Merge tag 'iommu-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
[linux/fpc-iii.git] / net / ipv6 / ndisc.c
blob76717478f1733fb753c067efaa5dd210320e0261
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Neighbour Discovery for IPv6
4 * Linux INET6 implementation
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Mike Shaver <shaver@ingenia.com>
9 */
12 * Changes:
14 * Alexey I. Froloff : RFC6106 (DNSSL) support
15 * Pierre Ynard : export userland ND options
16 * through netlink (RDNSS support)
17 * Lars Fenneberg : fixed MTU setting on receipt
18 * of an RA.
19 * Janos Farkas : kmalloc failure checks
20 * Alexey Kuznetsov : state machine reworked
21 * and moved to net/core.
22 * Pekka Savola : RFC2461 validation
23 * YOSHIFUJI Hideaki @USAGI : Verify ND options properly
26 #define pr_fmt(fmt) "ICMPv6: " fmt
28 #include <linux/module.h>
29 #include <linux/errno.h>
30 #include <linux/types.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/sched.h>
34 #include <linux/net.h>
35 #include <linux/in6.h>
36 #include <linux/route.h>
37 #include <linux/init.h>
38 #include <linux/rcupdate.h>
39 #include <linux/slab.h>
40 #ifdef CONFIG_SYSCTL
41 #include <linux/sysctl.h>
42 #endif
44 #include <linux/if_addr.h>
45 #include <linux/if_ether.h>
46 #include <linux/if_arp.h>
47 #include <linux/ipv6.h>
48 #include <linux/icmpv6.h>
49 #include <linux/jhash.h>
51 #include <net/sock.h>
52 #include <net/snmp.h>
54 #include <net/ipv6.h>
55 #include <net/protocol.h>
56 #include <net/ndisc.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #include <net/icmp.h>
61 #include <net/netlink.h>
62 #include <linux/rtnetlink.h>
64 #include <net/flow.h>
65 #include <net/ip6_checksum.h>
66 #include <net/inet_common.h>
67 #include <linux/proc_fs.h>
69 #include <linux/netfilter.h>
70 #include <linux/netfilter_ipv6.h>
72 static u32 ndisc_hash(const void *pkey,
73 const struct net_device *dev,
74 __u32 *hash_rnd);
75 static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
76 static bool ndisc_allow_add(const struct net_device *dev,
77 struct netlink_ext_ack *extack);
78 static int ndisc_constructor(struct neighbour *neigh);
79 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
80 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
81 static int pndisc_constructor(struct pneigh_entry *n);
82 static void pndisc_destructor(struct pneigh_entry *n);
83 static void pndisc_redo(struct sk_buff *skb);
84 static int ndisc_is_multicast(const void *pkey);
86 static const struct neigh_ops ndisc_generic_ops = {
87 .family = AF_INET6,
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 = {
95 .family = AF_INET6,
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 = {
104 .family = AF_INET6,
105 .output = neigh_direct_output,
106 .connected_output = neigh_direct_output,
109 struct neigh_table nd_tbl = {
110 .family = AF_INET6,
111 .key_len = sizeof(struct in6_addr),
112 .protocol = cpu_to_be16(ETH_P_IPV6),
113 .hash = ndisc_hash,
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,
121 .id = "ndisc_cache",
122 .parms = {
123 .tbl = &nd_tbl,
124 .reachable_time = ND_REACHABLE_TIME,
125 .data = {
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,
139 .gc_thresh1 = 128,
140 .gc_thresh2 = 512,
141 .gc_thresh3 = 1024,
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);
151 opt[0] = type;
152 opt[1] = space>>3;
154 memset(opt + 2, 0, pad);
155 opt += pad;
156 space -= pad;
158 memcpy(opt+2, data, data_len);
159 data_len += 2;
160 opt += data_len;
161 space -= data_len;
162 if (space > 0)
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,
176 void *ha,
177 const u8 *ops_data)
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)
186 int type;
187 if (!cur || !end || cur >= end)
188 return NULL;
189 type = cur->nd_opt_type;
190 do {
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)
211 return NULL;
212 do {
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)
225 return NULL;
226 memset(ndopts, 0, sizeof(*ndopts));
227 while (opt_len) {
228 int l;
229 if (opt_len < sizeof(struct nd_opt_hdr))
230 return NULL;
231 l = nd_opt->nd_opt_len << 3;
232 if (opt_len < l || l == 0)
233 return NULL;
234 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
235 goto next_opt;
236 switch (nd_opt->nd_opt_type) {
237 case ND_OPT_SOURCE_LL_ADDR:
238 case ND_OPT_TARGET_LL_ADDR:
239 case ND_OPT_MTU:
240 case ND_OPT_NONCE:
241 case ND_OPT_REDIRECT_HDR:
242 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
243 ND_PRINTK(2, warn,
244 "%s: duplicated ND6 option found: type=%d\n",
245 __func__, nd_opt->nd_opt_type);
246 } else {
247 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
249 break;
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;
254 break;
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;
260 break;
261 #endif
262 default:
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;
267 } else {
269 * Unknown options must be silently ignored,
270 * to accommodate future extension to the
271 * protocol.
