xmit,
126         .connected_output =     dev_queue_xmit,
127         .hh_output =            dev_queue_xmit,
128         .queue_xmit =           dev_queue_xmit,
129 };
130
131 struct neigh_table nd_tbl = {
132         .family =       AF_INET6,
133         .entry_size =   sizeof(struct neighbour) + sizeof(struct in6_addr),
134         .key_len =      sizeof(struct in6_addr),
135         .hash =         ndisc_hash,
136         .constructor =  ndisc_constructor,
137         .pconstructor = pndisc_constructor,
138         .pdestructor =  pndisc_destructor,
139         .proxy_redo =   pndisc_redo,
140         .id =           "ndisc_cache",
141         .parms = {
142                 .tbl =                  &nd_tbl,
143                 .base_reachable_time =  30 * HZ,
144                 .retrans_time =  1 * HZ,
145                 .gc_staletime = 60 * HZ,
146                 .reachable_time =               30 * HZ,
147                 .delay_probe_time =      5 * HZ,
148                 .queue_len =             3,
149                 .ucast_probes =  3,
150                 .mcast_probes =  3,
151                 .anycast_delay =         1 * HZ,
152                 .proxy_delay =          (8 * HZ) / 10,
153                 .proxy_qlen =           64,
154         },
155         .gc_interval =    30 * HZ,
156         .gc_thresh1 =    128,
157         .gc_thresh2 =    512,
158         .gc_thresh3 =   1024,
159 };
160
161 /* ND options */
162 struct ndisc_options {
163         struct nd_opt_hdr *nd_opt_array[__ND_OPT_ARRAY_MAX];
164 #ifdef CONFIG_IPV6_ROUTE_INFO
165         struct nd_opt_hdr *nd_opts_ri;
166         struct nd_opt_hdr *nd_opts_ri_end;
167 #endif
168         struct nd_opt_hdr *nd_useropts;
169         struct nd_opt_hdr *nd_useropts_end;
170 };
171
172 #define nd_opts_src_lladdr      nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
173 #define nd_opts_tgt_lladdr      nd_opt_array[ND_OPT_TARGET_LL_ADDR]
174 #define nd_opts_pi              nd_opt_array[ND_OPT_PREFIX_INFO]
175 #define nd_opts_pi_end          nd_opt_array[__ND_OPT_PREFIX_INFO_END]
176 #define nd_opts_rh              nd_opt_array[ND_OPT_REDIRECT_HDR]
177 #define nd_opts_mtu             nd_opt_array[ND_OPT_MTU]
178
179 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
180
181 /*
182  * Return the padding between the option length and the start of the
183  * link addr.  Currently only IP-over-InfiniBand needs this, although
184  * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
185  * also need a pad of 2.
186  */
187 static int ndisc_addr_option_pad(unsigned short type)
188 {
189         switch (type) {
190         case ARPHRD_INFINIBAND: return 2;
191         default:                return 0;
192         }
193 }
194
195 static inline int ndisc_opt_addr_space(struct net_device *dev)
196 {
197         return NDISC_OPT_SPACE(dev->addr_len + ndisc_addr_option_pad(dev->type));
198 }
199
200 static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
201                                   unsigned short addr_type)
202 {
203         int space = NDISC_OPT_SPACE(data_len);
204         int pad   = ndisc_addr_option_pad(addr_type);
205
206         opt[0] = type;
207         opt[1] = space>>3;
208
209         memset(opt + 2, 0, pad);
210         opt   += pad;
211         space -= pad;
212
213         memcpy(opt+2, data, data_len);
214         data_len += 2;
215         opt += data_len;
216         if ((space -= data_len) > 0)
217                 memset(opt, 0, space);
218         return opt + space;
219 }
220
221 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
222                                             struct nd_opt_hdr *end)
223 {
224         int type;
225         if (!cur || !end || cur >= end)
226                 return NULL;
227         type = cur->nd_opt_type;
228         do {
229                 cur = ((void *)cur) + (cur->nd_opt_len << 3);
230         } while(cur < end && cur->nd_opt_type != type);
231         return (cur <= end && cur->nd_opt_type == type ? cur : NULL);
232 }
233
234 static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
235 {
236         return (opt->nd_opt_type == ND_OPT_RDNSS);
237 }
238
239 static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
240                                              struct nd_opt_hdr *end)
241 {
242         if (!cur || !end || cur >= end)
243                 return NULL;
244         do {
245                 cur = ((void *)cur) + (cur->nd_opt_len << 3);
246         } while(cur < end && !ndisc_is_useropt(cur));
247         return (cur <= end && ndisc_is_useropt(cur) ? cur : NULL);
248 }
249
250 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
251                                                  struct ndisc_options *ndopts)
252 {
253         struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
254
255         if (!nd_opt || opt_len < 0 || !ndopts)
256                 return NULL;
257         memset(ndopts, 0, sizeof(*ndopts));
258         while (opt_len) {
259                 int l;
260                 if (opt_len < sizeof(struct nd_opt_hdr))
261                         return NULL;
262                 l = nd_opt->nd_opt_len << 3;
263                 if (opt_len < l || l == 0)
264                         return NULL;
265                 switch (nd_opt->nd_opt_type) {
266                 case ND_OPT_SOURCE_LL_ADDR:
267                 case ND_OPT_TARGET_LL_ADDR:
268                 case ND_OPT_MTU:
269                 case ND_OPT_REDIRECT_HDR:
270                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
271                                 ND_PRINTK2(KERN_WARNING
272                                            "%s(): duplicated ND6 option found: type=%d\n",
273                                            __FUNCTION__,
274                                            nd_opt->nd_opt_type);
275                         } else {
276                                 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
277                         }
278                         break;
279                 case ND_OPT_PREFIX_INFO:
280                         ndopts->nd_opts_pi_end = nd_opt;
281                         if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
282                                 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
283                         break;
284 #ifdef CONFIG_IPV6_ROUTE_INFO
285                 case ND_OPT_ROUTE_INFO:
286                         ndopts->nd_opts_ri_end = nd_opt;
287                         if (!ndopts->nd_opts_ri)
288                                 ndopts->nd_opts_ri = nd_opt;
289                         break;
290 #endif
291                 default:
292                         if (ndisc_is_useropt(nd_opt)) {
293                                 ndopts->nd_useropts_end = nd_opt;
294                                 if (!ndopts->nd_useropts)
295                                         ndopts->nd_useropts = nd_opt;
296                         } else {
297                                 /*
298                                  * Unknown options must be silently ignored,
299                                  * to accommodate future extension to the
300                                  * protocol.
