projects/linux·blob master · refs

2050 lines · 52.9 KiB

1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Neighbour Discovery for IPv6
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <[email protected]>
8 * Mike Shaver <[email protected]>
9 */
10
11/*
12 * Changes:
13 *
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
24 */
25
26#define pr_fmt(fmt) "ICMPv6: " fmt
27
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
43
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>
50
51#include <net/sock.h>
52#include <net/snmp.h>
53
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>
60
61#include <net/netlink.h>
62#include <linux/rtnetlink.h>
63
64#include <net/flow.h>
65#include <net/ip6_checksum.h>
66#include <net/inet_common.h>
67#include <linux/proc_fs.h>
68
69#include <linux/netfilter.h>
70#include <linux/netfilter_ipv6.h>
71
72static u32 ndisc_hash(const void *pkey,
73 const struct net_device *dev,
74 __u32 *hash_rnd);
75static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey);
76static bool ndisc_allow_add(const struct net_device *dev,
77 struct netlink_ext_ack *extack);
78static int ndisc_constructor(struct neighbour *neigh);
79static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb);
80static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb);
81static int pndisc_constructor(struct pneigh_entry *n);
82static void pndisc_destructor(struct pneigh_entry *n);
83static void pndisc_redo(struct sk_buff *skb);
84static int ndisc_is_multicast(const void *pkey);
85
86static 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,
92};
93
94static 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,
100};
101
102
103static const struct neigh_ops ndisc_direct_ops = {
104 .family = AF_INET6,
105 .output = neigh_direct_output,
106 .connected_output = neigh_direct_output,
107};
108
109struct 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_INTERVAL_PROBE_TIME_MS] = 5 * HZ,
132 [NEIGH_VAR_GC_STALETIME] = 60 * HZ,
133 [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_DEFAULT,
134 [NEIGH_VAR_PROXY_QLEN] = 64,
135 [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ,
136 [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10,
137 },
138 },
139 .gc_interval = 30 * HZ,
140 .gc_thresh1 = 128,
141 .gc_thresh2 = 512,
142 .gc_thresh3 = 1024,
143};
144EXPORT_SYMBOL_GPL(nd_tbl);
145
146void __ndisc_fill_addr_option(struct sk_buff *skb, int type, const void *data,
147 int data_len, int pad)
148{
149 int space = __ndisc_opt_addr_space(data_len, pad);
150 u8 *opt = skb_put(skb, space);
151
152 opt[0] = type;
153 opt[1] = space>>3;
154
155 memset(opt + 2, 0, pad);
156 opt += pad;
157 space -= pad;
158
159 memcpy(opt+2, data, data_len);
160 data_len += 2;
161 opt += data_len;
162 space -= data_len;
163 if (space > 0)
164 memset(opt, 0, space);
165}
166EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option);
167
168static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type,
169 const void *data, u8 icmp6_type)
170{
171 __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len,
172 ndisc_addr_option_pad(skb->dev->type));
173 ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type);
174}
175
176static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb,
177 void *ha,
178 const u8 *ops_data)
179{
180 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT);
181 ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data);
182}
183
184static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
185 struct nd_opt_hdr *end)
186{
187 int type;
188 if (!cur || !end || cur >= end)
189 return NULL;
190 type = cur->nd_opt_type;
191 do {
192 cur = ((void *)cur) + (cur->nd_opt_len << 3);
193 } while (cur < end && cur->nd_opt_type != type);
194 return cur <= end && cur->nd_opt_type == type ? cur : NULL;
195}
196
197static inline int ndisc_is_useropt(const struct net_device *dev,
198 struct nd_opt_hdr *opt)
199{
200 return opt->nd_opt_type == ND_OPT_PREFIX_INFO ||
201 opt->nd_opt_type == ND_OPT_RDNSS ||
202 opt->nd_opt_type == ND_OPT_DNSSL ||
203 opt->nd_opt_type == ND_OPT_6CO ||
204 opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
205 opt->nd_opt_type == ND_OPT_PREF64;
206}
207
208static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
209 struct nd_opt_hdr *cur,
210 struct nd_opt_hdr *end)
211{
212 if (!cur || !end || cur >= end)
213 return NULL;
214 do {
215 cur = ((void *)cur) + (cur->nd_opt_len << 3);
216 } while (cur < end && !ndisc_is_useropt(dev, cur));
217 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
218}
219
220struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
221 u8 *opt, int opt_len,
222 struct ndisc_options *ndopts)
223{
224 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
225
226 if (!nd_opt || opt_len < 0 || !ndopts)
227 return NULL;
228 memset(ndopts, 0, sizeof(*ndopts));
229 while (opt_len) {
230 bool unknown = false;
231 int l;
232 if (opt_len < sizeof(struct nd_opt_hdr))
233 return NULL;
234 l = nd_opt->nd_opt_len << 3;
235 if (opt_len < l || l == 0)
236 return NULL;
237 if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
238 goto next_opt;
239 switch (nd_opt->nd_opt_type) {
240 case ND_OPT_SOURCE_LL_ADDR:
241 case ND_OPT_TARGET_LL_ADDR:
242 case ND_OPT_MTU:
243 case ND_OPT_NONCE:
244 case ND_OPT_REDIRECT_HDR:
245 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
246 net_dbg_ratelimited("%s: duplicated ND6 option found: type=%d\n",
247 __func__, nd_opt->nd_opt_type);
248 } else {
249 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
250 }
251 break;
252 case ND_OPT_PREFIX_INFO:
253 ndopts->nd_opts_pi_end = nd_opt;
254 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
255 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
256 break;
257#ifdef CONFIG_IPV6_ROUTE_INFO
258 case ND_OPT_ROUTE_INFO:
259 ndopts->nd_opts_ri_end = nd_opt;
260 if (!ndopts->nd_opts_ri)
261 ndopts->nd_opts_ri = nd_opt;
262 break;
263#endif
264 default:
265 unknown = true;
266 }
267 if (ndisc_is_useropt(dev, nd_opt)) {
268 ndopts->nd_useropts_end = nd_opt;
269 if (!ndopts->nd_useropts)
270 ndopts->nd_useropts = nd_opt;
271 } else if (unknown) {
272 /*
273 * Unknown options must be silently ignored,
274 * to accommodate future extension to the
275 * protocol.
