1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Neighbour Discovery for IPv6
4 * Linux INET6 implementation
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_GC_STALETIME] = 60 * HZ,
132 [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
133 [NEIGH_VAR_PROXY_QLEN] = 64,
134 [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
135 [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
138 .gc_interval = 30 * HZ,
143 EXPORT_SYMBOL_GPL(nd_tbl);
145 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, void *data,
146 int data_len, int pad)
148 int space = __ndisc_opt_addr_space(data_len, pad);
149 u8 *opt = skb_put(skb, space);
154 memset(opt + 2, 0, pad);
158 memcpy(opt+2, data, data_len);
163 memset(opt, 0, space);
165 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
167 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
168 void *data, u8 icmp6_type)
170 __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
171 ndisc_addr_option_pad(skb->dev->type));
172 ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
175 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
179 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
180 ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
183 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
184 struct nd_opt_hdr *end)
187 if (!cur || !end || cur >= end)
189 type = cur->nd_opt_type;
191 cur = ((void *)cur) + (cur->nd_opt_len << 3);
192 } while (cur < end && cur->nd_opt_type != type);
193 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
196 static inline int ndisc_is_useropt(const struct net_device *dev,
197 struct nd_opt_hdr *opt)
199 return opt->nd_opt_type == ND_OPT_RDNSS ||
200 opt->nd_opt_type == ND_OPT_DNSSL ||
201 opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
202 opt->nd_opt_type == ND_OPT_PREF64 ||
203 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
206 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
207 struct nd_opt_hdr *cur,
208 struct nd_opt_hdr *end)
210 if (!cur || !end || cur >= end)
213 cur = ((void *)cur) + (cur->nd_opt_len << 3);
214 } while (cur < end && !ndisc_is_useropt(dev, cur));
215 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
218 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
219 u8 *opt, int opt_len,
220 struct ndisc_options *ndopts)
222 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
224 if (!nd_opt || opt_len < 0 || !ndopts)
226 memset(ndopts, 0, sizeof(*ndopts));
229 if (opt_len < sizeof(struct nd_opt_hdr))
231 l = nd_opt->nd_opt_len << 3;
232 if (opt_len < l || l == 0)
234 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
236 switch (nd_opt->nd_opt_type) {
237 case ND_OPT_SOURCE_LL_ADDR:
238 case ND_OPT_TARGET_LL_ADDR:
241 case ND_OPT_REDIRECT_HDR:
242 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
244 "%s: duplicated ND6 option found: type=%d\n",
245 __func__, nd_opt->nd_opt_type);
247 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
250 case ND_OPT_PREFIX_INFO:
251 ndopts->nd_opts_pi_end = nd_opt;
252 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
253 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
255 #ifdef CONFIG_IPV6_ROUTE_INFO
256 case ND_OPT_ROUTE_INFO:
257 ndopts->nd_opts_ri_end = nd_opt;
258 if (!ndopts->nd_opts_ri)
259 ndopts->nd_opts_ri = nd_opt;
263 if (ndisc_is_useropt(dev, nd_opt)) {
264 ndopts->nd_useropts_end = nd_opt;
265 if (!ndopts->nd_useropts)
266 ndopts->nd_useropts = nd_opt;
269 * Unknown options must be silently ignored,
270 * to accommodate future extension to the
274 "%s: ignored unsupported option; type=%d, len=%d\n",
282 nd_opt = ((void *)nd_opt) + l;
287 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
291 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
293 ipv6_eth_mc_map(addr, buf);
296 ipv6_arcnet_mc_map(addr, buf);
298 case ARPHRD_INFINIBAND:
299 ipv6_ib_mc_map(addr, dev->broadcast, buf);
302 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
305 memcpy(buf, dev->broadcast, dev->addr_len);
311 EXPORT_SYMBOL(ndisc_mc_map);
313 static u32 ndisc_hash(const void *pkey,
314 const struct net_device *dev,
317 return ndisc_hashfn(pkey, dev, hash_rnd);
320 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
322 return neigh_key_eq128(n, pkey);
325 static int ndisc_constructor(struct neighbour *neigh)
327 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
328 struct net_device *dev = neigh->dev;
329 struct inet6_dev *in6_dev;
330 struct neigh_parms *parms;