301                                  */
302                                 ND_PRINTK2(KERN_NOTICE
303                                            "%s(): ignored unsupported option; type=%d, len=%d\n",
304                                            __FUNCTION__,
305                                            nd_opt->nd_opt_type, nd_opt->nd_opt_len);
306                         }
307                 }
308                 opt_len -= l;
309                 nd_opt = ((void *)nd_opt) + l;
310         }
311         return ndopts;
312 }
313
314 static inline u8 *ndisc_opt_addr_data(struct nd_opt_hdr *p,
315                                       struct net_device *dev)
316 {
317         u8 *lladdr = (u8 *)(p + 1);
318         int lladdrlen = p->nd_opt_len << 3;
319         int prepad = ndisc_addr_option_pad(dev->type);
320         if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len + prepad))
321                 return NULL;
322         return (lladdr + prepad);
323 }
324
325 int ndisc_mc_map(struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
326 {
327         switch (dev->type) {
328         case ARPHRD_ETHER:
329         case ARPHRD_IEEE802:    /* Not sure. Check it later. --ANK */
330         case ARPHRD_FDDI:
331                 ipv6_eth_mc_map(addr, buf);
332                 return 0;
333         case ARPHRD_IEEE802_TR:
334                 ipv6_tr_mc_map(addr,buf);
335                 return 0;
336         case ARPHRD_ARCNET:
337                 ipv6_arcnet_mc_map(addr, buf);
338                 return 0;
339         case ARPHRD_INFINIBAND:
340                 ipv6_ib_mc_map(addr, dev->broadcast, buf);
341                 return 0;
342         default:
343                 if (dir) {
344                         memcpy(buf, dev->broadcast, dev->addr_len);
345                         return 0;
346                 }
347         }
348         return -EINVAL;
349 }
350
351 EXPORT_SYMBOL(ndisc_mc_map);
352
353 static u32 ndisc_hash(const void *pkey, const struct net_device *dev)
354 {
355         const u32 *p32 = pkey;
356         u32 addr_hash, i;
357
358         addr_hash = 0;
359         for (i = 0; i < (sizeof(struct in6_addr) / sizeof(u32)); i++)
360                 addr_hash ^= *p32++;
361
362         return jhash_2words(addr_hash, dev->ifindex, nd_tbl.hash_rnd);
363 }
364
365 static int ndisc_constructor(struct neighbour *neigh)
366 {
367         struct in6_addr *addr = (struct in6_addr*)&neigh->primary_key;
368         struct net_device *dev = neigh->dev;
369         struct inet6_dev *in6_dev;
370         struct neigh_parms *parms;
371         int is_multicast = ipv6_addr_is_multicast(addr);
372
373         rcu_read_lock();
374         in6_dev = in6_dev_get(dev);
375         if (in6_dev == NULL) {
376                 rcu_read_unlock();
377                 return -EINVAL;
378         }
379
380         parms = in6_dev->nd_parms;
381         __neigh_parms_put(neigh->parms);
382         neigh->parms = neigh_parms_clone(parms);
383         rcu_read_unlock();
384
385         neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
386         if (!dev->header_ops) {
387                 neigh->nud_state = NUD_NOARP;
388                 neigh->ops = &ndisc_direct_ops;
389                 neigh->output = neigh->ops->queue_xmit;
390         } else {
391                 if (is_multicast) {
392                         neigh->nud_state = NUD_NOARP;
393                         ndisc_mc_map(addr, neigh->ha, dev, 1);
394                 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
395                         neigh->nud_state = NUD_NOARP;
396                         memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
397                         if (dev->flags&IFF_LOOPBACK)
398                                 neigh->type = RTN_LOCAL;
399                 } else if (dev->flags&IFF_POINTOPOINT) {
400                         neigh->nud_state = NUD_NOARP;
401                         memcpy(neigh->ha, dev->broadcast, dev->addr_len);
402                 }
403                 if (dev->header_ops->cache)
404                         neigh->ops = &ndisc_hh_ops;
405                 else
406                         neigh->ops = &ndisc_generic_ops;
407                 if (neigh->nud_state&NUD_VALID)
408                         neigh->output = neigh->ops->connected_output;
409                 else
410                         neigh->output = neigh->ops->output;
411         }
412         in6_dev_put(in6_dev);
413         return 0;
414 }
415
416 static int pndisc_constructor(struct pneigh_entry *n)
417 {
418         struct in6_addr *addr = (struct in6_addr*)&n->key;
419         struct in6_addr maddr;
420         struct net_device *dev = n->dev;
421
422         if (dev == NULL || __in6_dev_get(dev) == NULL)
423                 return -EINVAL;
424         addrconf_addr_solict_mult(addr, &maddr);
425         ipv6_dev_mc_inc(dev, &maddr);
426         return 0;
427 }
428
429 static void pndisc_destructor(struct pneigh_entry *n)
430 {
431         struct in6_addr *addr = (struct in6_addr*)&n->key;
432         struct in6_addr maddr;
433         struct net_device *dev = n->dev;
434
435         if (dev == NULL || __in6_dev_get(dev) == NULL)
436                 return;
437         addrconf_addr_solict_mult(addr, &maddr);
438         ipv6_dev_mc_dec(dev, &maddr);
439 }
440
441 /*
442  *      Send a Neighbour Advertisement
443  */
444
445 static inline void ndisc_flow_init(struct flowi *fl, u8 type,
446                             struct in6_addr *saddr, struct in6_addr *daddr,
447                             int oif)
448 {
449         memset(fl, 0, sizeof(*fl));
450         ipv6_addr_copy(&fl->fl6_src, saddr);
451         ipv6_addr_copy(&fl->fl6_dst, daddr);
452         fl->proto               = IPPROTO_ICMPV6;
453         fl->fl_icmp_type        = type;
454         fl->fl_icmp_code        = 0;
455         fl->oif                 = oif;
456         security_sk_classify_flow(ndisc_socket->sk, fl);
457 }
458
459 static void __ndisc_send(struct net_device *dev,
460                          struct neighbour *neigh,
461                          struct in6_addr *daddr, struct in6_addr *saddr,
462                          struct icmp6hdr *icmp6h, struct in6_addr *target,
463                          int llinfo)
464 {
465         struct flowi fl;
466         struct dst_entry *dst;
467         struct sock *sk = ndisc_socket->sk;
468         struct sk_buff *skb;
469         struct icmp6hdr *hdr;
470         struct inet6_dev *idev;
471         int len;
472         int err;
473         u8 *opt, type;
474
475         type = icmp6h->icmp6_type;
476
477         ndisc_flow_init(&fl, type, saddr, daddr,
478                         dev->ifindex);
479
480         dst = ndisc_dst_alloc(dev, neigh, daddr, ip6_output);
481         if (!dst)
482                 return;
483
484         err = xfrm_lookup(&dst, &fl, NULL, 0);
485         if (err < 0)
486                 return;
487
488         if (!dev->addr_len)
489                 llinfo = 0;
490
491         len = sizeof(struct icmp6hdr) + (target ? sizeof(*target) : 0);
492         if (llinfo)
493                 len += ndisc_opt_addr_space(dev);
494
495         skb = sock_alloc_send_skb(sk,
496                                   (MAX_HEADER + sizeof(struct ipv6hdr) +
497                                    len + LL_RESERVED_SPACE(dev)),
498                                   1, &err);
499         if (!skb) {
500                 ND_PRINTK0(KERN_ERR
501                            "ICMPv6 ND: %s() failed to allocate an skb.\n",
502                            __FUNCTION__);
503                 dst_release(dst);
504                 return;
505         }
506
507         skb_reserve(skb, LL_RESERVED_SPACE(dev));
508         ip6_nd_hdr(sk, skb, dev, saddr, daddr, IPPROTO_ICMPV6, len);
509
510         skb->transport_header = skb->tail;
511         skb_put(skb, len);
512
513         hdr = (struct icmp6hdr *)skb_transport_header(skb);
