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_RDNSS ||
201 opt->nd_opt_type == ND_OPT_DNSSL ||
202 opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
203 opt->nd_opt_type == ND_OPT_PREF64 ||
204 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
207 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
208 struct nd_opt_hdr *cur,
209 struct nd_opt_hdr *end)
211 if (!cur || !end || cur >= end)
214 cur = ((void *)cur) + (cur->nd_opt_len << 3);
215 } while (cur < end && !ndisc_is_useropt(dev, cur));
216 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
219 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
220 u8 *opt, int opt_len,
221 struct ndisc_options *ndopts)
223 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
225 if (!nd_opt || opt_len < 0 || !ndopts)
227 memset(ndopts, 0, sizeof(*ndopts));
230 if (opt_len < sizeof(struct nd_opt_hdr))
232 l = nd_opt->nd_opt_len << 3;
233 if (opt_len < l || l == 0)
235 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
237 switch (nd_opt->nd_opt_type) {
238 case ND_OPT_SOURCE_LL_ADDR:
239 case ND_OPT_TARGET_LL_ADDR:
242 case ND_OPT_REDIRECT_HDR:
243 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
245 "%s: duplicated ND6 option found: type=%d\n",
246 __func__, nd_opt->nd_opt_type);
248 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
251 case ND_OPT_PREFIX_INFO:
252 ndopts->nd_opts_pi_end = nd_opt;
253 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
254 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
256 #ifdef CONFIG_IPV6_ROUTE_INFO
257 case ND_OPT_ROUTE_INFO:
258 ndopts->nd_opts_ri_end = nd_opt;
259 if (!ndopts->nd_opts_ri)
260 ndopts->nd_opts_ri = nd_opt;
264 if (ndisc_is_useropt(dev, nd_opt)) {
265 ndopts->nd_useropts_end = nd_opt;
266 if (!ndopts->nd_useropts)
267 ndopts->nd_useropts = nd_opt;
270 * Unknown options must be silently ignored,
271 * to accommodate future extension to the
275 "%s: ignored unsupported option; type=%d, len=%d\n",
283 nd_opt = ((void *)nd_opt) + l;
288 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
292 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
294 ipv6_eth_mc_map(addr, buf);
297 ipv6_arcnet_mc_map(addr, buf);
299 case ARPHRD_INFINIBAND:
300 ipv6_ib_mc_map(addr, dev->broadcast, buf);
303 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
306 memcpy(buf, dev->broadcast, dev->addr_len);
312 EXPORT_SYMBOL(ndisc_mc_map);
314 static u32 ndisc_hash(const void *pkey,
315 const struct net_device *dev,
318 return ndisc_hashfn(pkey, dev, hash_rnd);
321 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
323 return neigh_key_eq128(n, pkey);
326 static int ndisc_constructor(struct neighbour *neigh)
328 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
329 struct net_device *dev = neigh->dev;
330 struct inet6_dev *in6_dev;
331 struct neigh_parms *parms;
332 bool is_multicast = ipv6_addr_is_multicast(addr);
334 in6_dev = in6_dev_get(dev);
339 parms = in6_dev->nd_parms;
340 __neigh_parms_put(neigh->parms);
341 neigh->parms = neigh_parms_clone(parms);
343 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
344 if (!dev->header_ops) {
345 neigh->nud_state = NUD_NOARP;
346 neigh->ops = &ndisc_direct_ops;
347 neigh->output = neigh_direct_output;
350 neigh->nud_state = NUD_NOARP;
351 ndisc_mc_map(addr, neigh->ha, dev, 1);
352 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
353 neigh->nud_state = NUD_NOARP;
354 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
355 if (dev->flags&IFF_LOOPBACK)
356 neigh->type = RTN_LOCAL;
357 } else if (dev->flags&IFF_POINTOPOINT) {
358 neigh->nud_state = NUD_NOARP;
359 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
361 if (dev->header_ops->cache)
362 neigh->ops = &ndisc_hh_ops;
364 neigh->ops = &ndisc_generic_ops;
365 if (neigh->nud_state&NUD_VALID)
366 neigh->output = neigh->ops->connected_output;
368 neigh->output = neigh->ops->output;
370 in6_dev_put(in6_dev);
374 static int pndisc_constructor(struct pneigh_entry *n)
376 struct in6_addr *addr = (struct in6_addr *)&n->key;
377 struct in6_addr maddr;
378 struct net_device *dev = n->dev;
380 if (!dev || !__in6_dev_get(dev))
382 addrconf_addr_solict_mult(addr, &maddr);
383 ipv6_dev_mc_inc(dev, &maddr);
387 static void pndisc_destructor(struct pneigh_entry *n)
389 struct in6_addr *addr = (struct in6_addr *)&n->key;
390 struct in6_addr maddr;
391 struct net_device *dev = n->dev;
393 if (!dev || !__in6_dev_get(dev))
395 addrconf_addr_solict_mult(addr, &maddr);
396 ipv6_dev_mc_dec(dev, &maddr);
399 /* called with rtnl held */
400 static bool ndisc_allow_add(const struct net_device *dev,
401 struct netlink_ext_ack *extack)
403 struct inet6_dev *idev = __in6_dev_get(dev);
405 if (!idev || idev->cnf.disable_ipv6) {
406 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
413 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
416 int hlen = LL_RESERVED_SPACE(dev);
417 int tlen = dev->needed_tailroom;
418 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