331 bool is_multicast = ipv6_addr_is_multicast(addr);
333 in6_dev = in6_dev_get(dev);
338 parms = in6_dev->nd_parms;
339 __neigh_parms_put(neigh->parms);
340 neigh->parms = neigh_parms_clone(parms);
342 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
343 if (!dev->header_ops) {
344 neigh->nud_state = NUD_NOARP;
345 neigh->ops = &ndisc_direct_ops;
346 neigh->output = neigh_direct_output;
349 neigh->nud_state = NUD_NOARP;
350 ndisc_mc_map(addr, neigh->ha, dev, 1);
351 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
352 neigh->nud_state = NUD_NOARP;
353 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
354 if (dev->flags&IFF_LOOPBACK)
355 neigh->type = RTN_LOCAL;
356 } else if (dev->flags&IFF_POINTOPOINT) {
357 neigh->nud_state = NUD_NOARP;
358 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
360 if (dev->header_ops->cache)
361 neigh->ops = &ndisc_hh_ops;
363 neigh->ops = &ndisc_generic_ops;
364 if (neigh->nud_state&NUD_VALID)
365 neigh->output = neigh->ops->connected_output;
367 neigh->output = neigh->ops->output;
369 in6_dev_put(in6_dev);
373 static int pndisc_constructor(struct pneigh_entry *n)
375 struct in6_addr *addr = (struct in6_addr *)&n->key;
376 struct in6_addr maddr;
377 struct net_device *dev = n->dev;
379 if (!dev || !__in6_dev_get(dev))
381 addrconf_addr_solict_mult(addr, &maddr);
382 ipv6_dev_mc_inc(dev, &maddr);
386 static void pndisc_destructor(struct pneigh_entry *n)
388 struct in6_addr *addr = (struct in6_addr *)&n->key;
389 struct in6_addr maddr;
390 struct net_device *dev = n->dev;
392 if (!dev || !__in6_dev_get(dev))
394 addrconf_addr_solict_mult(addr, &maddr);
395 ipv6_dev_mc_dec(dev, &maddr);
398 /* called with rtnl held */
399 static bool ndisc_allow_add(const struct net_device *dev,
400 struct netlink_ext_ack *extack)
402 struct inet6_dev *idev = __in6_dev_get(dev);
404 if (!idev || idev->cnf.disable_ipv6) {
405 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
412 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
415 int hlen = LL_RESERVED_SPACE(dev);
416 int tlen = dev->needed_tailroom;
417 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
420 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
422 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
427 skb->protocol = htons(ETH_P_IPV6);
430 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
431 skb_reset_transport_header(skb);
433 /* Manually assign socket ownership as we avoid calling
434 * sock_alloc_send_pskb() to bypass wmem buffer limits
436 skb_set_owner_w(skb, sk);
441 static void ip6_nd_hdr(struct sk_buff *skb,
442 const struct in6_addr *saddr,
443 const struct in6_addr *daddr,
444 int hop_limit, int len)
447 struct inet6_dev *idev;
451 idev = __in6_dev_get(skb->dev);
452 tclass = idev ? idev->cnf.ndisc_tclass : 0;
455 skb_push(skb, sizeof(*hdr));
456 skb_reset_network_header(skb);
459 ip6_flow_hdr(hdr, tclass, 0);
461 hdr->payload_len = htons(len);
462 hdr->nexthdr = IPPROTO_ICMPV6;
463 hdr->hop_limit = hop_limit;
469 static void ndisc_send_skb(struct sk_buff *skb,
470 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);
519 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
520 const struct in6_addr *solicited_addr,
521 bool router, bool solicited, bool override, bool inc_opt)
524 struct in6_addr tmpaddr;
525 struct inet6_ifaddr *ifp;
526 const struct in6_addr *src_addr;
530 /* for anycast or proxy, solicited_addr != src_addr */
531 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
533 src_addr = solicited_addr;
534 if (ifp->flags & IFA_F_OPTIMISTIC)
536 inc_opt |= ifp->idev->cnf.force_tllao;
539 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
540 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
549 optlen += ndisc_opt_addr_space(dev,
550 NDISC_NEIGHBOUR_ADVERTISEMENT);
552 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
556 msg = skb_put(skb, sizeof(*msg));
557 *msg = (struct nd_msg) {
559 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
560 .icmp6_router = router,
561 .icmp6_solicited = solicited,
562 .icmp6_override = override,
564 .target = *solicited_addr,
568 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
570 NDISC_NEIGHBOUR_ADVERTISEMENT);
572 ndisc_send_skb(skb, daddr, src_addr);
575 static void ndisc_send_unsol_na(struct net_device *dev)
577 struct inet6_dev *idev;
578 struct inet6_ifaddr *ifa;
580 idev = in6_dev_get(dev);
584 read_lock_bh(&idev->lock);
585 list_for_each_entry(ifa, &idev->addr_list, if_list) {