514         memcpy(hdr, icmp6h, sizeof(*hdr));
515
516         opt = skb_transport_header(skb) + sizeof(struct icmp6hdr);
517         if (target) {
518                 ipv6_addr_copy((struct in6_addr *)opt, target);
519                 opt += sizeof(*target);
520         }
521
522         if (llinfo)
523                 ndisc_fill_addr_option(opt, llinfo, dev->dev_addr,
524                                        dev->addr_len, dev->type);
525
526         hdr->icmp6_cksum = csum_ipv6_magic(saddr, daddr, len,
527                                            IPPROTO_ICMPV6,
528                                            csum_partial((__u8 *) hdr,
529                                                         len, 0));
530
531         skb->dst = dst;
532
533         idev = in6_dev_get(dst->dev);
534         IP6_INC_STATS(idev, IPSTATS_MIB_OUTREQUESTS);
535
536         err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
537                       dst_output);
538         if (!err) {
539                 ICMP6MSGOUT_INC_STATS(idev, type);
540                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
541         }
542
543         if (likely(idev != NULL))
544                 in6_dev_put(idev);
545 }
546
547 static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
548                    struct in6_addr *daddr, struct in6_addr *solicited_addr,
549                    int router, int solicited, int override, int inc_opt)
550 {
551         struct in6_addr tmpaddr;
552         struct inet6_ifaddr *ifp;
553         struct in6_addr *src_addr;
554         struct icmp6hdr icmp6h = {
555                 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
556         };
557
558         /* for anycast or proxy, solicited_addr != src_addr */
559         ifp = ipv6_get_ifaddr(&init_net, solicited_addr, dev, 1);
560         if (ifp) {
561                 src_addr = solicited_addr;
562                 if (ifp->flags & IFA_F_OPTIMISTIC)
563                         override = 0;
564                 in6_ifa_put(ifp);
565         } else {
566                 if (ipv6_dev_get_saddr(dev, daddr, &tmpaddr))
567                         return;
568                 src_addr = &tmpaddr;
569         }
570
571         icmp6h.icmp6_router = router;
572         icmp6h.icmp6_solicited = solicited;
573         icmp6h.icmp6_override = override;
574
575         __ndisc_send(dev, neigh, daddr, src_addr,
576                      &icmp6h, solicited_addr,
577                      inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
578 }
579
580 void ndisc_send_ns(struct net_device *dev, struct neighbour *neigh,
581                    struct in6_addr *solicit,
582                    struct in6_addr *daddr, struct in6_addr *saddr)
583 {
584         struct in6_addr addr_buf;
585         struct icmp6hdr icmp6h = {
586                 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
587         };
588
589         if (saddr == NULL) {
590                 if (ipv6_get_lladdr(dev, &addr_buf,
591                                    (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
592                         return;
593                 saddr = &addr_buf;
594         }
595
596         __ndisc_send(dev, neigh, daddr, saddr,
597                      &icmp6h, solicit,
598                      !ipv6_addr_any(saddr) ? ND_OPT_SOURCE_LL_ADDR : 0);
599 }
600
601 void ndisc_send_rs(struct net_device *dev, struct in6_addr *saddr,
602                    struct in6_addr *daddr)
603 {
604         struct icmp6hdr icmp6h = {
605                 .icmp6_type = NDISC_ROUTER_SOLICITATION,
606         };
607         int send_sllao = dev->addr_len;
608
609 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
610         /*
611          * According to section 2.2 of RFC 4429, we must not
612          * send router solicitations with a sllao from
613          * optimistic addresses, but we may send the solicitation
614          * if we don't include the sllao.  So here we check
615          * if our address is optimistic, and if so, we
616          * suppress the inclusion of the sllao.
617          */
618         if (send_sllao) {
619                 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(&init_net, saddr,
620                                                            dev, 1);
621                 if (ifp) {
622                         if (ifp->flags & IFA_F_OPTIMISTIC)  {
623                                 send_sllao = 0;
624                         }
625                         in6_ifa_put(ifp);
626                 } else {
627                         send_sllao = 0;
628                 }
629         }
630 #endif
631         __ndisc_send(dev, NULL, daddr, saddr,
632                      &icmp6h, NULL,
633                      send_sllao ? ND_OPT_SOURCE_LL_ADDR : 0);
634 }
635
636
637 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
638 {
639         /*
640          *      "The sender MUST return an ICMP
641          *       destination unreachable"
642          */
643         dst_link_failure(skb);
644         kfree_skb(skb);
645 }
646
647 /* Called with locked neigh: either read or both */
648
649 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
650 {
651         struct in6_addr *saddr = NULL;
652         struct in6_addr mcaddr;
653         struct net_device *dev = neigh->dev;
654         struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
655         int probes = atomic_read(&neigh->probes);
656
657         if (skb && ipv6_chk_addr(&init_net, &ipv6_hdr(skb)->saddr, dev, 1))
658                 saddr = &ipv6_hdr(skb)->saddr;
659
660         if ((probes -= neigh->parms->ucast_probes) < 0) {
661                 if (!(neigh->nud_state & NUD_VALID)) {
662                         ND_PRINTK1(KERN_DEBUG
663                                    "%s(): trying to ucast probe in NUD_INVALID: "
664                                    NIP6_FMT "\n",
665                                    __FUNCTION__,
666                                    NIP6(*target));
667                 }
668                 ndisc_send_ns(dev, neigh, target, target, saddr);
669         } else if ((probes -= neigh->parms->app_probes) < 0) {
670 #ifdef CONFIG_ARPD
671                 neigh_app_ns(neigh);
672 #endif
673         } else {
674                 addrconf_addr_solict_mult(target, &mcaddr);
675                 ndisc_send_ns(dev, NULL, target, &mcaddr, saddr);
676         }
677 }
678
679 static struct pneigh_entry *pndisc_check_router(struct net_device *dev,
680                 struct in6_addr *addr, int *is_router)
681 {
682         struct pneigh_entry *n;
683
684         read_lock_bh(&nd_tbl.lock);
685         n = __pneigh_lookup(&nd_tbl, &init_net, addr, dev);
686         if (n != NULL)
687                 *is_router = (n->flags & NTF_ROUTER);
688         read_unlock_bh(&nd_tbl.lock);
689
690         return n;
691 }
692
693 static void ndisc_recv_ns(struct sk_buff *skb)
694 {
695         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
696         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
697         struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
698         u8 *lladdr = NULL;
699         u32 ndoptlen = skb->tail - (skb->transport_header +
700                                     offsetof(struct nd_msg, opt));
701         struct ndisc_options ndopts;
702         struct net_device *dev = skb->dev;
703         struct inet6_ifaddr *ifp;
704         struct inet6_dev *idev = NULL;
705         struct neighbour *neigh;
706         struct pneigh_entry *pneigh = NULL;
707         int dad = ipv6_addr_any(saddr);
708         int inc;
709         int is_router = 0;
710
711         if (ipv6_addr_is_multicast(&msg->target)) {
712                 ND_PRINTK2(KERN_WARNING
713                            "ICMPv6 NS: multicast target address");
714                 return;
715         }
716
717         /*
718          * RFC2461 7.1.1:
719          * DAD has to be destined for solicited node multicast address.
