2 * Neighbour Discovery for IPv6
3 * Linux INET6 implementation
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.
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
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>
45 #include <linux/sysctl.h>
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>
58 #include <net/protocol.h>
59 #include <net/ndisc.h>
60 #include <net/ip6_route.h>
61 #include <net/addrconf.h>
64 #include <net/netlink.h>
65 #include <linux/rtnetlink.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,
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 = {
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 = {
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 = {
105 .output = neigh_direct_output,
106 .connected_output = neigh_direct_output,
109 struct neigh_table nd_tbl = {
111 .key_len = sizeof(struct in6_addr),
112 .protocol = cpu_to_be16(ETH_P_IPV6),
114 .key_eq = ndisc_key_eq,
115 .constructor = ndisc_constructor,
116 .pconstructor = pndisc_constructor,
117 .pdestructor = pndisc_destructor,
118 .proxy_redo = pndisc_redo,
122 .reachable_time = ND_REACHABLE_TIME,
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,
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);
152 memset(opt + 2, 0, pad);
156 memcpy(opt+2, data, data_len);
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,
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)
185 if (!cur || !end || cur >= end)
187 type = cur->nd_opt_type;
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)
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)
222 memset(ndopts, 0, sizeof(*ndopts));
225 if (opt_len < sizeof(struct nd_opt_hdr))
227 l = nd_opt->nd_opt_len << 3;
228 if (opt_len < l || l == 0)
230 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
232 switch (nd_opt->nd_opt_type) {
233 case ND_OPT_SOURCE_LL_ADDR:
234 case ND_OPT_TARGET_LL_ADDR:
236 case ND_OPT_REDIRECT_HDR:
237 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
239 "%s: duplicated ND6 option found: type=%d\n",
240 __func__, nd_opt->nd_opt_type);
242 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
245 case ND_OPT_PREFIX_INFO:
246 ndopts->nd_opts_pi_end = nd_opt;
247 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
248 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
250 #ifdef CONFIG_IPV6_ROUTE_INFO
251 case ND_OPT_ROUTE_INFO:
252 ndopts->nd_opts_ri_end = nd_opt;
253 if (!ndopts->nd_opts_ri)
254 ndopts->nd_opts_ri = nd_opt;
258 if (ndisc_is_useropt(dev, nd_opt)) {
259 ndopts->nd_useropts_end = nd_opt;
260 if (!ndopts->nd_useropts)
261 ndopts->nd_useropts = nd_opt;
264 * Unknown options must be silently ignored,
265 * to accommodate future extension to the
269 "%s: ignored unsupported option; type=%d, len=%d\n",
277 nd_opt = ((void *)nd_opt) + l;
282 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
286 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
288 ipv6_eth_mc_map(addr, buf);
291 ipv6_arcnet_mc_map(addr, buf);
293 case ARPHRD_INFINIBAND:
294 ipv6_ib_mc_map(addr, dev->broadcast, buf);
297 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
300 memcpy(buf, dev->broadcast, dev->addr_len);
306 EXPORT_SYMBOL(ndisc_mc_map);
308 static u32 ndisc_hash(const void *pkey,
309 const struct net_device *dev,
312 return ndisc_hashfn(pkey, dev, hash_rnd);
315 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
317 return neigh_key_eq128(n, pkey);
320 static int ndisc_constructor(struct neighbour *neigh)
322 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
323 struct net_device *dev = neigh->dev;
324 struct inet6_dev *in6_dev;
325 struct neigh_parms *parms;
326 bool is_multicast = ipv6_addr_is_multicast(addr);
328 in6_dev = in6_dev_get(dev);
333 parms = in6_dev->nd_parms;
334 __neigh_parms_put(neigh->parms);
335 neigh->parms = neigh_parms_clone(parms);
337 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
338 if (!dev->header_ops) {
339 neigh->nud_state = NUD_NOARP;
340 neigh->ops = &ndisc_direct_ops;
341 neigh->output = neigh_direct_output;
344 neigh->nud_state = NUD_NOARP;
345 ndisc_mc_map(addr, neigh->ha, dev, 1);
346 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
347 neigh->nud_state = NUD_NOARP;
348 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