421 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
423 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
428 skb->protocol = htons(ETH_P_IPV6);
431 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
432 skb_reset_transport_header(skb);
434 /* Manually assign socket ownership as we avoid calling
435 * sock_alloc_send_pskb() to bypass wmem buffer limits
437 skb_set_owner_w(skb, sk);
442 static void ip6_nd_hdr(struct sk_buff *skb,
443 const struct in6_addr *saddr,
444 const struct in6_addr *daddr,
445 int hop_limit, int len)
448 struct inet6_dev *idev;
452 idev = __in6_dev_get(skb->dev);
453 tclass = idev ? idev->cnf.ndisc_tclass : 0;
456 skb_push(skb, sizeof(*hdr));
457 skb_reset_network_header(skb);
460 ip6_flow_hdr(hdr, tclass, 0);
462 hdr->payload_len = htons(len);
463 hdr->nexthdr = IPPROTO_ICMPV6;
464 hdr->hop_limit = hop_limit;
470 void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
471 const struct in6_addr *saddr)
473 struct dst_entry *dst = skb_dst(skb);
474 struct net *net = dev_net(skb->dev);
475 struct sock *sk = net->ipv6.ndisc_sk;
476 struct inet6_dev *idev;
478 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
481 type = icmp6h->icmp6_type;
485 int oif = skb->dev->ifindex;
487 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
488 dst = icmp6_dst_alloc(skb->dev, &fl6);
494 skb_dst_set(skb, dst);
497 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
502 ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
505 idev = __in6_dev_get(dst->dev);
506 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
508 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
509 net, sk, skb, NULL, dst->dev,
512 ICMP6MSGOUT_INC_STATS(net, idev, type);
513 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
518 EXPORT_SYMBOL(ndisc_send_skb);
520 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
521 const struct in6_addr *solicited_addr,
522 bool router, bool solicited, bool override, bool inc_opt)
525 struct in6_addr tmpaddr;
526 struct inet6_ifaddr *ifp;
527 const struct in6_addr *src_addr;
531 /* for anycast or proxy, solicited_addr != src_addr */
532 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
534 src_addr = solicited_addr;
535 if (ifp->flags & IFA_F_OPTIMISTIC)
537 inc_opt |= ifp->idev->cnf.force_tllao;
540 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
541 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
550 optlen += ndisc_opt_addr_space(dev,
551 NDISC_NEIGHBOUR_ADVERTISEMENT);
553 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
557 msg = skb_put(skb, sizeof(*msg));
558 *msg = (struct nd_msg) {
560 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
561 .icmp6_router = router,
562 .icmp6_solicited = solicited,
563 .icmp6_override = override,
565 .target = *solicited_addr,
569 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
571 NDISC_NEIGHBOUR_ADVERTISEMENT);
573 ndisc_send_skb(skb, daddr, src_addr);
576 static void ndisc_send_unsol_na(struct net_device *dev)
578 struct inet6_dev *idev;
579 struct inet6_ifaddr *ifa;
581 idev = in6_dev_get(dev);
585 read_lock_bh(&idev->lock);
586 list_for_each_entry(ifa, &idev->addr_list, if_list) {
587 /* skip tentative addresses until dad completes */
588 if (ifa->flags & IFA_F_TENTATIVE &&
589 !(ifa->flags & IFA_F_OPTIMISTIC))
592 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
593 /*router=*/ !!idev->cnf.forwarding,
594 /*solicited=*/ false, /*override=*/ true,
597 read_unlock_bh(&idev->lock);
602 struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
603 const struct in6_addr *saddr, u64 nonce)
605 int inc_opt = dev->addr_len;
613 if (ipv6_addr_any(saddr))
616 optlen += ndisc_opt_addr_space(dev,
617 NDISC_NEIGHBOUR_SOLICITATION);
621 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
625 msg = skb_put(skb, sizeof(*msg));
626 *msg = (struct nd_msg) {
628 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
634 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
636 NDISC_NEIGHBOUR_SOLICITATION);
638 u8 *opt = skb_put(skb, 8);
640 opt[0] = ND_OPT_NONCE;
642 memcpy(opt + 2, &nonce, 6);
647 EXPORT_SYMBOL(ndisc_ns_create);
649 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
650 const struct in6_addr *daddr, const struct in6_addr *saddr,
653 struct in6_addr addr_buf;
657 if (ipv6_get_lladdr(dev, &addr_buf,
658 (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
663 skb = ndisc_ns_create(dev, solicit, saddr, nonce);
666 ndisc_send_skb(skb, daddr, saddr);
669 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
670 const struct in6_addr *daddr)
674 int send_sllao = dev->addr_len;
677 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
679 * According to section 2.2 of RFC 4429, we must not
680 * send router solicitations with a sllao from
681 * optimistic addresses, but we may send the solicitation
682 * if we don't include the sllao. So here we check
683 * if our address is optimistic, and if so, we
684 * suppress the inclusion of the sllao.
