GNU Linux-libre 5.10.217-gnu1
[releases.git] / net / ipv6 / ndisc.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *      Neighbour Discovery for IPv6
4  *      Linux INET6 implementation
5  *
6  *      Authors:
7  *      Pedro Roque             <roque@di.fc.ul.pt>
8  *      Mike Shaver             <shaver@ingenia.com>
9  */
10
11 /*
12  *      Changes:
13  *
14  *      Alexey I. Froloff               :       RFC6106 (DNSSL) support
15  *      Pierre Ynard                    :       export userland ND options
16  *                                              through netlink (RDNSS support)
17  *      Lars Fenneberg                  :       fixed MTU setting on receipt
18  *                                              of an RA.
19  *      Janos Farkas                    :       kmalloc failure checks
20  *      Alexey Kuznetsov                :       state machine reworked
21  *                                              and moved to net/core.
22  *      Pekka Savola                    :       RFC2461 validation
23  *      YOSHIFUJI Hideaki @USAGI        :       Verify ND options properly
24  */
25
26 #define pr_fmt(fmt) "ICMPv6: " fmt
27
28 #include <linux/module.h>
29 #include <linux/errno.h>
30 #include <linux/types.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/sched.h>
34 #include <linux/net.h>
35 #include <linux/in6.h>
36 #include <linux/route.h>
37 #include <linux/init.h>
38 #include <linux/rcupdate.h>
39 #include <linux/slab.h>
40 #ifdef CONFIG_SYSCTL
41 #include <linux/sysctl.h>
42 #endif
43
44 #include <linux/if_addr.h>
45 #include <linux/if_ether.h>
46 #include <linux/if_arp.h>
47 #include <linux/ipv6.h>
48 #include <linux/icmpv6.h>
49 #include <linux/jhash.h>
50
51 #include <net/sock.h>
52 #include <net/snmp.h>
53
54 #include <net/ipv6.h>
55 #include <net/protocol.h>
56 #include <net/ndisc.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #include <net/icmp.h>
60
61 #include <net/netlink.h>
62 #include <linux/rtnetlink.h>
63
64 #include <net/flow.h>
65 #include <net/ip6_checksum.h>
66 #include <net/inet_common.h>
67 #include <linux/proc_fs.h>
68
69 #include <linux/netfilter.h>
70 #include <linux/netfilter_ipv6.h>
71
72 static u32 ndisc_hash(const void *pkey,
73                       const struct net_device *dev,
74                       __u32 *hash_rnd);
75 static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
76 static bool ndisc_allow_add(const struct net_device *dev,
77                             struct netlink_ext_ack *extack);
78 static int ndisc_constructor(struct neighbour *neigh);
79 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
80 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
81 static int pndisc_constructor(struct pneigh_entry *n);
82 static void pndisc_destructor(struct pneigh_entry *n);
83 static void pndisc_redo(struct sk_buff *skb);
84 static int ndisc_is_multicast(const void *pkey);
85
86 static const struct neigh_ops ndisc_generic_ops = {
87         .family =               AF_INET6,
88         .solicit =              ndisc_solicit,
89         .error_report =         ndisc_error_report,
90         .output =               neigh_resolve_output,
91         .connected_output =     neigh_connected_output,
92 };
93
94 static const struct neigh_ops ndisc_hh_ops = {
95         .family =               AF_INET6,
96         .solicit =              ndisc_solicit,
97         .error_report =         ndisc_error_report,
98         .output =               neigh_resolve_output,
99         .connected_output =     neigh_resolve_output,
100 };
101
102
103 static const struct neigh_ops ndisc_direct_ops = {
104         .family =               AF_INET6,
105         .output =               neigh_direct_output,
106         .connected_output =     neigh_direct_output,
107 };
108
109 struct neigh_table nd_tbl = {
110         .family =       AF_INET6,
111         .key_len =      sizeof(struct in6_addr),
112         .protocol =     cpu_to_be16(ETH_P_IPV6),
113         .hash =         ndisc_hash,
114         .key_eq =       ndisc_key_eq,
115         .constructor =  ndisc_constructor,
116         .pconstructor = pndisc_constructor,
117         .pdestructor =  pndisc_destructor,
118         .proxy_redo =   pndisc_redo,
119         .is_multicast = ndisc_is_multicast,
120         .allow_add  =   ndisc_allow_add,
121         .id =           "ndisc_cache",
122         .parms = {
123                 .tbl                    = &nd_tbl,
124                 .reachable_time         = ND_REACHABLE_TIME,
125                 .data = {
126                         [NEIGH_VAR_MCAST_PROBES] = 3,
127                         [NEIGH_VAR_UCAST_PROBES] = 3,
128                         [NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER,
129                         [NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME,
130                         [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
131                         [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
132                         [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
133                         [NEIGH_VAR_PROXY_QLEN] = 64,
134                         [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
135                         [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
136                 },
137         },
138         .gc_interval =    30 * HZ,
139         .gc_thresh1 =    128,
140         .gc_thresh2 =    512,
141         .gc_thresh3 =   1024,
142 };
143 EXPORT_SYMBOL_GPL(nd_tbl);
144
145 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, void *data,
146                               int data_len, int pad)
147 {
148         int space = __ndisc_opt_addr_space(data_len, pad);
149         u8 *opt = skb_put(skb, space);
150
151         opt[0] = type;
152         opt[1] = space>>3;
153
154         memset(opt + 2, 0, pad);
155         opt   += pad;
156         space -= pad;
157
158         memcpy(opt+2, data, data_len);
159         data_len += 2;
160         opt += data_len;
161         space -= data_len;
162         if (space > 0)
163                 memset(opt, 0, space);
164 }
165 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
166
167 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
168                                           void *data, u8 icmp6_type)
169 {
170         __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
171                                  ndisc_addr_option_pad(skb->dev->type));
172         ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
173 }
174
175 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
176                                                    void *ha,
177                                                    const u8 *ops_data)
178 {
179         ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
180         ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
181 }
182
183 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
184                                             struct nd_opt_hdr *end)
185 {
186         int type;
187         if (!cur || !end || cur >= end)
188                 return NULL;
189         type = cur->nd_opt_type;
190         do {
191                 cur = ((void *)cur) + (cur->nd_opt_len << 3);
192         } while (cur < end && cur->nd_opt_type != type);
193         return cur <= end && cur->nd_opt_type == type ? cur : NULL;
194 }
195
196 static inline int ndisc_is_useropt(const struct net_device *dev,
197                                    struct nd_opt_hdr *opt)
198 {
199         return opt->nd_opt_type == ND_OPT_PREFIX_INFO ||
200                 opt->nd_opt_type == ND_OPT_RDNSS ||
201                 opt->nd_opt_type == ND_OPT_DNSSL ||
202                 opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
203                 opt->nd_opt_type == ND_OPT_PREF64 ||
204                 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
205 }
206
207 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
208                                              struct nd_opt_hdr *cur,
209                                              struct nd_opt_hdr *end)
210 {
211         if (!cur || !end || cur >= end)
212                 return NULL;
213         do {
214                 cur = ((void *)cur) + (cur->nd_opt_len << 3);
215         } while (cur < end && !ndisc_is_useropt(dev, cur));
216         return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
217 }
218
219 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
220                                           u8 *opt, int opt_len,
221                                           struct ndisc_options *ndopts)
222 {
223         struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
224
225         if (!nd_opt || opt_len < 0 || !ndopts)
226                 return NULL;
227         memset(ndopts, 0, sizeof(*ndopts));
228         while (opt_len) {
229                 int l;
230                 if (opt_len < sizeof(struct nd_opt_hdr))
231                         return NULL;
232                 l = nd_opt->nd_opt_len << 3;
233                 if (opt_len < l || l == 0)
234                         return NULL;
235                 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
236                         goto next_opt;
237                 switch (nd_opt->nd_opt_type) {
238                 case ND_OPT_SOURCE_LL_ADDR:
239                 case ND_OPT_TARGET_LL_ADDR:
240                 case ND_OPT_MTU:
241                 case ND_OPT_NONCE:
242                 case ND_OPT_REDIRECT_HDR:
243                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
244                                 ND_PRINTK(2, warn,
245                                           "%s: duplicated ND6 option found: type=%d\n",
246                                           __func__, nd_opt->nd_opt_type);
247                         } else {
248                                 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
249                         }
250                         break;
251                 case ND_OPT_PREFIX_INFO:
252                         ndopts->nd_opts_pi_end = nd_opt;
253                         if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
254                                 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
255                         break;
256 #ifdef CONFIG_IPV6_ROUTE_INFO
257                 case ND_OPT_ROUTE_INFO:
258                         ndopts->nd_opts_ri_end = nd_opt;
259                         if (!ndopts->nd_opts_ri)
260                                 ndopts->nd_opts_ri = nd_opt;
261                         break;
262 #endif
263                 default:
264                         if (ndisc_is_useropt(dev, nd_opt)) {
265                                 ndopts->nd_useropts_end = nd_opt;
266                                 if (!ndopts->nd_useropts)
267                                         ndopts->nd_useropts = nd_opt;
268                         } else {
269                                 /*
270                                  * Unknown options must be silently ignored,
271                                  * to accommodate future extension to the
272                                  * protocol.
