Linux Audio

Check our new training course

Loading...
v4.6
 
   1/*
   2 *	UDP over IPv6
   3 *	Linux INET6 implementation
   4 *
   5 *	Authors:
   6 *	Pedro Roque		<roque@di.fc.ul.pt>
   7 *
   8 *	Based on linux/ipv4/udp.c
   9 *
  10 *	Fixes:
  11 *	Hideaki YOSHIFUJI	:	sin6_scope_id support
  12 *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
  13 *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
  14 *					a single port at the same time.
  15 *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
  16 *      YOSHIFUJI Hideaki @USAGI:	convert /proc/net/udp6 to seq_file.
  17 *
  18 *	This program is free software; you can redistribute it and/or
  19 *      modify it under the terms of the GNU General Public License
  20 *      as published by the Free Software Foundation; either version
  21 *      2 of the License, or (at your option) any later version.
  22 */
  23
 
  24#include <linux/errno.h>
  25#include <linux/types.h>
  26#include <linux/socket.h>
  27#include <linux/sockios.h>
  28#include <linux/net.h>
  29#include <linux/in6.h>
  30#include <linux/netdevice.h>
  31#include <linux/if_arp.h>
  32#include <linux/ipv6.h>
  33#include <linux/icmpv6.h>
  34#include <linux/init.h>
  35#include <linux/module.h>
  36#include <linux/skbuff.h>
  37#include <linux/slab.h>
  38#include <asm/uaccess.h>
 
  39
  40#include <net/addrconf.h>
  41#include <net/ndisc.h>
  42#include <net/protocol.h>
  43#include <net/transp_v6.h>
  44#include <net/ip6_route.h>
  45#include <net/raw.h>
 
  46#include <net/tcp_states.h>
  47#include <net/ip6_checksum.h>
 
 
  48#include <net/xfrm.h>
 
  49#include <net/inet6_hashtables.h>
  50#include <net/busy_poll.h>
  51#include <net/sock_reuseport.h>
 
  52
  53#include <linux/proc_fs.h>
  54#include <linux/seq_file.h>
  55#include <trace/events/skb.h>
  56#include "udp_impl.h"
  57
  58static u32 udp6_ehashfn(const struct net *net,
  59			const struct in6_addr *laddr,
  60			const u16 lport,
  61			const struct in6_addr *faddr,
  62			const __be16 fport)
 
 
  63{
  64	static u32 udp6_ehash_secret __read_mostly;
  65	static u32 udp_ipv6_hash_secret __read_mostly;
 
 
 
  66
 
 
 
 
 
 
 
  67	u32 lhash, fhash;
  68
  69	net_get_random_once(&udp6_ehash_secret,
  70			    sizeof(udp6_ehash_secret));
  71	net_get_random_once(&udp_ipv6_hash_secret,
  72			    sizeof(udp_ipv6_hash_secret));
  73
  74	lhash = (__force u32)laddr->s6_addr32[3];
  75	fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
  76
  77	return __inet6_ehashfn(lhash, lport, fhash, fport,
  78			       udp_ipv6_hash_secret + net_hash_mix(net));
  79}
  80
  81static u32 udp6_portaddr_hash(const struct net *net,
  82			      const struct in6_addr *addr6,
  83			      unsigned int port)
  84{
  85	unsigned int hash, mix = net_hash_mix(net);
  86
  87	if (ipv6_addr_any(addr6))
  88		hash = jhash_1word(0, mix);
  89	else if (ipv6_addr_v4mapped(addr6))
  90		hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
  91	else
  92		hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
  93
  94	return hash ^ port;
  95}
  96
  97int udp_v6_get_port(struct sock *sk, unsigned short snum)
  98{
  99	unsigned int hash2_nulladdr =
 100		udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
 101	unsigned int hash2_partial =
 102		udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
 103
 104	/* precompute partial secondary hash */
 105	udp_sk(sk)->udp_portaddr_hash = hash2_partial;
 106	return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
 107}
 108
 109static void udp_v6_rehash(struct sock *sk)
 110{
 111	u16 new_hash = udp6_portaddr_hash(sock_net(sk),
 112					  &sk->sk_v6_rcv_saddr,
 113					  inet_sk(sk)->inet_num);
 114
 115	udp_lib_rehash(sk, new_hash);
 116}
 117
 118static inline int compute_score(struct sock *sk, struct net *net,
 119				unsigned short hnum,
 120				const struct in6_addr *saddr, __be16 sport,
 121				const struct in6_addr *daddr, __be16 dport,
 122				int dif)
 123{
 124	int score;
 125	struct inet_sock *inet;
 126
 127	if (!net_eq(sock_net(sk), net) ||
 128	    udp_sk(sk)->udp_port_hash != hnum ||
 129	    sk->sk_family != PF_INET6)
 130		return -1;
 131
 132	score = 0;
 133	inet = inet_sk(sk);
 134
 135	if (inet->inet_dport) {
 136		if (inet->inet_dport != sport)
 137			return -1;
 138		score++;
 139	}
 140
 141	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
 142		if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
 143			return -1;
 144		score++;
 145	}
 146
 147	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
 148		if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
 149			return -1;
 150		score++;
 151	}
 152
 153	if (sk->sk_bound_dev_if) {
 154		if (sk->sk_bound_dev_if != dif)
 155			return -1;
 156		score++;
 157	}
 158
 159	if (sk->sk_incoming_cpu == raw_smp_processor_id())
 160		score++;
 161
 162	return score;
 163}
 164
 165static inline int compute_score2(struct sock *sk, struct net *net,
 166				 const struct in6_addr *saddr, __be16 sport,
 167				 const struct in6_addr *daddr,
 168				 unsigned short hnum, int dif)
 169{
 170	int score;
 171	struct inet_sock *inet;
 
 172
 173	if (!net_eq(sock_net(sk), net) ||
 174	    udp_sk(sk)->udp_port_hash != hnum ||
 175	    sk->sk_family != PF_INET6)
 176		return -1;
 177
 178	if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
 179		return -1;
 180
 181	score = 0;
 182	inet = inet_sk(sk);
 183
 184	if (inet->inet_dport) {
 185		if (inet->inet_dport != sport)
 186			return -1;
 187		score++;
 188	}
 189
 190	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
 191		if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
 192			return -1;
 193		score++;
 194	}
 195
 196	if (sk->sk_bound_dev_if) {
 197		if (sk->sk_bound_dev_if != dif)
 198			return -1;
 
 
 199		score++;
 200	}
 201
 202	if (sk->sk_incoming_cpu == raw_smp_processor_id())
 203		score++;
 204
 205	return score;
 206}
 207
 208/* called with read_rcu_lock() */
 209static struct sock *udp6_lib_lookup2(struct net *net,
 210		const struct in6_addr *saddr, __be16 sport,
 211		const struct in6_addr *daddr, unsigned int hnum, int dif,
 212		struct udp_hslot *hslot2, unsigned int slot2,
 213		struct sk_buff *skb)
 214{
 215	struct sock *sk, *result;
 216	struct hlist_nulls_node *node;
 217	int score, badness, matches = 0, reuseport = 0;
 218	bool select_ok = true;
 219	u32 hash = 0;
 220
 221begin:
 222	result = NULL;
 223	badness = -1;
 224	udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
 225		score = compute_score2(sk, net, saddr, sport,
 226				      daddr, hnum, dif);
 
 
 227		if (score > badness) {
 228			result = sk;
 229			badness = score;
 230			reuseport = sk->sk_reuseport;
 231			if (reuseport) {
 232				hash = udp6_ehashfn(net, daddr, hnum,
 233						    saddr, sport);
 234				if (select_ok) {
 235					struct sock *sk2;
 236
 237					sk2 = reuseport_select_sock(sk, hash, skb,
 238							sizeof(struct udphdr));
 239					if (sk2) {
 240						result = sk2;
 241						select_ok = false;
 242						goto found;
 243					}
 244				}
 245				matches = 1;
 246			}
 247		} else if (score == badness && reuseport) {
 248			matches++;
 249			if (reciprocal_scale(hash, matches) == 0)
 
 250				result = sk;
 251			hash = next_pseudo_random32(hash);
 252		}
 253	}
 254	/*
 255	 * if the nulls value we got at the end of this lookup is
 256	 * not the expected one, we must restart lookup.
 257	 * We probably met an item that was moved to another chain.
 258	 */
 259	if (get_nulls_value(node) != slot2)
 260		goto begin;
 261
 262	if (result) {
 263found:
 264		if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
 265			result = NULL;
 266		else if (unlikely(compute_score2(result, net, saddr, sport,
 267				  daddr, hnum, dif) < badness)) {
 268			sock_put(result);
 269			goto begin;
 
 
 
 
 
 
 
 
 270		}
 271	}
 272	return result;
 273}
 274
 
 275struct sock *__udp6_lib_lookup(struct net *net,
 276				      const struct in6_addr *saddr, __be16 sport,
 277				      const struct in6_addr *daddr, __be16 dport,
 278				      int dif, struct udp_table *udptable,
 279				      struct sk_buff *skb)
 280{
 281	struct sock *sk, *result;
 282	struct hlist_nulls_node *node;
 283	unsigned short hnum = ntohs(dport);
 284	unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
 285	struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
 286	int score, badness, matches = 0, reuseport = 0;
 287	bool select_ok = true;
 288	u32 hash = 0;
 289
 290	rcu_read_lock();
 291	if (hslot->count > 10) {
 292		hash2 = udp6_portaddr_hash(net, daddr, hnum);
 293		slot2 = hash2 & udptable->mask;
 294		hslot2 = &udptable->hash2[slot2];
 295		if (hslot->count < hslot2->count)
 296			goto begin;
 297
 298		result = udp6_lib_lookup2(net, saddr, sport,
 299					  daddr, hnum, dif,
 300					  hslot2, slot2, skb);
 301		if (!result) {
 302			hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
 303			slot2 = hash2 & udptable->mask;
 304			hslot2 = &udptable->hash2[slot2];
 305			if (hslot->count < hslot2->count)
 306				goto begin;
 307
 308			result = udp6_lib_lookup2(net, saddr, sport,
 309						  &in6addr_any, hnum, dif,
 310						  hslot2, slot2, skb);
 311		}
 312		rcu_read_unlock();
 313		return result;
 314	}
 315begin:
 316	result = NULL;
 317	badness = -1;
 318	sk_nulls_for_each_rcu(sk, node, &hslot->head) {
 319		score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
 320		if (score > badness) {
 321			result = sk;
 322			badness = score;
 323			reuseport = sk->sk_reuseport;
 324			if (reuseport) {
 325				hash = udp6_ehashfn(net, daddr, hnum,
 326						    saddr, sport);
 327				if (select_ok) {
 328					struct sock *sk2;
 329
 330					sk2 = reuseport_select_sock(sk, hash, skb,
 331							sizeof(struct udphdr));
 332					if (sk2) {
 333						result = sk2;
 334						select_ok = false;
 335						goto found;
 336					}
 337				}
 338				matches = 1;
 339			}
 340		} else if (score == badness && reuseport) {
 341			matches++;
 342			if (reciprocal_scale(hash, matches) == 0)
 343				result = sk;
 344			hash = next_pseudo_random32(hash);
 345		}
 346	}
 347	/*
 348	 * if the nulls value we got at the end of this lookup is
 349	 * not the expected one, we must restart lookup.
 350	 * We probably met an item that was moved to another chain.
 351	 */
 352	if (get_nulls_value(node) != slot)
 353		goto begin;
 354
 355	if (result) {
 356found:
 357		if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
 358			result = NULL;
 359		else if (unlikely(compute_score(result, net, hnum, saddr, sport,
 360					daddr, dport, dif) < badness)) {
 361			sock_put(result);
 362			goto begin;
 363		}
 364	}
 365	rcu_read_unlock();
 