586 /* skip tentative addresses until dad completes */
587 if (ifa->flags & IFA_F_TENTATIVE &&
588 !(ifa->flags & IFA_F_OPTIMISTIC))
591 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
592 /*router=*/ !!idev->cnf.forwarding,
593 /*solicited=*/ false, /*override=*/ true,
596 read_unlock_bh(&idev->lock);
601 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
602 const struct in6_addr *daddr, const struct in6_addr *saddr,
606 struct in6_addr addr_buf;
607 int inc_opt = dev->addr_len;
612 if (ipv6_get_lladdr(dev, &addr_buf,
613 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
618 if (ipv6_addr_any(saddr))
621 optlen += ndisc_opt_addr_space(dev,
622 NDISC_NEIGHBOUR_SOLICITATION);
626 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
630 msg = skb_put(skb, sizeof(*msg));
631 *msg = (struct nd_msg) {
633 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
639 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
641 NDISC_NEIGHBOUR_SOLICITATION);
643 u8 *opt = skb_put(skb, 8);
645 opt[0] = ND_OPT_NONCE;
647 memcpy(opt + 2, &nonce, 6);
650 ndisc_send_skb(skb, daddr, saddr);
653 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
654 const struct in6_addr *daddr)
658 int send_sllao = dev->addr_len;
661 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
663 * According to section 2.2 of RFC 4429, we must not
664 * send router solicitations with a sllao from
665 * optimistic addresses, but we may send the solicitation
666 * if we don't include the sllao. So here we check
667 * if our address is optimistic, and if so, we
668 * suppress the inclusion of the sllao.
671 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
674 if (ifp->flags & IFA_F_OPTIMISTIC) {
684 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
686 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
690 msg = skb_put(skb, sizeof(*msg));
691 *msg = (struct rs_msg) {
693 .icmp6_type = NDISC_ROUTER_SOLICITATION,
698 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
700 NDISC_ROUTER_SOLICITATION);
702 ndisc_send_skb(skb, daddr, saddr);
706 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
709 * "The sender MUST return an ICMP
710 * destination unreachable"
712 dst_link_failure(skb);
716 /* Called with locked neigh: either read or both */
718 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
720 struct in6_addr *saddr = NULL;
721 struct in6_addr mcaddr;
722 struct net_device *dev = neigh->dev;
723 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
724 int probes = atomic_read(&neigh->probes);
726 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
728 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
729 saddr = &ipv6_hdr(skb)->saddr;
730 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
732 if (!(neigh->nud_state & NUD_VALID)) {
734 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
737 ndisc_send_ns(dev, target, target, saddr, 0);
738 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
741 addrconf_addr_solict_mult(target, &mcaddr);
742 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
746 static int pndisc_is_router(const void *pkey,
747 struct net_device *dev)
749 struct pneigh_entry *n;
752 read_lock_bh(&nd_tbl.lock);
753 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
755 ret = !!(n->flags & NTF_ROUTER);
756 read_unlock_bh(&nd_tbl.lock);
761 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
762 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
763 struct ndisc_options *ndopts)
765 neigh_update(neigh, lladdr, new, flags, 0);
766 /* report ndisc ops about neighbour update */
767 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
770 static void ndisc_recv_ns(struct sk_buff *skb)
772 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
773 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
774 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
776 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
777 offsetof(struct nd_msg, opt));
778 struct ndisc_options ndopts;
779 struct net_device *dev = skb->dev;
780 struct inet6_ifaddr *ifp;
781 struct inet6_dev *idev = NULL;
782 struct neighbour *neigh;
783 int dad = ipv6_addr_any(saddr);
788 if (skb->len < sizeof(struct nd_msg)) {
789 ND_PRINTK(2, warn, "NS: packet too short\n");
793 if (ipv6_addr_is_multicast(&msg->target)) {
794 ND_PRINTK(2, warn, "NS: multicast target address\n");
800 * DAD has to be destined for solicited node multicast address.
