octeontx2-pf: Fix error return code in otx2_probe()
[linux/fpc-iii.git] / net / ipv6 / ndisc.c
blob2d09c4da03ee765d3dd50334f0c38050eebbbbb8
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);
85 static const struct neigh_ops ndisc_generic_ops = {
86 .family = AF_INET6,
87 .solicit = ndisc_solicit,
88 .error_report = ndisc_error_report,
89 .output = neigh_resolve_output,
90 .connected_output = neigh_connected_output,
93 static const struct neigh_ops ndisc_hh_ops = {
94 .family = AF_INET6,
95 .solicit = ndisc_solicit,
96 .error_report = ndisc_error_report,
97 .output = neigh_resolve_output,
98 .connected_output = neigh_resolve_output,
102 static const struct neigh_ops ndisc_direct_ops = {
103 .family = AF_INET6,
104 .output = neigh_direct_output,
105 .connected_output = neigh_direct_output,
108 struct neigh_table nd_tbl = {
109 .family = AF_INET6,
110 .key_len = sizeof(struct in6_addr),
111 .protocol = cpu_to_be16(ETH_P_IPV6),
112 .hash = ndisc_hash,
113 .key_eq = ndisc_key_eq,
114 .constructor = ndisc_constructor,
115 .pconstructor = pndisc_constructor,
116 .pdestructor = pndisc_destructor,
117 .proxy_redo = pndisc_redo,
118 .allow_add = ndisc_allow_add,
119 .id = "ndisc_cache",
120 .parms = {
121 .tbl = &nd_tbl,
122 .reachable_time = ND_REACHABLE_TIME,
123 .data = {
124 [NEIGH_VAR_MCAST_PROBES] = 3,
125 [NEIGH_VAR_UCAST_PROBES] = 3,
126 [NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER,
127 [NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME,
128 [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ,
129 [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
130 [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX,
131 [NEIGH_VAR_PROXY_QLEN] = 64,
132 [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
133 [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
136 .gc_interval = 30 * HZ,
137 .gc_thresh1 = 128,
138 .gc_thresh2 = 512,
139 .gc_thresh3 = 1024,
141 EXPORT_SYMBOL_GPL(nd_tbl);
143 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, void *data,
144 int data_len, int pad)
146 int space = __ndisc_opt_addr_space(data_len, pad);
147 u8 *opt = skb_put(skb, space);
149 opt[0] = type;
150 opt[1] = space>>3;
152 memset(opt + 2, 0, pad);
153 opt += pad;
154 space -= pad;
156 memcpy(opt+2, data, data_len);
157 data_len += 2;
158 opt += data_len;
159 space -= data_len;
160 if (space > 0)
161 memset(opt, 0, space);
163 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
165 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
166 void *data, u8 icmp6_type)
168 __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
169 ndisc_addr_option_pad(skb->dev->type));
170 ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
173 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
174 void *ha,
175 const u8 *ops_data)
177 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
178 ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
181 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
182 struct nd_opt_hdr *end)
184 int type;
185 if (!cur || !end || cur >= end)
186 return NULL;
187 type = cur->nd_opt_type;
188 do {
189 cur = ((void *)cur) + (cur->nd_opt_len << 3);
190 } while (cur < end && cur->nd_opt_type != type);
191 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
194 static inline int ndisc_is_useropt(const struct net_device *dev,
195 struct nd_opt_hdr *opt)
197 return opt->nd_opt_type == ND_OPT_RDNSS ||
198 opt->nd_opt_type == ND_OPT_DNSSL ||
199 opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
200 opt->nd_opt_type == ND_OPT_PREF64 ||
201 ndisc_ops_is_useropt(dev, opt->nd_opt_type);
204 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
205 struct nd_opt_hdr *cur,
206 struct nd_opt_hdr *end)
208 if (!cur || !end || cur >= end)
209 return NULL;
210 do {
211 cur = ((void *)cur) + (cur->nd_opt_len << 3);
212 } while (cur < end && !ndisc_is_useropt(dev, cur));
213 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
216 struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
217 u8 *opt, int opt_len,
218 struct ndisc_options *ndopts)
220 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
222 if (!nd_opt || opt_len < 0 || !ndopts)
223 return NULL;
224 memset(ndopts, 0, sizeof(*ndopts));
225 while (opt_len) {
226 int l;
227 if (opt_len < sizeof(struct nd_opt_hdr))
228 return NULL;
229 l = nd_opt->nd_opt_len << 3;
230 if (opt_len < l || l == 0)
231 return NULL;
232 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
233 goto next_opt;
234 switch (nd_opt->nd_opt_type) {
235 case ND_OPT_SOURCE_LL_ADDR:
236 case ND_OPT_TARGET_LL_ADDR:
237 case ND_OPT_MTU:
238 case ND_OPT_NONCE:
239 case ND_OPT_REDIRECT_HDR:
240 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
241 ND_PRINTK(2, warn,
242 "%s: duplicated ND6 option found: type=%d\n",
243 __func__, nd_opt->nd_opt_type);
244 } else {
245 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
247 break;
248 case ND_OPT_PREFIX_INFO:
249 ndopts->nd_opts_pi_end = nd_opt;
250 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
251 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
252 break;
253 #ifdef CONFIG_IPV6_ROUTE_INFO
254 case ND_OPT_ROUTE_INFO:
255 ndopts->nd_opts_ri_end = nd_opt;
256 if (!ndopts->nd_opts_ri)
257 ndopts->nd_opts_ri = nd_opt;
258 break;
259 #endif
260 default:
261 if (ndisc_is_useropt(dev, nd_opt)) {
262 ndopts->nd_useropts_end = nd_opt;
263 if (!ndopts->nd_useropts)
264 ndopts->nd_useropts = nd_opt;
265 } else {
267 * Unknown options must be silently ignored,
268 * to accommodate future extension to the
269 * protocol.