276 */
277 net_dbg_ratelimited("%s: ignored unsupported option; type=%d, len=%d\n",
278 __func__, nd_opt->nd_opt_type, nd_opt->nd_opt_len);
279 }
280next_opt:
281 opt_len -= l;
282 nd_opt = ((void *)nd_opt) + l;
283 }
284 return ndopts;
285}
286EXPORT_SYMBOL_GPL(ndisc_parse_options);
287
288int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
289{
290 switch (dev->type) {
291 case ARPHRD_ETHER:
292 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */
293 case ARPHRD_FDDI:
294 ipv6_eth_mc_map(addr, buf);
295 return 0;
296 case ARPHRD_ARCNET:
297 ipv6_arcnet_mc_map(addr, buf);
298 return 0;
299 case ARPHRD_INFINIBAND:
300 ipv6_ib_mc_map(addr, dev->broadcast, buf);
301 return 0;
302 case ARPHRD_IPGRE:
303 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
304 default:
305 if (dir) {
306 memcpy(buf, dev->broadcast, dev->addr_len);
307 return 0;
308 }
309 }
310 return -EINVAL;
311}
312EXPORT_SYMBOL(ndisc_mc_map);
313
314static u32 ndisc_hash(const void *pkey,
315 const struct net_device *dev,
316 __u32 *hash_rnd)
317{
318 return ndisc_hashfn(pkey, dev, hash_rnd);
319}
320
321static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
322{
323 return neigh_key_eq128(n, pkey);
324}
325
326static int ndisc_constructor(struct neighbour *neigh)
327{
328 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
329 struct net_device *dev = neigh->dev;
330 struct inet6_dev *in6_dev;
331 struct neigh_parms *parms;
332 bool is_multicast = ipv6_addr_is_multicast(addr);
333
334 in6_dev = in6_dev_get(dev);
335 if (!in6_dev) {
336 return -EINVAL;
337 }
338
339 parms = in6_dev->nd_parms;
340 __neigh_parms_put(neigh->parms);
341 neigh->parms = neigh_parms_clone(parms);
342
343 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
344 if (!dev->header_ops) {
345 neigh->nud_state = NUD_NOARP;
346 neigh->ops = &ndisc_direct_ops;
347 neigh->output = neigh_direct_output;
348 } else {
349 if (is_multicast) {
350 neigh->nud_state = NUD_NOARP;
351 ndisc_mc_map(addr, neigh->ha, dev, 1);
352 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
353 neigh->nud_state = NUD_NOARP;
354 memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
355 if (dev->flags&IFF_LOOPBACK)
356 neigh->type = RTN_LOCAL;
357 } else if (dev->flags&IFF_POINTOPOINT) {
358 neigh->nud_state = NUD_NOARP;
359 memcpy(neigh->ha, dev->broadcast, dev->addr_len);
360 }
361 if (dev->header_ops->cache)
362 neigh->ops = &ndisc_hh_ops;
363 else
364 neigh->ops = &ndisc_generic_ops;
365 if (neigh->nud_state&NUD_VALID)
366 neigh->output = neigh->ops->connected_output;
367 else
368 neigh->output = neigh->ops->output;
369 }
370 in6_dev_put(in6_dev);
371 return 0;
372}
373
374static int pndisc_constructor(struct pneigh_entry *n)
375{
376 struct in6_addr *addr = (struct in6_addr *)&n->key;
377 struct net_device *dev = n->dev;
378 struct in6_addr maddr;
379
380 if (!dev)
381 return -EINVAL;
382
383 addrconf_addr_solict_mult(addr, &maddr);
384 return ipv6_dev_mc_inc(dev, &maddr);
385}
386
387static void pndisc_destructor(struct pneigh_entry *n)
388{
389 struct in6_addr *addr = (struct in6_addr *)&n->key;
390 struct net_device *dev = n->dev;
391 struct in6_addr maddr;
392
393 if (!dev)
394 return;
395
396 addrconf_addr_solict_mult(addr, &maddr);
397 ipv6_dev_mc_dec(dev, &maddr);
398}
399
400/* called with rtnl held */
401static bool ndisc_allow_add(const struct net_device *dev,
402 struct netlink_ext_ack *extack)
403{
404 struct inet6_dev *idev = __in6_dev_get(dev);
405
406 if (!idev || idev->cnf.disable_ipv6) {
407 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
408 return false;
409 }
410
411 return true;
412}
413
414static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
415 int len)
416{
417 int hlen = LL_RESERVED_SPACE(dev);
418 int tlen = dev->needed_tailroom;
419 struct sk_buff *skb;
420
421 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
422 if (!skb)
423 return NULL;
424
425 skb->protocol = htons(ETH_P_IPV6);
426 skb->dev = dev;
427
428 skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
429 skb_reset_transport_header(skb);
430
431 /* Manually assign socket ownership as we avoid calling
432 * sock_alloc_send_pskb() to bypass wmem buffer limits
433 */
434 rcu_read_lock();
435 skb_set_owner_w(skb, dev_net_rcu(dev)->ipv6.ndisc_sk);
436 rcu_read_unlock();
437
438 return skb;
439}
440
441static 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)
445{
446 struct ipv6hdr *hdr;
447 struct inet6_dev *idev;
448 unsigned tclass;
449
450 rcu_read_lock();
451 idev = __in6_dev_get(skb->dev);
452 tclass = idev ? READ_ONCE(idev->cnf.ndisc_tclass) : 0;
453 rcu_read_unlock();
454
455 skb_push(skb, sizeof(*hdr));
456 skb_reset_network_header(skb);
457 hdr = ipv6_hdr(skb);
458
459 ip6_flow_hdr(hdr, tclass, 0);
460
461 hdr->payload_len = htons(len);
462 hdr->nexthdr = IPPROTO_ICMPV6;
463 hdr->hop_limit = hop_limit;
464
465 hdr->saddr = *saddr;
466 hdr->daddr = *daddr;
467}
468
469void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
470 const struct in6_addr *saddr)
471{
472 struct icmp6hdr *icmp6h = icmp6_hdr(skb);
473 struct dst_entry *dst = skb_dst(skb);
474 struct net_device *dev;
475 struct inet6_dev *idev;
476 struct net *net;
477 struct sock *sk;
478 int err;
479 u8 type;
480
481 type = icmp6h->icmp6_type;
482
483 rcu_read_lock();
484
485 net = dev_net_rcu(skb->dev);
486 sk = net->ipv6.ndisc_sk;
487 if (!dst) {
488 struct flowi6 fl6;
489 int oif = skb->dev->ifindex;
490
491 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
492 dst = icmp6_dst_alloc(skb->dev, &fl6);
493 if (IS_ERR(dst)) {
494 rcu_read_unlock();
495 kfree_skb(skb);
496 return;
497 }
498
499 skb_dst_set(skb, dst);
500 }
501
502 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
503 IPPROTO_ICMPV6,
504 csum_partial(icmp6h,
505 skb->len, 0));
506
507 ip6_nd_hdr(skb, saddr, daddr, READ_ONCE(inet6_sk(sk)->hop_limit), skb->len);
508
509 dev = dst_dev_rcu(dst);
510 idev = __in6_dev_get(dev);
511 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
512
513 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
514 net, sk, skb, NULL, dev,
515 dst_output);
516 if (!err) {
517 ICMP6MSGOUT_INC_STATS(net, idev, type);
518 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
519 }
520
521 rcu_read_unlock();
522}
523EXPORT_SYMBOL(ndisc_send_skb);
524
525void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
526 const struct in6_addr *solicited_addr,
527 bool router, bool solicited, bool override, bool inc_opt)
528{
529 struct sk_buff *skb;
530 struct in6_addr tmpaddr;
531 struct inet6_ifaddr *ifp;
532 const struct in6_addr *src_addr;
533 struct nd_msg *msg;
534 int optlen = 0;
535
536 /* for anycast or proxy, solicited_addr != src_addr */
537 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
538 if (ifp) {
539 src_addr = solicited_addr;
540 if (ifp->flags & IFA_F_OPTIMISTIC)
541 override = false;
542 inc_opt |= READ_ONCE(ifp->idev->cnf.force_tllao);
543 in6_ifa_put(ifp);
544 } else {
545 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
546 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
547 &tmpaddr))
548 return;
549 src_addr = &tmpaddr;
550 }
551
552 if (!dev->addr_len)
553 inc_opt = false;
554 if (inc_opt)
555 optlen += ndisc_opt_addr_space(dev,
556 NDISC_NEIGHBOUR_ADVERTISEMENT);
557
558 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
559 if (!skb)
560 return;
561
562 msg = skb_put(skb, sizeof(*msg));
563 *msg = (struct nd_msg) {
564 .icmph = {