349 if (dev->flags&IFF_LOOPBACK)
350 neigh->type = RTN_LOCAL;
351 } else if (dev->flags&IFF_POINTOPOINT) {
352 neigh->nud_state = NUD_NOARP;
353 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
355 if (dev->header_ops->cache)
356 neigh->ops = &ndisc_hh_ops;
358 neigh->ops = &ndisc_generic_ops;
359 if (neigh->nud_state&NUD_VALID)
360 neigh->output = neigh->ops->connected_output;
362 neigh->output = neigh->ops->output;
364 in6_dev_put(in6_dev);
368 static int pndisc_constructor(struct pneigh_entry *n)
370 struct in6_addr *addr = (struct in6_addr *)&n->key;
371 struct in6_addr maddr;
372 struct net_device *dev = n->dev;
374 if (!dev || !__in6_dev_get(dev))
376 addrconf_addr_solict_mult(addr, &maddr);
377 ipv6_dev_mc_inc(dev, &maddr);
381 static void pndisc_destructor(struct pneigh_entry *n)
383 struct in6_addr *addr = (struct in6_addr *)&n->key;
384 struct in6_addr maddr;
385 struct net_device *dev = n->dev;
387 if (!dev || !__in6_dev_get(dev))
389 addrconf_addr_solict_mult(addr, &maddr);
390 ipv6_dev_mc_dec(dev, &maddr);
393 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
396 int hlen = LL_RESERVED_SPACE(dev);
397 int tlen = dev->needed_tailroom;
398 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
401 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
403 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
408 skb->protocol = htons(ETH_P_IPV6);
411 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
412 skb_reset_transport_header(skb);
414 /* Manually assign socket ownership as we avoid calling
415 * sock_alloc_send_pskb() to bypass wmem buffer limits
417 skb_set_owner_w(skb, sk);
422 static void ip6_nd_hdr(struct sk_buff *skb,
423 const struct in6_addr *saddr,
424 const struct in6_addr *daddr,
425 int hop_limit, int len)
429 skb_push(skb, sizeof(*hdr));
430 skb_reset_network_header(skb);
433 ip6_flow_hdr(hdr, 0, 0);
435 hdr->payload_len = htons(len);
436 hdr->nexthdr = IPPROTO_ICMPV6;
437 hdr->hop_limit = hop_limit;
443 static void ndisc_send_skb(struct sk_buff *skb,
444 const struct in6_addr *daddr,
445 const struct in6_addr *saddr)
447 struct dst_entry *dst = skb_dst(skb);
448 struct net *net = dev_net(skb->dev);
449 struct sock *sk = net->ipv6.ndisc_sk;
450 struct inet6_dev *idev;
452 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
455 type = icmp6h->icmp6_type;
459 int oif = skb->dev->ifindex;
461 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
462 dst = icmp6_dst_alloc(skb->dev, &fl6);
468 skb_dst_set(skb, dst);
471 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
476 ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
479 idev = __in6_dev_get(dst->dev);
480 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
482 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
483 net, sk, skb, NULL, dst->dev,
486 ICMP6MSGOUT_INC_STATS(net, idev, type);
487 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
493 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
494 const struct in6_addr *solicited_addr,
495 bool router, bool solicited, bool override, bool inc_opt)
498 struct in6_addr tmpaddr;
499 struct inet6_ifaddr *ifp;
500 const struct in6_addr *src_addr;
504 /* for anycast or proxy, solicited_addr != src_addr */
505 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
507 src_addr = solicited_addr;
508 if (ifp->flags & IFA_F_OPTIMISTIC)
510 inc_opt |= ifp->idev->cnf.force_tllao;
513 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
514 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
523 optlen += ndisc_opt_addr_space(dev,
524 NDISC_NEIGHBOUR_ADVERTISEMENT);
526 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
530 msg = (struct nd_msg *)skb_put(skb, sizeof(*msg));
531 *msg = (struct nd_msg) {
533 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
534 .icmp6_router = router,
535 .icmp6_solicited = solicited,
536 .icmp6_override = override,
538 .target = *solicited_addr,
542 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
544 NDISC_NEIGHBOUR_ADVERTISEMENT);
546 ndisc_send_skb(skb, daddr, src_addr);
549 static void ndisc_send_unsol_na(struct net_device *dev)
551 struct inet6_dev *idev;
552 struct inet6_ifaddr *ifa;
554 idev = in6_dev_get(dev);
558 read_lock_bh(&idev->lock);