273 ND_PRINTK(2, notice,
274 "%s: ignored unsupported option; type=%d, len=%d\n",
275 __func__,
276 nd_opt->nd_opt_type,
277 nd_opt->nd_opt_len);
280 next_opt:
281 opt_len -= l;
282 nd_opt = ((void *)nd_opt) + l;
284 return ndopts;
287 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
289 switch (dev->type) {
290 case ARPHRD_ETHER:
291 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
292 case ARPHRD_FDDI:
293 ipv6_eth_mc_map(addr, buf);
294 return 0;
295 case ARPHRD_ARCNET:
296 ipv6_arcnet_mc_map(addr, buf);
297 return 0;
298 case ARPHRD_INFINIBAND:
299 ipv6_ib_mc_map(addr, dev->broadcast, buf);
300 return 0;
301 case ARPHRD_IPGRE:
302 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
303 default:
304 if (dir) {
305 memcpy(buf, dev->broadcast, dev->addr_len);
306 return 0;
309 return -EINVAL;
311 EXPORT_SYMBOL(ndisc_mc_map);
313 static u32 ndisc_hash(const void *pkey,
314 const struct net_device *dev,
315 __u32 *hash_rnd)
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);
334 if (!in6_dev) {
335 return -EINVAL;
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;
347 } else {
348 if (is_multicast) {
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;
362 else
363 neigh->ops = &ndisc_generic_ops;
364 if (neigh->nud_state&NUD_VALID)
365 neigh->output = neigh->ops->connected_output;
366 else
367 neigh->output = neigh->ops->output;
369 in6_dev_put(in6_dev);
370 return 0;
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))
380 return -EINVAL;
381 addrconf_addr_solict_mult(addr, &maddr);
382 ipv6_dev_mc_inc(dev, &maddr);
383 return 0;
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))
393 return;
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");
406 return false;
409 return true;
412 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
413 int len)
415 int hlen = LL_RESERVED_SPACE(dev);
416 int tlen = dev->needed_tailroom;
417 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
418 struct sk_buff *skb;
420 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
421 if (!skb) {
422 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
423 __func__);
424 return NULL;
427 skb->protocol = htons(ETH_P_IPV6);
428 skb->dev = dev;
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);
438 return skb;
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)
446 struct ipv6hdr *hdr;
447 struct inet6_dev *idev;
448 unsigned tclass;
450 rcu_read_lock();
451 idev = __in6_dev_get(skb->dev);
452 tclass = idev ? idev->cnf.ndisc_tclass : 0;
453 rcu_read_unlock();
455 skb_push(skb, sizeof(*hdr));
456 skb_reset_network_header(skb);
457 hdr = ipv6_hdr(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;
465 hdr->saddr = *saddr;
466 hdr->daddr = *daddr;
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;
477 int err;
478 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
479 u8 type;
481 type = icmp6h->icmp6_type;
483 if (!dst) {
484 struct flowi6 fl6;
485 int oif = skb->dev->ifindex;
487 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
488 dst = icmp6_dst_alloc(skb->dev, &fl6);
489 if (IS_ERR(dst)) {
490 kfree_skb(skb);
491 return;
494 skb_dst_set(skb, dst);
497 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
498 IPPROTO_ICMPV6,
499 csum_partial(icmp6h,
500 skb->len, 0));
502 ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
504 rcu_read_lock();
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,
510 dst_output);
511 if (!err) {
512 ICMP6MSGOUT_INC_STATS(net, idev, type);
513 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
516 rcu_read_unlock();
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)
523 struct sk_buff *skb;
524 struct in6_addr tmpaddr;
525 struct inet6_ifaddr *ifp;
526 const struct in6_addr *src_addr;
527 struct nd_msg *msg;
528 int optlen = 0;
530 /* for anycast or proxy, solicited_addr != src_addr */
531 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
532 if (ifp) {
533 src_addr = solicited_addr;
534 if (ifp->flags & IFA_F_OPTIMISTIC)
535 override = false;
536 inc_opt |= ifp->idev->cnf.force_tllao;
537 in6_ifa_put(ifp);
538 } else {
539 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
540 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
541 &tmpaddr))
542 return;
543 src_addr = &tmpaddr;
546 if (!dev->addr_len)
547 inc_opt = false;
548 if (inc_opt)
549 optlen += ndisc_opt_addr_space(dev,
550 NDISC_NEIGHBOUR_ADVERTISEMENT);
552 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