565 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
566 .icmp6_router = router,
567 .icmp6_solicited = solicited,
568 .icmp6_override = override,
569 },
570 .target = *solicited_addr,
571 };
572
573 if (inc_opt)
574 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
575 dev->dev_addr,
576 NDISC_NEIGHBOUR_ADVERTISEMENT);
577
578 ndisc_send_skb(skb, daddr, src_addr);
579}
580EXPORT_SYMBOL_GPL(ndisc_send_na);
581
582static void ndisc_send_unsol_na(struct net_device *dev)
583{
584 struct inet6_dev *idev;
585 struct inet6_ifaddr *ifa;
586
587 idev = in6_dev_get(dev);
588 if (!idev)
589 return;
590
591 read_lock_bh(&idev->lock);
592 list_for_each_entry(ifa, &idev->addr_list, if_list) {
593 /* skip tentative addresses until dad completes */
594 if (ifa->flags & IFA_F_TENTATIVE &&
595 !(ifa->flags & IFA_F_OPTIMISTIC))
596 continue;
597
598 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
599 /*router=*/ !!idev->cnf.forwarding,
600 /*solicited=*/ false, /*override=*/ true,
601 /*inc_opt=*/ true);
602 }
603 read_unlock_bh(&idev->lock);
604
605 in6_dev_put(idev);
606}
607
608struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
609 const struct in6_addr *saddr, u64 nonce)
610{
611 int inc_opt = dev->addr_len;
612 struct sk_buff *skb;
613 struct nd_msg *msg;
614 int optlen = 0;
615
616 if (!saddr)
617 return NULL;
618
619 if (ipv6_addr_any(saddr))
620 inc_opt = false;
621 if (inc_opt)
622 optlen += ndisc_opt_addr_space(dev,
623 NDISC_NEIGHBOUR_SOLICITATION);
624 if (nonce != 0)
625 optlen += 8;
626
627 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
628 if (!skb)
629 return NULL;
630
631 msg = skb_put(skb, sizeof(*msg));
632 *msg = (struct nd_msg) {
633 .icmph = {
634 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
635 },
636 .target = *solicit,
637 };
638
639 if (inc_opt)
640 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
641 dev->dev_addr,
642 NDISC_NEIGHBOUR_SOLICITATION);
643 if (nonce != 0) {
644 u8 *opt = skb_put(skb, 8);
645
646 opt[0] = ND_OPT_NONCE;
647 opt[1] = 8 >> 3;
648 memcpy(opt + 2, &nonce, 6);
649 }
650
651 return skb;
652}
653EXPORT_SYMBOL(ndisc_ns_create);
654
655void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
656 const struct in6_addr *daddr, const struct in6_addr *saddr,
657 u64 nonce)
658{
659 struct in6_addr addr_buf;
660 struct sk_buff *skb;
661
662 if (!saddr) {
663 if (ipv6_get_lladdr(dev, &addr_buf,
664 (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
665 return;
666 saddr = &addr_buf;
667 }
668
669 skb = ndisc_ns_create(dev, solicit, saddr, nonce);
670
671 if (skb)
672 ndisc_send_skb(skb, daddr, saddr);
673}
674
675void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
676 const struct in6_addr *daddr)
677{
678 struct sk_buff *skb;
679 struct rs_msg *msg;
680 int send_sllao = dev->addr_len;
681 int optlen = 0;
682
683#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
684 /*
685 * According to section 2.2 of RFC 4429, we must not
686 * send router solicitations with a sllao from
687 * optimistic addresses, but we may send the solicitation
688 * if we don't include the sllao. So here we check
689 * if our address is optimistic, and if so, we
690 * suppress the inclusion of the sllao.
691 */
692 if (send_sllao) {
693 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
694 dev, 1);
695 if (ifp) {
696 if (ifp->flags & IFA_F_OPTIMISTIC) {
697 send_sllao = 0;
698 }
699 in6_ifa_put(ifp);
700 } else {
701 send_sllao = 0;
702 }
703 }
704#endif
705 if (send_sllao)
706 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
707
708 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
709 if (!skb)
710 return;
711
712 msg = skb_put(skb, sizeof(*msg));
713 *msg = (struct rs_msg) {
714 .icmph = {
715 .icmp6_type = NDISC_ROUTER_SOLICITATION,
716 },
717 };
718
719 if (send_sllao)
720 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
721 dev->dev_addr,
722 NDISC_ROUTER_SOLICITATION);
723
724 ndisc_send_skb(skb, daddr, saddr);
725}
726
727
728static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
729{
730 /*
731 * "The sender MUST return an ICMP
732 * destination unreachable"
733 */
734 dst_link_failure(skb);
735 kfree_skb(skb);
736}
737
738/* Called with locked neigh: either read or both */
739
740static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
741{
742 struct in6_addr *saddr = NULL;
743 struct in6_addr mcaddr;
744 struct net_device *dev = neigh->dev;
745 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
746 int probes = atomic_read(&neigh->probes);
747
748 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
749 dev, false, 1,
750 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
751 saddr = &ipv6_hdr(skb)->saddr;
752 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
753 if (probes < 0) {
754 if (!(READ_ONCE(neigh->nud_state) & NUD_VALID)) {
755 net_dbg_ratelimited("%s: trying to ucast probe in NUD_INVALID: %pI6\n",
756 __func__, target);
757 }
758 ndisc_send_ns(dev, target, target, saddr, 0);
759 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
760 neigh_app_ns(neigh);
761 } else {
762 addrconf_addr_solict_mult(target, &mcaddr);
763 ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
764 }
765}
766
767static int pndisc_is_router(const void *pkey,
768 struct net_device *dev)
769{
770 struct pneigh_entry *n;
771 int ret = -1;
772
773 n = pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
774 if (n)
775 ret = !!(READ_ONCE(n->flags) & NTF_ROUTER);
776
777 return ret;
778}
779
780void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
781 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
782 struct ndisc_options *ndopts)
783{
784 neigh_update(neigh, lladdr, new, flags, 0);
785 /* report ndisc ops about neighbour update */
786 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
787}
788
789static enum skb_drop_reason ndisc_recv_ns(struct sk_buff *skb)
790{
791 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
792 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
793 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
794 u8 *lladdr = NULL;
795 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
796 offsetof(struct nd_msg, opt));
797 struct ndisc_options ndopts;
798 struct net_device *dev = skb->dev;
799 struct inet6_ifaddr *ifp;
800 struct inet6_dev *idev = NULL;
801 struct neighbour *neigh;
802 int dad = ipv6_addr_any(saddr);
803 int is_router = -1;
804 SKB_DR(reason);
805 u64 nonce = 0;
806 bool inc;
807
808 if (skb->len < sizeof(struct nd_msg))
809 return SKB_DROP_REASON_PKT_TOO_SMALL;
810
811 if (ipv6_addr_is_multicast(&msg->target)) {
812 net_dbg_ratelimited("NS: multicast target address\n");
813 return reason;
814 }
815
816 /*
817 * RFC2461 7.1.1:
818 * DAD has to be destined for solicited node multicast address.
819 */
820 if (dad && !ipv6_addr_is_solict_mult(daddr)) {
821 net_dbg_ratelimited("NS: bad DAD packet (wrong destination)\n");
822 return reason;
823 }
824
825 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
826 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
827
828 if (ndopts.nd_opts_src_lladdr) {
829 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
830 if (!lladdr) {
831 net_dbg_ratelimited("NS: invalid link-layer address length\n");
832 return reason;
833 }
834
835 /* RFC2461 7.1.1:
836 * If the IP source address is the unspecified address,
837 * there MUST NOT be source link-layer address option
838 * in the message.