559 list_for_each_entry(ifa, &idev->addr_list, if_list) {
560 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
561 /*router=*/ !!idev->cnf.forwarding,
562 /*solicited=*/ false, /*override=*/ true,
565 read_unlock_bh(&idev->lock);
570 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
571 const struct in6_addr *daddr, const struct in6_addr *saddr)
574 struct in6_addr addr_buf;
575 int inc_opt = dev->addr_len;
580 if (ipv6_get_lladdr(dev, &addr_buf,
581 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
586 if (ipv6_addr_any(saddr))
589 optlen += ndisc_opt_addr_space(dev,
590 NDISC_NEIGHBOUR_SOLICITATION);
592 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
596 msg = (struct nd_msg *)skb_put(skb, sizeof(*msg));
597 *msg = (struct nd_msg) {
599 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
605 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
607 NDISC_NEIGHBOUR_SOLICITATION);
609 ndisc_send_skb(skb, daddr, saddr);
612 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
613 const struct in6_addr *daddr)
617 int send_sllao = dev->addr_len;
620 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
622 * According to section 2.2 of RFC 4429, we must not
623 * send router solicitations with a sllao from
624 * optimistic addresses, but we may send the solicitation
625 * if we don't include the sllao. So here we check
626 * if our address is optimistic, and if so, we
627 * suppress the inclusion of the sllao.
630 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
633 if (ifp->flags & IFA_F_OPTIMISTIC) {
643 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
645 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
649 msg = (struct rs_msg *)skb_put(skb, sizeof(*msg));
650 *msg = (struct rs_msg) {
652 .icmp6_type = NDISC_ROUTER_SOLICITATION,
657 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
659 NDISC_ROUTER_SOLICITATION);
661 ndisc_send_skb(skb, daddr, saddr);
665 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
668 * "The sender MUST return an ICMP
669 * destination unreachable"
671 dst_link_failure(skb);
675 /* Called with locked neigh: either read or both */
677 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
679 struct in6_addr *saddr = NULL;
680 struct in6_addr mcaddr;
681 struct net_device *dev = neigh->dev;
682 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
683 int probes = atomic_read(&neigh->probes);
685 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
687 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
688 saddr = &ipv6_hdr(skb)->saddr;
689 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
691 if (!(neigh->nud_state & NUD_VALID)) {
693 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
696 ndisc_send_ns(dev, target, target, saddr);
697 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
700 addrconf_addr_solict_mult(target, &mcaddr);
701 ndisc_send_ns(dev, target, &mcaddr, saddr);
705 static int pndisc_is_router(const void *pkey,
706 struct net_device *dev)
708 struct pneigh_entry *n;
711 read_lock_bh(&nd_tbl.lock);
712 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
714 ret = !!(n->flags & NTF_ROUTER);
715 read_unlock_bh(&nd_tbl.lock);
720 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
721 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
722 struct ndisc_options *ndopts)
724 neigh_update(neigh, lladdr, new, flags);
725 /* report ndisc ops about neighbour update */
726 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
729 static void ndisc_recv_ns(struct sk_buff *skb)
731 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
732 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
733 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
735 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
736 offsetof(struct nd_msg, opt));
737 struct ndisc_options ndopts;
738 struct net_device *dev = skb->dev;
739 struct inet6_ifaddr *ifp;
740 struct inet6_dev *idev = NULL;
741 struct neighbour *neigh;
742 int dad = ipv6_addr_any(saddr);
746 if (skb->len < sizeof(struct nd_msg)) {
747 ND_PRINTK(2, warn, "NS: packet too short\n");
751 if (ipv6_addr_is_multicast(&msg->target)) {
752 ND_PRINTK(2, warn, "NS: multicast target address\n");
758 * DAD has to be destined for solicited node multicast address.