687 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
690 if (ifp->flags & IFA_F_OPTIMISTIC) {
700 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
702 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
706 msg = skb_put(skb, sizeof(*msg));
707 *msg = (struct rs_msg) {
709 .icmp6_type = NDISC_ROUTER_SOLICITATION,
714 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
716 NDISC_ROUTER_SOLICITATION);
718 ndisc_send_skb(skb, daddr, saddr);
722 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
725 * "The sender MUST return an ICMP
726 * destination unreachable"
728 dst_link_failure(skb);
732 /* Called with locked neigh: either read or both */
734 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
736 struct in6_addr *saddr = NULL;
737 struct in6_addr mcaddr;
738 struct net_device *dev = neigh->dev;
739 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
740 int probes = atomic_read(&neigh->probes);
742 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
744 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
745 saddr = &ipv6_hdr(skb)->saddr;
746 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
748 if (!(neigh->nud_state & NUD_VALID)) {
750 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
753 ndisc_send_ns(dev, target, target, saddr, 0);
754 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
757 addrconf_addr_solict_mult(target, &mcaddr);
758 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
762 static int pndisc_is_router(const void *pkey,
763 struct net_device *dev)
765 struct pneigh_entry *n;
768 read_lock_bh(&nd_tbl.lock);
769 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
771 ret = !!(n->flags & NTF_ROUTER);
772 read_unlock_bh(&nd_tbl.lock);
777 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
778 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
779 struct ndisc_options *ndopts)
781 neigh_update(neigh, lladdr, new, flags, 0);
782 /* report ndisc ops about neighbour update */
783 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
786 static void ndisc_recv_ns(struct sk_buff *skb)
788 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
789 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
790 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
792 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
793 offsetof(struct nd_msg, opt));
794 struct ndisc_options ndopts;
795 struct net_device *dev = skb->dev;
796 struct inet6_ifaddr *ifp;
797 struct inet6_dev *idev = NULL;
798 struct neighbour *neigh;
799 int dad = ipv6_addr_any(saddr);
804 if (skb->len < sizeof(struct nd_msg)) {
805 ND_PRINTK(2, warn, "NS: packet too short\n");
809 if (ipv6_addr_is_multicast(&msg->target)) {
810 ND_PRINTK(2, warn, "NS: multicast target address\n");
816 * DAD has to be destined for solicited node multicast address.
818 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
819 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
823 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
824 ND_PRINTK(2, warn, "NS: invalid ND options\n");
828 if (ndopts.nd_opts_src_lladdr) {
829 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
832 "NS: invalid link-layer address length\n");
837 * If the IP source address is the unspecified address,
838 * there MUST NOT be source link-layer address option
843 "NS: bad DAD packet (link-layer address option)\n");
847 if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
848 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
850 inc = ipv6_addr_is_multicast(daddr);
852 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
855 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
857 if (nonce != 0 && ifp->dad_nonce == nonce) {
858 u8 *np = (u8 *)&nonce;
859 /* Matching nonce if looped back */
861 "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
862 ifp->idev->dev->name,
867 * We are colliding with another node
869 * so fail our DAD process
871 addrconf_dad_failure(skb, ifp);
875 * This is not a dad solicitation.
876 * If we are an optimistic node,
878 * Otherwise, we should ignore it.
