ARM: dma-api: fix max_pfn off-by-one error in __dma_supported()
[linux/fpc-iii.git] / net / ipv6 / ndisc.c
blob53caf59c591e303c78cc70a88f77b9c2ec1403ba
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Neighbour Discovery for IPv6
4 * Linux INET6 implementation
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Mike Shaver <shaver@ingenia.com>
9 */
12 * Changes:
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
26 #define pr_fmt(fmt) "ICMPv6: " fmt
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
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>
51 #include <net/sock.h>
52 #include <net/snmp.h>
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>
61 #include <net/netlink.h>
62 #include <linux/rtnetlink.h>
64 #include <net/flow.h>
65 #include <net/ip6_checksum.h>
66 #include <net/inet_common.h>
67 #include <linux/proc_fs.h>
69 #include <linux/netfilter.h>
70 #include <linux/netfilter_ipv6.h>
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);
85 static const struct neigh_ops ndisc_generic_ops = {
86 .family = AF_INET6,
87 .solicit = ndisc_solicit,
88 .error_report = ndisc_error_report,
89 .output = neigh_resolve_output,
90 .connected_output = neigh_connected_output,
93 static const struct neigh_ops ndisc_hh_ops = {
94 .family = AF_INET6,
95 .solicit = ndisc_solicit,
96 .error_report = ndisc_error_report,
97 .output = neigh_resolve_output,
98 .connected_output = neigh_resolve_output,
102 static const struct neigh_ops ndisc_direct_ops = {
103 .family = AF_INET6,
104 .output = neigh_direct_output,
105 .connected_output = neigh_direct_output,
108 struct neigh_table nd_tbl = {
109 .family = AF_INET6,
110 .key_len = sizeof(struct in6_addr),
111 .protocol = cpu_to_be16(ETH_P_IPV6),
112 .hash = ndisc_hash,
113 .key_eq = ndisc_key_eq,
114 .constructor = ndisc_constructor,
115 .pconstructor = pndisc_constructor,
116 .pdestructor = pndisc_destructor,
117 .proxy_redo = pndisc_redo,
118 .allow_add = ndisc_allow_add,
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] = SK_WMEM_MAX,
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 opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
200 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
203 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
204 struct nd_opt_hdr *cur,
205 struct nd_opt_hdr *end)
207 if (!cur || !end || cur >= end)
208 return NULL;
209 do {
210 cur = ((void *)cur) + (cur->nd_opt_len << 3);
211 } while (cur < end && !ndisc_is_useropt(dev, cur));
212 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
215 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
216 u8 *opt, int opt_len,
217 struct ndisc_options *ndopts)
219 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
221 if (!nd_opt || opt_len < 0 || !ndopts)
222 return NULL;
223 memset(ndopts, 0, sizeof(*ndopts));
224 while (opt_len) {
225 int l;
226 if (opt_len < sizeof(struct nd_opt_hdr))
227 return NULL;
228 l = nd_opt->nd_opt_len << 3;
229 if (opt_len < l || l == 0)
230 return NULL;
231 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
232 goto next_opt;
233 switch (nd_opt->nd_opt_type) {
234 case ND_OPT_SOURCE_LL_ADDR:
235 case ND_OPT_TARGET_LL_ADDR:
236 case ND_OPT_MTU:
237 case ND_OPT_NONCE:
238 case ND_OPT_REDIRECT_HDR:
239 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
240 ND_PRINTK(2, warn,
241 "%s: duplicated ND6 option found: type=%d\n",
242 __func__, nd_opt->nd_opt_type);
243 } else {
244 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
246 break;
247 case ND_OPT_PREFIX_INFO:
248 ndopts->nd_opts_pi_end = nd_opt;
249 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
250 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
251 break;
252 #ifdef CONFIG_IPV6_ROUTE_INFO
253 case ND_OPT_ROUTE_INFO:
254 ndopts->nd_opts_ri_end = nd_opt;
255 if (!ndopts->nd_opts_ri)
256 ndopts->nd_opts_ri = nd_opt;
257 break;
258 #endif
259 default:
260 if (ndisc_is_useropt(dev, nd_opt)) {
261 ndopts->nd_useropts_end = nd_opt;
262 if (!ndopts->nd_useropts)
263 ndopts->nd_useropts = nd_opt;
264 } else {
266 * Unknown options must be silently ignored,
267 * to accommodate future extension to the
268 * protocol.