720          */
721         if (dad &&
722             !(daddr->s6_addr32[0] == htonl(0xff020000) &&
723               daddr->s6_addr32[1] == htonl(0x00000000) &&
724               daddr->s6_addr32[2] == htonl(0x00000001) &&
725               daddr->s6_addr [12] == 0xff )) {
726                 ND_PRINTK2(KERN_WARNING
727                            "ICMPv6 NS: bad DAD packet (wrong destination)\n");
728                 return;
729         }
730
731         if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
732                 ND_PRINTK2(KERN_WARNING
733                            "ICMPv6 NS: invalid ND options\n");
734                 return;
735         }
736
737         if (ndopts.nd_opts_src_lladdr) {
738                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
739                 if (!lladdr) {
740                         ND_PRINTK2(KERN_WARNING
741                                    "ICMPv6 NS: invalid link-layer address length\n");
742                         return;
743                 }
744
745                 /* RFC2461 7.1.1:
746                  *      If the IP source address is the unspecified address,
747                  *      there MUST NOT be source link-layer address option
748                  *      in the message.
749                  */
750                 if (dad) {
751                         ND_PRINTK2(KERN_WARNING
752                                    "ICMPv6 NS: bad DAD packet (link-layer address option)\n");
753                         return;
754                 }
755         }
756
757         inc = ipv6_addr_is_multicast(daddr);
758
759         if ((ifp = ipv6_get_ifaddr(&init_net, &msg->target, dev, 1)) != NULL) {
760
761                 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
762                         if (dad) {
763                                 if (dev->type == ARPHRD_IEEE802_TR) {
764                                         const unsigned char *sadr;
765                                         sadr = skb_mac_header(skb);
766                                         if (((sadr[8] ^ dev->dev_addr[0]) & 0x7f) == 0 &&
767                                             sadr[9] == dev->dev_addr[1] &&
768                                             sadr[10] == dev->dev_addr[2] &&
769                                             sadr[11] == dev->dev_addr[3] &&
770                                             sadr[12] == dev->dev_addr[4] &&
771                                             sadr[13] == dev->dev_addr[5]) {
772                                                 /* looped-back to us */
773                                                 goto out;
774                                         }
775                                 }
776
777                                 /*
778                                  * We are colliding with another node
779                                  * who is doing DAD
780                                  * so fail our DAD process
781                                  */
782                                 addrconf_dad_failure(ifp);
783                                 return;
784                         } else {
785                                 /*
786                                  * This is not a dad solicitation.
787                                  * If we are an optimistic node,
788                                  * we should respond.
789                                  * Otherwise, we should ignore it.
790                                  */
791                                 if (!(ifp->flags & IFA_F_OPTIMISTIC))
792                                         goto out;
793                         }
794                 }
795
796                 idev = ifp->idev;
797         } else {
798                 idev = in6_dev_get(dev);
799                 if (!idev) {
800                         /* XXX: count this drop? */
801                         return;
802                 }
803
804                 if (ipv6_chk_acast_addr(dev, &msg->target) ||
805                     (idev->cnf.forwarding &&
806                      (ipv6_devconf.proxy_ndp || idev->cnf.proxy_ndp) &&
807                      (pneigh = pndisc_check_router(dev, &msg->target,
808                                                   &is_router)) != NULL)) {
809                         if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
810                             skb->pkt_type != PACKET_HOST &&
811                             inc != 0 &&
812                             idev->nd_parms->proxy_delay != 0) {
813                                 /*
814                                  * for anycast or proxy,
815                                  * sender should delay its response
816                                  * by a random time between 0 and
817                                  * MAX_ANYCAST_DELAY_TIME seconds.
818                                  * (RFC2461) -- yoshfuji
819                                  */
820                                 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
821                                 if (n)
822                                         pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
823                                 goto out;
824                         }
825                 } else
826                         goto out;
827         }
828
829         is_router = !!(pneigh ? is_router : idev->cnf.forwarding);
830
831         if (dad) {
832                 struct in6_addr maddr;
833
834                 ipv6_addr_all_nodes(&maddr);
835                 ndisc_send_na(dev, NULL, &maddr, &msg->target,
836                               is_router, 0, (ifp != NULL), 1);
837                 goto out;
838         }
839
840         if (inc)
841                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
842         else
843                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
844
845         /*
846          *      update / create cache entry
847          *      for the source address
848          */
849         neigh = __neigh_lookup(&nd_tbl, saddr, dev,
850                                !inc || lladdr || !dev->addr_len);
851         if (neigh)
852                 neigh_update(neigh, lladdr, NUD_STALE,
853                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
854                              NEIGH_UPDATE_F_OVERRIDE);
855         if (neigh || !dev->header_ops) {
856                 ndisc_send_na(dev, neigh, saddr, &msg->target,
857                               is_router,
858                               1, (ifp != NULL && inc), inc);
859                 if (neigh)
860                         neigh_release(neigh);
861         }
862
863 out:
864         if (ifp)
865                 in6_ifa_put(ifp);
866         else
867                 in6_dev_put(idev);
868
869         return;
870 }
871
872 static void ndisc_recv_na(struct sk_buff *skb)
873 {
874         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
875         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
876         struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
877         u8 *lladdr = NULL;
878         u32 ndoptlen = skb->tail - (skb->transport_header +
879                                     offsetof(struct nd_msg, opt));
880         struct ndisc_options ndopts;
881         struct net_device *dev = skb->dev;
882         struct inet6_ifaddr *ifp;
883         struct neighbour *neigh;
884
885         if (skb->len < sizeof(struct nd_msg)) {
886                 ND_PRINTK2(KERN_WARNING
887                            "ICMPv6 NA: packet too short\n");
888                 return;
889         }
890
891         if (ipv6_addr_is_multicast(&msg->target)) {
892                 ND_PRINTK2(KERN_WARNING
893                            "ICMPv6 NA: target address is multicast.\n");
894                 return;
895         }
896
897         if (ipv6_addr_is_multicast(daddr) &&
898             msg->icmph.icmp6_solicited) {
899                 ND_PRINTK2(KERN_WARNING
900                            "ICMPv6 NA: solicited NA is multicasted.\n");
901                 return;
902         }
903
904         if (!ndisc_parse_options(msg->opt, ndoptlen, &ndopts)) {
905                 ND_PRINTK2(KERN_WARNING
906                            "ICMPv6 NS: invalid ND option\n");
907                 return;
908         }
909         if (ndopts.nd_opts_tgt_lladdr) {
910                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
911                 if (!lladdr) {
912                         ND_PRINTK2(KERN_WARNING
913                                    "ICMPv6 NA: invalid link-layer address length\n");
914                         return;
915                 }
916         }
917         if ((ifp = ipv6_get_ifaddr(&init_net, &msg->target, dev, 1))) {
918                 if (ifp->flags & IFA_F_TENTATIVE) {
919                         addrconf_dad_failure(ifp);
920                         return;
921                 }
922                 /* What should we make now? The advertisement
923                    is invalid, but ndisc specs say nothing
924                    about it. It could be misconfiguration, or
925                    an smart proxy agent tries to help us :-)
926                  */
927                 ND_PRINTK1(KERN_WARNING
928                            "ICMPv6 NA: someone advertises our address on %s!\n",
929                            ifp->idev->dev->name);
930                 in6_ifa_put(ifp);
931                 return;
932         }
933         neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
934
935         if (neigh) {
936                 u8 old_flags = neigh->flags;
937
938                 if (neigh->nud_state & NUD_FAILED)
939                         goto out;
940
941                 /*
942                  * Don't update the neighbor cache entry on a proxy NA from
943                  * ourselves because either the proxied node is off link or it
944                  * has already sent a NA to us.
