1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Neighbour Discovery for IPv6
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Mike Shaver <shaver@ingenia.com>
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
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>
41 #include <linux/sysctl.h>
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>
55 #include <net/protocol.h>
56 #include <net/ndisc.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
61 #include <net/netlink.h>
62 #include <linux/rtnetlink.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,
75 static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
76 static bool ndisc_allow_add(const struct net_device *dev,
77 struct netlink_ext_ack *extack);
78 static int ndisc_constructor(struct neighbour *neigh);
79 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
80 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
81 static int pndisc_constructor(struct pneigh_entry *n);
82 static void pndisc_destructor(struct pneigh_entry *n);
83 static void pndisc_redo(struct sk_buff *skb);
84 static int ndisc_is_multicast(const void *pkey);
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,
119 .is_multicast = ndisc_is_multicast,
120 .allow_add = ndisc_allow_add,
124 .reachable_time = ND_REACHABLE_TIME,
126 [NEIGH_VAR_MCAST_PROBES] = 3,
127 [NEIGH_VAR_UCAST_PROBES] = 3,
128 [NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER,
129 [NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME,
130 [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
131 [NEIGH_VAR_INTERVAL_PROBE_TIME_MS] = 5 * HZ,
132 [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
133 [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
134 [NEIGH_VAR_PROXY_QLEN] = 64,
135 [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
136 [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
139 .gc_interval = 30 * HZ,
144 EXPORT_SYMBOL_GPL(nd_tbl);
146 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, const void *data,
147 int data_len, int pad)
149 int space = __ndisc_opt_addr_space(data_len, pad);
150 u8 *opt = skb_put(skb, space);
155 memset(opt + 2, 0, pad);
159 memcpy(opt+2, data, data_len);
164 memset(opt, 0, space);
166 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
168 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
169 const void *data, u8 icmp6_type)
171 __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
172 ndisc_addr_option_pad(skb->dev->type));
173 ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
176 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
180 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
181 ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
184 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
185 struct nd_opt_hdr *end)
188 if (!cur || !end || cur >= end)
190 type = cur->nd_opt_type;
192 cur = ((void *)cur) + (cur->nd_opt_len << 3);
193 } while (cur < end && cur->nd_opt_type != type);
194 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
197 static inline int ndisc_is_useropt(const struct net_device *dev,
198 struct nd_opt_hdr *opt)
200 return opt->nd_opt_type == ND_OPT_PREFIX_INFO ||
201 opt->nd_opt_type == ND_OPT_RDNSS ||
202 opt->nd_opt_type == ND_OPT_DNSSL ||
203 opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
204 opt->nd_opt_type == ND_OPT_PREF64 ||
205 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
208 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
209 struct nd_opt_hdr *cur,
210 struct nd_opt_hdr *end)
212 if (!cur || !end || cur >= end)
215 cur = ((void *)cur) + (cur->nd_opt_len << 3);
216 } while (cur < end && !ndisc_is_useropt(dev, cur));
217 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
220 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
221 u8 *opt, int opt_len,
222 struct ndisc_options *ndopts)
224 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
226 if (!nd_opt || opt_len < 0 || !ndopts)
228 memset(ndopts, 0, sizeof(*ndopts));
231 if (opt_len < sizeof(struct nd_opt_hdr))
233 l = nd_opt->nd_opt_len << 3;
234 if (opt_len < l || l == 0)
236 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
238 switch (nd_opt->nd_opt_type) {
239 case ND_OPT_SOURCE_LL_ADDR:
240 case ND_OPT_TARGET_LL_ADDR:
243 case ND_OPT_REDIRECT_HDR:
244 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
246 "%s: duplicated ND6 option found: type=%d\n",
247 __func__, nd_opt->nd_opt_type);
249 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
252 case ND_OPT_PREFIX_INFO:
253 ndopts->nd_opts_pi_end = nd_opt;
254 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
255 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
257 #ifdef CONFIG_IPV6_ROUTE_INFO
258 case ND_OPT_ROUTE_INFO:
259 ndopts->nd_opts_ri_end = nd_opt;
260 if (!ndopts->nd_opts_ri)
261 ndopts->nd_opts_ri = nd_opt;
265 if (ndisc_is_useropt(dev, nd_opt)) {
266 ndopts->nd_useropts_end = nd_opt;
267 if (!ndopts->nd_useropts)
268 ndopts->nd_useropts = nd_opt;
271 * Unknown options must be silently ignored,
272 * to accommodate future extension to the
276 "%s: ignored unsupported option; type=%d, len=%d\n",
284 nd_opt = ((void *)nd_opt) + l;
289 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
293 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
295 ipv6_eth_mc_map(addr, buf);
