datagram.c

   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	common UDP/RAW code
   4 *	Linux INET6 implementation
   5 *
   6 *	Authors:
   7 *	Pedro Roque		<roque@di.fc.ul.pt>
   8 */
   9
  10#include <linux/capability.h>
  11#include <linux/errno.h>
  12#include <linux/types.h>
  13#include <linux/kernel.h>
  14#include <linux/interrupt.h>
  15#include <linux/socket.h>
  16#include <linux/sockios.h>
  17#include <linux/in6.h>
  18#include <linux/ipv6.h>
  19#include <linux/route.h>
  20#include <linux/slab.h>
  21#include <linux/export.h>
  22#include <linux/icmp.h>
  23
  24#include <net/ipv6.h>
  25#include <net/ndisc.h>
  26#include <net/addrconf.h>
  27#include <net/transp_v6.h>
  28#include <net/ip6_route.h>
  29#include <net/tcp_states.h>
  30#include <net/dsfield.h>
  31#include <net/sock_reuseport.h>
  32
  33#include <linux/errqueue.h>
  34#include <linux/uaccess.h>
  35
  36static bool ipv6_mapped_addr_any(const struct in6_addr *a)
  37{
  38	return ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0);
  39}
  40
  41static void ip6_datagram_flow_key_init(struct flowi6 *fl6,
  42				       const struct sock *sk)
  43{
  44	const struct inet_sock *inet = inet_sk(sk);
  45	const struct ipv6_pinfo *np = inet6_sk(sk);
  46	int oif = sk->sk_bound_dev_if;
  47
  48	memset(fl6, 0, sizeof(*fl6));
  49	fl6->flowi6_proto = sk->sk_protocol;
  50	fl6->daddr = sk->sk_v6_daddr;
  51	fl6->saddr = np->saddr;
  52	fl6->flowi6_mark = sk->sk_mark;
  53	fl6->fl6_dport = inet->inet_dport;
  54	fl6->fl6_sport = inet->inet_sport;
  55	fl6->flowlabel = ip6_make_flowinfo(np->tclass, np->flow_label);
  56	fl6->flowi6_uid = sk_uid(sk);
  57
  58	if (!oif)
  59		oif = np->sticky_pktinfo.ipi6_ifindex;
  60
  61	if (!oif) {
  62		if (ipv6_addr_is_multicast(&fl6->daddr))
  63			oif = READ_ONCE(np->mcast_oif);
  64		else
  65			oif = READ_ONCE(np->ucast_oif);
  66	}
  67
  68	fl6->flowi6_oif = oif;
  69	security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6));
  70}
  71
  72int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr)
  73{
  74	struct ip6_flowlabel *flowlabel = NULL;
  75	struct in6_addr *final_p, final;
  76	struct ipv6_txoptions *opt;
  77	struct dst_entry *dst;
  78	struct inet_sock *inet = inet_sk(sk);
  79	struct ipv6_pinfo *np = inet6_sk(sk);
  80	struct flowi6 fl6;
  81	int err = 0;
  82
  83	if (inet6_test_bit(SNDFLOW, sk) &&
  84	    (np->flow_label & IPV6_FLOWLABEL_MASK)) {
  85		flowlabel = fl6_sock_lookup(sk, np->flow_label);
  86		if (IS_ERR(flowlabel))
  87			return -EINVAL;
  88	}
  89	ip6_datagram_flow_key_init(&fl6, sk);
  90
  91	rcu_read_lock();
  92	opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt);
  93	final_p = fl6_update_dst(&fl6, opt, &final);
  94	rcu_read_unlock();
  95
  96	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
  97	if (IS_ERR(dst)) {
  98		err = PTR_ERR(dst);
  99		goto out;
 100	}
 101
 102	if (fix_sk_saddr) {
 103		if (ipv6_addr_any(&np->saddr))
 104			np->saddr = fl6.saddr;
 105
 106		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
 107			sk->sk_v6_rcv_saddr = fl6.saddr;
 108			inet->inet_rcv_saddr = LOOPBACK4_IPV6;
 109			if (sk->sk_prot->rehash)
 110				sk->sk_prot->rehash(sk);
 111		}
 112	}
 113
 114	ip6_sk_dst_store_flow(sk, dst, &fl6);
 115
 116out:
 117	fl6_sock_release(flowlabel);
 118	return err;
 119}
 120
 121void ip6_datagram_release_cb(struct sock *sk)
 122{
 123	struct dst_entry *dst;
 124
 125	if (ipv6_addr_v4mapped(&sk->sk_v6_daddr))
 126		return;
 127
 128	rcu_read_lock();
 129	dst = __sk_dst_get(sk);