270 ND_PRINTK(2, notice,
271 "%s: ignored unsupported option; type=%d, len=%d\n",
272 __func__,
273 nd_opt->nd_opt_type,
274 nd_opt->nd_opt_len);
277 next_opt:
278 opt_len -= l;
279 nd_opt = ((void *)nd_opt) + l;
281 return ndopts;
284 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
286 switch (dev->type) {
287 case ARPHRD_ETHER:
288 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
289 case ARPHRD_FDDI:
290 ipv6_eth_mc_map(addr, buf);
291 return 0;
292 case ARPHRD_ARCNET:
293 ipv6_arcnet_mc_map(addr, buf);
294 return 0;
295 case ARPHRD_INFINIBAND:
296 ipv6_ib_mc_map(addr, dev->broadcast, buf);
297 return 0;
298 case ARPHRD_IPGRE:
299 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
300 default:
301 if (dir) {
302 memcpy(buf, dev->broadcast, dev->addr_len);
303 return 0;
306 return -EINVAL;
308 EXPORT_SYMBOL(ndisc_mc_map);
310 static u32 ndisc_hash(const void *pkey,
311 const struct net_device *dev,
312 __u32 *hash_rnd)
314 return ndisc_hashfn(pkey, dev, hash_rnd);
317 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
319 return neigh_key_eq128(n, pkey);
322 static int ndisc_constructor(struct neighbour *neigh)
324 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
325 struct net_device *dev = neigh->dev;
326 struct inet6_dev *in6_dev;
327 struct neigh_parms *parms;
328 bool is_multicast = ipv6_addr_is_multicast(addr);
330 in6_dev = in6_dev_get(dev);
331 if (!in6_dev) {
332 return -EINVAL;
335 parms = in6_dev->nd_parms;
336 __neigh_parms_put(neigh->parms);
337 neigh->parms = neigh_parms_clone(parms);
339 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
340 if (!dev->header_ops) {
341 neigh->nud_state = NUD_NOARP;
342 neigh->ops = &ndisc_direct_ops;
343 neigh->output = neigh_direct_output;
344 } else {
345 if (is_multicast) {
346 neigh->nud_state = NUD_NOARP;
347 ndisc_mc_map(addr, neigh->ha, dev, 1);
348 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
349 neigh->nud_state = NUD_NOARP;
350 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
351 if (dev->flags&IFF_LOOPBACK)
352 neigh->type = RTN_LOCAL;
353 } else if (dev->flags&IFF_POINTOPOINT) {
354 neigh->nud_state = NUD_NOARP;
355 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
357 if (dev->header_ops->cache)
358 neigh->ops = &ndisc_hh_ops;
359 else
360 neigh->ops = &ndisc_generic_ops;
361 if (neigh->nud_state&NUD_VALID)
362 neigh->output = neigh->ops->connected_output;
363 else
364 neigh->output = neigh->ops->output;
366 in6_dev_put(in6_dev);
367 return 0;
370 static int pndisc_constructor(struct pneigh_entry *n)
372 struct in6_addr *addr = (struct in6_addr *)&n->key;
373 struct in6_addr maddr;
374 struct net_device *dev = n->dev;
376 if (!dev || !__in6_dev_get(dev))
377 return -EINVAL;
378 addrconf_addr_solict_mult(addr, &maddr);
379 ipv6_dev_mc_inc(dev, &maddr);
380 return 0;
383 static void pndisc_destructor(struct pneigh_entry *n)
385 struct in6_addr *addr = (struct in6_addr *)&n->key;
386 struct in6_addr maddr;
387 struct net_device *dev = n->dev;
389 if (!dev || !__in6_dev_get(dev))
390 return;
391 addrconf_addr_solict_mult(addr, &maddr);
392 ipv6_dev_mc_dec(dev, &maddr);
395 /* called with rtnl held */
396 static bool ndisc_allow_add(const struct net_device *dev,
397 struct netlink_ext_ack *extack)
399 struct inet6_dev *idev = __in6_dev_get(dev);
401 if (!idev || idev->cnf.disable_ipv6) {
402 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
403 return false;
406 return true;
409 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
410 int len)
412 int hlen = LL_RESERVED_SPACE(dev);
413 int tlen = dev->needed_tailroom;
414 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
415 struct sk_buff *skb;
417 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
418 if (!skb) {
419 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
420 __func__);
421 return NULL;
424 skb->protocol = htons(ETH_P_IPV6);
425 skb->dev = dev;
427 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
428 skb_reset_transport_header(skb);
430 /* Manually assign socket ownership as we avoid calling
431 * sock_alloc_send_pskb() to bypass wmem buffer limits
433 skb_set_owner_w(skb, sk);
435 return skb;
438 static void ip6_nd_hdr(struct sk_buff *skb,
439 const struct in6_addr *saddr,
440 const struct in6_addr *daddr,
441 int hop_limit, int len)
443 struct ipv6hdr *hdr;
444 struct inet6_dev *idev;
445 unsigned tclass;
447 rcu_read_lock();
448 idev = __in6_dev_get(skb->dev);
449 tclass = idev ? idev->cnf.ndisc_tclass : 0;
450 rcu_read_unlock();
452 skb_push(skb, sizeof(*hdr));
453 skb_reset_network_header(skb);
454 hdr = ipv6_hdr(skb);
456 ip6_flow_hdr(hdr, tclass, 0);
458 hdr->payload_len = htons(len);
459 hdr->nexthdr = IPPROTO_ICMPV6;
460 hdr->hop_limit = hop_limit;
462 hdr->saddr = *saddr;
463 hdr->daddr = *daddr;
466 static void ndisc_send_skb(struct sk_buff *skb,
467 const struct in6_addr *daddr,
468 const struct in6_addr *saddr)
470 struct dst_entry *dst = skb_dst(skb);
471 struct net *net = dev_net(skb->dev);
472 struct sock *sk = net->ipv6.ndisc_sk;
473 struct inet6_dev *idev;
474 int err;
475 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
476 u8 type;
478 type = icmp6h->icmp6_type;
480 if (!dst) {
481 struct flowi6 fl6;
482 int oif = skb->dev->ifindex;
484 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
485 dst = icmp6_dst_alloc(skb->dev, &fl6);
486 if (IS_ERR(dst)) {
487 kfree_skb(skb);
488 return;
491 skb_dst_set(skb, dst);
494 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
495 IPPROTO_ICMPV6,
496 csum_partial(icmp6h,
497 skb->len, 0));
499 ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
501 rcu_read_lock();
502 idev = __in6_dev_get(dst->dev);
503 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
505 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
506 net, sk, skb, NULL, dst->dev,
507 dst_output);
508 if (!err) {
509 ICMP6MSGOUT_INC_STATS(net, idev, type);
510 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
513 rcu_read_unlock();
516 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
517 const struct in6_addr *solicited_addr,
518 bool router, bool solicited, bool override, bool inc_opt)
520 struct sk_buff *skb;
521 struct in6_addr tmpaddr;
522 struct inet6_ifaddr *ifp;
523 const struct in6_addr *src_addr;
524 struct nd_msg *msg;
525 int optlen = 0;
527 /* for anycast or proxy, solicited_addr != src_addr */
528 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
529 if (ifp) {
530 src_addr = solicited_addr;
531 if (ifp->flags & IFA_F_OPTIMISTIC)
532 override = false;
533 inc_opt |= ifp->idev->cnf.force_tllao;
