ARC: defconfig: Cleanup from old Kconfig options
[linux/fpc-iii.git] / net / ipv6 / ndisc.c
blob0327c1f2e6fc5a8d22795afb476a1639bb341e99
1 /*
2 * Neighbour Discovery for IPv6
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Mike Shaver <shaver@ingenia.com>
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.
16 * Changes:
18 * Alexey I. Froloff : RFC6106 (DNSSL) support
19 * Pierre Ynard : export userland ND options
20 * through netlink (RDNSS support)
21 * Lars Fenneberg : fixed MTU setting on receipt
22 * of an RA.
23 * Janos Farkas : kmalloc failure checks
24 * Alexey Kuznetsov : state machine reworked
25 * and moved to net/core.
26 * Pekka Savola : RFC2461 validation
27 * YOSHIFUJI Hideaki @USAGI : Verify ND options properly
30 #define pr_fmt(fmt) "ICMPv6: " fmt
32 #include <linux/module.h>
33 #include <linux/errno.h>
34 #include <linux/types.h>
35 #include <linux/socket.h>
36 #include <linux/sockios.h>
37 #include <linux/sched.h>
38 #include <linux/net.h>
39 #include <linux/in6.h>
40 #include <linux/route.h>
41 #include <linux/init.h>
42 #include <linux/rcupdate.h>
43 #include <linux/slab.h>
44 #ifdef CONFIG_SYSCTL
45 #include <linux/sysctl.h>
46 #endif
48 #include <linux/if_addr.h>
49 #include <linux/if_arp.h>
50 #include <linux/ipv6.h>
51 #include <linux/icmpv6.h>
52 #include <linux/jhash.h>
54 #include <net/sock.h>
55 #include <net/snmp.h>
57 #include <net/ipv6.h>
58 #include <net/protocol.h>
59 #include <net/ndisc.h>
60 #include <net/ip6_route.h>
61 #include <net/addrconf.h>
62 #include <net/icmp.h>
64 #include <net/netlink.h>
65 #include <linux/rtnetlink.h>
67 #include <net/flow.h>
68 #include <net/ip6_checksum.h>
69 #include <net/inet_common.h>
70 #include <linux/proc_fs.h>
72 #include <linux/netfilter.h>
73 #include <linux/netfilter_ipv6.h>
75 static u32 ndisc_hash(const void *pkey,
76 const struct net_device *dev,
77 __u32 *hash_rnd);
78 static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
79 static int ndisc_constructor(struct neighbour *neigh);
80 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
81 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
82 static int pndisc_constructor(struct pneigh_entry *n);
83 static void pndisc_destructor(struct pneigh_entry *n);
84 static void pndisc_redo(struct sk_buff *skb);
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,
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,
103 static const struct neigh_ops ndisc_direct_ops = {
104 .family = AF_INET6,
105 .output = neigh_direct_output,
106 .connected_output = neigh_direct_output,
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 .id = "ndisc_cache",
120 .parms = {
121 .tbl = &nd_tbl,
122 .reachable_time = ND_REACHABLE_TIME,
123 .data = {
124 [NEIGH_VAR_MCAST_PROBES] = 3,
125 [NEIGH_VAR_UCAST_PROBES] = 3,
126 [NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER,
127 [NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME,
128 [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
129 [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
130 [NEIGH_VAR_QUEUE_LEN_BYTES] = 64 * 1024,
131 [NEIGH_VAR_PROXY_QLEN] = 64,
132 [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
133 [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
136 .gc_interval = 30 * HZ,
137 .gc_thresh1 = 128,
138 .gc_thresh2 = 512,
139 .gc_thresh3 = 1024,
141 EXPORT_SYMBOL_GPL(nd_tbl);
143 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, void *data,
144 int data_len, int pad)
146 int space = __ndisc_opt_addr_space(data_len, pad);
147 u8 *opt = skb_put(skb, space);
149 opt[0] = type;
150 opt[1] = space>>3;
152 memset(opt + 2, 0, pad);
153 opt += pad;
154 space -= pad;
156 memcpy(opt+2, data, data_len);
157 data_len += 2;
158 opt += data_len;
159 space -= data_len;
160 if (space > 0)
161 memset(opt, 0, space);
163 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
165 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
166 void *data, u8 icmp6_type)
168 __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
169 ndisc_addr_option_pad(skb->dev->type));
170 ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
173 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
174 void *ha,
175 const u8 *ops_data)
177 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
178 ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
181 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
182 struct nd_opt_hdr *end)
184 int type;
185 if (!cur || !end || cur >= end)
186 return NULL;
187 type = cur->nd_opt_type;
188 do {
189 cur = ((void *)cur) + (cur->nd_opt_len << 3);
190 } while (cur < end && cur->nd_opt_type != type);
191 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
194 static inline int ndisc_is_useropt(const struct net_device *dev,
195 struct nd_opt_hdr *opt)
197 return opt->nd_opt_type == ND_OPT_RDNSS ||
198 opt->nd_opt_type == ND_OPT_DNSSL ||
199 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
202 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
203 struct nd_opt_hdr *cur,
204 struct nd_opt_hdr *end)
206 if (!cur || !end || cur >= end)
207 return NULL;
208 do {
209 cur = ((void *)cur) + (cur->nd_opt_len << 3);
210 } while (cur < end && !ndisc_is_useropt(dev, cur));
211 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
214 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
215 u8 *opt, int opt_len,
216 struct ndisc_options *ndopts)
218 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
220 if (!nd_opt || opt_len < 0 || !ndopts)
221 return NULL;
222 memset(ndopts, 0, sizeof(*ndopts));
223 while (opt_len) {
224 int l;
225 if (opt_len < sizeof(struct nd_opt_hdr))
226 return NULL;
227 l = nd_opt->nd_opt_len << 3;
228 if (opt_len < l || l == 0)
229 return NULL;
230 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
231 goto next_opt;
232 switch (nd_opt->nd_opt_type) {
233 case ND_OPT_SOURCE_LL_ADDR:
234 case ND_OPT_TARGET_LL_ADDR:
235 case ND_OPT_MTU:
236 case ND_OPT_NONCE:
237 case ND_OPT_REDIRECT_HDR:
238 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
239 ND_PRINTK(2, warn,
240 "%s: duplicated ND6 option found: type=%d\n",
241 __func__, nd_opt->nd_opt_type);
242 } else {
243 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
245 break;
246 case ND_OPT_PREFIX_INFO:
247 ndopts->nd_opts_pi_end = nd_opt;
248 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
249 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
250 break;
251 #ifdef CONFIG_IPV6_ROUTE_INFO
252 case ND_OPT_ROUTE_INFO:
253 ndopts->nd_opts_ri_end = nd_opt;
254 if (!ndopts->nd_opts_ri)
255 ndopts->nd_opts_ri = nd_opt;
256 break;
257 #endif
258 default:
259 if (ndisc_is_useropt(dev, nd_opt)) {
260 ndopts->nd_useropts_end = nd_opt;
261 if (!ndopts->nd_useropts)
262 ndopts->nd_useropts = nd_opt;
263 } else {
265 * Unknown options must be silently ignored,
266 * to accommodate future extension to the
267 * protocol.
