Linux Audio

Check our new training course

Loading...
v6.13.7
   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		<roque@di.fc.ul.pt>
   8 *	Mike Shaver		<shaver@ingenia.com>
   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_MAX,
 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				ND_PRINTK(2, warn,
 247					  "%s: duplicated ND6 option found: type=%d\n",
 248					  __func__, nd_opt->nd_opt_type);
 249			} else {
 250				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
 251			}
 252			break;
 253		case ND_OPT_PREFIX_INFO:
 254			ndopts->nd_opts_pi_end = nd_opt;
 255			if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
 256				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
 257			break;
 258#ifdef CONFIG_IPV6_ROUTE_INFO
 259		case ND_OPT_ROUTE_INFO:
 260			ndopts->nd_opts_ri_end = nd_opt;
 261			if (!ndopts->nd_opts_ri)
 262				ndopts->nd_opts_ri = nd_opt;
 263			break;
 264#endif
 265		default:
 266			unknown = true;
 267		}
 268		if (ndisc_is_useropt(dev, nd_opt)) {
 269			ndopts->nd_useropts_end = nd_opt;
 270			if (!ndopts->nd_useropts)
 271				ndopts->nd_useropts = nd_opt;
 272		} else if (unknown) {
 273			/*
 274			 * Unknown options must be silently ignored,
 275			 * to accommodate future extension to the
 276			 * protocol.
 277			 */
 278			ND_PRINTK(2, notice,
 279				  "%s: ignored unsupported option; type=%d, len=%d\n",
 280				  __func__,
 281				  nd_opt->nd_opt_type,
 282				  nd_opt->nd_opt_len);
 283		}
 284next_opt:
 285		opt_len -= l;
 286		nd_opt = ((void *)nd_opt) + l;
 287	}
 288	return ndopts;
 289}
 290
 291int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir)
 292{
 293	switch (dev->type) {
 294	case ARPHRD_ETHER:
 295	case ARPHRD_IEEE802:	/* Not sure. Check it later. --ANK */
 296	case ARPHRD_FDDI:
 297		ipv6_eth_mc_map(addr, buf);
 298		return 0;
 299	case ARPHRD_ARCNET:
 300		ipv6_arcnet_mc_map(addr, buf);
 301		return 0;
 302	case ARPHRD_INFINIBAND:
 303		ipv6_ib_mc_map(addr, dev->broadcast, buf);
 304		return 0;
 305	case ARPHRD_IPGRE:
 306		return ipv6_ipgre_mc_map(addr, dev->broadcast, buf);
 307	default:
 308		if (dir) {
 309			memcpy(buf, dev->broadcast, dev->addr_len);
 310			return 0;
 311		}
 312	}
 313	return -EINVAL;
 314}
 315EXPORT_SYMBOL(ndisc_mc_map);
 316
 317static u32 ndisc_hash(const void *pkey,
 318		      const struct net_device *dev,
 319		      __u32 *hash_rnd)
 320{
 321	return ndisc_hashfn(pkey, dev, hash_rnd);
 322}
 323
 324static bool ndisc_key_eq(const struct neighbour *n, const void *pkey)
 325{
 326	return neigh_key_eq128(n, pkey);
 327}
 328
 329static int ndisc_constructor(struct neighbour *neigh)
 330{
 331	struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key;
 332	struct net_device *dev = neigh->dev;
 333	struct inet6_dev *in6_dev;
 334	struct neigh_parms *parms;
 335	bool is_multicast = ipv6_addr_is_multicast(addr);
 336
 337	in6_dev = in6_dev_get(dev);
 338	if (!in6_dev) {
 339		return -EINVAL;
 340	}
 341
 342	parms = in6_dev->nd_parms;
 343	__neigh_parms_put(neigh->parms);
 344	neigh->parms = neigh_parms_clone(parms);
 345
 346	neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST;
 347	if (!dev->header_ops) {
 348		neigh->nud_state = NUD_NOARP;
 349		neigh->ops = &ndisc_direct_ops;
 350		neigh->output = neigh_direct_output;
 351	} else {
 352		if (is_multicast) {
 353			neigh->nud_state = NUD_NOARP;
 354			ndisc_mc_map(addr, neigh->ha, dev, 1);
 355		} else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) {
 356			neigh->nud_state = NUD_NOARP;
 357			memcpy(neigh->ha, dev->dev_addr, dev->addr_len);
 358			if (dev->flags&IFF_LOOPBACK)
 359				neigh->type = RTN_LOCAL;
 360		} else if (dev->flags&IFF_POINTOPOINT) {
 361			neigh->nud_state = NUD_NOARP;
 362			memcpy(neigh->ha, dev->broadcast, dev->addr_len);
 363		}
 364		if (dev->header_ops->cache)
 365			neigh->ops = &ndisc_hh_ops;
 366		else
 367			neigh->ops = &ndisc_generic_ops;
 368		if (neigh->nud_state&NUD_VALID)
 369			neigh->output = neigh->ops->connected_output;
 370		else
 371			neigh->output = neigh->ops->output;
 372	}
 373	in6_dev_put(in6_dev);
 374	return 0;
 375}
 376
 377static int pndisc_constructor(struct pneigh_entry *n)
 378{
 379	struct in6_addr *addr = (struct in6_addr *)&n->key;
 380	struct in6_addr maddr;
 381	struct net_device *dev = n->dev;
 382
 383	if (!dev || !__in6_dev_get(dev))
 384		return -EINVAL;
 385	addrconf_addr_solict_mult(addr, &maddr);
 386	ipv6_dev_mc_inc(dev, &maddr);
 387	return 0;
 388}
 389
 390static void pndisc_destructor(struct pneigh_entry *n)
 391{
 392	struct in6_addr *addr = (struct in6_addr *)&n->key;
 393	struct in6_addr maddr;
 394	struct net_device *dev = n->dev;
 395
 396	if (!dev || !__in6_dev_get(dev))
 397		return;
 398	addrconf_addr_solict_mult(addr, &maddr);
 399	ipv6_dev_mc_dec(dev, &maddr);
 400}
 401
 402/* called with rtnl held */
 403static bool ndisc_allow_add(const struct net_device *dev,
 404			    struct netlink_ext_ack *extack)
 405{
 406	struct inet6_dev *idev = __in6_dev_get(dev);
 407
 408	if (!idev || idev->cnf.disable_ipv6) {
 409		NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
 410		return false;
 411	}
 412
 413	return true;
 414}
 415
 416static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
 417				       int len)
 418{
 419	int hlen = LL_RESERVED_SPACE(dev);
 420	int tlen = dev->needed_tailroom;
 
 421	struct sk_buff *skb;
 422
 423	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
 424	if (!skb)
 
 
 425		return NULL;
 
 426
 427	skb->protocol = htons(ETH_P_IPV6);
 428	skb->dev = dev;
 429
 430	skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
 431	skb_reset_transport_header(skb);
 432
 433	/* Manually assign socket ownership as we avoid calling
 434	 * sock_alloc_send_pskb() to bypass wmem buffer limits
 435	 */
 436	rcu_read_lock();
 437	skb_set_owner_w(skb, dev_net_rcu(dev)->ipv6.ndisc_sk);
 438	rcu_read_unlock();
 439
 440	return skb;
 441}
 442
 443static void ip6_nd_hdr(struct sk_buff *skb,
 444		       const struct in6_addr *saddr,
 445		       const struct in6_addr *daddr,
 446		       int hop_limit, int len)
 447{
 448	struct ipv6hdr *hdr;
 449	struct inet6_dev *idev;
 450	unsigned tclass;
 451
 452	rcu_read_lock();
 453	idev = __in6_dev_get(skb->dev);
 454	tclass = idev ? READ_ONCE(idev->cnf.ndisc_tclass) : 0;
 455	rcu_read_unlock();
 456
 457	skb_push(skb, sizeof(*hdr));
 458	skb_reset_network_header(skb);
 459	hdr = ipv6_hdr(skb);
 460
 461	ip6_flow_hdr(hdr, tclass, 0);
 462
 463	hdr->payload_len = htons(len);
 464	hdr->nexthdr = IPPROTO_ICMPV6;
 465	hdr->hop_limit = hop_limit;
 466
 467	hdr->saddr = *saddr;
 468	hdr->daddr = *daddr;
 469}
 470
 471void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
 472		    const struct in6_addr *saddr)
 473{
 474	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
 475	struct dst_entry *dst = skb_dst(skb);
 
 
 476	struct inet6_dev *idev;
 477	struct net *net;
 478	struct sock *sk;
 479	int err;
 
 480	u8 type;
 481
 482	type = icmp6h->icmp6_type;
 483
 484	rcu_read_lock();
 485
 486	net = dev_net_rcu(skb->dev);
 487	sk = net->ipv6.ndisc_sk;
 488	if (!dst) {
 489		struct flowi6 fl6;
 490		int oif = skb->dev->ifindex;
 491
 492		icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
 493		dst = icmp6_dst_alloc(skb->dev, &fl6);
 494		if (IS_ERR(dst)) {
 495			rcu_read_unlock();
 496			kfree_skb(skb);
 497			return;
 498		}
 499
 500		skb_dst_set(skb, dst);
 501	}
 502
 503	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
 504					      IPPROTO_ICMPV6,
 505					      csum_partial(icmp6h,
 506							   skb->len, 0));
 507
 508	ip6_nd_hdr(skb, saddr, daddr, READ_ONCE(inet6_sk(sk)->hop_limit), skb->len);
 509
 
 510	idev = __in6_dev_get(dst->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, dst->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}
 580
 581static void ndisc_send_unsol_na(struct net_device *dev)
 582{
 583	struct inet6_dev *idev;
 584	struct inet6_ifaddr *ifa;
 585
 586	idev = in6_dev_get(dev);
 587	if (!idev)
 588		return;
 589
 590	read_lock_bh(&idev->lock);
 591	list_for_each_entry(ifa, &idev->addr_list, if_list) {
 592		/* skip tentative addresses until dad completes */
 593		if (ifa->flags & IFA_F_TENTATIVE &&
 594		    !(ifa->flags & IFA_F_OPTIMISTIC))
 595			continue;
 596
 597		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
 598			      /*router=*/ !!idev->cnf.forwarding,
 599			      /*solicited=*/ false, /*override=*/ true,
 600			      /*inc_opt=*/ true);
 601	}
 602	read_unlock_bh(&idev->lock);
 603
 604	in6_dev_put(idev);
 605}
 606
 607struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
 608				const struct in6_addr *saddr, u64 nonce)
 609{
 610	int inc_opt = dev->addr_len;
 611	struct sk_buff *skb;
 612	struct nd_msg *msg;
 613	int optlen = 0;
 614
 615	if (!saddr)
 616		return NULL;
 617
 618	if (ipv6_addr_any(saddr))
 619		inc_opt = false;
 620	if (inc_opt)
 621		optlen += ndisc_opt_addr_space(dev,
 622					       NDISC_NEIGHBOUR_SOLICITATION);
 623	if (nonce != 0)
 624		optlen += 8;
 625
 626	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
 627	if (!skb)
 628		return NULL;
 629
 630	msg = skb_put(skb, sizeof(*msg));
 631	*msg = (struct nd_msg) {
 632		.icmph = {
 633			.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
 634		},
 635		.target = *solicit,
 636	};
 637
 638	if (inc_opt)
 639		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
 640				       dev->dev_addr,
 641				       NDISC_NEIGHBOUR_SOLICITATION);
 642	if (nonce != 0) {
 643		u8 *opt = skb_put(skb, 8);
 644
 645		opt[0] = ND_OPT_NONCE;
 646		opt[1] = 8 >> 3;
 647		memcpy(opt + 2, &nonce, 6);
 648	}
 649
 650	return skb;
 651}
 652EXPORT_SYMBOL(ndisc_ns_create);
 653
 654void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
 655		   const struct in6_addr *daddr, const struct in6_addr *saddr,
 656		   u64 nonce)
 657{
 658	struct in6_addr addr_buf;
 659	struct sk_buff *skb;
 660
 661	if (!saddr) {
 662		if (ipv6_get_lladdr(dev, &addr_buf,
 663				    (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
 664			return;
 665		saddr = &addr_buf;
 666	}
 667
 668	skb = ndisc_ns_create(dev, solicit, saddr, nonce);
 669
 670	if (skb)
 671		ndisc_send_skb(skb, daddr, saddr);
 672}
 673
 674void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
 675		   const struct in6_addr *daddr)
 676{
 677	struct sk_buff *skb;
 678	struct rs_msg *msg;
 679	int send_sllao = dev->addr_len;
 680	int optlen = 0;
 681
 682#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
 683	/*
 684	 * According to section 2.2 of RFC 4429, we must not
 685	 * send router solicitations with a sllao from
 686	 * optimistic addresses, but we may send the solicitation
 687	 * if we don't include the sllao.  So here we check
 688	 * if our address is optimistic, and if so, we
 689	 * suppress the inclusion of the sllao.
 690	 */
 691	if (send_sllao) {
 692		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
 693							   dev, 1);
 694		if (ifp) {
 695			if (ifp->flags & IFA_F_OPTIMISTIC)  {
 696				send_sllao = 0;
 697			}
 698			in6_ifa_put(ifp);
 699		} else {
 700			send_sllao = 0;
 701		}
 702	}
 703#endif
 704	if (send_sllao)
 705		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
 706
 707	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
 708	if (!skb)
 709		return;
 710
 711	msg = skb_put(skb, sizeof(*msg));
 712	*msg = (struct rs_msg) {
 713		.icmph = {
 714			.icmp6_type = NDISC_ROUTER_SOLICITATION,
 715		},
 716	};
 717
 718	if (send_sllao)
 719		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
 720				       dev->dev_addr,
 721				       NDISC_ROUTER_SOLICITATION);
 722
 723	ndisc_send_skb(skb, daddr, saddr);
 724}
 725
 726
 727static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
 728{
 729	/*
 730	 *	"The sender MUST return an ICMP
 731	 *	 destination unreachable"
 732	 */
 733	dst_link_failure(skb);
 734	kfree_skb(skb);
 735}
 736
 737/* Called with locked neigh: either read or both */
 738
 739static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
 740{
 741	struct in6_addr *saddr = NULL;
 742	struct in6_addr mcaddr;
 743	struct net_device *dev = neigh->dev;
 744	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
 745	int probes = atomic_read(&neigh->probes);
 746
 747	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
 748					   dev, false, 1,
 749					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
 750		saddr = &ipv6_hdr(skb)->saddr;
 751	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
 752	if (probes < 0) {
 753		if (!(READ_ONCE(neigh->nud_state) & NUD_VALID)) {
 754			ND_PRINTK(1, dbg,
 755				  "%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	read_lock_bh(&nd_tbl.lock);
 774	n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
 775	if (n)
 776		ret = !!(n->flags & NTF_ROUTER);
 777	read_unlock_bh(&nd_tbl.lock);
 778
 779	return ret;
 780}
 781
 782void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
 783		  const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
 784		  struct ndisc_options *ndopts)
 785{
 786	neigh_update(neigh, lladdr, new, flags, 0);
 787	/* report ndisc ops about neighbour update */
 788	ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
 789}
 790
 791static enum skb_drop_reason ndisc_recv_ns(struct sk_buff *skb)
 792{
 793	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
 794	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
 795	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
 796	u8 *lladdr = NULL;
 797	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
 798				    offsetof(struct nd_msg, opt));
 799	struct ndisc_options ndopts;
 800	struct net_device *dev = skb->dev;
 801	struct inet6_ifaddr *ifp;
 802	struct inet6_dev *idev = NULL;
 803	struct neighbour *neigh;
 804	int dad = ipv6_addr_any(saddr);
 