298 ipv6_arcnet_mc_map(addr, buf);
300 case ARPHRD_INFINIBAND:
301 ipv6_ib_mc_map(addr, dev->broadcast, buf);
304 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
307 memcpy(buf, dev->broadcast, dev->addr_len);
313 EXPORT_SYMBOL(ndisc_mc_map);
315 static u32 ndisc_hash(const void *pkey,
316 const struct net_device *dev,
319 return ndisc_hashfn(pkey, dev, hash_rnd);
322 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
324 return neigh_key_eq128(n, pkey);
327 static int ndisc_constructor(struct neighbour *neigh)
329 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
330 struct net_device *dev = neigh->dev;
331 struct inet6_dev *in6_dev;
332 struct neigh_parms *parms;
333 bool is_multicast = ipv6_addr_is_multicast(addr);
335 in6_dev = in6_dev_get(dev);
340 parms = in6_dev->nd_parms;
341 __neigh_parms_put(neigh->parms);
342 neigh->parms = neigh_parms_clone(parms);
344 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
345 if (!dev->header_ops) {
346 neigh->nud_state = NUD_NOARP;
347 neigh->ops = &ndisc_direct_ops;
348 neigh->output = neigh_direct_output;
351 neigh->nud_state = NUD_NOARP;
352 ndisc_mc_map(addr, neigh->ha, dev, 1);
353 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
354 neigh->nud_state = NUD_NOARP;
355 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
356 if (dev->flags&IFF_LOOPBACK)
357 neigh->type = RTN_LOCAL;
358 } else if (dev->flags&IFF_POINTOPOINT) {
359 neigh->nud_state = NUD_NOARP;
360 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
362 if (dev->header_ops->cache)
363 neigh->ops = &ndisc_hh_ops;
365 neigh->ops = &ndisc_generic_ops;
366 if (neigh->nud_state&NUD_VALID)
367 neigh->output = neigh->ops->connected_output;
369 neigh->output = neigh->ops->output;
371 in6_dev_put(in6_dev);
375 static int pndisc_constructor(struct pneigh_entry *n)
377 struct in6_addr *addr = (struct in6_addr *)&n->key;
378 struct in6_addr maddr;
379 struct net_device *dev = n->dev;
381 if (!dev || !__in6_dev_get(dev))
383 addrconf_addr_solict_mult(addr, &maddr);
384 ipv6_dev_mc_inc(dev, &maddr);
388 static void pndisc_destructor(struct pneigh_entry *n)
390 struct in6_addr *addr = (struct in6_addr *)&n->key;
391 struct in6_addr maddr;
392 struct net_device *dev = n->dev;
394 if (!dev || !__in6_dev_get(dev))
396 addrconf_addr_solict_mult(addr, &maddr);
397 ipv6_dev_mc_dec(dev, &maddr);
400 /* called with rtnl held */
401 static bool ndisc_allow_add(const struct net_device *dev,
402 struct netlink_ext_ack *extack)
404 struct inet6_dev *idev = __in6_dev_get(dev);
406 if (!idev || idev->cnf.disable_ipv6) {
407 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
414 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
417 int hlen = LL_RESERVED_SPACE(dev);
418 int tlen = dev->needed_tailroom;
419 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
422 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
424 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
429 skb->protocol = htons(ETH_P_IPV6);
432 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
433 skb_reset_transport_header(skb);
435 /* Manually assign socket ownership as we avoid calling
436 * sock_alloc_send_pskb() to bypass wmem buffer limits
438 skb_set_owner_w(skb, sk);
443 static void ip6_nd_hdr(struct sk_buff *skb,
444 const struct in6_addr *saddr,
445 const struct in6_addr *daddr,
446 int hop_limit, int len)
449 struct inet6_dev *idev;
453 idev = __in6_dev_get(skb->dev);
454 tclass = idev ? idev->cnf.ndisc_tclass : 0;
457 skb_push(skb, sizeof(*hdr));
458 skb_reset_network_header(skb);
461 ip6_flow_hdr(hdr, tclass, 0);
463 hdr->payload_len = htons(len);
464 hdr->nexthdr = IPPROTO_ICMPV6;
465 hdr->hop_limit = hop_limit;
471 void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
472 const struct in6_addr *saddr)
474 struct dst_entry *dst = skb_dst(skb);
475 struct net *net = dev_net(skb->dev);
476 struct sock *sk = net->ipv6.ndisc_sk;
477 struct inet6_dev *idev;
479 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
482 type = icmp6h->icmp6_type;
486 int oif = skb->dev->ifindex;
488 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
489 dst = icmp6_dst_alloc(skb->dev, &fl6);
495 skb_dst_set(skb, dst);
498 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
503 ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
506 idev = __in6_dev_get(dst->dev);
507 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
509 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
510 net, sk, skb, NULL, dst->dev,
513 ICMP6MSGOUT_INC_STATS(net, idev, type);
514 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
519 EXPORT_SYMBOL(ndisc_send_skb);
521 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
522 const struct in6_addr *solicited_addr,
523 bool router, bool solicited, bool override, bool inc_opt)
526 struct in6_addr tmpaddr;
527 struct inet6_ifaddr *ifp;
528 const struct in6_addr *src_addr;
532 /* for anycast or proxy, solicited_addr != src_addr */
533 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
535 src_addr = solicited_addr;
536 if (ifp->flags & IFA_F_OPTIMISTIC)
538 inc_opt |= ifp->idev->cnf.force_tllao;
541 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
542 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
551 optlen += ndisc_opt_addr_space(dev,
552 NDISC_NEIGHBOUR_ADVERTISEMENT);