534 in6_ifa_put(ifp);
535 } else {
536 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
537 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
538 &tmpaddr))
539 return;
540 src_addr = &tmpaddr;
543 if (!dev->addr_len)
544 inc_opt = false;
545 if (inc_opt)
546 optlen += ndisc_opt_addr_space(dev,
547 NDISC_NEIGHBOUR_ADVERTISEMENT);
549 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
550 if (!skb)
551 return;
553 msg = skb_put(skb, sizeof(*msg));
554 *msg = (struct nd_msg) {
555 .icmph = {
556 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
557 .icmp6_router = router,
558 .icmp6_solicited = solicited,
559 .icmp6_override = override,
561 .target = *solicited_addr,
564 if (inc_opt)
565 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
566 dev->dev_addr,
567 NDISC_NEIGHBOUR_ADVERTISEMENT);
569 ndisc_send_skb(skb, daddr, src_addr);
572 static void ndisc_send_unsol_na(struct net_device *dev)
574 struct inet6_dev *idev;
575 struct inet6_ifaddr *ifa;
577 idev = in6_dev_get(dev);
578 if (!idev)
579 return;
581 read_lock_bh(&idev->lock);
582 list_for_each_entry(ifa, &idev->addr_list, if_list) {
583 /* skip tentative addresses until dad completes */
584 if (ifa->flags & IFA_F_TENTATIVE &&
585 !(ifa->flags & IFA_F_OPTIMISTIC))
586 continue;
588 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
589 /*router=*/ !!idev->cnf.forwarding,
590 /*solicited=*/ false, /*override=*/ true,
591 /*inc_opt=*/ true);
593 read_unlock_bh(&idev->lock);
595 in6_dev_put(idev);
598 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
599 const struct in6_addr *daddr, const struct in6_addr *saddr,
600 u64 nonce)
602 struct sk_buff *skb;
603 struct in6_addr addr_buf;
604 int inc_opt = dev->addr_len;
605 int optlen = 0;
606 struct nd_msg *msg;
608 if (!saddr) {
609 if (ipv6_get_lladdr(dev, &addr_buf,
610 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
611 return;
612 saddr = &addr_buf;
615 if (ipv6_addr_any(saddr))
616 inc_opt = false;
617 if (inc_opt)
618 optlen += ndisc_opt_addr_space(dev,
619 NDISC_NEIGHBOUR_SOLICITATION);
620 if (nonce != 0)
621 optlen += 8;
623 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
624 if (!skb)
625 return;
627 msg = skb_put(skb, sizeof(*msg));
628 *msg = (struct nd_msg) {
629 .icmph = {
630 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
632 .target = *solicit,
635 if (inc_opt)
636 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
637 dev->dev_addr,
638 NDISC_NEIGHBOUR_SOLICITATION);
639 if (nonce != 0) {
640 u8 *opt = skb_put(skb, 8);
642 opt[0] = ND_OPT_NONCE;
643 opt[1] = 8 >> 3;
644 memcpy(opt + 2, &nonce, 6);
647 ndisc_send_skb(skb, daddr, saddr);
650 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
651 const struct in6_addr *daddr)
653 struct sk_buff *skb;
654 struct rs_msg *msg;
655 int send_sllao = dev->addr_len;
656 int optlen = 0;
658 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
660 * According to section 2.2 of RFC 4429, we must not
661 * send router solicitations with a sllao from
662 * optimistic addresses, but we may send the solicitation
663 * if we don't include the sllao. So here we check
664 * if our address is optimistic, and if so, we
665 * suppress the inclusion of the sllao.
667 if (send_sllao) {
668 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
669 dev, 1);
670 if (ifp) {
671 if (ifp->flags & IFA_F_OPTIMISTIC) {
672 send_sllao = 0;
674 in6_ifa_put(ifp);
675 } else {
676 send_sllao = 0;
679 #endif
680 if (send_sllao)
681 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
683 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
684 if (!skb)
685 return;
687 msg = skb_put(skb, sizeof(*msg));
688 *msg = (struct rs_msg) {
689 .icmph = {
690 .icmp6_type = NDISC_ROUTER_SOLICITATION,
694 if (send_sllao)
695 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
696 dev->dev_addr,
697 NDISC_ROUTER_SOLICITATION);
699 ndisc_send_skb(skb, daddr, saddr);
703 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
706 * "The sender MUST return an ICMP
707 * destination unreachable"
709 dst_link_failure(skb);
710 kfree_skb(skb);
713 /* Called with locked neigh: either read or both */
715 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
717 struct in6_addr *saddr = NULL;
718 struct in6_addr mcaddr;
719 struct net_device *dev = neigh->dev;
720 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
721 int probes = atomic_read(&neigh->probes);
723 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
724 dev, false, 1,
725 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
726 saddr = &ipv6_hdr(skb)->saddr;
727 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
728 if (probes < 0) {
729 if (!(neigh->nud_state & NUD_VALID)) {
730 ND_PRINTK(1, dbg,
731 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
732 __func__, target);
734 ndisc_send_ns(dev, target, target, saddr, 0);
735 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
736 neigh_app_ns(neigh);
737 } else {
738 addrconf_addr_solict_mult(target, &mcaddr);
739 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
743 static int pndisc_is_router(const void *pkey,
744 struct net_device *dev)
746 struct pneigh_entry *n;
747 int ret = -1;
749 read_lock_bh(&nd_tbl.lock);
750 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
751 if (n)
752 ret = !!(n->flags & NTF_ROUTER);
753 read_unlock_bh(&nd_tbl.lock);
755 return ret;
758 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
759 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
760 struct ndisc_options *ndopts)
762 neigh_update(neigh, lladdr, new, flags, 0);
763 /* report ndisc ops about neighbour update */
764 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
767 static void ndisc_recv_ns(struct sk_buff *skb)
769 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
770 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
771 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
772 u8 *lladdr = NULL;
773 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
774 offsetof(struct nd_msg, opt));
775 struct ndisc_options ndopts;
776 struct net_device *dev = skb->dev;
777 struct inet6_ifaddr *ifp;
778 struct inet6_dev *idev = NULL;
779 struct neighbour *neigh;
780 int dad = ipv6_addr_any(saddr);
781 bool inc;
782 int is_router = -1;
783 u64 nonce = 0;
785 if (skb->len < sizeof(struct nd_msg)) {
786 ND_PRINTK(2, warn, "NS: packet too short\n");
787 return;
790 if (ipv6_addr_is_multicast(&msg->target)) {
791 ND_PRINTK(2, warn, "NS: multicast target address\n");
792 return;
796 * RFC2461 7.1.1:
797 * DAD has to be destined for solicited node multicast address.