553 if (!skb)
554 return;
556 msg = skb_put(skb, sizeof(*msg));
557 *msg = (struct nd_msg) {
558 .icmph = {
559 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
560 .icmp6_router = router,
561 .icmp6_solicited = solicited,
562 .icmp6_override = override,
564 .target = *solicited_addr,
567 if (inc_opt)
568 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
569 dev->dev_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);
581 if (!idev)
582 return;
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))
589 continue;
591 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
592 /*router=*/ !!idev->cnf.forwarding,
593 /*solicited=*/ false, /*override=*/ true,
594 /*inc_opt=*/ true);
596 read_unlock_bh(&idev->lock);
598 in6_dev_put(idev);
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,
603 u64 nonce)
605 struct sk_buff *skb;
606 struct in6_addr addr_buf;
607 int inc_opt = dev->addr_len;
608 int optlen = 0;
609 struct nd_msg *msg;
611 if (!saddr) {
612 if (ipv6_get_lladdr(dev, &addr_buf,
613 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
614 return;
615 saddr = &addr_buf;
618 if (ipv6_addr_any(saddr))
619 inc_opt = false;
620 if (inc_opt)
621 optlen += ndisc_opt_addr_space(dev,
622 NDISC_NEIGHBOUR_SOLICITATION);
623 if (nonce != 0)
624 optlen += 8;
626 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
627 if (!skb)
628 return;
630 msg = skb_put(skb, sizeof(*msg));
631 *msg = (struct nd_msg) {
632 .icmph = {
633 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
635 .target = *solicit,
638 if (inc_opt)
639 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
640 dev->dev_addr,
641 NDISC_NEIGHBOUR_SOLICITATION);
642 if (nonce != 0) {
643 u8 *opt = skb_put(skb, 8);
645 opt[0] = ND_OPT_NONCE;
646 opt[1] = 8 >> 3;
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)
656 struct sk_buff *skb;
657 struct rs_msg *msg;
658 int send_sllao = dev->addr_len;
659 int optlen = 0;
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.
670 if (send_sllao) {
671 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
672 dev, 1);
673 if (ifp) {
674 if (ifp->flags & IFA_F_OPTIMISTIC) {
675 send_sllao = 0;
677 in6_ifa_put(ifp);
678 } else {
679 send_sllao = 0;
682 #endif
683 if (send_sllao)
684 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
686 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
687 if (!skb)
688 return;
690 msg = skb_put(skb, sizeof(*msg));
691 *msg = (struct rs_msg) {
692 .icmph = {
693 .icmp6_type = NDISC_ROUTER_SOLICITATION,
697 if (send_sllao)
698 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
699 dev->dev_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);
713 kfree_skb(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,
727 dev, false, 1,
728 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
729 saddr = &ipv6_hdr(skb)->saddr;
730 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
731 if (probes < 0) {
732 if (!(neigh->nud_state & NUD_VALID)) {
733 ND_PRINTK(1, dbg,
734 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
735 __func__, target);
737 ndisc_send_ns(dev, target, target, saddr, 0);
738 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
739 neigh_app_ns(neigh);
740 } else {
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;
750 int ret = -1;
752 read_lock_bh(&nd_tbl.lock);
753 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
754 if (n)
755 ret = !!(n->flags & NTF_ROUTER);
756 read_unlock_bh(&nd_tbl.lock);
758 return ret;
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;
775 u8 *lladdr = NULL;
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);
784 bool inc;
785 int is_router = -1;
786 u64 nonce = 0;
788 if (skb->len < sizeof(struct nd_msg)) {
789 ND_PRINTK(2, warn, "NS: packet too short\n");
790 return;
793 if (ipv6_addr_is_multicast(&msg->target)) {
794 ND_PRINTK(2, warn, "NS: multicast target address\n");
795 return;
799 * RFC2461 7.1.1:
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");
804 return;
807 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
808 ND_PRINTK(2, warn, "NS: invalid ND options\n");
809 return;
812 if (ndopts.nd_opts_src_lladdr) {
813 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
814 if (!lladdr) {
815 ND_PRINTK(2, warn,
816 "NS: invalid link-layer address length\n");
817 return;
820 /* RFC2461 7.1.1:
821 * If the IP source address is the unspecified address,
822 * there MUST NOT be source link-layer address option
823 * in the message.