554 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
558 msg = skb_put(skb, sizeof(*msg));
559 *msg = (struct nd_msg) {
561 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
562 .icmp6_router = router,
563 .icmp6_solicited = solicited,
564 .icmp6_override = override,
566 .target = *solicited_addr,
570 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
572 NDISC_NEIGHBOUR_ADVERTISEMENT);
574 ndisc_send_skb(skb, daddr, src_addr);
577 static void ndisc_send_unsol_na(struct net_device *dev)
579 struct inet6_dev *idev;
580 struct inet6_ifaddr *ifa;
582 idev = in6_dev_get(dev);
586 read_lock_bh(&idev->lock);
587 list_for_each_entry(ifa, &idev->addr_list, if_list) {
588 /* skip tentative addresses until dad completes */
589 if (ifa->flags & IFA_F_TENTATIVE &&
590 !(ifa->flags & IFA_F_OPTIMISTIC))
593 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
594 /*router=*/ !!idev->cnf.forwarding,
595 /*solicited=*/ false, /*override=*/ true,
598 read_unlock_bh(&idev->lock);
603 struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
604 const struct in6_addr *saddr, u64 nonce)
606 int inc_opt = dev->addr_len;
614 if (ipv6_addr_any(saddr))
617 optlen += ndisc_opt_addr_space(dev,
618 NDISC_NEIGHBOUR_SOLICITATION);
622 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
626 msg = skb_put(skb, sizeof(*msg));
627 *msg = (struct nd_msg) {
629 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
635 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
637 NDISC_NEIGHBOUR_SOLICITATION);
639 u8 *opt = skb_put(skb, 8);
641 opt[0] = ND_OPT_NONCE;
643 memcpy(opt + 2, &nonce, 6);
648 EXPORT_SYMBOL(ndisc_ns_create);
650 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
651 const struct in6_addr *daddr, const struct in6_addr *saddr,
654 struct in6_addr addr_buf;
658 if (ipv6_get_lladdr(dev, &addr_buf,
659 (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
664 skb = ndisc_ns_create(dev, solicit, saddr, nonce);
667 ndisc_send_skb(skb, daddr, saddr);
670 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
671 const struct in6_addr *daddr)
675 int send_sllao = dev->addr_len;
678 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
680 * According to section 2.2 of RFC 4429, we must not
681 * send router solicitations with a sllao from
682 * optimistic addresses, but we may send the solicitation
683 * if we don't include the sllao. So here we check
684 * if our address is optimistic, and if so, we
685 * suppress the inclusion of the sllao.
688 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
691 if (ifp->flags & IFA_F_OPTIMISTIC) {
701 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
703 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
707 msg = skb_put(skb, sizeof(*msg));
708 *msg = (struct rs_msg) {
710 .icmp6_type = NDISC_ROUTER_SOLICITATION,
715 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
717 NDISC_ROUTER_SOLICITATION);
719 ndisc_send_skb(skb, daddr, saddr);
723 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
726 * "The sender MUST return an ICMP
727 * destination unreachable"
729 dst_link_failure(skb);
733 /* Called with locked neigh: either read or both */
735 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
737 struct in6_addr *saddr = NULL;
738 struct in6_addr mcaddr;
739 struct net_device *dev = neigh->dev;
740 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
741 int probes = atomic_read(&neigh->probes);
743 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
745 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
746 saddr = &ipv6_hdr(skb)->saddr;
747 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
749 if (!(READ_ONCE(neigh->nud_state) & NUD_VALID)) {
751 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
754 ndisc_send_ns(dev, target, target, saddr, 0);
755 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
758 addrconf_addr_solict_mult(target, &mcaddr);
759 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
763 static int pndisc_is_router(const void *pkey,
764 struct net_device *dev)
766 struct pneigh_entry *n;
769 read_lock_bh(&nd_tbl.lock);
770 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
772 ret = !!(n->flags & NTF_ROUTER);
773 read_unlock_bh(&nd_tbl.lock);
778 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
779 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
780 struct ndisc_options *ndopts)
782 neigh_update(neigh, lladdr, new, flags, 0);
783 /* report ndisc ops about neighbour update */
784 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
787 static void ndisc_recv_ns(struct sk_buff *skb)
789 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
790 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
791 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
793 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
794 offsetof(struct nd_msg, opt));
795 struct ndisc_options ndopts;
796 struct net_device *dev = skb->dev;
797 struct inet6_ifaddr *ifp;
798 struct inet6_dev *idev = NULL;
799 struct neighbour *neigh;
800 int dad = ipv6_addr_any(saddr);
805 if (skb->len < sizeof(struct nd_msg)) {
806 ND_PRINTK(2, warn, "NS: packet too short\n");
810 if (ipv6_addr_is_multicast(&msg->target)) {
811 ND_PRINTK(2, warn, "NS: multicast target address\n");
817 * DAD has to be destined for solicited node multicast address.