 130	if (!dst || !READ_ONCE(dst->obsolete) ||
 131	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) {
 132		rcu_read_unlock();
 133		return;
 134	}
 135	rcu_read_unlock();
 136
 137	ip6_datagram_dst_update(sk, false);
 138}
 139EXPORT_SYMBOL_GPL(ip6_datagram_release_cb);
 140
 141int __ip6_datagram_connect(struct sock *sk, struct sockaddr_unsized *uaddr,
 142			   int addr_len)
 143{
 144	struct sockaddr_in6	*usin = (struct sockaddr_in6 *) uaddr;
 145	struct inet_sock	*inet = inet_sk(sk);
 146	struct ipv6_pinfo	*np = inet6_sk(sk);
 147	struct in6_addr		*daddr, old_daddr;
 148	__be32			fl6_flowlabel = 0;
 149	__be32			old_fl6_flowlabel;
 150	__be16			old_dport;
 151	int			addr_type;
 152	int			err;
 153
 154	if (usin->sin6_family == AF_INET) {
 155		if (ipv6_only_sock(sk))
 156			return -EAFNOSUPPORT;
 157		err = __ip4_datagram_connect(sk, uaddr, addr_len);
 158		goto ipv4_connected;
 159	}
 160
 161	if (addr_len < SIN6_LEN_RFC2133)
 162		return -EINVAL;
 163
 164	if (usin->sin6_family != AF_INET6)
 165		return -EAFNOSUPPORT;
 166
 167	if (inet6_test_bit(SNDFLOW, sk))
 168		fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
 169
 170	if (ipv6_addr_any(&usin->sin6_addr)) {
 171		/*
 172		 *	connect to self
 173		 */
 174		if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr))
 175			ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
 176					       &usin->sin6_addr);
 177		else
 178			usin->sin6_addr = in6addr_loopback;
 179	}
 180
 181	addr_type = ipv6_addr_type(&usin->sin6_addr);
 182
 183	daddr = &usin->sin6_addr;
 184
 185	if (addr_type & IPV6_ADDR_MAPPED) {
 186		struct sockaddr_in sin;
 187
 188		if (ipv6_only_sock(sk)) {
 189			err = -ENETUNREACH;
 190			goto out;
 191		}
 192		sin.sin_family = AF_INET;
 193		sin.sin_addr.s_addr = daddr->s6_addr32[3];
 194		sin.sin_port = usin->sin6_port;
 195
 196		err = __ip4_datagram_connect(sk,
 197					     (struct sockaddr_unsized *)&sin,
 198					     sizeof(sin));
 199
 200ipv4_connected:
 201		if (err)
 202			goto out;
 203
 204		ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr);
 205
 206		if (ipv6_addr_any(&np->saddr) ||
 207		    ipv6_mapped_addr_any(&np->saddr))
 208			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
 209
 210		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
 211		    ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) {
 212			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
 213					       &sk->sk_v6_rcv_saddr);
 214			if (sk->sk_prot->rehash)
 215				sk->sk_prot->rehash(sk);
 216		}
 217
 218		goto out;
 219	}
 220
 221	if (__ipv6_addr_needs_scope_id(addr_type)) {
 222		if (addr_len >= sizeof(struct sockaddr_in6) &&
 223		    usin->sin6_scope_id) {
 224			if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) {
 225				err = -EINVAL;
 226				goto out;
 227			}
 228			WRITE_ONCE(sk->sk_bound_dev_if, usin->sin6_scope_id);
 229		}
 230
 231		if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
 232			WRITE_ONCE(sk->sk_bound_dev_if, READ_ONCE(np->mcast_oif));
 233
 234		/* Connect to link-local address requires an interface */
 235		if (!sk->sk_bound_dev_if) {
 236			err = -EINVAL;
 237			goto out;
 238		}
 239	}
 240
 241	/* save the current peer information before updating it */
 242	old_daddr = sk->sk_v6_daddr;
 243	old_fl6_flowlabel = np->flow_label;
 244	old_dport = inet->inet_dport;
 245
 246	sk->sk_v6_daddr = *daddr;
 247	np->flow_label = fl6_flowlabel;
 248	inet->inet_dport = usin->sin6_port;
 249
 250	/*
 251	 *	Check for a route to destination an obtain the
 252	 *	destination cache for it.