273                                  */
274                                 ND_PRINTK(2, notice,
275                                           "%s: ignored unsupported option; type=%d, len=%d\n",
276                                           __func__,
277                                           nd_opt->nd_opt_type,
278                                           nd_opt->nd_opt_len);
279                         }
280                 }
281 next_opt:
282                 opt_len -= l;
283                 nd_opt = ((void *)nd_opt) + l;
284         }
285         return ndopts;
286 }
287
288 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
289 {
290         switch (dev->type) {
291         case ARPHRD_ETHER:
292         case ARPHRD_IEEE802:    /* Not sure. Check it later. --ANK */
293         case ARPHRD_FDDI:
294                 ipv6_eth_mc_map(addr, buf);
295                 return 0;
296         case ARPHRD_ARCNET:
297                 ipv6_arcnet_mc_map(addr, buf);
298                 return 0;
299         case ARPHRD_INFINIBAND:
300                 ipv6_ib_mc_map(addr, dev->broadcast, buf);
301                 return 0;
302         case ARPHRD_IPGRE:
303                 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
304         default:
305                 if (dir) {
306                         memcpy(buf, dev->broadcast, dev->addr_len);
307                         return 0;
308                 }
309         }
310         return -EINVAL;
311 }
312 EXPORT_SYMBOL(ndisc_mc_map);
313
314 static u32 ndisc_hash(const void *pkey,
315                       const struct net_device *dev,
316                       __u32 *hash_rnd)
317 {
318         return ndisc_hashfn(pkey, dev, hash_rnd);
319 }
320
321 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
322 {
323         return neigh_key_eq128(n, pkey);
324 }
325
326 static int ndisc_constructor(struct neighbour *neigh)
327 {
328         struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
329         struct net_device *dev = neigh->dev;
330         struct inet6_dev *in6_dev;
331         struct neigh_parms *parms;
332         bool is_multicast = ipv6_addr_is_multicast(addr);
333
334         in6_dev = in6_dev_get(dev);
335         if (!in6_dev) {
336                 return -EINVAL;
337         }
338
339         parms = in6_dev->nd_parms;
340         __neigh_parms_put(neigh->parms);
341         neigh->parms = neigh_parms_clone(parms);
342
343         neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
344         if (!dev->header_ops) {
345                 neigh->nud_state = NUD_NOARP;
346                 neigh->ops = &ndisc_direct_ops;
347                 neigh->output = neigh_direct_output;
348         } else {
349                 if (is_multicast) {
350                         neigh->nud_state = NUD_NOARP;
351                         ndisc_mc_map(addr, neigh->ha, dev, 1);
352                 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
353                         neigh->nud_state = NUD_NOARP;
354                         memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
355                         if (dev->flags&IFF_LOOPBACK)
356                                 neigh->type = RTN_LOCAL;
357                 } else if (dev->flags&IFF_POINTOPOINT) {
358                         neigh->nud_state = NUD_NOARP;
359                         memcpy(neigh->ha, dev->broadcast, dev->addr_len);
360                 }
361                 if (dev->header_ops->cache)
362                         neigh->ops = &ndisc_hh_ops;
363                 else
364                         neigh->ops = &ndisc_generic_ops;
365                 if (neigh->nud_state&NUD_VALID)
366                         neigh->output = neigh->ops->connected_output;
367                 else
368                         neigh->output = neigh->ops->output;
369         }
370         in6_dev_put(in6_dev);
371         return 0;
372 }
373
374 static int pndisc_constructor(struct pneigh_entry *n)
375 {
376         struct in6_addr *addr = (struct in6_addr *)&n->key;
377         struct in6_addr maddr;
378         struct net_device *dev = n->dev;
379
380         if (!dev || !__in6_dev_get(dev))
381                 return -EINVAL;
382         addrconf_addr_solict_mult(addr, &maddr);
383         ipv6_dev_mc_inc(dev, &maddr);
384         return 0;
385 }
386
387 static void pndisc_destructor(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 || !__in6_dev_get(dev))
394                 return;
395         addrconf_addr_solict_mult(addr, &maddr);
396         ipv6_dev_mc_dec(dev, &maddr);
397 }
398
399 /* called with rtnl held */
400 static bool ndisc_allow_add(const struct net_device *dev,
401                             struct netlink_ext_ack *extack)
402 {
403         struct inet6_dev *idev = __in6_dev_get(dev);
404
405         if (!idev || idev->cnf.disable_ipv6) {
406                 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
407                 return false;
408         }
409
410         return true;
411 }
412
413 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
414                                        int len)
415 {
416         int hlen = LL_RESERVED_SPACE(dev);
417         int tlen = dev->needed_tailroom;
418         struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
419         struct sk_buff *skb;
420
421         skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
422         if (!skb) {
423                 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
424                           __func__);
425                 return NULL;
426         }
427
428         skb->protocol = htons(ETH_P_IPV6);
429         skb->dev = dev;
430
431         skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
432         skb_reset_transport_header(skb);
433
434         /* Manually assign socket ownership as we avoid calling
435          * sock_alloc_send_pskb() to bypass wmem buffer limits
436          */
437         skb_set_owner_w(skb, sk);
438
439         return skb;
440 }
441
442 static void ip6_nd_hdr(struct sk_buff *skb,
443                        const struct in6_addr *saddr,
444                        const struct in6_addr *daddr,
445                        int hop_limit, int len)
446 {
447         struct ipv6hdr *hdr;
448         struct inet6_dev *idev;
449         unsigned tclass;
450
451         rcu_read_lock();
452         idev = __in6_dev_get(skb->dev);
453         tclass = idev ? idev->cnf.ndisc_tclass : 0;
454         rcu_read_unlock();
455
456         skb_push(skb, sizeof(*hdr));
457         skb_reset_network_header(skb);
458         hdr = ipv6_hdr(skb);
459
460         ip6_flow_hdr(hdr, tclass, 0);
461
462         hdr->payload_len = htons(len);
463         hdr->nexthdr = IPPROTO_ICMPV6;
464         hdr->hop_limit = hop_limit;
465
466         hdr->saddr = *saddr;
467         hdr->daddr = *daddr;
468 }
469
470 static void ndisc_send_skb(struct sk_buff *skb,
471                            const struct in6_addr *daddr,
472                            const struct in6_addr *saddr)
473 {
474         struct dst_entry *dst = skb_dst(skb);
475         struct net *net = dev_net(skb->dev);
476         struct sock *sk = net->ipv6.ndisc_sk;
477         struct inet6_dev *idev;