880 if (!(ifp->flags & IFA_F_OPTIMISTIC))
887 struct net *net = dev_net(dev);
889 /* perhaps an address on the master device */
890 if (netif_is_l3_slave(dev)) {
891 struct net_device *mdev;
893 mdev = netdev_master_upper_dev_get_rcu(dev);
895 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
901 idev = in6_dev_get(dev);
903 /* XXX: count this drop? */
907 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
908 (idev->cnf.forwarding &&
909 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
910 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
911 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
912 skb->pkt_type != PACKET_HOST &&
914 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
916 * for anycast or proxy,
917 * sender should delay its response
918 * by a random time between 0 and
919 * MAX_ANYCAST_DELAY_TIME seconds.
920 * (RFC2461) -- yoshfuji
922 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
924 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
932 is_router = idev->cnf.forwarding;
935 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
936 !!is_router, false, (ifp != NULL), true);
941 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
943 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
946 * update / create cache entry
947 * for the source address
949 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
950 !inc || lladdr || !dev->addr_len);
952 ndisc_update(dev, neigh, lladdr, NUD_STALE,
953 NEIGH_UPDATE_F_WEAK_OVERRIDE|
954 NEIGH_UPDATE_F_OVERRIDE,
955 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
956 if (neigh || !dev->header_ops) {
957 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
958 true, (ifp != NULL && inc), inc);
960 neigh_release(neigh);
970 static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
972 struct inet6_dev *idev = __in6_dev_get(dev);
974 switch (idev->cnf.accept_untracked_na) {
975 case 0: /* Don't accept untracked na (absent in neighbor cache) */
977 case 1: /* Create new entries from na if currently untracked */
979 case 2: /* Create new entries from untracked na only if saddr is in the
980 * same subnet as an address configured on the interface that
983 return !!ipv6_chk_prefix(saddr, dev);
989 static void ndisc_recv_na(struct sk_buff *skb)
991 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
992 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
993 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
995 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
996 offsetof(struct nd_msg, opt));
997 struct ndisc_options ndopts;
998 struct net_device *dev = skb->dev;
999 struct inet6_dev *idev = __in6_dev_get(dev);
1000 struct inet6_ifaddr *ifp;
1001 struct neighbour *neigh;
1004 if (skb->len < sizeof(struct nd_msg)) {
1005 ND_PRINTK(2, warn, "NA: packet too short\n");
1009 if (ipv6_addr_is_multicast(&msg->target)) {
1010 ND_PRINTK(2, warn, "NA: target address is multicast\n");
1014 if (ipv6_addr_is_multicast(daddr) &&
1015 msg->icmph.icmp6_solicited) {
1016 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
1020 /* For some 802.11 wireless deployments (and possibly other networks),
1021 * there will be a NA proxy and unsolicitd packets are attacks
1022 * and thus should not be accepted.
1023 * drop_unsolicited_na takes precedence over accept_untracked_na
1025 if (!msg->icmph.icmp6_solicited && idev &&
1026 idev->cnf.drop_unsolicited_na)
1029 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
1030 ND_PRINTK(2, warn, "NS: invalid ND option\n");
1033 if (ndopts.nd_opts_tgt_lladdr) {
1034 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1037 "NA: invalid link-layer address length\n");
1041 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1043 if (skb->pkt_type != PACKET_LOOPBACK
1044 && (ifp->flags & IFA_F_TENTATIVE)) {
1045 addrconf_dad_failure(skb, ifp);
1048 /* What should we make now? The advertisement
1049 is invalid, but ndisc specs say nothing
1050 about it. It could be misconfiguration, or
1051 an smart proxy agent tries to help us :-)
1053 We should not print the error if NA has been
1054 received from loopback - it is just our own
1055 unsolicited advertisement.
1057 if (skb->pkt_type != PACKET_LOOPBACK)
1059 "NA: %pM advertised our address %pI6c on %s!\n",
1060 eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1065 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1067 /* RFC 9131 updates original Neighbour Discovery RFC 4861.
1068 * NAs with Target LL Address option without a corresponding
1069 * entry in the neighbour cache can now create a STALE neighbour
1070 * cache entry on routers.
1072 * entry accept fwding solicited behaviour
1073 * ------- ------ ------ --------- ----------------------
1074 * present X X 0 Set state to STALE
1075 * present X X 1 Set state to REACHABLE
1076 * absent 0 X X Do nothing
1077 * absent 1 0 X Do nothing
1078 * absent 1 1 X Add a new STALE entry
1080 * Note that we don't do a (daddr == all-routers-mcast) check.