 805	int is_router = -1;
 806	SKB_DR(reason);
 807	u64 nonce = 0;
 808	bool inc;
 809
 810	if (skb->len < sizeof(struct nd_msg))
 811		return SKB_DROP_REASON_PKT_TOO_SMALL;
 
 
 812
 813	if (ipv6_addr_is_multicast(&msg->target)) {
 814		ND_PRINTK(2, warn, "NS: multicast target address\n");
 815		return reason;
 816	}
 817
 818	/*
 819	 * RFC2461 7.1.1:
 820	 * DAD has to be destined for solicited node multicast address.
 821	 */
 822	if (dad && !ipv6_addr_is_solict_mult(daddr)) {
 823		ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
 824		return reason;
 825	}
 826
 827	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
 828		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
 
 
 829
 830	if (ndopts.nd_opts_src_lladdr) {
 831		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
 832		if (!lladdr) {
 833			ND_PRINTK(2, warn,
 834				  "NS: invalid link-layer address length\n");
 835			return reason;
 836		}
 837
 838		/* RFC2461 7.1.1:
 839		 *	If the IP source address is the unspecified address,
 840		 *	there MUST NOT be source link-layer address option
 841		 *	in the message.
 842		 */
 843		if (dad) {
 844			ND_PRINTK(2, warn,
 845				  "NS: bad DAD packet (link-layer address option)\n");
 846			return reason;
 847		}
 848	}
 849	if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
 850		memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
 851
 852	inc = ipv6_addr_is_multicast(daddr);
 853
 854	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
 855	if (ifp) {
 856have_ifp:
 857		if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
 858			if (dad) {
 859				if (nonce != 0 && ifp->dad_nonce == nonce) {
 860					u8 *np = (u8 *)&nonce;
 861					/* Matching nonce if looped back */
 862					ND_PRINTK(2, notice,
 863						  "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
 864						  ifp->idev->dev->name,
 865						  &ifp->addr, np);
 866					goto out;
 867				}
 868				/*
 869				 * We are colliding with another node
 870				 * who is doing DAD
 871				 * so fail our DAD process
 872				 */
 873				addrconf_dad_failure(skb, ifp);
 874				return reason;
 875			} else {
 876				/*
 877				 * This is not a dad solicitation.
 878				 * If we are an optimistic node,
 879				 * we should respond.
 880				 * Otherwise, we should ignore it.
 881				 */
 882				if (!(ifp->flags & IFA_F_OPTIMISTIC))
 883					goto out;
 884			}
 885		}
 886
 887		idev = ifp->idev;
 888	} else {
 889		struct net *net = dev_net(dev);
 890
 891		/* perhaps an address on the master device */
 892		if (netif_is_l3_slave(dev)) {
 893			struct net_device *mdev;
 894
 895			mdev = netdev_master_upper_dev_get_rcu(dev);
 896			if (mdev) {
 897				ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
 898				if (ifp)
 899					goto have_ifp;
 900			}
 901		}
 902
 903		idev = in6_dev_get(dev);
 904		if (!idev) {
 905			/* XXX: count this drop? */
 906			return reason;
 907		}
 908
 909		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
 910		    (READ_ONCE(idev->cnf.forwarding) &&
 911		     (READ_ONCE(net->ipv6.devconf_all->proxy_ndp) ||
 912		      READ_ONCE(idev->cnf.proxy_ndp)) &&
 913		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
 914			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
 915			    skb->pkt_type != PACKET_HOST &&
 916			    inc &&
 917			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
 918				/*
 919				 * for anycast or proxy,
 920				 * sender should delay its response
 921				 * by a random time between 0 and
 922				 * MAX_ANYCAST_DELAY_TIME seconds.
 923				 * (RFC2461) -- yoshfuji
 924				 */
 925				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
 926				if (n)
 927					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
 928				goto out;
 929			}
 930		} else {
 931			SKB_DR_SET(reason, IPV6_NDISC_NS_OTHERHOST);
 932			goto out;
 933		}
 934	}
 935
 936	if (is_router < 0)
 937		is_router = READ_ONCE(idev->cnf.forwarding);
 938
 939	if (dad) {
 940		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
 941			      !!is_router, false, (ifp != NULL), true);
 942		goto out;
 943	}
 944
 945	if (inc)
 946		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
 947	else
 948		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
 949
 950	/*
 951	 *	update / create cache entry
 952	 *	for the source address
 953	 */
 954	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
 955			       !inc || lladdr || !dev->addr_len);
 956	if (neigh)
 957		ndisc_update(dev, neigh, lladdr, NUD_STALE,
 958			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
 959			     NEIGH_UPDATE_F_OVERRIDE,
 960			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
 961	if (neigh || !dev->header_ops) {
 962		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
 963			      true, (ifp != NULL && inc), inc);
 964		if (neigh)
 965			neigh_release(neigh);
 966		reason = SKB_CONSUMED;
 967	}
 968
 969out:
 970	if (ifp)
 971		in6_ifa_put(ifp);
 972	else
 973		in6_dev_put(idev);
 974	return reason;
 975}
 976
 977static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
 978{
 979	struct inet6_dev *idev = __in6_dev_get(dev);
 980
 981	switch (READ_ONCE(idev->cnf.accept_untracked_na)) {
 982	case 0: /* Don't accept untracked na (absent in neighbor cache) */
 983		return 0;
 984	case 1: /* Create new entries from na if currently untracked */
 985		return 1;
 986	case 2: /* Create new entries from untracked na only if saddr is in the
 987		 * same subnet as an address configured on the interface that
 988		 * received the na
 989		 */
 990		return !!ipv6_chk_prefix(saddr, dev);
 991	default:
 992		return 0;
 993	}
 994}
 995
 996static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)
 997{
 998	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
 999	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1000	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
1001	u8 *lladdr = NULL;
1002	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1003				    offsetof(struct nd_msg, opt));
1004	struct ndisc_options ndopts;
1005	struct net_device *dev = skb->dev;
1006	struct inet6_dev *idev = __in6_dev_get(dev);
1007	struct inet6_ifaddr *ifp;
1008	struct neighbour *neigh;
1009	SKB_DR(reason);
1010	u8 new_state;
1011
1012	if (skb->len < sizeof(struct nd_msg))
1013		return SKB_DROP_REASON_PKT_TOO_SMALL;
 
 
1014
1015	if (ipv6_addr_is_multicast(&msg->target)) {
1016		ND_PRINTK(2, warn, "NA: target address is multicast\n");
1017		return reason;
1018	}
1019
1020	if (ipv6_addr_is_multicast(daddr) &&
1021	    msg->icmph.icmp6_solicited) {
1022		ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
1023		return reason;
1024	}
1025
1026	/* For some 802.11 wireless deployments (and possibly other networks),
1027	 * there will be a NA proxy and unsolicitd packets are attacks
1028	 * and thus should not be accepted.
1029	 * drop_unsolicited_na takes precedence over accept_untracked_na
1030	 */
1031	if (!msg->icmph.icmp6_solicited && idev &&
1032	    READ_ONCE(idev->cnf.drop_unsolicited_na))
1033		return reason;
1034
1035	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts))
1036		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1037
 
 
 
 
1038	if (ndopts.nd_opts_tgt_lladdr) {
1039		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1040		if (!lladdr) {
1041			ND_PRINTK(2, warn,
1042				  "NA: invalid link-layer address length\n");
1043			return reason;
1044		}
1045	}
1046	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1047	if (ifp) {
1048		if (skb->pkt_type != PACKET_LOOPBACK
1049		    && (ifp->flags & IFA_F_TENTATIVE)) {
1050				addrconf_dad_failure(skb, ifp);
1051				return reason;
1052		}
1053		/* What should we make now? The advertisement
1054		   is invalid, but ndisc specs say nothing
1055		   about it. It could be misconfiguration, or
1056		   an smart proxy agent tries to help us :-)
1057
1058		   We should not print the error if NA has been
1059		   received from loopback - it is just our own
1060		   unsolicited advertisement.
1061		 */
1062		if (skb->pkt_type != PACKET_LOOPBACK)
1063			ND_PRINTK(1, warn,
1064				  "NA: %pM advertised our address %pI6c on %s!\n",
1065				  eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1066		in6_ifa_put(ifp);
1067		return reason;
1068	}
1069
1070	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1071
1072	/* RFC 9131 updates original Neighbour Discovery RFC 4861.
1073	 * NAs with Target LL Address option without a corresponding
1074	 * entry in the neighbour cache can now create a STALE neighbour
1075	 * cache entry on routers.
1076	 *
1077	 *   entry accept  fwding  solicited        behaviour
1078	 * ------- ------  ------  ---------    ----------------------
1079	 * present      X       X         0     Set state to STALE
1080	 * present      X       X         1     Set state to REACHABLE
1081	 *  absent      0       X         X     Do nothing
1082	 *  absent      1       0         X     Do nothing
1083	 *  absent      1       1         X     Add a new STALE entry
1084	 *
1085	 * Note that we don't do a (daddr == all-routers-mcast) check.
1086	 */
1087	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1088	if (!neigh && lladdr && idev && READ_ONCE(idev->cnf.forwarding)) {
1089		if (accept_untracked_na(dev, saddr)) {
1090			neigh = neigh_create(&nd_tbl, &msg->target, dev);
1091			new_state = NUD_STALE;
1092		}
1093	}
1094
1095	if (neigh && !IS_ERR(neigh)) {
1096		u8 old_flags = neigh->flags;
1097		struct net *net = dev_net(dev);
1098
1099		if (READ_ONCE(neigh->nud_state) & NUD_FAILED)
1100			goto out;
1101
1102		/*
1103		 * Don't update the neighbor cache entry on a proxy NA from
1104		 * ourselves because either the proxied node is off link or it
1105		 * has already sent a NA to us.
1106		 */
1107		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1108		    READ_ONCE(net->ipv6.devconf_all->forwarding) &&
1109		    READ_ONCE(net->ipv6.devconf_all->proxy_ndp) &&
1110		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1111			/* XXX: idev->cnf.proxy_ndp */
1112			goto out;
1113		}
1114
1115		ndisc_update(dev, neigh, lladdr,
1116			     new_state,
1117			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1118			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1119			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1120			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1121			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1122
1123		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1124			/*
1125			 * Change: router to host
1126			 */
1127			rt6_clean_tohost(dev_net(dev),  saddr);
1128		}
1129		reason = SKB_CONSUMED;
1130out:
1131		neigh_release(neigh);
1132	}
1133	return reason;
1134}
1135
1136static enum skb_drop_reason ndisc_recv_rs(struct sk_buff *skb)
1137{
1138	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1139	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1140	struct neighbour *neigh;
1141	struct inet6_dev *idev;
1142	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1143	struct ndisc_options ndopts;
1144	u8 *lladdr = NULL;
1145	SKB_DR(reason);
1146
1147	if (skb->len < sizeof(*rs_msg))
1148		return SKB_DROP_REASON_PKT_TOO_SMALL;
1149
1150	idev = __in6_dev_get(skb->dev);
1151	if (!idev) {
1152		ND_PRINTK(1, err, "RS: can't find in6 device\n");
1153		return reason;
1154	}
1155
1156	/* Don't accept RS if we're not in router mode */
1157	if (!READ_ONCE(idev->cnf.forwarding))
1158		goto out;
1159
1160	/*
1161	 * Don't update NCE if src = ::;
1162	 * this implies that the source node has no ip address assigned yet.
1163	 */
1164	if (ipv6_addr_any(saddr))
1165		goto out;
1166
1167	/* Parse ND options */
1168	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts))
1169		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
 