799 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
800 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
801 return;
804 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
805 ND_PRINTK(2, warn, "NS: invalid ND options\n");
806 return;
809 if (ndopts.nd_opts_src_lladdr) {
810 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
811 if (!lladdr) {
812 ND_PRINTK(2, warn,
813 "NS: invalid link-layer address length\n");
814 return;
817 /* RFC2461 7.1.1:
818 * If the IP source address is the unspecified address,
819 * there MUST NOT be source link-layer address option
820 * in the message.
822 if (dad) {
823 ND_PRINTK(2, warn,
824 "NS: bad DAD packet (link-layer address option)\n");
825 return;
828 if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
829 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
831 inc = ipv6_addr_is_multicast(daddr);
833 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
834 if (ifp) {
835 have_ifp:
836 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
837 if (dad) {
838 if (nonce != 0 && ifp->dad_nonce == nonce) {
839 u8 *np = (u8 *)&nonce;
840 /* Matching nonce if looped back */
841 ND_PRINTK(2, notice,
842 "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
843 ifp->idev->dev->name,
844 &ifp->addr, np);
845 goto out;
848 * We are colliding with another node
849 * who is doing DAD
850 * so fail our DAD process
852 addrconf_dad_failure(skb, ifp);
853 return;
854 } else {
856 * This is not a dad solicitation.
857 * If we are an optimistic node,
858 * we should respond.
859 * Otherwise, we should ignore it.
861 if (!(ifp->flags & IFA_F_OPTIMISTIC))
862 goto out;
866 idev = ifp->idev;
867 } else {
868 struct net *net = dev_net(dev);
870 /* perhaps an address on the master device */
871 if (netif_is_l3_slave(dev)) {
872 struct net_device *mdev;
874 mdev = netdev_master_upper_dev_get_rcu(dev);
875 if (mdev) {
876 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
877 if (ifp)
878 goto have_ifp;
882 idev = in6_dev_get(dev);
883 if (!idev) {
884 /* XXX: count this drop? */
885 return;
888 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
889 (idev->cnf.forwarding &&
890 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
891 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
892 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
893 skb->pkt_type != PACKET_HOST &&
894 inc &&
895 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
897 * for anycast or proxy,
898 * sender should delay its response
899 * by a random time between 0 and
900 * MAX_ANYCAST_DELAY_TIME seconds.
901 * (RFC2461) -- yoshfuji
903 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
904 if (n)
905 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
906 goto out;
908 } else
909 goto out;
912 if (is_router < 0)
913 is_router = idev->cnf.forwarding;
915 if (dad) {
916 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
917 !!is_router, false, (ifp != NULL), true);
918 goto out;
921 if (inc)
922 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
923 else
924 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
927 * update / create cache entry
928 * for the source address
930 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
931 !inc || lladdr || !dev->addr_len);
932 if (neigh)
933 ndisc_update(dev, neigh, lladdr, NUD_STALE,
934 NEIGH_UPDATE_F_WEAK_OVERRIDE|
935 NEIGH_UPDATE_F_OVERRIDE,
936 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
937 if (neigh || !dev->header_ops) {
938 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
939 true, (ifp != NULL && inc), inc);
940 if (neigh)
941 neigh_release(neigh);
944 out:
945 if (ifp)
946 in6_ifa_put(ifp);
947 else
948 in6_dev_put(idev);
951 static void ndisc_recv_na(struct sk_buff *skb)
953 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
954 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
955 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
956 u8 *lladdr = NULL;
957 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
958 offsetof(struct nd_msg, opt));
959 struct ndisc_options ndopts;
960 struct net_device *dev = skb->dev;
961 struct inet6_dev *idev = __in6_dev_get(dev);
962 struct inet6_ifaddr *ifp;
963 struct neighbour *neigh;
965 if (skb->len < sizeof(struct nd_msg)) {
966 ND_PRINTK(2, warn, "NA: packet too short\n");
967 return;
970 if (ipv6_addr_is_multicast(&msg->target)) {
971 ND_PRINTK(2, warn, "NA: target address is multicast\n");
972 return;
975 if (ipv6_addr_is_multicast(daddr) &&
976 msg->icmph.icmp6_solicited) {
977 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
978 return;
981 /* For some 802.11 wireless deployments (and possibly other networks),
982 * there will be a NA proxy and unsolicitd packets are attacks
983 * and thus should not be accepted.
985 if (!msg->icmph.icmp6_solicited && idev &&
986 idev->cnf.drop_unsolicited_na)
987 return;
989 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
990 ND_PRINTK(2, warn, "NS: invalid ND option\n");
991 return;
993 if (ndopts.nd_opts_tgt_lladdr) {
994 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
995 if (!lladdr) {
996 ND_PRINTK(2, warn,
997 "NA: invalid link-layer address length\n");
998 return;
1001 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1002 if (ifp) {
1003 if (skb->pkt_type != PACKET_LOOPBACK
1004 && (ifp->flags & IFA_F_TENTATIVE)) {
1005 addrconf_dad_failure(skb, ifp);
1006 return;
1008 /* What should we make now? The advertisement
1009 is invalid, but ndisc specs say nothing
1010 about it. It could be misconfiguration, or
1011 an smart proxy agent tries to help us :-)
1013 We should not print the error if NA has been
1014 received from loopback - it is just our own
1015 unsolicited advertisement.