 253	 */
 254
 255	err = ip6_datagram_dst_update(sk, true);
 256	if (err) {
 257		/* Restore the socket peer info, to keep it consistent with
 258		 * the old socket state
 259		 */
 260		sk->sk_v6_daddr = old_daddr;
 261		np->flow_label = old_fl6_flowlabel;
 262		inet->inet_dport = old_dport;
 263		goto out;
 264	}
 265
 266	reuseport_has_conns_set(sk);
 267	sk->sk_state = TCP_ESTABLISHED;
 268	sk_set_txhash(sk);
 269out:
 270	return err;
 271}
 272EXPORT_SYMBOL_GPL(__ip6_datagram_connect);
 273
 274int ip6_datagram_connect(struct sock *sk, struct sockaddr_unsized *uaddr, int addr_len)
 275{
 276	int res;
 277
 278	lock_sock(sk);
 279	res = __ip6_datagram_connect(sk, uaddr, addr_len);
 280	release_sock(sk);
 281	return res;
 282}
 283EXPORT_SYMBOL_GPL(ip6_datagram_connect);
 284
 285int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr_unsized *uaddr,
 286				 int addr_len)
 287{
 288	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr);
 289	if (sin6->sin6_family != AF_INET6)
 290		return -EAFNOSUPPORT;
 291	return ip6_datagram_connect(sk, uaddr, addr_len);
 292}
 293EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only);
 294
 295static void ipv6_icmp_error_rfc4884(const struct sk_buff *skb,
 296				    struct sock_ee_data_rfc4884 *out)
 297{
 298	switch (icmp6_hdr(skb)->icmp6_type) {
 299	case ICMPV6_TIME_EXCEED:
 300	case ICMPV6_DEST_UNREACH:
 301		ip_icmp_error_rfc4884(skb, out, sizeof(struct icmp6hdr),
 302				      icmp6_hdr(skb)->icmp6_datagram_len * 8);
 303	}
 304}
 305
 306void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
 307		     __be16 port, u32 info, u8 *payload)
 308{
 309	struct icmp6hdr *icmph = icmp6_hdr(skb);
 310	struct sock_exterr_skb *serr;
 311
 312	if (!inet6_test_bit(RECVERR6, sk))
 313		return;
 314
 315	skb = skb_clone(skb, GFP_ATOMIC);
 316	if (!skb)
 317		return;
 318
 319	skb->protocol = htons(ETH_P_IPV6);
 320
 321	serr = SKB_EXT_ERR(skb);
 322	serr->ee.ee_errno = err;
 323	serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
 324	serr->ee.ee_type = icmph->icmp6_type;
 325	serr->ee.ee_code = icmph->icmp6_code;
 326	serr->ee.ee_pad = 0;
 327	serr->ee.ee_info = info;
 328	serr->ee.ee_data = 0;
 329	serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
 330				  skb_network_header(skb);
 331	serr->port = port;
 332
 333	__skb_pull(skb, payload - skb->data);
 334
 335	if (inet6_test_bit(RECVERR6_RFC4884, sk))
 336		ipv6_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
 337
 338	skb_reset_transport_header(skb);
 339
 340	if (sock_queue_err_skb(sk, skb))
 341		kfree_skb(skb);
 342}
 343EXPORT_SYMBOL_GPL(ipv6_icmp_error);
 344
 345void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info)
 346{
 347	struct sock_exterr_skb *serr;
 348	struct ipv6hdr *iph;
 349	struct sk_buff *skb;
 350
 351	if (!inet6_test_bit(RECVERR6, sk))
 352		return;
 353
 354	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
 355	if (!skb)
 356		return;
 357
 358	skb->protocol = htons(ETH_P_IPV6);
 359
 360	skb_put(skb, sizeof(struct ipv6hdr));
 361	skb_reset_network_header(skb);
 362	iph = ipv6_hdr(skb);
 363	iph->daddr = fl6->daddr;
 364	ip6_flow_hdr(iph, 0, 0);
 365
 366	serr = SKB_EXT_ERR(skb);
 367	serr->ee.ee_errno = err;
 368	serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
 369	serr->ee.ee_type = 0;
 370	serr->ee.ee_code = 0;
 371	serr->ee.ee_pad = 0;
 372	serr->ee.ee_info = info;
 373	serr->ee.ee_data = 0;
 374	serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
 375	serr->port = fl6->fl6_dport;
 376
 377	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
 378	skb_reset_transport_header(skb);
 379
 380	if (sock_queue_err_skb(sk, skb))
 381		kfree_skb(skb);
 382}
 383
 384void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu)
 385{
 386	struct ipv6_pinfo *np = inet6_sk(sk);
 387	struct ipv6hdr *iph;
 388	struct sk_buff *skb;
 389	struct ip6_mtuinfo *mtu_info;
 390
 391	if (!np->rxopt.bits.rxpmtu)