1082 new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1083 if (!neigh && lladdr && idev && idev->cnf.forwarding) {
1084 if (accept_untracked_na(dev, saddr)) {
1085 neigh = neigh_create(&nd_tbl, &msg->target, dev);
1086 new_state = NUD_STALE;
1090 if (neigh && !IS_ERR(neigh)) {
1091 u8 old_flags = neigh->flags;
1092 struct net *net = dev_net(dev);
1094 if (neigh->nud_state & NUD_FAILED)
1098 * Don't update the neighbor cache entry on a proxy NA from
1099 * ourselves because either the proxied node is off link or it
1100 * has already sent a NA to us.
1102 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1103 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1104 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1105 /* XXX: idev->cnf.proxy_ndp */
1109 ndisc_update(dev, neigh, lladdr,
1111 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1112 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1113 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1114 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1115 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1117 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1119 * Change: router to host
1121 rt6_clean_tohost(dev_net(dev), saddr);
1125 neigh_release(neigh);
1129 static void ndisc_recv_rs(struct sk_buff *skb)
1131 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1132 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1133 struct neighbour *neigh;
1134 struct inet6_dev *idev;
1135 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1136 struct ndisc_options ndopts;
1139 if (skb->len < sizeof(*rs_msg))
1142 idev = __in6_dev_get(skb->dev);
1144 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1148 /* Don't accept RS if we're not in router mode */
1149 if (!idev->cnf.forwarding)
1153 * Don't update NCE if src = ::;
1154 * this implies that the source node has no ip address assigned yet.
1156 if (ipv6_addr_any(saddr))
1159 /* Parse ND options */
1160 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1161 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1165 if (ndopts.nd_opts_src_lladdr) {
1166 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1172 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1174 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1175 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1176 NEIGH_UPDATE_F_OVERRIDE|
1177 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1178 NDISC_ROUTER_SOLICITATION, &ndopts);
1179 neigh_release(neigh);
1185 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1187 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1188 struct sk_buff *skb;
1189 struct nlmsghdr *nlh;
1190 struct nduseroptmsg *ndmsg;
1191 struct net *net = dev_net(ra->dev);
1193 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1194 + (opt->nd_opt_len << 3));
1195 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1197 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1203 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1205 goto nla_put_failure;
1208 ndmsg = nlmsg_data(nlh);
1209 ndmsg->nduseropt_family = AF_INET6;
1210 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1211 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1212 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1213 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1215 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1217 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1218 goto nla_put_failure;
1219 nlmsg_end(skb, nlh);
1221 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1228 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1231 static void ndisc_router_discovery(struct sk_buff *skb)
1233 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1234 struct neighbour *neigh = NULL;
1235 struct inet6_dev *in6_dev;
1236 struct fib6_info *rt = NULL;
1237 u32 defrtr_usr_metric;
1240 struct ndisc_options ndopts;
1242 unsigned int pref = 0;
1244 bool send_ifinfo_notify = false;
1246 __u8 *opt = (__u8 *)(ra_msg + 1);
1248 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1249 sizeof(struct ra_msg);
1252 "RA: %s, dev: %s\n",
1253 __func__, skb->dev->name);
1254 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1255 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1259 ND_PRINTK(2, warn, "RA: packet too short\n");
1263 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1264 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1265 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1271 * set the RA_RECV flag in the interface
1274 in6_dev = __in6_dev_get(skb->dev);
1276 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1281 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1282 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1286 if (!ipv6_accept_ra(in6_dev)) {
1288 "RA: %s, did not accept ra for dev: %s\n",
1289 __func__, skb->dev->name);
1290 goto skip_linkparms;
1293 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1294 /* skip link-specific parameters from interior routers */
1295 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1297 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1298 __func__, skb->dev->name);
1299 goto skip_linkparms;
1303 if (in6_dev->if_flags & IF_RS_SENT) {
1305 * flag that an RA was received after an RS was sent
1306 * out on this interface.
1308 in6_dev->if_flags |= IF_RA_RCVD;
1312 * Remember the managed/otherconf flags from most recently
1313 * received RA message (RFC 2462) -- yoshfuji
1315 old_if_flags = in6_dev->if_flags;
1316 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1318 (ra_msg->icmph.icmp6_addrconf_managed ?
1319 IF_RA_MANAGED : 0) |
1320 (ra_msg->icmph.icmp6_addrconf_other ?