 
 
 
 366	return result;
 367}
 368EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
 369
 370static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
 371					  __be16 sport, __be16 dport,
 372					  struct udp_table *udptable)
 373{
 374	struct sock *sk;
 375	const struct ipv6hdr *iph = ipv6_hdr(skb);
 376
 377	sk = skb_steal_sock(skb);
 378	if (unlikely(sk))
 379		return sk;
 380	return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
 381				 &iph->daddr, dport, inet6_iif(skb),
 382				 udptable, skb);
 383}
 384
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 385struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
 386			     const struct in6_addr *daddr, __be16 dport, int dif)
 387{
 388	return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table, NULL);
 
 
 
 
 
 
 389}
 390EXPORT_SYMBOL_GPL(udp6_lib_lookup);
 
 
 
 
 
 
 
 
 
 
 391
 392/*
 393 *	This should be easy, if there is something there we
 394 *	return it, otherwise we block.
 395 */
 396
 397int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 398		  int noblock, int flags, int *addr_len)
 399{
 400	struct ipv6_pinfo *np = inet6_sk(sk);
 401	struct inet_sock *inet = inet_sk(sk);
 402	struct sk_buff *skb;
 403	unsigned int ulen, copied;
 404	int peeked, off = 0;
 405	int err;
 406	int is_udplite = IS_UDPLITE(sk);
 
 407	bool checksum_valid = false;
 408	int is_udp4;
 409	bool slow;
 410
 411	if (flags & MSG_ERRQUEUE)
 412		return ipv6_recv_error(sk, msg, len, addr_len);
 413
 414	if (np->rxpmtu && np->rxopt.bits.rxpmtu)
 415		return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
 416
 417try_again:
 418	skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
 419				  &peeked, &off, &err);
 420	if (!skb)
 421		goto out;
 422
 423	ulen = skb->len - sizeof(struct udphdr);
 424	copied = len;
 425	if (copied > ulen)
 426		copied = ulen;
 427	else if (copied < ulen)
 428		msg->msg_flags |= MSG_TRUNC;
 429
 430	is_udp4 = (skb->protocol == htons(ETH_P_IP));
 
 431
 432	/*
 433	 * If checksum is needed at all, try to do it while copying the
 434	 * data.  If the data is truncated, or if we only want a partial
 435	 * coverage checksum (UDP-Lite), do it before the copy.
 436	 */
 437
 438	if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
 439		checksum_valid = !udp_lib_checksum_complete(skb);
 
 
 440		if (!checksum_valid)
 441			goto csum_copy_err;
 442	}
 443
 444	if (checksum_valid || skb_csum_unnecessary(skb))
 445		err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
 446					    msg, copied);
 447	else {
 448		err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr), msg);
 
 
 449		if (err == -EINVAL)
 450			goto csum_copy_err;
 451	}
 452	if (unlikely(err)) {
 453		trace_kfree_skb(skb, udpv6_recvmsg);
 454		if (!peeked) {
 455			atomic_inc(&sk->sk_drops);
 456			if (is_udp4)
 457				UDP_INC_STATS_USER(sock_net(sk),
 458						   UDP_MIB_INERRORS,
 459						   is_udplite);
 460			else
 461				UDP6_INC_STATS_USER(sock_net(sk),
 462						    UDP_MIB_INERRORS,
 463						    is_udplite);
 464		}
 465		goto out_free;
 466	}
 467	if (!peeked) {
 468		if (is_udp4)
 469			UDP_INC_STATS_USER(sock_net(sk),
 470					UDP_MIB_INDATAGRAMS, is_udplite);
 471		else
 472			UDP6_INC_STATS_USER(sock_net(sk),
 473					UDP_MIB_INDATAGRAMS, is_udplite);
 474	}
 
 
 475
 476	sock_recv_ts_and_drops(msg, sk, skb);
 477
 478	/* Copy the address. */
 479	if (msg->msg_name) {
 480		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 481		sin6->sin6_family = AF_INET6;
 482		sin6->sin6_port = udp_hdr(skb)->source;
 483		sin6->sin6_flowinfo = 0;
 484
 485		if (is_udp4) {
 486			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
 487					       &sin6->sin6_addr);
 488			sin6->sin6_scope_id = 0;
 489		} else {
 490			sin6->sin6_addr = ipv6_hdr(skb)->saddr;
 491			sin6->sin6_scope_id =
 492				ipv6_iface_scope_id(&sin6->sin6_addr,
 493						    inet6_iif(skb));
 494		}
 495		*addr_len = sizeof(*sin6);
 
 
 
 
 496	}
 497
 
 
 
 498	if (np->rxopt.all)
 499		ip6_datagram_recv_common_ctl(sk, msg, skb);
 500
 501	if (is_udp4) {
 502		if (inet->cmsg_flags)
 503			ip_cmsg_recv(msg, skb);
 
 504	} else {
 505		if (np->rxopt.all)
 506			ip6_datagram_recv_specific_ctl(sk, msg, skb);
 507	}
 508
 509	err = copied;
 510	if (flags & MSG_TRUNC)
 511		err = ulen;
 512
 513out_free:
 514	skb_free_datagram_locked(sk, skb);
 515out:
 516	return err;
 517
 518csum_copy_err:
 519	slow = lock_sock_fast(sk);
 520	if (!skb_kill_datagram(sk, skb, flags)) {
 521		if (is_udp4) {
 522			UDP_INC_STATS_USER(sock_net(sk),
 523					UDP_MIB_CSUMERRORS, is_udplite);
 524			UDP_INC_STATS_USER(sock_net(sk),
 525					UDP_MIB_INERRORS, is_udplite);
 526		} else {
 527			UDP6_INC_STATS_USER(sock_net(sk),
 528					UDP_MIB_CSUMERRORS, is_udplite);
 529			UDP6_INC_STATS_USER(sock_net(sk),
 530					UDP_MIB_INERRORS, is_udplite);
 531		}
 532	}
 533	unlock_sock_fast(sk, slow);
 534
 535	/* starting over for a new packet, but check if we need to yield */
 536	cond_resched();
 537	msg->msg_flags &= ~MSG_TRUNC;
 538	goto try_again;
 539}
 540
 541void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 542		    u8 type, u8 code, int offset, __be32 info,
 543		    struct udp_table *udptable)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 544{
 545	struct ipv6_pinfo *np;
 546	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
 547	const struct in6_addr *saddr = &hdr->saddr;
 548	const struct in6_addr *daddr = &hdr->daddr;
 549	struct udphdr *uh = (struct udphdr *)(skb->data+offset);
 
 550	struct sock *sk;
 551	int harderr;
 552	int err;
 553	struct net *net = dev_net(skb->dev);
 554
 555	sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
 556			       inet6_iif(skb), udptable, skb);
 557	if (!sk) {
 558		ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
 559				   ICMP6_MIB_INERRORS);
 560		return;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 561	}
 562
 563	harderr = icmpv6_err_convert(type, code, &err);
 564	np = inet6_sk(sk);
 565
 566	if (type == ICMPV6_PKT_TOOBIG) {
 567		if (!ip6_sk_accept_pmtu(sk))
 568			goto out;
 569		ip6_sk_update_pmtu(skb, sk, info);
 570		if (np->pmtudisc != IPV6_PMTUDISC_DONT)
 571			harderr = 1;
 572	}
 573	if (type == NDISC_REDIRECT) {
 574		ip6_sk_redirect(skb, sk);
 
 
 
 
 
 
 
 
 
 
 
 
 
 575		goto out;
 576	}
 577
 578	if (!np->recverr) {
 579		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 580			goto out;
 581	} else {
 582		ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
 583	}
 584
 585	sk->sk_err = err;
 586	sk->sk_error_report(sk);
 587out:
 588	sock_put(sk);
 589}
 590
 591static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 592{
 593	int rc;
 594
 595	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
 596		sock_rps_save_rxhash(sk, skb);
 597		sk_mark_napi_id(sk, skb);
 598		sk_incoming_cpu_update(sk);
 
 
 599	}
 600
 601	rc = sock_queue_rcv_skb(sk, skb);
 602	if (rc < 0) {
 603		int is_udplite = IS_UDPLITE(sk);
 
 604
 605		/* Note that an ENOMEM error is charged twice */
 606		if (rc == -ENOMEM)
 607			UDP6_INC_STATS_BH(sock_net(sk),
 608					UDP_MIB_RCVBUFERRORS, is_udplite);
 609		UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 610		kfree_skb(skb);
 
 
 
 
 
 
 
 611		return -1;
 612	}
 613	return 0;
 614}
 615
 616static __inline__ void udpv6_err(struct sk_buff *skb,
 617				 struct inet6_skb_parm *opt, u8 type,
 618				 u8 code, int offset, __be32 info)
 619{
 620	__udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
 621}
 622
 623static struct static_key udpv6_encap_needed __read_mostly;
 624void udpv6_encap_enable(void)
 
 625{
 626	if (!static_key_enabled(&udpv6_encap_needed))
 627		static_key_slow_inc(&udpv6_encap_needed);
 628}
 629EXPORT_SYMBOL(udpv6_encap_enable);
 630
 631int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 632{
 
 633	struct udp_sock *up = udp_sk(sk);
 634	int rc;
 635	int is_udplite = IS_UDPLITE(sk);
 636
 637	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
 
 638		goto drop;
 
 
 639
 640	if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
 
 641		int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
 642
 643		/*
 644		 * This is an encapsulation socket so pass the skb to
 645		 * the socket's udp_encap_rcv() hook. Otherwise, just
 646		 * fall through and pass this up the UDP socket.
 647		 * up->encap_rcv() returns the following value:
 648		 * =0 if skb was successfully passed to the encap
 649		 *    handler or was discarded by it.
 650		 * >0 if skb should be passed on to UDP.
 651		 * <0 if skb should be resubmitted as proto -N
 652		 */
 653
 654		/* if we're overly short, let UDP handle it */
 655		encap_rcv = ACCESS_ONCE(up->encap_rcv);
 656		if (skb->len > sizeof(struct udphdr) && encap_rcv) {
 657			int ret;
 658
 659			/* Verify checksum before giving to encap */
 660			if (udp_lib_checksum_complete(skb))
 661				goto csum_error;
 662
 663			ret = encap_rcv(sk, skb);
 664			if (ret <= 0) {
 665				UDP_INC_STATS_BH(sock_net(sk),
 666						 UDP_MIB_INDATAGRAMS,
 667						 is_udplite);
 668				return -ret;
 669			}
 670		}
 671
 672		/* FALLTHROUGH -- it's a UDP Packet */
 673	}
 674
 675	/*
 676	 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
 677	 */
 678	if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
 
 679
 680		if (up->pcrlen == 0) {          /* full coverage was set  */
 681			net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
 682					    UDP_SKB_CB(skb)->cscov, skb->len);
 683			goto drop;
 684		}
 685		if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
 686			net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
 687					    UDP_SKB_CB(skb)->cscov, up->pcrlen);
 688			goto drop;
 689		}
 690	}
 691
 692	if (rcu_access_pointer(sk->sk_filter)) {
 693		if (udp_lib_checksum_complete(skb))
 694			goto csum_error;
 695	}
 696
 697	if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
 698		UDP6_INC_STATS_BH(sock_net(sk),
 699				  UDP_MIB_RCVBUFERRORS, is_udplite);
 700		goto drop;
 701	}
 702
 703	skb_dst_drop(skb);
 704
 705	bh_lock_sock(sk);
 706	rc = 0;
 707	if (!sock_owned_by_user(sk))
 708		rc = __udpv6_queue_rcv_skb(sk, skb);
 709	else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
 710		bh_unlock_sock(sk);
 711		goto drop;
 712	}
 713	bh_unlock_sock(sk);
 714
 715	return rc;
 716
 717csum_error:
 718	UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
 