839 */
840 if (dad) {
841 net_dbg_ratelimited("NS: bad DAD packet (link-layer address option)\n");
842 return reason;
843 }
844 }
845 if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
846 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
847
848 inc = ipv6_addr_is_multicast(daddr);
849
850 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
851 if (ifp) {
852have_ifp:
853 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
854 if (dad) {
855 if (nonce != 0 && ifp->dad_nonce == nonce) {
856 u8 *np = (u8 *)&nonce;
857 /* Matching nonce if looped back */
858 net_dbg_ratelimited("%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
859 ifp->idev->dev->name, &ifp->addr, np);
860 goto out;
861 }
862 /*
863 * We are colliding with another node
864 * who is doing DAD
865 * so fail our DAD process
866 */
867 addrconf_dad_failure(skb, ifp);
868 return reason;
869 } else {
870 /*
871 * This is not a dad solicitation.
872 * If we are an optimistic node,
873 * we should respond.
874 * Otherwise, we should ignore it.
875 */
876 if (!(ifp->flags & IFA_F_OPTIMISTIC))
877 goto out;
878 }
879 }
880
881 idev = ifp->idev;
882 } else {
883 struct net *net = dev_net(dev);
884
885 /* perhaps an address on the master device */
886 if (netif_is_l3_slave(dev)) {
887 struct net_device *mdev;
888
889 mdev = netdev_master_upper_dev_get_rcu(dev);
890 if (mdev) {
891 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
892 if (ifp)
893 goto have_ifp;
894 }
895 }
896
897 idev = in6_dev_get(dev);
898 if (!idev) {
899 /* XXX: count this drop? */
900 return reason;
901 }
902
903 if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
904 (READ_ONCE(idev->cnf.forwarding) &&
905 (READ_ONCE(net->ipv6.devconf_all->proxy_ndp) ||
906 READ_ONCE(idev->cnf.proxy_ndp)) &&
907 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
908 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
909 skb->pkt_type != PACKET_HOST &&
910 inc &&
911 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
912 /*
913 * for anycast or proxy,
914 * sender should delay its response
915 * by a random time between 0 and
916 * MAX_ANYCAST_DELAY_TIME seconds.
917 * (RFC2461) -- yoshfuji
918 */
919 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
920 if (n)
921 pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
922 goto out;
923 }
924 } else {
925 SKB_DR_SET(reason, IPV6_NDISC_NS_OTHERHOST);
926 goto out;
927 }
928 }
929
930 if (is_router < 0)
931 is_router = READ_ONCE(idev->cnf.forwarding);
932
933 if (dad) {
934 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
935 !!is_router, false, (ifp != NULL), true);
936 goto out;
937 }
938
939 if (inc)
940 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
941 else
942 NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
943
944 /*
945 * update / create cache entry
946 * for the source address
947 */
948 neigh = __neigh_lookup(&nd_tbl, saddr, dev,
949 !inc || lladdr || !dev->addr_len);
950 if (neigh)
951 ndisc_update(dev, neigh, lladdr, NUD_STALE,
952 NEIGH_UPDATE_F_WEAK_OVERRIDE|
953 NEIGH_UPDATE_F_OVERRIDE,
954 NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
955 if (neigh || !dev->header_ops) {
956 ndisc_send_na(dev, saddr, &msg->target, !!is_router,
957 true, (ifp != NULL && inc), inc);
958 if (neigh)
959 neigh_release(neigh);
960 reason = SKB_CONSUMED;
961 }
962
963out:
964 if (ifp)
965 in6_ifa_put(ifp);
966 else
967 in6_dev_put(idev);
968 return reason;
969}
970
971static int accept_untracked_na(struct inet6_dev *idev, struct in6_addr *saddr)
972{
973 switch (READ_ONCE(idev->cnf.accept_untracked_na)) {
974 case 0: /* Don't accept untracked na (absent in neighbor cache) */
975 return 0;
976 case 1: /* Create new entries from na if currently untracked */
977 return 1;
978 case 2: /* Create new entries from untracked na only if saddr is in the
979 * same subnet as an address configured on the interface that
980 * received the na
981 */
982 return !!ipv6_chk_prefix(saddr, idev->dev);
983 default:
984 return 0;
985 }
986}
987
988static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)
989{
990 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
991 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
992 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
993 u8 *lladdr = NULL;
994 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
995 offsetof(struct nd_msg, opt));
996 struct ndisc_options ndopts;
997 struct net_device *dev = skb->dev;
998 struct inet6_dev *idev = __in6_dev_get(dev);
999 struct inet6_ifaddr *ifp;
1000 struct neighbour *neigh;
1001 SKB_DR(reason);
1002 u8 new_state;
1003
1004 if (skb->len < sizeof(struct nd_msg))
1005 return SKB_DROP_REASON_PKT_TOO_SMALL;
1006
1007 if (ipv6_addr_is_multicast(&msg->target)) {
1008 net_dbg_ratelimited("NA: target address is multicast\n");
1009 return reason;
1010 }
1011
1012 if (ipv6_addr_is_multicast(daddr) &&
1013 msg->icmph.icmp6_solicited) {
1014 net_dbg_ratelimited("NA: solicited NA is multicasted\n");
1015 return reason;
1016 }
1017
1018 /* For some 802.11 wireless deployments (and possibly other networks),
1019 * there will be a NA proxy and unsolicitd packets are attacks
1020 * and thus should not be accepted.
1021 * drop_unsolicited_na takes precedence over accept_untracked_na
1022 */
1023 if (!msg->icmph.icmp6_solicited && idev &&
1024 READ_ONCE(idev->cnf.drop_unsolicited_na))
1025 return reason;
1026
1027 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
1028 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1029
1030 if (ndopts.nd_opts_tgt_lladdr) {
1031 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1032 if (!lladdr) {
1033 net_dbg_ratelimited("NA: invalid link-layer address length\n");
1034 return reason;
1035 }
1036 }
1037 ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1038 if (ifp) {
1039 if (skb->pkt_type != PACKET_LOOPBACK
1040 && (ifp->flags & IFA_F_TENTATIVE)) {
1041 addrconf_dad_failure(skb, ifp);
1042 return reason;
1043 }
1044 /* What should we make now? The advertisement
1045 is invalid, but ndisc specs say nothing
1046 about it. It could be misconfiguration, or
1047 an smart proxy agent tries to help us :-)
1048
1049 We should not print the error if NA has been
1050 received from loopback - it is just our own
1051 unsolicited advertisement.
1052 */
1053 if (skb->pkt_type != PACKET_LOOPBACK)
1054 net_warn_ratelimited("NA: %pM advertised our address %pI6c on %s!\n",
1055 eth_hdr(skb)->h_source, &ifp->addr,
1056 ifp->idev->dev->name);
1057 in6_ifa_put(ifp);
1058 return reason;
1059 }
1060
1061 neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1062
1063 /* RFC 9131 updates original Neighbour Discovery RFC 4861.
1064 * NAs with Target LL Address option without a corresponding
1065 * entry in the neighbour cache can now create a STALE neighbour
1066 * cache entry on routers.
1067 *
1068 * entry accept fwding solicited behaviour
1069 * ------- ------ ------ --------- ----------------------
1070 * present X X 0 Set state to STALE
1071 * present X X 1 Set state to REACHABLE
1072 * absent 0 X X Do nothing
1073 * absent 1 0 X Do nothing
1074 * absent 1 1 X Add a new STALE entry
1075 *
1076 * Note that we don't do a (daddr == all-routers-mcast) check.
1077 */
1078 new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1079 if (!neigh && lladdr && idev && READ_ONCE(idev->cnf.forwarding)) {
1080 if (accept_untracked_na(idev, saddr)) {
1081 neigh = neigh_create(&nd_tbl, &msg->target, dev);
1082 new_state = NUD_STALE;
1083 }
1084 }
1085
1086 if (neigh && !IS_ERR(neigh)) {
1087 u8 old_flags = neigh->flags;
1088 struct net *net = dev_net(dev);
1089
1090 if (READ_ONCE(neigh->nud_state) & NUD_FAILED)
1091 goto out;
1092
1093 /*
1094 * Don't update the neighbor cache entry on a proxy NA from
1095 * ourselves because either the proxied node is off link or it
1096 * has already sent a NA to us.