478         int err;
479         struct icmp6hdr *icmp6h = icmp6_hdr(skb);
480         u8 type;
481
482         type = icmp6h->icmp6_type;
483
484         if (!dst) {
485                 struct flowi6 fl6;
486                 int oif = skb->dev->ifindex;
487
488                 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
489                 dst = icmp6_dst_alloc(skb->dev, &fl6);
490                 if (IS_ERR(dst)) {
491                         kfree_skb(skb);
492                         return;
493                 }
494
495                 skb_dst_set(skb, dst);
496         }
497
498         icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
499                                               IPPROTO_ICMPV6,
500                                               csum_partial(icmp6h,
501                                                            skb->len, 0));
502
503         ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
504
505         rcu_read_lock();
506         idev = __in6_dev_get(dst->dev);
507         IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
508
509         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
510                       net, sk, skb, NULL, dst->dev,
511                       dst_output);
512         if (!err) {
513                 ICMP6MSGOUT_INC_STATS(net, idev, type);
514                 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
515         }
516
517         rcu_read_unlock();
518 }
519
520 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
521                    const struct in6_addr *solicited_addr,
522                    bool router, bool solicited, bool override, bool inc_opt)
523 {
524         struct sk_buff *skb;
525         struct in6_addr tmpaddr;
526         struct inet6_ifaddr *ifp;
527         const struct in6_addr *src_addr;
528         struct nd_msg *msg;
529         int optlen = 0;
530
531         /* for anycast or proxy, solicited_addr != src_addr */
532         ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
533         if (ifp) {
534                 src_addr = solicited_addr;
535                 if (ifp->flags & IFA_F_OPTIMISTIC)
536                         override = false;
537                 inc_opt |= ifp->idev->cnf.force_tllao;
538                 in6_ifa_put(ifp);
539         } else {
540                 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
541                                        inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
542                                        &tmpaddr))
543                         return;
544                 src_addr = &tmpaddr;
545         }
546
547         if (!dev->addr_len)
548                 inc_opt = false;
549         if (inc_opt)
550                 optlen += ndisc_opt_addr_space(dev,
551                                                NDISC_NEIGHBOUR_ADVERTISEMENT);
552
553         skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
554         if (!skb)
555                 return;
556
557         msg = skb_put(skb, sizeof(*msg));
558         *msg = (struct nd_msg) {
559                 .icmph = {
560                         .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
561                         .icmp6_router = router,
562                         .icmp6_solicited = solicited,
563                         .icmp6_override = override,
564                 },
565                 .target = *solicited_addr,
566         };
567
568         if (inc_opt)
569                 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
570                                        dev->dev_addr,
571                                        NDISC_NEIGHBOUR_ADVERTISEMENT);
572
573         ndisc_send_skb(skb, daddr, src_addr);
574 }
575
576 static void ndisc_send_unsol_na(struct net_device *dev)
577 {
578         struct inet6_dev *idev;
579         struct inet6_ifaddr *ifa;
580
581         idev = in6_dev_get(dev);
582         if (!idev)
583                 return;
584
585         read_lock_bh(&idev->lock);
586         list_for_each_entry(ifa, &idev->addr_list, if_list) {
587                 /* skip tentative addresses until dad completes */
588                 if (ifa->flags & IFA_F_TENTATIVE &&
589                     !(ifa->flags & IFA_F_OPTIMISTIC))
590                         continue;
591
592                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
593                               /*router=*/ !!idev->cnf.forwarding,
594                               /*solicited=*/ false, /*override=*/ true,
595                               /*inc_opt=*/ true);
596         }
597         read_unlock_bh(&idev->lock);
598
599         in6_dev_put(idev);
600 }
601
602 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
603                    const struct in6_addr *daddr, const struct in6_addr *saddr,
604                    u64 nonce)
605 {
606         struct sk_buff *skb;
607         struct in6_addr addr_buf;
608         int inc_opt = dev->addr_len;
609         int optlen = 0;
610         struct nd_msg *msg;
611
612         if (!saddr) {
613                 if (ipv6_get_lladdr(dev, &addr_buf,
614                                    (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
615                         return;
616                 saddr = &addr_buf;
617         }
618
619         if (ipv6_addr_any(saddr))
620                 inc_opt = false;
621         if (inc_opt)
622                 optlen += ndisc_opt_addr_space(dev,
623                                                NDISC_NEIGHBOUR_SOLICITATION);
624         if (nonce != 0)
625                 optlen += 8;
626
627         skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
628         if (!skb)
629                 return;
630
631         msg = skb_put(skb, sizeof(*msg));
632         *msg = (struct nd_msg) {
633                 .icmph = {
634                         .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
635                 },
636                 .target = *solicit,
637         };
638
639         if (inc_opt)
640                 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
641                                        dev->dev_addr,
642                                        NDISC_NEIGHBOUR_SOLICITATION);
643         if (nonce != 0) {
644                 u8 *opt = skb_put(skb, 8);
645
646                 opt[0] = ND_OPT_NONCE;
647                 opt[1] = 8 >> 3;
648                 memcpy(opt + 2, &nonce, 6);
649         }
650
651         ndisc_send_skb(skb, daddr, saddr);
652 }
653
654 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
655                    const struct in6_addr *daddr)
656 {
657         struct sk_buff *skb;
658         struct rs_msg *msg;
659         int send_sllao = dev->addr_len;
660         int optlen = 0;
661
662 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
663         /*
664          * According to section 2.2 of RFC 4429, we must not
665          * send router solicitations with a sllao from
666          * optimistic addresses, but we may send the solicitation
667          * if we don't include the sllao.  So here we check
668          * if our address is optimistic, and if so, we
669          * suppress the inclusion of the sllao.