 719drop:
 720	UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 721	atomic_inc(&sk->sk_drops);
 722	kfree_skb(skb);
 723	return -1;
 724}
 725
 726static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 727				   __be16 loc_port, const struct in6_addr *loc_addr,
 728				   __be16 rmt_port, const struct in6_addr *rmt_addr,
 729				   int dif, unsigned short hnum)
 730{
 731	struct inet_sock *inet = inet_sk(sk);
 732
 733	if (!net_eq(sock_net(sk), net))
 734		return false;
 735
 736	if (udp_sk(sk)->udp_port_hash != hnum ||
 737	    sk->sk_family != PF_INET6 ||
 738	    (inet->inet_dport && inet->inet_dport != rmt_port) ||
 739	    (!ipv6_addr_any(&sk->sk_v6_daddr) &&
 740		    !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
 741	    (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
 742	    (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 743		    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
 744		return false;
 745	if (!inet6_mc_check(sk, loc_addr, rmt_addr))
 746		return false;
 747	return true;
 748}
 749
 750static void flush_stack(struct sock **stack, unsigned int count,
 751			struct sk_buff *skb, unsigned int final)
 752{
 753	struct sk_buff *skb1 = NULL;
 754	struct sock *sk;
 755	unsigned int i;
 756
 757	for (i = 0; i < count; i++) {
 758		sk = stack[i];
 759		if (likely(!skb1))
 760			skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
 761		if (!skb1) {
 762			atomic_inc(&sk->sk_drops);
 763			UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
 764					  IS_UDPLITE(sk));
 765			UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
 766					  IS_UDPLITE(sk));
 767		}
 768
 769		if (skb1 && udpv6_queue_rcv_skb(sk, skb1) <= 0)
 770			skb1 = NULL;
 771		sock_put(sk);
 772	}
 773	if (unlikely(skb1))
 774		kfree_skb(skb1);
 775}
 776
 777static void udp6_csum_zero_error(struct sk_buff *skb)
 778{
 779	/* RFC 2460 section 8.1 says that we SHOULD log
 780	 * this error. Well, it is reasonable.
 781	 */
 782	net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
 783			    &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
 784			    &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
 785}
 786
 787/*
 788 * Note: called only from the BH handler context,
 789 * so we don't need to lock the hashes.
 790 */
 791static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
 792		const struct in6_addr *saddr, const struct in6_addr *daddr,
 793		struct udp_table *udptable, int proto)
 794{
 795	struct sock *sk, *stack[256 / sizeof(struct sock *)];
 796	const struct udphdr *uh = udp_hdr(skb);
 797	struct hlist_nulls_node *node;
 798	unsigned short hnum = ntohs(uh->dest);
 799	struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
 800	int dif = inet6_iif(skb);
 801	unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
 802	unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
 803	bool inner_flushed = false;
 
 
 
 804
 805	if (use_hash2) {
 806		hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
 807			    udp_table.mask;
 808		hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
 809start_lookup:
 810		hslot = &udp_table.hash2[hash2];
 811		offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
 812	}
 813
 814	spin_lock(&hslot->lock);
 815	sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
 816		if (__udp_v6_is_mcast_sock(net, sk,
 817					   uh->dest, daddr,
 818					   uh->source, saddr,
 819					   dif, hnum) &&
 820		    /* If zero checksum and no_check is not on for
 821		     * the socket then skip it.
 822		     */
 823		    (uh->check || udp_sk(sk)->no_check6_rx)) {
 824			if (unlikely(count == ARRAY_SIZE(stack))) {
 825				flush_stack(stack, count, skb, ~0);
 826				inner_flushed = true;
 827				count = 0;
 828			}
 829			stack[count++] = sk;
 830			sock_hold(sk);
 
 
 
 
 
 831		}
 832	}
 833
 834	spin_unlock(&hslot->lock);
 
 
 835
 836	/* Also lookup *:port if we are using hash2 and haven't done so yet. */
 837	if (use_hash2 && hash2 != hash2_any) {
 838		hash2 = hash2_any;
 839		goto start_lookup;
 840	}
 841
 842	if (count) {
 843		flush_stack(stack, count, skb, count - 1);
 
 844	} else {
 845		if (!inner_flushed)
 846			UDP6_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
 847					  proto == IPPROTO_UDPLITE);
 848		consume_skb(skb);
 849	}
 850	return 0;
 851}
 852
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 853int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
 854		   int proto)
 855{
 
 
 856	struct net *net = dev_net(skb->dev);
 857	struct sock *sk;
 858	struct udphdr *uh;
 859	const struct in6_addr *saddr, *daddr;
 
 860	u32 ulen = 0;
 861
 862	if (!pskb_may_pull(skb, sizeof(struct udphdr)))
 863		goto discard;
 864
 865	saddr = &ipv6_hdr(skb)->saddr;
 866	daddr = &ipv6_hdr(skb)->daddr;
 867	uh = udp_hdr(skb);
 868
 869	ulen = ntohs(uh->len);
 870	if (ulen > skb->len)
 871		goto short_packet;
 872
 873	if (proto == IPPROTO_UDP) {
 874		/* UDP validates ulen. */
 875
 876		/* Check for jumbo payload */
 877		if (ulen == 0)
 878			ulen = skb->len;
 879
 880		if (ulen < sizeof(*uh))
 881			goto short_packet;
 882
 883		if (ulen < skb->len) {
 884			if (pskb_trim_rcsum(skb, ulen))
 885				goto short_packet;
 886			saddr = &ipv6_hdr(skb)->saddr;
 887			daddr = &ipv6_hdr(skb)->daddr;
 888			uh = udp_hdr(skb);
 889		}
 890	}
 891
 892	if (udp6_csum_init(skb, uh, proto))
 893		goto csum_error;
 894
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 895	/*
 896	 *	Multicast receive code
 897	 */
 898	if (ipv6_addr_is_multicast(daddr))
 899		return __udp6_lib_mcast_deliver(net, skb,
 900				saddr, daddr, udptable, proto);
 901
 902	/* Unicast */
 903
 904	/*
 905	 * check socket cache ... must talk to Alan about his plans
 906	 * for sock caches... i'll skip this for now.
 907	 */
 908	sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
 909	if (sk) {
 910		int ret;
 911
 912		if (!uh->check && !udp_sk(sk)->no_check6_rx) {
 913			sock_put(sk);
 914			udp6_csum_zero_error(skb);
 915			goto csum_error;
 916		}
 917
 918		if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
 919			skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
 920						 ip6_compute_pseudo);
 921
 922		ret = udpv6_queue_rcv_skb(sk, skb);
 923		sock_put(sk);
 924
 925		/* a return value > 0 means to resubmit the input */
 926		if (ret > 0)
 927			return ret;
 928
 929		return 0;
 930	}
 
 
 931
 932	if (!uh->check) {
 933		udp6_csum_zero_error(skb);
 934		goto csum_error;
 935	}
 936
 937	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 938		goto discard;
 
 939
 940	if (udp_lib_checksum_complete(skb))
 941		goto csum_error;
 942
 943	UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
 944	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 945
 946	kfree_skb(skb);
 947	return 0;
 948
 949short_packet:
 
 
 950	net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
 951			    proto == IPPROTO_UDPLITE ? "-Lite" : "",
 952			    saddr, ntohs(uh->source),
 953			    ulen, skb->len,
 954			    daddr, ntohs(uh->dest));
 955	goto discard;
 
 
 
 956csum_error:
 957	UDP6_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
 
 
 958discard:
 959	UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
 960	kfree_skb(skb);
 961	return 0;
 962}
 963
 964static __inline__ int udpv6_rcv(struct sk_buff *skb)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 965{
 966	return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
 967}
 968
 969/*
 970 * Throw away all pending data and cancel the corking. Socket is locked.
 971 */
 972static void udp_v6_flush_pending_frames(struct sock *sk)
 973{
 974	struct udp_sock *up = udp_sk(sk);
 975
 976	if (up->pending == AF_INET)
 977		udp_flush_pending_frames(sk);
 978	else if (up->pending) {
 979		up->len = 0;
 980		up->pending = 0;
 981		ip6_flush_pending_frames(sk);
 982	}
 983}
 984
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 985/**
 986 *	udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
 987 *	@sk:	socket we are sending on
 988 *	@skb:	sk_buff containing the filled-in UDP header
 989 *		(checksum field must be zeroed out)
 
 
 
 990 */
 991static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
 992				 const struct in6_addr *saddr,
 993				 const struct in6_addr *daddr, int len)
 994{
 995	unsigned int offset;
 996	struct udphdr *uh = udp_hdr(skb);
 997	struct sk_buff *frags = skb_shinfo(skb)->frag_list;
 998	__wsum csum = 0;
 999
1000	if (!frags) {
1001		/* Only one fragment on the socket.  */
1002		skb->csum_start = skb_transport_header(skb) - skb->head;
1003		skb->csum_offset = offsetof(struct udphdr, check);
1004		uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
1005	} else {
1006		/*
1007		 * HW-checksum won't work as there are two or more
1008		 * fragments on the socket so that all csums of sk_buffs
1009		 * should be together
1010		 */
1011		offset = skb_transport_offset(skb);
1012		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
 
1013
1014		skb->ip_summed = CHECKSUM_NONE;
1015
1016		do {
1017			csum = csum_add(csum, frags->csum);
1018		} while ((frags = frags->next));
1019
1020		uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1021					    csum);
1022		if (uh->check == 0)
1023			uh->check = CSUM_MANGLED_0;
1024	}
1025}
1026
1027/*
1028 *	Sending
1029 */
1030
1031static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
 
1032{
1033	struct sock *sk = skb->sk;
1034	struct udphdr *uh;
1035	int err = 0;
1036	int is_udplite = IS_UDPLITE(sk);
1037	__wsum csum = 0;
1038	int offset = skb_transport_offset(skb);
1039	int len = skb->len - offset;
 
1040
1041	/*
1042	 * Create a UDP header
1043	 */
1044	uh = udp_hdr(skb);
1045	uh->source = fl6->fl6_sport;
1046	uh->dest = fl6->fl6_dport;
1047	uh->len = htons(len);
1048	uh->check = 0;
1049
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1050	if (is_udplite)
1051		csum = udplite_csum(skb);
1052	else if (udp_sk(sk)->no_check6_tx) {   /* UDP csum disabled */
1053		skb->ip_summed = CHECKSUM_NONE;
1054		goto send;
1055	} else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
 
1056		udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
1057		goto send;
1058	} else
1059		csum = udp_csum(skb);
1060
1061	/* add protocol-dependent pseudo-header */
1062	uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1063				    len, fl6->flowi6_proto, csum);
1064	if (uh->check == 0)
1065		uh->check = CSUM_MANGLED_0;
1066
1067send:
1068	err = ip6_send_skb(skb);
1069	if (err) {
1070		if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1071			UDP6_INC_STATS_USER(sock_net(sk),
1072					    UDP_MIB_SNDBUFERRORS, is_udplite);
1073			err = 0;
1074		}
1075	} else
1076		UDP6_INC_STATS_USER(sock_net(sk),
1077				    UDP_MIB_OUTDATAGRAMS, is_udplite);
 