1321 IF_RA_OTHERCONF : 0);
1323 if (old_if_flags != in6_dev->if_flags)
1324 send_ifinfo_notify = true;
1326 if (!in6_dev->cnf.accept_ra_defrtr) {
1328 "RA: %s, defrtr is false for dev: %s\n",
1329 __func__, skb->dev->name);
1333 /* Do not accept RA with source-addr found on local machine unless
1334 * accept_ra_from_local is set to true.
1336 net = dev_net(in6_dev->dev);
1337 if (!in6_dev->cnf.accept_ra_from_local &&
1338 ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1340 "RA from local address detected on dev: %s: default router ignored\n",
1345 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1347 #ifdef CONFIG_IPV6_ROUTER_PREF
1348 pref = ra_msg->icmph.icmp6_router_pref;
1349 /* 10b is handled as if it were 00b (medium) */
1350 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1351 !in6_dev->cnf.accept_ra_rtr_pref)
1352 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1354 /* routes added from RAs do not use nexthop objects */
1355 rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1357 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1358 rt->fib6_nh->fib_nh_dev, NULL,
1359 &ipv6_hdr(skb)->saddr);
1362 "RA: %s got default router without neighbour\n",
1364 fib6_info_release(rt);
1368 /* Set default route metric as specified by user */
1369 defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1370 /* delete the route if lifetime is 0 or if metric needs change */
1371 if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1372 ip6_del_rt(net, rt, false);
1376 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, metric: %d, for dev: %s\n",
1377 rt, lifetime, defrtr_usr_metric, skb->dev->name);
1378 if (!rt && lifetime) {
1379 ND_PRINTK(3, info, "RA: adding default router\n");
1381 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1382 skb->dev, pref, defrtr_usr_metric);
1385 "RA: %s failed to add default route\n",
1390 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1391 rt->fib6_nh->fib_nh_dev, NULL,
1392 &ipv6_hdr(skb)->saddr);
1395 "RA: %s got default router without neighbour\n",
1397 fib6_info_release(rt);
1400 neigh->flags |= NTF_ROUTER;
1401 } else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1402 struct nl_info nlinfo = {
1405 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1406 inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1410 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1411 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1412 ra_msg->icmph.icmp6_hop_limit) {
1413 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1414 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1415 fib6_metric_set(rt, RTAX_HOPLIMIT,
1416 ra_msg->icmph.icmp6_hop_limit);
1418 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1425 * Update Reachable Time and Retrans Timer
1428 if (in6_dev->nd_parms) {
1429 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1431 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1432 rtime = (rtime*HZ)/1000;
1435 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1436 in6_dev->tstamp = jiffies;
1437 send_ifinfo_notify = true;
1440 rtime = ntohl(ra_msg->reachable_time);
1441 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1442 rtime = (rtime*HZ)/1000;
1447 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1448 NEIGH_VAR_SET(in6_dev->nd_parms,
1449 BASE_REACHABLE_TIME, rtime);
1450 NEIGH_VAR_SET(in6_dev->nd_parms,
1451 GC_STALETIME, 3 * rtime);
1452 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1453 in6_dev->tstamp = jiffies;
1454 send_ifinfo_notify = true;
1466 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1470 if (ndopts.nd_opts_src_lladdr) {
1471 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1475 "RA: invalid link-layer address length\n");
1479 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1480 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1481 NEIGH_UPDATE_F_OVERRIDE|
1482 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1483 NEIGH_UPDATE_F_ISROUTER,
1484 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1487 if (!ipv6_accept_ra(in6_dev)) {
1489 "RA: %s, accept_ra is false for dev: %s\n",
1490 __func__, skb->dev->name);
1494 #ifdef CONFIG_IPV6_ROUTE_INFO
1495 if (!in6_dev->cnf.accept_ra_from_local &&
1496 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1499 "RA from local address detected on dev: %s: router info ignored.\n",
1501 goto skip_routeinfo;
1504 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1505 struct nd_opt_hdr *p;
1506 for (p = ndopts.nd_opts_ri;
1508 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1509 struct route_info *ri = (struct route_info *)p;
1510 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1511 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1512 ri->prefix_len == 0)
1515 if (ri->prefix_len == 0 &&
1516 !in6_dev->cnf.accept_ra_defrtr)
1518 if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1520 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1522 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1523 &ipv6_hdr(skb)->saddr);