 
1170
1171	if (ndopts.nd_opts_src_lladdr) {
1172		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1173					     skb->dev);
1174		if (!lladdr)
1175			goto out;
1176	}
1177
1178	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1179	if (neigh) {
1180		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1181			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1182			     NEIGH_UPDATE_F_OVERRIDE|
1183			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1184			     NDISC_ROUTER_SOLICITATION, &ndopts);
1185		neigh_release(neigh);
1186		reason = SKB_CONSUMED;
1187	}
1188out:
1189	return reason;
1190}
1191
1192static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1193{
1194	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1195	struct sk_buff *skb;
1196	struct nlmsghdr *nlh;
1197	struct nduseroptmsg *ndmsg;
1198	struct net *net = dev_net(ra->dev);
1199	int err;
1200	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1201				    + (opt->nd_opt_len << 3));
1202	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1203
1204	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1205	if (!skb) {
1206		err = -ENOBUFS;
1207		goto errout;
1208	}
1209
1210	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1211	if (!nlh) {
1212		goto nla_put_failure;
1213	}
1214
1215	ndmsg = nlmsg_data(nlh);
1216	ndmsg->nduseropt_family = AF_INET6;
1217	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1218	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1219	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1220	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1221
1222	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1223
1224	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1225		goto nla_put_failure;
1226	nlmsg_end(skb, nlh);
1227
1228	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1229	return;
1230
1231nla_put_failure:
1232	nlmsg_free(skb);
1233	err = -EMSGSIZE;
1234errout:
1235	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1236}
1237
1238static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb)
1239{
1240	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
1241	bool send_ifinfo_notify = false;
1242	struct neighbour *neigh = NULL;
1243	struct ndisc_options ndopts;
1244	struct fib6_info *rt = NULL;
1245	struct inet6_dev *in6_dev;
1246	struct fib6_table *table;
1247	u32 defrtr_usr_metric;
1248	unsigned int pref = 0;
1249	__u32 old_if_flags;
1250	struct net *net;
1251	SKB_DR(reason);
1252	int lifetime;
 
1253	int optlen;
 
 
 
1254
1255	__u8 *opt = (__u8 *)(ra_msg + 1);
1256
1257	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1258		sizeof(struct ra_msg);
1259
1260	ND_PRINTK(2, info,
1261		  "RA: %s, dev: %s\n",
1262		  __func__, skb->dev->name);
1263	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1264		ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1265		return reason;
 
 
 
 
1266	}
1267	if (optlen < 0)
1268		return SKB_DROP_REASON_PKT_TOO_SMALL;
1269
1270#ifdef CONFIG_IPV6_NDISC_NODETYPE
1271	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1272		ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1273		return reason;
1274	}
1275#endif
1276
 
 
 
 
1277	in6_dev = __in6_dev_get(skb->dev);
1278	if (!in6_dev) {
1279		ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1280			  skb->dev->name);
1281		return reason;
1282	}
1283
1284	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts))
1285		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
 
 
1286
1287	if (!ipv6_accept_ra(in6_dev)) {
1288		ND_PRINTK(2, info,
1289			  "RA: %s, did not accept ra for dev: %s\n",
1290			  __func__, skb->dev->name);
1291		goto skip_linkparms;
1292	}
1293
1294#ifdef CONFIG_IPV6_NDISC_NODETYPE
1295	/* skip link-specific parameters from interior routers */
1296	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1297		ND_PRINTK(2, info,
1298			  "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1299			  __func__, skb->dev->name);
1300		goto skip_linkparms;
1301	}
1302#endif
1303
1304	if (in6_dev->if_flags & IF_RS_SENT) {
1305		/*
1306		 *	flag that an RA was received after an RS was sent
1307		 *	out on this interface.
1308		 */
1309		in6_dev->if_flags |= IF_RA_RCVD;
1310	}
1311
1312	/*
1313	 * Remember the managed/otherconf flags from most recently
1314	 * received RA message (RFC 2462) -- yoshfuji
1315	 */
1316	old_if_flags = in6_dev->if_flags;
1317	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1318				IF_RA_OTHERCONF)) |
1319				(ra_msg->icmph.icmp6_addrconf_managed ?
1320					IF_RA_MANAGED : 0) |
1321				(ra_msg->icmph.icmp6_addrconf_other ?
1322					IF_RA_OTHERCONF : 0);
1323
1324	if (old_if_flags != in6_dev->if_flags)
1325		send_ifinfo_notify = true;
1326
1327	if (!READ_ONCE(in6_dev->cnf.accept_ra_defrtr)) {
1328		ND_PRINTK(2, info,
1329			  "RA: %s, defrtr is false for dev: %s\n",
1330			  __func__, skb->dev->name);
1331		goto skip_defrtr;
1332	}
1333
1334	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1335	if (lifetime != 0 &&
1336	    lifetime < READ_ONCE(in6_dev->cnf.accept_ra_min_lft)) {
1337		ND_PRINTK(2, info,
1338			  "RA: router lifetime (%ds) is too short: %s\n",
1339			  lifetime, skb->dev->name);
1340		goto skip_defrtr;
1341	}
1342
1343	/* Do not accept RA with source-addr found on local machine unless
1344	 * accept_ra_from_local is set to true.
1345	 */
1346	net = dev_net(in6_dev->dev);
1347	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1348	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1349		ND_PRINTK(2, info,
1350			  "RA from local address detected on dev: %s: default router ignored\n",
1351			  skb->dev->name);
1352		goto skip_defrtr;
1353	}
1354
 
 
1355#ifdef CONFIG_IPV6_ROUTER_PREF
1356	pref = ra_msg->icmph.icmp6_router_pref;
1357	/* 10b is handled as if it were 00b (medium) */
1358	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1359	    !READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref))
1360		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1361#endif
1362	/* routes added from RAs do not use nexthop objects */
1363	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1364	if (rt) {
1365		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1366					 rt->fib6_nh->fib_nh_dev, NULL,
1367					  &ipv6_hdr(skb)->saddr);
1368		if (!neigh) {
1369			ND_PRINTK(0, err,
1370				  "RA: %s got default router without neighbour\n",
1371				  __func__);
1372			fib6_info_release(rt);
1373			return reason;
1374		}
1375	}
1376	/* Set default route metric as specified by user */
1377	defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1378	/* delete the route if lifetime is 0 or if metric needs change */
1379	if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1380		ip6_del_rt(net, rt, false);
1381		rt = NULL;
1382	}
1383
1384	ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, metric: %d, for dev: %s\n",
1385		  rt, lifetime, defrtr_usr_metric, skb->dev->name);
1386	if (!rt && lifetime) {
1387		ND_PRINTK(3, info, "RA: adding default router\n");
1388
1389		if (neigh)
1390			neigh_release(neigh);
1391
1392		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1393					 skb->dev, pref, defrtr_usr_metric,
1394					 lifetime);
1395		if (!rt) {
1396			ND_PRINTK(0, err,
1397				  "RA: %s failed to add default route\n",
1398				  __func__);
1399			return reason;
1400		}
1401
1402		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1403					 rt->fib6_nh->fib_nh_dev, NULL,
1404					  &ipv6_hdr(skb)->saddr);
1405		if (!neigh) {
1406			ND_PRINTK(0, err,
1407				  "RA: %s got default router without neighbour\n",
1408				  __func__);
1409			fib6_info_release(rt);
1410			return reason;
1411		}
1412		neigh->flags |= NTF_ROUTER;
1413	} else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1414		struct nl_info nlinfo = {
1415			.nl_net = net,
1416		};
1417		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1418		inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1419	}
1420
1421	if (rt) {
1422		table = rt->fib6_table;
1423		spin_lock_bh(&table->tb6_lock);
1424
1425		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1426		fib6_add_gc_list(rt);
1427
1428		spin_unlock_bh(&table->tb6_lock);
1429	}
1430	if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) < 256 &&
1431	    ra_msg->icmph.icmp6_hop_limit) {
1432		if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) <=
1433		    ra_msg->icmph.icmp6_hop_limit) {
1434			WRITE_ONCE(in6_dev->cnf.hop_limit,
1435				   ra_msg->icmph.icmp6_hop_limit);
1436			fib6_metric_set(rt, RTAX_HOPLIMIT,
1437					ra_msg->icmph.icmp6_hop_limit);
1438		} else {
1439			ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1440		}
1441	}
1442
1443skip_defrtr:
1444
1445	/*
1446	 *	Update Reachable Time and Retrans Timer
1447	 */
1448
1449	if (in6_dev->nd_parms) {
1450		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1451
1452		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1453			rtime = (rtime*HZ)/1000;
1454			if (rtime < HZ/100)
1455				rtime = HZ/100;
1456			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1457			in6_dev->tstamp = jiffies;
1458			send_ifinfo_notify = true;
1459		}
1460
1461		rtime = ntohl(ra_msg->reachable_time);
1462		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1463			rtime = (rtime*HZ)/1000;
1464
1465			if (rtime < HZ/10)
1466				rtime = HZ/10;
1467
1468			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1469				NEIGH_VAR_SET(in6_dev->nd_parms,
1470					      BASE_REACHABLE_TIME, rtime);
1471				NEIGH_VAR_SET(in6_dev->nd_parms,
1472					      GC_STALETIME, 3 * rtime);
1473				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1474				in6_dev->tstamp = jiffies;
1475				send_ifinfo_notify = true;
1476			}
1477		}
1478	}
1479
1480skip_linkparms:
1481
1482	/*
1483	 *	Process options.
1484	 */
1485
1486	if (!neigh)
1487		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1488				       skb->dev, 1);
1489	if (neigh) {
1490		u8 *lladdr = NULL;
1491		if (ndopts.nd_opts_src_lladdr) {
1492			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1493						     skb->dev);
1494			if (!lladdr) {
1495				ND_PRINTK(2, warn,
1496					  "RA: invalid link-layer address length\n");
1497				goto out;
1498			}
1499		}
1500		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1501			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1502			     NEIGH_UPDATE_F_OVERRIDE|
1503			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1504			     NEIGH_UPDATE_F_ISROUTER,
1505			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
1506		reason = SKB_CONSUMED;
1507	}
1508
1509	if (!ipv6_accept_ra(in6_dev)) {
1510		ND_PRINTK(2, info,
1511			  "RA: %s, accept_ra is false for dev: %s\n",
1512			  __func__, skb->dev->name);
1513		goto out;
1514	}
1515
1516#ifdef CONFIG_IPV6_ROUTE_INFO
1517	if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) &&
1518	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1519			  in6_dev->dev, 0)) {
1520		ND_PRINTK(2, info,
1521			  "RA from local address detected on dev: %s: router info ignored.\n",
1522			  skb->dev->name);
1523		goto skip_routeinfo;
1524	}
1525
1526	if (READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref) && ndopts.nd_opts_ri) {
1527		struct nd_opt_hdr *p;
1528		for (p = ndopts.nd_opts_ri;
1529		     p;
1530		     p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) {
1531			struct route_info *ri = (struct route_info *)p;
1532#ifdef CONFIG_IPV6_NDISC_NODETYPE
1533			if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT &&
1534			    ri->prefix_len == 0)
1535				continue;
1536#endif
1537			if (ri->prefix_len == 0 &&
1538			    !READ_ONCE(in6_dev->cnf.accept_ra_defrtr))
1539				continue;
1540			if (ri->lifetime != 0 &&
1541			    ntohl(ri->lifetime) < READ_ONCE(in6_dev->cnf.accept_ra_min_lft))
1542				continue;
1543			if (ri->prefix_len < READ_ONCE(in6_dev->cnf.accept_ra_rt_info_min_plen))
1544				continue;
1545			if (ri->prefix_len > READ_ONCE(in6_dev->cnf.accept_ra_rt_info_max_plen))
1546				continue;
1547			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1548				      &ipv6_hdr(skb)->saddr);
1549		}
1550	}
1551
1552skip_routeinfo:
1553#endif
1554
1555#ifdef CONFIG_IPV6_NDISC_NODETYPE
1556	/* skip link-specific ndopts from interior routers */
1557	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1558		ND_PRINTK(2, info,
1559			  "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1560			  __func__, skb->dev->name);
1561		goto out;
1562	}
1563#endif
1564
1565	if (READ_ONCE(in6_dev->cnf.accept_ra_pinfo) && ndopts.nd_opts_pi) {
1566		struct nd_opt_hdr *p;
1567		for (p = ndopts.nd_opts_pi;
1568		     p;
1569		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1570			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1571					    (p->nd_opt_len) << 3,
1572					    ndopts.nd_opts_src_lladdr != NULL);
1573		}
1574	}
1575
1576	if (ndopts.nd_opts_mtu && READ_ONCE(in6_dev->cnf.accept_ra_mtu)) {
1577		__be32 n;
1578		u32 mtu;
1579
1580		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1581		mtu = ntohl(n);
1582
1583		if (in6_dev->ra_mtu != mtu) {
1584			in6_dev->ra_mtu = mtu;
1585			send_ifinfo_notify = true;
1586		}
1587
1588		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1589			ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1590		} else if (READ_ONCE(in6_dev->cnf.mtu6) != mtu) {
1591			WRITE_ONCE(in6_dev->cnf.mtu6, mtu);
1592			fib6_metric_set(rt, RTAX_MTU, mtu);
1593			rt6_mtu_change(skb->dev, mtu);
1594		}
1595	}
1596
1597	if (ndopts.nd_useropts) {
1598		struct nd_opt_hdr *p;
1599		for (p = ndopts.nd_useropts;
1600		     p;
1601		     p = ndisc_next_useropt(skb->dev, p,
1602					    ndopts.nd_useropts_end)) {
1603			ndisc_ra_useropt(skb, p);
1604		}
1605	}
1606
1607	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1608		ND_PRINTK(2, warn, "RA: invalid RA options\n");
1609	}
1610out:
1611	/* Send a notify if RA changed managed/otherconf flags or
1612	 * timer settings or ra_mtu value
1613	 */
1614	if (send_ifinfo_notify)
1615		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1616
1617	fib6_info_release(rt);
1618	if (neigh)
1619		neigh_release(neigh);
1620	return reason;
1621}
1622
1623static enum skb_drop_reason ndisc_redirect_rcv(struct sk_buff *skb)
1624{
 