 392		return;
 393
 394	skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
 395	if (!skb)
 396		return;
 397
 398	skb_put(skb, sizeof(struct ipv6hdr));
 399	skb_reset_network_header(skb);
 400	iph = ipv6_hdr(skb);
 401	iph->daddr = fl6->daddr;
 402
 403	mtu_info = IP6CBMTU(skb);
 404
 405	mtu_info->ip6m_mtu = mtu;
 406	mtu_info->ip6m_addr.sin6_family = AF_INET6;
 407	mtu_info->ip6m_addr.sin6_port = 0;
 408	mtu_info->ip6m_addr.sin6_flowinfo = 0;
 409	mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif;
 410	mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr;
 411
 412	__skb_pull(skb, skb_tail_pointer(skb) - skb->data);
 413	skb_reset_transport_header(skb);
 414
 415	skb = xchg(&np->rxpmtu, skb);
 416	kfree_skb(skb);
 417}
 418
 419/* For some errors we have valid addr_offset even with zero payload and
 420 * zero port. Also, addr_offset should be supported if port is set.
 421 */
 422static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr)
 423{
 424	return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 ||
 425	       serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
 426	       serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
 427}
 428
 429/* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL.
 430 *
 431 * At one point, excluding local errors was a quick test to identify icmp/icmp6
 432 * errors. This is no longer true, but the test remained, so the v6 stack,
 433 * unlike v4, also honors cmsg requests on all wifi and timestamp errors.
 434 */
 435static bool ip6_datagram_support_cmsg(struct sk_buff *skb,
 436				      struct sock_exterr_skb *serr)
 437{
 438	if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
 439	    serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6)
 440		return true;
 441
 442	if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL)
 443		return false;
 444
 445	if (!IP6CB(skb)->iif)
 446		return false;
 447
 448	return true;
 449}
 450
 451/*
 452 *	Handle MSG_ERRQUEUE
 453 */
 454int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 455{
 456	struct ipv6_pinfo *np = inet6_sk(sk);
 457	struct sock_exterr_skb *serr;
 458	struct sk_buff *skb;
 459	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
 460	struct {
 461		struct sock_extended_err ee;
 462		struct sockaddr_in6	 offender;
 463	} errhdr;
 464	int err;
 465	int copied;
 466
 467	err = -EAGAIN;
 468	skb = sock_dequeue_err_skb(sk);
 469	if (!skb)
 470		goto out;
 471
 472	copied = skb->len;
 473	if (copied > len) {
 474		msg->msg_flags |= MSG_TRUNC;
 475		copied = len;
 476	}
 477	err = skb_copy_datagram_msg(skb, 0, msg, copied);
 478	if (unlikely(err)) {
 479		kfree_skb(skb);
 480		return err;
 481	}
 482	sock_recv_timestamp(msg, sk, skb);
 483
 484	serr = SKB_EXT_ERR(skb);
 485
 486	if (sin && ipv6_datagram_support_addr(serr)) {
 487		const unsigned char *nh = skb_network_header(skb);
 488		sin->sin6_family = AF_INET6;
 489		sin->sin6_flowinfo = 0;
 490		sin->sin6_port = serr->port;
 491		if (skb->protocol == htons(ETH_P_IPV6)) {
 492			const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset),
 493								  struct ipv6hdr, daddr);
 494			sin->sin6_addr = ip6h->daddr;
 495			if (inet6_test_bit(SNDFLOW, sk))
 496				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
 497			sin->sin6_scope_id =
 498				ipv6_iface_scope_id(&sin->sin6_addr,
 499						    IP6CB(skb)->iif);
 500		} else {
 501			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
 502					       &sin->sin6_addr);
 503			sin->sin6_scope_id = 0;
 504		}
 505		*addr_len = sizeof(*sin);
 506	}
 507
 508	memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
 509	sin = &errhdr.offender;
 510	memset(sin, 0, sizeof(*sin));
 511
 512	if (ip6_datagram_support_cmsg(skb, serr)) {
 513		sin->sin6_family = AF_INET6;
 514		if (np->rxopt.all)
 515			ip6_datagram_recv_common_ctl(sk, msg, skb);
 516		if (skb->protocol == htons(ETH_P_IPV6)) {