945                  */
946                 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
947                     ipv6_devconf.forwarding && ipv6_devconf.proxy_ndp &&
948                     pneigh_lookup(&nd_tbl, &init_net, &msg->target, dev, 0)) {
949                         /* XXX: idev->cnf.prixy_ndp */
950                         goto out;
951                 }
952
953                 neigh_update(neigh, lladdr,
954                              msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
955                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
956                              (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
957                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
958                              (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0));
959
960                 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
961                         /*
962                          * Change: router to host
963                          */
964                         struct rt6_info *rt;
965                         rt = rt6_get_dflt_router(saddr, dev);
966                         if (rt)
967                                 ip6_del_rt(rt);
968                 }
969
970 out:
971                 neigh_release(neigh);
972         }
973 }
974
975 static void ndisc_recv_rs(struct sk_buff *skb)
976 {
977         struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
978         unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
979         struct neighbour *neigh;
980         struct inet6_dev *idev;
981         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
982         struct ndisc_options ndopts;
983         u8 *lladdr = NULL;
984
985         if (skb->len < sizeof(*rs_msg))
986                 return;
987
988         idev = in6_dev_get(skb->dev);
989         if (!idev) {
990                 if (net_ratelimit())
991                         ND_PRINTK1("ICMP6 RS: can't find in6 device\n");
992                 return;
993         }
994
995         /* Don't accept RS if we're not in router mode */
996         if (!idev->cnf.forwarding)
997                 goto out;
998
999         /*
1000          * Don't update NCE if src = ::;
1001          * this implies that the source node has no ip address assigned yet.
1002          */
1003         if (ipv6_addr_any(saddr))
1004                 goto out;
1005
1006         /* Parse ND options */
1007         if (!ndisc_parse_options(rs_msg->opt, ndoptlen, &ndopts)) {
1008                 if (net_ratelimit())
1009                         ND_PRINTK2("ICMP6 NS: invalid ND option, ignored\n");
1010                 goto out;
1011         }
1012
1013         if (ndopts.nd_opts_src_lladdr) {
1014                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1015                                              skb->dev);
1016                 if (!lladdr)
1017                         goto out;
1018         }
1019
1020         neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1021         if (neigh) {
1022                 neigh_update(neigh, lladdr, NUD_STALE,
1023                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1024                              NEIGH_UPDATE_F_OVERRIDE|
1025                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1026                 neigh_release(neigh);
1027         }
1028 out:
1029         in6_dev_put(idev);
1030 }
1031
1032 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1033 {
1034         struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1035         struct sk_buff *skb;
1036         struct nlmsghdr *nlh;
1037         struct nduseroptmsg *ndmsg;
1038         int err;
1039         int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1040                                     + (opt->nd_opt_len << 3));
1041         size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1042
1043         skb = nlmsg_new(msg_size, GFP_ATOMIC);
1044         if (skb == NULL) {
1045                 err = -ENOBUFS;
1046                 goto errout;
1047         }
1048
1049         nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1050         if (nlh == NULL) {
1051                 goto nla_put_failure;
1052         }
1053
1054         ndmsg = nlmsg_data(nlh);
1055         ndmsg->nduseropt_family = AF_INET6;
1056         ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1057         ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1058         ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1059         ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1060
1061         memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1062
1063         NLA_PUT(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
1064                 &ipv6_hdr(ra)->saddr);
1065         nlmsg_end(skb, nlh);
1066
1067         err = rtnl_notify(skb, &init_net, 0, RTNLGRP_ND_USEROPT, NULL,
1068                           GFP_ATOMIC);
1069         if (err < 0)
1070                 goto errout;
1071
1072         return;
1073
1074 nla_put_failure:
1075         nlmsg_free(skb);
1076         err = -EMSGSIZE;
1077 errout:
1078         rtnl_set_sk_err(&init_net, RTNLGRP_ND_USEROPT, err);
1079 }
1080
1081 static void ndisc_router_discovery(struct sk_buff *skb)
1082 {
1083         struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1084         struct neighbour *neigh = NULL;
1085         struct inet6_dev *in6_dev;
1086         struct rt6_info *rt = NULL;
1087         int lifetime;
1088         struct ndisc_options ndopts;
1089         int optlen;
1090         unsigned int pref = 0;
1091
1092         __u8 * opt = (__u8 *)(ra_msg + 1);
1093
1094         optlen = (skb->tail - skb->transport_header) - sizeof(struct ra_msg);
1095
1096         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1097                 ND_PRINTK2(KERN_WARNING
1098                            "ICMPv6 RA: source address is not link-local.\n");
1099                 return;
1100         }
1101         if (optlen < 0) {
1102                 ND_PRINTK2(KERN_WARNING
1103                            "ICMPv6 RA: packet too short\n");
1104                 return;
1105         }
1106
1107         /*
1108          *      set the RA_RECV flag in the interface
1109          */
1110
1111         in6_dev = in6_dev_get(skb->dev);
1112         if (in6_dev == NULL) {
1113                 ND_PRINTK0(KERN_ERR
1114                            "ICMPv6 RA: can't find inet6 device for %s.\n",
1115                            skb->dev->name);
1116                 return;
1117         }
1118         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_ra) {
1119                 in6_dev_put(in6_dev);
1120                 return;
1121         }
1122
1123         if (!ndisc_parse_options(opt, optlen, &ndopts)) {
1124                 in6_dev_put(in6_dev);
1125                 ND_PRINTK2(KERN_WARNING
1126                            "ICMP6 RA: invalid ND options\n");
1127                 return;
1128         }
1129
1130         if (in6_dev->if_flags & IF_RS_SENT) {
1131                 /*
1132                  *      flag that an RA was received after an RS was sent
1133                  *      out on this interface.
1134                  */
1135                 in6_dev->if_flags |= IF_RA_RCVD;
1136         }
1137
1138         /*
1139          * Remember the managed/otherconf flags from most recently
1140          * received RA message (RFC 2462) -- yoshfuji
1141          */
1142         in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1143                                 IF_RA_OTHERCONF)) |
1144                                 (ra_msg->icmph.icmp6_addrconf_managed ?
1145                                         IF_RA_MANAGED : 0) |
1146                                 (ra_msg->icmph.icmp6_addrconf_other ?