1017 if (skb->pkt_type != PACKET_LOOPBACK)
1018 ND_PRINTK(1, warn,
1019 "NA: %pM advertised our address %pI6c on %s!\n",
1020 eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1021 in6_ifa_put(ifp);
1022 return;
1024 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1026 if (neigh) {
1027 u8 old_flags = neigh->flags;
1028 struct net *net = dev_net(dev);
1030 if (neigh->nud_state & NUD_FAILED)
1031 goto out;
1034 * Don't update the neighbor cache entry on a proxy NA from
1035 * ourselves because either the proxied node is off link or it
1036 * has already sent a NA to us.
1038 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1039 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1040 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1041 /* XXX: idev->cnf.proxy_ndp */
1042 goto out;
1045 ndisc_update(dev, neigh, lladdr,
1046 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1047 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1048 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1049 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1050 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1051 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1053 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1055 * Change: router to host
1057 rt6_clean_tohost(dev_net(dev), saddr);
1060 out:
1061 neigh_release(neigh);
1065 static void ndisc_recv_rs(struct sk_buff *skb)
1067 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1068 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1069 struct neighbour *neigh;
1070 struct inet6_dev *idev;
1071 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1072 struct ndisc_options ndopts;
1073 u8 *lladdr = NULL;
1075 if (skb->len < sizeof(*rs_msg))
1076 return;
1078 idev = __in6_dev_get(skb->dev);
1079 if (!idev) {
1080 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1081 return;
1084 /* Don't accept RS if we're not in router mode */
1085 if (!idev->cnf.forwarding)
1086 goto out;
1089 * Don't update NCE if src = ::;
1090 * this implies that the source node has no ip address assigned yet.
1092 if (ipv6_addr_any(saddr))
1093 goto out;
1095 /* Parse ND options */
1096 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1097 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1098 goto out;
1101 if (ndopts.nd_opts_src_lladdr) {
1102 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1103 skb->dev);
1104 if (!lladdr)
1105 goto out;
1108 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1109 if (neigh) {
1110 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1111 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1112 NEIGH_UPDATE_F_OVERRIDE|
1113 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1114 NDISC_ROUTER_SOLICITATION, &ndopts);
1115 neigh_release(neigh);
1117 out:
1118 return;
1121 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1123 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1124 struct sk_buff *skb;
1125 struct nlmsghdr *nlh;
1126 struct nduseroptmsg *ndmsg;
1127 struct net *net = dev_net(ra->dev);
1128 int err;
1129 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1130 + (opt->nd_opt_len << 3));
1131 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1133 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1134 if (!skb) {
1135 err = -ENOBUFS;
1136 goto errout;
1139 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1140 if (!nlh) {
1141 goto nla_put_failure;
1144 ndmsg = nlmsg_data(nlh);
1145 ndmsg->nduseropt_family = AF_INET6;
1146 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1147 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1148 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1149 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1151 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1153 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1154 goto nla_put_failure;
1155 nlmsg_end(skb, nlh);
1157 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1158 return;
1160 nla_put_failure:
1161 nlmsg_free(skb);
1162 err = -EMSGSIZE;
1163 errout:
1164 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1167 static void ndisc_router_discovery(struct sk_buff *skb)
1169 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1170 struct neighbour *neigh = NULL;
1171 struct inet6_dev *in6_dev;
1172 struct fib6_info *rt = NULL;
1173 struct net *net;
1174 int lifetime;
1175 struct ndisc_options ndopts;
1176 int optlen;
1177 unsigned int pref = 0;
1178 __u32 old_if_flags;
1179 bool send_ifinfo_notify = false;
1181 __u8 *opt = (__u8 *)(ra_msg + 1);
1183 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1184 sizeof(struct ra_msg);
1186 ND_PRINTK(2, info,
1187 "RA: %s, dev: %s\n",
1188 __func__, skb->dev->name);
1189 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1190 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1191 return;
1193 if (optlen < 0) {
1194 ND_PRINTK(2, warn, "RA: packet too short\n");
1195 return;
1198 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1199 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1200 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1201 return;
1203 #endif
1206 * set the RA_RECV flag in the interface
1209 in6_dev = __in6_dev_get(skb->dev);
1210 if (!in6_dev) {
1211 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1212 skb->dev->name);
1213 return;
1216 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1217 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1218 return;
1221 if (!ipv6_accept_ra(in6_dev)) {
1222 ND_PRINTK(2, info,
1223 "RA: %s, did not accept ra for dev: %s\n",
1224 __func__, skb->dev->name);
1225 goto skip_linkparms;
1228 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1229 /* skip link-specific parameters from interior routers */
1230 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1231 ND_PRINTK(2, info,
1232 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1233 __func__, skb->dev->name);
1234 goto skip_linkparms;
1236 #endif
1238 if (in6_dev->if_flags & IF_RS_SENT) {
1240 * flag that an RA was received after an RS was sent
1241 * out on this interface.
1243 in6_dev->if_flags |= IF_RA_RCVD;
1247 * Remember the managed/otherconf flags from most recently
1248 * received RA message (RFC 2462) -- yoshfuji
1250 old_if_flags = in6_dev->if_flags;
1251 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1252 IF_RA_OTHERCONF)) |
1253 (ra_msg->icmph.icmp6_addrconf_managed ?
1254 IF_RA_MANAGED : 0) |
1255 (ra_msg->icmph.icmp6_addrconf_other ?