1530 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1531 /* skip link-specific ndopts from interior routers */
1532 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1534 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1535 __func__, skb->dev->name);
1540 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1541 struct nd_opt_hdr *p;
1542 for (p = ndopts.nd_opts_pi;
1544 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1545 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1546 (p->nd_opt_len) << 3,
1547 ndopts.nd_opts_src_lladdr != NULL);
1551 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1555 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1558 if (in6_dev->ra_mtu != mtu) {
1559 in6_dev->ra_mtu = mtu;
1560 send_ifinfo_notify = true;
1563 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1564 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1565 } else if (in6_dev->cnf.mtu6 != mtu) {
1566 in6_dev->cnf.mtu6 = mtu;
1567 fib6_metric_set(rt, RTAX_MTU, mtu);
1568 rt6_mtu_change(skb->dev, mtu);
1572 if (ndopts.nd_useropts) {
1573 struct nd_opt_hdr *p;
1574 for (p = ndopts.nd_useropts;
1576 p = ndisc_next_useropt(skb->dev, p,
1577 ndopts.nd_useropts_end)) {
1578 ndisc_ra_useropt(skb, p);
1582 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1583 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1586 /* Send a notify if RA changed managed/otherconf flags or
1587 * timer settings or ra_mtu value
1589 if (send_ifinfo_notify)
1590 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1592 fib6_info_release(rt);
1594 neigh_release(neigh);
1597 static void ndisc_redirect_rcv(struct sk_buff *skb)
1600 struct ndisc_options ndopts;
1601 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1602 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1603 offsetof(struct rd_msg, opt));
1605 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1606 switch (skb->ndisc_nodetype) {
1607 case NDISC_NODETYPE_HOST:
1608 case NDISC_NODETYPE_NODEFAULT:
1610 "Redirect: from host or unauthorized router\n");
1615 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1617 "Redirect: source address is not link-local\n");
1621 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1624 if (!ndopts.nd_opts_rh) {
1625 ip6_redirect_no_header(skb, dev_net(skb->dev),
1630 hdr = (u8 *)ndopts.nd_opts_rh;
1632 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1635 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1638 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1639 struct sk_buff *orig_skb,
1642 u8 *opt = skb_put(skb, rd_len);
1645 *(opt++) = ND_OPT_REDIRECT_HDR;
1646 *(opt++) = (rd_len >> 3);
1649 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1653 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1655 struct net_device *dev = skb->dev;
1656 struct net *net = dev_net(dev);
1657 struct sock *sk = net->ipv6.ndisc_sk;
1659 struct inet_peer *peer;
1660 struct sk_buff *buff;
1662 struct in6_addr saddr_buf;
1663 struct rt6_info *rt;
1664 struct dst_entry *dst;
1667 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1668 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1671 if (netif_is_l3_master(skb->dev)) {
1672 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1677 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1678 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1683 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1684 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1686 "Redirect: target address is not link-local unicast\n");
1690 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1691 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1693 dst = ip6_route_output(net, NULL, &fl6);
1698 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1702 rt = (struct rt6_info *) dst;
1704 if (rt->rt6i_flags & RTF_GATEWAY) {
1706 "Redirect: destination is not a neighbour\n");
1709 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1710 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1716 if (dev->addr_len) {
1717 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1720 "Redirect: no neigh for target address\n");
1724 read_lock_bh(&neigh->lock);
1725 if (neigh->nud_state & NUD_VALID) {
1726 memcpy(ha_buf, neigh->ha, dev->addr_len);
1727 read_unlock_bh(&neigh->lock);
1729 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1733 read_unlock_bh(&neigh->lock);
1735 neigh_release(neigh);
1738 rd_len = min_t(unsigned int,
1739 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1744 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1748 msg = skb_put(buff, sizeof(*msg));
1749 *msg = (struct rd_msg) {
1751 .icmp6_type = NDISC_REDIRECT,
1754 .dest = ipv6_hdr(skb)->daddr,
1758 * include target_address option
1762 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1765 * build redirect option and copy skb over to the new packet.