825 if (dad) {
826 ND_PRINTK(2, warn,
827 "NS: bad DAD packet (link-layer address option)\n");
828 return;
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);
837 if (ifp) {
838 have_ifp:
839 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
840 if (dad) {
841 if (nonce != 0 && ifp->dad_nonce == nonce) {
842 u8 *np = (u8 *)&nonce;
843 /* Matching nonce if looped back */
844 ND_PRINTK(2, notice,
845 "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
846 ifp->idev->dev->name,
847 &ifp->addr, np);
848 goto out;
851 * We are colliding with another node
852 * who is doing DAD
853 * so fail our DAD process
855 addrconf_dad_failure(skb, ifp);
856 return;
857 } else {
859 * This is not a dad solicitation.
860 * If we are an optimistic node,
861 * we should respond.
862 * Otherwise, we should ignore it.
864 if (!(ifp->flags & IFA_F_OPTIMISTIC))
865 goto out;
869 idev = ifp->idev;
870 } else {
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);
878 if (mdev) {
879 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
880 if (ifp)
881 goto have_ifp;
885 idev = in6_dev_get(dev);
886 if (!idev) {
887 /* XXX: count this drop? */
888 return;
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 &&
897 inc &&
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);
907 if (n)
908 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
909 goto out;
911 } else
912 goto out;
915 if (is_router < 0)
916 is_router = idev->cnf.forwarding;
918 if (dad) {
919 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
920 !!is_router, false, (ifp != NULL), true);
921 goto out;
924 if (inc)
925 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
926 else
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);
935 if (neigh)
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);
943 if (neigh)
944 neigh_release(neigh);
947 out:
948 if (ifp)
949 in6_ifa_put(ifp);
950 else
951 in6_dev_put(idev);
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;
959 u8 *lladdr = NULL;
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");
970 return;
973 if (ipv6_addr_is_multicast(&msg->target)) {
974 ND_PRINTK(2, warn, "NA: target address is multicast\n");
975 return;
978 if (ipv6_addr_is_multicast(daddr) &&
979 msg->icmph.icmp6_solicited) {
980 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
981 return;
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)
990 return;
992 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
993 ND_PRINTK(2, warn, "NS: invalid ND option\n");
994 return;
996 if (ndopts.nd_opts_tgt_lladdr) {
997 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
998 if (!lladdr) {
999 ND_PRINTK(2, warn,
1000 "NA: invalid link-layer address length\n");
1001 return;
1004 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1005 if (ifp) {
1006 if (skb->pkt_type != PACKET_LOOPBACK
1007 && (ifp->flags & IFA_F_TENTATIVE)) {
1008 addrconf_dad_failure(skb, ifp);
1009 return;
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)
1021 ND_PRINTK(1, warn,
1022 "NA: %pM advertised our address %pI6c on %s!\n",
1023 eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1024 in6_ifa_put(ifp);
1025 return;
1027 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1029 if (neigh) {
1030 u8 old_flags = neigh->flags;
1031 struct net *net = dev_net(dev);
1033 if (neigh->nud_state & NUD_FAILED)
1034 goto out;
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 */
1045 goto out;
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);
1063 out:
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;
1076 u8 *lladdr = NULL;
1078 if (skb->len < sizeof(*rs_msg))
1079 return;
1081 idev = __in6_dev_get(skb->dev);
1082 if (!idev) {
1083 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1084 return;
1087 /* Don't accept RS if we're not in router mode */
1088 if (!idev->cnf.forwarding)
1089 goto out;
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))
1096 goto out;
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");
1101 goto out;
1104 if (ndopts.nd_opts_src_lladdr) {
1105 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1106 skb->dev);
1107 if (!lladdr)
1108 goto out;
1111 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1112 if (neigh) {
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);
1120 out:
1121 return;
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);
1131 int err;
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);
1137 if (!skb) {
1138 err = -ENOBUFS;
1139 goto errout;
1142 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1143 if (!nlh) {
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);
1161 return;
1163 nla_put_failure:
1164 nlmsg_free(skb);
1165 err = -EMSGSIZE;
1166 errout:
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 struct net *net;
1177 int lifetime;
1178 struct ndisc_options ndopts;
1179 int optlen;
1180 unsigned int pref = 0;
1181 __u32 old_if_flags;
1182 bool send_ifinfo_notify = false;
1184 __u8 *opt = (__u8 *)(ra_msg + 1);
1186 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1187 sizeof(struct ra_msg);
1189 ND_PRINTK(2, info,
1190 "RA: %s, dev: %s\n",
1191 __func__, skb->dev->name);
1192 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1193 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1194 return;
1196 if (optlen < 0) {
1197 ND_PRINTK(2, warn, "RA: packet too short\n");
1198 return;
1201 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1202 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1203 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1204 return;
1206 #endif
1209 * set the RA_RECV flag in the interface
1212 in6_dev = __in6_dev_get(skb->dev);
1213 if (!in6_dev) {
1214 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1215 skb->dev->name);
1216 return;
1219 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1220 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1221 return;
1224 if (!ipv6_accept_ra(in6_dev)) {
1225 ND_PRINTK(2, info,
1226 "RA: %s, did not accept ra for dev: %s\n",
1227 __func__, skb->dev->name);
1228 goto skip_linkparms;
1231 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1232 /* skip link-specific parameters from interior routers */
1233 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1234 ND_PRINTK(2, info,
1235 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1236 __func__, skb->dev->name);
1237 goto skip_linkparms;
1239 #endif
1241 if (in6_dev->if_flags & IF_RS_SENT) {
1243 * flag that an RA was received after an RS was sent
1244 * out on this interface.