1097 */
1098 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1099 READ_ONCE(net->ipv6.devconf_all->forwarding) &&
1100 READ_ONCE(net->ipv6.devconf_all->proxy_ndp) &&
1101 pneigh_lookup(&nd_tbl, net, &msg->target, dev)) {
1102 /* XXX: idev->cnf.proxy_ndp */
1103 goto out;
1104 }
1105
1106 ndisc_update(dev, neigh, lladdr,
1107 new_state,
1108 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1109 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1110 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1111 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1112 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1113
1114 if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1115 /*
1116 * Change: router to host
1117 */
1118 rt6_clean_tohost(dev_net(dev), saddr);
1119 }
1120 reason = SKB_CONSUMED;
1121out:
1122 neigh_release(neigh);
1123 }
1124 return reason;
1125}
1126
1127static enum skb_drop_reason ndisc_recv_rs(struct sk_buff *skb)
1128{
1129 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1130 unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1131 struct neighbour *neigh;
1132 struct inet6_dev *idev;
1133 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1134 struct ndisc_options ndopts;
1135 u8 *lladdr = NULL;
1136 SKB_DR(reason);
1137
1138 if (skb->len < sizeof(*rs_msg))
1139 return SKB_DROP_REASON_PKT_TOO_SMALL;
1140
1141 idev = __in6_dev_get(skb->dev);
1142 if (!idev) {
1143 net_err_ratelimited("RS: can't find in6 device\n");
1144 return reason;
1145 }
1146
1147 /* Don't accept RS if we're not in router mode */
1148 if (!READ_ONCE(idev->cnf.forwarding))
1149 goto out;
1150
1151 /*
1152 * Don't update NCE if src = ::;
1153 * this implies that the source node has no ip address assigned yet.
1154 */
1155 if (ipv6_addr_any(saddr))
1156 goto out;
1157
1158 /* Parse ND options */
1159 if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts))
1160 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1161
1162 if (ndopts.nd_opts_src_lladdr) {
1163 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1164 skb->dev);
1165 if (!lladdr)
1166 goto out;
1167 }
1168
1169 neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1170 if (neigh) {
1171 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1172 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1173 NEIGH_UPDATE_F_OVERRIDE|
1174 NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1175 NDISC_ROUTER_SOLICITATION, &ndopts);
1176 neigh_release(neigh);
1177 reason = SKB_CONSUMED;
1178 }
1179out:
1180 return reason;
1181}
1182
1183static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1184{
1185 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1186 struct sk_buff *skb;
1187 struct nlmsghdr *nlh;
1188 struct nduseroptmsg *ndmsg;
1189 struct net *net = dev_net(ra->dev);
1190 int err;
1191 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1192 + (opt->nd_opt_len << 3));
1193 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1194
1195 skb = nlmsg_new(msg_size, GFP_ATOMIC);
1196 if (!skb) {
1197 err = -ENOBUFS;
1198 goto errout;
1199 }
1200
1201 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1202 if (!nlh) {
1203 goto nla_put_failure;
1204 }
1205
1206 ndmsg = nlmsg_data(nlh);
1207 ndmsg->nduseropt_family = AF_INET6;
1208 ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1209 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1210 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1211 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1212 ndmsg->nduseropt_pad1 = 0;
1213 ndmsg->nduseropt_pad2 = 0;
1214 ndmsg->nduseropt_pad3 = 0;
1215
1216 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1217
1218 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1219 goto nla_put_failure;
1220 nlmsg_end(skb, nlh);
1221
1222 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1223 return;
1224
1225nla_put_failure:
1226 nlmsg_free(skb);
1227 err = -EMSGSIZE;
1228errout:
1229 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1230}
1231
1232static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb)
1233{
1234 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1235 bool send_ifinfo_notify = false;
1236 struct neighbour *neigh = NULL;
1237 struct ndisc_options ndopts;
1238 struct fib6_info *rt = NULL;
1239 struct inet6_dev *in6_dev;
1240 struct fib6_table *table;
1241 u32 defrtr_usr_metric;
1242 unsigned int pref = 0;
1243 __u32 old_if_flags;
1244 struct net *net;
1245 SKB_DR(reason);
1246 int lifetime;
1247 int optlen;
1248
1249 __u8 *opt = (__u8 *)(ra_msg + 1);
1250
1251 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1252 sizeof(struct ra_msg);
1253
1254 net_dbg_ratelimited("RA: %s, dev: %s\n", __func__, skb->dev->name);
1255 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1256 net_dbg_ratelimited("RA: source address is not link-local\n");
1257 return reason;
1258 }
1259 if (optlen < 0)
1260 return SKB_DROP_REASON_PKT_TOO_SMALL;
1261
1262#ifdef CONFIG_IPV6_NDISC_NODETYPE
1263 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1264 net_dbg_ratelimited("RA: from host or unauthorized router\n");
1265 return reason;
1266 }
1267#endif
1268
1269 in6_dev = __in6_dev_get(skb->dev);
1270 if (!in6_dev) {
1271 net_err_ratelimited("RA: can't find inet6 device for %s\n", skb->dev->name);
1272 return reason;
1273 }
1274
1275 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts))
1276 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1277
1278 if (!ipv6_accept_ra(in6_dev)) {
1279 net_dbg_ratelimited("RA: %s, did not accept ra for dev: %s\n", __func__,
1280 skb->dev->name);
1281 goto skip_linkparms;
1282 }
1283
1284#ifdef CONFIG_IPV6_NDISC_NODETYPE
1285 /* skip link-specific parameters from interior routers */
1286 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1287 net_dbg_ratelimited("RA: %s, nodetype is NODEFAULT, dev: %s\n", __func__,
1288 skb->dev->name);
1289 goto skip_linkparms;
1290 }
1291#endif
1292
1293 if (in6_dev->if_flags & IF_RS_SENT) {
1294 /*
1295 * flag that an RA was received after an RS was sent
1296 * out on this interface.
1297 */
1298 in6_dev->if_flags |= IF_RA_RCVD;
1299 }
1300
1301 /*
1302 * Remember the managed/otherconf flags from most recently
1303 * received RA message (RFC 2462) -- yoshfuji
1304 */
1305 old_if_flags = in6_dev->if_flags;
1306 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1307 IF_RA_OTHERCONF)) |
1308 (ra_msg->icmph.icmp6_addrconf_managed ?
1309 IF_RA_MANAGED : 0) |
1310 (ra_msg->icmph.icmp6_addrconf_other ?
1311 IF_RA_OTHERCONF : 0);
1312
1313 if (old_if_flags != in6_dev->if_flags)
1314 send_ifinfo_notify = true;
1315
1316 if (!READ_ONCE(in6_dev->cnf.accept_ra_defrtr)) {
1317 net_dbg_ratelimited("RA: %s, defrtr is false for dev: %s\n", __func__,
1318 skb->dev->name);
1319 goto skip_defrtr;
1320 }
1321
1322 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1323 if (lifetime != 0 &&
1324 lifetime < READ_ONCE(in6_dev->cnf.accept_ra_min_lft)) {
1325 net_dbg_ratelimited("RA: router lifetime (%ds) is too short: %s\n", lifetime,
1326 skb->dev->name);
1327 goto skip_defrtr;
1328 }
1329
1330 /* Do not accept RA with source-addr found on local machine unless
1331 * accept_ra_from_local is set to true.