819 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
820 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
824 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
825 ND_PRINTK(2, warn, "NS: invalid ND options\n");
829 if (ndopts.nd_opts_src_lladdr) {
830 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
833 "NS: invalid link-layer address length\n");
838 * If the IP source address is the unspecified address,
839 * there MUST NOT be source link-layer address option
844 "NS: bad DAD packet (link-layer address option)\n");
848 if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
849 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
851 inc = ipv6_addr_is_multicast(daddr);
853 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
856 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
858 if (nonce != 0 && ifp->dad_nonce == nonce) {
859 u8 *np = (u8 *)&nonce;
860 /* Matching nonce if looped back */
862 "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
863 ifp->idev->dev->name,
868 * We are colliding with another node
870 * so fail our DAD process
872 addrconf_dad_failure(skb, ifp);
876 * This is not a dad solicitation.
877 * If we are an optimistic node,
879 * Otherwise, we should ignore it.
881 if (!(ifp->flags & IFA_F_OPTIMISTIC))
888 struct net *net = dev_net(dev);
890 /* perhaps an address on the master device */
891 if (netif_is_l3_slave(dev)) {
892 struct net_device *mdev;
894 mdev = netdev_master_upper_dev_get_rcu(dev);
896 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
902 idev = in6_dev_get(dev);
904 /* XXX: count this drop? */
908 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
909 (idev->cnf.forwarding &&
910 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
911 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
912 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
913 skb->pkt_type != PACKET_HOST &&
915 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
917 * for anycast or proxy,
918 * sender should delay its response
919 * by a random time between 0 and
920 * MAX_ANYCAST_DELAY_TIME seconds.
921 * (RFC2461) -- yoshfuji
923 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
925 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
933 is_router = idev->cnf.forwarding;
936 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
937 !!is_router, false, (ifp != NULL), true);
942 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
944 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
947 * update / create cache entry
948 * for the source address
950 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
951 !inc || lladdr || !dev->addr_len);
953 ndisc_update(dev, neigh, lladdr, NUD_STALE,
954 NEIGH_UPDATE_F_WEAK_OVERRIDE|
955 NEIGH_UPDATE_F_OVERRIDE,
956 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
957 if (neigh || !dev->header_ops) {
958 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
959 true, (ifp != NULL && inc), inc);
961 neigh_release(neigh);
971 static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
973 struct inet6_dev *idev = __in6_dev_get(dev);
975 switch (idev->cnf.accept_untracked_na) {
976 case 0: /* Don't accept untracked na (absent in neighbor cache) */
978 case 1: /* Create new entries from na if currently untracked */
980 case 2: /* Create new entries from untracked na only if saddr is in the
981 * same subnet as an address configured on the interface that
984 return !!ipv6_chk_prefix(saddr, dev);
990 static void ndisc_recv_na(struct sk_buff *skb)
992 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
993 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
994 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
996 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
997 offsetof(struct nd_msg, opt));
998 struct ndisc_options ndopts;
999 struct net_device *dev = skb->dev;
1000 struct inet6_dev *idev = __in6_dev_get(dev);
1001 struct inet6_ifaddr *ifp;
1002 struct neighbour *neigh;
1005 if (skb->len < sizeof(struct nd_msg)) {
1006 ND_PRINTK(2, warn, "NA: packet too short\n");
1010 if (ipv6_addr_is_multicast(&msg->target)) {
1011 ND_PRINTK(2, warn, "NA: target address is multicast\n");
1015 if (ipv6_addr_is_multicast(daddr) &&
1016 msg->icmph.icmp6_solicited) {
1017 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
1021 /* For some 802.11 wireless deployments (and possibly other networks),
1022 * there will be a NA proxy and unsolicitd packets are attacks
1023 * and thus should not be accepted.
1024 * drop_unsolicited_na takes precedence over accept_untracked_na
1026 if (!msg->icmph.icmp6_solicited && idev &&
1027 idev->cnf.drop_unsolicited_na)
1030 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
1031 ND_PRINTK(2, warn, "NS: invalid ND option\n");
1034 if (ndopts.nd_opts_tgt_lladdr) {
1035 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1038 "NA: invalid link-layer address length\n");
1042 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1044 if (skb->pkt_type != PACKET_LOOPBACK
1045 && (ifp->flags & IFA_F_TENTATIVE)) {
1046 addrconf_dad_failure(skb, ifp);
1049 /* What should we make now? The advertisement
1050 is invalid, but ndisc specs say nothing
1051 about it. It could be misconfiguration, or
1052 an smart proxy agent tries to help us :-)
1054 We should not print the error if NA has been
1055 received from loopback - it is just our own
1056 unsolicited advertisement.