1246 in6_dev->if_flags |= IF_RA_RCVD;
1250 * Remember the managed/otherconf flags from most recently
1251 * received RA message (RFC 2462) -- yoshfuji
1253 old_if_flags = in6_dev->if_flags;
1254 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1255 IF_RA_OTHERCONF)) |
1256 (ra_msg->icmph.icmp6_addrconf_managed ?
1257 IF_RA_MANAGED : 0) |
1258 (ra_msg->icmph.icmp6_addrconf_other ?
1259 IF_RA_OTHERCONF : 0);
1261 if (old_if_flags != in6_dev->if_flags)
1262 send_ifinfo_notify = true;
1264 if (!in6_dev->cnf.accept_ra_defrtr) {
1265 ND_PRINTK(2, info,
1266 "RA: %s, defrtr is false for dev: %s\n",
1267 __func__, skb->dev->name);
1268 goto skip_defrtr;
1271 /* Do not accept RA with source-addr found on local machine unless
1272 * accept_ra_from_local is set to true.
1274 net = dev_net(in6_dev->dev);
1275 if (!in6_dev->cnf.accept_ra_from_local &&
1276 ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1277 ND_PRINTK(2, info,
1278 "RA from local address detected on dev: %s: default router ignored\n",
1279 skb->dev->name);
1280 goto skip_defrtr;
1283 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1285 #ifdef CONFIG_IPV6_ROUTER_PREF
1286 pref = ra_msg->icmph.icmp6_router_pref;
1287 /* 10b is handled as if it were 00b (medium) */
1288 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1289 !in6_dev->cnf.accept_ra_rtr_pref)
1290 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1291 #endif
1292 /* routes added from RAs do not use nexthop objects */
1293 rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1294 if (rt) {
1295 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1296 rt->fib6_nh->fib_nh_dev, NULL,
1297 &ipv6_hdr(skb)->saddr);
1298 if (!neigh) {
1299 ND_PRINTK(0, err,
1300 "RA: %s got default router without neighbour\n",
1301 __func__);
1302 fib6_info_release(rt);
1303 return;
1306 if (rt && lifetime == 0) {
1307 ip6_del_rt(net, rt, false);
1308 rt = NULL;
1311 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, for dev: %s\n",
1312 rt, lifetime, skb->dev->name);
1313 if (!rt && lifetime) {
1314 ND_PRINTK(3, info, "RA: adding default router\n");
1316 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1317 skb->dev, pref);
1318 if (!rt) {
1319 ND_PRINTK(0, err,
1320 "RA: %s failed to add default route\n",
1321 __func__);
1322 return;
1325 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1326 rt->fib6_nh->fib_nh_dev, NULL,
1327 &ipv6_hdr(skb)->saddr);
1328 if (!neigh) {
1329 ND_PRINTK(0, err,
1330 "RA: %s got default router without neighbour\n",
1331 __func__);
1332 fib6_info_release(rt);
1333 return;
1335 neigh->flags |= NTF_ROUTER;
1336 } else if (rt) {
1337 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1340 if (rt)
1341 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1342 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1343 ra_msg->icmph.icmp6_hop_limit) {
1344 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1345 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1346 fib6_metric_set(rt, RTAX_HOPLIMIT,
1347 ra_msg->icmph.icmp6_hop_limit);
1348 } else {
1349 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1353 skip_defrtr:
1356 * Update Reachable Time and Retrans Timer
1359 if (in6_dev->nd_parms) {
1360 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1362 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1363 rtime = (rtime*HZ)/1000;
1364 if (rtime < HZ/100)
1365 rtime = HZ/100;
1366 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1367 in6_dev->tstamp = jiffies;
1368 send_ifinfo_notify = true;
1371 rtime = ntohl(ra_msg->reachable_time);
1372 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1373 rtime = (rtime*HZ)/1000;
1375 if (rtime < HZ/10)
1376 rtime = HZ/10;
1378 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1379 NEIGH_VAR_SET(in6_dev->nd_parms,
1380 BASE_REACHABLE_TIME, rtime);
1381 NEIGH_VAR_SET(in6_dev->nd_parms,
1382 GC_STALETIME, 3 * rtime);
1383 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1384 in6_dev->tstamp = jiffies;
1385 send_ifinfo_notify = true;
1391 * Send a notify if RA changed managed/otherconf flags or timer settings
1393 if (send_ifinfo_notify)
1394 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1396 skip_linkparms:
1399 * Process options.