1332 */
1333 net = dev_net(in6_dev->dev);
1334 if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1335 ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1336 net_dbg_ratelimited("RA from local address detected on dev: %s: default router ignored\n",
1337 skb->dev->name);
1338 goto skip_defrtr;
1339 }
1340
1341#ifdef CONFIG_IPV6_ROUTER_PREF
1342 pref = ra_msg->icmph.icmp6_router_pref;
1343 /* 10b is handled as if it were 00b (medium) */
1344 if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1345 !READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref))
1346 pref = ICMPV6_ROUTER_PREF_MEDIUM;
1347#endif
1348 /* routes added from RAs do not use nexthop objects */
1349 rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1350 if (rt) {
1351 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1352 rt->fib6_nh->fib_nh_dev, NULL,
1353 &ipv6_hdr(skb)->saddr);
1354 if (!neigh) {
1355 net_err_ratelimited("RA: %s got default router without neighbour\n",
1356 __func__);
1357 fib6_info_release(rt);
1358 return reason;
1359 }
1360 }
1361 /* Set default route metric as specified by user */
1362 defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1363 /* delete the route if lifetime is 0 or if metric needs change */
1364 if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1365 ip6_del_rt_reason(net, rt,
1366 lifetime == 0 ? RT_DEL_REASON_RA_WITHDRAWN :
1367 RT_DEL_REASON_UNSPEC);
1368 rt = NULL;
1369 }
1370
1371 net_dbg_ratelimited("RA: rt: %p lifetime: %d, metric: %d, for dev: %s\n", rt, lifetime,
1372 defrtr_usr_metric, skb->dev->name);
1373 if (!rt && lifetime) {
1374 net_dbg_ratelimited("RA: adding default router\n");
1375
1376 if (neigh)
1377 neigh_release(neigh);
1378
1379 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1380 skb->dev, pref, defrtr_usr_metric,
1381 lifetime);
1382 if (!rt) {
1383 net_err_ratelimited("RA: %s failed to add default route\n", __func__);
1384 return reason;
1385 }
1386
1387 neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1388 rt->fib6_nh->fib_nh_dev, NULL,
1389 &ipv6_hdr(skb)->saddr);
1390 if (!neigh) {
1391 net_err_ratelimited("RA: %s got default router without neighbour\n",
1392 __func__);
1393 fib6_info_release(rt);
1394 return reason;
1395 }
1396 neigh->flags |= NTF_ROUTER;
1397 } else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1398 struct nl_info nlinfo = {
1399 .nl_net = net,
1400 };
1401 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1402 inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1403 }
1404
1405 if (rt) {
1406 table = rt->fib6_table;
1407 spin_lock_bh(&table->tb6_lock);
1408
1409 fib6_set_expires(rt, jiffies + (HZ * lifetime));
1410 fib6_add_gc_list(rt);
1411
1412 spin_unlock_bh(&table->tb6_lock);
1413 }
1414 if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) < 256 &&
1415 ra_msg->icmph.icmp6_hop_limit) {
1416 if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) <=
1417 ra_msg->icmph.icmp6_hop_limit) {
1418 WRITE_ONCE(in6_dev->cnf.hop_limit,
1419 ra_msg->icmph.icmp6_hop_limit);
1420 fib6_metric_set(rt, RTAX_HOPLIMIT,
1421 ra_msg->icmph.icmp6_hop_limit);
1422 } else {
1423 net_dbg_ratelimited("RA: Got route advertisement with lower hop_limit than minimum\n");
1424 }
1425 }
1426
1427skip_defrtr:
1428
1429 /*
1430 * Update Reachable Time and Retrans Timer
1431 */
1432
1433 if (in6_dev->nd_parms) {
1434 unsigned long rtime = ntohl(ra_msg->retrans_timer);
1435
1436 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1437 rtime = (rtime*HZ)/1000;
1438 if (rtime < HZ/100)
1439 rtime = HZ/100;
1440 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1441 in6_dev->tstamp = jiffies;
1442 send_ifinfo_notify = true;
1443 }
1444
1445 rtime = ntohl(ra_msg->reachable_time);
1446 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1447 rtime = (rtime*HZ)/1000;
1448
1449 if (rtime < HZ/10)
1450 rtime = HZ/10;
1451
1452 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1453 NEIGH_VAR_SET(in6_dev->nd_parms,
1454 BASE_REACHABLE_TIME, rtime);
1455 NEIGH_VAR_SET(in6_dev->nd_parms,
1456 GC_STALETIME, 3 * rtime);
1457 neigh_set_reach_time(in6_dev->nd_parms);
1458 in6_dev->tstamp = jiffies;
1459 send_ifinfo_notify = true;
1460 }
1461 }
1462 }
1463
1464skip_linkparms:
1465
1466 /*
1467 * Process options.
1468 */
1469
1470 if (!neigh)
1471 neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1472 skb->dev, 1);
1473 if (neigh) {
1474 u8 *lladdr = NULL;
1475 if (ndopts.nd_opts_src_lladdr) {
1476 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1477 skb->dev);
1478 if (!lladdr) {
1479 net_dbg_ratelimited("RA: invalid link-layer address length\n");
1480 goto out;
1481 }
1482 }
1483 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1484 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1485 NEIGH_UPDATE_F_OVERRIDE|
1486 NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1487 NEIGH_UPDATE_F_ISROUTER,
1488 NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1489 reason = SKB_CONSUMED;
1490 }
1491
1492 if (!ipv6_accept_ra(in6_dev)) {
1493 net_dbg_ratelimited("RA: %s, accept_ra is false for dev: %s\n", __func__,
1494 skb->dev->name);
1495 goto out;
1496 }
1497
1498#ifdef CONFIG_IPV6_ROUTE_INFO
1499 if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1500 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1501 in6_dev->dev, 0)) {
1502 net_dbg_ratelimited("RA from local address detected on dev: %s: router info ignored.\n",
1503 skb->dev->name);
1504 goto skip_routeinfo;
1505 }
1506
1507 if (READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref) && ndopts.nd_opts_ri) {
1508 struct nd_opt_hdr *p;
1509 for (p = ndopts.nd_opts_ri;
1510 p;
1511 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1512 struct route_info *ri = (struct route_info *)p;
1513#ifdef CONFIG_IPV6_NDISC_NODETYPE
1514 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1515 ri->prefix_len == 0)
1516 continue;
1517#endif
1518 if (ri->prefix_len == 0 &&
1519 !READ_ONCE(in6_dev->cnf.accept_ra_defrtr))
1520 continue;
1521 if (ri->lifetime != 0 &&
1522 ntohl(ri->lifetime) < READ_ONCE(in6_dev->cnf.accept_ra_min_lft))
1523 continue;
1524 if (ri->prefix_len < READ_ONCE(in6_dev->cnf.accept_ra_rt_info_min_plen))
1525 continue;
1526 if (ri->prefix_len > READ_ONCE(in6_dev->cnf.accept_ra_rt_info_max_plen))
1527 continue;
1528 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1529 &ipv6_hdr(skb)->saddr);
1530 }
1531 }
1532
1533skip_routeinfo:
1534#endif
1535
1536#ifdef CONFIG_IPV6_NDISC_NODETYPE