271 ND_PRINTK(2, notice,
272 "%s: ignored unsupported option; type=%d, len=%d\n",
273 __func__,
274 nd_opt->nd_opt_type,
275 nd_opt->nd_opt_len);
278 next_opt:
279 opt_len -= l;
280 nd_opt = ((void *)nd_opt) + l;
282 return ndopts;
285 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
287 switch (dev->type) {
288 case ARPHRD_ETHER:
289 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
290 case ARPHRD_FDDI:
291 ipv6_eth_mc_map(addr, buf);
292 return 0;
293 case ARPHRD_ARCNET:
294 ipv6_arcnet_mc_map(addr, buf);
295 return 0;
296 case ARPHRD_INFINIBAND:
297 ipv6_ib_mc_map(addr, dev->broadcast, buf);
298 return 0;
299 case ARPHRD_IPGRE:
300 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
301 default:
302 if (dir) {
303 memcpy(buf, dev->broadcast, dev->addr_len);
304 return 0;
307 return -EINVAL;
309 EXPORT_SYMBOL(ndisc_mc_map);
311 static u32 ndisc_hash(const void *pkey,
312 const struct net_device *dev,
313 __u32 *hash_rnd)
315 return ndisc_hashfn(pkey, dev, hash_rnd);
318 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
320 return neigh_key_eq128(n, pkey);
323 static int ndisc_constructor(struct neighbour *neigh)
325 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
326 struct net_device *dev = neigh->dev;
327 struct inet6_dev *in6_dev;
328 struct neigh_parms *parms;
329 bool is_multicast = ipv6_addr_is_multicast(addr);
331 in6_dev = in6_dev_get(dev);
332 if (!in6_dev) {
333 return -EINVAL;
336 parms = in6_dev->nd_parms;
337 __neigh_parms_put(neigh->parms);
338 neigh->parms = neigh_parms_clone(parms);
340 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
341 if (!dev->header_ops) {
342 neigh->nud_state = NUD_NOARP;
343 neigh->ops = &ndisc_direct_ops;
344 neigh->output = neigh_direct_output;
345 } else {
346 if (is_multicast) {
347 neigh->nud_state = NUD_NOARP;
348 ndisc_mc_map(addr, neigh->ha, dev, 1);
349 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
350 neigh->nud_state = NUD_NOARP;
351 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
352 if (dev->flags&IFF_LOOPBACK)
353 neigh->type = RTN_LOCAL;
354 } else if (dev->flags&IFF_POINTOPOINT) {
355 neigh->nud_state = NUD_NOARP;
356 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
358 if (dev->header_ops->cache)
359 neigh->ops = &ndisc_hh_ops;
360 else
361 neigh->ops = &ndisc_generic_ops;
362 if (neigh->nud_state&NUD_VALID)
363 neigh->output = neigh->ops->connected_output;
364 else
365 neigh->output = neigh->ops->output;
367 in6_dev_put(in6_dev);
368 return 0;
371 static int pndisc_constructor(struct pneigh_entry *n)
373 struct in6_addr *addr = (struct in6_addr *)&n->key;
374 struct in6_addr maddr;
375 struct net_device *dev = n->dev;
377 if (!dev || !__in6_dev_get(dev))
378 return -EINVAL;
379 addrconf_addr_solict_mult(addr, &maddr);
380 ipv6_dev_mc_inc(dev, &maddr);
381 return 0;
384 static void pndisc_destructor(struct pneigh_entry *n)
386 struct in6_addr *addr = (struct in6_addr *)&n->key;
387 struct in6_addr maddr;
388 struct net_device *dev = n->dev;
390 if (!dev || !__in6_dev_get(dev))
391 return;
392 addrconf_addr_solict_mult(addr, &maddr);
393 ipv6_dev_mc_dec(dev, &maddr);
396 /* called with rtnl held */
397 static bool ndisc_allow_add(const struct net_device *dev,
398 struct netlink_ext_ack *extack)
400 struct inet6_dev *idev = __in6_dev_get(dev);
402 if (!idev || idev->cnf.disable_ipv6) {
403 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
404 return false;
407 return true;
410 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
411 int len)
413 int hlen = LL_RESERVED_SPACE(dev);
414 int tlen = dev->needed_tailroom;
415 struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
416 struct sk_buff *skb;
418 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
419 if (!skb) {
420 ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
421 __func__);
422 return NULL;
425 skb->protocol = htons(ETH_P_IPV6);
426 skb->dev = dev;
428 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
429 skb_reset_transport_header(skb);
431 /* Manually assign socket ownership as we avoid calling
432 * sock_alloc_send_pskb() to bypass wmem buffer limits
434 skb_set_owner_w(skb, sk);
436 return skb;
439 static void ip6_nd_hdr(struct sk_buff *skb,
440 const struct in6_addr *saddr,
441 const struct in6_addr *daddr,
442 int hop_limit, int len)
444 struct ipv6hdr *hdr;
445 struct inet6_dev *idev;
446 unsigned tclass;
448 rcu_read_lock();
449 idev = __in6_dev_get(skb->dev);
450 tclass = idev ? idev->cnf.ndisc_tclass : 0;
451 rcu_read_unlock();
453 skb_push(skb, sizeof(*hdr));
454 skb_reset_network_header(skb);
455 hdr = ipv6_hdr(skb);
457 ip6_flow_hdr(hdr, tclass, 0);
459 hdr->payload_len = htons(len);
460 hdr->nexthdr = IPPROTO_ICMPV6;
461 hdr->hop_limit = hop_limit;
463 hdr->saddr = *saddr;
464 hdr->daddr = *daddr;
467 static void ndisc_send_skb(struct sk_buff *skb,
468 const struct in6_addr *daddr,
469 const struct in6_addr *saddr)