1402 if (!neigh)
1403 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1404 skb->dev, 1);
1405 if (neigh) {
1406 u8 *lladdr = NULL;
1407 if (ndopts.nd_opts_src_lladdr) {
1408 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1409 skb->dev);
1410 if (!lladdr) {
1411 ND_PRINTK(2, warn,
1412 "RA: invalid link-layer address length\n");
1413 goto out;
1416 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1417 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1418 NEIGH_UPDATE_F_OVERRIDE|
1419 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1420 NEIGH_UPDATE_F_ISROUTER,
1421 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1424 if (!ipv6_accept_ra(in6_dev)) {
1425 ND_PRINTK(2, info,
1426 "RA: %s, accept_ra is false for dev: %s\n",
1427 __func__, skb->dev->name);
1428 goto out;
1431 #ifdef CONFIG_IPV6_ROUTE_INFO
1432 if (!in6_dev->cnf.accept_ra_from_local &&
1433 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1434 in6_dev->dev, 0)) {
1435 ND_PRINTK(2, info,
1436 "RA from local address detected on dev: %s: router info ignored.\n",
1437 skb->dev->name);
1438 goto skip_routeinfo;
1441 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1442 struct nd_opt_hdr *p;
1443 for (p = ndopts.nd_opts_ri;
1445 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1446 struct route_info *ri = (struct route_info *)p;
1447 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1448 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1449 ri->prefix_len == 0)
1450 continue;
1451 #endif
1452 if (ri->prefix_len == 0 &&
1453 !in6_dev->cnf.accept_ra_defrtr)
1454 continue;
1455 if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1456 continue;
1457 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1458 continue;
1459 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1460 &ipv6_hdr(skb)->saddr);
1464 skip_routeinfo:
1465 #endif
1467 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1468 /* skip link-specific ndopts from interior routers */
1469 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1470 ND_PRINTK(2, info,
1471 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1472 __func__, skb->dev->name);
1473 goto out;
1475 #endif
1477 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1478 struct nd_opt_hdr *p;
1479 for (p = ndopts.nd_opts_pi;
1481 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1482 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1483 (p->nd_opt_len) << 3,
1484 ndopts.nd_opts_src_lladdr != NULL);
1488 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1489 __be32 n;
1490 u32 mtu;
1492 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1493 mtu = ntohl(n);
1495 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1496 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1497 } else if (in6_dev->cnf.mtu6 != mtu) {
1498 in6_dev->cnf.mtu6 = mtu;
1499 fib6_metric_set(rt, RTAX_MTU, mtu);
1500 rt6_mtu_change(skb->dev, mtu);
1504 if (ndopts.nd_useropts) {
1505 struct nd_opt_hdr *p;
1506 for (p = ndopts.nd_useropts;
1508 p = ndisc_next_useropt(skb->dev, p,
1509 ndopts.nd_useropts_end)) {
1510 ndisc_ra_useropt(skb, p);
1514 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1515 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1517 out:
1518 fib6_info_release(rt);
1519 if (neigh)
1520 neigh_release(neigh);
1523 static void ndisc_redirect_rcv(struct sk_buff *skb)
1525 u8 *hdr;
1526 struct ndisc_options ndopts;
1527 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1528 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1529 offsetof(struct rd_msg, opt));
1531 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1532 switch (skb->ndisc_nodetype) {
1533 case NDISC_NODETYPE_HOST:
1534 case NDISC_NODETYPE_NODEFAULT:
1535 ND_PRINTK(2, warn,
1536 "Redirect: from host or unauthorized router\n");
1537 return;
1539 #endif
1541 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1542 ND_PRINTK(2, warn,
1543 "Redirect: source address is not link-local\n");
1544 return;
1547 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1548 return;
1550 if (!ndopts.nd_opts_rh) {
1551 ip6_redirect_no_header(skb, dev_net(skb->dev),
1552 skb->dev->ifindex);
1553 return;
1556 hdr = (u8 *)ndopts.nd_opts_rh;
1557 hdr += 8;
1558 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1559 return;
1561 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1564 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1565 struct sk_buff *orig_skb,
1566 int rd_len)
1568 u8 *opt = skb_put(skb, rd_len);
1570 memset(opt, 0, 8);
1571 *(opt++) = ND_OPT_REDIRECT_HDR;
1572 *(opt++) = (rd_len >> 3);
1573 opt += 6;