1058 if (skb->pkt_type != PACKET_LOOPBACK)
1060 "NA: %pM advertised our address %pI6c on %s!\n",
1061 eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1066 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1068 /* RFC 9131 updates original Neighbour Discovery RFC 4861.
1069 * NAs with Target LL Address option without a corresponding
1070 * entry in the neighbour cache can now create a STALE neighbour
1071 * cache entry on routers.
1073 * entry accept fwding solicited behaviour
1074 * ------- ------ ------ --------- ----------------------
1075 * present X X 0 Set state to STALE
1076 * present X X 1 Set state to REACHABLE
1077 * absent 0 X X Do nothing
1078 * absent 1 0 X Do nothing
1079 * absent 1 1 X Add a new STALE entry
1081 * Note that we don't do a (daddr == all-routers-mcast) check.
1083 new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1084 if (!neigh && lladdr && idev && idev->cnf.forwarding) {
1085 if (accept_untracked_na(dev, saddr)) {
1086 neigh = neigh_create(&nd_tbl, &msg->target, dev);
1087 new_state = NUD_STALE;
1091 if (neigh && !IS_ERR(neigh)) {
1092 u8 old_flags = neigh->flags;
1093 struct net *net = dev_net(dev);
1095 if (READ_ONCE(neigh->nud_state) & NUD_FAILED)
1099 * Don't update the neighbor cache entry on a proxy NA from
1100 * ourselves because either the proxied node is off link or it
1101 * has already sent a NA to us.
1103 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1104 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1105 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1106 /* XXX: idev->cnf.proxy_ndp */
1110 ndisc_update(dev, neigh, lladdr,
1112 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1113 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1114 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1115 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1116 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1118 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1120 * Change: router to host
1122 rt6_clean_tohost(dev_net(dev), saddr);
1126 neigh_release(neigh);
1130 static void ndisc_recv_rs(struct sk_buff *skb)
1132 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1133 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1134 struct neighbour *neigh;
1135 struct inet6_dev *idev;
1136 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1137 struct ndisc_options ndopts;
1140 if (skb->len < sizeof(*rs_msg))
1143 idev = __in6_dev_get(skb->dev);
1145 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1149 /* Don't accept RS if we're not in router mode */
1150 if (!idev->cnf.forwarding)
1154 * Don't update NCE if src = ::;
1155 * this implies that the source node has no ip address assigned yet.
1157 if (ipv6_addr_any(saddr))
1160 /* Parse ND options */
1161 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1162 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1166 if (ndopts.nd_opts_src_lladdr) {
1167 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1173 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1175 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1176 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1177 NEIGH_UPDATE_F_OVERRIDE|
1178 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1179 NDISC_ROUTER_SOLICITATION, &ndopts);
1180 neigh_release(neigh);
1186 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1188 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1189 struct sk_buff *skb;
1190 struct nlmsghdr *nlh;
1191 struct nduseroptmsg *ndmsg;
1192 struct net *net = dev_net(ra->dev);
1194 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1195 + (opt->nd_opt_len << 3));
1196 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1198 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1204 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1206 goto nla_put_failure;
1209 ndmsg = nlmsg_data(nlh);
1210 ndmsg->nduseropt_family = AF_INET6;
1211 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1212 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1213 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1214 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1216 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1218 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1219 goto nla_put_failure;
1220 nlmsg_end(skb, nlh);
1222 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1229 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1232 static void ndisc_router_discovery(struct sk_buff *skb)
1234 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1235 struct neighbour *neigh = NULL;
1236 struct inet6_dev *in6_dev;
1237 struct fib6_info *rt = NULL;
1238 u32 defrtr_usr_metric;
1241 struct ndisc_options ndopts;
1243 unsigned int pref = 0;
1245 bool send_ifinfo_notify = false;
1247 __u8 *opt = (__u8 *)(ra_msg + 1);
1249 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1250 sizeof(struct ra_msg);
1253 "RA: %s, dev: %s\n",
1254 __func__, skb->dev->name);
1255 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1256 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1260 ND_PRINTK(2, warn, "RA: packet too short\n");
1264 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1265 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1266 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1272 * set the RA_RECV flag in the interface
1275 in6_dev = __in6_dev_get(skb->dev);
1277 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1282 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1283 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1287 if (!ipv6_accept_ra(in6_dev)) {
1289 "RA: %s, did not accept ra for dev: %s\n",
1290 __func__, skb->dev->name);
1291 goto skip_linkparms;
1294 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1295 /* skip link-specific parameters from interior routers */
1296 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1298 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1299 __func__, skb->dev->name);
1300 goto skip_linkparms;
1304 if (in6_dev->if_flags & IF_RS_SENT) {
1306 * flag that an RA was received after an RS was sent
1307 * out on this interface.