670          */
671         if (send_sllao) {
672                 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
673                                                            dev, 1);
674                 if (ifp) {
675                         if (ifp->flags & IFA_F_OPTIMISTIC)  {
676                                 send_sllao = 0;
677                         }
678                         in6_ifa_put(ifp);
679                 } else {
680                         send_sllao = 0;
681                 }
682         }
683 #endif
684         if (send_sllao)
685                 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
686
687         skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
688         if (!skb)
689                 return;
690
691         msg = skb_put(skb, sizeof(*msg));
692         *msg = (struct rs_msg) {
693                 .icmph = {
694                         .icmp6_type = NDISC_ROUTER_SOLICITATION,
695                 },
696         };
697
698         if (send_sllao)
699                 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
700                                        dev->dev_addr,
701                                        NDISC_ROUTER_SOLICITATION);
702
703         ndisc_send_skb(skb, daddr, saddr);
704 }
705
706
707 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
708 {
709         /*
710          *      "The sender MUST return an ICMP
711          *       destination unreachable"
712          */
713         dst_link_failure(skb);
714         kfree_skb(skb);
715 }
716
717 /* Called with locked neigh: either read or both */
718
719 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
720 {
721         struct in6_addr *saddr = NULL;
722         struct in6_addr mcaddr;
723         struct net_device *dev = neigh->dev;
724         struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
725         int probes = atomic_read(&neigh->probes);
726
727         if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
728                                            dev, false, 1,
729                                            IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
730                 saddr = &ipv6_hdr(skb)->saddr;
731         probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
732         if (probes < 0) {
733                 if (!(neigh->nud_state & NUD_VALID)) {
734                         ND_PRINTK(1, dbg,
735                                   "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
736                                   __func__, target);
737                 }
738                 ndisc_send_ns(dev, target, target, saddr, 0);
739         } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
740                 neigh_app_ns(neigh);
741         } else {
742                 addrconf_addr_solict_mult(target, &mcaddr);
743                 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
744         }
745 }
746
747 static int pndisc_is_router(const void *pkey,
748                             struct net_device *dev)
749 {
750         struct pneigh_entry *n;
751         int ret = -1;
752
753         read_lock_bh(&nd_tbl.lock);
754         n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
755         if (n)
756                 ret = !!(n->flags & NTF_ROUTER);
757         read_unlock_bh(&nd_tbl.lock);
758
759         return ret;
760 }
761
762 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
763                   const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
764                   struct ndisc_options *ndopts)
765 {
766         neigh_update(neigh, lladdr, new, flags, 0);
767         /* report ndisc ops about neighbour update */
768         ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
769 }
770
771 static void ndisc_recv_ns(struct sk_buff *skb)
772 {
773         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
774         const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
775         const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
776         u8 *lladdr = NULL;
777         u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
778                                     offsetof(struct nd_msg, opt));
779         struct ndisc_options ndopts;
780         struct net_device *dev = skb->dev;
781         struct inet6_ifaddr *ifp;
782         struct inet6_dev *idev = NULL;
783         struct neighbour *neigh;
784         int dad = ipv6_addr_any(saddr);
785         bool inc;
786         int is_router = -1;
787         u64 nonce = 0;
788
789         if (skb->len < sizeof(struct nd_msg)) {
790                 ND_PRINTK(2, warn, "NS: packet too short\n");
791                 return;
792         }
793
794         if (ipv6_addr_is_multicast(&msg->target)) {
795                 ND_PRINTK(2, warn, "NS: multicast target address\n");
796                 return;
797         }
798
799         /*
800          * RFC2461 7.1.1:
801          * DAD has to be destined for solicited node multicast address.
802          */
803         if (dad && !ipv6_addr_is_solict_mult(daddr)) {
804                 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
805                 return;
806         }
807
808         if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
809                 ND_PRINTK(2, warn, "NS: invalid ND options\n");
810                 return;
811         }
812
813         if (ndopts.nd_opts_src_lladdr) {
814                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
815                 if (!lladdr) {
816                         ND_PRINTK(2, warn,
817                                   "NS: invalid link-layer address length\n");
818                         return;
819                 }
820
821                 /* RFC2461 7.1.1:
822                  *      If the IP source address is the unspecified address,
823                  *      there MUST NOT be source link-layer address option
824                  *      in the message.
825                  */
826                 if (dad) {
827                         ND_PRINTK(2, warn,
828                                   "NS: bad DAD packet (link-layer address option)\n");
829                         return;
830                 }
831         }
832         if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
833                 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
834
835         inc = ipv6_addr_is_multicast(daddr);
836
837         ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
838         if (ifp) {
839 have_ifp:
840                 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
841                         if (dad) {
842                                 if (nonce != 0 && ifp->dad_nonce == nonce) {
843                                         u8 *np = (u8 *)&nonce;
844                                         /* Matching nonce if looped back */
845                                         ND_PRINTK(2, notice,
846                                                   "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
847                                                   ifp->idev->dev->name,
848                                                   &ifp->addr, np);
849                                         goto out;
850                                 }
851                                 /*
852                                  * We are colliding with another node
853                                  * who is doing DAD
854                                  * so fail our DAD process
855                                  */
856                                 addrconf_dad_failure(skb, ifp);
857                                 return;
858                         } else {
859                                 /*
860                                  * This is not a dad solicitation.
861                                  * If we are an optimistic node,
862                                  * we should respond.
863                                  * Otherwise, we should ignore it.
864                                  */
865                                 if (!(ifp->flags & IFA_F_OPTIMISTIC))
866                                         goto out;
867                         }
868                 }
869
870                 idev = ifp->idev;
871         } else {
872                 struct net *net = dev_net(dev);
873
874                 /* perhaps an address on the master device */
875                 if (netif_is_l3_slave(dev)) {
876                         struct net_device *mdev;
877
878                         mdev = netdev_master_upper_dev_get_rcu(dev);
879                         if (mdev) {
880                                 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
881                                 if (ifp)
882                                         goto have_ifp;
883                         }
884                 }
885
886                 idev = in6_dev_get(dev);
887                 if (!idev) {
888                         /* XXX: count this drop? */
889                         return;
890                 }
891
892                 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
893                     (idev->cnf.forwarding &&
894                      (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
895                      (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
896                         if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
897                             skb->pkt_type != PACKET_HOST &&
898                             inc &&
899                             NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
900                                 /*
901                                  * for anycast or proxy,
902                                  * sender should delay its response
903                                  * by a random time between 0 and
904                                  * MAX_ANYCAST_DELAY_TIME seconds.
905                                  * (RFC2461) -- yoshfuji
906                                  */
907                                 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
908                                 if (n)
909                                         pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
910                                 goto out;
911                         }
912                 } else
913                         goto out;
914         }
915
916         if (is_router < 0)
917                 is_router = idev->cnf.forwarding;
918
919         if (dad) {
920                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
921                               !!is_router, false, (ifp != NULL), true);
922                 goto out;
923         }
924
925         if (inc)
926                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
927         else
928                 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
929
930         /*
931          *      update / create cache entry
932          *      for the source address
933          */
934         neigh = __neigh_lookup(&nd_tbl, saddr, dev,
935                                !inc || lladdr || !dev->addr_len);
936         if (neigh)
937                 ndisc_update(dev, neigh, lladdr, NUD_STALE,
938                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
939                              NEIGH_UPDATE_F_OVERRIDE,
940                              NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
941         if (neigh || !dev->header_ops) {
942                 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
943                               true, (ifp != NULL && inc), inc);
944                 if (neigh)
945                         neigh_release(neigh);
946         }
947
948 out:
949         if (ifp)
950                 in6_ifa_put(ifp);
951         else
952                 in6_dev_put(idev);
953 }
954
955 static void ndisc_recv_na(struct sk_buff *skb)
956 {
957         struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
958         struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
959         const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
960         u8 *lladdr = NULL;
961         u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
962                                     offsetof(struct nd_msg, opt));
963         struct ndisc_options ndopts;
964         struct net_device *dev = skb->dev;
965         struct inet6_dev *idev = __in6_dev_get(dev);
966         struct inet6_ifaddr *ifp;
967         struct neighbour *neigh;
968
969         if (skb->len < sizeof(struct nd_msg)) {
970                 ND_PRINTK(2, warn, "NA: packet too short\n");
971                 return;
972         }
973
974         if (ipv6_addr_is_multicast(&msg->target)) {
975                 ND_PRINTK(2, warn, "NA: target address is multicast\n");
976                 return;
977         }
978
979         if (ipv6_addr_is_multicast(daddr) &&
980             msg->icmph.icmp6_solicited) {
981                 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
982                 return;
983         }
984
985         /* For some 802.11 wireless deployments (and possibly other networks),
986          * there will be a NA proxy and unsolicitd packets are attacks
987          * and thus should not be accepted.