760 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
761 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
765 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
766 ND_PRINTK(2, warn, "NS: invalid ND options\n");
770 if (ndopts.nd_opts_src_lladdr) {
771 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
774 "NS: invalid link-layer address length\n");
779 * If the IP source address is the unspecified address,
780 * there MUST NOT be source link-layer address option
785 "NS: bad DAD packet (link-layer address option)\n");
790 inc = ipv6_addr_is_multicast(daddr);
792 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
795 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
798 * We are colliding with another node
800 * so fail our DAD process
802 addrconf_dad_failure(ifp);
806 * This is not a dad solicitation.
807 * If we are an optimistic node,
809 * Otherwise, we should ignore it.
811 if (!(ifp->flags & IFA_F_OPTIMISTIC))
818 struct net *net = dev_net(dev);
820 /* perhaps an address on the master device */
821 if (netif_is_l3_slave(dev)) {
822 struct net_device *mdev;
824 mdev = netdev_master_upper_dev_get_rcu(dev);
826 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
832 idev = in6_dev_get(dev);
834 /* XXX: count this drop? */
838 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
839 (idev->cnf.forwarding &&
840 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
841 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
842 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
843 skb->pkt_type != PACKET_HOST &&
845 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
847 * for anycast or proxy,
848 * sender should delay its response
849 * by a random time between 0 and
850 * MAX_ANYCAST_DELAY_TIME seconds.
851 * (RFC2461) -- yoshfuji
853 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
855 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
863 is_router = idev->cnf.forwarding;
866 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
867 !!is_router, false, (ifp != NULL), true);
872 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
874 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
877 * update / create cache entry
878 * for the source address
880 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
881 !inc || lladdr || !dev->addr_len);
883 ndisc_update(dev, neigh, lladdr, NUD_STALE,
884 NEIGH_UPDATE_F_WEAK_OVERRIDE|
885 NEIGH_UPDATE_F_OVERRIDE,
886 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
887 if (neigh || !dev->header_ops) {
888 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
889 true, (ifp != NULL && inc), inc);
891 neigh_release(neigh);
901 static void ndisc_recv_na(struct sk_buff *skb)
903 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
904 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
905 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
907 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
908 offsetof(struct nd_msg, opt));
909 struct ndisc_options ndopts;
910 struct net_device *dev = skb->dev;
911 struct inet6_dev *idev = __in6_dev_get(dev);
912 struct inet6_ifaddr *ifp;
913 struct neighbour *neigh;
915 if (skb->len < sizeof(struct nd_msg)) {
916 ND_PRINTK(2, warn, "NA: packet too short\n");
920 if (ipv6_addr_is_multicast(&msg->target)) {
921 ND_PRINTK(2, warn, "NA: target address is multicast\n");
925 if (ipv6_addr_is_multicast(daddr) &&
926 msg->icmph.icmp6_solicited) {
927 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
931 /* For some 802.11 wireless deployments (and possibly other networks),
932 * there will be a NA proxy and unsolicitd packets are attacks
933 * and thus should not be accepted.
935 if (!msg->icmph.icmp6_solicited && idev &&
936 idev->cnf.drop_unsolicited_na)
939 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
940 ND_PRINTK(2, warn, "NS: invalid ND option\n");
943 if (ndopts.nd_opts_tgt_lladdr) {
944 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
947 "NA: invalid link-layer address length\n");
951 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
953 if (skb->pkt_type != PACKET_LOOPBACK
954 && (ifp->flags & IFA_F_TENTATIVE)) {
955 addrconf_dad_failure(ifp);
958 /* What should we make now? The advertisement
959 is invalid, but ndisc specs say nothing
960 about it. It could be misconfiguration, or
961 an smart proxy agent tries to help us :-)
963 We should not print the error if NA has been
964 received from loopback - it is just our own
965 unsolicited advertisement.
967 if (skb->pkt_type != PACKET_LOOPBACK)
969 "NA: someone advertises our address %pI6 on %s!\n",
970 &ifp->addr, ifp->idev->dev->name);
974 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
977 u8 old_flags = neigh->flags;
978 struct net *net = dev_net(dev);
980 if (neigh->nud_state & NUD_FAILED)
984 * Don't update the neighbor cache entry on a proxy NA from
985 * ourselves because either the proxied node is off link or it
986 * has already sent a NA to us.