1309 in6_dev->if_flags |= IF_RA_RCVD;
1313 * Remember the managed/otherconf flags from most recently
1314 * received RA message (RFC 2462) -- yoshfuji
1316 old_if_flags = in6_dev->if_flags;
1317 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1319 (ra_msg->icmph.icmp6_addrconf_managed ?
1320 IF_RA_MANAGED : 0) |
1321 (ra_msg->icmph.icmp6_addrconf_other ?
1322 IF_RA_OTHERCONF : 0);
1324 if (old_if_flags != in6_dev->if_flags)
1325 send_ifinfo_notify = true;
1327 if (!in6_dev->cnf.accept_ra_defrtr) {
1329 "RA: %s, defrtr is false for dev: %s\n",
1330 __func__, skb->dev->name);
1334 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1335 if (lifetime != 0 && lifetime < in6_dev->cnf.accept_ra_min_lft) {
1337 "RA: router lifetime (%ds) is too short: %s\n",
1338 lifetime, skb->dev->name);
1342 /* Do not accept RA with source-addr found on local machine unless
1343 * accept_ra_from_local is set to true.
1345 net = dev_net(in6_dev->dev);
1346 if (!in6_dev->cnf.accept_ra_from_local &&
1347 ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1349 "RA from local address detected on dev: %s: default router ignored\n",
1354 #ifdef CONFIG_IPV6_ROUTER_PREF
1355 pref = ra_msg->icmph.icmp6_router_pref;
1356 /* 10b is handled as if it were 00b (medium) */
1357 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1358 !in6_dev->cnf.accept_ra_rtr_pref)
1359 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1361 /* routes added from RAs do not use nexthop objects */
1362 rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1364 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1365 rt->fib6_nh->fib_nh_dev, NULL,
1366 &ipv6_hdr(skb)->saddr);
1369 "RA: %s got default router without neighbour\n",
1371 fib6_info_release(rt);
1375 /* Set default route metric as specified by user */
1376 defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1377 /* delete the route if lifetime is 0 or if metric needs change */
1378 if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1379 ip6_del_rt(net, rt, false);
1383 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, metric: %d, for dev: %s\n",
1384 rt, lifetime, defrtr_usr_metric, skb->dev->name);
1385 if (!rt && lifetime) {
1386 ND_PRINTK(3, info, "RA: adding default router\n");
1389 neigh_release(neigh);
1391 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1392 skb->dev, pref, defrtr_usr_metric);
1395 "RA: %s failed to add default route\n",
1400 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1401 rt->fib6_nh->fib_nh_dev, NULL,
1402 &ipv6_hdr(skb)->saddr);
1405 "RA: %s got default router without neighbour\n",
1407 fib6_info_release(rt);
1410 neigh->flags |= NTF_ROUTER;
1411 } else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1412 struct nl_info nlinfo = {
1415 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1416 inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1420 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1421 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1422 ra_msg->icmph.icmp6_hop_limit) {
1423 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1424 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1425 fib6_metric_set(rt, RTAX_HOPLIMIT,
1426 ra_msg->icmph.icmp6_hop_limit);
1428 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1435 * Update Reachable Time and Retrans Timer
1438 if (in6_dev->nd_parms) {
1439 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1441 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1442 rtime = (rtime*HZ)/1000;
1445 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1446 in6_dev->tstamp = jiffies;
1447 send_ifinfo_notify = true;
1450 rtime = ntohl(ra_msg->reachable_time);
1451 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1452 rtime = (rtime*HZ)/1000;
1457 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1458 NEIGH_VAR_SET(in6_dev->nd_parms,
1459 BASE_REACHABLE_TIME, rtime);
1460 NEIGH_VAR_SET(in6_dev->nd_parms,
1461 GC_STALETIME, 3 * rtime);
1462 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1463 in6_dev->tstamp = jiffies;
1464 send_ifinfo_notify = true;
1476 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1480 if (ndopts.nd_opts_src_lladdr) {
1481 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1485 "RA: invalid link-layer address length\n");
1489 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1490 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1491 NEIGH_UPDATE_F_OVERRIDE|
1492 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1493 NEIGH_UPDATE_F_ISROUTER,
1494 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1497 if (!ipv6_accept_ra(in6_dev)) {
1499 "RA: %s, accept_ra is false for dev: %s\n",
1500 __func__, skb->dev->name);
1504 #ifdef CONFIG_IPV6_ROUTE_INFO
1505 if (!in6_dev->cnf.accept_ra_from_local &&
1506 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1509 "RA from local address detected on dev: %s: router info ignored.\n",
1511 goto skip_routeinfo;
1514 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1515 struct nd_opt_hdr *p;
1516 for (p = ndopts.nd_opts_ri;
1518 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1519 struct route_info *ri = (struct route_info *)p;
1520 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1521 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1522 ri->prefix_len == 0)