802 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
803 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
807 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
808 ND_PRINTK(2, warn, "NS: invalid ND options\n");
812 if (ndopts.nd_opts_src_lladdr) {
813 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
816 "NS: invalid link-layer address length\n");
821 * If the IP source address is the unspecified address,
822 * there MUST NOT be source link-layer address option
827 "NS: bad DAD packet (link-layer address option)\n");
831 if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
832 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
834 inc = ipv6_addr_is_multicast(daddr);
836 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
839 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
841 if (nonce != 0 && ifp->dad_nonce == nonce) {
842 u8 *np = (u8 *)&nonce;
843 /* Matching nonce if looped back */
845 "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
846 ifp->idev->dev->name,
851 * We are colliding with another node
853 * so fail our DAD process
855 addrconf_dad_failure(skb, ifp);
859 * This is not a dad solicitation.
860 * If we are an optimistic node,
862 * Otherwise, we should ignore it.
864 if (!(ifp->flags & IFA_F_OPTIMISTIC))
871 struct net *net = dev_net(dev);
873 /* perhaps an address on the master device */
874 if (netif_is_l3_slave(dev)) {
875 struct net_device *mdev;
877 mdev = netdev_master_upper_dev_get_rcu(dev);
879 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
885 idev = in6_dev_get(dev);
887 /* XXX: count this drop? */
891 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
892 (idev->cnf.forwarding &&
893 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
894 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
895 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
896 skb->pkt_type != PACKET_HOST &&
898 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
900 * for anycast or proxy,
901 * sender should delay its response
902 * by a random time between 0 and
903 * MAX_ANYCAST_DELAY_TIME seconds.
904 * (RFC2461) -- yoshfuji
906 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
908 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
916 is_router = idev->cnf.forwarding;
919 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
920 !!is_router, false, (ifp != NULL), true);
925 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
927 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
930 * update / create cache entry
931 * for the source address
933 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
934 !inc || lladdr || !dev->addr_len);
936 ndisc_update(dev, neigh, lladdr, NUD_STALE,
937 NEIGH_UPDATE_F_WEAK_OVERRIDE|
938 NEIGH_UPDATE_F_OVERRIDE,
939 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
940 if (neigh || !dev->header_ops) {
941 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
942 true, (ifp != NULL && inc), inc);
944 neigh_release(neigh);
954 static void ndisc_recv_na(struct sk_buff *skb)
956 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
957 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
958 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
960 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
961 offsetof(struct nd_msg, opt));
962 struct ndisc_options ndopts;
963 struct net_device *dev = skb->dev;
964 struct inet6_dev *idev = __in6_dev_get(dev);
965 struct inet6_ifaddr *ifp;
966 struct neighbour *neigh;
968 if (skb->len < sizeof(struct nd_msg)) {
969 ND_PRINTK(2, warn, "NA: packet too short\n");
973 if (ipv6_addr_is_multicast(&msg->target)) {
974 ND_PRINTK(2, warn, "NA: target address is multicast\n");
978 if (ipv6_addr_is_multicast(daddr) &&
979 msg->icmph.icmp6_solicited) {
980 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
984 /* For some 802.11 wireless deployments (and possibly other networks),
985 * there will be a NA proxy and unsolicitd packets are attacks
986 * and thus should not be accepted.
988 if (!msg->icmph.icmp6_solicited && idev &&
989 idev->cnf.drop_unsolicited_na)
992 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
993 ND_PRINTK(2, warn, "NS: invalid ND option\n");
996 if (ndopts.nd_opts_tgt_lladdr) {
997 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1000 "NA: invalid link-layer address length\n");
1004 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1006 if (skb->pkt_type != PACKET_LOOPBACK
1007 && (ifp->flags & IFA_F_TENTATIVE)) {
1008 addrconf_dad_failure(skb, ifp);
1011 /* What should we make now? The advertisement
1012 is invalid, but ndisc specs say nothing
1013 about it. It could be misconfiguration, or
1014 an smart proxy agent tries to help us :-)
1016 We should not print the error if NA has been
1017 received from loopback - it is just our own
1018 unsolicited advertisement.