988 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
989 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
990 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
991 /* XXX: idev->cnf.proxy_ndp */
995 ndisc_update(dev, neigh, lladdr,
996 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
997 NEIGH_UPDATE_F_WEAK_OVERRIDE|
998 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
999 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1000 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1001 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1003 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1005 * Change: router to host
1007 rt6_clean_tohost(dev_net(dev), saddr);
1011 neigh_release(neigh);
1015 static void ndisc_recv_rs(struct sk_buff *skb)
1017 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1018 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1019 struct neighbour *neigh;
1020 struct inet6_dev *idev;
1021 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1022 struct ndisc_options ndopts;
1025 if (skb->len < sizeof(*rs_msg))
1028 idev = __in6_dev_get(skb->dev);
1030 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1034 /* Don't accept RS if we're not in router mode */
1035 if (!idev->cnf.forwarding)
1039 * Don't update NCE if src = ::;
1040 * this implies that the source node has no ip address assigned yet.
1042 if (ipv6_addr_any(saddr))
1045 /* Parse ND options */
1046 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1047 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1051 if (ndopts.nd_opts_src_lladdr) {
1052 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1058 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1060 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1061 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1062 NEIGH_UPDATE_F_OVERRIDE|
1063 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1064 NDISC_ROUTER_SOLICITATION, &ndopts);
1065 neigh_release(neigh);
1071 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1073 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1074 struct sk_buff *skb;
1075 struct nlmsghdr *nlh;
1076 struct nduseroptmsg *ndmsg;
1077 struct net *net = dev_net(ra->dev);
1079 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1080 + (opt->nd_opt_len << 3));
1081 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1083 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1089 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1091 goto nla_put_failure;
1094 ndmsg = nlmsg_data(nlh);
1095 ndmsg->nduseropt_family = AF_INET6;
1096 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1097 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1098 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1099 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1101 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1103 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1104 goto nla_put_failure;
1105 nlmsg_end(skb, nlh);
1107 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1114 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1117 static void ndisc_router_discovery(struct sk_buff *skb)
1119 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1120 struct neighbour *neigh = NULL;
1121 struct inet6_dev *in6_dev;
1122 struct rt6_info *rt = NULL;
1124 struct ndisc_options ndopts;
1126 unsigned int pref = 0;
1128 bool send_ifinfo_notify = false;
1130 __u8 *opt = (__u8 *)(ra_msg + 1);
1132 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1133 sizeof(struct ra_msg);
1136 "RA: %s, dev: %s\n",
1137 __func__, skb->dev->name);
1138 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1139 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1143 ND_PRINTK(2, warn, "RA: packet too short\n");
1147 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1148 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1149 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1155 * set the RA_RECV flag in the interface
1158 in6_dev = __in6_dev_get(skb->dev);
1160 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1165 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1166 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1170 if (!ipv6_accept_ra(in6_dev)) {
1172 "RA: %s, did not accept ra for dev: %s\n",
1173 __func__, skb->dev->name);
1174 goto skip_linkparms;
1177 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1178 /* skip link-specific parameters from interior routers */
1179 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1181 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1182 __func__, skb->dev->name);
1183 goto skip_linkparms;
1187 if (in6_dev->if_flags & IF_RS_SENT) {
1189 * flag that an RA was received after an RS was sent
1190 * out on this interface.
1192 in6_dev->if_flags |= IF_RA_RCVD;
1196 * Remember the managed/otherconf flags from most recently
1197 * received RA message (RFC 2462) -- yoshfuji
1199 old_if_flags = in6_dev->if_flags;
1200 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1202 (ra_msg->icmph.icmp6_addrconf_managed ?
1203 IF_RA_MANAGED : 0) |
1204 (ra_msg->icmph.icmp6_addrconf_other ?