269 ND_PRINTK(2, notice,
270 "%s: ignored unsupported option; type=%d, len=%d\n",
271 __func__,
272 nd_opt->nd_opt_type,
273 nd_opt->nd_opt_len);
276 next_opt:
277 opt_len -= l;
278 nd_opt = ((void *)nd_opt) + l;
280 return ndopts;
283 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
285 switch (dev->type) {
286 case ARPHRD_ETHER:
287 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
288 case ARPHRD_FDDI:
289 ipv6_eth_mc_map(addr, buf);
290 return 0;
291 case ARPHRD_ARCNET:
292 ipv6_arcnet_mc_map(addr, buf);
293 return 0;
294 case ARPHRD_INFINIBAND:
295 ipv6_ib_mc_map(addr, dev->broadcast, buf);
296 return 0;
297 case ARPHRD_IPGRE:
298 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
299 default:
300 if (dir) {
301 memcpy(buf, dev->broadcast, dev->addr_len);
302 return 0;
305 return -EINVAL;
307 EXPORT_SYMBOL(ndisc_mc_map);
309 static u32 ndisc_hash(const void *pkey,
310 const struct net_device *dev,
311 __u32 *hash_rnd)
313 return ndisc_hashfn(pkey, dev, hash_rnd);
316 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
318 return neigh_key_eq128(n, pkey);
321 static int ndisc_constructor(struct neighbour *neigh)
323 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
324 struct net_device *dev = neigh->dev;
325 struct inet6_dev *in6_dev;
326 struct neigh_parms *parms;
327 bool is_multicast = ipv6_addr_is_multicast(addr);
329 in6_dev = in6_dev_get(dev);
330 if (!in6_dev) {
331 return -EINVAL;
334 parms = in6_dev->nd_parms;
335 __neigh_parms_put(neigh->parms);
336 neigh->parms = neigh_parms_clone(parms);
338 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
339 if (!dev->header_ops) {
340 neigh->nud_state = NUD_NOARP;
341 neigh->ops = &ndisc_direct_ops;
342 neigh->output = neigh_direct_output;
343 } else {
344 if (is_multicast) {
345 neigh->nud_state = NUD_NOARP;
346 ndisc_mc_map(addr, neigh->ha, dev, 1);
347 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
348 neigh->nud_state = NUD_NOARP;
349 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
350 if (dev->flags&IFF_LOOPBACK)
351 neigh->type = RTN_LOCAL;
352 } else if (dev->flags&IFF_POINTOPOINT) {
353 neigh->nud_state = NUD_NOARP;
354 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
356 if (dev->header_ops->cache)
357 neigh->ops = &ndisc_hh_ops;
358 else
359 neigh->ops = &ndisc_generic_ops;
360 if (neigh->nud_state&NUD_VALID)
361 neigh->output = neigh->ops->connected_output;
362 else
363 neigh->output = neigh->ops->output;
365 in6_dev_put(in6_dev);
366 return 0;
369 static int pndisc_constructor(struct pneigh_entry *n)
371 struct in6_addr *addr = (struct in6_addr *)&n->key;
372 struct in6_addr maddr;
373 struct net_device *dev = n->dev;
375 if (!dev || !__in6_dev_get(dev))
376 return -EINVAL;
377 addrconf_addr_solict_mult(addr, &maddr);
378 ipv6_dev_mc_inc(dev, &maddr);
379 return 0;
382 static void pndisc_destructor(struct pneigh_entry *n)
384 struct in6_addr *addr = (struct in6_addr *)&n->key;
385 struct in6_addr maddr;
386 struct net_device *dev = n->dev;
388 if (!dev || !__in6_dev_get(dev))
389 return;
390 addrconf_addr_solict_mult(addr, &maddr);
391 ipv6_dev_mc_dec(dev, &maddr);
394 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
395 int len)
397 int hlen = LL_RESERVED_SPACE(dev);
398 int tlen = dev->needed_tailroom;
399 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
400 struct sk_buff *skb;
402 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
403 if (!skb) {
404 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
405 __func__);
406 return NULL;
409 skb->protocol = htons(ETH_P_IPV6);
410 skb->dev = dev;
412 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
413 skb_reset_transport_header(skb);
415 /* Manually assign socket ownership as we avoid calling
416 * sock_alloc_send_pskb() to bypass wmem buffer limits
418 skb_set_owner_w(skb, sk);
420 return skb;
423 static void ip6_nd_hdr(struct sk_buff *skb,
424 const struct in6_addr *saddr,
425 const struct in6_addr *daddr,
426 int hop_limit, int len)
428 struct ipv6hdr *hdr;
430 skb_push(skb, sizeof(*hdr));
431 skb_reset_network_header(skb);
432 hdr = ipv6_hdr(skb);
434 ip6_flow_hdr(hdr, 0, 0);
436 hdr->payload_len = htons(len);
437 hdr->nexthdr = IPPROTO_ICMPV6;
438 hdr->hop_limit = hop_limit;
440 hdr->saddr = *saddr;
441 hdr->daddr = *daddr;
444 static void ndisc_send_skb(struct sk_buff *skb,
445 const struct in6_addr *daddr,
446 const struct in6_addr *saddr)
448 struct dst_entry *dst = skb_dst(skb);
449 struct net *net = dev_net(skb->dev);
450 struct sock *sk = net->ipv6.ndisc_sk;
451 struct inet6_dev *idev;
452 int err;