 517			sin->sin6_addr = ipv6_hdr(skb)->saddr;
 518			if (np->rxopt.all)
 519				ip6_datagram_recv_specific_ctl(sk, msg, skb);
 520			sin->sin6_scope_id =
 521				ipv6_iface_scope_id(&sin->sin6_addr,
 522						    IP6CB(skb)->iif);
 523		} else {
 524			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
 525					       &sin->sin6_addr);
 526			if (inet_cmsg_flags(inet_sk(sk)))
 527				ip_cmsg_recv(msg, skb);
 528		}
 529	}
 530
 531	put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
 532
 533	/* Now we could try to dump offended packet options */
 534
 535	msg->msg_flags |= MSG_ERRQUEUE;
 536	err = copied;
 537
 538	consume_skb(skb);
 539out:
 540	return err;
 541}
 542EXPORT_SYMBOL_GPL(ipv6_recv_error);
 543
 544/*
 545 *	Handle IPV6_RECVPATHMTU
 546 */
 547int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
 548		     int *addr_len)
 549{
 550	struct ipv6_pinfo *np = inet6_sk(sk);
 551	struct sk_buff *skb;
 552	struct ip6_mtuinfo mtu_info;
 553	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name);
 554	int err;
 555	int copied;
 556
 557	err = -EAGAIN;
 558	skb = xchg(&np->rxpmtu, NULL);
 559	if (!skb)
 560		goto out;
 561
 562	copied = skb->len;
 563	if (copied > len) {
 564		msg->msg_flags |= MSG_TRUNC;
 565		copied = len;
 566	}
 567	err = skb_copy_datagram_msg(skb, 0, msg, copied);
 568	if (err)
 569		goto out_free_skb;
 570
 571	sock_recv_timestamp(msg, sk, skb);
 572
 573	memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info));
 574
 575	if (sin) {
 576		sin->sin6_family = AF_INET6;
 577		sin->sin6_flowinfo = 0;
 578		sin->sin6_port = 0;
 579		sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id;
 580		sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr;
 581		*addr_len = sizeof(*sin);
 582	}
 583
 584	put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info);
 585
 586	err = copied;
 587
 588out_free_skb:
 589	kfree_skb(skb);
 590out:
 591	return err;
 592}
 593
 594
 595void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg,
 596				 struct sk_buff *skb)
 597{
 598	struct ipv6_pinfo *np = inet6_sk(sk);
 599	bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6);
 600
 601	if (np->rxopt.bits.rxinfo) {
 602		struct in6_pktinfo src_info;
 603
 604		if (is_ipv6) {
 605			src_info.ipi6_ifindex = IP6CB(skb)->iif;
 606			src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
 607		} else {
 608			src_info.ipi6_ifindex =
 609				PKTINFO_SKB_CB(skb)->ipi_ifindex;
 610			ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr,
 611					       &src_info.ipi6_addr);
 612		}
 613
 614		if (src_info.ipi6_ifindex >= 0)
 615			put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO,
 616				 sizeof(src_info), &src_info);
 617	}
 618}
 619
 620void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg,
 621				    struct sk_buff *skb)
 622{
 623	struct ipv6_pinfo *np = inet6_sk(sk);
 624	struct inet6_skb_parm *opt = IP6CB(skb);
 625	unsigned char *nh = skb_network_header(skb);
 626
 627	if (np->rxopt.bits.rxhlim) {
 628		int hlim = ipv6_hdr(skb)->hop_limit;
 629		put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 630	}
 631
 632	if (np->rxopt.bits.rxtclass) {
 633		int tclass = ipv6_get_dsfield(ipv6_hdr(skb));
 634		put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
 635	}
 636
 637	if (np->rxopt.bits.rxflow) {
 638		__be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh);
 639		if (flowinfo)
 640			put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
 641	}
 642
 643	/* HbH is allowed only once */
 644	if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
 645		u8 *ptr = nh + sizeof(struct ipv6hdr);
 646		put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
 647	}
 648
 649	if (opt->lastopt &&
 650	    (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
 651		/*
 652		 * Silly enough, but we need to reparse in order to
 653		 * report extension headers (except for HbH)
 654		 * in order.