471 struct dst_entry *dst = skb_dst(skb);
472 struct net *net = dev_net(skb->dev);
473 struct sock *sk = net->ipv6.ndisc_sk;
474 struct inet6_dev *idev;
475 int err;
476 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
477 u8 type;
479 type = icmp6h->icmp6_type;
481 if (!dst) {
482 struct flowi6 fl6;
483 int oif = skb->dev->ifindex;
485 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
486 dst = icmp6_dst_alloc(skb->dev, &fl6);
487 if (IS_ERR(dst)) {
488 kfree_skb(skb);
489 return;
492 skb_dst_set(skb, dst);
495 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
496 IPPROTO_ICMPV6,
497 csum_partial(icmp6h,
498 skb->len, 0));
500 ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
502 rcu_read_lock();
503 idev = __in6_dev_get(dst->dev);
504 IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
506 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
507 net, sk, skb, NULL, dst->dev,
508 dst_output);
509 if (!err) {
510 ICMP6MSGOUT_INC_STATS(net, idev, type);
511 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
514 rcu_read_unlock();
517 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
518 const struct in6_addr *solicited_addr,
519 bool router, bool solicited, bool override, bool inc_opt)
521 struct sk_buff *skb;
522 struct in6_addr tmpaddr;
523 struct inet6_ifaddr *ifp;
524 const struct in6_addr *src_addr;
525 struct nd_msg *msg;
526 int optlen = 0;
528 /* for anycast or proxy, solicited_addr != src_addr */
529 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
530 if (ifp) {
531 src_addr = solicited_addr;
532 if (ifp->flags & IFA_F_OPTIMISTIC)
533 override = false;
534 inc_opt |= ifp->idev->cnf.force_tllao;
535 in6_ifa_put(ifp);
536 } else {
537 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
538 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
539 &tmpaddr))
540 return;
541 src_addr = &tmpaddr;
544 if (!dev->addr_len)
545 inc_opt = false;
546 if (inc_opt)
547 optlen += ndisc_opt_addr_space(dev,
548 NDISC_NEIGHBOUR_ADVERTISEMENT);
550 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
551 if (!skb)
552 return;
554 msg = skb_put(skb, sizeof(*msg));
555 *msg = (struct nd_msg) {
556 .icmph = {
557 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
558 .icmp6_router = router,
559 .icmp6_solicited = solicited,
560 .icmp6_override = override,
562 .target = *solicited_addr,
565 if (inc_opt)
566 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
567 dev->dev_addr,
568 NDISC_NEIGHBOUR_ADVERTISEMENT);
570 ndisc_send_skb(skb, daddr, src_addr);
573 static void ndisc_send_unsol_na(struct net_device *dev)
575 struct inet6_dev *idev;
576 struct inet6_ifaddr *ifa;
578 idev = in6_dev_get(dev);
579 if (!idev)
580 return;
582 read_lock_bh(&idev->lock);
583 list_for_each_entry(ifa, &idev->addr_list, if_list) {
584 /* skip tentative addresses until dad completes */
585 if (ifa->flags & IFA_F_TENTATIVE &&
586 !(ifa->flags & IFA_F_OPTIMISTIC))
587 continue;
589 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
590 /*router=*/ !!idev->cnf.forwarding,
591 /*solicited=*/ false, /*override=*/ true,
592 /*inc_opt=*/ true);
594 read_unlock_bh(&idev->lock);
596 in6_dev_put(idev);
599 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
600 const struct in6_addr *daddr, const struct in6_addr *saddr,
601 u64 nonce)
603 struct sk_buff *skb;
604 struct in6_addr addr_buf;
605 int inc_opt = dev->addr_len;
606 int optlen = 0;
607 struct nd_msg *msg;
609 if (!saddr) {
610 if (ipv6_get_lladdr(dev, &addr_buf,
611 (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)))
612 return;
613 saddr = &addr_buf;
616 if (ipv6_addr_any(saddr))
617 inc_opt = false;
618 if (inc_opt)
619 optlen += ndisc_opt_addr_space(dev,
620 NDISC_NEIGHBOUR_SOLICITATION);
621 if (nonce != 0)
622 optlen += 8;
624 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
625 if (!skb)
626 return;
628 msg = skb_put(skb, sizeof(*msg));
629 *msg = (struct nd_msg) {
630 .icmph = {
631 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
633 .target = *solicit,
636 if (inc_opt)
637 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
638 dev->dev_addr,
639 NDISC_NEIGHBOUR_SOLICITATION);
640 if (nonce != 0) {
641 u8 *opt = skb_put(skb, 8);
643 opt[0] = ND_OPT_NONCE;
644 opt[1] = 8 >> 3;
645 memcpy(opt + 2, &nonce, 6);
648 ndisc_send_skb(skb, daddr, saddr);
651 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
652 const struct in6_addr *daddr)
654 struct sk_buff *skb;
655 struct rs_msg *msg;
656 int send_sllao = dev->addr_len;
657 int optlen = 0;
659 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
661 * According to section 2.2 of RFC 4429, we must not
662 * send router solicitations with a sllao from
663 * optimistic addresses, but we may send the solicitation
664 * if we don't include the sllao. So here we check
665 * if our address is optimistic, and if so, we
666 * suppress the inclusion of the sllao.