1205 IF_RA_OTHERCONF : 0);
1207 if (old_if_flags != in6_dev->if_flags)
1208 send_ifinfo_notify = true;
1210 if (!in6_dev->cnf.accept_ra_defrtr) {
1212 "RA: %s, defrtr is false for dev: %s\n",
1213 __func__, skb->dev->name);
1217 /* Do not accept RA with source-addr found on local machine unless
1218 * accept_ra_from_local is set to true.
1220 if (!in6_dev->cnf.accept_ra_from_local &&
1221 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1224 "RA from local address detected on dev: %s: default router ignored\n",
1229 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1231 #ifdef CONFIG_IPV6_ROUTER_PREF
1232 pref = ra_msg->icmph.icmp6_router_pref;
1233 /* 10b is handled as if it were 00b (medium) */
1234 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1235 !in6_dev->cnf.accept_ra_rtr_pref)
1236 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1239 rt = rt6_get_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev);
1242 neigh = dst_neigh_lookup(&rt->dst, &ipv6_hdr(skb)->saddr);
1245 "RA: %s got default router without neighbour\n",
1251 if (rt && lifetime == 0) {
1256 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, for dev: %s\n",
1257 rt, lifetime, skb->dev->name);
1258 if (!rt && lifetime) {
1259 ND_PRINTK(3, info, "RA: adding default router\n");
1261 rt = rt6_add_dflt_router(&ipv6_hdr(skb)->saddr, skb->dev, pref);
1264 "RA: %s failed to add default route\n",
1269 neigh = dst_neigh_lookup(&rt->dst, &ipv6_hdr(skb)->saddr);
1272 "RA: %s got default router without neighbour\n",
1277 neigh->flags |= NTF_ROUTER;
1279 rt->rt6i_flags = (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1283 rt6_set_expires(rt, jiffies + (HZ * lifetime));
1284 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1285 ra_msg->icmph.icmp6_hop_limit) {
1286 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1287 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1289 dst_metric_set(&rt->dst, RTAX_HOPLIMIT,
1290 ra_msg->icmph.icmp6_hop_limit);
1292 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1299 * Update Reachable Time and Retrans Timer
1302 if (in6_dev->nd_parms) {
1303 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1305 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1306 rtime = (rtime*HZ)/1000;
1309 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1310 in6_dev->tstamp = jiffies;
1311 send_ifinfo_notify = true;
1314 rtime = ntohl(ra_msg->reachable_time);
1315 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1316 rtime = (rtime*HZ)/1000;
1321 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1322 NEIGH_VAR_SET(in6_dev->nd_parms,
1323 BASE_REACHABLE_TIME, rtime);
1324 NEIGH_VAR_SET(in6_dev->nd_parms,
1325 GC_STALETIME, 3 * rtime);
1326 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1327 in6_dev->tstamp = jiffies;
1328 send_ifinfo_notify = true;
1334 * Send a notify if RA changed managed/otherconf flags or timer settings
1336 if (send_ifinfo_notify)
1337 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1346 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1350 if (ndopts.nd_opts_src_lladdr) {
1351 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1355 "RA: invalid link-layer address length\n");
1359 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1360 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1361 NEIGH_UPDATE_F_OVERRIDE|
1362 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1363 NEIGH_UPDATE_F_ISROUTER,
1364 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1367 if (!ipv6_accept_ra(in6_dev)) {
1369 "RA: %s, accept_ra is false for dev: %s\n",
1370 __func__, skb->dev->name);
1374 #ifdef CONFIG_IPV6_ROUTE_INFO
1375 if (!in6_dev->cnf.accept_ra_from_local &&
1376 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1379 "RA from local address detected on dev: %s: router info ignored.\n",
1381 goto skip_routeinfo;
1384 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1385 struct nd_opt_hdr *p;
1386 for (p = ndopts.nd_opts_ri;
1388 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1389 struct route_info *ri = (struct route_info *)p;
1390 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1391 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1392 ri->prefix_len == 0)
1395 if (ri->prefix_len == 0 &&
1396 !in6_dev->cnf.accept_ra_defrtr)
1398 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1400 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1401 &ipv6_hdr(skb)->saddr);