988          */
989         if (!msg->icmph.icmp6_solicited && idev &&
990             idev->cnf.drop_unsolicited_na)
991                 return;
992
993         if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
994                 ND_PRINTK(2, warn, "NS: invalid ND option\n");
995                 return;
996         }
997         if (ndopts.nd_opts_tgt_lladdr) {
998                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
999                 if (!lladdr) {
1000                         ND_PRINTK(2, warn,
1001                                   "NA: invalid link-layer address length\n");
1002                         return;
1003                 }
1004         }
1005         ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1006         if (ifp) {
1007                 if (skb->pkt_type != PACKET_LOOPBACK
1008                     && (ifp->flags & IFA_F_TENTATIVE)) {
1009                                 addrconf_dad_failure(skb, ifp);
1010                                 return;
1011                 }
1012                 /* What should we make now? The advertisement
1013                    is invalid, but ndisc specs say nothing
1014                    about it. It could be misconfiguration, or
1015                    an smart proxy agent tries to help us :-)
1016
1017                    We should not print the error if NA has been
1018                    received from loopback - it is just our own
1019                    unsolicited advertisement.
1020                  */
1021                 if (skb->pkt_type != PACKET_LOOPBACK)
1022                         ND_PRINTK(1, warn,
1023                                   "NA: %pM advertised our address %pI6c on %s!\n",
1024                                   eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1025                 in6_ifa_put(ifp);
1026                 return;
1027         }
1028         neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1029
1030         if (neigh) {
1031                 u8 old_flags = neigh->flags;
1032                 struct net *net = dev_net(dev);
1033
1034                 if (neigh->nud_state & NUD_FAILED)
1035                         goto out;
1036
1037                 /*
1038                  * Don't update the neighbor cache entry on a proxy NA from
1039                  * ourselves because either the proxied node is off link or it
1040                  * has already sent a NA to us.
1041                  */
1042                 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1043                     net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1044                     pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1045                         /* XXX: idev->cnf.proxy_ndp */
1046                         goto out;
1047                 }
1048
1049                 ndisc_update(dev, neigh, lladdr,
1050                              msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1051                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1052                              (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1053                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1054                              (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1055                              NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1056
1057                 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1058                         /*
1059                          * Change: router to host
1060                          */
1061                         rt6_clean_tohost(dev_net(dev),  saddr);
1062                 }
1063
1064 out:
1065                 neigh_release(neigh);
1066         }
1067 }
1068
1069 static void ndisc_recv_rs(struct sk_buff *skb)
1070 {
1071         struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1072         unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1073         struct neighbour *neigh;
1074         struct inet6_dev *idev;
1075         const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1076         struct ndisc_options ndopts;
1077         u8 *lladdr = NULL;
1078
1079         if (skb->len < sizeof(*rs_msg))
1080                 return;
1081
1082         idev = __in6_dev_get(skb->dev);
1083         if (!idev) {
1084                 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1085                 return;
1086         }
1087
1088         /* Don't accept RS if we're not in router mode */
1089         if (!idev->cnf.forwarding)
1090                 goto out;
1091
1092         /*
1093          * Don't update NCE if src = ::;
1094          * this implies that the source node has no ip address assigned yet.
1095          */
1096         if (ipv6_addr_any(saddr))
1097                 goto out;
1098
1099         /* Parse ND options */
1100         if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1101                 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1102                 goto out;
1103         }
1104
1105         if (ndopts.nd_opts_src_lladdr) {
1106                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1107                                              skb->dev);
1108                 if (!lladdr)
1109                         goto out;
1110         }
1111
1112         neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1113         if (neigh) {
1114                 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1115                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1116                              NEIGH_UPDATE_F_OVERRIDE|
1117                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1118                              NDISC_ROUTER_SOLICITATION, &ndopts);
1119                 neigh_release(neigh);
1120         }
1121 out:
1122         return;
1123 }
1124
1125 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1126 {
1127         struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1128         struct sk_buff *skb;
1129         struct nlmsghdr *nlh;
1130         struct nduseroptmsg *ndmsg;
1131         struct net *net = dev_net(ra->dev);
1132         int err;
1133         int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1134                                     + (opt->nd_opt_len << 3));
1135         size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1136
1137         skb = nlmsg_new(msg_size, GFP_ATOMIC);
1138         if (!skb) {
1139                 err = -ENOBUFS;
1140                 goto errout;
1141         }
1142
1143         nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1144         if (!nlh) {
1145                 goto nla_put_failure;
1146         }
1147
1148         ndmsg = nlmsg_data(nlh);
1149         ndmsg->nduseropt_family = AF_INET6;
1150         ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1151         ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1152         ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1153         ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1154
1155         memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1156
1157         if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1158                 goto nla_put_failure;
1159         nlmsg_end(skb, nlh);
1160
1161         rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1162         return;
1163
1164 nla_put_failure:
1165         nlmsg_free(skb);
1166         err = -EMSGSIZE;
1167 errout:
1168         rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1169 }
1170
1171 static void ndisc_router_discovery(struct sk_buff *skb)
1172 {
1173         struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1174         struct neighbour *neigh = NULL;
1175         struct inet6_dev *in6_dev;
1176         struct fib6_info *rt = NULL;
1177         struct net *net;
1178         int lifetime;
1179         struct ndisc_options ndopts;
1180         int optlen;
1181         unsigned int pref = 0;
1182         __u32 old_if_flags;
1183         bool send_ifinfo_notify = false;
1184
1185         __u8 *opt = (__u8 *)(ra_msg + 1);
1186
1187         optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1188                 sizeof(struct ra_msg);
1189
1190         ND_PRINTK(2, info,
1191                   "RA: %s, dev: %s\n",
1192                   __func__, skb->dev->name);
1193         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1194                 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1195                 return;
1196         }
1197         if (optlen < 0) {
1198                 ND_PRINTK(2, warn, "RA: packet too short\n");
1199                 return;
1200         }
1201
1202 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1203         if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1204                 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1205                 return;
1206         }
1207 #endif
1208
1209         /*
1210          *      set the RA_RECV flag in the interface
1211          */
1212
1213         in6_dev = __in6_dev_get(skb->dev);
1214         if (!in6_dev) {
1215                 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1216                           skb->dev->name);
1217                 return;
1218         }
1219
1220         if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1221                 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1222                 return;
1223         }
1224
1225         if (!ipv6_accept_ra(in6_dev)) {
1226                 ND_PRINTK(2, info,
1227                           "RA: %s, did not accept ra for dev: %s\n",
1228                           __func__, skb->dev->name);
1229                 goto skip_linkparms;
1230         }
1231
1232 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1233         /* skip link-specific parameters from interior routers */
1234         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1235                 ND_PRINTK(2, info,
1236                           "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1237                           __func__, skb->dev->name);
1238                 goto skip_linkparms;
1239         }
1240 #endif
1241
1242         if (in6_dev->if_flags & IF_RS_SENT) {
1243                 /*
1244                  *      flag that an RA was received after an RS was sent
1245                  *      out on this interface.