668 if (send_sllao) {
669 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
670 dev, 1);
671 if (ifp) {
672 if (ifp->flags & IFA_F_OPTIMISTIC) {
673 send_sllao = 0;
675 in6_ifa_put(ifp);
676 } else {
677 send_sllao = 0;
680 #endif
681 if (send_sllao)
682 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
684 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
685 if (!skb)
686 return;
688 msg = skb_put(skb, sizeof(*msg));
689 *msg = (struct rs_msg) {
690 .icmph = {
691 .icmp6_type = NDISC_ROUTER_SOLICITATION,
695 if (send_sllao)
696 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
697 dev->dev_addr,
698 NDISC_ROUTER_SOLICITATION);
700 ndisc_send_skb(skb, daddr, saddr);
704 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
707 * "The sender MUST return an ICMP
708 * destination unreachable"
710 dst_link_failure(skb);
711 kfree_skb(skb);
714 /* Called with locked neigh: either read or both */
716 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
718 struct in6_addr *saddr = NULL;
719 struct in6_addr mcaddr;
720 struct net_device *dev = neigh->dev;
721 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
722 int probes = atomic_read(&neigh->probes);
724 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
725 dev, false, 1,
726 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
727 saddr = &ipv6_hdr(skb)->saddr;
728 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
729 if (probes < 0) {
730 if (!(neigh->nud_state & NUD_VALID)) {
731 ND_PRINTK(1, dbg,
732 "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
733 __func__, target);
735 ndisc_send_ns(dev, target, target, saddr, 0);
736 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
737 neigh_app_ns(neigh);
738 } else {
739 addrconf_addr_solict_mult(target, &mcaddr);
740 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
744 static int pndisc_is_router(const void *pkey,
745 struct net_device *dev)
747 struct pneigh_entry *n;
748 int ret = -1;
750 read_lock_bh(&nd_tbl.lock);
751 n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
752 if (n)
753 ret = !!(n->flags & NTF_ROUTER);
754 read_unlock_bh(&nd_tbl.lock);
756 return ret;
759 void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
760 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
761 struct ndisc_options *ndopts)
763 neigh_update(neigh, lladdr, new, flags, 0);
764 /* report ndisc ops about neighbour update */
765 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
768 static void ndisc_recv_ns(struct sk_buff *skb)
770 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
771 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
772 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
773 u8 *lladdr = NULL;
774 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
775 offsetof(struct nd_msg, opt));
776 struct ndisc_options ndopts;
777 struct net_device *dev = skb->dev;
778 struct inet6_ifaddr *ifp;
779 struct inet6_dev *idev = NULL;
780 struct neighbour *neigh;
781 int dad = ipv6_addr_any(saddr);
782 bool inc;
783 int is_router = -1;
784 u64 nonce = 0;
786 if (skb->len < sizeof(struct nd_msg)) {
787 ND_PRINTK(2, warn, "NS: packet too short\n");
788 return;
791 if (ipv6_addr_is_multicast(&msg->target)) {
792 ND_PRINTK(2, warn, "NS: multicast target address\n");
793 return;
797 * RFC2461 7.1.1:
798 * DAD has to be destined for solicited node multicast address.
800 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
801 ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
802 return;
805 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
806 ND_PRINTK(2, warn, "NS: invalid ND options\n");
807 return;
810 if (ndopts.nd_opts_src_lladdr) {
811 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
812 if (!lladdr) {
813 ND_PRINTK(2, warn,
814 "NS: invalid link-layer address length\n");
815 return;
818 /* RFC2461 7.1.1:
819 * If the IP source address is the unspecified address,
820 * there MUST NOT be source link-layer address option
821 * in the message.
823 if (dad) {
824 ND_PRINTK(2, warn,
825 "NS: bad DAD packet (link-layer address option)\n");
826 return;
829 if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
830 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
832 inc = ipv6_addr_is_multicast(daddr);
834 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
835 if (ifp) {
836 have_ifp:
837 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
838 if (dad) {
839 if (nonce != 0 && ifp->dad_nonce == nonce) {
840 u8 *np = (u8 *)&nonce;
841 /* Matching nonce if looped back */
842 ND_PRINTK(2, notice,
843 "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
844 ifp->idev->dev->name,
845 &ifp->addr, np);
846 goto out;
849 * We are colliding with another node
850 * who is doing DAD
851 * so fail our DAD process
853 addrconf_dad_failure(skb, ifp);
854 return;
855 } else {
857 * This is not a dad solicitation.
858 * If we are an optimistic node,
859 * we should respond.
860 * Otherwise, we should ignore it.
862 if (!(ifp->flags & IFA_F_OPTIMISTIC))
863 goto out;
867 idev = ifp->idev;
868 } else {
869 struct net *net = dev_net(dev);
871 /* perhaps an address on the master device */
872 if (netif_is_l3_slave(dev)) {
873 struct net_device *mdev;
875 mdev = netdev_master_upper_dev_get_rcu(dev);
876 if (mdev) {
877 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
878 if (ifp)
879 goto have_ifp;
883 idev = in6_dev_get(dev);
884 if (!idev) {
885 /* XXX: count this drop? */
886 return;
889 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
890 (idev->cnf.forwarding &&
891 (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
892 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
893 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
894 skb->pkt_type != PACKET_HOST &&
895 inc &&
896 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
898 * for anycast or proxy,
899 * sender should delay its response
900 * by a random time between 0 and
901 * MAX_ANYCAST_DELAY_TIME seconds.
902 * (RFC2461) -- yoshfuji
904 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
905 if (n)
906 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
907 goto out;
909 } else
910 goto out;
913 if (is_router < 0)
914 is_router = idev->cnf.forwarding;
916 if (dad) {
917 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
918 !!is_router, false, (ifp != NULL), true);
919 goto out;
922 if (inc)
923 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
924 else
925 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
928 * update / create cache entry
929 * for the source address
931 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
932 !inc || lladdr || !dev->addr_len);
933 if (neigh)
934 ndisc_update(dev, neigh, lladdr, NUD_STALE,
935 NEIGH_UPDATE_F_WEAK_OVERRIDE|
936 NEIGH_UPDATE_F_OVERRIDE,
937 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
938 if (neigh || !dev->header_ops) {
939 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
940 true, (ifp != NULL && inc), inc);
941 if (neigh)
942 neigh_release(neigh);
945 out:
946 if (ifp)
947 in6_ifa_put(ifp);
948 else
949 in6_dev_put(idev);
952 static void ndisc_recv_na(struct sk_buff *skb)
954 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
955 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
956 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
957 u8 *lladdr = NULL;
958 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
959 offsetof(struct nd_msg, opt));
960 struct ndisc_options ndopts;
961 struct net_device *dev = skb->dev;
962 struct inet6_dev *idev = __in6_dev_get(dev);
963 struct inet6_ifaddr *ifp;
964 struct neighbour *neigh;
966 if (skb->len < sizeof(struct nd_msg)) {
967 ND_PRINTK(2, warn, "NA: packet too short\n");
968 return;
971 if (ipv6_addr_is_multicast(&msg->target)) {
972 ND_PRINTK(2, warn, "NA: target address is multicast\n");
973 return;
976 if (ipv6_addr_is_multicast(daddr) &&
977 msg->icmph.icmp6_solicited) {
978 ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
979 return;
982 /* For some 802.11 wireless deployments (and possibly other networks),
983 * there will be a NA proxy and unsolicitd packets are attacks
984 * and thus should not be accepted.