1256 IF_RA_OTHERCONF : 0);
1258 if (old_if_flags != in6_dev->if_flags)
1259 send_ifinfo_notify = true;
1261 if (!in6_dev->cnf.accept_ra_defrtr) {
1262 ND_PRINTK(2, info,
1263 "RA: %s, defrtr is false for dev: %s\n",
1264 __func__, skb->dev->name);
1265 goto skip_defrtr;
1268 /* Do not accept RA with source-addr found on local machine unless
1269 * accept_ra_from_local is set to true.
1271 net = dev_net(in6_dev->dev);
1272 if (!in6_dev->cnf.accept_ra_from_local &&
1273 ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1274 ND_PRINTK(2, info,
1275 "RA from local address detected on dev: %s: default router ignored\n",
1276 skb->dev->name);
1277 goto skip_defrtr;
1280 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1282 #ifdef CONFIG_IPV6_ROUTER_PREF
1283 pref = ra_msg->icmph.icmp6_router_pref;
1284 /* 10b is handled as if it were 00b (medium) */
1285 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1286 !in6_dev->cnf.accept_ra_rtr_pref)
1287 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1288 #endif
1289 /* routes added from RAs do not use nexthop objects */
1290 rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1291 if (rt) {
1292 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1293 rt->fib6_nh->fib_nh_dev, NULL,
1294 &ipv6_hdr(skb)->saddr);
1295 if (!neigh) {
1296 ND_PRINTK(0, err,
1297 "RA: %s got default router without neighbour\n",
1298 __func__);
1299 fib6_info_release(rt);
1300 return;
1303 if (rt && lifetime == 0) {
1304 ip6_del_rt(net, rt);
1305 rt = NULL;
1308 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, for dev: %s\n",
1309 rt, lifetime, skb->dev->name);
1310 if (!rt && lifetime) {
1311 ND_PRINTK(3, info, "RA: adding default router\n");
1313 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1314 skb->dev, pref);
1315 if (!rt) {
1316 ND_PRINTK(0, err,
1317 "RA: %s failed to add default route\n",
1318 __func__);
1319 return;
1322 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1323 rt->fib6_nh->fib_nh_dev, NULL,
1324 &ipv6_hdr(skb)->saddr);
1325 if (!neigh) {
1326 ND_PRINTK(0, err,
1327 "RA: %s got default router without neighbour\n",
1328 __func__);
1329 fib6_info_release(rt);
1330 return;
1332 neigh->flags |= NTF_ROUTER;
1333 } else if (rt) {
1334 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1337 if (rt)
1338 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1339 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1340 ra_msg->icmph.icmp6_hop_limit) {
1341 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1342 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1343 fib6_metric_set(rt, RTAX_HOPLIMIT,
1344 ra_msg->icmph.icmp6_hop_limit);
1345 } else {
1346 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1350 skip_defrtr:
1353 * Update Reachable Time and Retrans Timer
1356 if (in6_dev->nd_parms) {
1357 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1359 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1360 rtime = (rtime*HZ)/1000;
1361 if (rtime < HZ/10)
1362 rtime = HZ/10;
1363 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1364 in6_dev->tstamp = jiffies;
1365 send_ifinfo_notify = true;
1368 rtime = ntohl(ra_msg->reachable_time);
1369 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1370 rtime = (rtime*HZ)/1000;
1372 if (rtime < HZ/10)
1373 rtime = HZ/10;
1375 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1376 NEIGH_VAR_SET(in6_dev->nd_parms,
1377 BASE_REACHABLE_TIME, rtime);
1378 NEIGH_VAR_SET(in6_dev->nd_parms,
1379 GC_STALETIME, 3 * rtime);
1380 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1381 in6_dev->tstamp = jiffies;
1382 send_ifinfo_notify = true;
1388 * Send a notify if RA changed managed/otherconf flags or timer settings
1390 if (send_ifinfo_notify)
1391 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1393 skip_linkparms:
1396 * Process options.
1399 if (!neigh)
1400 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1401 skb->dev, 1);
1402 if (neigh) {
1403 u8 *lladdr = NULL;
1404 if (ndopts.nd_opts_src_lladdr) {
1405 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1406 skb->dev);
1407 if (!lladdr) {
1408 ND_PRINTK(2, warn,
1409 "RA: invalid link-layer address length\n");
1410 goto out;
1413 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1414 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1415 NEIGH_UPDATE_F_OVERRIDE|
1416 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1417 NEIGH_UPDATE_F_ISROUTER,
1418 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1421 if (!ipv6_accept_ra(in6_dev)) {
1422 ND_PRINTK(2, info,
1423 "RA: %s, accept_ra is false for dev: %s\n",
1424 __func__, skb->dev->name);
1425 goto out;
1428 #ifdef CONFIG_IPV6_ROUTE_INFO
1429 if (!in6_dev->cnf.accept_ra_from_local &&
1430 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1431 in6_dev->dev, 0)) {
1432 ND_PRINTK(2, info,
1433 "RA from local address detected on dev: %s: router info ignored.\n",
1434 skb->dev->name);
1435 goto skip_routeinfo;
1438 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1439 struct nd_opt_hdr *p;
1440 for (p = ndopts.nd_opts_ri;
1442 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1443 struct route_info *ri = (struct route_info *)p;
1444 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1445 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1446 ri->prefix_len == 0)
1447 continue;
1448 #endif
1449 if (ri->prefix_len == 0 &&
1450 !in6_dev->cnf.accept_ra_defrtr)
1451 continue;
1452 if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1453 continue;
1454 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1455 continue;
1456 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1457 &ipv6_hdr(skb)->saddr);