1408 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1409 /* skip link-specific ndopts from interior routers */
1410 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1412 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1413 __func__, skb->dev->name);
1418 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1419 struct nd_opt_hdr *p;
1420 for (p = ndopts.nd_opts_pi;
1422 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1423 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1424 (p->nd_opt_len) << 3,
1425 ndopts.nd_opts_src_lladdr != NULL);
1429 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1433 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1436 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1437 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1438 } else if (in6_dev->cnf.mtu6 != mtu) {
1439 in6_dev->cnf.mtu6 = mtu;
1442 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
1444 rt6_mtu_change(skb->dev, mtu);
1448 if (ndopts.nd_useropts) {
1449 struct nd_opt_hdr *p;
1450 for (p = ndopts.nd_useropts;
1452 p = ndisc_next_useropt(skb->dev, p,
1453 ndopts.nd_useropts_end)) {
1454 ndisc_ra_useropt(skb, p);
1458 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1459 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1464 neigh_release(neigh);
1467 static void ndisc_redirect_rcv(struct sk_buff *skb)
1470 struct ndisc_options ndopts;
1471 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1472 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1473 offsetof(struct rd_msg, opt));
1475 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1476 switch (skb->ndisc_nodetype) {
1477 case NDISC_NODETYPE_HOST:
1478 case NDISC_NODETYPE_NODEFAULT:
1480 "Redirect: from host or unauthorized router\n");
1485 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1487 "Redirect: source address is not link-local\n");
1491 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1494 if (!ndopts.nd_opts_rh) {
1495 ip6_redirect_no_header(skb, dev_net(skb->dev),
1496 skb->dev->ifindex, 0);
1500 hdr = (u8 *)ndopts.nd_opts_rh;
1502 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1505 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1508 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1509 struct sk_buff *orig_skb,
1512 u8 *opt = skb_put(skb, rd_len);
1515 *(opt++) = ND_OPT_REDIRECT_HDR;
1516 *(opt++) = (rd_len >> 3);
1519 memcpy(opt, ipv6_hdr(orig_skb), rd_len - 8);
1522 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1524 struct net_device *dev = skb->dev;
1525 struct net *net = dev_net(dev);
1526 struct sock *sk = net->ipv6.ndisc_sk;
1528 struct inet_peer *peer;
1529 struct sk_buff *buff;
1531 struct in6_addr saddr_buf;
1532 struct rt6_info *rt;
1533 struct dst_entry *dst;
1536 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1537 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1540 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1541 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1546 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1547 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1549 "Redirect: target address is not link-local unicast\n");
1553 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1554 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1556 dst = ip6_route_output(net, NULL, &fl6);
1561 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1565 rt = (struct rt6_info *) dst;
1567 if (rt->rt6i_flags & RTF_GATEWAY) {
1569 "Redirect: destination is not a neighbour\n");
1572 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1573 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1579 if (dev->addr_len) {
1580 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1583 "Redirect: no neigh for target address\n");
1587 read_lock_bh(&neigh->lock);
1588 if (neigh->nud_state & NUD_VALID) {
1589 memcpy(ha_buf, neigh->ha, dev->addr_len);
1590 read_unlock_bh(&neigh->lock);
1592 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1596 read_unlock_bh(&neigh->lock);
1598 neigh_release(neigh);
1601 rd_len = min_t(unsigned int,
1602 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1607 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1611 msg = (struct rd_msg *)skb_put(buff, sizeof(*msg));
1612 *msg = (struct rd_msg) {
1614 .icmp6_type = NDISC_REDIRECT,
1617 .dest = ipv6_hdr(skb)->daddr,
1621 * include target_address option
1625 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1628 * build redirect option and copy skb over to the new packet.