1147                                         IF_RA_OTHERCONF : 0);
1148
1149         if (!in6_dev->cnf.accept_ra_defrtr)
1150                 goto skip_defrtr;
1151
1152         lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1153
1154 #ifdef CONFIG_IPV6_ROUTER_PREF
1155         pref = ra_msg->icmph.icmp6_router_pref;
1156         /* 10b is handled as if it were 00b (medium) */
1157         if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1158             !in6_dev->cnf.accept_ra_rtr_pref)
1159                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1160 #endif
1161
1162         rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1163
1164         if (rt)
1165                 neigh = rt->rt6i_nexthop;
1166
1167         if (rt && lifetime == 0) {
1168                 neigh_clone(neigh);
1169                 ip6_del_rt(rt);
1170                 rt = NULL;
1171         }
1172
1173         if (rt == NULL && lifetime) {
1174                 ND_PRINTK3(KERN_DEBUG
1175                            "ICMPv6 RA: adding default router.\n");
1176
1177                 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1178                 if (rt == NULL) {
1179                         ND_PRINTK0(KERN_ERR
1180                                    "ICMPv6 RA: %s() failed to add default route.\n",
1181                                    __FUNCTION__);
1182                         in6_dev_put(in6_dev);
1183                         return;
1184                 }
1185
1186                 neigh = rt->rt6i_nexthop;
1187                 if (neigh == NULL) {
1188                         ND_PRINTK0(KERN_ERR
1189                                    "ICMPv6 RA: %s() got default router without neighbour.\n",
1190                                    __FUNCTION__);
1191                         dst_release(&rt->u.dst);
1192                         in6_dev_put(in6_dev);
1193                         return;
1194                 }
1195                 neigh->flags |= NTF_ROUTER;
1196         } else if (rt) {
1197                 rt->rt6i_flags |= (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1198         }
1199
1200         if (rt)
1201                 rt->rt6i_expires = jiffies + (HZ * lifetime);
1202
1203         if (ra_msg->icmph.icmp6_hop_limit) {
1204                 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1205                 if (rt)
1206                         rt->u.dst.metrics[RTAX_HOPLIMIT-1] = ra_msg->icmph.icmp6_hop_limit;
1207         }
1208
1209 skip_defrtr:
1210
1211         /*
1212          *      Update Reachable Time and Retrans Timer
1213          */
1214
1215         if (in6_dev->nd_parms) {
1216                 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1217
1218                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1219                         rtime = (rtime*HZ)/1000;
1220                         if (rtime < HZ/10)
1221                                 rtime = HZ/10;
1222                         in6_dev->nd_parms->retrans_time = rtime;
1223                         in6_dev->tstamp = jiffies;
1224                         inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1225                 }
1226
1227                 rtime = ntohl(ra_msg->reachable_time);
1228                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1229                         rtime = (rtime*HZ)/1000;
1230
1231                         if (rtime < HZ/10)
1232                                 rtime = HZ/10;
1233
1234                         if (rtime != in6_dev->nd_parms->base_reachable_time) {
1235                                 in6_dev->nd_parms->base_reachable_time = rtime;
1236                                 in6_dev->nd_parms->gc_staletime = 3 * rtime;
1237                                 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1238                                 in6_dev->tstamp = jiffies;
1239                                 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1240                         }
1241                 }
1242         }
1243
1244         /*
1245          *      Process options.
1246          */
1247
1248         if (!neigh)
1249                 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1250                                        skb->dev, 1);
1251         if (neigh) {
1252                 u8 *lladdr = NULL;
1253                 if (ndopts.nd_opts_src_lladdr) {
1254                         lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1255                                                      skb->dev);
1256                         if (!lladdr) {
1257                                 ND_PRINTK2(KERN_WARNING
1258                                            "ICMPv6 RA: invalid link-layer address length\n");
1259                                 goto out;
1260                         }
1261                 }
1262                 neigh_update(neigh, lladdr, NUD_STALE,
1263                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1264                              NEIGH_UPDATE_F_OVERRIDE|
1265                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1266                              NEIGH_UPDATE_F_ISROUTER);
1267         }
1268
1269 #ifdef CONFIG_IPV6_ROUTE_INFO
1270         if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1271                 struct nd_opt_hdr *p;
1272                 for (p = ndopts.nd_opts_ri;
1273                      p;
1274                      p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1275                         if (((struct route_info *)p)->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1276                                 continue;
1277                         rt6_route_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3,
1278                                       &ipv6_hdr(skb)->saddr);
1279                 }
1280         }
1281 #endif
1282
1283         if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1284                 struct nd_opt_hdr *p;
1285                 for (p = ndopts.nd_opts_pi;
1286                      p;
1287                      p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1288                         addrconf_prefix_rcv(skb->dev, (u8*)p, (p->nd_opt_len) << 3);
1289                 }
1290         }
1291
1292         if (ndopts.nd_opts_mtu) {
1293                 __be32 n;
1294                 u32 mtu;
1295
1296                 memcpy(&n, ((u8*)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1297                 mtu = ntohl(n);
1298
1299                 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1300                         ND_PRINTK2(KERN_WARNING
1301                                    "ICMPv6 RA: invalid mtu: %d\n",
1302                                    mtu);
1303                 } else if (in6_dev->cnf.mtu6 != mtu) {
1304                         in6_dev->cnf.mtu6 = mtu;
1305
1306                         if (rt)
1307                                 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
1308
1309                         rt6_mtu_change(skb->dev, mtu);
1310                 }
1311         }
1312
1313         if (ndopts.nd_useropts) {
1314                 struct nd_opt_hdr *p;
1315                 for (p = ndopts.nd_useropts;
1316                      p;
1317                      p = ndisc_next_useropt(p, ndopts.nd_useropts_end)) {
1318                         ndisc_ra_useropt(skb, p);
1319                 }
1320         }
1321
1322         if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1323                 ND_PRINTK2(KERN_WARNING
1324                            "ICMPv6 RA: invalid RA options");
1325         }
1326 out:
1327         if (rt)
1328                 dst_release(&rt->u.dst);
1329         else if (neigh)
1330                 neigh_release(neigh);
1331         in6_dev_put(in6_dev);
1332 }
1333
1334 static void ndisc_redirect_rcv(struct sk_buff *skb)
1335 {
1336         struct inet6_dev *in6_dev;
1337         struct icmp6hdr *icmph;
1338         struct in6_addr *dest;
1339         struct in6_addr *target;        /* new first hop to destination */
1340         struct neighbour *neigh;
1341         int on_link = 0;
1342         struct ndisc_options ndopts;
1343         int optlen;
1344         u8 *lladdr = NULL;
1345
1346         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1347                 ND_PRINTK2(KERN_WARNING
1348                            "ICMPv6 Redirect: source address is not link-local.\n");
1349                 return;
1350         }
1351
1352         optlen = skb->tail - skb->transport_header;
1353         optlen -= sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1354
1355         if (optlen < 0) {
1356                 ND_PRINTK2(KERN_WARNING
1357                            "ICMPv6 Redirect: packet too short\n");
1358                 return;
1359         }
1360
1361         icmph = icmp6_hdr(skb);
1362         target = (struct in6_addr *) (icmph + 1);
1363         dest = target + 1;
1364
1365         if (ipv6_addr_is_multicast(dest)) {
1366                 ND_PRINTK2(KERN_WARNING
1367                            "ICMPv6 Redirect: destination address is multicast.\n");
1368                 return;
1369         }
1370
1371         if (ipv6_addr_equal(dest, target)) {
1372                 on_link = 1;
1373         } else if (ipv6_addr_type(target) !=
1374                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1375                 ND_PRINTK2(KERN_WARNING
1376                            "ICMPv6 Redirect: target address is not link-local unicast.\n");
1377                 return;
1378         }
1379
1380         in6_dev = in6_dev_get(skb->dev);
1381         if (!in6_dev)
1382                 return;
1383         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects) {
1384                 in6_dev_put(in6_dev);
1385                 return;
1386         }
1387
1388         /* RFC2461 8.1:
1389          *      The IP source address of the Redirect MUST be the same as the current
1390          *      first-hop router for the specified ICMP Destination Address.