986 if (!msg->icmph.icmp6_solicited && idev &&
987 idev->cnf.drop_unsolicited_na)
988 return;
990 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
991 ND_PRINTK(2, warn, "NS: invalid ND option\n");
992 return;
994 if (ndopts.nd_opts_tgt_lladdr) {
995 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
996 if (!lladdr) {
997 ND_PRINTK(2, warn,
998 "NA: invalid link-layer address length\n");
999 return;
1002 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1003 if (ifp) {
1004 if (skb->pkt_type != PACKET_LOOPBACK
1005 && (ifp->flags & IFA_F_TENTATIVE)) {
1006 addrconf_dad_failure(skb, ifp);
1007 return;
1009 /* What should we make now? The advertisement
1010 is invalid, but ndisc specs say nothing
1011 about it. It could be misconfiguration, or
1012 an smart proxy agent tries to help us :-)
1014 We should not print the error if NA has been
1015 received from loopback - it is just our own
1016 unsolicited advertisement.
1018 if (skb->pkt_type != PACKET_LOOPBACK)
1019 ND_PRINTK(1, warn,
1020 "NA: %pM advertised our address %pI6c on %s!\n",
1021 eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1022 in6_ifa_put(ifp);
1023 return;
1025 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1027 if (neigh) {
1028 u8 old_flags = neigh->flags;
1029 struct net *net = dev_net(dev);
1031 if (neigh->nud_state & NUD_FAILED)
1032 goto out;
1035 * Don't update the neighbor cache entry on a proxy NA from
1036 * ourselves because either the proxied node is off link or it
1037 * has already sent a NA to us.
1039 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1040 net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
1041 pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1042 /* XXX: idev->cnf.proxy_ndp */
1043 goto out;
1046 ndisc_update(dev, neigh, lladdr,
1047 msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE,
1048 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1049 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1050 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1051 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1052 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1054 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1056 * Change: router to host
1058 rt6_clean_tohost(dev_net(dev), saddr);
1061 out:
1062 neigh_release(neigh);
1066 static void ndisc_recv_rs(struct sk_buff *skb)
1068 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1069 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1070 struct neighbour *neigh;
1071 struct inet6_dev *idev;
1072 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1073 struct ndisc_options ndopts;
1074 u8 *lladdr = NULL;
1076 if (skb->len < sizeof(*rs_msg))
1077 return;
1079 idev = __in6_dev_get(skb->dev);
1080 if (!idev) {
1081 ND_PRINTK(1, err, "RS: can't find in6 device\n");
1082 return;
1085 /* Don't accept RS if we're not in router mode */
1086 if (!idev->cnf.forwarding)
1087 goto out;
1090 * Don't update NCE if src = ::;
1091 * this implies that the source node has no ip address assigned yet.
1093 if (ipv6_addr_any(saddr))
1094 goto out;
1096 /* Parse ND options */
1097 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1098 ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1099 goto out;
1102 if (ndopts.nd_opts_src_lladdr) {
1103 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1104 skb->dev);
1105 if (!lladdr)
1106 goto out;
1109 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1110 if (neigh) {
1111 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1112 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1113 NEIGH_UPDATE_F_OVERRIDE|
1114 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1115 NDISC_ROUTER_SOLICITATION, &ndopts);
1116 neigh_release(neigh);
1118 out:
1119 return;
1122 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1124 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1125 struct sk_buff *skb;
1126 struct nlmsghdr *nlh;
1127 struct nduseroptmsg *ndmsg;
1128 struct net *net = dev_net(ra->dev);
1129 int err;
1130 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1131 + (opt->nd_opt_len << 3));
1132 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1134 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1135 if (!skb) {
1136 err = -ENOBUFS;
1137 goto errout;
1140 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1141 if (!nlh) {
1142 goto nla_put_failure;
1145 ndmsg = nlmsg_data(nlh);
1146 ndmsg->nduseropt_family = AF_INET6;
1147 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1148 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1149 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1150 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1152 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1154 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1155 goto nla_put_failure;
1156 nlmsg_end(skb, nlh);
1158 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1159 return;
1161 nla_put_failure:
1162 nlmsg_free(skb);
1163 err = -EMSGSIZE;
1164 errout:
1165 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1168 static void ndisc_router_discovery(struct sk_buff *skb)
1170 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1171 struct neighbour *neigh = NULL;
1172 struct inet6_dev *in6_dev;
1173 struct fib6_info *rt = NULL;
1174 struct net *net;
1175 int lifetime;
1176 struct ndisc_options ndopts;
1177 int optlen;
1178 unsigned int pref = 0;
1179 __u32 old_if_flags;
1180 bool send_ifinfo_notify = false;
1182 __u8 *opt = (__u8 *)(ra_msg + 1);
1184 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1185 sizeof(struct ra_msg);
1187 ND_PRINTK(2, info,
1188 "RA: %s, dev: %s\n",
1189 __func__, skb->dev->name);
1190 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1191 ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1192 return;
1194 if (optlen < 0) {
1195 ND_PRINTK(2, warn, "RA: packet too short\n");
1196 return;
1199 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1200 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1201 ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1202 return;
1204 #endif
1207 * set the RA_RECV flag in the interface
1210 in6_dev = __in6_dev_get(skb->dev);
1211 if (!in6_dev) {
1212 ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1213 skb->dev->name);
1214 return;
1217 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1218 ND_PRINTK(2, warn, "RA: invalid ND options\n");
1219 return;
1222 if (!ipv6_accept_ra(in6_dev)) {
1223 ND_PRINTK(2, info,
1224 "RA: %s, did not accept ra for dev: %s\n",
1225 __func__, skb->dev->name);
1226 goto skip_linkparms;
1229 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1230 /* skip link-specific parameters from interior routers */
1231 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1232 ND_PRINTK(2, info,
1233 "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1234 __func__, skb->dev->name);
1235 goto skip_linkparms;
1237 #endif
1239 if (in6_dev->if_flags & IF_RS_SENT) {
1241 * flag that an RA was received after an RS was sent
1242 * out on this interface.
1244 in6_dev->if_flags |= IF_RA_RCVD;
1248 * Remember the managed/otherconf flags from most recently
1249 * received RA message (RFC 2462) -- yoshfuji
1251 old_if_flags = in6_dev->if_flags;
1252 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1253 IF_RA_OTHERCONF)) |
1254 (ra_msg->icmph.icmp6_addrconf_managed ?