 
1625	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1626	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1627				    offsetof(struct rd_msg, opt));
1628	struct ndisc_options ndopts;
1629	SKB_DR(reason);
1630	u8 *hdr;
1631
1632#ifdef CONFIG_IPV6_NDISC_NODETYPE
1633	switch (skb->ndisc_nodetype) {
1634	case NDISC_NODETYPE_HOST:
1635	case NDISC_NODETYPE_NODEFAULT:
1636		ND_PRINTK(2, warn,
1637			  "Redirect: from host or unauthorized router\n");
1638		return reason;
1639	}
1640#endif
1641
1642	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1643		ND_PRINTK(2, warn,
1644			  "Redirect: source address is not link-local\n");
1645		return reason;
1646	}
1647
1648	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1649		return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS;
1650
1651	if (!ndopts.nd_opts_rh) {
1652		ip6_redirect_no_header(skb, dev_net(skb->dev),
1653					skb->dev->ifindex);
1654		return reason;
1655	}
1656
1657	hdr = (u8 *)ndopts.nd_opts_rh;
1658	hdr += 8;
1659	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1660		return SKB_DROP_REASON_PKT_TOO_SMALL;
1661
1662	return icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1663}
1664
1665static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1666					   struct sk_buff *orig_skb,
1667					   int rd_len)
1668{
1669	u8 *opt = skb_put(skb, rd_len);
1670
1671	memset(opt, 0, 8);
1672	*(opt++) = ND_OPT_REDIRECT_HDR;
1673	*(opt++) = (rd_len >> 3);
1674	opt += 6;
1675
1676	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1677		      rd_len - 8);
1678}
1679
1680void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1681{
1682	struct net_device *dev = skb->dev;
1683	struct net *net = dev_net(dev);
1684	struct sock *sk = net->ipv6.ndisc_sk;
1685	int optlen = 0;
1686	struct inet_peer *peer;
1687	struct sk_buff *buff;
1688	struct rd_msg *msg;
1689	struct in6_addr saddr_buf;
1690	struct rt6_info *rt;
1691	struct dst_entry *dst;
1692	struct flowi6 fl6;
1693	int rd_len;
1694	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1695	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1696	bool ret;
1697
1698	if (netif_is_l3_master(skb->dev)) {
1699		dev = dev_get_by_index_rcu(dev_net(skb->dev), IPCB(skb)->iif);
1700		if (!dev)
1701			return;
1702	}
1703
1704	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1705		ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1706			  dev->name);
1707		return;
1708	}
1709
1710	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1711	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1712		ND_PRINTK(2, warn,
1713			  "Redirect: target address is not link-local unicast\n");
1714		return;
1715	}
1716
1717	icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT,
1718			 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex);
1719
1720	dst = ip6_route_output(net, NULL, &fl6);
1721	if (dst->error) {
1722		dst_release(dst);
1723		return;
1724	}
1725	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
1726	if (IS_ERR(dst))
1727		return;
1728
1729	rt = dst_rt6_info(dst);
1730
1731	if (rt->rt6i_flags & RTF_GATEWAY) {
1732		ND_PRINTK(2, warn,
1733			  "Redirect: destination is not a neighbour\n");
1734		goto release;
1735	}
1736
1737	rcu_read_lock();
1738	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr);
1739	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1740	rcu_read_unlock();
1741
1742	if (!ret)
1743		goto release;
1744
1745	if (dev->addr_len) {
1746		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1747		if (!neigh) {
1748			ND_PRINTK(2, warn,
1749				  "Redirect: no neigh for target address\n");
1750			goto release;
1751		}
1752
1753		read_lock_bh(&neigh->lock);
1754		if (neigh->nud_state & NUD_VALID) {
1755			memcpy(ha_buf, neigh->ha, dev->addr_len);
1756			read_unlock_bh(&neigh->lock);
1757			ha = ha_buf;
1758			optlen += ndisc_redirect_opt_addr_space(dev, neigh,
1759								ops_data_buf,
1760								&ops_data);
1761		} else
1762			read_unlock_bh(&neigh->lock);
1763
1764		neigh_release(neigh);
1765	}
1766
1767	rd_len = min_t(unsigned int,
1768		       IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen,
1769		       skb->len + 8);
1770	rd_len &= ~0x7;
1771	optlen += rd_len;
1772
1773	buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
1774	if (!buff)
1775		goto release;
1776
1777	msg = skb_put(buff, sizeof(*msg));
1778	*msg = (struct rd_msg) {
1779		.icmph = {
1780			.icmp6_type = NDISC_REDIRECT,
1781		},
1782		.target = *target,
1783		.dest = ipv6_hdr(skb)->daddr,
1784	};
1785
1786	/*
1787	 *	include target_address option
1788	 */
1789
1790	if (ha)
1791		ndisc_fill_redirect_addr_option(buff, ha, ops_data);
1792
1793	/*
1794	 *	build redirect option and copy skb over to the new packet.
1795	 */
1796
1797	if (rd_len)
1798		ndisc_fill_redirect_hdr_option(buff, skb, rd_len);
1799
1800	skb_dst_set(buff, dst);
1801	ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf);
1802	return;
1803
1804release:
1805	dst_release(dst);
1806}
1807
1808static void pndisc_redo(struct sk_buff *skb)
1809{
1810	enum skb_drop_reason reason = ndisc_recv_ns(skb);
1811
1812	kfree_skb_reason(skb, reason);
1813}
1814
1815static int ndisc_is_multicast(const void *pkey)
1816{
1817	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1818}
1819
1820static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1821{
1822	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1823
1824	if (!idev)
1825		return true;
1826	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1827	    READ_ONCE(idev->cnf.suppress_frag_ndisc)) {
1828		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1829		return true;
1830	}
1831	return false;
1832}
1833
1834enum skb_drop_reason ndisc_rcv(struct sk_buff *skb)
1835{
1836	struct nd_msg *msg;
1837	SKB_DR(reason);
1838
1839	if (ndisc_suppress_frag_ndisc(skb))
1840		return SKB_DROP_REASON_IPV6_NDISC_FRAG;
1841
1842	if (skb_linearize(skb))
1843		return SKB_DROP_REASON_NOMEM;
1844
1845	msg = (struct nd_msg *)skb_transport_header(skb);
1846
1847	__skb_push(skb, skb->data - skb_transport_header(skb));
1848
1849	if (ipv6_hdr(skb)->hop_limit != 255) {
1850		ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1851			  ipv6_hdr(skb)->hop_limit);
1852		return SKB_DROP_REASON_IPV6_NDISC_HOP_LIMIT;
1853	}
1854
1855	if (msg->icmph.icmp6_code != 0) {
1856		ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1857			  msg->icmph.icmp6_code);
1858		return SKB_DROP_REASON_IPV6_NDISC_BAD_CODE;
1859	}
1860
1861	switch (msg->icmph.icmp6_type) {
1862	case NDISC_NEIGHBOUR_SOLICITATION:
1863		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1864		reason = ndisc_recv_ns(skb);
1865		break;
1866
1867	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1868		reason = ndisc_recv_na(skb);
1869		break;
1870
1871	case NDISC_ROUTER_SOLICITATION:
1872		reason = ndisc_recv_rs(skb);
1873		break;
1874
1875	case NDISC_ROUTER_ADVERTISEMENT:
1876		reason = ndisc_router_discovery(skb);
1877		break;
1878
1879	case NDISC_REDIRECT:
1880		reason = ndisc_redirect_rcv(skb);
1881		break;
1882	}
1883
1884	return reason;
1885}
1886
1887static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1888{
1889	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1890	struct netdev_notifier_change_info *change_info;
1891	struct net *net = dev_net(dev);
1892	struct inet6_dev *idev;
1893	bool evict_nocarrier;
1894
1895	switch (event) {
1896	case NETDEV_CHANGEADDR:
1897		neigh_changeaddr(&nd_tbl, dev);
1898		fib6_run_gc(0, net, false);
1899		fallthrough;
1900	case NETDEV_UP:
1901		idev = in6_dev_get(dev);
1902		if (!idev)
1903			break;
1904		if (READ_ONCE(idev->cnf.ndisc_notify) ||
1905		    READ_ONCE(net->ipv6.devconf_all->ndisc_notify))
1906			ndisc_send_unsol_na(dev);
1907		in6_dev_put(idev);
1908		break;
1909	case NETDEV_CHANGE:
1910		idev = in6_dev_get(dev);
1911		if (!idev)
1912			evict_nocarrier = true;
1913		else {
1914			evict_nocarrier = READ_ONCE(idev->cnf.ndisc_evict_nocarrier) &&
1915					  READ_ONCE(net->ipv6.devconf_all->ndisc_evict_nocarrier);
1916			in6_dev_put(idev);
1917		}
1918
1919		change_info = ptr;
1920		if (change_info->flags_changed & IFF_NOARP)
1921			neigh_changeaddr(&nd_tbl, dev);
1922		if (evict_nocarrier && !netif_carrier_ok(dev))
1923			neigh_carrier_down(&nd_tbl, dev);
1924		break;
1925	case NETDEV_DOWN:
1926		neigh_ifdown(&nd_tbl, dev);
1927		fib6_run_gc(0, net, false);
1928		break;
1929	case NETDEV_NOTIFY_PEERS:
1930		ndisc_send_unsol_na(dev);
1931		break;
1932	default:
1933		break;
1934	}
1935
1936	return NOTIFY_DONE;
1937}
1938
1939static struct notifier_block ndisc_netdev_notifier = {
1940	.notifier_call = ndisc_netdev_event,
1941	.priority = ADDRCONF_NOTIFY_PRIORITY - 5,
1942};
1943
1944#ifdef CONFIG_SYSCTL
1945static void ndisc_warn_deprecated_sysctl(const struct ctl_table *ctl,
1946					 const char *func, const char *dev_name)
1947{
1948	static char warncomm[TASK_COMM_LEN];
1949	static int warned;
1950	if (strcmp(warncomm, current->comm) && warned < 5) {
1951		strscpy(warncomm, current->comm);
1952		pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n",
1953			warncomm, func,
1954			dev_name, ctl->procname,
1955			dev_name, ctl->procname);
1956		warned++;
1957	}
1958}
1959
1960int ndisc_ifinfo_sysctl_change(const struct ctl_table *ctl, int write, void *buffer,
1961		size_t *lenp, loff_t *ppos)
1962{
1963	struct net_device *dev = ctl->extra1;
1964	struct inet6_dev *idev;
1965	int ret;
1966
1967	if ((strcmp(ctl->procname, "retrans_time") == 0) ||
1968	    (strcmp(ctl->procname, "base_reachable_time") == 0))
1969		ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default");
1970
1971	if (strcmp(ctl->procname, "retrans_time") == 0)
1972		ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos);
1973
1974	else if (strcmp(ctl->procname, "base_reachable_time") == 0)
1975		ret = neigh_proc_dointvec_jiffies(ctl, write,
1976						  buffer, lenp, ppos);
1977
1978	else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) ||
1979		 (strcmp(ctl->procname, "base_reachable_time_ms") == 0))
1980		ret = neigh_proc_dointvec_ms_jiffies(ctl, write,
1981						     buffer, lenp, ppos);
1982	else
1983		ret = -1;
1984
1985	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
1986		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
1987			idev->nd_parms->reachable_time =
1988					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1989		WRITE_ONCE(idev->tstamp, jiffies);
1990		inet6_ifinfo_notify(RTM_NEWLINK, idev);
1991		in6_dev_put(idev);
1992	}
1993	return ret;
1994}
1995
1996
1997#endif
1998
1999static int __net_init ndisc_net_init(struct net *net)
2000{
2001	struct ipv6_pinfo *np;
2002	struct sock *sk;
2003	int err;
2004
2005	err = inet_ctl_sock_create(&sk, PF_INET6,
2006				   SOCK_RAW, IPPROTO_ICMPV6, net);
2007	if (err < 0) {
2008		ND_PRINTK(0, err,
2009			  "NDISC: Failed to initialize the control socket (err %d)\n",
2010			  err);
2011		return err;
2012	}
2013
2014	net->ipv6.ndisc_sk = sk;
2015
2016	np = inet6_sk(sk);
2017	np->hop_limit = 255;
2018	/* Do not loopback ndisc messages */
2019	inet6_clear_bit(MC6_LOOP, sk);
2020
2021	return 0;
2022}
2023
2024static void __net_exit ndisc_net_exit(struct net *net)
2025{
2026	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
2027}
2028
2029static struct pernet_operations ndisc_net_ops = {
2030	.init = ndisc_net_init,
2031	.exit = ndisc_net_exit,
2032};
2033
2034int __init ndisc_init(void)
2035{
2036	int err;
2037
2038	err = register_pernet_subsys(&ndisc_net_ops);
2039	if (err)
2040		return err;
2041	/*
2042	 * Initialize the neighbour table
2043	 */
2044	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2045
2046#ifdef CONFIG_SYSCTL
2047	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2048				    ndisc_ifinfo_sysctl_change);
2049	if (err)
2050		goto out_unregister_pernet;
2051out:
2052#endif
2053	return err;
2054
2055#ifdef CONFIG_SYSCTL
2056out_unregister_pernet:
2057	unregister_pernet_subsys(&ndisc_net_ops);
2058	goto out;
2059#endif
2060}
2061
2062int __init ndisc_late_init(void)
2063{
2064	return register_netdevice_notifier(&ndisc_netdev_notifier);
2065}
2066
2067void ndisc_late_cleanup(void)
2068{
2069	unregister_netdevice_notifier(&ndisc_netdev_notifier);
2070}
2071
2072void ndisc_cleanup(void)
2073{
2074#ifdef CONFIG_SYSCTL
2075	neigh_sysctl_unregister(&nd_tbl.parms);
2076#endif
2077	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2078	unregister_pernet_subsys(&ndisc_net_ops);
2079}
v6.2
   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		<roque@di.fc.ul.pt>
   8 *	Mike Shaver		<shaver@ingenia.com>
   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_MAX,
 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_RDNSS ||
 