1020 if (skb->pkt_type != PACKET_LOOPBACK)
1022 "NA: %pM advertised our address %pI6c on %s!\n",
1023 eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1027 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1030 u8 old_flags = neigh->flags;
1031 struct net *net = dev_net(dev);
1033 if (neigh->nud_state & NUD_FAILED)
1037 * Don't update the neighbor cache entry on a proxy NA from
1038 * ourselves because either the proxied node is off link or it
1039 * has already sent a NA to us.
1041 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1042 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1043 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1044 /* XXX: idev->cnf.proxy_ndp */
1048 ndisc_update(dev, neigh, lladdr,
1049 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1050 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1051 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1052 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1053 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1054 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1056 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1058 * Change: router to host
1060 rt6_clean_tohost(dev_net(dev), saddr);
1064 neigh_release(neigh);
1068 static void ndisc_recv_rs(struct sk_buff *skb)
1070 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1071 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1072 struct neighbour *neigh;
1073 struct inet6_dev *idev;
1074 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1075 struct ndisc_options ndopts;
1078 if (skb->len < sizeof(*rs_msg))
1081 idev = __in6_dev_get(skb->dev);
1083 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1087 /* Don't accept RS if we're not in router mode */
1088 if (!idev->cnf.forwarding)
1092 * Don't update NCE if src = ::;
1093 * this implies that the source node has no ip address assigned yet.
1095 if (ipv6_addr_any(saddr))
1098 /* Parse ND options */
1099 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1100 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1104 if (ndopts.nd_opts_src_lladdr) {
1105 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1111 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1113 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1114 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1115 NEIGH_UPDATE_F_OVERRIDE|
1116 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1117 NDISC_ROUTER_SOLICITATION, &ndopts);
1118 neigh_release(neigh);
1124 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1126 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1127 struct sk_buff *skb;
1128 struct nlmsghdr *nlh;
1129 struct nduseroptmsg *ndmsg;
1130 struct net *net = dev_net(ra->dev);
1132 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1133 + (opt->nd_opt_len << 3));
1134 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1136 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1142 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1144 goto nla_put_failure;
1147 ndmsg = nlmsg_data(nlh);
1148 ndmsg->nduseropt_family = AF_INET6;
1149 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1150 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1151 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1152 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1154 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1156 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1157 goto nla_put_failure;
1158 nlmsg_end(skb, nlh);
1160 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1167 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1170 static void ndisc_router_discovery(struct sk_buff *skb)
1172 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1173 struct neighbour *neigh = NULL;
1174 struct inet6_dev *in6_dev;
1175 struct fib6_info *rt = NULL;
1176 u32 defrtr_usr_metric;
1179 struct ndisc_options ndopts;
1181 unsigned int pref = 0;
1183 bool send_ifinfo_notify = false;
1185 __u8 *opt = (__u8 *)(ra_msg + 1);
1187 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1188 sizeof(struct ra_msg);
1191 "RA: %s, dev: %s\n",
1192 __func__, skb->dev->name);
1193 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1194 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1198 ND_PRINTK(2, warn, "RA: packet too short\n");
1202 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1203 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1204 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1210 * set the RA_RECV flag in the interface
1213 in6_dev = __in6_dev_get(skb->dev);
1215 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1220 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1221 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1225 if (!ipv6_accept_ra(in6_dev)) {
1227 "RA: %s, did not accept ra for dev: %s\n",
1228 __func__, skb->dev->name);
1229 goto skip_linkparms;
1232 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1233 /* skip link-specific parameters from interior routers */
1234 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1236 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1237 __func__, skb->dev->name);
1238 goto skip_linkparms;
1242 if (in6_dev->if_flags & IF_RS_SENT) {
1244 * flag that an RA was received after an RS was sent
1245 * out on this interface.
1247 in6_dev->if_flags |= IF_RA_RCVD;
1251 * Remember the managed/otherconf flags from most recently
1252 * received RA message (RFC 2462) -- yoshfuji
1254 old_if_flags = in6_dev->if_flags;
1255 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1257 (ra_msg->icmph.icmp6_addrconf_managed ?
1258 IF_RA_MANAGED : 0) |
1259 (ra_msg->icmph.icmp6_addrconf_other ?