453 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
454 u8 type;
456 type = icmp6h->icmp6_type;
458 if (!dst) {
459 struct flowi6 fl6;
460 int oif = skb->dev->ifindex;
462 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
463 dst = icmp6_dst_alloc(skb->dev, &fl6);
464 if (IS_ERR(dst)) {
465 kfree_skb(skb);
466 return;
469 skb_dst_set(skb, dst);
472 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
473 IPPROTO_ICMPV6,
474 csum_partial(icmp6h,
475 skb->len, 0));
477 ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
479 rcu_read_lock();
480 idev = __in6_dev_get(dst->dev);
481 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
483 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
484 net, sk, skb, NULL, dst->dev,
485 dst_output);
486 if (!err) {
487 ICMP6MSGOUT_INC_STATS(net, idev, type);
488 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
491 rcu_read_unlock();
494 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
495 const struct in6_addr *solicited_addr,
496 bool router, bool solicited, bool override, bool inc_opt)
498 struct sk_buff *skb;
499 struct in6_addr tmpaddr;
500 struct inet6_ifaddr *ifp;
501 const struct in6_addr *src_addr;
502 struct nd_msg *msg;
503 int optlen = 0;
505 /* for anycast or proxy, solicited_addr != src_addr */
506 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
507 if (ifp) {
508 src_addr = solicited_addr;
509 if (ifp->flags & IFA_F_OPTIMISTIC)
510 override = false;
511 inc_opt |= ifp->idev->cnf.force_tllao;
512 in6_ifa_put(ifp);
513 } else {
514 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
515 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
516 &tmpaddr))
517 return;
518 src_addr = &tmpaddr;
521 if (!dev->addr_len)
522 inc_opt = 0;
523 if (inc_opt)
524 optlen += ndisc_opt_addr_space(dev,
525 NDISC_NEIGHBOUR_ADVERTISEMENT);
527 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
528 if (!skb)
529 return;
531 msg = skb_put(skb, sizeof(*msg));
532 *msg = (struct nd_msg) {
533 .icmph = {
534 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
535 .icmp6_router = router,
536 .icmp6_solicited = solicited,
537 .icmp6_override = override,
539 .target = *solicited_addr,
542 if (inc_opt)
543 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
544 dev->dev_addr,
545 NDISC_NEIGHBOUR_ADVERTISEMENT);
547 ndisc_send_skb(skb, daddr, src_addr);
550 static void ndisc_send_unsol_na(struct net_device *dev)
552 struct inet6_dev *idev;
553 struct inet6_ifaddr *ifa;
555 idev = in6_dev_get(dev);
556 if (!idev)
557 return;
559 read_lock_bh(&idev->lock);
560 list_for_each_entry(ifa, &idev->addr_list, if_list) {
561 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
562 /*router=*/ !!idev->cnf.forwarding,
563 /*solicited=*/ false, /*override=*/ true,
564 /*inc_opt=*/ true);
566 read_unlock_bh(&idev->lock);
568 in6_dev_put(idev);
571 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
572 const struct in6_addr *daddr, const struct in6_addr *saddr,
573 u64 nonce)
575 struct sk_buff *skb;
576 struct in6_addr addr_buf;
577 int inc_opt = dev->addr_len;
578 int optlen = 0;
579 struct nd_msg *msg;
581 if (!saddr) {
582 if (ipv6_get_lladdr(dev, &addr_buf,
583 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
584 return;
585 saddr = &addr_buf;
588 if (ipv6_addr_any(saddr))
589 inc_opt = false;
590 if (inc_opt)
591 optlen += ndisc_opt_addr_space(dev,
592 NDISC_NEIGHBOUR_SOLICITATION);
593 if (nonce != 0)
594 optlen += 8;
596 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
597 if (!skb)
598 return;
600 msg = skb_put(skb, sizeof(*msg));
601 *msg = (struct nd_msg) {
602 .icmph = {
603 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
605 .target = *solicit,
608 if (inc_opt)
609 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
610 dev->dev_addr,
611 NDISC_NEIGHBOUR_SOLICITATION);
612 if (nonce != 0) {
613 u8 *opt = skb_put(skb, 8);
615 opt[0] = ND_OPT_NONCE;
616 opt[1] = 8 >> 3;
617 memcpy(opt + 2, &nonce, 6);
620 ndisc_send_skb(skb, daddr, saddr);
623 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
624 const struct in6_addr *daddr)
626 struct sk_buff *skb;
627 struct rs_msg *msg;
628 int send_sllao = dev->addr_len;
629 int optlen = 0;
631 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
633 * According to section 2.2 of RFC 4429, we must not
634 * send router solicitations with a sllao from
635 * optimistic addresses, but we may send the solicitation
636 * if we don't include the sllao. So here we check
637 * if our address is optimistic, and if so, we
638 * suppress the inclusion of the sllao.