1246                  */
1247                 in6_dev->if_flags |= IF_RA_RCVD;
1248         }
1249
1250         /*
1251          * Remember the managed/otherconf flags from most recently
1252          * received RA message (RFC 2462) -- yoshfuji
1253          */
1254         old_if_flags = in6_dev->if_flags;
1255         in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1256                                 IF_RA_OTHERCONF)) |
1257                                 (ra_msg->icmph.icmp6_addrconf_managed ?
1258                                         IF_RA_MANAGED : 0) |
1259                                 (ra_msg->icmph.icmp6_addrconf_other ?
1260                                         IF_RA_OTHERCONF : 0);
1261
1262         if (old_if_flags != in6_dev->if_flags)
1263                 send_ifinfo_notify = true;
1264
1265         if (!in6_dev->cnf.accept_ra_defrtr) {
1266                 ND_PRINTK(2, info,
1267                           "RA: %s, defrtr is false for dev: %s\n",
1268                           __func__, skb->dev->name);
1269                 goto skip_defrtr;
1270         }
1271
1272         lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1273         if (lifetime != 0 && lifetime < in6_dev->cnf.accept_ra_min_lft) {
1274                 ND_PRINTK(2, info,
1275                           "RA: router lifetime (%ds) is too short: %s\n",
1276                           lifetime, skb->dev->name);
1277                 goto skip_defrtr;
1278         }
1279
1280         /* Do not accept RA with source-addr found on local machine unless
1281          * accept_ra_from_local is set to true.
1282          */
1283         net = dev_net(in6_dev->dev);
1284         if (!in6_dev->cnf.accept_ra_from_local &&
1285             ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1286                 ND_PRINTK(2, info,
1287                           "RA from local address detected on dev: %s: default router ignored\n",
1288                           skb->dev->name);
1289                 goto skip_defrtr;
1290         }
1291
1292 #ifdef CONFIG_IPV6_ROUTER_PREF
1293         pref = ra_msg->icmph.icmp6_router_pref;
1294         /* 10b is handled as if it were 00b (medium) */
1295         if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1296             !in6_dev->cnf.accept_ra_rtr_pref)
1297                 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1298 #endif
1299         /* routes added from RAs do not use nexthop objects */
1300         rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1301         if (rt) {
1302                 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1303                                          rt->fib6_nh->fib_nh_dev, NULL,
1304                                           &ipv6_hdr(skb)->saddr);
1305                 if (!neigh) {
1306                         ND_PRINTK(0, err,
1307                                   "RA: %s got default router without neighbour\n",
1308                                   __func__);
1309                         fib6_info_release(rt);
1310                         return;
1311                 }
1312         }
1313         if (rt && lifetime == 0) {
1314                 ip6_del_rt(net, rt, false);
1315                 rt = NULL;
1316         }
1317
1318         ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, for dev: %s\n",
1319                   rt, lifetime, skb->dev->name);
1320         if (!rt && lifetime) {
1321                 ND_PRINTK(3, info, "RA: adding default router\n");
1322
1323                 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1324                                          skb->dev, pref);
1325                 if (!rt) {
1326                         ND_PRINTK(0, err,
1327                                   "RA: %s failed to add default route\n",
1328                                   __func__);
1329                         return;
1330                 }
1331
1332                 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1333                                          rt->fib6_nh->fib_nh_dev, NULL,
1334                                           &ipv6_hdr(skb)->saddr);
1335                 if (!neigh) {
1336                         ND_PRINTK(0, err,
1337                                   "RA: %s got default router without neighbour\n",
1338                                   __func__);
1339                         fib6_info_release(rt);
1340                         return;
1341                 }
1342                 neigh->flags |= NTF_ROUTER;
1343         } else if (rt) {
1344                 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1345         }
1346
1347         if (rt)
1348                 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1349         if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1350             ra_msg->icmph.icmp6_hop_limit) {
1351                 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1352                         in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1353                         fib6_metric_set(rt, RTAX_HOPLIMIT,
1354                                         ra_msg->icmph.icmp6_hop_limit);
1355                 } else {
1356                         ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1357                 }
1358         }
1359
1360 skip_defrtr:
1361
1362         /*
1363          *      Update Reachable Time and Retrans Timer
1364          */
1365
1366         if (in6_dev->nd_parms) {
1367                 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1368
1369                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1370                         rtime = (rtime*HZ)/1000;
1371                         if (rtime < HZ/100)
1372                                 rtime = HZ/100;
1373                         NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1374                         in6_dev->tstamp = jiffies;
1375                         send_ifinfo_notify = true;
1376                 }
1377
1378                 rtime = ntohl(ra_msg->reachable_time);
1379                 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1380                         rtime = (rtime*HZ)/1000;
1381
1382                         if (rtime < HZ/10)
1383                                 rtime = HZ/10;
1384
1385                         if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1386                                 NEIGH_VAR_SET(in6_dev->nd_parms,
1387                                               BASE_REACHABLE_TIME, rtime);
1388                                 NEIGH_VAR_SET(in6_dev->nd_parms,
1389                                               GC_STALETIME, 3 * rtime);
1390                                 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1391                                 in6_dev->tstamp = jiffies;
1392                                 send_ifinfo_notify = true;
1393                         }
1394                 }
1395         }
1396
1397         /*
1398          *      Send a notify if RA changed managed/otherconf flags or timer settings
1399          */
1400         if (send_ifinfo_notify)
1401                 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1402
1403 skip_linkparms:
1404
1405         /*
1406          *      Process options.