 201		opt->nd_opt_type == ND_OPT_DNSSL ||
 
 202		opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL ||
 203		opt->nd_opt_type == ND_OPT_PREF64 ||
 204		ndisc_ops_is_useropt(dev, opt->nd_opt_type);
 205}
 206
 207static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev,
 208					     struct nd_opt_hdr *cur,
 209					     struct nd_opt_hdr *end)
 210{
 211	if (!cur || !end || cur >= end)
 212		return NULL;
 213	do {
 214		cur = ((void *)cur) + (cur->nd_opt_len << 3);
 215	} while (cur < end && !ndisc_is_useropt(dev, cur));
 216	return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL;
 217}
 218
 219struct ndisc_options *ndisc_parse_options(const struct net_device *dev,
 220					  u8 *opt, int opt_len,
 221					  struct ndisc_options *ndopts)
 222{
 223	struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt;
 224
 225	if (!nd_opt || opt_len < 0 || !ndopts)
 226		return NULL;
 227	memset(ndopts, 0, sizeof(*ndopts));
 228	while (opt_len) {
 
 229		int l;
 230		if (opt_len < sizeof(struct nd_opt_hdr))
 231			return NULL;
 232		l = nd_opt->nd_opt_len << 3;
 233		if (opt_len < l || l == 0)
 234			return NULL;
 235		if (ndisc_ops_parse_options(dev, nd_opt, ndopts))
 236			goto next_opt;
 237		switch (nd_opt->nd_opt_type) {
 238		case ND_OPT_SOURCE_LL_ADDR:
 239		case ND_OPT_TARGET_LL_ADDR:
 240		case ND_OPT_MTU:
 241		case ND_OPT_NONCE:
 242		case ND_OPT_REDIRECT_HDR:
 243			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
 244				ND_PRINTK(2, warn,
 245					  "%s: duplicated ND6 option found: type=%d\n",
 246					  __func__, nd_opt->nd_opt_type);
 247			} else {
 248				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
 249			}
 250			break;
 251		case ND_OPT_PREFIX_INFO:
 252			ndopts->nd_opts_pi_end = nd_opt;
 253			if (!ndopts->nd_opt_array[nd_opt->nd_opt_type])
 254				ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt;
 255			break;
 256#ifdef CONFIG_IPV6_ROUTE_INFO
 257		case ND_OPT_ROUTE_INFO:
 258			ndopts->nd_opts_ri_end = nd_opt;
 259			if (!ndopts->nd_opts_ri)
 260				ndopts->nd_opts_ri = nd_opt;
 261			break;
 262#endif
 263		default:
 264			if (ndisc_is_useropt(dev, nd_opt)) {
 265				ndopts->nd_useropts_end = nd_opt;
 266				if (!ndopts->nd_useropts)
 267					ndopts->nd_useropts = nd_opt;
 268			} else {
 269				/*
 270				 * Unknown options must be silently ignored,
 271				 * to accommodate future extension to the
 272				 * protocol.
 273				 */
 274				ND_PRINTK(2, notice,
 275					  "%s: ignored unsupported option; type=%d, len=%d\n",
 276					  __func__,
 277					  nd_opt->nd_opt_type,
 278					  nd_opt->nd_opt_len);
 279			}
 
 280		}
 281next_opt:
 282		opt_len -= l;
 283		nd_opt = ((void *)nd_opt) + l;
 284	}
 285	return ndopts;
 286}
 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 in6_addr maddr;
 378	struct net_device *dev = n->dev;
 379
 380	if (!dev || !__in6_dev_get(dev))
 381		return -EINVAL;
 382	addrconf_addr_solict_mult(addr, &maddr);
 383	ipv6_dev_mc_inc(dev, &maddr);
 384	return 0;
 385}
 386
 387static void pndisc_destructor(struct pneigh_entry *n)
 388{
 389	struct in6_addr *addr = (struct in6_addr *)&n->key;
 390	struct in6_addr maddr;
 391	struct net_device *dev = n->dev;
 392
 393	if (!dev || !__in6_dev_get(dev))
 394		return;
 395	addrconf_addr_solict_mult(addr, &maddr);
 396	ipv6_dev_mc_dec(dev, &maddr);
 397}
 398
 399/* called with rtnl held */
 400static bool ndisc_allow_add(const struct net_device *dev,
 401			    struct netlink_ext_ack *extack)
 402{
 403	struct inet6_dev *idev = __in6_dev_get(dev);
 404
 405	if (!idev || idev->cnf.disable_ipv6) {
 406		NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device");
 407		return false;
 408	}
 409
 410	return true;
 411}
 412
 413static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
 414				       int len)
 415{
 416	int hlen = LL_RESERVED_SPACE(dev);
 417	int tlen = dev->needed_tailroom;
 418	struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
 419	struct sk_buff *skb;
 420
 421	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
 422	if (!skb) {
 423		ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
 424			  __func__);
 425		return NULL;
 426	}
 427
 428	skb->protocol = htons(ETH_P_IPV6);
 429	skb->dev = dev;
 430
 431	skb_reserve(skb, hlen + sizeof(struct ipv6hdr));
 432	skb_reset_transport_header(skb);
 433
 434	/* Manually assign socket ownership as we avoid calling
 435	 * sock_alloc_send_pskb() to bypass wmem buffer limits
 436	 */
 437	skb_set_owner_w(skb, sk);
 
 
 438
 439	return skb;
 440}
 441
 442static void ip6_nd_hdr(struct sk_buff *skb,
 443		       const struct in6_addr *saddr,
 444		       const struct in6_addr *daddr,
 445		       int hop_limit, int len)
 446{
 447	struct ipv6hdr *hdr;
 448	struct inet6_dev *idev;
 449	unsigned tclass;
 450
 451	rcu_read_lock();
 452	idev = __in6_dev_get(skb->dev);
 453	tclass = idev ? idev->cnf.ndisc_tclass : 0;
 454	rcu_read_unlock();
 455
 456	skb_push(skb, sizeof(*hdr));
 457	skb_reset_network_header(skb);
 458	hdr = ipv6_hdr(skb);
 459
 460	ip6_flow_hdr(hdr, tclass, 0);
 461
 462	hdr->payload_len = htons(len);
 463	hdr->nexthdr = IPPROTO_ICMPV6;
 464	hdr->hop_limit = hop_limit;
 465
 466	hdr->saddr = *saddr;
 467	hdr->daddr = *daddr;
 468}
 469
 470void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr,
 471		    const struct in6_addr *saddr)
 472{
 
 473	struct dst_entry *dst = skb_dst(skb);
 474	struct net *net = dev_net(skb->dev);
 475	struct sock *sk = net->ipv6.ndisc_sk;
 476	struct inet6_dev *idev;
 
 
 477	int err;
 478	struct icmp6hdr *icmp6h = icmp6_hdr(skb);
 479	u8 type;
 480
 481	type = icmp6h->icmp6_type;
 482
 
 
 
 
 483	if (!dst) {
 484		struct flowi6 fl6;
 485		int oif = skb->dev->ifindex;
 486
 487		icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif);
 488		dst = icmp6_dst_alloc(skb->dev, &fl6);
 489		if (IS_ERR(dst)) {
 