1461 skip_routeinfo:
1462 #endif
1464 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1465 /* skip link-specific ndopts from interior routers */
1466 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1467 ND_PRINTK(2, info,
1468 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1469 __func__, skb->dev->name);
1470 goto out;
1472 #endif
1474 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1475 struct nd_opt_hdr *p;
1476 for (p = ndopts.nd_opts_pi;
1478 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1479 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1480 (p->nd_opt_len) << 3,
1481 ndopts.nd_opts_src_lladdr != NULL);
1485 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1486 __be32 n;
1487 u32 mtu;
1489 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1490 mtu = ntohl(n);
1492 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1493 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1494 } else if (in6_dev->cnf.mtu6 != mtu) {
1495 in6_dev->cnf.mtu6 = mtu;
1496 fib6_metric_set(rt, RTAX_MTU, mtu);
1497 rt6_mtu_change(skb->dev, mtu);
1501 if (ndopts.nd_useropts) {
1502 struct nd_opt_hdr *p;
1503 for (p = ndopts.nd_useropts;
1505 p = ndisc_next_useropt(skb->dev, p,
1506 ndopts.nd_useropts_end)) {
1507 ndisc_ra_useropt(skb, p);
1511 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1512 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1514 out:
1515 fib6_info_release(rt);
1516 if (neigh)
1517 neigh_release(neigh);
1520 static void ndisc_redirect_rcv(struct sk_buff *skb)
1522 u8 *hdr;
1523 struct ndisc_options ndopts;
1524 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1525 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1526 offsetof(struct rd_msg, opt));
1528 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1529 switch (skb->ndisc_nodetype) {
1530 case NDISC_NODETYPE_HOST:
1531 case NDISC_NODETYPE_NODEFAULT:
1532 ND_PRINTK(2, warn,
1533 "Redirect: from host or unauthorized router\n");
1534 return;
1536 #endif
1538 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1539 ND_PRINTK(2, warn,
1540 "Redirect: source address is not link-local\n");
1541 return;
1544 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1545 return;
1547 if (!ndopts.nd_opts_rh) {
1548 ip6_redirect_no_header(skb, dev_net(skb->dev),
1549 skb->dev->ifindex);
1550 return;
1553 hdr = (u8 *)ndopts.nd_opts_rh;
1554 hdr += 8;
1555 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1556 return;
1558 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1561 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1562 struct sk_buff *orig_skb,
1563 int rd_len)
1565 u8 *opt = skb_put(skb, rd_len);
1567 memset(opt, 0, 8);
1568 *(opt++) = ND_OPT_REDIRECT_HDR;
1569 *(opt++) = (rd_len >> 3);
1570 opt += 6;
1572 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1573 rd_len - 8);
1576 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1578 struct net_device *dev = skb->dev;
1579 struct net *net = dev_net(dev);
1580 struct sock *sk = net->ipv6.ndisc_sk;
1581 int optlen = 0;
1582 struct inet_peer *peer;
1583 struct sk_buff *buff;
1584 struct rd_msg *msg;
1585 struct in6_addr saddr_buf;
1586 struct rt6_info *rt;
1587 struct dst_entry *dst;
1588 struct flowi6 fl6;
1589 int rd_len;
1590 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1591 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1592 bool ret;
1594 if (netif_is_l3_master(skb->dev)) {
1595 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1596 if (!dev)
1597 return;
1600 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1601 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1602 dev->name);
1603 return;
1606 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1607 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1608 ND_PRINTK(2, warn,
1609 "Redirect: target address is not link-local unicast\n");
1610 return;
1613 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1614 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1616 dst = ip6_route_output(net, NULL, &fl6);
1617 if (dst->error) {
1618 dst_release(dst);
1619 return;
1621 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1622 if (IS_ERR(dst))
1623 return;
1625 rt = (struct rt6_info *) dst;
1627 if (rt->rt6i_flags & RTF_GATEWAY) {
1628 ND_PRINTK(2, warn,
1629 "Redirect: destination is not a neighbour\n");
1630 goto release;
1632 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1633 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1634 if (peer)
1635 inet_putpeer(peer);
1636 if (!ret)
1637 goto release;
1639 if (dev->addr_len) {
1640 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1641 if (!neigh) {
1642 ND_PRINTK(2, warn,
1643 "Redirect: no neigh for target address\n");
1644 goto release;
1647 read_lock_bh(&neigh->lock);
1648 if (neigh->nud_state & NUD_VALID) {
1649 memcpy(ha_buf, neigh->ha, dev->addr_len);
1650 read_unlock_bh(&neigh->lock);
1651 ha = ha_buf;
1652 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1653 ops_data_buf,
1654 &ops_data);
1655 } else
1656 read_unlock_bh(&neigh->lock);
1658 neigh_release(neigh);
1661 rd_len = min_t(unsigned int,
1662 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1663 skb->len + 8);
1664 rd_len &= ~0x7;
1665 optlen += rd_len;
1667 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1668 if (!buff)
1669 goto release;
1671 msg = skb_put(buff, sizeof(*msg));
1672 *msg = (struct rd_msg) {
1673 .icmph = {
1674 .icmp6_type = NDISC_REDIRECT,
1676 .target = *target,
1677 .dest = ipv6_hdr(skb)->daddr,
1681 * include target_address option
1684 if (ha)
1685 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1688 * build redirect option and copy skb over to the new packet.