1537 /* skip link-specific ndopts from interior routers */
1538 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1539 net_dbg_ratelimited("RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1540 __func__, skb->dev->name);
1541 goto out;
1542 }
1543#endif
1544
1545 if (READ_ONCE(in6_dev->cnf.accept_ra_pinfo) && ndopts.nd_opts_pi) {
1546 struct nd_opt_hdr *p;
1547 for (p = ndopts.nd_opts_pi;
1548 p;
1549 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1550 addrconf_prefix_rcv(skb->dev, (u8 *)p,
1551 (p->nd_opt_len) << 3,
1552 ndopts.nd_opts_src_lladdr != NULL);
1553 }
1554 }
1555
1556 if (ndopts.nd_opts_mtu && READ_ONCE(in6_dev->cnf.accept_ra_mtu)) {
1557 __be32 n;
1558 u32 mtu;
1559
1560 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1561 mtu = ntohl(n);
1562
1563 if (READ_ONCE(in6_dev->ra_mtu) != mtu) {
1564 WRITE_ONCE(in6_dev->ra_mtu, mtu);
1565 send_ifinfo_notify = true;
1566 }
1567
1568 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1569 net_dbg_ratelimited("RA: invalid mtu: %d\n", mtu);
1570 } else if (READ_ONCE(in6_dev->cnf.mtu6) != mtu) {
1571 WRITE_ONCE(in6_dev->cnf.mtu6, mtu);
1572 fib6_metric_set(rt, RTAX_MTU, mtu);
1573 rt6_mtu_change(skb->dev, mtu);
1574 }
1575 }
1576
1577 if (ndopts.nd_useropts) {
1578 struct nd_opt_hdr *p;
1579 for (p = ndopts.nd_useropts;
1580 p;
1581 p = ndisc_next_useropt(skb->dev, p,
1582 ndopts.nd_useropts_end)) {
1583 ndisc_ra_useropt(skb, p);
1584 }
1585 }
1586
1587 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1588 net_dbg_ratelimited("RA: invalid RA options\n");
1589 }
1590out:
1591 /* Send a notify if RA changed managed/otherconf flags or
1592 * timer settings or ra_mtu value
1593 */
1594 if (send_ifinfo_notify)
1595 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1596
1597 fib6_info_release(rt);
1598 if (neigh)
1599 neigh_release(neigh);
1600 return reason;
1601}
1602
1603static enum skb_drop_reason ndisc_redirect_rcv(struct sk_buff *skb)
1604{
1605 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1606 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1607 offsetof(struct rd_msg, opt));
1608 struct ndisc_options ndopts;
1609 SKB_DR(reason);
1610 u8 *hdr;
1611
1612#ifdef CONFIG_IPV6_NDISC_NODETYPE
1613 switch (skb->ndisc_nodetype) {
1614 case NDISC_NODETYPE_HOST:
1615 case NDISC_NODETYPE_NODEFAULT:
1616 net_dbg_ratelimited("Redirect: from host or unauthorized router\n");
1617 return reason;
1618 }
1619#endif
1620
1621 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1622 net_dbg_ratelimited("Redirect: source address is not link-local\n");
1623 return reason;
1624 }
1625
1626 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1627 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1628
1629 if (!ndopts.nd_opts_rh) {
1630 ip6_redirect_no_header(skb, dev_net(skb->dev),
1631 skb->dev->ifindex);
1632 return reason;
1633 }
1634
1635 hdr = (u8 *)ndopts.nd_opts_rh;
1636 hdr += 8;
1637 if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1638 return SKB_DROP_REASON_PKT_TOO_SMALL;
1639
1640 return icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1641}
1642
1643static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1644 struct sk_buff *orig_skb,
1645 int rd_len)
1646{
1647 u8 *opt = skb_put(skb, rd_len);
1648
1649 memset(opt, 0, 8);
1650 *(opt++) = ND_OPT_REDIRECT_HDR;
1651 *(opt++) = (rd_len >> 3);
1652 opt += 6;
1653
1654 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1655 rd_len - 8);
1656}
1657
1658void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1659{
1660 struct net_device *dev = skb->dev;
1661 struct net *net = dev_net_rcu(dev);
1662 struct sock *sk = net->ipv6.ndisc_sk;
1663 int optlen = 0;
1664 struct inet_peer *peer;
1665 struct sk_buff *buff;
1666 struct rd_msg *msg;
1667 struct in6_addr saddr_buf;
1668 struct rt6_info *rt;
1669 struct dst_entry *dst;
1670 struct flowi6 fl6;
1671 int rd_len;
1672 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1673 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1674 bool ret;
1675
1676 if (netif_is_l3_master(dev)) {
1677 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
1678 if (!dev)
1679 return;
1680 }
1681
1682 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1683 net_dbg_ratelimited("Redirect: no link-local address on %s\n", dev->name);
1684 return;
1685 }
1686
1687 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1688 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1689 net_dbg_ratelimited("Redirect: target address is not link-local unicast\n");
1690 return;
1691 }
1692
1693 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1694 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1695
1696 dst = ip6_route_output(net, NULL, &fl6);
1697 if (dst->error) {
1698 dst_release(dst);
1699 return;
1700 }
1701 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1702 if (IS_ERR(dst))
1703 return;
1704
1705 rt = dst_rt6_info(dst);
1706
1707 if (rt->rt6i_flags & RTF_GATEWAY) {
1708 net_dbg_ratelimited("Redirect: destination is not a neighbour\n");
1709 goto release;
1710 }
1711
1712 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr);
1713 if (!peer)
1714 goto release;
1715 ret = inet_peer_xrlim_allow(peer, 1*HZ);
1716
1717 if (!ret)
1718 goto release;
1719
1720 if (dev->addr_len) {
1721 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1722 if (!neigh) {
1723 net_dbg_ratelimited("Redirect: no neigh for target address\n");
1724 goto release;
1725 }
1726
1727 read_lock_bh(&neigh->lock);
1728 if (neigh->nud_state & NUD_VALID) {
1729 memcpy(ha_buf, neigh->ha, dev->addr_len);
1730 read_unlock_bh(&neigh->lock);
1731 ha = ha_buf;
1732 optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1733 ops_data_buf,
1734 &ops_data);
1735 } else
1736 read_unlock_bh(&neigh->lock);
1737
1738 neigh_release(neigh);
1739 }
1740
1741 rd_len = min_t(unsigned int,
1742 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1743 skb->len + 8);
1744 rd_len &= ~0x7;
1745 optlen += rd_len;
1746
1747 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1748 if (!buff)
1749 goto release;
1750
1751 msg = skb_put(buff, sizeof(*msg));
1752 *msg = (struct rd_msg) {
1753 .icmph = {
1754 .icmp6_type = NDISC_REDIRECT,
1755 },
1756 .target = *target,
1757 .dest = ipv6_hdr(skb)->daddr,
1758 };
1759
1760 /*
1761 * include target_address option
1762 */
1763
1764 if (ha)
1765 ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1766
1767 /*
1768 * build redirect option and copy skb over to the new packet.