640 if (send_sllao) {
641 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
642 dev, 1);
643 if (ifp) {
644 if (ifp->flags & IFA_F_OPTIMISTIC) {
645 send_sllao = 0;
647 in6_ifa_put(ifp);
648 } else {
649 send_sllao = 0;
652 #endif
653 if (send_sllao)
654 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
656 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
657 if (!skb)
658 return;
660 msg = skb_put(skb, sizeof(*msg));
661 *msg = (struct rs_msg) {
662 .icmph = {
663 .icmp6_type = NDISC_ROUTER_SOLICITATION,
667 if (send_sllao)
668 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
669 dev->dev_addr,
670 NDISC_ROUTER_SOLICITATION);
672 ndisc_send_skb(skb, daddr, saddr);
676 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
679 * "The sender MUST return an ICMP
680 * destination unreachable"
682 dst_link_failure(skb);
683 kfree_skb(skb);
686 /* Called with locked neigh: either read or both */
688 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
690 struct in6_addr *saddr = NULL;
691 struct in6_addr mcaddr;
692 struct net_device *dev = neigh->dev;
693 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
694 int probes = atomic_read(&neigh->probes);
696 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
697 dev, 1,
698 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
699 saddr = &ipv6_hdr(skb)->saddr;
700 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
701 if (probes < 0) {
702 if (!(neigh->nud_state & NUD_VALID)) {
703 ND_PRINTK(1, dbg,
704 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
705 __func__, target);
707 ndisc_send_ns(dev, target, target, saddr, 0);
708 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
709 neigh_app_ns(neigh);
710 } else {
711 addrconf_addr_solict_mult(target, &mcaddr);
712 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
716 static int pndisc_is_router(const void *pkey,
717 struct net_device *dev)
719 struct pneigh_entry *n;
720 int ret = -1;
722 read_lock_bh(&nd_tbl.lock);
723 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
724 if (n)
725 ret = !!(n->flags & NTF_ROUTER);
726 read_unlock_bh(&nd_tbl.lock);
728 return ret;
731 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
732 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
733 struct ndisc_options *ndopts)
735 neigh_update(neigh, lladdr, new, flags, 0);
736 /* report ndisc ops about neighbour update */
737 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
740 static void ndisc_recv_ns(struct sk_buff *skb)
742 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
743 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
744 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
745 u8 *lladdr = NULL;
746 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
747 offsetof(struct nd_msg, opt));
748 struct ndisc_options ndopts;
749 struct net_device *dev = skb->dev;
750 struct inet6_ifaddr *ifp;
751 struct inet6_dev *idev = NULL;
752 struct neighbour *neigh;
753 int dad = ipv6_addr_any(saddr);
754 bool inc;
755 int is_router = -1;
756 u64 nonce = 0;
758 if (skb->len < sizeof(struct nd_msg)) {
759 ND_PRINTK(2, warn, "NS: packet too short\n");
760 return;
763 if (ipv6_addr_is_multicast(&msg->target)) {
764 ND_PRINTK(2, warn, "NS: multicast target address\n");
765 return;
769 * RFC2461 7.1.1:
770 * DAD has to be destined for solicited node multicast address.
772 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
773 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
774 return;
777 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
778 ND_PRINTK(2, warn, "NS: invalid ND options\n");
779 return;
782 if (ndopts.nd_opts_src_lladdr) {
783 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
784 if (!lladdr) {
785 ND_PRINTK(2, warn,
786 "NS: invalid link-layer address length\n");
787 return;
790 /* RFC2461 7.1.1:
791 * If the IP source address is the unspecified address,
792 * there MUST NOT be source link-layer address option
793 * in the message.
795 if (dad) {
796 ND_PRINTK(2, warn,
797 "NS: bad DAD packet (link-layer address option)\n");
798 return;
801 if (ndopts.nd_opts_nonce)
802 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
804 inc = ipv6_addr_is_multicast(daddr);
806 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
807 if (ifp) {
808 have_ifp:
809 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
810 if (dad) {
811 if (nonce != 0 && ifp->dad_nonce == nonce) {
812 u8 *np = (u8 *)&nonce;
813 /* Matching nonce if looped back */
814 ND_PRINTK(2, notice,
815 "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
816 ifp->idev->dev->name,
817 &ifp->addr, np);
818 goto out;
821 * We are colliding with another node
822 * who is doing DAD
823 * so fail our DAD process
825 addrconf_dad_failure(ifp);
826 return;
827 } else {
829 * This is not a dad solicitation.
830 * If we are an optimistic node,
831 * we should respond.
832 * Otherwise, we should ignore it.
834 if (!(ifp->flags & IFA_F_OPTIMISTIC))
835 goto out;
839 idev = ifp->idev;
840 } else {
841 struct net *net = dev_net(dev);
843 /* perhaps an address on the master device */
844 if (netif_is_l3_slave(dev)) {
845 struct net_device *mdev;
847 mdev = netdev_master_upper_dev_get_rcu(dev);
848 if (mdev) {
849 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
850 if (ifp)
851 goto have_ifp;
855 idev = in6_dev_get(dev);
856 if (!idev) {
857 /* XXX: count this drop? */
858 return;
861 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
862 (idev->cnf.forwarding &&
863 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
864 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
865 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
866 skb->pkt_type != PACKET_HOST &&
867 inc &&
868 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
870 * for anycast or proxy,
871 * sender should delay its response
872 * by a random time between 0 and
873 * MAX_ANYCAST_DELAY_TIME seconds.
874 * (RFC2461) -- yoshfuji
876 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
877 if (n)
878 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
879 goto out;
881 } else
882 goto out;
885 if (is_router < 0)
886 is_router = idev->cnf.forwarding;
888 if (dad) {
889 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
890 !!is_router, false, (ifp != NULL), true);
891 goto out;
894 if (inc)
895 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
896 else
897 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
900 * update / create cache entry
901 * for the source address
903 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
904 !inc || lladdr || !dev->addr_len);
905 if (neigh)
906 ndisc_update(dev, neigh, lladdr, NUD_STALE,
907 NEIGH_UPDATE_F_WEAK_OVERRIDE|
908 NEIGH_UPDATE_F_OVERRIDE,
909 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
910 if (neigh || !dev->header_ops) {
911 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
912 true, (ifp != NULL && inc), inc);
913 if (neigh)
914 neigh_release(neigh);
917 out:
918 if (ifp)
919 in6_ifa_put(ifp);
920 else
921 in6_dev_put(idev);
924 static void ndisc_recv_na(struct sk_buff *skb)
926 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
927 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
928 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
929 u8 *lladdr = NULL;
930 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
931 offsetof(struct nd_msg, opt));
932 struct ndisc_options ndopts;
933 struct net_device *dev = skb->dev;
934 struct inet6_dev *idev = __in6_dev_get(dev);
935 struct inet6_ifaddr *ifp;
936 struct neighbour *neigh;
938 if (skb->len < sizeof(struct nd_msg)) {
939 ND_PRINTK(2, warn, "NA: packet too short\n");
940 return;
943 if (ipv6_addr_is_multicast(&msg->target)) {
944 ND_PRINTK(2, warn, "NA: target address is multicast\n");
945 return;
948 if (ipv6_addr_is_multicast(daddr) &&
949 msg->icmph.icmp6_solicited) {
950 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
951 return;
954 /* For some 802.11 wireless deployments (and possibly other networks),
955 * there will be a NA proxy and unsolicitd packets are attacks
956 * and thus should not be accepted.