1525 if (ri->prefix_len == 0 &&
1526 !in6_dev->cnf.accept_ra_defrtr)
1528 if (ri->lifetime != 0 &&
1529 ntohl(ri->lifetime) < in6_dev->cnf.accept_ra_min_lft)
1531 if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1533 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1535 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1536 &ipv6_hdr(skb)->saddr);
1543 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1544 /* skip link-specific ndopts from interior routers */
1545 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1547 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1548 __func__, skb->dev->name);
1553 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1554 struct nd_opt_hdr *p;
1555 for (p = ndopts.nd_opts_pi;
1557 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1558 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1559 (p->nd_opt_len) << 3,
1560 ndopts.nd_opts_src_lladdr != NULL);
1564 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1568 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1571 if (in6_dev->ra_mtu != mtu) {
1572 in6_dev->ra_mtu = mtu;
1573 send_ifinfo_notify = true;
1576 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1577 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1578 } else if (in6_dev->cnf.mtu6 != mtu) {
1579 in6_dev->cnf.mtu6 = mtu;
1580 fib6_metric_set(rt, RTAX_MTU, mtu);
1581 rt6_mtu_change(skb->dev, mtu);
1585 if (ndopts.nd_useropts) {
1586 struct nd_opt_hdr *p;
1587 for (p = ndopts.nd_useropts;
1589 p = ndisc_next_useropt(skb->dev, p,
1590 ndopts.nd_useropts_end)) {
1591 ndisc_ra_useropt(skb, p);
1595 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1596 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1599 /* Send a notify if RA changed managed/otherconf flags or
1600 * timer settings or ra_mtu value
1602 if (send_ifinfo_notify)
1603 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1605 fib6_info_release(rt);
1607 neigh_release(neigh);
1610 static void ndisc_redirect_rcv(struct sk_buff *skb)
1613 struct ndisc_options ndopts;
1614 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1615 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1616 offsetof(struct rd_msg, opt));
1618 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1619 switch (skb->ndisc_nodetype) {
1620 case NDISC_NODETYPE_HOST:
1621 case NDISC_NODETYPE_NODEFAULT:
1623 "Redirect: from host or unauthorized router\n");
1628 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1630 "Redirect: source address is not link-local\n");
1634 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1637 if (!ndopts.nd_opts_rh) {
1638 ip6_redirect_no_header(skb, dev_net(skb->dev),
1643 hdr = (u8 *)ndopts.nd_opts_rh;
1645 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1648 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1651 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1652 struct sk_buff *orig_skb,
1655 u8 *opt = skb_put(skb, rd_len);
1658 *(opt++) = ND_OPT_REDIRECT_HDR;
1659 *(opt++) = (rd_len >> 3);
1662 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1666 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1668 struct net_device *dev = skb->dev;
1669 struct net *net = dev_net(dev);
1670 struct sock *sk = net->ipv6.ndisc_sk;
1672 struct inet_peer *peer;
1673 struct sk_buff *buff;
1675 struct in6_addr saddr_buf;
1676 struct rt6_info *rt;
1677 struct dst_entry *dst;
1680 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1681 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1684 if (netif_is_l3_master(skb->dev)) {
1685 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1690 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1691 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1696 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1697 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1699 "Redirect: target address is not link-local unicast\n");
1703 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1704 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1706 dst = ip6_route_output(net, NULL, &fl6);
1711 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1715 rt = (struct rt6_info *) dst;
1717 if (rt->rt6i_flags & RTF_GATEWAY) {
1719 "Redirect: destination is not a neighbour\n");
1722 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1723 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1729 if (dev->addr_len) {
1730 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1733 "Redirect: no neigh for target address\n");
1737 read_lock_bh(&neigh->lock);
1738 if (neigh->nud_state & NUD_VALID) {
1739 memcpy(ha_buf, neigh->ha, dev->addr_len);
1740 read_unlock_bh(&neigh->lock);
1742 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1746 read_unlock_bh(&neigh->lock);
1748 neigh_release(neigh);
1751 rd_len = min_t(unsigned int,
1752 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1757 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1761 msg = skb_put(buff, sizeof(*msg));
1762 *msg = (struct rd_msg) {
1764 .icmp6_type = NDISC_REDIRECT,
1767 .dest = ipv6_hdr(skb)->daddr,
1771 * include target_address option
1775 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1778 * build redirect option and copy skb over to the new packet.