 655		 *
 656		 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
 657		 * (and WILL NOT be) defined because
 658		 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
 659		 */
 660		unsigned int off = sizeof(struct ipv6hdr);
 661		u8 nexthdr = ipv6_hdr(skb)->nexthdr;
 662
 663		while (off <= opt->lastopt) {
 664			unsigned int len;
 665			u8 *ptr = nh + off;
 666
 667			switch (nexthdr) {
 668			case IPPROTO_DSTOPTS:
 669				nexthdr = ptr[0];
 670				len = (ptr[1] + 1) << 3;
 671				if (np->rxopt.bits.dstopts)
 672					put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
 673				break;
 674			case IPPROTO_ROUTING:
 675				nexthdr = ptr[0];
 676				len = (ptr[1] + 1) << 3;
 677				if (np->rxopt.bits.srcrt)
 678					put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
 679				break;
 680			case IPPROTO_AH:
 681				nexthdr = ptr[0];
 682				len = (ptr[1] + 2) << 2;
 683				break;
 684			default:
 685				nexthdr = ptr[0];
 686				len = (ptr[1] + 1) << 3;
 687				break;
 688			}
 689
 690			off += len;
 691		}
 692	}
 693
 694	/* socket options in old style */
 695	if (np->rxopt.bits.rxoinfo) {
 696		struct in6_pktinfo src_info;
 697
 698		src_info.ipi6_ifindex = opt->iif;
 699		src_info.ipi6_addr = ipv6_hdr(skb)->daddr;
 700		put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 701	}
 702	if (np->rxopt.bits.rxohlim) {
 703		int hlim = ipv6_hdr(skb)->hop_limit;
 704		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 705	}
 706	if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) {
 707		u8 *ptr = nh + sizeof(struct ipv6hdr);
 708		put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
 709	}
 710	if (np->rxopt.bits.odstopts && opt->dst0) {
 711		u8 *ptr = nh + opt->dst0;
 712		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
 713	}
 714	if (np->rxopt.bits.osrcrt && opt->srcrt) {
 715		struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
 716		put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
 717	}
 718	if (np->rxopt.bits.odstopts && opt->dst1) {
 719		u8 *ptr = nh + opt->dst1;
 720		put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
 721	}
 722	if (np->rxopt.bits.rxorigdstaddr) {
 723		struct sockaddr_in6 sin6;
 724		__be16 _ports[2], *ports;
 725
 726		ports = skb_header_pointer(skb, skb_transport_offset(skb),
 727					   sizeof(_ports), &_ports);
 728		if (ports) {
 729			/* All current transport protocols have the port numbers in the
 730			 * first four bytes of the transport header and this function is
 731			 * written with this assumption in mind.
 732			 */
 733			sin6.sin6_family = AF_INET6;
 734			sin6.sin6_addr = ipv6_hdr(skb)->daddr;
 735			sin6.sin6_port = ports[1];
 736			sin6.sin6_flowinfo = 0;
 737			sin6.sin6_scope_id =
 738				ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr,
 739						    opt->iif);
 740
 741			put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6);
 742		}
 743	}
 744	if (np->rxopt.bits.recvfragsize && opt->frag_max_size) {
 745		int val = opt->frag_max_size;
 746
 747		put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val);
 748	}
 749}
 750
 751void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg,
 752			  struct sk_buff *skb)
 753{
 754	ip6_datagram_recv_common_ctl(sk, msg, skb);
 755	ip6_datagram_recv_specific_ctl(sk, msg, skb);
 756}
 757EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl);
 758
 759int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
 760			  struct msghdr *msg, struct flowi6 *fl6,
 761			  struct ipcm6_cookie *ipc6)
 762{
 763	struct in6_pktinfo *src_info;
 764	struct cmsghdr *cmsg;
 765	struct ipv6_rt_hdr *orthdr;
 766	struct ipv6_rt_hdr *rthdr;
 767	struct ipv6_opt_hdr *hdr;
 768	struct ipv6_txoptions *opt = ipc6->opt;
 769	int len;
 770	int err = 0;
 771
 772	for_each_cmsghdr(cmsg, msg) {
 773		int addr_type;
 774
 775		if (!CMSG_OK(msg, cmsg)) {
 776			err = -EINVAL;
 777			goto exit_f;
 778		}
 779
 780		if (cmsg->cmsg_level == SOL_SOCKET) {