1260 IF_RA_OTHERCONF : 0);
1262 if (old_if_flags != in6_dev->if_flags)
1263 send_ifinfo_notify = true;
1265 if (!in6_dev->cnf.accept_ra_defrtr) {
1267 "RA: %s, defrtr is false for dev: %s\n",
1268 __func__, skb->dev->name);
1272 /* Do not accept RA with source-addr found on local machine unless
1273 * accept_ra_from_local is set to true.
1275 net = dev_net(in6_dev->dev);
1276 if (!in6_dev->cnf.accept_ra_from_local &&
1277 ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1279 "RA from local address detected on dev: %s: default router ignored\n",
1284 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1286 #ifdef CONFIG_IPV6_ROUTER_PREF
1287 pref = ra_msg->icmph.icmp6_router_pref;
1288 /* 10b is handled as if it were 00b (medium) */
1289 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1290 !in6_dev->cnf.accept_ra_rtr_pref)
1291 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1293 /* routes added from RAs do not use nexthop objects */
1294 rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1296 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1297 rt->fib6_nh->fib_nh_dev, NULL,
1298 &ipv6_hdr(skb)->saddr);
1301 "RA: %s got default router without neighbour\n",
1303 fib6_info_release(rt);
1307 /* Set default route metric as specified by user */
1308 defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1309 /* delete the route if lifetime is 0 or if metric needs change */
1310 if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1311 ip6_del_rt(net, rt, false);
1315 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, metric: %d, for dev: %s\n",
1316 rt, lifetime, defrtr_usr_metric, skb->dev->name);
1317 if (!rt && lifetime) {
1318 ND_PRINTK(3, info, "RA: adding default router\n");
1320 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1321 skb->dev, pref, defrtr_usr_metric);
1324 "RA: %s failed to add default route\n",
1329 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1330 rt->fib6_nh->fib_nh_dev, NULL,
1331 &ipv6_hdr(skb)->saddr);
1334 "RA: %s got default router without neighbour\n",
1336 fib6_info_release(rt);
1339 neigh->flags |= NTF_ROUTER;
1341 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1345 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1346 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1347 ra_msg->icmph.icmp6_hop_limit) {
1348 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1349 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1350 fib6_metric_set(rt, RTAX_HOPLIMIT,
1351 ra_msg->icmph.icmp6_hop_limit);
1353 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1360 * Update Reachable Time and Retrans Timer
1363 if (in6_dev->nd_parms) {
1364 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1366 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1367 rtime = (rtime*HZ)/1000;
1370 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1371 in6_dev->tstamp = jiffies;
1372 send_ifinfo_notify = true;
1375 rtime = ntohl(ra_msg->reachable_time);
1376 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1377 rtime = (rtime*HZ)/1000;
1382 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1383 NEIGH_VAR_SET(in6_dev->nd_parms,
1384 BASE_REACHABLE_TIME, rtime);
1385 NEIGH_VAR_SET(in6_dev->nd_parms,
1386 GC_STALETIME, 3 * rtime);
1387 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1388 in6_dev->tstamp = jiffies;
1389 send_ifinfo_notify = true;
1395 * Send a notify if RA changed managed/otherconf flags or timer settings
1397 if (send_ifinfo_notify)
1398 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1407 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1411 if (ndopts.nd_opts_src_lladdr) {
1412 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1416 "RA: invalid link-layer address length\n");
1420 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1421 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1422 NEIGH_UPDATE_F_OVERRIDE|
1423 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1424 NEIGH_UPDATE_F_ISROUTER,
1425 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1428 if (!ipv6_accept_ra(in6_dev)) {
1430 "RA: %s, accept_ra is false for dev: %s\n",
1431 __func__, skb->dev->name);
1435 #ifdef CONFIG_IPV6_ROUTE_INFO
1436 if (!in6_dev->cnf.accept_ra_from_local &&
1437 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1440 "RA from local address detected on dev: %s: router info ignored.\n",
1442 goto skip_routeinfo;
1445 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1446 struct nd_opt_hdr *p;
1447 for (p = ndopts.nd_opts_ri;
1449 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1450 struct route_info *ri = (struct route_info *)p;
1451 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1452 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1453 ri->prefix_len == 0)
1456 if (ri->prefix_len == 0 &&