1078	return err;
1079}
1080
1081static int udp_v6_push_pending_frames(struct sock *sk)
1082{
1083	struct sk_buff *skb;
1084	struct udp_sock  *up = udp_sk(sk);
1085	struct flowi6 fl6;
1086	int err = 0;
1087
1088	if (up->pending == AF_INET)
1089		return udp_push_pending_frames(sk);
1090
1091	/* ip6_finish_skb will release the cork, so make a copy of
1092	 * fl6 here.
1093	 */
1094	fl6 = inet_sk(sk)->cork.fl.u.ip6;
1095
1096	skb = ip6_finish_skb(sk);
1097	if (!skb)
1098		goto out;
1099
1100	err = udp_v6_send_skb(skb, &fl6);
1101
1102out:
1103	up->len = 0;
1104	up->pending = 0;
1105	return err;
1106}
1107
1108int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1109{
1110	struct ipv6_txoptions opt_space;
1111	struct udp_sock *up = udp_sk(sk);
1112	struct inet_sock *inet = inet_sk(sk);
1113	struct ipv6_pinfo *np = inet6_sk(sk);
1114	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1115	struct in6_addr *daddr, *final_p, final;
1116	struct ipv6_txoptions *opt = NULL;
1117	struct ipv6_txoptions *opt_to_free = NULL;
1118	struct ip6_flowlabel *flowlabel = NULL;
1119	struct flowi6 fl6;
 
1120	struct dst_entry *dst;
 
1121	int addr_len = msg->msg_namelen;
 
1122	int ulen = len;
1123	int hlimit = -1;
1124	int tclass = -1;
1125	int dontfrag = -1;
1126	int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1127	int err;
1128	int connected = 0;
1129	int is_udplite = IS_UDPLITE(sk);
1130	int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1131
 
 
 
 
 
1132	/* destination address check */
1133	if (sin6) {
1134		if (addr_len < offsetof(struct sockaddr, sa_data))
1135			return -EINVAL;
1136
1137		switch (sin6->sin6_family) {
1138		case AF_INET6:
1139			if (addr_len < SIN6_LEN_RFC2133)
1140				return -EINVAL;
1141			daddr = &sin6->sin6_addr;
 
 
 
 
1142			break;
1143		case AF_INET:
1144			goto do_udp_sendmsg;
1145		case AF_UNSPEC:
1146			msg->msg_name = sin6 = NULL;
1147			msg->msg_namelen = addr_len = 0;
1148			daddr = NULL;
1149			break;
1150		default:
1151			return -EINVAL;
1152		}
1153	} else if (!up->pending) {
1154		if (sk->sk_state != TCP_ESTABLISHED)
1155			return -EDESTADDRREQ;
1156		daddr = &sk->sk_v6_daddr;
1157	} else
1158		daddr = NULL;
1159
1160	if (daddr) {
1161		if (ipv6_addr_v4mapped(daddr)) {
1162			struct sockaddr_in sin;
1163			sin.sin_family = AF_INET;
1164			sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1165			sin.sin_addr.s_addr = daddr->s6_addr32[3];
1166			msg->msg_name = &sin;
1167			msg->msg_namelen = sizeof(sin);
1168do_udp_sendmsg:
1169			if (__ipv6_only_sock(sk))
1170				return -ENETUNREACH;
1171			return udp_sendmsg(sk, msg, len);
 
 
1172		}
1173	}
1174
1175	if (up->pending == AF_INET)
1176		return udp_sendmsg(sk, msg, len);
1177
1178	/* Rough check on arithmetic overflow,
1179	   better check is made in ip6_append_data().
1180	   */
1181	if (len > INT_MAX - sizeof(struct udphdr))
1182		return -EMSGSIZE;
1183
1184	getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1185	if (up->pending) {
 
 
1186		/*
1187		 * There are pending frames.
1188		 * The socket lock must be held while it's corked.
1189		 */
1190		lock_sock(sk);
1191		if (likely(up->pending)) {
1192			if (unlikely(up->pending != AF_INET6)) {
1193				release_sock(sk);
1194				return -EAFNOSUPPORT;
1195			}
1196			dst = NULL;
1197			goto do_append_data;
1198		}
1199		release_sock(sk);
1200	}
1201	ulen += sizeof(struct udphdr);
1202
1203	memset(&fl6, 0, sizeof(fl6));
1204
1205	if (sin6) {
1206		if (sin6->sin6_port == 0)
1207			return -EINVAL;
1208
1209		fl6.fl6_dport = sin6->sin6_port;
1210		daddr = &sin6->sin6_addr;
1211
1212		if (np->sndflow) {
1213			fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1214			if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1215				flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1216				if (!flowlabel)
1217					return -EINVAL;
1218			}
1219		}
1220
1221		/*
1222		 * Otherwise it will be difficult to maintain
1223		 * sk->sk_dst_cache.
1224		 */
1225		if (sk->sk_state == TCP_ESTABLISHED &&
1226		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1227			daddr = &sk->sk_v6_daddr;
1228
1229		if (addr_len >= sizeof(struct sockaddr_in6) &&
1230		    sin6->sin6_scope_id &&
1231		    __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1232			fl6.flowi6_oif = sin6->sin6_scope_id;
1233	} else {
1234		if (sk->sk_state != TCP_ESTABLISHED)
1235			return -EDESTADDRREQ;
1236
1237		fl6.fl6_dport = inet->inet_dport;
1238		daddr = &sk->sk_v6_daddr;
1239		fl6.flowlabel = np->flow_label;
1240		connected = 1;
1241	}
1242
1243	if (!fl6.flowi6_oif)
1244		fl6.flowi6_oif = sk->sk_bound_dev_if;
1245
1246	if (!fl6.flowi6_oif)
1247		fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1248
1249	fl6.flowi6_mark = sk->sk_mark;
1250
1251	if (msg->msg_controllen) {
1252		opt = &opt_space;
1253		memset(opt, 0, sizeof(struct ipv6_txoptions));
1254		opt->tot_len = sizeof(*opt);
 
1255
1256		err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1257					    &hlimit, &tclass, &dontfrag);
 
 
 
 
1258		if (err < 0) {
1259			fl6_sock_release(flowlabel);
1260			return err;
1261		}
1262		if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1263			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1264			if (!flowlabel)
1265				return -EINVAL;
1266		}
1267		if (!(opt->opt_nflen|opt->opt_flen))
1268			opt = NULL;
1269		connected = 0;
1270	}
1271	if (!opt) {
1272		opt = txopt_get(np);
1273		opt_to_free = opt;
1274	}
1275	if (flowlabel)
1276		opt = fl6_merge_options(&opt_space, flowlabel, opt);
1277	opt = ipv6_fixup_options(&opt_space, opt);
 
1278
1279	fl6.flowi6_proto = sk->sk_protocol;
1280	if (!ipv6_addr_any(daddr))
1281		fl6.daddr = *daddr;
1282	else
1283		fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1284	if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1285		fl6.saddr = np->saddr;
1286	fl6.fl6_sport = inet->inet_sport;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1287
1288	final_p = fl6_update_dst(&fl6, opt, &final);
1289	if (final_p)
1290		connected = 0;
 
 
 
 
 
 
 
 
1291
1292	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1293		fl6.flowi6_oif = np->mcast_oif;
1294		connected = 0;
1295	} else if (!fl6.flowi6_oif)
1296		fl6.flowi6_oif = np->ucast_oif;
1297
1298	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1299
1300	dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
1301	if (IS_ERR(dst)) {
1302		err = PTR_ERR(dst);
1303		dst = NULL;
1304		goto out;
1305	}
1306
1307	if (hlimit < 0)
1308		hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1309
1310	if (tclass < 0)
1311		tclass = np->tclass;
1312
1313	if (msg->msg_flags&MSG_CONFIRM)
1314		goto do_confirm;
1315back_from_confirm:
1316
1317	/* Lockless fast path for the non-corking case */
1318	if (!corkreq) {
1319		struct sk_buff *skb;
1320
1321		skb = ip6_make_skb(sk, getfrag, msg, ulen,
1322				   sizeof(struct udphdr), hlimit, tclass, opt,
1323				   &fl6, (struct rt6_info *)dst,
1324				   msg->msg_flags, dontfrag);
1325		err = PTR_ERR(skb);
1326		if (!IS_ERR_OR_NULL(skb))
1327			err = udp_v6_send_skb(skb, &fl6);
1328		goto release_dst;
 
1329	}
1330
1331	lock_sock(sk);
1332	if (unlikely(up->pending)) {
1333		/* The socket is already corked while preparing it. */
1334		/* ... which is an evident application bug. --ANK */
1335		release_sock(sk);
1336
1337		net_dbg_ratelimited("udp cork app bug 2\n");
1338		err = -EINVAL;
1339		goto out;
1340	}
1341
1342	up->pending = AF_INET6;
1343
1344do_append_data:
1345	if (dontfrag < 0)
1346		dontfrag = np->dontfrag;
1347	up->len += ulen;
1348	err = ip6_append_data(sk, getfrag, msg, ulen,
1349		sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1350		(struct rt6_info *)dst,
1351		corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1352	if (err)
1353		udp_v6_flush_pending_frames(sk);
1354	else if (!corkreq)
1355		err = udp_v6_push_pending_frames(sk);
1356	else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1357		up->pending = 0;
1358
1359	if (err > 0)
1360		err = np->recverr ? net_xmit_errno(err) : 0;
1361	release_sock(sk);
1362
1363release_dst:
1364	if (dst) {
1365		if (connected) {
1366			ip6_dst_store(sk, dst,
1367				      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1368				      &sk->sk_v6_daddr : NULL,
1369#ifdef CONFIG_IPV6_SUBTREES
1370				      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1371				      &np->saddr :
1372#endif
1373				      NULL);
1374		} else {
1375			dst_release(dst);
1376		}
1377		dst = NULL;
1378	}
1379
1380out:
1381	dst_release(dst);
 
1382	fl6_sock_release(flowlabel);
1383	txopt_put(opt_to_free);
1384	if (!err)
1385		return len;
1386	/*
1387	 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1388	 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1389	 * we don't have a good statistic (IpOutDiscards but it can be too many
1390	 * things).  We could add another new stat but at least for now that
1391	 * seems like overkill.
1392	 */
1393	if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1394		UDP6_INC_STATS_USER(sock_net(sk),
1395				UDP_MIB_SNDBUFERRORS, is_udplite);
1396	}
1397	return err;
1398
1399do_confirm:
1400	dst_confirm(dst);
 
1401	if (!(msg->msg_flags&MSG_PROBE) || len)
1402		goto back_from_confirm;
1403	err = 0;
1404	goto out;
1405}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1406
1407void udpv6_destroy_sock(struct sock *sk)
1408{
1409	struct udp_sock *up = udp_sk(sk);
1410	lock_sock(sk);
 
 
 
1411	udp_v6_flush_pending_frames(sk);
1412	release_sock(sk);
1413
1414	if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1415		void (*encap_destroy)(struct sock *sk);
1416		encap_destroy = ACCESS_ONCE(up->encap_destroy);
1417		if (encap_destroy)
1418			encap_destroy(sk);
 
 
 
 
 
 
1419	}
1420
1421	inet6_destroy_sock(sk);
1422}
1423
1424/*
1425 *	Socket option code for UDP
1426 */
1427int udpv6_setsockopt(struct sock *sk, int level, int optname,
1428		     char __user *optval, unsigned int optlen)
1429{
1430	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1431		return udp_lib_setsockopt(sk, level, optname, optval, optlen,
 