1782 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1784 skb_dst_set(buff, dst);
1785 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1792 static void pndisc_redo(struct sk_buff *skb)
1798 static int ndisc_is_multicast(const void *pkey)
1800 return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1803 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1805 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1809 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1810 idev->cnf.suppress_frag_ndisc) {
1811 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1817 int ndisc_rcv(struct sk_buff *skb)
1821 if (ndisc_suppress_frag_ndisc(skb))
1824 if (skb_linearize(skb))
1827 msg = (struct nd_msg *)skb_transport_header(skb);
1829 __skb_push(skb, skb->data - skb_transport_header(skb));
1831 if (ipv6_hdr(skb)->hop_limit != 255) {
1832 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1833 ipv6_hdr(skb)->hop_limit);
1837 if (msg->icmph.icmp6_code != 0) {
1838 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1839 msg->icmph.icmp6_code);
1843 switch (msg->icmph.icmp6_type) {
1844 case NDISC_NEIGHBOUR_SOLICITATION:
1845 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1849 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1853 case NDISC_ROUTER_SOLICITATION:
1857 case NDISC_ROUTER_ADVERTISEMENT:
1858 ndisc_router_discovery(skb);
1861 case NDISC_REDIRECT:
1862 ndisc_redirect_rcv(skb);
1869 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1871 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1872 struct netdev_notifier_change_info *change_info;
1873 struct net *net = dev_net(dev);
1874 struct inet6_dev *idev;
1875 bool evict_nocarrier;
1878 case NETDEV_CHANGEADDR:
1879 neigh_changeaddr(&nd_tbl, dev);
1880 fib6_run_gc(0, net, false);
1883 idev = in6_dev_get(dev);
1886 if (idev->cnf.ndisc_notify ||
1887 net->ipv6.devconf_all->ndisc_notify)
1888 ndisc_send_unsol_na(dev);
1892 idev = in6_dev_get(dev);
1894 evict_nocarrier = true;
1896 evict_nocarrier = idev->cnf.ndisc_evict_nocarrier &&
1897 net->ipv6.devconf_all->ndisc_evict_nocarrier;
1902 if (change_info->flags_changed & IFF_NOARP)
1903 neigh_changeaddr(&nd_tbl, dev);
1904 if (evict_nocarrier && !netif_carrier_ok(dev))
1905 neigh_carrier_down(&nd_tbl, dev);
1908 neigh_ifdown(&nd_tbl, dev);
1909 fib6_run_gc(0, net, false);
1911 case NETDEV_NOTIFY_PEERS:
1912 ndisc_send_unsol_na(dev);
1921 static struct notifier_block ndisc_netdev_notifier = {
1922 .notifier_call = ndisc_netdev_event,
1923 .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1926 #ifdef CONFIG_SYSCTL
1927 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1928 const char *func, const char *dev_name)
1930 static char warncomm[TASK_COMM_LEN];
1932 if (strcmp(warncomm, current->comm) && warned < 5) {
1933 strcpy(warncomm, current->comm);
1934 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1936 dev_name, ctl->procname,
1937 dev_name, ctl->procname);
1942 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1943 size_t *lenp, loff_t *ppos)
1945 struct net_device *dev = ctl->extra1;
1946 struct inet6_dev *idev;
1949 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1950 (strcmp(ctl->procname, "base_reachable_time") == 0))
1951 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1953 if (strcmp(ctl->procname, "retrans_time") == 0)
1954 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1956 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1957 ret = neigh_proc_dointvec_jiffies(ctl, write,
1958 buffer, lenp, ppos);
1960 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1961 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1962 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1963 buffer, lenp, ppos);
1967 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1968 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1969 idev->nd_parms->reachable_time =
1970 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1971 idev->tstamp = jiffies;
1972 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1981 static int __net_init ndisc_net_init(struct net *net)
1983 struct ipv6_pinfo *np;
1987 err = inet_ctl_sock_create(&sk, PF_INET6,
1988 SOCK_RAW, IPPROTO_ICMPV6, net);
1991 "NDISC: Failed to initialize the control socket (err %d)\n",
1996 net->ipv6.ndisc_sk = sk;
1999 np->hop_limit = 255;
2000 /* Do not loopback ndisc messages */
2006 static void __net_exit ndisc_net_exit(struct net *net)
2008 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
2011 static struct pernet_operations ndisc_net_ops = {
2012 .init = ndisc_net_init,
2013 .exit = ndisc_net_exit,
2016 int __init ndisc_init(void)
2020 err = register_pernet_subsys(&ndisc_net_ops);
2024 * Initialize the neighbour table
2026 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2028 #ifdef CONFIG_SYSCTL
2029 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2030 ndisc_ifinfo_sysctl_change);
2032 goto out_unregister_pernet;
2037 #ifdef CONFIG_SYSCTL
2038 out_unregister_pernet:
2039 unregister_pernet_subsys(&ndisc_net_ops);
2044 int __init ndisc_late_init(void)
2046 return register_netdevice_notifier(&ndisc_netdev_notifier);
2049 void ndisc_late_cleanup(void)
2051 unregister_netdevice_notifier(&ndisc_netdev_notifier);
2054 void ndisc_cleanup(void)
2056 #ifdef CONFIG_SYSCTL
2057 neigh_sysctl_unregister(&nd_tbl.parms);
2059 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2060 unregister_pernet_subsys(&ndisc_net_ops);