 490			kfree_skb(skb);
 491			return;
 492		}
 493
 494		skb_dst_set(skb, dst);
 495	}
 496
 497	icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len,
 498					      IPPROTO_ICMPV6,
 499					      csum_partial(icmp6h,
 500							   skb->len, 0));
 501
 502	ip6_nd_hdr(skb, saddr, daddr, inet6_sk(sk)->hop_limit, skb->len);
 503
 504	rcu_read_lock();
 505	idev = __in6_dev_get(dst->dev);
 506	IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
 507
 508	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
 509		      net, sk, skb, NULL, dst->dev,
 510		      dst_output);
 511	if (!err) {
 512		ICMP6MSGOUT_INC_STATS(net, idev, type);
 513		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
 514	}
 515
 516	rcu_read_unlock();
 517}
 518EXPORT_SYMBOL(ndisc_send_skb);
 519
 520void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
 521		   const struct in6_addr *solicited_addr,
 522		   bool router, bool solicited, bool override, bool inc_opt)
 523{
 524	struct sk_buff *skb;
 525	struct in6_addr tmpaddr;
 526	struct inet6_ifaddr *ifp;
 527	const struct in6_addr *src_addr;
 528	struct nd_msg *msg;
 529	int optlen = 0;
 530
 531	/* for anycast or proxy, solicited_addr != src_addr */
 532	ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1);
 533	if (ifp) {
 534		src_addr = solicited_addr;
 535		if (ifp->flags & IFA_F_OPTIMISTIC)
 536			override = false;
 537		inc_opt |= ifp->idev->cnf.force_tllao;
 538		in6_ifa_put(ifp);
 539	} else {
 540		if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr,
 541				       inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs,
 542				       &tmpaddr))
 543			return;
 544		src_addr = &tmpaddr;
 545	}
 546
 547	if (!dev->addr_len)
 548		inc_opt = false;
 549	if (inc_opt)
 550		optlen += ndisc_opt_addr_space(dev,
 551					       NDISC_NEIGHBOUR_ADVERTISEMENT);
 552
 553	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
 554	if (!skb)
 555		return;
 556
 557	msg = skb_put(skb, sizeof(*msg));
 558	*msg = (struct nd_msg) {
 559		.icmph = {
 560			.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT,
 561			.icmp6_router = router,
 562			.icmp6_solicited = solicited,
 563			.icmp6_override = override,
 564		},
 565		.target = *solicited_addr,
 566	};
 567
 568	if (inc_opt)
 569		ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR,
 570				       dev->dev_addr,
 571				       NDISC_NEIGHBOUR_ADVERTISEMENT);
 572
 573	ndisc_send_skb(skb, daddr, src_addr);
 574}
 575
 576static void ndisc_send_unsol_na(struct net_device *dev)
 577{
 578	struct inet6_dev *idev;
 579	struct inet6_ifaddr *ifa;
 580
 581	idev = in6_dev_get(dev);
 582	if (!idev)
 583		return;
 584
 585	read_lock_bh(&idev->lock);
 586	list_for_each_entry(ifa, &idev->addr_list, if_list) {
 587		/* skip tentative addresses until dad completes */
 588		if (ifa->flags & IFA_F_TENTATIVE &&
 589		    !(ifa->flags & IFA_F_OPTIMISTIC))
 590			continue;
 591
 592		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr,
 593			      /*router=*/ !!idev->cnf.forwarding,
 594			      /*solicited=*/ false, /*override=*/ true,
 595			      /*inc_opt=*/ true);
 596	}
 597	read_unlock_bh(&idev->lock);
 598
 599	in6_dev_put(idev);
 600}
 601
 602struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit,
 603				const struct in6_addr *saddr, u64 nonce)
 604{
 605	int inc_opt = dev->addr_len;
 606	struct sk_buff *skb;
 607	struct nd_msg *msg;
 608	int optlen = 0;
 609
 610	if (!saddr)
 611		return NULL;
 612
 613	if (ipv6_addr_any(saddr))
 614		inc_opt = false;
 615	if (inc_opt)
 616		optlen += ndisc_opt_addr_space(dev,
 617					       NDISC_NEIGHBOUR_SOLICITATION);
 618	if (nonce != 0)
 619		optlen += 8;
 620
 621	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
 622	if (!skb)
 623		return NULL;
 624
 625	msg = skb_put(skb, sizeof(*msg));
 626	*msg = (struct nd_msg) {
 627		.icmph = {
 628			.icmp6_type = NDISC_NEIGHBOUR_SOLICITATION,
 629		},
 630		.target = *solicit,
 631	};
 632
 633	if (inc_opt)
 634		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
 635				       dev->dev_addr,
 636				       NDISC_NEIGHBOUR_SOLICITATION);
 637	if (nonce != 0) {
 638		u8 *opt = skb_put(skb, 8);
 639
 640		opt[0] = ND_OPT_NONCE;
 641		opt[1] = 8 >> 3;
 642		memcpy(opt + 2, &nonce, 6);
 643	}
 644
 645	return skb;
 646}
 647EXPORT_SYMBOL(ndisc_ns_create);
 648
 649void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit,
 650		   const struct in6_addr *daddr, const struct in6_addr *saddr,
 651		   u64 nonce)
 652{
 653	struct in6_addr addr_buf;
 654	struct sk_buff *skb;
 655
 656	if (!saddr) {
 657		if (ipv6_get_lladdr(dev, &addr_buf,
 658				    (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC)))
 659			return;
 660		saddr = &addr_buf;
 661	}
 662
 663	skb = ndisc_ns_create(dev, solicit, saddr, nonce);
 664
 665	if (skb)
 666		ndisc_send_skb(skb, daddr, saddr);
 667}
 668
 669void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr,
 670		   const struct in6_addr *daddr)
 671{
 672	struct sk_buff *skb;
 673	struct rs_msg *msg;
 674	int send_sllao = dev->addr_len;
 675	int optlen = 0;
 676
 677#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
 678	/*
 679	 * According to section 2.2 of RFC 4429, we must not
 680	 * send router solicitations with a sllao from
 681	 * optimistic addresses, but we may send the solicitation
 682	 * if we don't include the sllao.  So here we check
 683	 * if our address is optimistic, and if so, we
 684	 * suppress the inclusion of the sllao.
 685	 */
 686	if (send_sllao) {
 687		struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr,
 688							   dev, 1);
 689		if (ifp) {
 690			if (ifp->flags & IFA_F_OPTIMISTIC)  {
 691				send_sllao = 0;
 692			}
 693			in6_ifa_put(ifp);
 694		} else {
 695			send_sllao = 0;
 696		}
 697	}
 698#endif
 699	if (send_sllao)
 700		optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION);
 701
 702	skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen);
 703	if (!skb)
 704		return;
 705
 706	msg = skb_put(skb, sizeof(*msg));
 707	*msg = (struct rs_msg) {
 708		.icmph = {
 709			.icmp6_type = NDISC_ROUTER_SOLICITATION,
 710		},
 711	};
 712
 713	if (send_sllao)
 714		ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR,
 715				       dev->dev_addr,
 716				       NDISC_ROUTER_SOLICITATION);
 717
 718	ndisc_send_skb(skb, daddr, saddr);
 719}
 720
 721
 722static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb)
 723{
 724	/*
 725	 *	"The sender MUST return an ICMP
 726	 *	 destination unreachable"
 727	 */
 728	dst_link_failure(skb);
 729	kfree_skb(skb);
 730}
 731
 732/* Called with locked neigh: either read or both */
 733
 734static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb)
 735{
 736	struct in6_addr *saddr = NULL;
 737	struct in6_addr mcaddr;
 738	struct net_device *dev = neigh->dev;
 739	struct in6_addr *target = (struct in6_addr *)&neigh->primary_key;
 740	int probes = atomic_read(&neigh->probes);
 741
 742	if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr,
 743					   dev, false, 1,
 744					   IFA_F_TENTATIVE|IFA_F_OPTIMISTIC))
 745		saddr = &ipv6_hdr(skb)->saddr;
 746	probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES);
 747	if (probes < 0) {
 748		if (!(neigh->nud_state & NUD_VALID)) {
 749			ND_PRINTK(1, dbg,
 750				  "%s: trying to ucast probe in NUD_INVALID: %pI6\n",
 751				  __func__, target);
 752		}
 753		ndisc_send_ns(dev, target, target, saddr, 0);
 754	} else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) {
 755		neigh_app_ns(neigh);
 756	} else {
 757		addrconf_addr_solict_mult(target, &mcaddr);
 758		ndisc_send_ns(dev, target, &mcaddr, saddr, 0);
 759	}
 760}
 761
 762static int pndisc_is_router(const void *pkey,
 763			    struct net_device *dev)
 764{
 765	struct pneigh_entry *n;
 766	int ret = -1;
 767
 768	read_lock_bh(&nd_tbl.lock);
 769	n = __pneigh_lookup(&nd_tbl, dev_net(dev), pkey, dev);
 770	if (n)
 771		ret = !!(n->flags & NTF_ROUTER);
 772	read_unlock_bh(&nd_tbl.lock);
 773
 774	return ret;
 775}
 776
 777void ndisc_update(const struct net_device *dev, struct neighbour *neigh,
 778		  const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type,
 779		  struct ndisc_options *ndopts)
 780{
 781	neigh_update(neigh, lladdr, new, flags, 0);
 782	/* report ndisc ops about neighbour update */
 783	ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts);
 784}
 785
 786static void ndisc_recv_ns(struct sk_buff *skb)
 787{
 788	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
 789	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
 790	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
 791	u8 *lladdr = NULL;
 792	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
 793				    offsetof(struct nd_msg, opt));
 794	struct ndisc_options ndopts;
 795	struct net_device *dev = skb->dev;
 796	struct inet6_ifaddr *ifp;
 797	struct inet6_dev *idev = NULL;
 798	struct neighbour *neigh;
 799	int dad = ipv6_addr_any(saddr);
 800	bool inc;
 801	int is_router = -1;
 
 802	u64 nonce = 0;
 
 803
 804	if (skb->len < sizeof(struct nd_msg)) {
 805		ND_PRINTK(2, warn, "NS: packet too short\n");
 806		return;
 807	}
 808
 809	if (ipv6_addr_is_multicast(&msg->target)) {
 810		ND_PRINTK(2, warn, "NS: multicast target address\n");
 811		return;
 812	}
 813
 814	/*
 815	 * RFC2461 7.1.1:
 816	 * DAD has to be destined for solicited node multicast address.
 817	 */
 818	if (dad && !ipv6_addr_is_solict_mult(daddr)) {
 819		ND_PRINTK(2, warn, "NS: bad DAD packet (wrong destination)\n");
 820		return;
 821	}
 822
 823	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
 824		ND_PRINTK(2, warn, "NS: invalid ND options\n");
 825		return;
 826	}
 827
 828	if (ndopts.nd_opts_src_lladdr) {
 829		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
 830		if (!lladdr) {
 831			ND_PRINTK(2, warn,
 832				  "NS: invalid link-layer address length\n");
 833			return;
 834		}
 835
 836		/* RFC2461 7.1.1:
 837		 *	If the IP source address is the unspecified address,
 838		 *	there MUST NOT be source link-layer address option
 839		 *	in the message.
 840		 */
 841		if (dad) {
 842			ND_PRINTK(2, warn,
 843				  "NS: bad DAD packet (link-layer address option)\n");
 844			return;
 845		}
 846	}
 847	if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1)
 848		memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6);
 849
 850	inc = ipv6_addr_is_multicast(daddr);
 851
 852	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
 853	if (ifp) {
 854have_ifp:
 855		if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) {
 856			if (dad) {
 857				if (nonce != 0 && ifp->dad_nonce == nonce) {
 858					u8 *np = (u8 *)&nonce;
 859					/* Matching nonce if looped back */
 860					ND_PRINTK(2, notice,
 861						  "%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n",
 862						  ifp->idev->dev->name,
 863						  &ifp->addr, np);
 864					goto out;
 865				}
 866				/*
 867				 * We are colliding with another node
 868				 * who is doing DAD
 869				 * so fail our DAD process
 870				 */
 871				addrconf_dad_failure(skb, ifp);
 872				return;
 873			} else {
 874				/*
 875				 * This is not a dad solicitation.
 876				 * If we are an optimistic node,
 877				 * we should respond.
 878				 * Otherwise, we should ignore it.
 879				 */
 880				if (!(ifp->flags & IFA_F_OPTIMISTIC))
 881					goto out;
 882			}
 883		}
 884
 885		idev = ifp->idev;
 886	} else {
 887		struct net *net = dev_net(dev);
 888
 889		/* perhaps an address on the master device */
 890		if (netif_is_l3_slave(dev)) {
 891			struct net_device *mdev;
 892
 893			mdev = netdev_master_upper_dev_get_rcu(dev);
 894			if (mdev) {
 895				ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1);
 896				if (ifp)
 897					goto have_ifp;
 898			}
 899		}
 900
 901		idev = in6_dev_get(dev);
 902		if (!idev) {
 903			/* XXX: count this drop? */
 904			return;
 905		}
 906
 907		if (ipv6_chk_acast_addr(net, dev, &msg->target) ||
 908		    (idev->cnf.forwarding &&
 909		     (net->ipv6.devconf_all->proxy_ndp || idev->cnf.proxy_ndp) &&
 
 910		     (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) {
 911			if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) &&
 912			    skb->pkt_type != PACKET_HOST &&
 913			    inc &&
 914			    NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) {
 915				/*
 916				 * for anycast or proxy,
 917				 * sender should delay its response
 918				 * by a random time between 0 and
 919				 * MAX_ANYCAST_DELAY_TIME seconds.
 920				 * (RFC2461) -- yoshfuji
 921				 */
 922				struct sk_buff *n = skb_clone(skb, GFP_ATOMIC);
 923				if (n)
 924					pneigh_enqueue(&nd_tbl, idev->nd_parms, n);
 925				goto out;
 926			}
 927		} else
 
 928			goto out;
 
 929	}
 930
 931	if (is_router < 0)
 932		is_router = idev->cnf.forwarding;
 933
 934	if (dad) {
 935		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target,
 936			      !!is_router, false, (ifp != NULL), true);
 937		goto out;
 938	}
 939
 940	if (inc)
 941		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_mcast);
 942	else
 943		NEIGH_CACHE_STAT_INC(&nd_tbl, rcv_probes_ucast);
 944
 945	/*
 946	 *	update / create cache entry
 947	 *	for the source address
 948	 */
 949	neigh = __neigh_lookup(&nd_tbl, saddr, dev,
 950			       !inc || lladdr || !dev->addr_len);
 951	if (neigh)
 952		ndisc_update(dev, neigh, lladdr, NUD_STALE,
 953			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
 954			     NEIGH_UPDATE_F_OVERRIDE,
 955			     NDISC_NEIGHBOUR_SOLICITATION, &ndopts);
 956	if (neigh || !dev->header_ops) {
 957		ndisc_send_na(dev, saddr, &msg->target, !!is_router,
 958			      true, (ifp != NULL && inc), inc);
 959		if (neigh)
 960			neigh_release(neigh);
 
 961	}
 962
 963out:
 964	if (ifp)
 965		in6_ifa_put(ifp);
 966	else
 967		in6_dev_put(idev);
 
 968}
 969
 970static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
 971{
 972	struct inet6_dev *idev = __in6_dev_get(dev);
 973
 974	switch (idev->cnf.accept_untracked_na) {
 975	case 0: /* Don't accept untracked na (absent in neighbor cache) */
 976		return 0;
 977	case 1: /* Create new entries from na if currently untracked */
 978		return 1;
 979	case 2: /* Create new entries from untracked na only if saddr is in the
 980		 * same subnet as an address configured on the interface that
 981		 * received the na
 982		 */
 983		return !!ipv6_chk_prefix(saddr, dev);
 984	default:
 985		return 0;
 986	}
 987}
 988
 989static void ndisc_recv_na(struct sk_buff *skb)
 990{
 991	struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb);
 992	struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
 993	const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
 994	u8 *lladdr = NULL;
 995	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
 996				    offsetof(struct nd_msg, opt));
 997	struct ndisc_options ndopts;
 998	struct net_device *dev = skb->dev;
 999	struct inet6_dev *idev = __in6_dev_get(dev);
1000	struct inet6_ifaddr *ifp;
1001	struct neighbour *neigh;
 
1002	u8 new_state;
1003
1004	if (skb->len < sizeof(struct nd_msg)) {
1005		ND_PRINTK(2, warn, "NA: packet too short\n");
1006		return;
1007	}
1008
1009	if (ipv6_addr_is_multicast(&msg->target)) {
1010		ND_PRINTK(2, warn, "NA: target address is multicast\n");
1011		return;
1012	}
1013
1014	if (ipv6_addr_is_multicast(daddr) &&
1015	    msg->icmph.icmp6_solicited) {
1016		ND_PRINTK(2, warn, "NA: solicited NA is multicasted\n");
1017		return;
1018	}
1019
1020	/* For some 802.11 wireless deployments (and possibly other networks),
1021	 * there will be a NA proxy and unsolicitd packets are attacks
1022	 * and thus should not be accepted.
1023	 * drop_unsolicited_na takes precedence over accept_untracked_na
1024	 */
1025	if (!msg->icmph.icmp6_solicited && idev &&
1026	    idev->cnf.drop_unsolicited_na)
1027		return;
 
 
 