1432					  udp_v6_push_pending_frames);
1433	return ipv6_setsockopt(sk, level, optname, optval, optlen);
1434}
1435
1436#ifdef CONFIG_COMPAT
1437int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1438			    char __user *optval, unsigned int optlen)
1439{
1440	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1441		return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1442					  udp_v6_push_pending_frames);
1443	return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1444}
1445#endif
1446
1447int udpv6_getsockopt(struct sock *sk, int level, int optname,
1448		     char __user *optval, int __user *optlen)
1449{
1450	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1451		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1452	return ipv6_getsockopt(sk, level, optname, optval, optlen);
1453}
1454
1455#ifdef CONFIG_COMPAT
1456int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1457			    char __user *optval, int __user *optlen)
1458{
1459	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1460		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1461	return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1462}
1463#endif
1464
1465static const struct inet6_protocol udpv6_protocol = {
1466	.handler	=	udpv6_rcv,
1467	.err_handler	=	udpv6_err,
1468	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1469};
1470
1471/* ------------------------------------------------------------------------ */
1472#ifdef CONFIG_PROC_FS
1473int udp6_seq_show(struct seq_file *seq, void *v)
1474{
1475	if (v == SEQ_START_TOKEN) {
1476		seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1477	} else {
1478		int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1479		struct inet_sock *inet = inet_sk(v);
1480		__u16 srcp = ntohs(inet->inet_sport);
1481		__u16 destp = ntohs(inet->inet_dport);
1482		ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
 
1483	}
1484	return 0;
1485}
1486
1487static const struct file_operations udp6_afinfo_seq_fops = {
1488	.owner    = THIS_MODULE,
1489	.open     = udp_seq_open,
1490	.read     = seq_read,
1491	.llseek   = seq_lseek,
1492	.release  = seq_release_net
1493};
 
1494
1495static struct udp_seq_afinfo udp6_seq_afinfo = {
1496	.name		= "udp6",
1497	.family		= AF_INET6,
1498	.udp_table	= &udp_table,
1499	.seq_fops	= &udp6_afinfo_seq_fops,
1500	.seq_ops	= {
1501		.show		= udp6_seq_show,
1502	},
1503};
1504
1505int __net_init udp6_proc_init(struct net *net)
1506{
1507	return udp_proc_register(net, &udp6_seq_afinfo);
 
 
 
1508}
1509
1510void udp6_proc_exit(struct net *net)
1511{
1512	udp_proc_unregister(net, &udp6_seq_afinfo);
1513}
1514#endif /* CONFIG_PROC_FS */
1515
1516void udp_v6_clear_sk(struct sock *sk, int size)
1517{
1518	struct inet_sock *inet = inet_sk(sk);
1519
1520	/* we do not want to clear pinet6 field, because of RCU lookups */
1521	sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1522
1523	size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1524	memset(&inet->pinet6 + 1, 0, size);
1525}
1526
1527/* ------------------------------------------------------------------------ */
1528
1529struct proto udpv6_prot = {
1530	.name		   = "UDPv6",
1531	.owner		   = THIS_MODULE,
1532	.close		   = udp_lib_close,
1533	.connect	   = ip6_datagram_connect,
1534	.disconnect	   = udp_disconnect,
1535	.ioctl		   = udp_ioctl,
1536	.destroy	   = udpv6_destroy_sock,
1537	.setsockopt	   = udpv6_setsockopt,
1538	.getsockopt	   = udpv6_getsockopt,
1539	.sendmsg	   = udpv6_sendmsg,
1540	.recvmsg	   = udpv6_recvmsg,
1541	.backlog_rcv	   = __udpv6_queue_rcv_skb,
1542	.release_cb	   = ip6_datagram_release_cb,
1543	.hash		   = udp_lib_hash,
1544	.unhash		   = udp_lib_unhash,
1545	.rehash		   = udp_v6_rehash,
1546	.get_port	   = udp_v6_get_port,
1547	.memory_allocated  = &udp_memory_allocated,
1548	.sysctl_mem	   = sysctl_udp_mem,
1549	.sysctl_wmem	   = &sysctl_udp_wmem_min,
1550	.sysctl_rmem	   = &sysctl_udp_rmem_min,
1551	.obj_size	   = sizeof(struct udp6_sock),
1552	.slab_flags	   = SLAB_DESTROY_BY_RCU,
1553	.h.udp_table	   = &udp_table,
1554#ifdef CONFIG_COMPAT
1555	.compat_setsockopt = compat_udpv6_setsockopt,
1556	.compat_getsockopt = compat_udpv6_getsockopt,
1557#endif
1558	.clear_sk	   = udp_v6_clear_sk,
 
 
 
 
 
 
 
 
 
 
1559};
1560
1561static struct inet_protosw udpv6_protosw = {
1562	.type =      SOCK_DGRAM,
1563	.protocol =  IPPROTO_UDP,
1564	.prot =      &udpv6_prot,
1565	.ops =       &inet6_dgram_ops,
1566	.flags =     INET_PROTOSW_PERMANENT,
1567};
1568
1569int __init udpv6_init(void)
1570{
1571	int ret;
1572
1573	ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
 
 
 
 
 
1574	if (ret)
1575		goto out;
1576
1577	ret = inet6_register_protosw(&udpv6_protosw);
1578	if (ret)
1579		goto out_udpv6_protocol;
1580out:
1581	return ret;
1582
1583out_udpv6_protocol:
1584	inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1585	goto out;
1586}
1587
1588void udpv6_exit(void)
1589{
1590	inet6_unregister_protosw(&udpv6_protosw);
1591	inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1592}
v6.9.4
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	UDP over IPv6
   4 *	Linux INET6 implementation
   5 *
   6 *	Authors:
   7 *	Pedro Roque		<roque@di.fc.ul.pt>
   8 *
   9 *	Based on linux/ipv4/udp.c
  10 *
  11 *	Fixes:
  12 *	Hideaki YOSHIFUJI	:	sin6_scope_id support
  13 *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
  14 *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
  15 *					a single port at the same time.
  16 *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
  17 *      YOSHIFUJI Hideaki @USAGI:	convert /proc/net/udp6 to seq_file.
 
 
 
 
 
  18 */
  19
  20#include <linux/bpf-cgroup.h>
  21#include <linux/errno.h>
  22#include <linux/types.h>
  23#include <linux/socket.h>
  24#include <linux/sockios.h>
  25#include <linux/net.h>
  26#include <linux/in6.h>
  27#include <linux/netdevice.h>
  28#include <linux/if_arp.h>
  29#include <linux/ipv6.h>
  30#include <linux/icmpv6.h>
  31#include <linux/init.h>
  32#include <linux/module.h>
  33#include <linux/skbuff.h>
  34#include <linux/slab.h>
  35#include <linux/uaccess.h>
  36#include <linux/indirect_call_wrapper.h>
  37
  38#include <net/addrconf.h>
  39#include <net/ndisc.h>
  40#include <net/protocol.h>
  41#include <net/transp_v6.h>
  42#include <net/ip6_route.h>
  43#include <net/raw.h>
  44#include <net/seg6.h>
  45#include <net/tcp_states.h>
  46#include <net/ip6_checksum.h>
  47#include <net/ip6_tunnel.h>
  48#include <trace/events/udp.h>
  49#include <net/xfrm.h>
  50#include <net/inet_hashtables.h>
  51#include <net/inet6_hashtables.h>
  52#include <net/busy_poll.h>
  53#include <net/sock_reuseport.h>
  54#include <net/gro.h>
  55
  56#include <linux/proc_fs.h>
  57#include <linux/seq_file.h>
  58#include <trace/events/skb.h>
  59#include "udp_impl.h"
  60
  61static void udpv6_destruct_sock(struct sock *sk)
  62{
  63	udp_destruct_common(sk);
  64	inet6_sock_destruct(sk);
  65}
  66
  67int udpv6_init_sock(struct sock *sk)
  68{
  69	udp_lib_init_sock(sk);
  70	sk->sk_destruct = udpv6_destruct_sock;
  71	set_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags);
  72	return 0;
  73}
  74
  75INDIRECT_CALLABLE_SCOPE
  76u32 udp6_ehashfn(const struct net *net,
  77		 const struct in6_addr *laddr,
  78		 const u16 lport,
  79		 const struct in6_addr *faddr,
  80		 const __be16 fport)
  81{
  82	u32 lhash, fhash;
  83
  84	net_get_random_once(&udp6_ehash_secret,
  85			    sizeof(udp6_ehash_secret));
  86	net_get_random_once(&udp_ipv6_hash_secret,
  87			    sizeof(udp_ipv6_hash_secret));
  88
  89	lhash = (__force u32)laddr->s6_addr32[3];
  90	fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
  91
  92	return __inet6_ehashfn(lhash, lport, fhash, fport,
  93			       udp6_ehash_secret + net_hash_mix(net));
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  94}
  95
  96int udp_v6_get_port(struct sock *sk, unsigned short snum)
  97{
  98	unsigned int hash2_nulladdr =
  99		ipv6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
 100	unsigned int hash2_partial =
 101		ipv6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
 102
 103	/* precompute partial secondary hash */
 104	udp_sk(sk)->udp_portaddr_hash = hash2_partial;
 105	return udp_lib_get_port(sk, snum, hash2_nulladdr);
 106}
 107
 108void udp_v6_rehash(struct sock *sk)
 109{
 110	u16 new_hash = ipv6_portaddr_hash(sock_net(sk),
 111					  &sk->sk_v6_rcv_saddr,
 112					  inet_sk(sk)->inet_num);
 113
 114	udp_lib_rehash(sk, new_hash);
 115}
 116
 117static int compute_score(struct sock *sk, struct net *net,
 118			 const struct in6_addr *saddr, __be16 sport,
 119			 const struct in6_addr *daddr, unsigned short hnum,
 120			 int dif, int sdif)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 121{
 122	int bound_dev_if, score;
 123	struct inet_sock *inet;
 124	bool dev_match;
 125
 126	if (!net_eq(sock_net(sk), net) ||
 127	    udp_sk(sk)->udp_port_hash != hnum ||
 128	    sk->sk_family != PF_INET6)
 129		return -1;
 130
 131	if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
 132		return -1;
 133
 134	score = 0;
 135	inet = inet_sk(sk);
 136
 137	if (inet->inet_dport) {
 138		if (inet->inet_dport != sport)
 139			return -1;
 140		score++;
 141	}
 142
 143	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
 144		if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
 145			return -1;
 146		score++;
 147	}
 148
 149	bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
 150	dev_match = udp_sk_bound_dev_eq(net, bound_dev_if, dif, sdif);
 151	if (!dev_match)
 152		return -1;
 153	if (bound_dev_if)
 154		score++;
 
 155
 156	if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id())
 157		score++;
 158
 159	return score;
 160}
 161
 162/* called with rcu_read_lock() */
 163static struct sock *udp6_lib_lookup2(struct net *net,
 164		const struct in6_addr *saddr, __be16 sport,
 165		const struct in6_addr *daddr, unsigned int hnum,
 166		int dif, int sdif, struct udp_hslot *hslot2,
 167		struct sk_buff *skb)
 168{
 169	struct sock *sk, *result;
 170	int score, badness;
 171	bool need_rescore;
 
 
 172
 
 173	result = NULL;
 174	badness = -1;
 175	udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
 176		need_rescore = false;
 177rescore:
 178		score = compute_score(need_rescore ? result : sk, net, saddr,
 179				      sport, daddr, hnum, dif, sdif);
 180		if (score > badness) {
 