1575 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1576 rd_len - 8);
1579 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1581 struct net_device *dev = skb->dev;
1582 struct net *net = dev_net(dev);
1583 struct sock *sk = net->ipv6.ndisc_sk;
1584 int optlen = 0;
1585 struct inet_peer *peer;
1586 struct sk_buff *buff;
1587 struct rd_msg *msg;
1588 struct in6_addr saddr_buf;
1589 struct rt6_info *rt;
1590 struct dst_entry *dst;
1591 struct flowi6 fl6;
1592 int rd_len;
1593 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1594 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1595 bool ret;
1597 if (netif_is_l3_master(skb->dev)) {
1598 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1599 if (!dev)
1600 return;
1603 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1604 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1605 dev->name);
1606 return;
1609 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1610 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1611 ND_PRINTK(2, warn,
1612 "Redirect: target address is not link-local unicast\n");
1613 return;
1616 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1617 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1619 dst = ip6_route_output(net, NULL, &fl6);
1620 if (dst->error) {
1621 dst_release(dst);
1622 return;
1624 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1625 if (IS_ERR(dst))
1626 return;
1628 rt = (struct rt6_info *) dst;
1630 if (rt->rt6i_flags & RTF_GATEWAY) {
1631 ND_PRINTK(2, warn,
1632 "Redirect: destination is not a neighbour\n");
1633 goto release;
1635 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1636 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1637 if (peer)
1638 inet_putpeer(peer);
1639 if (!ret)
1640 goto release;
1642 if (dev->addr_len) {
1643 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1644 if (!neigh) {
1645 ND_PRINTK(2, warn,
1646 "Redirect: no neigh for target address\n");
1647 goto release;
1650 read_lock_bh(&neigh->lock);
1651 if (neigh->nud_state & NUD_VALID) {
1652 memcpy(ha_buf, neigh->ha, dev->addr_len);
1653 read_unlock_bh(&neigh->lock);
1654 ha = ha_buf;
1655 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1656 ops_data_buf,
1657 &ops_data);
1658 } else
1659 read_unlock_bh(&neigh->lock);
1661 neigh_release(neigh);
1664 rd_len = min_t(unsigned int,
1665 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1666 skb->len + 8);
1667 rd_len &= ~0x7;
1668 optlen += rd_len;
1670 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1671 if (!buff)
1672 goto release;
1674 msg = skb_put(buff, sizeof(*msg));
1675 *msg = (struct rd_msg) {
1676 .icmph = {
1677 .icmp6_type = NDISC_REDIRECT,
1679 .target = *target,
1680 .dest = ipv6_hdr(skb)->daddr,
1684 * include target_address option
1687 if (ha)
1688 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1691 * build redirect option and copy skb over to the new packet.
1694 if (rd_len)
1695 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1697 skb_dst_set(buff, dst);
1698 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1699 return;
1701 release:
1702 dst_release(dst);
1705 static void pndisc_redo(struct sk_buff *skb)
1707 ndisc_recv_ns(skb);
1708 kfree_skb(skb);
1711 static int ndisc_is_multicast(const void *pkey)
1713 return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1716 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1718 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1720 if (!idev)
1721 return true;
1722 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1723 idev->cnf.suppress_frag_ndisc) {
1724 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1725 return true;
1727 return false;
1730 int ndisc_rcv(struct sk_buff *skb)
1732 struct nd_msg *msg;
1734 if (ndisc_suppress_frag_ndisc(skb))
1735 return 0;
1737 if (skb_linearize(skb))
1738 return 0;
1740 msg = (struct nd_msg *)skb_transport_header(skb);
1742 __skb_push(skb, skb->data - skb_transport_header(skb));
1744 if (ipv6_hdr(skb)->hop_limit != 255) {
1745 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1746 ipv6_hdr(skb)->hop_limit);
1747 return 0;
1750 if (msg->icmph.icmp6_code != 0) {
1751 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1752 msg->icmph.icmp6_code);
1753 return 0;
1756 switch (msg->icmph.icmp6_type) {
1757 case NDISC_NEIGHBOUR_SOLICITATION:
1758 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1759 ndisc_recv_ns(skb);
1760 break;
1762 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1763 ndisc_recv_na(skb);
1764 break;
1766 case NDISC_ROUTER_SOLICITATION:
1767 ndisc_recv_rs(skb);
1768 break;
1770 case NDISC_ROUTER_ADVERTISEMENT:
1771 ndisc_router_discovery(skb);