1407          */
1408
1409         if (!neigh)
1410                 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1411                                        skb->dev, 1);
1412         if (neigh) {
1413                 u8 *lladdr = NULL;
1414                 if (ndopts.nd_opts_src_lladdr) {
1415                         lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1416                                                      skb->dev);
1417                         if (!lladdr) {
1418                                 ND_PRINTK(2, warn,
1419                                           "RA: invalid link-layer address length\n");
1420                                 goto out;
1421                         }
1422                 }
1423                 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1424                              NEIGH_UPDATE_F_WEAK_OVERRIDE|
1425                              NEIGH_UPDATE_F_OVERRIDE|
1426                              NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1427                              NEIGH_UPDATE_F_ISROUTER,
1428                              NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1429         }
1430
1431         if (!ipv6_accept_ra(in6_dev)) {
1432                 ND_PRINTK(2, info,
1433                           "RA: %s, accept_ra is false for dev: %s\n",
1434                           __func__, skb->dev->name);
1435                 goto out;
1436         }
1437
1438 #ifdef CONFIG_IPV6_ROUTE_INFO
1439         if (!in6_dev->cnf.accept_ra_from_local &&
1440             ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1441                           in6_dev->dev, 0)) {
1442                 ND_PRINTK(2, info,
1443                           "RA from local address detected on dev: %s: router info ignored.\n",
1444                           skb->dev->name);
1445                 goto skip_routeinfo;
1446         }
1447
1448         if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1449                 struct nd_opt_hdr *p;
1450                 for (p = ndopts.nd_opts_ri;
1451                      p;
1452                      p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1453                         struct route_info *ri = (struct route_info *)p;
1454 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1455                         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1456                             ri->prefix_len == 0)
1457                                 continue;
1458 #endif
1459                         if (ri->prefix_len == 0 &&
1460                             !in6_dev->cnf.accept_ra_defrtr)
1461                                 continue;
1462                         if (ri->lifetime != 0 &&
1463                             ntohl(ri->lifetime) < in6_dev->cnf.accept_ra_min_lft)
1464                                 continue;
1465                         if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1466                                 continue;
1467                         if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1468                                 continue;
1469                         rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1470                                       &ipv6_hdr(skb)->saddr);
1471                 }
1472         }
1473
1474 skip_routeinfo:
1475 #endif
1476
1477 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1478         /* skip link-specific ndopts from interior routers */
1479         if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1480                 ND_PRINTK(2, info,
1481                           "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1482                           __func__, skb->dev->name);
1483                 goto out;
1484         }
1485 #endif
1486
1487         if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1488                 struct nd_opt_hdr *p;
1489                 for (p = ndopts.nd_opts_pi;
1490                      p;
1491                      p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1492                         addrconf_prefix_rcv(skb->dev, (u8 *)p,
1493                                             (p->nd_opt_len) << 3,
1494                                             ndopts.nd_opts_src_lladdr != NULL);
1495                 }
1496         }
1497
1498         if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1499                 __be32 n;
1500                 u32 mtu;
1501
1502                 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1503                 mtu = ntohl(n);
1504
1505                 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1506                         ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1507                 } else if (in6_dev->cnf.mtu6 != mtu) {
1508                         in6_dev->cnf.mtu6 = mtu;
1509                         fib6_metric_set(rt, RTAX_MTU, mtu);
1510                         rt6_mtu_change(skb->dev, mtu);
1511                 }
1512         }
1513
1514         if (ndopts.nd_useropts) {
1515                 struct nd_opt_hdr *p;
1516                 for (p = ndopts.nd_useropts;
1517                      p;
1518                      p = ndisc_next_useropt(skb->dev, p,
1519                                             ndopts.nd_useropts_end)) {
1520                         ndisc_ra_useropt(skb, p);
1521                 }
1522         }
1523
1524         if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1525                 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1526         }
1527 out:
1528         fib6_info_release(rt);
1529         if (neigh)
1530                 neigh_release(neigh);
1531 }
1532
1533 static void ndisc_redirect_rcv(struct sk_buff *skb)
1534 {
1535         u8 *hdr;
1536         struct ndisc_options ndopts;
1537         struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1538         u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1539                                     offsetof(struct rd_msg, opt));
1540
1541 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1542         switch (skb->ndisc_nodetype) {
1543         case NDISC_NODETYPE_HOST:
1544         case NDISC_NODETYPE_NODEFAULT:
1545                 ND_PRINTK(2, warn,
1546                           "Redirect: from host or unauthorized router\n");
1547                 return;
1548         }
1549 #endif
1550
1551         if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1552                 ND_PRINTK(2, warn,
1553                           "Redirect: source address is not link-local\n");
1554                 return;
1555         }
1556
1557         if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1558                 return;
1559
1560         if (!ndopts.nd_opts_rh) {
1561                 ip6_redirect_no_header(skb, dev_net(skb->dev),
1562                                         skb->dev->ifindex);
1563                 return;
1564         }
1565
1566         hdr = (u8 *)ndopts.nd_opts_rh;
1567         hdr += 8;
1568         if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1569                 return;
1570
1571         icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1572 }
1573
1574 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1575                                            struct sk_buff *orig_skb,
1576                                            int rd_len)
1577 {
1578         u8 *opt = skb_put(skb, rd_len);
1579
1580         memset(opt, 0, 8);
1581         *(opt++) = ND_OPT_REDIRECT_HDR;
1582         *(opt++) = (rd_len >> 3);
1583         opt += 6;
1584
1585         skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1586                       rd_len - 8);
1587 }
1588
1589 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1590 {
1591         struct net_device *dev = skb->dev;
1592         struct net *net = dev_net(dev);
1593         struct sock *sk = net->ipv6.ndisc_sk;
1594         int optlen = 0;
1595         struct inet_peer *peer;
1596         struct sk_buff *buff;
1597         struct rd_msg *msg;
1598         struct in6_addr saddr_buf;
1599         struct rt6_info *rt;
1600         struct dst_entry *dst;
1601         struct flowi6 fl6;
1602         int rd_len;
1603         u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1604            ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1605         bool ret;
1606
1607         if (netif_is_l3_master(skb->dev)) {
1608                 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1609                 if (!dev)
1610                         return;
1611         }
1612
1613         if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1614                 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1615                           dev->name);
1616                 return;
1617         }
1618
1619         if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1620             ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1621                 ND_PRINTK(2, warn,
1622                           "Redirect: target address is not link-local unicast\n");
1623                 return;
1624         }
1625
1626         icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1627                          &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1628
1629         dst = ip6_route_output(net, NULL, &fl6);
1630         if (dst->error) {
1631                 dst_release(dst);
1632                 return;
1633         }
1634         dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1635         if (IS_ERR(dst))
1636                 return;
1637
1638         rt = (struct rt6_info *) dst;
1639
1640         if (rt->rt6i_flags & RTF_GATEWAY) {
1641                 ND_PRINTK(2, warn,
1642                           "Redirect: destination is not a neighbour\n");
1643                 goto release;
1644         }
1645         peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1646         ret = inet_peer_xrlim_allow(peer, 1*HZ);
1647         if (peer)
1648                 inet_putpeer(peer);
1649         if (!ret)
1650                 goto release;
1651
1652         if (dev->addr_len) {
1653                 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1654                 if (!neigh) {
1655                         ND_PRINTK(2, warn,
1656                                   "Redirect: no neigh for target address\n");
1657                         goto release;
1658                 }
1659
1660                 read_lock_bh(&neigh->lock);
1661                 if (neigh->nud_state & NUD_VALID) {
1662                         memcpy(ha_buf, neigh->ha, dev->addr_len);
1663                         read_unlock_bh(&neigh->lock);
1664                         ha = ha_buf;
1665                         optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1666                                                                 ops_data_buf,
1667                                                                 &ops_data);
1668                 } else
1669                         read_unlock_bh(&neigh->lock);
1670
1671                 neigh_release(neigh);
1672         }
1673
1674         rd_len = min_t(unsigned int,
1675                        IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1676                        skb->len + 8);
1677         rd_len &= ~0x7;
1678         optlen += rd_len;
1679
1680         buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1681         if (!buff)
1682                 goto release;
1683
1684         msg = skb_put(buff, sizeof(*msg));
1685         *msg = (struct rd_msg) {
1686                 .icmph = {
1687                         .icmp6_type = NDISC_REDIRECT,
1688                 },
1689                 .target = *target,
1690                 .dest = ipv6_hdr(skb)->daddr,
1691         };
1692
1693         /*
1694          *      include target_address option
1695          */
1696
1697         if (ha)
1698                 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1699
1700         /*
1701          *      build redirect option and copy skb over to the new packet.