1028
1029	if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) {
1030		ND_PRINTK(2, warn, "NS: invalid ND option\n");
1031		return;
1032	}
1033	if (ndopts.nd_opts_tgt_lladdr) {
1034		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev);
1035		if (!lladdr) {
1036			ND_PRINTK(2, warn,
1037				  "NA: invalid link-layer address length\n");
1038			return;
1039		}
1040	}
1041	ifp = ipv6_get_ifaddr(dev_net(dev), &msg->target, dev, 1);
1042	if (ifp) {
1043		if (skb->pkt_type != PACKET_LOOPBACK
1044		    && (ifp->flags & IFA_F_TENTATIVE)) {
1045				addrconf_dad_failure(skb, ifp);
1046				return;
1047		}
1048		/* What should we make now? The advertisement
1049		   is invalid, but ndisc specs say nothing
1050		   about it. It could be misconfiguration, or
1051		   an smart proxy agent tries to help us :-)
1052
1053		   We should not print the error if NA has been
1054		   received from loopback - it is just our own
1055		   unsolicited advertisement.
1056		 */
1057		if (skb->pkt_type != PACKET_LOOPBACK)
1058			ND_PRINTK(1, warn,
1059				  "NA: %pM advertised our address %pI6c on %s!\n",
1060				  eth_hdr(skb)->h_source, &ifp->addr, ifp->idev->dev->name);
1061		in6_ifa_put(ifp);
1062		return;
1063	}
1064
1065	neigh = neigh_lookup(&nd_tbl, &msg->target, dev);
1066
1067	/* RFC 9131 updates original Neighbour Discovery RFC 4861.
1068	 * NAs with Target LL Address option without a corresponding
1069	 * entry in the neighbour cache can now create a STALE neighbour
1070	 * cache entry on routers.
1071	 *
1072	 *   entry accept  fwding  solicited        behaviour
1073	 * ------- ------  ------  ---------    ----------------------
1074	 * present      X       X         0     Set state to STALE
1075	 * present      X       X         1     Set state to REACHABLE
1076	 *  absent      0       X         X     Do nothing
1077	 *  absent      1       0         X     Do nothing
1078	 *  absent      1       1         X     Add a new STALE entry
1079	 *
1080	 * Note that we don't do a (daddr == all-routers-mcast) check.
1081	 */
1082	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
1083	if (!neigh && lladdr && idev && idev->cnf.forwarding) {
1084		if (accept_untracked_na(dev, saddr)) {
1085			neigh = neigh_create(&nd_tbl, &msg->target, dev);
1086			new_state = NUD_STALE;
1087		}
1088	}
1089
1090	if (neigh && !IS_ERR(neigh)) {
1091		u8 old_flags = neigh->flags;
1092		struct net *net = dev_net(dev);
1093
1094		if (neigh->nud_state & NUD_FAILED)
1095			goto out;
1096
1097		/*
1098		 * Don't update the neighbor cache entry on a proxy NA from
1099		 * ourselves because either the proxied node is off link or it
1100		 * has already sent a NA to us.
1101		 */
1102		if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) &&
1103		    net->ipv6.devconf_all->forwarding && net->ipv6.devconf_all->proxy_ndp &&
 
1104		    pneigh_lookup(&nd_tbl, net, &msg->target, dev, 0)) {
1105			/* XXX: idev->cnf.proxy_ndp */
1106			goto out;
1107		}
1108
1109		ndisc_update(dev, neigh, lladdr,
1110			     new_state,
1111			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1112			     (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)|
1113			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1114			     (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0),
1115			     NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts);
1116
1117		if ((old_flags & ~neigh->flags) & NTF_ROUTER) {
1118			/*
1119			 * Change: router to host
1120			 */
1121			rt6_clean_tohost(dev_net(dev),  saddr);
1122		}
1123
1124out:
1125		neigh_release(neigh);
1126	}
 
1127}
1128
1129static void ndisc_recv_rs(struct sk_buff *skb)
1130{
1131	struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb);
1132	unsigned long ndoptlen = skb->len - sizeof(*rs_msg);
1133	struct neighbour *neigh;
1134	struct inet6_dev *idev;
1135	const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr;
1136	struct ndisc_options ndopts;
1137	u8 *lladdr = NULL;
 
1138
1139	if (skb->len < sizeof(*rs_msg))
1140		return;
1141
1142	idev = __in6_dev_get(skb->dev);
1143	if (!idev) {
1144		ND_PRINTK(1, err, "RS: can't find in6 device\n");
1145		return;
1146	}
1147
1148	/* Don't accept RS if we're not in router mode */
1149	if (!idev->cnf.forwarding)
1150		goto out;
1151
1152	/*
1153	 * Don't update NCE if src = ::;
1154	 * this implies that the source node has no ip address assigned yet.
1155	 */
1156	if (ipv6_addr_any(saddr))
1157		goto out;
1158
1159	/* Parse ND options */
1160	if (!ndisc_parse_options(skb->dev, rs_msg->opt, ndoptlen, &ndopts)) {
1161		ND_PRINTK(2, notice, "NS: invalid ND option, ignored\n");
1162		goto out;
1163	}
1164
1165	if (ndopts.nd_opts_src_lladdr) {
1166		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1167					     skb->dev);
1168		if (!lladdr)
1169			goto out;
1170	}
1171
1172	neigh = __neigh_lookup(&nd_tbl, saddr, skb->dev, 1);
1173	if (neigh) {
1174		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1175			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1176			     NEIGH_UPDATE_F_OVERRIDE|
1177			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER,
1178			     NDISC_ROUTER_SOLICITATION, &ndopts);
1179		neigh_release(neigh);
 
1180	}
1181out:
1182	return;
1183}
1184
1185static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
1186{
1187	struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
1188	struct sk_buff *skb;
1189	struct nlmsghdr *nlh;
1190	struct nduseroptmsg *ndmsg;
1191	struct net *net = dev_net(ra->dev);
1192	int err;
1193	int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
1194				    + (opt->nd_opt_len << 3));
1195	size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
1196
1197	skb = nlmsg_new(msg_size, GFP_ATOMIC);
1198	if (!skb) {
1199		err = -ENOBUFS;
1200		goto errout;
1201	}
1202
1203	nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
1204	if (!nlh) {
1205		goto nla_put_failure;
1206	}
1207
1208	ndmsg = nlmsg_data(nlh);
1209	ndmsg->nduseropt_family = AF_INET6;
1210	ndmsg->nduseropt_ifindex = ra->dev->ifindex;
1211	ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
1212	ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
1213	ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
1214
1215	memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
1216
1217	if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr))
1218		goto nla_put_failure;
1219	nlmsg_end(skb, nlh);
1220
1221	rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
1222	return;
1223
1224nla_put_failure:
1225	nlmsg_free(skb);
1226	err = -EMSGSIZE;
1227errout:
1228	rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err);
1229}
1230
1231static void ndisc_router_discovery(struct sk_buff *skb)
1232{
1233	struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
 
1234	struct neighbour *neigh = NULL;
 
 
1235	struct inet6_dev *in6_dev;
1236	struct fib6_info *rt = NULL;
1237	u32 defrtr_usr_metric;
 
 
1238	struct net *net;
 
1239	int lifetime;
1240	struct ndisc_options ndopts;
1241	int optlen;
1242	unsigned int pref = 0;
1243	__u32 old_if_flags;
1244	bool send_ifinfo_notify = false;
1245
1246	__u8 *opt = (__u8 *)(ra_msg + 1);
1247
1248	optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) -
1249		sizeof(struct ra_msg);
1250
1251	ND_PRINTK(2, info,
1252		  "RA: %s, dev: %s\n",
1253		  __func__, skb->dev->name);
1254	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1255		ND_PRINTK(2, warn, "RA: source address is not link-local\n");
1256		return;
1257	}
1258	if (optlen < 0) {
1259		ND_PRINTK(2, warn, "RA: packet too short\n");
1260		return;
1261	}
 
 
1262
1263#ifdef CONFIG_IPV6_NDISC_NODETYPE
1264	if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) {
1265		ND_PRINTK(2, warn, "RA: from host or unauthorized router\n");
1266		return;
1267	}
1268#endif
1269
1270	/*
1271	 *	set the RA_RECV flag in the interface
1272	 */
1273
1274	in6_dev = __in6_dev_get(skb->dev);
1275	if (!in6_dev) {
1276		ND_PRINTK(0, err, "RA: can't find inet6 device for %s\n",
1277			  skb->dev->name);
1278		return;
1279	}
1280
1281	if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) {
1282		ND_PRINTK(2, warn, "RA: invalid ND options\n");
1283		return;
1284	}
1285
1286	if (!ipv6_accept_ra(in6_dev)) {
1287		ND_PRINTK(2, info,
1288			  "RA: %s, did not accept ra for dev: %s\n",
1289			  __func__, skb->dev->name);
1290		goto skip_linkparms;
1291	}
1292
1293#ifdef CONFIG_IPV6_NDISC_NODETYPE
1294	/* skip link-specific parameters from interior routers */
1295	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1296		ND_PRINTK(2, info,
1297			  "RA: %s, nodetype is NODEFAULT, dev: %s\n",
1298			  __func__, skb->dev->name);
1299		goto skip_linkparms;
1300	}
1301#endif
1302
1303	if (in6_dev->if_flags & IF_RS_SENT) {
1304		/*
1305		 *	flag that an RA was received after an RS was sent
1306		 *	out on this interface.
1307		 */
1308		in6_dev->if_flags |= IF_RA_RCVD;
1309	}
1310
1311	/*
1312	 * Remember the managed/otherconf flags from most recently
1313	 * received RA message (RFC 2462) -- yoshfuji
1314	 */
1315	old_if_flags = in6_dev->if_flags;
1316	in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED |
1317				IF_RA_OTHERCONF)) |
1318				(ra_msg->icmph.icmp6_addrconf_managed ?
1319					IF_RA_MANAGED : 0) |
1320				(ra_msg->icmph.icmp6_addrconf_other ?
1321					IF_RA_OTHERCONF : 0);
1322
1323	if (old_if_flags != in6_dev->if_flags)
1324		send_ifinfo_notify = true;
1325
1326	if (!in6_dev->cnf.accept_ra_defrtr) {
1327		ND_PRINTK(2, info,
1328			  "RA: %s, defrtr is false for dev: %s\n",
1329			  __func__, skb->dev->name);
1330		goto skip_defrtr;
1331	}
1332
 
 
 
 
 
 
 
 
 
1333	/* Do not accept RA with source-addr found on local machine unless
1334	 * accept_ra_from_local is set to true.
1335	 */
1336	net = dev_net(in6_dev->dev);
1337	if (!in6_dev->cnf.accept_ra_from_local &&
1338	    ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) {
1339		ND_PRINTK(2, info,
1340			  "RA from local address detected on dev: %s: default router ignored\n",
1341			  skb->dev->name);
1342		goto skip_defrtr;
1343	}
1344
1345	lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime);
1346
1347#ifdef CONFIG_IPV6_ROUTER_PREF
1348	pref = ra_msg->icmph.icmp6_router_pref;
1349	/* 10b is handled as if it were 00b (medium) */
1350	if (pref == ICMPV6_ROUTER_PREF_INVALID ||
1351	    !in6_dev->cnf.accept_ra_rtr_pref)
1352		pref = ICMPV6_ROUTER_PREF_MEDIUM;
1353#endif
1354	/* routes added from RAs do not use nexthop objects */
1355	rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev);
1356	if (rt) {
1357		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1358					 rt->fib6_nh->fib_nh_dev, NULL,
1359					  &ipv6_hdr(skb)->saddr);
1360		if (!neigh) {
1361			ND_PRINTK(0, err,
1362				  "RA: %s got default router without neighbour\n",
1363				  __func__);
1364			fib6_info_release(rt);
1365			return;
1366		}
1367	}
1368	/* Set default route metric as specified by user */
1369	defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric;
1370	/* delete the route if lifetime is 0 or if metric needs change */
1371	if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) {
1372		ip6_del_rt(net, rt, false);
1373		rt = NULL;
1374	}
1375
1376	ND_PRINTK(3, info, "RA: rt: %p  lifetime: %d, metric: %d, for dev: %s\n",
1377		  rt, lifetime, defrtr_usr_metric, skb->dev->name);
1378	if (!rt && lifetime) {
1379		ND_PRINTK(3, info, "RA: adding default router\n");
1380
1381		if (neigh)
1382			neigh_release(neigh);
1383
1384		rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr,
1385					 skb->dev, pref, defrtr_usr_metric);
 
1386		if (!rt) {
1387			ND_PRINTK(0, err,
1388				  "RA: %s failed to add default route\n",
1389				  __func__);
1390			return;
1391		}
1392
1393		neigh = ip6_neigh_lookup(&rt->fib6_nh->fib_nh_gw6,
1394					 rt->fib6_nh->fib_nh_dev, NULL,
1395					  &ipv6_hdr(skb)->saddr);
1396		if (!neigh) {
1397			ND_PRINTK(0, err,
1398				  "RA: %s got default router without neighbour\n",
1399				  __func__);
1400			fib6_info_release(rt);
1401			return;
1402		}
1403		neigh->flags |= NTF_ROUTER;
1404	} else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) {
1405		struct nl_info nlinfo = {
1406			.nl_net = net,
1407		};
1408		rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
1409		inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE);
1410	}
1411
1412	if (rt)
 
 
 
1413		fib6_set_expires(rt, jiffies + (HZ * lifetime));
1414	if (in6_dev->cnf.accept_ra_min_hop_limit < 256 &&
 
 
 
 
1415	    ra_msg->icmph.icmp6_hop_limit) {
1416		if (in6_dev->cnf.accept_ra_min_hop_limit <= ra_msg->icmph.icmp6_hop_limit) {
1417			in6_dev->cnf.hop_limit = ra_msg->icmph.icmp6_hop_limit;
 