1691 if (rd_len)
1692 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1694 skb_dst_set(buff, dst);
1695 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1696 return;
1698 release:
1699 dst_release(dst);
1702 static void pndisc_redo(struct sk_buff *skb)
1704 ndisc_recv_ns(skb);
1705 kfree_skb(skb);
1708 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1710 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1712 if (!idev)
1713 return true;
1714 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1715 idev->cnf.suppress_frag_ndisc) {
1716 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1717 return true;
1719 return false;
1722 int ndisc_rcv(struct sk_buff *skb)
1724 struct nd_msg *msg;
1726 if (ndisc_suppress_frag_ndisc(skb))
1727 return 0;
1729 if (skb_linearize(skb))
1730 return 0;
1732 msg = (struct nd_msg *)skb_transport_header(skb);
1734 __skb_push(skb, skb->data - skb_transport_header(skb));
1736 if (ipv6_hdr(skb)->hop_limit != 255) {
1737 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1738 ipv6_hdr(skb)->hop_limit);
1739 return 0;
1742 if (msg->icmph.icmp6_code != 0) {
1743 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1744 msg->icmph.icmp6_code);
1745 return 0;
1748 switch (msg->icmph.icmp6_type) {
1749 case NDISC_NEIGHBOUR_SOLICITATION:
1750 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1751 ndisc_recv_ns(skb);
1752 break;
1754 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1755 ndisc_recv_na(skb);
1756 break;
1758 case NDISC_ROUTER_SOLICITATION:
1759 ndisc_recv_rs(skb);
1760 break;
1762 case NDISC_ROUTER_ADVERTISEMENT:
1763 ndisc_router_discovery(skb);
1764 break;
1766 case NDISC_REDIRECT:
1767 ndisc_redirect_rcv(skb);
1768 break;
1771 return 0;
1774 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1776 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1777 struct netdev_notifier_change_info *change_info;
1778 struct net *net = dev_net(dev);
1779 struct inet6_dev *idev;
1781 switch (event) {
1782 case NETDEV_CHANGEADDR:
1783 neigh_changeaddr(&nd_tbl, dev);
1784 fib6_run_gc(0, net, false);
1785 /* fallthrough */
1786 case NETDEV_UP:
1787 idev = in6_dev_get(dev);
1788 if (!idev)
1789 break;
1790 if (idev->cnf.ndisc_notify ||
1791 net->ipv6.devconf_all->ndisc_notify)
1792 ndisc_send_unsol_na(dev);
1793 in6_dev_put(idev);
1794 break;
1795 case NETDEV_CHANGE:
1796 change_info = ptr;
1797 if (change_info->flags_changed & IFF_NOARP)
1798 neigh_changeaddr(&nd_tbl, dev);
1799 if (!netif_carrier_ok(dev))
1800 neigh_carrier_down(&nd_tbl, dev);
1801 break;
1802 case NETDEV_DOWN:
1803 neigh_ifdown(&nd_tbl, dev);
1804 fib6_run_gc(0, net, false);
1805 break;
1806 case NETDEV_NOTIFY_PEERS:
1807 ndisc_send_unsol_na(dev);
1808 break;
1809 default:
1810 break;
1813 return NOTIFY_DONE;
1816 static struct notifier_block ndisc_netdev_notifier = {
1817 .notifier_call = ndisc_netdev_event,
1818 .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1821 #ifdef CONFIG_SYSCTL
1822 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1823 const char *func, const char *dev_name)
1825 static char warncomm[TASK_COMM_LEN];
1826 static int warned;
1827 if (strcmp(warncomm, current->comm) && warned < 5) {
1828 strcpy(warncomm, current->comm);
1829 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1830 warncomm, func,
1831 dev_name, ctl->procname,
1832 dev_name, ctl->procname);
1833 warned++;
1837 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1839 struct net_device *dev = ctl->extra1;
1840 struct inet6_dev *idev;
1841 int ret;
1843 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1844 (strcmp(ctl->procname, "base_reachable_time") == 0))
1845 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1847 if (strcmp(ctl->procname, "retrans_time") == 0)
1848 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1850 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1851 ret = neigh_proc_dointvec_jiffies(ctl, write,
1852 buffer, lenp, ppos);
1854 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1855 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1856 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1857 buffer, lenp, ppos);
1858 else
1859 ret = -1;
1861 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1862 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1863 idev->nd_parms->reachable_time =
1864 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1865 idev->tstamp = jiffies;
1866 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1867 in6_dev_put(idev);
1869 return ret;
1873 #endif
1875 static int __net_init ndisc_net_init(struct net *net)
1877 struct ipv6_pinfo *np;
1878 struct sock *sk;
1879 int err;
1881 err = inet_ctl_sock_create(&sk, PF_INET6,
1882 SOCK_RAW, IPPROTO_ICMPV6, net);
1883 if (err < 0) {
1884 ND_PRINTK(0, err,
1885 "NDISC: Failed to initialize the control socket (err %d)\n",
1886 err);
1887 return err;
1890 net->ipv6.ndisc_sk = sk;
1892 np = inet6_sk(sk);
1893 np->hop_limit = 255;
1894 /* Do not loopback ndisc messages */
1895 np->mc_loop = 0;
1897 return 0;
1900 static void __net_exit ndisc_net_exit(struct net *net)
1902 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1905 static struct pernet_operations ndisc_net_ops = {
1906 .init = ndisc_net_init,
1907 .exit = ndisc_net_exit,
1910 int __init ndisc_init(void)
1912 int err;
1914 err = register_pernet_subsys(&ndisc_net_ops);
1915 if (err)
1916 return err;
1918 * Initialize the neighbour table
1920 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1922 #ifdef CONFIG_SYSCTL
1923 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1924 ndisc_ifinfo_sysctl_change);
1925 if (err)
1926 goto out_unregister_pernet;
1927 out:
1928 #endif
1929 return err;
1931 #ifdef CONFIG_SYSCTL
1932 out_unregister_pernet:
1933 unregister_pernet_subsys(&ndisc_net_ops);
1934 goto out;
1935 #endif
1938 int __init ndisc_late_init(void)
1940 return register_netdevice_notifier(&ndisc_netdev_notifier);
1943 void ndisc_late_cleanup(void)
1945 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1948 void ndisc_cleanup(void)
1950 #ifdef CONFIG_SYSCTL
1951 neigh_sysctl_unregister(&nd_tbl.parms);
1952 #endif
1953 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1954 unregister_pernet_subsys(&ndisc_net_ops);