958 if (!msg->icmph.icmp6_solicited && idev &&
959 idev->cnf.drop_unsolicited_na)
960 return;
962 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
963 ND_PRINTK(2, warn, "NS: invalid ND option\n");
964 return;
966 if (ndopts.nd_opts_tgt_lladdr) {
967 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
968 if (!lladdr) {
969 ND_PRINTK(2, warn,
970 "NA: invalid link-layer address length\n");
971 return;
974 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
975 if (ifp) {
976 if (skb->pkt_type != PACKET_LOOPBACK
977 && (ifp->flags & IFA_F_TENTATIVE)) {
978 addrconf_dad_failure(ifp);
979 return;
981 /* What should we make now? The advertisement
982 is invalid, but ndisc specs say nothing
983 about it. It could be misconfiguration, or
984 an smart proxy agent tries to help us :-)
986 We should not print the error if NA has been
987 received from loopback - it is just our own
988 unsolicited advertisement.
990 if (skb->pkt_type != PACKET_LOOPBACK)
991 ND_PRINTK(1, warn,
992 "NA: someone advertises our address %pI6 on %s!\n",
993 &ifp->addr, ifp->idev->dev->name);
994 in6_ifa_put(ifp);
995 return;
997 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
999 if (neigh) {
1000 u8 old_flags = neigh->flags;
1001 struct net *net = dev_net(dev);
1003 if (neigh->nud_state & NUD_FAILED)
1004 goto out;
1007 * Don't update the neighbor cache entry on a proxy NA from
1008 * ourselves because either the proxied node is off link or it
1009 * has already sent a NA to us.
1011 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1012 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1013 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1014 /* XXX: idev->cnf.proxy_ndp */
1015 goto out;
1018 ndisc_update(dev, neigh, lladdr,
1019 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1020 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1021 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1022 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1023 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1024 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1026 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1028 * Change: router to host
1030 rt6_clean_tohost(dev_net(dev), saddr);
1033 out:
1034 neigh_release(neigh);
1038 static void ndisc_recv_rs(struct sk_buff *skb)
1040 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1041 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1042 struct neighbour *neigh;
1043 struct inet6_dev *idev;
1044 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1045 struct ndisc_options ndopts;
1046 u8 *lladdr = NULL;
1048 if (skb->len < sizeof(*rs_msg))
1049 return;
1051 idev = __in6_dev_get(skb->dev);
1052 if (!idev) {
1053 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1054 return;
1057 /* Don't accept RS if we're not in router mode */
1058 if (!idev->cnf.forwarding)
1059 goto out;
1062 * Don't update NCE if src = ::;
1063 * this implies that the source node has no ip address assigned yet.
1065 if (ipv6_addr_any(saddr))
1066 goto out;
1068 /* Parse ND options */
1069 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1070 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1071 goto out;
1074 if (ndopts.nd_opts_src_lladdr) {
1075 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1076 skb->dev);
1077 if (!lladdr)
1078 goto out;
1081 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1082 if (neigh) {
1083 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1084 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1085 NEIGH_UPDATE_F_OVERRIDE|
1086 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1087 NDISC_ROUTER_SOLICITATION, &ndopts);
1088 neigh_release(neigh);
1090 out:
1091 return;
1094 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1096 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1097 struct sk_buff *skb;
1098 struct nlmsghdr *nlh;
1099 struct nduseroptmsg *ndmsg;
1100 struct net *net = dev_net(ra->dev);
1101 int err;
1102 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1103 + (opt->nd_opt_len << 3));
1104 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1106 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1107 if (!skb) {
1108 err = -ENOBUFS;
1109 goto errout;
1112 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1113 if (!nlh) {
1114 goto nla_put_failure;
1117 ndmsg = nlmsg_data(nlh);
1118 ndmsg->nduseropt_family = AF_INET6;
1119 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1120 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1121 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1122 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1124 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1126 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1127 goto nla_put_failure;
1128 nlmsg_end(skb, nlh);
1130 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1131 return;
1133 nla_put_failure:
1134 nlmsg_free(skb);
1135 err = -EMSGSIZE;
1136 errout:
1137 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1140 static void ndisc_router_discovery(struct sk_buff *skb)
1142 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1143 struct neighbour *neigh = NULL;
1144 struct inet6_dev *in6_dev;
1145 struct rt6_info *rt = NULL;
1146 int lifetime;
1147 struct ndisc_options ndopts;
1148 int optlen;
1149 unsigned int pref = 0;
1150 __u32 old_if_flags;
1151 bool send_ifinfo_notify = false;
1153 __u8 *opt = (__u8 *)(ra_msg + 1);
1155 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1156 sizeof(struct ra_msg);
1158 ND_PRINTK(2, info,
1159 "RA: %s, dev: %s\n",
1160 __func__, skb->dev->name);
1161 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1162 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1163 return;
1165 if (optlen < 0) {
1166 ND_PRINTK(2, warn, "RA: packet too short\n");
1167 return;
1170 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1171 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1172 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1173 return;
1175 #endif
1178 * set the RA_RECV flag in the interface
1181 in6_dev = __in6_dev_get(skb->dev);
1182 if (!in6_dev) {
1183 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1184 skb->dev->name);
1185 return;
1188 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1189 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1190 return;
1193 if (!ipv6_accept_ra(in6_dev)) {
1194 ND_PRINTK(2, info,
1195 "RA: %s, did not accept ra for dev: %s\n",
1196 __func__, skb->dev->name);
1197 goto skip_linkparms;
1200 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1201 /* skip link-specific parameters from interior routers */
1202 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1203 ND_PRINTK(2, info,
1204 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1205 __func__, skb->dev->name);
1206 goto skip_linkparms;
1208 #endif
1210 if (in6_dev->if_flags & IF_RS_SENT) {
1212 * flag that an RA was received after an RS was sent
1213 * out on this interface.