1391          */
1392
1393         if (!ndisc_parse_options((u8*)(dest + 1), optlen, &ndopts)) {
1394                 ND_PRINTK2(KERN_WARNING
1395                            "ICMPv6 Redirect: invalid ND options\n");
1396                 in6_dev_put(in6_dev);
1397                 return;
1398         }
1399         if (ndopts.nd_opts_tgt_lladdr) {
1400                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1401                                              skb->dev);
1402                 if (!lladdr) {
1403                         ND_PRINTK2(KERN_WARNING
1404                                    "ICMPv6 Redirect: invalid link-layer address length\n");
1405                         in6_dev_put(in6_dev);
1406                         return;
1407                 }
1408         }
1409
1410         neigh = __neigh_lookup(&nd_tbl, target, skb->dev, 1);
1411         if (neigh) {
1412                 rt6_redirect(dest, &ipv6_hdr(skb)->daddr,
1413                              &ipv6_hdr(skb)->saddr, neigh, lladdr,
1414                              on_link);
1415                 neigh_release(neigh);
1416         }
1417         in6_dev_put(in6_dev);
1418 }
1419
1420 void ndisc_send_redirect(struct sk_buff *skb, struct neighbour *neigh,
1421                          struct in6_addr *target)
1422 {
1423         struct sock *sk = ndisc_socket->sk;
1424         int len = sizeof(struct icmp6hdr) + 2 * sizeof(struct in6_addr);
1425         struct sk_buff *buff;
1426         struct icmp6hdr *icmph;
1427         struct in6_addr saddr_buf;
1428         struct in6_addr *addrp;
1429         struct net_device *dev;
1430         struct rt6_info *rt;
1431         struct dst_entry *dst;
1432         struct inet6_dev *idev;
1433         struct flowi fl;
1434         u8 *opt;
1435         int rd_len;
1436         int err;
1437         u8 ha_buf[MAX_ADDR_LEN], *ha = NULL;
1438
1439         dev = skb->dev;
1440
1441         if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1442                 ND_PRINTK2(KERN_WARNING
1443                            "ICMPv6 Redirect: no link-local address on %s\n",
1444                            dev->name);
1445                 return;
1446         }
1447
1448         if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1449             ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1450                 ND_PRINTK2(KERN_WARNING
1451                         "ICMPv6 Redirect: target address is not link-local unicast.\n");
1452                 return;
1453         }
1454
1455         ndisc_flow_init(&fl, NDISC_REDIRECT, &saddr_buf, &ipv6_hdr(skb)->saddr,
1456                         dev->ifindex);
1457
1458         dst = ip6_route_output(NULL, &fl);
1459         if (dst == NULL)
1460                 return;
1461
1462         err = xfrm_lookup(&dst, &fl, NULL, 0);
1463         if (err)
1464                 return;
1465
1466         rt = (struct rt6_info *) dst;
1467
1468         if (rt->rt6i_flags & RTF_GATEWAY) {
1469                 ND_PRINTK2(KERN_WARNING
1470                            "ICMPv6 Redirect: destination is not a neighbour.\n");
1471                 dst_release(dst);
1472                 return;
1473         }
1474         if (!xrlim_allow(dst, 1*HZ)) {
1475                 dst_release(dst);
1476                 return;
1477         }
1478
1479         if (dev->addr_len) {
1480                 read_lock_bh(&neigh->lock);
1481                 if (neigh->nud_state & NUD_VALID) {
1482                         memcpy(ha_buf, neigh->ha, dev->addr_len);
1483                         read_unlock_bh(&neigh->lock);
1484                         ha = ha_buf;
1485                         len += ndisc_opt_addr_space(dev);
1486                 } else
1487                         read_unlock_bh(&neigh->lock);
1488         }
1489
1490         rd_len = min_t(unsigned int,
1491                      IPV6_MIN_MTU-sizeof(struct ipv6hdr)-len, skb->len + 8);
1492         rd_len &= ~0x7;
1493         len += rd_len;
1494
1495         buff = sock_alloc_send_skb(sk,
1496                                    (MAX_HEADER + sizeof(struct ipv6hdr) +
1497                                     len + LL_RESERVED_SPACE(dev)),
1498                                    1, &err);
1499         if (buff == NULL) {
1500                 ND_PRINTK0(KERN_ERR
1501                            "ICMPv6 Redirect: %s() failed to allocate an skb.\n",
1502                            __FUNCTION__);
1503                 dst_release(dst);
1504                 return;
1505         }
1506
1507
1508         skb_reserve(buff, LL_RESERVED_SPACE(dev));
1509         ip6_nd_hdr(sk, buff, dev, &saddr_buf, &ipv6_hdr(skb)->saddr,
1510                    IPPROTO_ICMPV6, len);
1511
1512         skb_set_transport_header(buff, skb_tail_pointer(buff) - buff->data);
1513         skb_put(buff, len);
1514         icmph = icmp6_hdr(buff);
1515
1516         memset(icmph, 0, sizeof(struct icmp6hdr));
1517         icmph->icmp6_type = NDISC_REDIRECT;
1518
1519         /*
1520          *      copy target and destination addresses
1521          */
1522
1523         addrp = (struct in6_addr *)(icmph + 1);
1524         ipv6_addr_copy(addrp, target);
1525         addrp++;
1526         ipv6_addr_copy(addrp, &ipv6_hdr(skb)->daddr);
1527
1528         opt = (u8*) (addrp + 1);
1529
1530         /*
1531          *      include target_address option
1532          */
1533
1534         if (ha)
1535                 opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, ha,
1536                                              dev->addr_len, dev->type);
1537
1538         /*
1539          *      build redirect option and copy skb over to the new packet.