1457 !in6_dev->cnf.accept_ra_defrtr)
1459 if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1461 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1463 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1464 &ipv6_hdr(skb)->saddr);
1471 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1472 /* skip link-specific ndopts from interior routers */
1473 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1475 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1476 __func__, skb->dev->name);
1481 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1482 struct nd_opt_hdr *p;
1483 for (p = ndopts.nd_opts_pi;
1485 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1486 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1487 (p->nd_opt_len) << 3,
1488 ndopts.nd_opts_src_lladdr != NULL);
1492 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1496 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1499 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1500 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1501 } else if (in6_dev->cnf.mtu6 != mtu) {
1502 in6_dev->cnf.mtu6 = mtu;
1503 fib6_metric_set(rt, RTAX_MTU, mtu);
1504 rt6_mtu_change(skb->dev, mtu);
1508 if (ndopts.nd_useropts) {
1509 struct nd_opt_hdr *p;
1510 for (p = ndopts.nd_useropts;
1512 p = ndisc_next_useropt(skb->dev, p,
1513 ndopts.nd_useropts_end)) {
1514 ndisc_ra_useropt(skb, p);
1518 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1519 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1522 fib6_info_release(rt);
1524 neigh_release(neigh);
1527 static void ndisc_redirect_rcv(struct sk_buff *skb)
1530 struct ndisc_options ndopts;
1531 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1532 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1533 offsetof(struct rd_msg, opt));
1535 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1536 switch (skb->ndisc_nodetype) {
1537 case NDISC_NODETYPE_HOST:
1538 case NDISC_NODETYPE_NODEFAULT:
1540 "Redirect: from host or unauthorized router\n");
1545 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1547 "Redirect: source address is not link-local\n");
1551 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1554 if (!ndopts.nd_opts_rh) {
1555 ip6_redirect_no_header(skb, dev_net(skb->dev),
1560 hdr = (u8 *)ndopts.nd_opts_rh;
1562 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1565 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1568 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1569 struct sk_buff *orig_skb,
1572 u8 *opt = skb_put(skb, rd_len);
1575 *(opt++) = ND_OPT_REDIRECT_HDR;
1576 *(opt++) = (rd_len >> 3);
1579 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1583 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1585 struct net_device *dev = skb->dev;
1586 struct net *net = dev_net(dev);
1587 struct sock *sk = net->ipv6.ndisc_sk;
1589 struct inet_peer *peer;
1590 struct sk_buff *buff;
1592 struct in6_addr saddr_buf;
1593 struct rt6_info *rt;
1594 struct dst_entry *dst;
1597 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1598 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1601 if (netif_is_l3_master(skb->dev)) {
1602 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1607 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1608 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1613 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1614 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1616 "Redirect: target address is not link-local unicast\n");
1620 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1621 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1623 dst = ip6_route_output(net, NULL, &fl6);
1628 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1632 rt = (struct rt6_info *) dst;
1634 if (rt->rt6i_flags & RTF_GATEWAY) {
1636 "Redirect: destination is not a neighbour\n");
1639 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1640 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1646 if (dev->addr_len) {
1647 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1650 "Redirect: no neigh for target address\n");
1654 read_lock_bh(&neigh->lock);
1655 if (neigh->nud_state & NUD_VALID) {
1656 memcpy(ha_buf, neigh->ha, dev->addr_len);
1657 read_unlock_bh(&neigh->lock);
1659 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1663 read_unlock_bh(&neigh->lock);
1665 neigh_release(neigh);
1668 rd_len = min_t(unsigned int,
1669 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1674 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1678 msg = skb_put(buff, sizeof(*msg));
1679 *msg = (struct rd_msg) {
1681 .icmp6_type = NDISC_REDIRECT,
1684 .dest = ipv6_hdr(skb)->daddr,
1688 * include target_address option
1692 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1695 * build redirect option and copy skb over to the new packet.