1769 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1771 skb_dst_set(buff, dst);
1772 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1779 static void pndisc_redo(struct sk_buff *skb)
1785 static int ndisc_is_multicast(const void *pkey)
1787 return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1790 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1792 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1796 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1797 idev->cnf.suppress_frag_ndisc) {
1798 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1804 int ndisc_rcv(struct sk_buff *skb)
1808 if (ndisc_suppress_frag_ndisc(skb))
1811 if (skb_linearize(skb))
1814 msg = (struct nd_msg *)skb_transport_header(skb);
1816 __skb_push(skb, skb->data - skb_transport_header(skb));
1818 if (ipv6_hdr(skb)->hop_limit != 255) {
1819 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1820 ipv6_hdr(skb)->hop_limit);
1824 if (msg->icmph.icmp6_code != 0) {
1825 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1826 msg->icmph.icmp6_code);
1830 switch (msg->icmph.icmp6_type) {
1831 case NDISC_NEIGHBOUR_SOLICITATION:
1832 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1836 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1840 case NDISC_ROUTER_SOLICITATION:
1844 case NDISC_ROUTER_ADVERTISEMENT:
1845 ndisc_router_discovery(skb);
1848 case NDISC_REDIRECT:
1849 ndisc_redirect_rcv(skb);
1856 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1858 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1859 struct netdev_notifier_change_info *change_info;
1860 struct net *net = dev_net(dev);
1861 struct inet6_dev *idev;
1862 bool evict_nocarrier;
1865 case NETDEV_CHANGEADDR:
1866 neigh_changeaddr(&nd_tbl, dev);
1867 fib6_run_gc(0, net, false);
1870 idev = in6_dev_get(dev);
1873 if (idev->cnf.ndisc_notify ||
1874 net->ipv6.devconf_all->ndisc_notify)
1875 ndisc_send_unsol_na(dev);
1879 idev = in6_dev_get(dev);
1881 evict_nocarrier = true;
1883 evict_nocarrier = idev->cnf.ndisc_evict_nocarrier &&
1884 net->ipv6.devconf_all->ndisc_evict_nocarrier;
1889 if (change_info->flags_changed & IFF_NOARP)
1890 neigh_changeaddr(&nd_tbl, dev);
1891 if (evict_nocarrier && !netif_carrier_ok(dev))
1892 neigh_carrier_down(&nd_tbl, dev);
1895 neigh_ifdown(&nd_tbl, dev);
1896 fib6_run_gc(0, net, false);
1898 case NETDEV_NOTIFY_PEERS:
1899 ndisc_send_unsol_na(dev);
1908 static struct notifier_block ndisc_netdev_notifier = {
1909 .notifier_call = ndisc_netdev_event,
1910 .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1913 #ifdef CONFIG_SYSCTL
1914 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1915 const char *func, const char *dev_name)
1917 static char warncomm[TASK_COMM_LEN];
1919 if (strcmp(warncomm, current->comm) && warned < 5) {
1920 strcpy(warncomm, current->comm);
1921 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1923 dev_name, ctl->procname,
1924 dev_name, ctl->procname);
1929 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1930 size_t *lenp, loff_t *ppos)
1932 struct net_device *dev = ctl->extra1;
1933 struct inet6_dev *idev;
1936 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1937 (strcmp(ctl->procname, "base_reachable_time") == 0))
1938 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1940 if (strcmp(ctl->procname, "retrans_time") == 0)
1941 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1943 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1944 ret = neigh_proc_dointvec_jiffies(ctl, write,
1945 buffer, lenp, ppos);
1947 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1948 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1949 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1950 buffer, lenp, ppos);
1954 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1955 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1956 idev->nd_parms->reachable_time =
1957 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1958 idev->tstamp = jiffies;
1959 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1968 static int __net_init ndisc_net_init(struct net *net)
1970 struct ipv6_pinfo *np;
1974 err = inet_ctl_sock_create(&sk, PF_INET6,
1975 SOCK_RAW, IPPROTO_ICMPV6, net);
1978 "NDISC: Failed to initialize the control socket (err %d)\n",
1983 net->ipv6.ndisc_sk = sk;
1986 np->hop_limit = 255;
1987 /* Do not loopback ndisc messages */
1993 static void __net_exit ndisc_net_exit(struct net *net)
1995 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1998 static struct pernet_operations ndisc_net_ops = {
1999 .init = ndisc_net_init,
2000 .exit = ndisc_net_exit,
2003 int __init ndisc_init(void)
2007 err = register_pernet_subsys(&ndisc_net_ops);
2011 * Initialize the neighbour table
2013 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2015 #ifdef CONFIG_SYSCTL
2016 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2017 ndisc_ifinfo_sysctl_change);
2019 goto out_unregister_pernet;
2024 #ifdef CONFIG_SYSCTL
2025 out_unregister_pernet:
2026 unregister_pernet_subsys(&ndisc_net_ops);
2031 int __init ndisc_late_init(void)
2033 return register_netdevice_notifier(&ndisc_netdev_notifier);
2036 void ndisc_late_cleanup(void)
2038 unregister_netdevice_notifier(&ndisc_netdev_notifier);
2041 void ndisc_cleanup(void)
2043 #ifdef CONFIG_SYSCTL
2044 neigh_sysctl_unregister(&nd_tbl.parms);
2046 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2047 unregister_pernet_subsys(&ndisc_net_ops);