 181			badness = score;
 182
 183			if (need_rescore)
 184				continue;
 185
 186			if (sk->sk_state == TCP_ESTABLISHED) {
 187				result = sk;
 188				continue;
 
 
 
 
 
 
 
 
 
 189			}
 190
 191			result = inet6_lookup_reuseport(net, sk, skb, sizeof(struct udphdr),
 192							saddr, sport, daddr, hnum, udp6_ehashfn);
 193			if (!result) {
 194				result = sk;
 195				continue;
 196			}
 
 
 
 
 
 
 
 
 197
 198			/* Fall back to scoring if group has connections */
 199			if (!reuseport_has_conns(sk))
 200				return result;
 201
 202			/* Reuseport logic returned an error, keep original score. */
 203			if (IS_ERR(result))
 204				continue;
 205
 206			/* compute_score is too long of a function to be
 207			 * inlined, and calling it again here yields
 208			 * measureable overhead for some
 209			 * workloads. Work around it by jumping
 210			 * backwards to rescore 'result'.
 211			 */
 212			need_rescore = true;
 213			goto rescore;
 214		}
 215	}
 216	return result;
 217}
 218
 219/* rcu_read_lock() must be held */
 220struct sock *__udp6_lib_lookup(struct net *net,
 221			       const struct in6_addr *saddr, __be16 sport,
 222			       const struct in6_addr *daddr, __be16 dport,
 223			       int dif, int sdif, struct udp_table *udptable,
 224			       struct sk_buff *skb)
 225{
 
 
 226	unsigned short hnum = ntohs(dport);
 227	unsigned int hash2, slot2;
 228	struct udp_hslot *hslot2;
 229	struct sock *result, *sk;
 230
 231	hash2 = ipv6_portaddr_hash(net, daddr, hnum);
 232	slot2 = hash2 & udptable->mask;
 233	hslot2 = &udptable->hash2[slot2];
 234
 235	/* Lookup connected or non-wildcard sockets */
 236	result = udp6_lib_lookup2(net, saddr, sport,
 237				  daddr, hnum, dif, sdif,
 238				  hslot2, skb);
 239	if (!IS_ERR_OR_NULL(result) && result->sk_state == TCP_ESTABLISHED)
 240		goto done;
 241
 242	/* Lookup redirect from BPF */
 243	if (static_branch_unlikely(&bpf_sk_lookup_enabled) &&
 244	    udptable == net->ipv4.udp_table) {
 245		sk = inet6_lookup_run_sk_lookup(net, IPPROTO_UDP, skb, sizeof(struct udphdr),
 246						saddr, sport, daddr, hnum, dif,
 247						udp6_ehashfn);
 248		if (sk) {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 249			result = sk;
 250			goto done;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 251		}
 252	}
 
 
 
 
 
 
 
 253
 254	/* Got non-wildcard socket or error on first lookup */
 255	if (result)
 256		goto done;
 257
 258	/* Lookup wildcard sockets */
 259	hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum);
 260	slot2 = hash2 & udptable->mask;
 261	hslot2 = &udptable->hash2[slot2];
 262
 263	result = udp6_lib_lookup2(net, saddr, sport,
 264				  &in6addr_any, hnum, dif, sdif,
 265				  hslot2, skb);
 266done:
 267	if (IS_ERR(result))
 268		return NULL;
 269	return result;
 270}
 271EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
 272
 273static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
 274					  __be16 sport, __be16 dport,
 275					  struct udp_table *udptable)
 276{
 
 277	const struct ipv6hdr *iph = ipv6_hdr(skb);
 278
 279	return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
 
 
 
 280				 &iph->daddr, dport, inet6_iif(skb),
 281				 inet6_sdif(skb), udptable, skb);
 282}
 283
 284struct sock *udp6_lib_lookup_skb(const struct sk_buff *skb,
 285				 __be16 sport, __be16 dport)
 286{
 287	const u16 offset = NAPI_GRO_CB(skb)->network_offsets[skb->encapsulation];
 288	const struct ipv6hdr *iph = (struct ipv6hdr *)(skb->data + offset);
 289	struct net *net = dev_net(skb->dev);
 290	int iif, sdif;
 291
 292	inet6_get_iif_sdif(skb, &iif, &sdif);
 293
 294	return __udp6_lib_lookup(net, &iph->saddr, sport,
 295				 &iph->daddr, dport, iif,
 296				 sdif, net->ipv4.udp_table, NULL);
 297}
 298
 299/* Must be called under rcu_read_lock().
 300 * Does increment socket refcount.
 301 */
 302#if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6)
 303struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
 304			     const struct in6_addr *daddr, __be16 dport, int dif)
 305{
 306	struct sock *sk;
 307
 308	sk =  __udp6_lib_lookup(net, saddr, sport, daddr, dport,
 309				dif, 0, net->ipv4.udp_table, NULL);
 310	if (sk && !refcount_inc_not_zero(&sk->sk_refcnt))
 311		sk = NULL;
 312	return sk;
 313}
 314EXPORT_SYMBOL_GPL(udp6_lib_lookup);
 315#endif
 316
 317/* do not use the scratch area len for jumbogram: their length execeeds the
 318 * scratch area space; note that the IP6CB flags is still in the first
 319 * cacheline, so checking for jumbograms is cheap
 320 */
 321static int udp6_skb_len(struct sk_buff *skb)
 322{
 323	return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb);
 324}
 325
 326/*
 327 *	This should be easy, if there is something there we
 328 *	return it, otherwise we block.
 329 */
 330
 331int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 332		  int flags, int *addr_len)
 333{
 334	struct ipv6_pinfo *np = inet6_sk(sk);
 335	struct inet_sock *inet = inet_sk(sk);
 336	struct sk_buff *skb;
 337	unsigned int ulen, copied;
 338	int off, err, peeking = flags & MSG_PEEK;
 
 339	int is_udplite = IS_UDPLITE(sk);
 340	struct udp_mib __percpu *mib;
 341	bool checksum_valid = false;
 342	int is_udp4;
 
 343
 344	if (flags & MSG_ERRQUEUE)
 345		return ipv6_recv_error(sk, msg, len, addr_len);
 346
 347	if (np->rxpmtu && np->rxopt.bits.rxpmtu)
 348		return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
 349
 350try_again:
 351	off = sk_peek_offset(sk, flags);
 352	skb = __skb_recv_udp(sk, flags, &off, &err);
 353	if (!skb)
 354		return err;
 355
 356	ulen = udp6_skb_len(skb);
 357	copied = len;
 358	if (copied > ulen - off)
 359		copied = ulen - off;
 360	else if (copied < ulen)
 361		msg->msg_flags |= MSG_TRUNC;
 362
 363	is_udp4 = (skb->protocol == htons(ETH_P_IP));
 364	mib = __UDPX_MIB(sk, is_udp4);
 365
 366	/*
 367	 * If checksum is needed at all, try to do it while copying the
 368	 * data.  If the data is truncated, or if we only want a partial
 369	 * coverage checksum (UDP-Lite), do it before the copy.
 370	 */
 371
 372	if (copied < ulen || peeking ||
 373	    (is_udplite && UDP_SKB_CB(skb)->partial_cov)) {
 374		checksum_valid = udp_skb_csum_unnecessary(skb) ||
 375				!__udp_lib_checksum_complete(skb);
 376		if (!checksum_valid)
 377			goto csum_copy_err;
 378	}
 379
 380	if (checksum_valid || udp_skb_csum_unnecessary(skb)) {
 381		if (udp_skb_is_linear(skb))
 382			err = copy_linear_skb(skb, copied, off, &msg->msg_iter);
 383		else
 384			err = skb_copy_datagram_msg(skb, off, msg, copied);
 385	} else {
 386		err = skb_copy_and_csum_datagram_msg(skb, off, msg);
 387		if (err == -EINVAL)
 388			goto csum_copy_err;
 389	}
 390	if (unlikely(err)) {
 391		if (!peeking) {
 
 392			atomic_inc(&sk->sk_drops);
 393			SNMP_INC_STATS(mib, UDP_MIB_INERRORS);
 394		}
 395		kfree_skb(skb);
 396		return err;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 397	}
 398	if (!peeking)
 399		SNMP_INC_STATS(mib, UDP_MIB_INDATAGRAMS);
 400
 401	sock_recv_cmsgs(msg, sk, skb);
 402
 403	/* Copy the address. */
 404	if (msg->msg_name) {
 405		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 406		sin6->sin6_family = AF_INET6;
 407		sin6->sin6_port = udp_hdr(skb)->source;
 408		sin6->sin6_flowinfo = 0;
 409
 410		if (is_udp4) {
 411			ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
 412					       &sin6->sin6_addr);
 413			sin6->sin6_scope_id = 0;
 414		} else {
 415			sin6->sin6_addr = ipv6_hdr(skb)->saddr;
 416			sin6->sin6_scope_id =
 417				ipv6_iface_scope_id(&sin6->sin6_addr,
 418						    inet6_iif(skb));
 419		}
 420		*addr_len = sizeof(*sin6);
 421
 422		BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk,
 423						      (struct sockaddr *)sin6,
 424						      addr_len);
 425	}
 426
 427	if (udp_test_bit(GRO_ENABLED, sk))
 428		udp_cmsg_recv(msg, sk, skb);
 429
 430	if (np->rxopt.all)
 431		ip6_datagram_recv_common_ctl(sk, msg, skb);
 432
 433	if (is_udp4) {
 434		if (inet_cmsg_flags(inet))
 435			ip_cmsg_recv_offset(msg, sk, skb,
 436					    sizeof(struct udphdr), off);
 437	} else {
 438		if (np->rxopt.all)
 439			ip6_datagram_recv_specific_ctl(sk, msg, skb);
 440	}
 441
 442	err = copied;
 443	if (flags & MSG_TRUNC)
 444		err = ulen;
 445
 446	skb_consume_udp(sk, skb, peeking ? -err : err);
 
 
 447	return err;
 448
 449csum_copy_err:
 450	if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags,
 451				 udp_skb_destructor)) {
 452		SNMP_INC_STATS(mib, UDP_MIB_CSUMERRORS);
 453		SNMP_INC_STATS(mib, UDP_MIB_INERRORS);
 
 
 
 
 
 
 
 
 