1255 IF_RA_MANAGED : 0) |
1256 (ra_msg->icmph.icmp6_addrconf_other ?
1257 IF_RA_OTHERCONF : 0);
1259 if (old_if_flags != in6_dev->if_flags)
1260 send_ifinfo_notify = true;
1262 if (!in6_dev->cnf.accept_ra_defrtr) {
1263 ND_PRINTK(2, info,
1264 "RA: %s, defrtr is false for dev: %s\n",
1265 __func__, skb->dev->name);
1266 goto skip_defrtr;
1269 /* Do not accept RA with source-addr found on local machine unless
1270 * accept_ra_from_local is set to true.
1272 net = dev_net(in6_dev->dev);
1273 if (!in6_dev->cnf.accept_ra_from_local &&
1274 ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1275 ND_PRINTK(2, info,
1276 "RA from local address detected on dev: %s: default router ignored\n",
1277 skb->dev->name);
1278 goto skip_defrtr;
1281 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1283 #ifdef CONFIG_IPV6_ROUTER_PREF
1284 pref = ra_msg->icmph.icmp6_router_pref;
1285 /* 10b is handled as if it were 00b (medium) */
1286 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1287 !in6_dev->cnf.accept_ra_rtr_pref)
1288 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1289 #endif
1290 /* routes added from RAs do not use nexthop objects */
1291 rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1292 if (rt) {
1293 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1294 rt->fib6_nh->fib_nh_dev, NULL,
1295 &ipv6_hdr(skb)->saddr);
1296 if (!neigh) {
1297 ND_PRINTK(0, err,
1298 "RA: %s got default router without neighbour\n",
1299 __func__);
1300 fib6_info_release(rt);
1301 return;
1304 if (rt && lifetime == 0) {
1305 ip6_del_rt(net, rt, false);
1306 rt = NULL;
1309 ND_PRINTK(3, info, "RA: rt: %p lifetime: %d, for dev: %s\n",
1310 rt, lifetime, skb->dev->name);
1311 if (!rt && lifetime) {
1312 ND_PRINTK(3, info, "RA: adding default router\n");
1314 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1315 skb->dev, pref);
1316 if (!rt) {
1317 ND_PRINTK(0, err,
1318 "RA: %s failed to add default route\n",
1319 __func__);
1320 return;
1323 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1324 rt->fib6_nh->fib_nh_dev, NULL,
1325 &ipv6_hdr(skb)->saddr);
1326 if (!neigh) {
1327 ND_PRINTK(0, err,
1328 "RA: %s got default router without neighbour\n",
1329 __func__);
1330 fib6_info_release(rt);
1331 return;
1333 neigh->flags |= NTF_ROUTER;
1334 } else if (rt) {
1335 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1338 if (rt)
1339 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1340 if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
1341 ra_msg->icmph.icmp6_hop_limit) {
1342 if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1343 in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
1344 fib6_metric_set(rt, RTAX_HOPLIMIT,
1345 ra_msg->icmph.icmp6_hop_limit);
1346 } else {
1347 ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1351 skip_defrtr:
1354 * Update Reachable Time and Retrans Timer
1357 if (in6_dev->nd_parms) {
1358 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1360 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1361 rtime = (rtime*HZ)/1000;
1362 if (rtime < HZ/100)
1363 rtime = HZ/100;
1364 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1365 in6_dev->tstamp = jiffies;
1366 send_ifinfo_notify = true;
1369 rtime = ntohl(ra_msg->reachable_time);
1370 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1371 rtime = (rtime*HZ)/1000;
1373 if (rtime < HZ/10)
1374 rtime = HZ/10;
1376 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1377 NEIGH_VAR_SET(in6_dev->nd_parms,
1378 BASE_REACHABLE_TIME, rtime);
1379 NEIGH_VAR_SET(in6_dev->nd_parms,
1380 GC_STALETIME, 3 * rtime);
1381 in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1382 in6_dev->tstamp = jiffies;
1383 send_ifinfo_notify = true;
1389 * Send a notify if RA changed managed/otherconf flags or timer settings
1391 if (send_ifinfo_notify)
1392 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1394 skip_linkparms:
1397 * Process options.