 
1418			fib6_metric_set(rt, RTAX_HOPLIMIT,
1419					ra_msg->icmph.icmp6_hop_limit);
1420		} else {
1421			ND_PRINTK(2, warn, "RA: Got route advertisement with lower hop_limit than minimum\n");
1422		}
1423	}
1424
1425skip_defrtr:
1426
1427	/*
1428	 *	Update Reachable Time and Retrans Timer
1429	 */
1430
1431	if (in6_dev->nd_parms) {
1432		unsigned long rtime = ntohl(ra_msg->retrans_timer);
1433
1434		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) {
1435			rtime = (rtime*HZ)/1000;
1436			if (rtime < HZ/100)
1437				rtime = HZ/100;
1438			NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime);
1439			in6_dev->tstamp = jiffies;
1440			send_ifinfo_notify = true;
1441		}
1442
1443		rtime = ntohl(ra_msg->reachable_time);
1444		if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) {
1445			rtime = (rtime*HZ)/1000;
1446
1447			if (rtime < HZ/10)
1448				rtime = HZ/10;
1449
1450			if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) {
1451				NEIGH_VAR_SET(in6_dev->nd_parms,
1452					      BASE_REACHABLE_TIME, rtime);
1453				NEIGH_VAR_SET(in6_dev->nd_parms,
1454					      GC_STALETIME, 3 * rtime);
1455				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
1456				in6_dev->tstamp = jiffies;
1457				send_ifinfo_notify = true;
1458			}
1459		}
1460	}
1461
1462skip_linkparms:
1463
1464	/*
1465	 *	Process options.
1466	 */
1467
1468	if (!neigh)
1469		neigh = __neigh_lookup(&nd_tbl, &ipv6_hdr(skb)->saddr,
1470				       skb->dev, 1);
1471	if (neigh) {
1472		u8 *lladdr = NULL;
1473		if (ndopts.nd_opts_src_lladdr) {
1474			lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr,
1475						     skb->dev);
1476			if (!lladdr) {
1477				ND_PRINTK(2, warn,
1478					  "RA: invalid link-layer address length\n");
1479				goto out;
1480			}
1481		}
1482		ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1483			     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1484			     NEIGH_UPDATE_F_OVERRIDE|
1485			     NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1486			     NEIGH_UPDATE_F_ISROUTER,
1487			     NDISC_ROUTER_ADVERTISEMENT, &ndopts);
 
1488	}
1489
1490	if (!ipv6_accept_ra(in6_dev)) {
1491		ND_PRINTK(2, info,
1492			  "RA: %s, accept_ra is false for dev: %s\n",
1493			  __func__, skb->dev->name);
1494		goto out;
1495	}
1496
1497#ifdef CONFIG_IPV6_ROUTE_INFO
1498	if (!in6_dev->cnf.accept_ra_from_local &&
1499	    ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr,
1500			  in6_dev->dev, 0)) {
1501		ND_PRINTK(2, info,
1502			  "RA from local address detected on dev: %s: router info ignored.\n",
1503			  skb->dev->name);
1504		goto skip_routeinfo;
1505	}
1506
1507	if (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			    !in6_dev->cnf.accept_ra_defrtr)
 
 
 
1520				continue;
1521			if (ri->prefix_len < in6_dev->cnf.accept_ra_rt_info_min_plen)
1522				continue;
1523			if (ri->prefix_len > in6_dev->cnf.accept_ra_rt_info_max_plen)
1524				continue;
1525			rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3,
1526				      &ipv6_hdr(skb)->saddr);
1527		}
1528	}
1529
1530skip_routeinfo:
1531#endif
1532
1533#ifdef CONFIG_IPV6_NDISC_NODETYPE
1534	/* skip link-specific ndopts from interior routers */
1535	if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) {
1536		ND_PRINTK(2, info,
1537			  "RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n",
1538			  __func__, skb->dev->name);
1539		goto out;
1540	}
1541#endif
1542
1543	if (in6_dev->cnf.accept_ra_pinfo && ndopts.nd_opts_pi) {
1544		struct nd_opt_hdr *p;
1545		for (p = ndopts.nd_opts_pi;
1546		     p;
1547		     p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) {
1548			addrconf_prefix_rcv(skb->dev, (u8 *)p,
1549					    (p->nd_opt_len) << 3,
1550					    ndopts.nd_opts_src_lladdr != NULL);
1551		}
1552	}
1553
1554	if (ndopts.nd_opts_mtu && in6_dev->cnf.accept_ra_mtu) {
1555		__be32 n;
1556		u32 mtu;
1557
1558		memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu));
1559		mtu = ntohl(n);
1560
1561		if (in6_dev->ra_mtu != mtu) {
1562			in6_dev->ra_mtu = mtu;
1563			send_ifinfo_notify = true;
1564		}
1565
1566		if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) {
1567			ND_PRINTK(2, warn, "RA: invalid mtu: %d\n", mtu);
1568		} else if (in6_dev->cnf.mtu6 != mtu) {
1569			in6_dev->cnf.mtu6 = mtu;
1570			fib6_metric_set(rt, RTAX_MTU, mtu);
1571			rt6_mtu_change(skb->dev, mtu);
1572		}
1573	}
1574
1575	if (ndopts.nd_useropts) {
1576		struct nd_opt_hdr *p;
1577		for (p = ndopts.nd_useropts;
1578		     p;
1579		     p = ndisc_next_useropt(skb->dev, p,
1580					    ndopts.nd_useropts_end)) {
1581			ndisc_ra_useropt(skb, p);
1582		}
1583	}
1584
1585	if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
1586		ND_PRINTK(2, warn, "RA: invalid RA options\n");
1587	}
1588out:
1589	/* Send a notify if RA changed managed/otherconf flags or
1590	 * timer settings or ra_mtu value
1591	 */
1592	if (send_ifinfo_notify)
1593		inet6_ifinfo_notify(RTM_NEWLINK, in6_dev);
1594
1595	fib6_info_release(rt);
1596	if (neigh)
1597		neigh_release(neigh);
 
1598}
1599
1600static void ndisc_redirect_rcv(struct sk_buff *skb)
1601{
1602	u8 *hdr;
1603	struct ndisc_options ndopts;
1604	struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb);
1605	u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) +
1606				    offsetof(struct rd_msg, opt));
 
 
 
1607
1608#ifdef CONFIG_IPV6_NDISC_NODETYPE
1609	switch (skb->ndisc_nodetype) {
1610	case NDISC_NODETYPE_HOST:
1611	case NDISC_NODETYPE_NODEFAULT:
1612		ND_PRINTK(2, warn,
1613			  "Redirect: from host or unauthorized router\n");
1614		return;
1615	}
1616#endif
1617
1618	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) {
1619		ND_PRINTK(2, warn,
1620			  "Redirect: source address is not link-local\n");
1621		return;
1622	}
1623
1624	if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts))
1625		return;
1626
1627	if (!ndopts.nd_opts_rh) {
1628		ip6_redirect_no_header(skb, dev_net(skb->dev),
1629					skb->dev->ifindex);
1630		return;
1631	}
1632
1633	hdr = (u8 *)ndopts.nd_opts_rh;
1634	hdr += 8;
1635	if (!pskb_pull(skb, hdr - skb_transport_header(skb)))
1636		return;
1637
1638	icmpv6_notify(skb, NDISC_REDIRECT, 0, 0);
1639}
1640
1641static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb,
1642					   struct sk_buff *orig_skb,
1643					   int rd_len)
1644{
1645	u8 *opt = skb_put(skb, rd_len);
1646
1647	memset(opt, 0, 8);
1648	*(opt++) = ND_OPT_REDIRECT_HDR;
1649	*(opt++) = (rd_len >> 3);
1650	opt += 6;
1651
1652	skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt,
1653		      rd_len - 8);
1654}
1655
1656void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target)
1657{
1658	struct net_device *dev = skb->dev;
1659	struct net *net = dev_net(dev);
1660	struct sock *sk = net->ipv6.ndisc_sk;
1661	int optlen = 0;
1662	struct inet_peer *peer;
1663	struct sk_buff *buff;
1664	struct rd_msg *msg;
1665	struct in6_addr saddr_buf;
1666	struct rt6_info *rt;
1667	struct dst_entry *dst;
1668	struct flowi6 fl6;
1669	int rd_len;
1670	u8 ha_buf[MAX_ADDR_LEN], *ha = NULL,
1671	   ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL;
1672	bool ret;
1673
1674	if (netif_is_l3_master(skb->dev)) {
1675		dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
1676		if (!dev)
1677			return;
1678	}
1679
1680	if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) {
1681		ND_PRINTK(2, warn, "Redirect: no link-local address on %s\n",
1682			  dev->name);
1683		return;
1684	}
1685
1686	if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) &&
1687	    ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1688		ND_PRINTK(2, warn,
1689			  "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 = (struct rt6_info *) dst;
1706
1707	if (rt->rt6i_flags & RTF_GATEWAY) {
1708		ND_PRINTK(2, warn,
1709			  "Redirect: destination is not a neighbour\n");
1710		goto release;
1711	}
1712	peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr, 1);
 
 
1713	ret = inet_peer_xrlim_allow(peer, 1*HZ);
1714	if (peer)
1715		inet_putpeer(peer);
1716	if (!ret)
1717		goto release;
1718
1719	if (dev->addr_len) {
1720		struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target);
1721		if (!neigh) {
1722			ND_PRINTK(2, warn,
1723				  "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	ndisc_recv_ns(skb);
1785	kfree_skb(skb);
 
1786}
1787
1788static int ndisc_is_multicast(const void *pkey)
1789{
1790	return ipv6_addr_is_multicast((struct in6_addr *)pkey);
1791}
1792
1793static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb)
1794{
1795	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1796
1797	if (!idev)
1798		return true;
1799	if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED &&
1800	    idev->cnf.suppress_frag_ndisc) {
1801		net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n");
1802		return true;
1803	}
1804	return false;
1805}
1806
1807int ndisc_rcv(struct sk_buff *skb)
1808{
1809	struct nd_msg *msg;
 
1810
1811	if (ndisc_suppress_frag_ndisc(skb))
1812		return 0;
1813
1814	if (skb_linearize(skb))
1815		return 0;
1816
1817	msg = (struct nd_msg *)skb_transport_header(skb);
1818
1819	__skb_push(skb, skb->data - skb_transport_header(skb));
1820
1821	if (ipv6_hdr(skb)->hop_limit != 255) {
1822		ND_PRINTK(2, warn, "NDISC: invalid hop-limit: %d\n",
1823			  ipv6_hdr(skb)->hop_limit);
1824		return 0;
1825	}
1826
1827	if (msg->icmph.icmp6_code != 0) {
1828		ND_PRINTK(2, warn, "NDISC: invalid ICMPv6 code: %d\n",
1829			  msg->icmph.icmp6_code);
1830		return 0;
1831	}
1832
1833	switch (msg->icmph.icmp6_type) {
1834	case NDISC_NEIGHBOUR_SOLICITATION:
1835		memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb));
1836		ndisc_recv_ns(skb);
1837		break;
1838
1839	case NDISC_NEIGHBOUR_ADVERTISEMENT:
1840		ndisc_recv_na(skb);
1841		break;
1842
1843	case NDISC_ROUTER_SOLICITATION:
1844		ndisc_recv_rs(skb);
1845		break;
1846
1847	case NDISC_ROUTER_ADVERTISEMENT:
1848		ndisc_router_discovery(skb);
1849		break;
1850
1851	case NDISC_REDIRECT:
1852		ndisc_redirect_rcv(skb);
1853		break;
1854	}
1855
1856	return 0;
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 (idev->cnf.ndisc_notify ||
1877		    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 = idev->cnf.ndisc_evict_nocarrier &&
1887					  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(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		strcpy(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(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(idev->nd_parms, BASE_REACHABLE_TIME))
1959			idev->nd_parms->reachable_time =
1960					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
1961		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		ND_PRINTK(0, err,
1981			  "NDISC: Failed to initialize the control socket (err %d)\n",
1982			  err);
1983		return err;
1984	}
1985
1986	net->ipv6.ndisc_sk = sk;
1987
1988	np = inet6_sk(sk);
1989	np->hop_limit = 255;
1990	/* Do not loopback ndisc messages */
1991	np->mc_loop = 0;
1992
1993	return 0;
1994}
1995
1996static void __net_exit ndisc_net_exit(struct net *net)
1997{
1998	inet_ctl_sock_destroy(net->ipv6.ndisc_sk);
1999}
2000
2001static struct pernet_operations ndisc_net_ops = {
2002	.init = ndisc_net_init,
2003	.exit = ndisc_net_exit,
2004};
2005
2006int __init ndisc_init(void)
2007{
2008	int err;
2009
2010	err = register_pernet_subsys(&ndisc_net_ops);
2011	if (err)
2012		return err;
2013	/*
2014	 * Initialize the neighbour table
2015	 */
2016	neigh_table_init(NEIGH_ND_TABLE, &nd_tbl);
2017
2018#ifdef CONFIG_SYSCTL
2019	err = neigh_sysctl_register(NULL, &nd_tbl.parms,
2020				    ndisc_ifinfo_sysctl_change);
2021	if (err)
2022		goto out_unregister_pernet;
2023out:
2024#endif
2025	return err;
2026
2027#ifdef CONFIG_SYSCTL
2028out_unregister_pernet:
2029	unregister_pernet_subsys(&ndisc_net_ops);
2030	goto out;
2031#endif
2032}
2033
2034int __init ndisc_late_init(void)
2035{
2036	return register_netdevice_notifier(&ndisc_netdev_notifier);
2037}
2038
2039void ndisc_late_cleanup(void)
2040{
2041	unregister_netdevice_notifier(&ndisc_netdev_notifier);
2042}
2043
2044void ndisc_cleanup(void)
2045{
2046#ifdef CONFIG_SYSCTL
2047	neigh_sysctl_unregister(&nd_tbl.parms);
2048#endif
2049	neigh_table_clear(NEIGH_ND_TABLE, &nd_tbl);
2050	unregister_pernet_subsys(&ndisc_net_ops);
2051}