 454	}
 455	kfree_skb(skb);
 456
 457	/* starting over for a new packet, but check if we need to yield */
 458	cond_resched();
 459	msg->msg_flags &= ~MSG_TRUNC;
 460	goto try_again;
 461}
 462
 463DECLARE_STATIC_KEY_FALSE(udpv6_encap_needed_key);
 464void udpv6_encap_enable(void)
 465{
 466	static_branch_inc(&udpv6_encap_needed_key);
 467}
 468EXPORT_SYMBOL(udpv6_encap_enable);
 469
 470/* Handler for tunnels with arbitrary destination ports: no socket lookup, go
 471 * through error handlers in encapsulations looking for a match.
 472 */
 473static int __udp6_lib_err_encap_no_sk(struct sk_buff *skb,
 474				      struct inet6_skb_parm *opt,
 475				      u8 type, u8 code, int offset, __be32 info)
 476{
 477	int i;
 478
 479	for (i = 0; i < MAX_IPTUN_ENCAP_OPS; i++) {
 480		int (*handler)(struct sk_buff *skb, struct inet6_skb_parm *opt,
 481			       u8 type, u8 code, int offset, __be32 info);
 482		const struct ip6_tnl_encap_ops *encap;
 483
 484		encap = rcu_dereference(ip6tun_encaps[i]);
 485		if (!encap)
 486			continue;
 487		handler = encap->err_handler;
 488		if (handler && !handler(skb, opt, type, code, offset, info))
 489			return 0;
 490	}
 491
 492	return -ENOENT;
 493}
 494
 495/* Try to match ICMP errors to UDP tunnels by looking up a socket without
 496 * reversing source and destination port: this will match tunnels that force the
 497 * same destination port on both endpoints (e.g. VXLAN, GENEVE). Note that
 498 * lwtunnels might actually break this assumption by being configured with
 499 * different destination ports on endpoints, in this case we won't be able to
 500 * trace ICMP messages back to them.
 501 *
 502 * If this doesn't match any socket, probe tunnels with arbitrary destination
 503 * ports (e.g. FoU, GUE): there, the receiving socket is useless, as the port
 504 * we've sent packets to won't necessarily match the local destination port.
 505 *
 506 * Then ask the tunnel implementation to match the error against a valid
 507 * association.
 508 *
 509 * Return an error if we can't find a match, the socket if we need further
 510 * processing, zero otherwise.
 511 */
 512static struct sock *__udp6_lib_err_encap(struct net *net,
 513					 const struct ipv6hdr *hdr, int offset,
 514					 struct udphdr *uh,
 515					 struct udp_table *udptable,
 516					 struct sock *sk,
 517					 struct sk_buff *skb,
 518					 struct inet6_skb_parm *opt,
 519					 u8 type, u8 code, __be32 info)
 520{
 521	int (*lookup)(struct sock *sk, struct sk_buff *skb);
 522	int network_offset, transport_offset;
 523	struct udp_sock *up;
 524
 525	network_offset = skb_network_offset(skb);
 526	transport_offset = skb_transport_offset(skb);
 527
 528	/* Network header needs to point to the outer IPv6 header inside ICMP */
 529	skb_reset_network_header(skb);
 530
 531	/* Transport header needs to point to the UDP header */
 532	skb_set_transport_header(skb, offset);
 533
 534	if (sk) {
 535		up = udp_sk(sk);
 536
 537		lookup = READ_ONCE(up->encap_err_lookup);
 538		if (lookup && lookup(sk, skb))
 539			sk = NULL;
 540
 541		goto out;
 542	}
 543
 544	sk = __udp6_lib_lookup(net, &hdr->daddr, uh->source,
 545			       &hdr->saddr, uh->dest,
 546			       inet6_iif(skb), 0, udptable, skb);
 547	if (sk) {
 548		up = udp_sk(sk);
 549
 550		lookup = READ_ONCE(up->encap_err_lookup);
 551		if (!lookup || lookup(sk, skb))
 552			sk = NULL;
 553	}
 554
 555out:
 556	if (!sk) {
 557		sk = ERR_PTR(__udp6_lib_err_encap_no_sk(skb, opt, type, code,
 558							offset, info));
 559	}
 560
 561	skb_set_transport_header(skb, transport_offset);
 562	skb_set_network_header(skb, network_offset);
 563
 564	return sk;
 565}
 566
 567int __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 568		   u8 type, u8 code, int offset, __be32 info,
 569		   struct udp_table *udptable)
 570{
 571	struct ipv6_pinfo *np;
 572	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
 573	const struct in6_addr *saddr = &hdr->saddr;
 574	const struct in6_addr *daddr = seg6_get_daddr(skb, opt) ? : &hdr->daddr;
 575	struct udphdr *uh = (struct udphdr *)(skb->data+offset);
 576	bool tunnel = false;
 577	struct sock *sk;
 578	int harderr;
 579	int err;
 580	struct net *net = dev_net(skb->dev);
 581
 582	sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
 583			       inet6_iif(skb), inet6_sdif(skb), udptable, NULL);
 584
 585	if (!sk || READ_ONCE(udp_sk(sk)->encap_type)) {
 586		/* No socket for error: try tunnels before discarding */
 587		if (static_branch_unlikely(&udpv6_encap_needed_key)) {
 588			sk = __udp6_lib_err_encap(net, hdr, offset, uh,
 589						  udptable, sk, skb,
 590						  opt, type, code, info);
 591			if (!sk)
 592				return 0;
 593		} else
 594			sk = ERR_PTR(-ENOENT);
 595
 596		if (IS_ERR(sk)) {
 597			__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
 598					  ICMP6_MIB_INERRORS);
 599			return PTR_ERR(sk);
 600		}
 601
 602		tunnel = true;
 603	}
 604
 605	harderr = icmpv6_err_convert(type, code, &err);
 606	np = inet6_sk(sk);
 607
 608	if (type == ICMPV6_PKT_TOOBIG) {
 609		if (!ip6_sk_accept_pmtu(sk))
 610			goto out;
 611		ip6_sk_update_pmtu(skb, sk, info);
 612		if (READ_ONCE(np->pmtudisc) != IPV6_PMTUDISC_DONT)
 613			harderr = 1;
 614	}
 615	if (type == NDISC_REDIRECT) {
 616		if (tunnel) {
 617			ip6_redirect(skb, sock_net(sk), inet6_iif(skb),
 618				     READ_ONCE(sk->sk_mark), sk->sk_uid);
 619		} else {
 620			ip6_sk_redirect(skb, sk);
 621		}
 622		goto out;
 623	}
 624
 625	/* Tunnels don't have an application socket: don't pass errors back */
 626	if (tunnel) {
 627		if (udp_sk(sk)->encap_err_rcv)
 628			udp_sk(sk)->encap_err_rcv(sk, skb, err, uh->dest,
 629						  ntohl(info), (u8 *)(uh+1));
 630		goto out;
 631	}
 632
 633	if (!inet6_test_bit(RECVERR6, sk)) {
 634		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 635			goto out;
 636	} else {
 637		ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
 638	}
 639
 640	sk->sk_err = err;
 641	sk_error_report(sk);
 642out:
 643	return 0;
 644}
 645
 646static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 647{
 648	int rc;
 649
 650	if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
 651		sock_rps_save_rxhash(sk, skb);
 652		sk_mark_napi_id(sk, skb);
 653		sk_incoming_cpu_update(sk);
 654	} else {
 655		sk_mark_napi_id_once(sk, skb);
 656	}
 657
 658	rc = __udp_enqueue_schedule_skb(sk, skb);
 659	if (rc < 0) {
 660		int is_udplite = IS_UDPLITE(sk);
 661		enum skb_drop_reason drop_reason;
 662
 663		/* Note that an ENOMEM error is charged twice */
 664		if (rc == -ENOMEM) {
 665			UDP6_INC_STATS(sock_net(sk),
 666					 UDP_MIB_RCVBUFERRORS, is_udplite);
 667			drop_reason = SKB_DROP_REASON_SOCKET_RCVBUFF;
 668		} else {
 669			UDP6_INC_STATS(sock_net(sk),
 670				       UDP_MIB_MEMERRORS, is_udplite);
 671			drop_reason = SKB_DROP_REASON_PROTO_MEM;
 672		}
 673		UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 674		kfree_skb_reason(skb, drop_reason);
 675		trace_udp_fail_queue_rcv_skb(rc, sk);
 676		return -1;
 677	}
 
 
 678
 679	return 0;
 
 
 
 
 680}
 681
 682static __inline__ int udpv6_err(struct sk_buff *skb,
 683				struct inet6_skb_parm *opt, u8 type,
 684				u8 code, int offset, __be32 info)
 685{
 686	return __udp6_lib_err(skb, opt, type, code, offset, info,
 687			      dev_net(skb->dev)->ipv4.udp_table);
 688}
 
 689
 690static int udpv6_queue_rcv_one_skb(struct sock *sk, struct sk_buff *skb)
 691{
 692	enum skb_drop_reason drop_reason = SKB_DROP_REASON_NOT_SPECIFIED;
 693	struct udp_sock *up = udp_sk(sk);
 
 694	int is_udplite = IS_UDPLITE(sk);
 695
 696	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
 697		drop_reason = SKB_DROP_REASON_XFRM_POLICY;
 698		goto drop;
 699	}
 700	nf_reset_ct(skb);
 701
 702	if (static_branch_unlikely(&udpv6_encap_needed_key) &&
 703	    READ_ONCE(up->encap_type)) {
 704		int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
 705
 706		/*
 707		 * This is an encapsulation socket so pass the skb to
 708		 * the socket's udp_encap_rcv() hook. Otherwise, just
 709		 * fall through and pass this up the UDP socket.
 710		 * up->encap_rcv() returns the following value:
 711		 * =0 if skb was successfully passed to the encap
 712		 *    handler or was discarded by it.
 713		 * >0 if skb should be passed on to UDP.
 714		 * <0 if skb should be resubmitted as proto -N
 715		 */
 716
 717		/* if we're overly short, let UDP handle it */
 718		encap_rcv = READ_ONCE(up->encap_rcv);
 719		if (encap_rcv) {
 720			int ret;
 721
 722			/* Verify checksum before giving to encap */
 723			if (udp_lib_checksum_complete(skb))
 724				goto csum_error;
 725
 726			ret = encap_rcv(sk, skb);
 727			if (ret <= 0) {
 728				__UDP6_INC_STATS(sock_net(sk),
 729						 UDP_MIB_INDATAGRAMS,
 730						 is_udplite);
 731				return -ret;
 732			}
 733		}
 734
 735		/* FALLTHROUGH -- it's a UDP Packet */
 736	}
 737
 738	/*
 739	 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
 740	 */
 741	if (udp_test_bit(UDPLITE_RECV_CC, sk) && UDP_SKB_CB(skb)->partial_cov) {
 742		u16 pcrlen = READ_ONCE(up->pcrlen);
 743
 744		if (pcrlen == 0) {          /* full coverage was set  */
 745			net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
 746					    UDP_SKB_CB(skb)->cscov, skb->len);
 747			goto drop;
 748		}
 749		if (UDP_SKB_CB(skb)->cscov < pcrlen) {
 750			net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
 751					    UDP_SKB_CB(skb)->cscov, pcrlen);
 752			goto drop;
 753		}
 754	}
 755
 756	prefetch(&sk->sk_rmem_alloc);
 757	if (rcu_access_pointer(sk->sk_filter) &&
 758	    udp_lib_checksum_complete(skb))
 759		goto csum_error;
 760
 761	if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr))) {
 762		drop_reason = SKB_DROP_REASON_SOCKET_FILTER;
 
 763		goto drop;
 764	}
 765
 766	udp_csum_pull_header(skb);
 767
 768	skb_dst_drop(skb);
 
 
 
 
 
 
 