1702          */
1703
1704         if (rd_len)
1705                 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1706
1707         skb_dst_set(buff, dst);
1708         ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1709         return;
1710
1711 release:
1712         dst_release(dst);
1713 }
1714
1715 static void pndisc_redo(struct sk_buff *skb)
1716 {
1717         ndisc_recv_ns(skb);
1718         kfree_skb(skb);
1719 }
1720
1721 static int ndisc_is_multicast(const void *pkey)
1722 {
1723         return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1724 }
1725
1726 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1727 {
1728         struct inet6_dev *idev = __in6_dev_get(skb->dev);
1729
1730         if (!idev)
1731                 return true;
1732         if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1733             idev->cnf.suppress_frag_ndisc) {
1734                 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1735                 return true;
1736         }
1737         return false;
1738 }
1739
1740 int ndisc_rcv(struct sk_buff *skb)
1741 {
1742         struct nd_msg *msg;
1743
1744         if (ndisc_suppress_frag_ndisc(skb))
1745                 return 0;
1746
1747         if (skb_linearize(skb))
1748                 return 0;
1749
1750         msg = (struct nd_msg *)skb_transport_header(skb);
1751
1752         __skb_push(skb, skb->data - skb_transport_header(skb));
1753
1754         if (ipv6_hdr(skb)->hop_limit != 255) {
1755                 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1756                           ipv6_hdr(skb)->hop_limit);
1757                 return 0;
1758         }
1759
1760         if (msg->icmph.icmp6_code != 0) {
1761                 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1762                           msg->icmph.icmp6_code);
1763                 return 0;
1764         }
1765
1766         switch (msg->icmph.icmp6_type) {
1767         case NDISC_NEIGHBOUR_SOLICITATION:
1768                 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1769                 ndisc_recv_ns(skb);
1770                 break;
1771
1772         case NDISC_NEIGHBOUR_ADVERTISEMENT:
1773                 ndisc_recv_na(skb);
1774                 break;
1775
1776         case NDISC_ROUTER_SOLICITATION:
1777                 ndisc_recv_rs(skb);
1778                 break;
1779
1780         case NDISC_ROUTER_ADVERTISEMENT:
1781                 ndisc_router_discovery(skb);
1782                 break;
1783
1784         case NDISC_REDIRECT:
1785                 ndisc_redirect_rcv(skb);
1786                 break;
1787         }
1788
1789         return 0;
1790 }
1791
1792 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1793 {
1794         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1795         struct netdev_notifier_change_info *change_info;
1796         struct net *net = dev_net(dev);
1797         struct inet6_dev *idev;
1798
1799         switch (event) {
1800         case NETDEV_CHANGEADDR:
1801                 neigh_changeaddr(&nd_tbl, dev);
1802                 fib6_run_gc(0, net, false);
1803                 fallthrough;
1804         case NETDEV_UP:
1805                 idev = in6_dev_get(dev);
1806                 if (!idev)
1807                         break;
1808                 if (idev->cnf.ndisc_notify ||
1809                     net->ipv6.devconf_all->ndisc_notify)
1810                         ndisc_send_unsol_na(dev);
1811                 in6_dev_put(idev);
1812                 break;
1813         case NETDEV_CHANGE:
1814                 change_info = ptr;
1815                 if (change_info->flags_changed & IFF_NOARP)
1816                         neigh_changeaddr(&nd_tbl, dev);
1817                 if (!netif_carrier_ok(dev))
1818                         neigh_carrier_down(&nd_tbl, dev);
1819                 break;
1820         case NETDEV_DOWN:
1821                 neigh_ifdown(&nd_tbl, dev);
1822                 fib6_run_gc(0, net, false);
1823                 break;
1824         case NETDEV_NOTIFY_PEERS:
1825                 ndisc_send_unsol_na(dev);
1826                 break;
1827         default:
1828                 break;
1829         }
1830
1831         return NOTIFY_DONE;
1832 }
1833
1834 static struct notifier_block ndisc_netdev_notifier = {
1835         .notifier_call = ndisc_netdev_event,
1836         .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1837 };
1838
1839 #ifdef CONFIG_SYSCTL
1840 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1841                                          const char *func, const char *dev_name)
1842 {
1843         static char warncomm[TASK_COMM_LEN];
1844         static int warned;
1845         if (strcmp(warncomm, current->comm) && warned < 5) {
1846                 strcpy(warncomm, current->comm);
1847                 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1848                         warncomm, func,
1849                         dev_name, ctl->procname,
1850                         dev_name, ctl->procname);
1851                 warned++;
1852         }
1853 }
1854
1855 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1856                 size_t *lenp, loff_t *ppos)
1857 {
1858         struct net_device *dev = ctl->extra1;
1859         struct inet6_dev *idev;
1860         int ret;
1861
1862         if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1863             (strcmp(ctl->procname, "base_reachable_time") == 0))
1864                 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1865
1866         if (strcmp(ctl->procname, "retrans_time") == 0)
1867                 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1868
1869         else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1870                 ret = neigh_proc_dointvec_jiffies(ctl, write,
1871                                                   buffer, lenp, ppos);
1872
1873         else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1874                  (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1875                 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1876                                                      buffer, lenp, ppos);
1877         else
1878                 ret = -1;
1879
1880         if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1881                 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1882                         idev->nd_parms->reachable_time =
1883                                         neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1884                 idev->tstamp = jiffies;
1885                 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1886                 in6_dev_put(idev);
1887         }
1888         return ret;
1889 }
1890
1891
1892 #endif
1893
1894 static int __net_init ndisc_net_init(struct net *net)
1895 {
1896         struct ipv6_pinfo *np;
1897         struct sock *sk;
1898         int err;
1899
1900         err = inet_ctl_sock_create(&sk, PF_INET6,
1901                                    SOCK_RAW, IPPROTO_ICMPV6, net);
1902         if (err < 0) {
1903                 ND_PRINTK(0, err,
1904                           "NDISC: Failed to initialize the control socket (err %d)\n",
1905                           err);
1906                 return err;
1907         }
1908
1909         net->ipv6.ndisc_sk = sk;
1910
1911         np = inet6_sk(sk);
1912         np->hop_limit = 255;
1913         /* Do not loopback ndisc messages */
1914         np->mc_loop = 0;
1915
1916         return 0;
1917 }
1918
1919 static void __net_exit ndisc_net_exit(struct net *net)
1920 {
1921         inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1922 }
1923
1924 static struct pernet_operations ndisc_net_ops = {
1925         .init = ndisc_net_init,
1926         .exit = ndisc_net_exit,
1927 };
1928
1929 int __init ndisc_init(void)
1930 {
1931         int err;
1932
1933         err = register_pernet_subsys(&ndisc_net_ops);
1934         if (err)
1935                 return err;
1936         /*
1937          * Initialize the neighbour table
1938          */
1939         neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1940
1941 #ifdef CONFIG_SYSCTL
1942         err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1943                                     ndisc_ifinfo_sysctl_change);
1944         if (err)
1945                 goto out_unregister_pernet;
1946 out:
1947 #endif
1948         return err;
1949
1950 #ifdef CONFIG_SYSCTL
1951 out_unregister_pernet:
1952         unregister_pernet_subsys(&ndisc_net_ops);
1953         goto out;
1954 #endif
1955 }
1956
1957 int __init ndisc_late_init(void)
1958 {
1959         return register_netdevice_notifier(&ndisc_netdev_notifier);
1960 }
1961
1962 void ndisc_late_cleanup(void)
1963 {
1964         unregister_netdevice_notifier(&ndisc_netdev_notifier);
1965 }
1966
1967 void ndisc_cleanup(void)
1968 {
1969 #ifdef CONFIG_SYSCTL
1970         neigh_sysctl_unregister(&nd_tbl.parms);
1971 #endif
1972         neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1973         unregister_pernet_subsys(&ndisc_net_ops);
1974 }