1215 in6_dev->if_flags |= IF_RA_RCVD;
1219 * Remember the managed/otherconf flags from most recently
1220 * received RA message (RFC 2462) -- yoshfuji
1222 old_if_flags = in6_dev->if_flags;
1223 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1224 IF_RA_OTHERCONF)) |
1225 (ra_msg->icmph.icmp6_addrconf_managed ?
1226 IF_RA_MANAGED : 0) |
1227 (ra_msg->icmph.icmp6_addrconf_other ?
1228 IF_RA_OTHERCONF : 0);
1230 if (old_if_flags != in6_dev->if_flags)
1231 send_ifinfo_notify = true;
1233 if (!in6_dev->cnf.accept_ra_defrtr) {
1234 ND_PRINTK(2, info,
1235 "RA: %s, defrtr is false for dev: %s\n",
1236 __func__, skb->dev->name);
1237 goto skip_defrtr;
1240 /* Do not accept RA with source-addr found on local machine unless
1241 * accept_ra_from_local is set to true.
1243 if (!in6_dev->cnf.accept_ra_from_local &&
1244 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1245 in6_dev->dev, 0)) {
1246 ND_PRINTK(2, info,
1247 "RA from local address detected on dev: %s: default router ignored\n",
1248 skb->dev->name);
1249 goto skip_defrtr;
1252 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1254 #ifdef CONFIG_IPV6_ROUTER_PREF
1255 pref = ra_msg->icmph.icmp6_router_pref;
1256 /* 10b is handled as if it were 00b (medium) */
1257 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1258 !in6_dev->cnf.accept_ra_rtr_pref)
1259 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1260 #endif
1262 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1264 if (rt) {
1265 neigh = dst_neigh_lookup(&rt->dst, &ipv6_hdr(skb)->saddr);
1266 if (!neigh) {
1267 ND_PRINTK(0, err,
1268 "RA: %s got default router without neighbour\n",
1269 __func__);
1270 ip6_rt_put(rt);
1271 return;
1274 if (rt && lifetime == 0) {
1275 ip6_del_rt(rt);
1276 rt = NULL;
1279 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, for dev: %s\n",
1280 rt, lifetime, skb->dev->name);
1281 if (!rt && lifetime) {
1282 ND_PRINTK(3, info, "RA: adding default router\n");
1284 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1285 if (!rt) {
1286 ND_PRINTK(0, err,
1287 "RA: %s failed to add default route\n",
1288 __func__);
1289 return;
1292 neigh = dst_neigh_lookup(&rt->dst, &ipv6_hdr(skb)->saddr);
1293 if (!neigh) {
1294 ND_PRINTK(0, err,
1295 "RA: %s got default router without neighbour\n",
1296 __func__);
1297 ip6_rt_put(rt);
1298 return;
1300 neigh->flags |= NTF_ROUTER;
1301 } else if (rt) {
1302 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1305 if (rt)
1306 rt6_set_expires(rt, jiffies + (HZ * lifetime));
1307 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1308 ra_msg->icmph.icmp6_hop_limit) {
1309 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1310 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1311 if (rt)
1312 dst_metric_set(&rt->dst, RTAX_HOPLIMIT,
1313 ra_msg->icmph.icmp6_hop_limit);
1314 } else {
1315 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1319 skip_defrtr:
1322 * Update Reachable Time and Retrans Timer
1325 if (in6_dev->nd_parms) {
1326 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1328 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1329 rtime = (rtime*HZ)/1000;
1330 if (rtime < HZ/10)
1331 rtime = HZ/10;
1332 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1333 in6_dev->tstamp = jiffies;
1334 send_ifinfo_notify = true;
1337 rtime = ntohl(ra_msg->reachable_time);
1338 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1339 rtime = (rtime*HZ)/1000;
1341 if (rtime < HZ/10)
1342 rtime = HZ/10;
1344 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1345 NEIGH_VAR_SET(in6_dev->nd_parms,
1346 BASE_REACHABLE_TIME, rtime);
1347 NEIGH_VAR_SET(in6_dev->nd_parms,
1348 GC_STALETIME, 3 * rtime);
1349 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1350 in6_dev->tstamp = jiffies;
1351 send_ifinfo_notify = true;
1357 * Send a notify if RA changed managed/otherconf flags or timer settings
1359 if (send_ifinfo_notify)
1360 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1362 skip_linkparms:
1365 * Process options.