 781			err = __sock_cmsg_send(sk, cmsg, &ipc6->sockc);
 782			if (err)
 783				return err;
 784			continue;
 785		}
 786
 787		if (cmsg->cmsg_level != SOL_IPV6)
 788			continue;
 789
 790		switch (cmsg->cmsg_type) {
 791		case IPV6_PKTINFO:
 792		case IPV6_2292PKTINFO:
 793		    {
 794			struct net_device *dev = NULL;
 795			int src_idx;
 796
 797			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
 798				err = -EINVAL;
 799				goto exit_f;
 800			}
 801
 802			src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
 803			src_idx = src_info->ipi6_ifindex;
 804
 805			if (src_idx) {
 806				if (fl6->flowi6_oif &&
 807				    src_idx != fl6->flowi6_oif &&
 808				    (READ_ONCE(sk->sk_bound_dev_if) != fl6->flowi6_oif ||
 809				     !sk_dev_equal_l3scope(sk, src_idx)))
 810					return -EINVAL;
 811				fl6->flowi6_oif = src_idx;
 812			}
 813
 814			addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
 815
 816			rcu_read_lock();
 817			if (fl6->flowi6_oif) {
 818				dev = dev_get_by_index_rcu(net, fl6->flowi6_oif);
 819				if (!dev) {
 820					rcu_read_unlock();
 821					return -ENODEV;
 822				}
 823			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
 824				rcu_read_unlock();
 825				return -EINVAL;
 826			}
 827
 828			if (addr_type != IPV6_ADDR_ANY) {
 829				int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
 830				if (!ipv6_can_nonlocal_bind(net, inet_sk(sk)) &&
 831				    !ipv6_chk_addr_and_flags(net, &src_info->ipi6_addr,
 832							     dev, !strict, 0,
 833							     IFA_F_TENTATIVE) &&
 834				    !ipv6_chk_acast_addr_src(net, dev,
 835							     &src_info->ipi6_addr))
 836					err = -EINVAL;
 837				else
 838					fl6->saddr = src_info->ipi6_addr;
 839			}
 840
 841			rcu_read_unlock();
 842
 843			if (err)
 844				goto exit_f;
 845
 846			break;
 847		    }
 848
 849		case IPV6_FLOWINFO:
 850			if (cmsg->cmsg_len < CMSG_LEN(4)) {
 851				err = -EINVAL;
 852				goto exit_f;
 853			}
 854
 855			if (fl6->flowlabel&IPV6_FLOWINFO_MASK) {
 856				if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
 857					err = -EINVAL;
 858					goto exit_f;
 859				}
 860			}
 861			fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
 862			break;
 863
 864		case IPV6_2292HOPOPTS:
 865		case IPV6_HOPOPTS:
 866			if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
 867				err = -EINVAL;
 868				goto exit_f;
 869			}
 870
 871			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
 872			len = ((hdr->hdrlen + 1) << 3);
 873			if (cmsg->cmsg_len < CMSG_LEN(len)) {
 874				err = -EINVAL;
 875				goto exit_f;
 876			}
 877			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
 878				err = -EPERM;
 879				goto exit_f;
 880			}
 881			opt->opt_nflen += len;
 882			opt->hopopt = hdr;
 883			break;
 884
 885		case IPV6_2292DSTOPTS:
 886			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
 887				err = -EINVAL;
 888				goto exit_f;
 889			}
 890
 891			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
 892			len = ((hdr->hdrlen + 1) << 3);
 893			if (cmsg->cmsg_len < CMSG_LEN(len)) {
 894				err = -EINVAL;
 895				goto exit_f;
 896			}
 897			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
 898				err = -EPERM;
 899				goto exit_f;
 900			}
 901			if (opt->dst1opt) {
 902				err = -EINVAL;
 903				goto exit_f;
 904			}
 905			opt->opt_flen += len;
 906			opt->dst1opt = hdr;
 907			break;
 908
 909		case IPV6_DSTOPTS:
 910		case IPV6_RTHDRDSTOPTS:
 911			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
 912				err = -EINVAL;
 913				goto exit_f;
 914			}
 915
 916			hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
 917			len = ((hdr->hdrlen + 1) << 3);
 918			if (cmsg->cmsg_len < CMSG_LEN(len)) {
 919				err = -EINVAL;
 920				goto exit_f;
 921			}
 922			if (!ns_capable(net->user_ns, CAP_NET_RAW)) {
 923				err = -EPERM;
 924				goto exit_f;
 925			}
 926			if (cmsg->cmsg_type == IPV6_DSTOPTS) {
 927				if (opt->dst1opt)