1772 break;
1774 case NDISC_REDIRECT:
1775 ndisc_redirect_rcv(skb);
1776 break;
1779 return 0;
1782 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1784 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1785 struct netdev_notifier_change_info *change_info;
1786 struct net *net = dev_net(dev);
1787 struct inet6_dev *idev;
1789 switch (event) {
1790 case NETDEV_CHANGEADDR:
1791 neigh_changeaddr(&nd_tbl, dev);
1792 fib6_run_gc(0, net, false);
1793 fallthrough;
1794 case NETDEV_UP:
1795 idev = in6_dev_get(dev);
1796 if (!idev)
1797 break;
1798 if (idev->cnf.ndisc_notify ||
1799 net->ipv6.devconf_all->ndisc_notify)
1800 ndisc_send_unsol_na(dev);
1801 in6_dev_put(idev);
1802 break;
1803 case NETDEV_CHANGE:
1804 change_info = ptr;
1805 if (change_info->flags_changed & IFF_NOARP)
1806 neigh_changeaddr(&nd_tbl, dev);
1807 if (!netif_carrier_ok(dev))
1808 neigh_carrier_down(&nd_tbl, dev);
1809 break;
1810 case NETDEV_DOWN:
1811 neigh_ifdown(&nd_tbl, dev);
1812 fib6_run_gc(0, net, false);
1813 break;
1814 case NETDEV_NOTIFY_PEERS:
1815 ndisc_send_unsol_na(dev);
1816 break;
1817 default:
1818 break;
1821 return NOTIFY_DONE;
1824 static struct notifier_block ndisc_netdev_notifier = {
1825 .notifier_call = ndisc_netdev_event,
1826 .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1829 #ifdef CONFIG_SYSCTL
1830 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1831 const char *func, const char *dev_name)
1833 static char warncomm[TASK_COMM_LEN];
1834 static int warned;
1835 if (strcmp(warncomm, current->comm) && warned < 5) {
1836 strcpy(warncomm, current->comm);
1837 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1838 warncomm, func,
1839 dev_name, ctl->procname,
1840 dev_name, ctl->procname);
1841 warned++;
1845 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void *buffer,
1846 size_t *lenp, loff_t *ppos)
1848 struct net_device *dev = ctl->extra1;
1849 struct inet6_dev *idev;
1850 int ret;
1852 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1853 (strcmp(ctl->procname, "base_reachable_time") == 0))
1854 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1856 if (strcmp(ctl->procname, "retrans_time") == 0)
1857 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1859 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1860 ret = neigh_proc_dointvec_jiffies(ctl, write,
1861 buffer, lenp, ppos);
1863 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1864 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1865 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1866 buffer, lenp, ppos);
1867 else
1868 ret = -1;
1870 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1871 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1872 idev->nd_parms->reachable_time =
1873 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1874 idev->tstamp = jiffies;
1875 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1876 in6_dev_put(idev);
1878 return ret;
1882 #endif
1884 static int __net_init ndisc_net_init(struct net *net)
1886 struct ipv6_pinfo *np;
1887 struct sock *sk;
1888 int err;
1890 err = inet_ctl_sock_create(&sk, PF_INET6,
1891 SOCK_RAW, IPPROTO_ICMPV6, net);
1892 if (err < 0) {
1893 ND_PRINTK(0, err,
1894 "NDISC: Failed to initialize the control socket (err %d)\n",
1895 err);
1896 return err;
1899 net->ipv6.ndisc_sk = sk;
1901 np = inet6_sk(sk);
1902 np->hop_limit = 255;
1903 /* Do not loopback ndisc messages */
1904 np->mc_loop = 0;
1906 return 0;
1909 static void __net_exit ndisc_net_exit(struct net *net)
1911 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1914 static struct pernet_operations ndisc_net_ops = {
1915 .init = ndisc_net_init,
1916 .exit = ndisc_net_exit,
1919 int __init ndisc_init(void)
1921 int err;
1923 err = register_pernet_subsys(&ndisc_net_ops);
1924 if (err)
1925 return err;
1927 * Initialize the neighbour table
1929 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1931 #ifdef CONFIG_SYSCTL
1932 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1933 ndisc_ifinfo_sysctl_change);
1934 if (err)
1935 goto out_unregister_pernet;
1936 out:
1937 #endif
1938 return err;
1940 #ifdef CONFIG_SYSCTL
1941 out_unregister_pernet:
1942 unregister_pernet_subsys(&ndisc_net_ops);
1943 goto out;
1944 #endif
1947 int __init ndisc_late_init(void)
1949 return register_netdevice_notifier(&ndisc_netdev_notifier);
1952 void ndisc_late_cleanup(void)
1954 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1957 void ndisc_cleanup(void)
1959 #ifdef CONFIG_SYSCTL
1960 neigh_sysctl_unregister(&nd_tbl.parms);
1961 #endif
1962 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1963 unregister_pernet_subsys(&ndisc_net_ops);