1400 if (!neigh)
1401 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1402 skb->dev, 1);
1403 if (neigh) {
1404 u8 *lladdr = NULL;
1405 if (ndopts.nd_opts_src_lladdr) {
1406 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1407 skb->dev);
1408 if (!lladdr) {
1409 ND_PRINTK(2, warn,
1410 "RA: invalid link-layer address length\n");
1411 goto out;
1414 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1415 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1416 NEIGH_UPDATE_F_OVERRIDE|
1417 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1418 NEIGH_UPDATE_F_ISROUTER,
1419 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1422 if (!ipv6_accept_ra(in6_dev)) {
1423 ND_PRINTK(2, info,
1424 "RA: %s, accept_ra is false for dev: %s\n",
1425 __func__, skb->dev->name);
1426 goto out;
1429 #ifdef CONFIG_IPV6_ROUTE_INFO
1430 if (!in6_dev->cnf.accept_ra_from_local &&
1431 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1432 in6_dev->dev, 0)) {
1433 ND_PRINTK(2, info,
1434 "RA from local address detected on dev: %s: router info ignored.\n",
1435 skb->dev->name);
1436 goto skip_routeinfo;
1439 if (in6_dev->cnf.accept_ra_rtr_pref && ndopts.nd_opts_ri) {
1440 struct nd_opt_hdr *p;
1441 for (p = ndopts.nd_opts_ri;
1443 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1444 struct route_info *ri = (struct route_info *)p;
1445 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1446 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1447 ri->prefix_len == 0)
1448 continue;
1449 #endif
1450 if (ri->prefix_len == 0 &&
1451 !in6_dev->cnf.accept_ra_defrtr)
1452 continue;
1453 if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1454 continue;
1455 if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1456 continue;
1457 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1458 &ipv6_hdr(skb)->saddr);
1462 skip_routeinfo:
1463 #endif
1465 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1466 /* skip link-specific ndopts from interior routers */
1467 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1468 ND_PRINTK(2, info,
1469 "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1470 __func__, skb->dev->name);
1471 goto out;
1473 #endif
1475 if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1476 struct nd_opt_hdr *p;
1477 for (p = ndopts.nd_opts_pi;
1479 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1480 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1481 (p->nd_opt_len) << 3,
1482 ndopts.nd_opts_src_lladdr != NULL);
1486 if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1487 __be32 n;
1488 u32 mtu;
1490 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1491 mtu = ntohl(n);
1493 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1494 ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1495 } else if (in6_dev->cnf.mtu6 != mtu) {
1496 in6_dev->cnf.mtu6 = mtu;
1497 fib6_metric_set(rt, RTAX_MTU, mtu);
1498 rt6_mtu_change(skb->dev, mtu);
1502 if (ndopts.nd_useropts) {
1503 struct nd_opt_hdr *p;
1504 for (p = ndopts.nd_useropts;
1506 p = ndisc_next_useropt(skb->dev, p,
1507 ndopts.nd_useropts_end)) {
1508 ndisc_ra_useropt(skb, p);
1512 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1513 ND_PRINTK(2, warn, "RA: invalid RA options\n");
1515 out:
1516 fib6_info_release(rt);
1517 if (neigh)
1518 neigh_release(neigh);
1521 static void ndisc_redirect_rcv(struct sk_buff *skb)
1523 u8 *hdr;
1524 struct ndisc_options ndopts;
1525 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1526 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1527 offsetof(struct rd_msg, opt));
1529 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1530 switch (skb->ndisc_nodetype) {
1531 case NDISC_NODETYPE_HOST:
1532 case NDISC_NODETYPE_NODEFAULT:
1533 ND_PRINTK(2, warn,
1534 "Redirect: from host or unauthorized router\n");
1535 return;
1537 #endif
1539 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1540 ND_PRINTK(2, warn,
1541 "Redirect: source address is not link-local\n");
1542 return;
1545 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1546 return;
1548 if (!ndopts.nd_opts_rh) {
1549 ip6_redirect_no_header(skb, dev_net(skb->dev),
1550 skb->dev->ifindex);
1551 return;
1554 hdr = (u8 *)ndopts.nd_opts_rh;
1555 hdr += 8;
1556 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1557 return;
1559 icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1562 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1563 struct sk_buff *orig_skb,
1564 int rd_len)
1566 u8 *opt = skb_put(skb, rd_len);
1568 memset(opt, 0, 8);
1569 *(opt++) = ND_OPT_REDIRECT_HDR;
1570 *(opt++) = (rd_len >> 3);
1571 opt += 6;
1573 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1574 rd_len - 8);
1577 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1579 struct net_device *dev = skb->dev;
1580 struct net *net = dev_net(dev);
1581 struct sock *sk = net->ipv6.ndisc_sk;
1582 int optlen = 0;
1583 struct inet_peer *peer;
1584 struct sk_buff *buff;
1585 struct rd_msg *msg;
1586 struct in6_addr saddr_buf;
1587 struct rt6_info *rt;
1588 struct dst_entry *dst;
1589 struct flowi6 fl6;
1590 int rd_len;
1591 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1592 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1593 bool ret;
1595 if (netif_is_l3_master(skb->dev)) {
1596 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1597 if (!dev)
1598 return;
1601 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1602 ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1603 dev->name);
1604 return;
1607 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1608 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1609 ND_PRINTK(2, warn,
1610 "Redirect: target address is not link-local unicast\n");
1611 return;
1614 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1615 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1617 dst = ip6_route_output(net, NULL, &fl6);
1618 if (dst->error) {
1619 dst_release(dst);
1620 return;
1622 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1623 if (IS_ERR(dst))
1624 return;
1626 rt = (struct rt6_info *) dst;
1628 if (rt->rt6i_flags & RTF_GATEWAY) {
1629 ND_PRINTK(2, warn,
1630 "Redirect: destination is not a neighbour\n");
1631 goto release;
1633 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
1634 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1635 if (peer)
1636 inet_putpeer(peer);
1637 if (!ret)
1638 goto release;
1640 if (dev->addr_len) {
1641 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1642 if (!neigh) {
1643 ND_PRINTK(2, warn,
1644 "Redirect: no neigh for target address\n");
1645 goto release;
1648 read_lock_bh(&neigh->lock);
1649 if (neigh->nud_state & NUD_VALID) {
1650 memcpy(ha_buf, neigh->ha, dev->addr_len);
1651 read_unlock_bh(&neigh->lock);
1652 ha = ha_buf;
1653 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1654 ops_data_buf,
1655 &ops_data);
1656 } else
1657 read_unlock_bh(&neigh->lock);
1659 neigh_release(neigh);
1662 rd_len = min_t(unsigned int,
1663 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1664 skb->len + 8);
1665 rd_len &= ~0x7;
1666 optlen += rd_len;
1668 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1669 if (!buff)
1670 goto release;
1672 msg = skb_put(buff, sizeof(*msg));
1673 *msg = (struct rd_msg) {
1674 .icmph = {
1675 .icmp6_type = NDISC_REDIRECT,
1677 .target = *target,
1678 .dest = ipv6_hdr(skb)->daddr,
1682 * include target_address option
1685 if (ha)
1686 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1689 * build redirect option and copy skb over to the new packet.