 928					opt->opt_flen -= ipv6_optlen(opt->dst1opt);
 929				opt->opt_flen += len;
 930				opt->dst1opt = hdr;
 931			} else {
 932				if (opt->dst0opt)
 933					opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
 934				opt->opt_nflen += len;
 935				opt->dst0opt = hdr;
 936			}
 937			break;
 938
 939		case IPV6_2292RTHDR:
 940		case IPV6_RTHDR:
 941			if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
 942				err = -EINVAL;
 943				goto exit_f;
 944			}
 945
 946			rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
 947
 948			switch (rthdr->type) {
 949#if IS_ENABLED(CONFIG_IPV6_MIP6)
 950			case IPV6_SRCRT_TYPE_2:
 951				if (rthdr->hdrlen != 2 ||
 952				    rthdr->segments_left != 1) {
 953					err = -EINVAL;
 954					goto exit_f;
 955				}
 956				break;
 957#endif
 958			default:
 959				err = -EINVAL;
 960				goto exit_f;
 961			}
 962
 963			len = ((rthdr->hdrlen + 1) << 3);
 964
 965			if (cmsg->cmsg_len < CMSG_LEN(len)) {
 966				err = -EINVAL;
 967				goto exit_f;
 968			}
 969
 970			/* segments left must also match */
 971			if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
 972				err = -EINVAL;
 973				goto exit_f;
 974			}
 975
 976			orthdr = opt->srcrt;
 977			if (orthdr)
 978				opt->opt_nflen -= ((orthdr->hdrlen + 1) << 3);
 979			opt->opt_nflen += len;
 980			opt->srcrt = rthdr;
 981
 982			if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
 983				int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
 984
 985				if (opt->dst0opt)
 986					opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
 987				opt->opt_nflen += dsthdrlen;
 988				opt->dst0opt = opt->dst1opt;
 989				opt->dst1opt = NULL;
 990				opt->opt_flen -= dsthdrlen;
 991			}
 992
 993			break;
 994
 995		case IPV6_2292HOPLIMIT:
 996		case IPV6_HOPLIMIT:
 997			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
 998				err = -EINVAL;
 999				goto exit_f;
1000			}
1001
1002			ipc6->hlimit = *(int *)CMSG_DATA(cmsg);
1003			if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) {
1004				err = -EINVAL;
1005				goto exit_f;
1006			}
1007
1008			break;
1009
1010		case IPV6_TCLASS:
1011		    {
1012			int tc;
1013
1014			err = -EINVAL;
1015			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
1016				goto exit_f;
1017
1018			tc = *(int *)CMSG_DATA(cmsg);
1019			if (tc < -1 || tc > 0xff)
1020				goto exit_f;
1021
1022			err = 0;
1023			ipc6->tclass = tc;
1024
1025			break;
1026		    }
1027
1028		case IPV6_DONTFRAG:
1029		    {
1030			int df;
1031
1032			err = -EINVAL;
1033			if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
1034				goto exit_f;
1035
1036			df = *(int *)CMSG_DATA(cmsg);
1037			if (df < 0 || df > 1)
1038				goto exit_f;
1039
1040			err = 0;
1041			ipc6->dontfrag = df;
1042
1043			break;
1044		    }
1045		default:
1046			net_dbg_ratelimited("invalid cmsg type: %d\n",
1047					    cmsg->cmsg_type);
1048			err = -EINVAL;
1049			goto exit_f;
1050		}
1051	}
1052
1053exit_f:
1054	return err;
1055}
1056EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl);
1057
1058void __ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp,
1059			       __u16 srcp, __u16 destp, int rqueue, int bucket)
1060{
1061	const struct in6_addr *dest, *src;
1062
1063	dest  = &sp->sk_v6_daddr;
1064	src   = &sp->sk_v6_rcv_saddr;
1065	seq_printf(seq,
1066		   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1067		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1068		   bucket,
1069		   src->s6_addr32[0], src->s6_addr32[1],
1070		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1071		   dest->s6_addr32[0], dest->s6_addr32[1],
1072		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1073		   sp->sk_state,
1074		   sk_wmem_alloc_get(sp),
1075		   rqueue,
1076		   0, 0L, 0,
1077		   from_kuid_munged(seq_user_ns(seq), sk_uid(sp)),
1078		   0,
1079		   sock_i_ino(sp),
1080		   refcount_read(&sp->sk_refcnt), sp,
1081		   sk_drops_read(sp));
1082}