1368 if (!neigh)
1369 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1370 skb->dev, 1);
1371 if (neigh) {
1372 u8 *lladdr = NULL;
1373 if (ndopts.nd_opts_src_lladdr) {
1374 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1375 skb->dev);
1376 if (!lladdr) {
1377 ND_PRINTK(2, warn,
1378 "RA: invalid link-layer address length\n");
1379 goto out;
1382 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1383 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1384 NEIGH_UPDATE_F_OVERRIDE|
1385 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1386 NEIGH_UPDATE_F_ISROUTER,
1387 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1390 if (!ipv6_accept_ra(in6_dev)) {
1391 ND_PRINTK(2, info,
1392 "RA: %s, accept_ra is false for dev: %s\n",
1393 __func__, skb->dev->name);
1394 goto out;
1397 #ifdef CONFIG_IPV6_ROUTE_INFO
1398 if (!in6_dev->cnf.accept_ra_from_local &&
1399 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1400 in6_dev->dev, 0)) {
1401 ND_PRINTK(2, info,
1402 "RA from local address detected on dev: %s: router info ignored.\n",
1403 skb->dev->name);
1404 goto skip_routeinfo;
1407 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1408 struct nd_opt_hdr *p;
1409 for (p = ndopts.nd_opts_ri;
1411 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1412 struct route_info *ri = (struct route_info *)p;
1413 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1414 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1415 ri->prefix_len == 0)
1416 continue;
1417 #endif
1418 if (ri->prefix_len == 0 &&
1419 !in6_dev->cnf.accept_ra_defrtr)
1420 continue;
1421 if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1422 continue;
1423 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1424 continue;
1425 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1426 &ipv6_hdr(skb)->saddr);
1430 skip_routeinfo:
1431 #endif
1433 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1434 /* skip link-specific ndopts from interior routers */
1435 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1436 ND_PRINTK(2, info,
1437 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1438 __func__, skb->dev->name);
1439 goto out;
1441 #endif
1443 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1444 struct nd_opt_hdr *p;
1445 for (p = ndopts.nd_opts_pi;
1447 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1448 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1449 (p->nd_opt_len) << 3,
1450 ndopts.nd_opts_src_lladdr != NULL);
1454 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1455 __be32 n;
1456 u32 mtu;
1458 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1459 mtu = ntohl(n);
1461 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1462 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1463 } else if (in6_dev->cnf.mtu6 != mtu) {
1464 in6_dev->cnf.mtu6 = mtu;
1466 if (rt)
1467 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
1469 rt6_mtu_change(skb->dev, mtu);
1473 if (ndopts.nd_useropts) {
1474 struct nd_opt_hdr *p;
1475 for (p = ndopts.nd_useropts;
1477 p = ndisc_next_useropt(skb->dev, p,
1478 ndopts.nd_useropts_end)) {
1479 ndisc_ra_useropt(skb, p);
1483 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1484 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1486 out:
1487 ip6_rt_put(rt);
1488 if (neigh)
1489 neigh_release(neigh);
1492 static void ndisc_redirect_rcv(struct sk_buff *skb)
1494 u8 *hdr;
1495 struct ndisc_options ndopts;
1496 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1497 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1498 offsetof(struct rd_msg, opt));
1500 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1501 switch (skb->ndisc_nodetype) {
1502 case NDISC_NODETYPE_HOST:
1503 case NDISC_NODETYPE_NODEFAULT:
1504 ND_PRINTK(2, warn,
1505 "Redirect: from host or unauthorized router\n");
1506 return;
1508 #endif
1510 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1511 ND_PRINTK(2, warn,
1512 "Redirect: source address is not link-local\n");
1513 return;
1516 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1517 return;
1519 if (!ndopts.nd_opts_rh) {
1520 ip6_redirect_no_header(skb, dev_net(skb->dev),
1521 skb->dev->ifindex, 0);
1522 return;
1525 hdr = (u8 *)ndopts.nd_opts_rh;
1526 hdr += 8;
1527 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1528 return;
1530 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1533 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1534 struct sk_buff *orig_skb,
1535 int rd_len)
1537 u8 *opt = skb_put(skb, rd_len);
1539 memset(opt, 0, 8);
1540 *(opt++) = ND_OPT_REDIRECT_HDR;
1541 *(opt++) = (rd_len >> 3);
1542 opt += 6;
1544 memcpy(opt, ipv6_hdr(orig_skb), rd_len - 8);
1547 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1549 struct net_device *dev = skb->dev;
1550 struct net *net = dev_net(dev);
1551 struct sock *sk = net->ipv6.ndisc_sk;
1552 int optlen = 0;
1553 struct inet_peer *peer;
1554 struct sk_buff *buff;
1555 struct rd_msg *msg;
1556 struct in6_addr saddr_buf;
1557 struct rt6_info *rt;
1558 struct dst_entry *dst;
1559 struct flowi6 fl6;
1560 int rd_len;
1561 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1562 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1563 bool ret;
1565 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1566 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1567 dev->name);
1568 return;
1571 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1572 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1573 ND_PRINTK(2, warn,
1574 "Redirect: target address is not link-local unicast\n");
1575 return;
1578 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1579 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1581 dst = ip6_route_output(net, NULL, &fl6);
1582 if (dst->error) {
1583 dst_release(dst);
1584 return;
1586 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1587 if (IS_ERR(dst))
1588 return;
1590 rt = (struct rt6_info *) dst;
1592 if (rt->rt6i_flags & RTF_GATEWAY) {
1593 ND_PRINTK(2, warn,
1594 "Redirect: destination is not a neighbour\n");
1595 goto release;
1597 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1598 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1599 if (peer)
1600 inet_putpeer(peer);
1601 if (!ret)
1602 goto release;
1604 if (dev->addr_len) {
1605 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1606 if (!neigh) {
1607 ND_PRINTK(2, warn,
1608 "Redirect: no neigh for target address\n");
1609 goto release;
1612 read_lock_bh(&neigh->lock);
1613 if (neigh->nud_state & NUD_VALID) {
1614 memcpy(ha_buf, neigh->ha, dev->addr_len);
1615 read_unlock_bh(&neigh->lock);
1616 ha = ha_buf;
1617 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1618 ops_data_buf,
1619 &ops_data);
1620 } else
1621 read_unlock_bh(&neigh->lock);
1623 neigh_release(neigh);
1626 rd_len = min_t(unsigned int,
1627 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1628 skb->len + 8);
1629 rd_len &= ~0x7;
1630 optlen += rd_len;
1632 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1633 if (!buff)
1634 goto release;
1636 msg = skb_put(buff, sizeof(*msg));
1637 *msg = (struct rd_msg) {
1638 .icmph = {
1639 .icmp6_type = NDISC_REDIRECT,
1641 .target = *target,
1642 .dest = ipv6_hdr(skb)->daddr,
1646 * include target_address option
1649 if (ha)
1650 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1653 * build redirect option and copy skb over to the new packet.