1632 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1634 skb_dst_set(buff, dst);
1635 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1642 static void pndisc_redo(struct sk_buff *skb)
1648 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1650 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1654 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1655 idev->cnf.suppress_frag_ndisc) {
1656 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1662 int ndisc_rcv(struct sk_buff *skb)
1666 if (ndisc_suppress_frag_ndisc(skb))
1669 if (skb_linearize(skb))
1672 msg = (struct nd_msg *)skb_transport_header(skb);
1674 __skb_push(skb, skb->data - skb_transport_header(skb));
1676 if (ipv6_hdr(skb)->hop_limit != 255) {
1677 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1678 ipv6_hdr(skb)->hop_limit);
1682 if (msg->icmph.icmp6_code != 0) {
1683 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1684 msg->icmph.icmp6_code);
1688 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1690 switch (msg->icmph.icmp6_type) {
1691 case NDISC_NEIGHBOUR_SOLICITATION:
1695 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1699 case NDISC_ROUTER_SOLICITATION:
1703 case NDISC_ROUTER_ADVERTISEMENT:
1704 ndisc_router_discovery(skb);
1707 case NDISC_REDIRECT:
1708 ndisc_redirect_rcv(skb);
1715 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1717 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1718 struct netdev_notifier_change_info *change_info;
1719 struct net *net = dev_net(dev);
1720 struct inet6_dev *idev;
1723 case NETDEV_CHANGEADDR:
1724 neigh_changeaddr(&nd_tbl, dev);
1725 fib6_run_gc(0, net, false);
1726 idev = in6_dev_get(dev);
1729 if (idev->cnf.ndisc_notify)
1730 ndisc_send_unsol_na(dev);
1735 if (change_info->flags_changed & IFF_NOARP)
1736 neigh_changeaddr(&nd_tbl, dev);
1739 neigh_ifdown(&nd_tbl, dev);
1740 fib6_run_gc(0, net, false);
1742 case NETDEV_NOTIFY_PEERS:
1743 ndisc_send_unsol_na(dev);
1752 static struct notifier_block ndisc_netdev_notifier = {
1753 .notifier_call = ndisc_netdev_event,
1756 #ifdef CONFIG_SYSCTL
1757 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1758 const char *func, const char *dev_name)
1760 static char warncomm[TASK_COMM_LEN];
1762 if (strcmp(warncomm, current->comm) && warned < 5) {
1763 strcpy(warncomm, current->comm);
1764 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1766 dev_name, ctl->procname,
1767 dev_name, ctl->procname);
1772 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1774 struct net_device *dev = ctl->extra1;
1775 struct inet6_dev *idev;
1778 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1779 (strcmp(ctl->procname, "base_reachable_time") == 0))
1780 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1782 if (strcmp(ctl->procname, "retrans_time") == 0)
1783 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1785 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1786 ret = neigh_proc_dointvec_jiffies(ctl, write,
1787 buffer, lenp, ppos);
1789 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1790 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1791 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1792 buffer, lenp, ppos);
1796 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1797 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1798 idev->nd_parms->reachable_time =
1799 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1800 idev->tstamp = jiffies;
1801 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1810 static int __net_init ndisc_net_init(struct net *net)
1812 struct ipv6_pinfo *np;
1816 err = inet_ctl_sock_create(&sk, PF_INET6,
1817 SOCK_RAW, IPPROTO_ICMPV6, net);
1820 "NDISC: Failed to initialize the control socket (err %d)\n",
1825 net->ipv6.ndisc_sk = sk;
1828 np->hop_limit = 255;
1829 /* Do not loopback ndisc messages */
1835 static void __net_exit ndisc_net_exit(struct net *net)
1837 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1840 static struct pernet_operations ndisc_net_ops = {
1841 .init = ndisc_net_init,
1842 .exit = ndisc_net_exit,
1845 int __init ndisc_init(void)
1849 err = register_pernet_subsys(&ndisc_net_ops);
1853 * Initialize the neighbour table
1855 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1857 #ifdef CONFIG_SYSCTL
1858 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1859 ndisc_ifinfo_sysctl_change);
1861 goto out_unregister_pernet;
1866 #ifdef CONFIG_SYSCTL
1867 out_unregister_pernet:
1868 unregister_pernet_subsys(&ndisc_net_ops);
1873 int __init ndisc_late_init(void)
1875 return register_netdevice_notifier(&ndisc_netdev_notifier);
1878 void ndisc_late_cleanup(void)
1880 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1883 void ndisc_cleanup(void)
1885 #ifdef CONFIG_SYSCTL
1886 neigh_sysctl_unregister(&nd_tbl.parms);
1888 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1889 unregister_pernet_subsys(&ndisc_net_ops);