1540          */
1541
1542         memset(opt, 0, 8);
1543         *(opt++) = ND_OPT_REDIRECT_HDR;
1544         *(opt++) = (rd_len >> 3);
1545         opt += 6;
1546
1547         memcpy(opt, ipv6_hdr(skb), rd_len - 8);
1548
1549         icmph->icmp6_cksum = csum_ipv6_magic(&saddr_buf, &ipv6_hdr(skb)->saddr,
1550                                              len, IPPROTO_ICMPV6,
1551                                              csum_partial((u8 *) icmph, len, 0));
1552
1553         buff->dst = dst;
1554         idev = in6_dev_get(dst->dev);
1555         IP6_INC_STATS(idev, IPSTATS_MIB_OUTREQUESTS);
1556         err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, buff, NULL, dst->dev,
1557                       dst_output);
1558         if (!err) {
1559                 ICMP6MSGOUT_INC_STATS(idev, NDISC_REDIRECT);
1560                 ICMP6_INC_STATS(idev, ICMP6_MIB_OUTMSGS);
1561         }
1562
1563         if (likely(idev != NULL))
1564                 in6_dev_put(idev);
1565 }
1566
1567 static void pndisc_redo(struct sk_buff *skb)
1568 {
1569         ndisc_recv_ns(skb);
1570         kfree_skb(skb);
1571 }
1572
1573 int ndisc_rcv(struct sk_buff *skb)
1574 {
1575         struct nd_msg *msg;
1576
1577         if (!pskb_may_pull(skb, skb->len))
1578                 return 0;
1579
1580         msg = (struct nd_msg *)skb_transport_header(skb);
1581
1582         __skb_push(skb, skb->data - skb_transport_header(skb));
1583
1584         if (ipv6_hdr(skb)->hop_limit != 255) {
1585                 ND_PRINTK2(KERN_WARNING
1586                            "ICMPv6 NDISC: invalid hop-limit: %d\n",
1587                            ipv6_hdr(skb)->hop_limit);
1588                 return 0;
1589         }
1590
1591         if (msg->icmph.icmp6_code != 0) {
1592                 ND_PRINTK2(KERN_WARNING
1593                            "ICMPv6 NDISC: invalid ICMPv6 code: %d\n",
1594                            msg->icmph.icmp6_code);
1595                 return 0;
1596         }
1597
1598         memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1599
1600         switch (msg->icmph.icmp6_type) {
1601         case NDISC_NEIGHBOUR_SOLICITATION:
1602                 ndisc_recv_ns(skb);
1603                 break;
1604
1605         case NDISC_NEIGHBOUR_ADVERTISEMENT:
1606                 ndisc_recv_na(skb);
1607                 break;
1608
1609         case NDISC_ROUTER_SOLICITATION:
1610                 ndisc_recv_rs(skb);
1611                 break;
1612
1613         case NDISC_ROUTER_ADVERTISEMENT:
1614                 ndisc_router_discovery(skb);
1615                 break;
1616
1617         case NDISC_REDIRECT:
1618                 ndisc_redirect_rcv(skb);
1619                 break;
1620         }
1621
1622         return 0;
1623 }
1624
1625 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1626 {
1627         struct net_device *dev = ptr;
1628
1629         if (dev->nd_net != &init_net)
1630                 return NOTIFY_DONE;
1631
1632         switch (event) {
1633         case NETDEV_CHANGEADDR:
1634                 neigh_changeaddr(&nd_tbl, dev);
1635                 fib6_run_gc(~0UL);
1636                 break;
1637         case NETDEV_DOWN:
1638                 neigh_ifdown(&nd_tbl, dev);
1639                 fib6_run_gc(~0UL);
1640                 break;
1641         default:
1642                 break;
1643         }
1644
1645         return NOTIFY_DONE;
1646 }
1647
1648 static struct notifier_block ndisc_netdev_notifier = {
1649         .notifier_call = ndisc_netdev_event,
1650 };
1651
1652 #ifdef CONFIG_SYSCTL
1653 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1654                                          const char *func, const char *dev_name)
1655 {
1656         static char warncomm[TASK_COMM_LEN];
1657         static int warned;
1658         if (strcmp(warncomm, current->comm) && warned < 5) {
1659                 strcpy(warncomm, current->comm);
1660                 printk(KERN_WARNING
1661                         "process `%s' is using deprecated sysctl (%s) "
1662                         "net.ipv6.neigh.%s.%s; "
1663                         "Use net.ipv6.neigh.%s.%s_ms "
1664                         "instead.\n",
1665                         warncomm, func,
1666                         dev_name, ctl->procname,
1667                         dev_name, ctl->procname);
1668                 warned++;
1669         }
1670 }
1671
1672 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, struct file * filp, void __user *buffer, size_t *lenp, loff_t *ppos)
1673 {
1674         struct net_device *dev = ctl->extra1;
1675         struct inet6_dev *idev;
1676         int ret;
1677
1678         if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1679             (strcmp(ctl->procname, "base_reachable_time") == 0))
1680                 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1681
1682         if (strcmp(ctl->procname, "retrans_time") == 0)
1683                 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1684
1685         else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1686                 ret = proc_dointvec_jiffies(ctl, write,
1687                                             filp, buffer, lenp, ppos);
1688
1689         else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1690                  (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1691                 ret = proc_dointvec_ms_jiffies(ctl, write,
1692                                                filp, buffer, lenp, ppos);
1693         else
1694                 ret = -1;
1695
1696         if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1697                 if (ctl->data == &idev->nd_parms->base_reachable_time)
1698                         idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1699                 idev->tstamp = jiffies;
1700                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1701                 in6_dev_put(idev);
1702         }
1703         return ret;
1704 }
1705
1706 static int ndisc_ifinfo_sysctl_strategy(ctl_table *ctl, int __user *name,
1707                                         int nlen, void __user *oldval,
1708                                         size_t __user *oldlenp,
1709                                         void __user *newval, size_t newlen)
1710 {
1711         struct net_device *dev = ctl->extra1;
1712         struct inet6_dev *idev;
1713         int ret;
1714
1715         if (ctl->ctl_name == NET_NEIGH_RETRANS_TIME ||
1716             ctl->ctl_name == NET_NEIGH_REACHABLE_TIME)
1717                 ndisc_warn_deprecated_sysctl(ctl, "procfs", dev ? dev->name : "default");
1718
1719         switch (ctl->ctl_name) {
1720         case NET_NEIGH_REACHABLE_TIME:
1721                 ret = sysctl_jiffies(ctl, name, nlen,
1722                                      oldval, oldlenp, newval, newlen);
1723                 break;
1724         case NET_NEIGH_RETRANS_TIME_MS:
1725         case NET_NEIGH_REACHABLE_TIME_MS:
1726                  ret = sysctl_ms_jiffies(ctl, name, nlen,
1727                                          oldval, oldlenp, newval, newlen);
1728                  break;
1729         default:
1730                 ret = 0;
1731         }
1732
1733         if (newval && newlen && ret > 0 &&
1734             dev && (idev = in6_dev_get(dev)) != NULL) {
1735                 if (ctl->ctl_name == NET_NEIGH_REACHABLE_TIME ||
1736                     ctl->ctl_name == NET_NEIGH_REACHABLE_TIME_MS)
1737                         idev->nd_parms->reachable_time = neigh_rand_reach_time(idev->nd_parms->base_reachable_time);
1738                 idev->tstamp = jiffies;
1739                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1740                 in6_dev_put(idev);
1741         }
1742
1743         return ret;
1744 }
1745
1746 #endif
1747
1748 int __init ndisc_init(struct net_proto_family *ops)
1749 {
1750         struct ipv6_pinfo *np;
1751         struct sock *sk;
1752         int err;
1753
1754         err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, &ndisc_socket);
1755         if (err < 0) {
1756                 ND_PRINTK0(KERN_ERR
1757                            "ICMPv6 NDISC: Failed to initialize the control socket (err %d).\n",
1758                            err);
1759                 ndisc_socket = NULL; /* For safety. */
1760                 return err;
1761         }
1762
1763         sk = ndisc_socket->sk;
1764         np = inet6_sk(sk);
1765         sk->sk_allocation = GFP_ATOMIC;
1766         np->hop_limit = 255;
1767         /* Do not loopback ndisc messages */
1768         np->mc_loop = 0;
1769         sk->sk_prot->unhash(sk);
1770
1771         /*
1772          * Initialize the neighbour table
1773          */
1774
1775         neigh_table_init(&nd_tbl);
1776
1777 #ifdef CONFIG_SYSCTL
1778         neigh_sysctl_register(NULL, &nd_tbl.parms, NET_IPV6, NET_IPV6_NEIGH,
1779                               "ipv6",
1780                               &ndisc_ifinfo_sysctl_change,
1781                               &ndisc_ifinfo_sysctl_strategy);
1782 #endif
1783
1784         register_netdevice_notifier(&ndisc_netdev_notifier);
1785         return 0;
1786 }
1787
1788 void ndisc_cleanup(void)
1789 {
1790         unregister_netdevice_notifier(&ndisc_netdev_notifier);
1791 #ifdef CONFIG_SYSCTL
1792         neigh_sysctl_unregister(&nd_tbl.parms);
1793 #endif
1794         neigh_table_clear(&nd_tbl);
1795         sock_release(ndisc_socket);
1796         ndisc_socket = NULL; /* For safety. */
1797 }