1656 if (rd_len)
1657 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1659 skb_dst_set(buff, dst);
1660 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1661 return;
1663 release:
1664 dst_release(dst);
1667 static void pndisc_redo(struct sk_buff *skb)
1669 ndisc_recv_ns(skb);
1670 kfree_skb(skb);
1673 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1675 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1677 if (!idev)
1678 return true;
1679 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1680 idev->cnf.suppress_frag_ndisc) {
1681 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1682 return true;
1684 return false;
1687 int ndisc_rcv(struct sk_buff *skb)
1689 struct nd_msg *msg;
1691 if (ndisc_suppress_frag_ndisc(skb))
1692 return 0;
1694 if (skb_linearize(skb))
1695 return 0;
1697 msg = (struct nd_msg *)skb_transport_header(skb);
1699 __skb_push(skb, skb->data - skb_transport_header(skb));
1701 if (ipv6_hdr(skb)->hop_limit != 255) {
1702 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1703 ipv6_hdr(skb)->hop_limit);
1704 return 0;
1707 if (msg->icmph.icmp6_code != 0) {
1708 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1709 msg->icmph.icmp6_code);
1710 return 0;
1713 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1715 switch (msg->icmph.icmp6_type) {
1716 case NDISC_NEIGHBOUR_SOLICITATION:
1717 ndisc_recv_ns(skb);
1718 break;
1720 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1721 ndisc_recv_na(skb);
1722 break;
1724 case NDISC_ROUTER_SOLICITATION:
1725 ndisc_recv_rs(skb);
1726 break;
1728 case NDISC_ROUTER_ADVERTISEMENT:
1729 ndisc_router_discovery(skb);
1730 break;
1732 case NDISC_REDIRECT:
1733 ndisc_redirect_rcv(skb);
1734 break;
1737 return 0;
1740 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1742 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1743 struct netdev_notifier_change_info *change_info;
1744 struct net *net = dev_net(dev);
1745 struct inet6_dev *idev;
1747 switch (event) {
1748 case NETDEV_CHANGEADDR:
1749 neigh_changeaddr(&nd_tbl, dev);
1750 fib6_run_gc(0, net, false);
1751 /* fallthrough */
1752 case NETDEV_UP:
1753 idev = in6_dev_get(dev);
1754 if (!idev)
1755 break;
1756 if (idev->cnf.ndisc_notify ||
1757 net->ipv6.devconf_all->ndisc_notify)
1758 ndisc_send_unsol_na(dev);
1759 in6_dev_put(idev);
1760 break;
1761 case NETDEV_CHANGE:
1762 change_info = ptr;
1763 if (change_info->flags_changed & IFF_NOARP)
1764 neigh_changeaddr(&nd_tbl, dev);
1765 break;
1766 case NETDEV_DOWN:
1767 neigh_ifdown(&nd_tbl, dev);
1768 fib6_run_gc(0, net, false);
1769 break;
1770 case NETDEV_NOTIFY_PEERS:
1771 ndisc_send_unsol_na(dev);
1772 break;
1773 default:
1774 break;
1777 return NOTIFY_DONE;
1780 static struct notifier_block ndisc_netdev_notifier = {
1781 .notifier_call = ndisc_netdev_event,
1784 #ifdef CONFIG_SYSCTL
1785 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1786 const char *func, const char *dev_name)
1788 static char warncomm[TASK_COMM_LEN];
1789 static int warned;
1790 if (strcmp(warncomm, current->comm) && warned < 5) {
1791 strcpy(warncomm, current->comm);
1792 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1793 warncomm, func,
1794 dev_name, ctl->procname,
1795 dev_name, ctl->procname);
1796 warned++;
1800 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1802 struct net_device *dev = ctl->extra1;
1803 struct inet6_dev *idev;
1804 int ret;
1806 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1807 (strcmp(ctl->procname, "base_reachable_time") == 0))
1808 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1810 if (strcmp(ctl->procname, "retrans_time") == 0)
1811 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1813 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1814 ret = neigh_proc_dointvec_jiffies(ctl, write,
1815 buffer, lenp, ppos);
1817 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1818 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1819 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1820 buffer, lenp, ppos);
1821 else
1822 ret = -1;
1824 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1825 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1826 idev->nd_parms->reachable_time =
1827 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1828 idev->tstamp = jiffies;
1829 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1830 in6_dev_put(idev);
1832 return ret;
1836 #endif
1838 static int __net_init ndisc_net_init(struct net *net)
1840 struct ipv6_pinfo *np;
1841 struct sock *sk;
1842 int err;
1844 err = inet_ctl_sock_create(&sk, PF_INET6,
1845 SOCK_RAW, IPPROTO_ICMPV6, net);
1846 if (err < 0) {
1847 ND_PRINTK(0, err,
1848 "NDISC: Failed to initialize the control socket (err %d)\n",
1849 err);
1850 return err;
1853 net->ipv6.ndisc_sk = sk;
1855 np = inet6_sk(sk);
1856 np->hop_limit = 255;
1857 /* Do not loopback ndisc messages */
1858 np->mc_loop = 0;
1860 return 0;
1863 static void __net_exit ndisc_net_exit(struct net *net)
1865 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1868 static struct pernet_operations ndisc_net_ops = {
1869 .init = ndisc_net_init,
1870 .exit = ndisc_net_exit,
1873 int __init ndisc_init(void)
1875 int err;
1877 err = register_pernet_subsys(&ndisc_net_ops);
1878 if (err)
1879 return err;
1881 * Initialize the neighbour table
1883 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1885 #ifdef CONFIG_SYSCTL
1886 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1887 ndisc_ifinfo_sysctl_change);
1888 if (err)
1889 goto out_unregister_pernet;
1890 out:
1891 #endif
1892 return err;
1894 #ifdef CONFIG_SYSCTL
1895 out_unregister_pernet:
1896 unregister_pernet_subsys(&ndisc_net_ops);
1897 goto out;
1898 #endif
1901 int __init ndisc_late_init(void)
1903 return register_netdevice_notifier(&ndisc_netdev_notifier);
1906 void ndisc_late_cleanup(void)
1908 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1911 void ndisc_cleanup(void)
1913 #ifdef CONFIG_SYSCTL
1914 neigh_sysctl_unregister(&nd_tbl.parms);
1915 #endif
1916 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1917 unregister_pernet_subsys(&ndisc_net_ops);