1692 if (rd_len)
1693 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1695 skb_dst_set(buff, dst);
1696 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1697 return;
1699 release:
1700 dst_release(dst);
1703 static void pndisc_redo(struct sk_buff *skb)
1705 ndisc_recv_ns(skb);
1706 kfree_skb(skb);
1709 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1711 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1713 if (!idev)
1714 return true;
1715 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1716 idev->cnf.suppress_frag_ndisc) {
1717 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1718 return true;
1720 return false;
1723 int ndisc_rcv(struct sk_buff *skb)
1725 struct nd_msg *msg;
1727 if (ndisc_suppress_frag_ndisc(skb))
1728 return 0;
1730 if (skb_linearize(skb))
1731 return 0;
1733 msg = (struct nd_msg *)skb_transport_header(skb);
1735 __skb_push(skb, skb->data - skb_transport_header(skb));
1737 if (ipv6_hdr(skb)->hop_limit != 255) {
1738 ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1739 ipv6_hdr(skb)->hop_limit);
1740 return 0;
1743 if (msg->icmph.icmp6_code != 0) {
1744 ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1745 msg->icmph.icmp6_code);
1746 return 0;
1749 switch (msg->icmph.icmp6_type) {
1750 case NDISC_NEIGHBOUR_SOLICITATION:
1751 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1752 ndisc_recv_ns(skb);
1753 break;
1755 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1756 ndisc_recv_na(skb);
1757 break;
1759 case NDISC_ROUTER_SOLICITATION:
1760 ndisc_recv_rs(skb);
1761 break;
1763 case NDISC_ROUTER_ADVERTISEMENT:
1764 ndisc_router_discovery(skb);
1765 break;
1767 case NDISC_REDIRECT:
1768 ndisc_redirect_rcv(skb);
1769 break;
1772 return 0;
1775 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1777 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1778 struct netdev_notifier_change_info *change_info;
1779 struct net *net = dev_net(dev);
1780 struct inet6_dev *idev;
1782 switch (event) {
1783 case NETDEV_CHANGEADDR:
1784 neigh_changeaddr(&nd_tbl, dev);
1785 fib6_run_gc(0, net, false);
1786 fallthrough;
1787 case NETDEV_UP:
1788 idev = in6_dev_get(dev);
1789 if (!idev)
1790 break;
1791 if (idev->cnf.ndisc_notify ||
1792 net->ipv6.devconf_all->ndisc_notify)
1793 ndisc_send_unsol_na(dev);
1794 in6_dev_put(idev);
1795 break;
1796 case NETDEV_CHANGE:
1797 change_info = ptr;
1798 if (change_info->flags_changed & IFF_NOARP)
1799 neigh_changeaddr(&nd_tbl, dev);
1800 if (!netif_carrier_ok(dev))
1801 neigh_carrier_down(&nd_tbl, dev);
1802 break;
1803 case NETDEV_DOWN:
1804 neigh_ifdown(&nd_tbl, dev);
1805 fib6_run_gc(0, net, false);
1806 break;
1807 case NETDEV_NOTIFY_PEERS:
1808 ndisc_send_unsol_na(dev);
1809 break;
1810 default:
1811 break;
1814 return NOTIFY_DONE;
1817 static struct notifier_block ndisc_netdev_notifier = {
1818 .notifier_call = ndisc_netdev_event,
1819 .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1822 #ifdef CONFIG_SYSCTL
1823 static void ndisc_warn_deprecated_sysctl(struct ctl_table *ctl,
1824 const char *func, const char *dev_name)
1826 static char warncomm[TASK_COMM_LEN];
1827 static int warned;
1828 if (strcmp(warncomm, current->comm) && warned < 5) {
1829 strcpy(warncomm, current->comm);
1830 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1831 warncomm, func,
1832 dev_name, ctl->procname,
1833 dev_name, ctl->procname);
1834 warned++;
1838 int ndisc_ifinfo_sysctl_change(struct ctl_table *ctl, int write, void __user *buffer, size_t *lenp, loff_t *ppos)
1840 struct net_device *dev = ctl->extra1;
1841 struct inet6_dev *idev;
1842 int ret;
1844 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1845 (strcmp(ctl->procname, "base_reachable_time") == 0))
1846 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1848 if (strcmp(ctl->procname, "retrans_time") == 0)
1849 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1851 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1852 ret = neigh_proc_dointvec_jiffies(ctl, write,
1853 buffer, lenp, ppos);
1855 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1856 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1857 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1858 buffer, lenp, ppos);
1859 else
1860 ret = -1;
1862 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1863 if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1864 idev->nd_parms->reachable_time =
1865 neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1866 idev->tstamp = jiffies;
1867 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1868 in6_dev_put(idev);
1870 return ret;
1874 #endif
1876 static int __net_init ndisc_net_init(struct net *net)
1878 struct ipv6_pinfo *np;
1879 struct sock *sk;
1880 int err;
1882 err = inet_ctl_sock_create(&sk, PF_INET6,
1883 SOCK_RAW, IPPROTO_ICMPV6, net);
1884 if (err < 0) {
1885 ND_PRINTK(0, err,
1886 "NDISC: Failed to initialize the control socket (err %d)\n",
1887 err);
1888 return err;
1891 net->ipv6.ndisc_sk = sk;
1893 np = inet6_sk(sk);
1894 np->hop_limit = 255;
1895 /* Do not loopback ndisc messages */
1896 np->mc_loop = 0;
1898 return 0;
1901 static void __net_exit ndisc_net_exit(struct net *net)
1903 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1906 static struct pernet_operations ndisc_net_ops = {
1907 .init = ndisc_net_init,
1908 .exit = ndisc_net_exit,
1911 int __init ndisc_init(void)
1913 int err;
1915 err = register_pernet_subsys(&ndisc_net_ops);
1916 if (err)
1917 return err;
1919 * Initialize the neighbour table
1921 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
1923 #ifdef CONFIG_SYSCTL
1924 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
1925 ndisc_ifinfo_sysctl_change);
1926 if (err)
1927 goto out_unregister_pernet;
1928 out:
1929 #endif
1930 return err;
1932 #ifdef CONFIG_SYSCTL
1933 out_unregister_pernet:
1934 unregister_pernet_subsys(&ndisc_net_ops);
1935 goto out;
1936 #endif
1939 int __init ndisc_late_init(void)
1941 return register_netdevice_notifier(&ndisc_netdev_notifier);
1944 void ndisc_late_cleanup(void)
1946 unregister_netdevice_notifier(&ndisc_netdev_notifier);
1949 void ndisc_cleanup(void)
1951 #ifdef CONFIG_SYSCTL
1952 neigh_sysctl_unregister(&nd_tbl.parms);
1953 #endif
1954 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
1955 unregister_pernet_subsys(&ndisc_net_ops);