1769 */
1770
1771 if (rd_len)
1772 ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1773
1774 skb_dst_set(buff, dst);
1775 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1776 return;
1777
1778release:
1779 dst_release(dst);
1780}
1781
1782static void pndisc_redo(struct sk_buff *skb)
1783{
1784 enum skb_drop_reason reason = ndisc_recv_ns(skb);
1785
1786 kfree_skb_reason(skb, reason);
1787}
1788
1789static int ndisc_is_multicast(const void *pkey)
1790{
1791 return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1792}
1793
1794static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1795{
1796 struct inet6_dev *idev = __in6_dev_get(skb->dev);
1797
1798 if (!idev)
1799 return true;
1800 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1801 READ_ONCE(idev->cnf.suppress_frag_ndisc)) {
1802 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1803 return true;
1804 }
1805 return false;
1806}
1807
1808enum skb_drop_reason ndisc_rcv(struct sk_buff *skb)
1809{
1810 struct nd_msg *msg;
1811 SKB_DR(reason);
1812
1813 if (ndisc_suppress_frag_ndisc(skb))
1814 return SKB_DROP_REASON_IPV6_NDISC_FRAG;
1815
1816 if (skb_linearize(skb))
1817 return SKB_DROP_REASON_NOMEM;
1818
1819 msg = (struct nd_msg *)skb_transport_header(skb);
1820
1821 __skb_push(skb, skb->data - skb_transport_header(skb));
1822
1823 if (ipv6_hdr(skb)->hop_limit != 255) {
1824 net_dbg_ratelimited("NDISC: invalid hop-limit: %d\n", ipv6_hdr(skb)->hop_limit);
1825 return SKB_DROP_REASON_IPV6_NDISC_HOP_LIMIT;
1826 }
1827
1828 if (msg->icmph.icmp6_code != 0) {
1829 net_dbg_ratelimited("NDISC: invalid ICMPv6 code: %d\n", msg->icmph.icmp6_code);
1830 return SKB_DROP_REASON_IPV6_NDISC_BAD_CODE;
1831 }
1832
1833 switch (msg->icmph.icmp6_type) {
1834 case NDISC_NEIGHBOUR_SOLICITATION:
1835 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1836 reason = ndisc_recv_ns(skb);
1837 break;
1838
1839 case NDISC_NEIGHBOUR_ADVERTISEMENT:
1840 reason = ndisc_recv_na(skb);
1841 break;
1842
1843 case NDISC_ROUTER_SOLICITATION:
1844 reason = ndisc_recv_rs(skb);
1845 break;
1846
1847 case NDISC_ROUTER_ADVERTISEMENT:
1848 reason = ndisc_router_discovery(skb);
1849 break;
1850
1851 case NDISC_REDIRECT:
1852 reason = ndisc_redirect_rcv(skb);
1853 break;
1854 }
1855
1856 return reason;
1857}
1858
1859static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1860{
1861 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1862 struct netdev_notifier_change_info *change_info;
1863 struct net *net = dev_net(dev);
1864 struct inet6_dev *idev;
1865 bool evict_nocarrier;
1866
1867 switch (event) {
1868 case NETDEV_CHANGEADDR:
1869 neigh_changeaddr(&nd_tbl, dev);
1870 fib6_run_gc(0, net, false);
1871 fallthrough;
1872 case NETDEV_UP:
1873 idev = in6_dev_get(dev);
1874 if (!idev)
1875 break;
1876 if (READ_ONCE(idev->cnf.ndisc_notify) ||
1877 READ_ONCE(net->ipv6.devconf_all->ndisc_notify))
1878 ndisc_send_unsol_na(dev);
1879 in6_dev_put(idev);
1880 break;
1881 case NETDEV_CHANGE:
1882 idev = in6_dev_get(dev);
1883 if (!idev)
1884 evict_nocarrier = true;
1885 else {
1886 evict_nocarrier = READ_ONCE(idev->cnf.ndisc_evict_nocarrier) &&
1887 READ_ONCE(net->ipv6.devconf_all->ndisc_evict_nocarrier);
1888 in6_dev_put(idev);
1889 }
1890
1891 change_info = ptr;
1892 if (change_info->flags_changed & IFF_NOARP)
1893 neigh_changeaddr(&nd_tbl, dev);
1894 if (evict_nocarrier && !netif_carrier_ok(dev))
1895 neigh_carrier_down(&nd_tbl, dev);
1896 break;
1897 case NETDEV_DOWN:
1898 neigh_ifdown(&nd_tbl, dev);
1899 fib6_run_gc(0, net, false);
1900 break;
1901 case NETDEV_NOTIFY_PEERS:
1902 ndisc_send_unsol_na(dev);
1903 break;
1904 default:
1905 break;
1906 }
1907
1908 return NOTIFY_DONE;
1909}
1910
1911static struct notifier_block ndisc_netdev_notifier = {
1912 .notifier_call = ndisc_netdev_event,
1913 .priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1914};
1915
1916#ifdef CONFIG_SYSCTL
1917static void ndisc_warn_deprecated_sysctl(const struct ctl_table *ctl,
1918 const char *func, const char *dev_name)
1919{
1920 static char warncomm[TASK_COMM_LEN];
1921 static int warned;
1922 if (strcmp(warncomm, current->comm) && warned < 5) {
1923 strscpy(warncomm, current->comm);
1924 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1925 warncomm, func,
1926 dev_name, ctl->procname,
1927 dev_name, ctl->procname);
1928 warned++;
1929 }
1930}
1931
1932int ndisc_ifinfo_sysctl_change(const struct ctl_table *ctl, int write, void *buffer,
1933 size_t *lenp, loff_t *ppos)
1934{
1935 struct net_device *dev = ctl->extra1;
1936 struct inet6_dev *idev;
1937 int ret;
1938
1939 if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1940 (strcmp(ctl->procname, "base_reachable_time") == 0))
1941 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1942
1943 if (strcmp(ctl->procname, "retrans_time") == 0)
1944 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1945
1946 else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1947 ret = neigh_proc_dointvec_jiffies(ctl, write,
1948 buffer, lenp, ppos);
1949
1950 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1951 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1952 ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1953 buffer, lenp, ppos);
1954 else
1955 ret = -1;
1956
1957 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1958 if (ctl->data == NEIGH_VAR_PTR(idev->nd_parms, BASE_REACHABLE_TIME))
1959 neigh_set_reach_time(idev->nd_parms);
1960
1961 WRITE_ONCE(idev->tstamp, jiffies);
1962 inet6_ifinfo_notify(RTM_NEWLINK, idev);
1963 in6_dev_put(idev);
1964 }
1965 return ret;
1966}
1967
1968
1969#endif
1970
1971static int __net_init ndisc_net_init(struct net *net)
1972{
1973 struct ipv6_pinfo *np;
1974 struct sock *sk;
1975 int err;
1976
1977 err = inet_ctl_sock_create(&sk, PF_INET6,
1978 SOCK_RAW, IPPROTO_ICMPV6, net);
1979 if (err < 0) {
1980 net_err_ratelimited("NDISC: Failed to initialize the control socket (err %d)\n",
1981 err);
1982 return err;
1983 }
1984
1985 net->ipv6.ndisc_sk = sk;
1986
1987 np = inet6_sk(sk);
1988 np->hop_limit = 255;
1989 /* Do not loopback ndisc messages */
1990 inet6_clear_bit(MC6_LOOP, sk);
1991
1992 return 0;
1993}
1994
1995static void __net_exit ndisc_net_exit(struct net *net)
1996{
1997 inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1998}
1999
2000static struct pernet_operations ndisc_net_ops = {
2001 .init = ndisc_net_init,
2002 .exit = ndisc_net_exit,
2003};
2004
2005int __init ndisc_init(void)
2006{
2007 int err;
2008
2009 err = register_pernet_subsys(&ndisc_net_ops);
2010 if (err)
2011 return err;
2012 /*
2013 * Initialize the neighbour table
2014 */
2015 neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2016
2017#ifdef CONFIG_SYSCTL
2018 err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2019 ndisc_ifinfo_sysctl_change);
2020 if (err)
2021 goto out_unregister_pernet;
2022out:
2023#endif
2024 return err;
2025
2026#ifdef CONFIG_SYSCTL
2027out_unregister_pernet:
2028 unregister_pernet_subsys(&ndisc_net_ops);
2029 goto out;
2030#endif
2031}
2032
2033int __init ndisc_late_init(void)
2034{
2035 return register_netdevice_notifier(&ndisc_netdev_notifier);
2036}
2037
2038void ndisc_late_cleanup(void)
2039{
2040 unregister_netdevice_notifier(&ndisc_netdev_notifier);
2041}
2042
2043void ndisc_cleanup(void)
2044{
2045#ifdef CONFIG_SYSCTL
2046 neigh_sysctl_unregister(&nd_tbl.parms);
2047#endif
2048 neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2049 unregister_pernet_subsys(&ndisc_net_ops);
2050}