1699 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1701 skb_dst_set(buff, dst);
1702 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1709 static void pndisc_redo(struct sk_buff *skb)
1715 static int ndisc_is_multicast(const void *pkey)
1717 return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1720 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1722 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1726 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1727 idev->cnf.suppress_frag_ndisc) {
1728 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1734 int ndisc_rcv(struct sk_buff *skb)
1738 if (ndisc_suppress_frag_ndisc(skb))
1741 if (skb_linearize(skb))
1744 msg = (struct nd_msg *)skb_transport_header(skb);
1746 __skb_push(skb, skb->data - skb_transport_header(skb));
1748 if (ipv6_hdr(skb)->hop_limit != 255) {
1749 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1750 ipv6_hdr(skb)->hop_limit);
1754 if (msg->icmph.icmp6_code != 0) {
1755 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1756 msg->icmph.icmp6_code);
1760 switch (msg->icmph.icmp6_type) {
1761 case NDISC_NEIGHBOUR_SOLICITATION:
1762 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1766 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1770 case NDISC_ROUTER_SOLICITATION:
1774 case NDISC_ROUTER_ADVERTISEMENT:
1775 ndisc_router_discovery(skb);
1778 case NDISC_REDIRECT:
1779 ndisc_redirect_rcv(skb);
1786 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1788 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1789 struct netdev_notifier_change_info *change_info;
1790 struct net *net = dev_net(dev);
1791 struct inet6_dev *idev;
1794 case NETDEV_CHANGEADDR:
1795 neigh_changeaddr(&nd_tbl, dev);
1796 fib6_run_gc(0, net, false);
1799 idev = in6_dev_get(dev);
1802 if (idev->cnf.ndisc_notify ||
1803 net->ipv6.devconf_all->ndisc_notify)
1804 ndisc_send_unsol_na(dev);
1809 if (change_info->flags_changed & IFF_NOARP)
1810 neigh_changeaddr(&nd_tbl, dev);
1811 if (!netif_carrier_ok(dev))
1812 neigh_carrier_down(&nd_tbl, dev);
1815 neigh_ifdown(&nd_tbl, dev);
1816 fib6_run_gc(0, net, false);
1818 case NETDEV_NOTIFY_PEERS:
1819 ndisc_send_unsol_na(dev);
1828 static struct notifier_block ndisc_netdev_notifier = {
1829 .notifier_call = ndisc_netdev_event,
1830 .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1833 #ifdef CONFIG_SYSCTL
1834 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1835 const char *func, const char *dev_name)
1837 static char warncomm[TASK_COMM_LEN];
1839 if (strcmp(warncomm, current->comm) && warned < 5) {
1840 strcpy(warncomm, current->comm);
1841 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1843 dev_name, ctl->procname,
1844 dev_name, ctl->procname);
1849 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1850 size_t *lenp, loff_t *ppos)
1852 struct net_device *dev = ctl->extra1;
1853 struct inet6_dev *idev;
1856 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1857 (strcmp(ctl->procname, "base_reachable_time") == 0))
1858 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1860 if (strcmp(ctl->procname, "retrans_time") == 0)
1861 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1863 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1864 ret = neigh_proc_dointvec_jiffies(ctl, write,
1865 buffer, lenp, ppos);
1867 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1868 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1869 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1870 buffer, lenp, ppos);
1874 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1875 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1876 idev->nd_parms->reachable_time =
1877 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1878 idev->tstamp = jiffies;
1879 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1888 static int __net_init ndisc_net_init(struct net *net)
1890 struct ipv6_pinfo *np;
1894 err = inet_ctl_sock_create(&sk, PF_INET6,
1895 SOCK_RAW, IPPROTO_ICMPV6, net);
1898 "NDISC: Failed to initialize the control socket (err %d)\n",
1903 net->ipv6.ndisc_sk = sk;
1906 np->hop_limit = 255;
1907 /* Do not loopback ndisc messages */
1913 static void __net_exit ndisc_net_exit(struct net *net)
1915 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1918 static struct pernet_operations ndisc_net_ops = {
1919 .init = ndisc_net_init,
1920 .exit = ndisc_net_exit,
1923 int __init ndisc_init(void)
1927 err = register_pernet_subsys(&ndisc_net_ops);
1931 * Initialize the neighbour table
1933 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1935 #ifdef CONFIG_SYSCTL
1936 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1937 ndisc_ifinfo_sysctl_change);
1939 goto out_unregister_pernet;
1944 #ifdef CONFIG_SYSCTL
1945 out_unregister_pernet:
1946 unregister_pernet_subsys(&ndisc_net_ops);
1951 int __init ndisc_late_init(void)
1953 return register_netdevice_notifier(&ndisc_netdev_notifier);
1956 void ndisc_late_cleanup(void)
1958 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1961 void ndisc_cleanup(void)
1963 #ifdef CONFIG_SYSCTL
1964 neigh_sysctl_unregister(&nd_tbl.parms);
1966 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1967 unregister_pernet_subsys(&ndisc_net_ops);