 
 769
 770	return __udpv6_queue_rcv_skb(sk, skb);
 771
 772csum_error:
 773	drop_reason = SKB_DROP_REASON_UDP_CSUM;
 774	__UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
 775drop:
 776	__UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 777	atomic_inc(&sk->sk_drops);
 778	kfree_skb_reason(skb, drop_reason);
 779	return -1;
 780}
 781
 782static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 783{
 784	struct sk_buff *next, *segs;
 785	int ret;
 786
 787	if (likely(!udp_unexpected_gso(sk, skb)))
 788		return udpv6_queue_rcv_one_skb(sk, skb);
 789
 790	__skb_push(skb, -skb_mac_offset(skb));
 791	segs = udp_rcv_segment(sk, skb, false);
 792	skb_list_walk_safe(segs, skb, next) {
 793		__skb_pull(skb, skb_transport_offset(skb));
 794
 795		udp_post_segment_fix_csum(skb);
 796		ret = udpv6_queue_rcv_one_skb(sk, skb);
 797		if (ret > 0)
 798			ip6_protocol_deliver_rcu(dev_net(skb->dev), skb, ret,
 799						 true);
 800	}
 801	return 0;
 802}
 803
 804static bool __udp_v6_is_mcast_sock(struct net *net, const struct sock *sk,
 805				   __be16 loc_port, const struct in6_addr *loc_addr,
 806				   __be16 rmt_port, const struct in6_addr *rmt_addr,
 807				   int dif, int sdif, unsigned short hnum)
 808{
 809	const struct inet_sock *inet = inet_sk(sk);
 810
 811	if (!net_eq(sock_net(sk), net))
 812		return false;
 813
 814	if (udp_sk(sk)->udp_port_hash != hnum ||
 815	    sk->sk_family != PF_INET6 ||
 816	    (inet->inet_dport && inet->inet_dport != rmt_port) ||
 817	    (!ipv6_addr_any(&sk->sk_v6_daddr) &&
 818		    !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
 819	    !udp_sk_bound_dev_eq(net, READ_ONCE(sk->sk_bound_dev_if), dif, sdif) ||
 820	    (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 821		    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
 822		return false;
 823	if (!inet6_mc_check(sk, loc_addr, rmt_addr))
 824		return false;
 825	return true;
 826}
 827
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 828static void udp6_csum_zero_error(struct sk_buff *skb)
 829{
 830	/* RFC 2460 section 8.1 says that we SHOULD log
 831	 * this error. Well, it is reasonable.
 832	 */
 833	net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
 834			    &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
 835			    &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
 836}
 837
 838/*
 839 * Note: called only from the BH handler context,
 840 * so we don't need to lock the hashes.
 841 */
 842static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
 843		const struct in6_addr *saddr, const struct in6_addr *daddr,
 844		struct udp_table *udptable, int proto)
 845{
 846	struct sock *sk, *first = NULL;
 847	const struct udphdr *uh = udp_hdr(skb);
 
 848	unsigned short hnum = ntohs(uh->dest);
 849	struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
 850	unsigned int offset = offsetof(typeof(*sk), sk_node);
 
 851	unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
 852	int dif = inet6_iif(skb);
 853	int sdif = inet6_sdif(skb);
 854	struct hlist_node *node;
 855	struct sk_buff *nskb;
 856
 857	if (use_hash2) {
 858		hash2_any = ipv6_portaddr_hash(net, &in6addr_any, hnum) &
 859			    udptable->mask;
 860		hash2 = ipv6_portaddr_hash(net, daddr, hnum) & udptable->mask;
 861start_lookup:
 862		hslot = &udptable->hash2[hash2];
 863		offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
 864	}
 865
 866	sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
 867		if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
 868					    uh->source, saddr, dif, sdif,
 869					    hnum))
 870			continue;
 871		/* If zero checksum and no_check is not on for
 872		 * the socket then skip it.
 873		 */
 874		if (!uh->check && !udp_get_no_check6_rx(sk))
 875			continue;
 876		if (!first) {
 877			first = sk;
 878			continue;
 879		}
 880		nskb = skb_clone(skb, GFP_ATOMIC);
 881		if (unlikely(!nskb)) {
 882			atomic_inc(&sk->sk_drops);
 883			__UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
 884					 IS_UDPLITE(sk));
 885			__UDP6_INC_STATS(net, UDP_MIB_INERRORS,
 886					 IS_UDPLITE(sk));
 887			continue;
 888		}
 
 889
 890		if (udpv6_queue_rcv_skb(sk, nskb) > 0)
 891			consume_skb(nskb);
 892	}
 893
 894	/* Also lookup *:port if we are using hash2 and haven't done so yet. */
 895	if (use_hash2 && hash2 != hash2_any) {
 896		hash2 = hash2_any;
 897		goto start_lookup;
 898	}
 899
 900	if (first) {
 901		if (udpv6_queue_rcv_skb(first, skb) > 0)
 902			consume_skb(skb);
 903	} else {
 904		kfree_skb(skb);
 905		__UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
 906				 proto == IPPROTO_UDPLITE);
 
 907	}
 908	return 0;
 909}
 910
 911static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
 912{
 913	if (udp_sk_rx_dst_set(sk, dst))
 914		sk->sk_rx_dst_cookie = rt6_get_cookie(dst_rt6_info(dst));
 915}
 916
 917/* wrapper for udp_queue_rcv_skb tacking care of csum conversion and
 918 * return code conversion for ip layer consumption
 919 */
 920static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb,
 921				struct udphdr *uh)
 922{
 923	int ret;
 924
 925	if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
 926		skb_checksum_try_convert(skb, IPPROTO_UDP, ip6_compute_pseudo);
 927
 928	ret = udpv6_queue_rcv_skb(sk, skb);
 929
 930	/* a return value > 0 means to resubmit the input */
 931	if (ret > 0)
 932		return ret;
 933	return 0;
 934}
 935
 936int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
 937		   int proto)
 938{
 939	enum skb_drop_reason reason = SKB_DROP_REASON_NOT_SPECIFIED;
 940	const struct in6_addr *saddr, *daddr;
 941	struct net *net = dev_net(skb->dev);
 
 942	struct udphdr *uh;
 943	struct sock *sk;
 944	bool refcounted;
 945	u32 ulen = 0;
 946
 947	if (!pskb_may_pull(skb, sizeof(struct udphdr)))
 948		goto discard;
 949
 950	saddr = &ipv6_hdr(skb)->saddr;
 951	daddr = &ipv6_hdr(skb)->daddr;
 952	uh = udp_hdr(skb);
 953
 954	ulen = ntohs(uh->len);
 955	if (ulen > skb->len)
 956		goto short_packet;
 957
 958	if (proto == IPPROTO_UDP) {
 959		/* UDP validates ulen. */
 960
 961		/* Check for jumbo payload */
 962		if (ulen == 0)
 963			ulen = skb->len;
 964
 965		if (ulen < sizeof(*uh))
 966			goto short_packet;
 967
 968		if (ulen < skb->len) {
 969			if (pskb_trim_rcsum(skb, ulen))
 970				goto short_packet;
 971			saddr = &ipv6_hdr(skb)->saddr;
 972			daddr = &ipv6_hdr(skb)->daddr;
 973			uh = udp_hdr(skb);
 974		}
 975	}
 976
 977	if (udp6_csum_init(skb, uh, proto))
 978		goto csum_error;
 979
 980	/* Check if the socket is already available, e.g. due to early demux */
 981	sk = inet6_steal_sock(net, skb, sizeof(struct udphdr), saddr, uh->source, daddr, uh->dest,
 982			      &refcounted, udp6_ehashfn);
 983	if (IS_ERR(sk))
 984		goto no_sk;
 985
 986	if (sk) {
 987		struct dst_entry *dst = skb_dst(skb);
 988		int ret;
 989
 990		if (unlikely(rcu_dereference(sk->sk_rx_dst) != dst))
 991			udp6_sk_rx_dst_set(sk, dst);
 992
 993		if (!uh->check && !udp_get_no_check6_rx(sk)) {
 994			if (refcounted)
 995				sock_put(sk);
 996			goto report_csum_error;
 997		}
 998
 999		ret = udp6_unicast_rcv_skb(sk, skb, uh);
1000		if (refcounted)
1001			sock_put(sk);
1002		return ret;
1003	}
1004
1005	/*
1006	 *	Multicast receive code
1007	 */
1008	if (ipv6_addr_is_multicast(daddr))
1009		return __udp6_lib_mcast_deliver(net, skb,
1010				saddr, daddr, udptable, proto);
1011
1012	/* Unicast */
 
 
 
 
 
1013	sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
1014	if (sk) {
1015		if (!uh->check && !udp_get_no_check6_rx(sk))
1016			goto report_csum_error;
1017		return udp6_unicast_rcv_skb(sk, skb, uh);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1018	}
1019no_sk:
1020	reason = SKB_DROP_REASON_NO_SOCKET;
1021
1022	if (!uh->check)
1023		goto report_csum_error;
 
 
1024
1025	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1026		goto discard;
1027	nf_reset_ct(skb);
1028
1029	if (udp_lib_checksum_complete(skb))
1030		goto csum_error;
1031
1032	__UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
1033	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
1034
1035	kfree_skb_reason(skb, reason);
1036	return 0;
1037
1038short_packet:
1039	if (reason == SKB_DROP_REASON_NOT_SPECIFIED)
1040		reason = SKB_DROP_REASON_PKT_TOO_SMALL;
1041	net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
1042			    proto == IPPROTO_UDPLITE ? "-Lite" : "",
1043			    saddr, ntohs(uh->source),
1044			    ulen, skb->len,
1045			    daddr, ntohs(uh->dest));
1046	goto discard;
1047
1048report_csum_error:
1049	udp6_csum_zero_error(skb);
1050csum_error:
1051	if (reason == SKB_DROP_REASON_NOT_SPECIFIED)
1052		reason = SKB_DROP_REASON_UDP_CSUM;
1053	__UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1054discard:
1055	__UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1056	kfree_skb_reason(skb, reason);
1057	return 0;
1058}
1059
1060
1061static struct sock *__udp6_lib_demux_lookup(struct net *net,
1062			__be16 loc_port, const struct in6_addr *loc_addr,
1063			__be16 rmt_port, const struct in6_addr *rmt_addr,
1064			int dif, int sdif)
1065{
1066	struct udp_table *udptable = net->ipv4.udp_table;
1067	unsigned short hnum = ntohs(loc_port);
1068	unsigned int hash2, slot2;
1069	struct udp_hslot *hslot2;
1070	__portpair ports;
1071	struct sock *sk;
1072
1073	hash2 = ipv6_portaddr_hash(net, loc_addr, hnum);
1074	slot2 = hash2 & udptable->mask;
1075	hslot2 = &udptable->hash2[slot2];
1076	ports = INET_COMBINED_PORTS(rmt_port, hnum);
1077
1078	udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
1079		if (sk->sk_state == TCP_ESTABLISHED &&
1080		    inet6_match(net, sk, rmt_addr, loc_addr, ports, dif, sdif))
1081			return sk;
1082		/* Only check first socket in chain */
1083		break;
1084	}
1085	return NULL;
1086}
1087
1088void udp_v6_early_demux(struct sk_buff *skb)
1089{
1090	struct net *net = dev_net(skb->dev);
1091	const struct udphdr *uh;
1092	struct sock *sk;
1093	struct dst_entry *dst;
1094	int dif = skb->dev->ifindex;
1095	int sdif = inet6_sdif(skb);
1096
1097	if (!pskb_may_pull(skb, skb_transport_offset(skb) +
1098	    sizeof(struct udphdr)))
1099		return;
1100
1101	uh = udp_hdr(skb);
1102
1103	if (skb->pkt_type == PACKET_HOST)
1104		sk = __udp6_lib_demux_lookup(net, uh->dest,
1105					     &ipv6_hdr(skb)->daddr,
1106					     uh->source, &ipv6_hdr(skb)->saddr,
1107					     dif, sdif);
1108	else
1109		return;
1110
1111	if (!sk)
1112		return;
1113
1114	skb->sk = sk;
1115	DEBUG_NET_WARN_ON_ONCE(sk_is_refcounted(sk));
1116	skb->destructor = sock_pfree;
1117	dst = rcu_dereference(sk->sk_rx_dst);
1118
1119	if (dst)
1120		dst = dst_check(dst, sk->sk_rx_dst_cookie);
1121	if (dst) {
1122		/* set noref for now.
1123		 * any place which wants to hold dst has to call
1124		 * dst_hold_safe()
1125		 */
1126		skb_dst_set_noref(skb, dst);
1127	}
1128}
1129
1130INDIRECT_CALLABLE_SCOPE int udpv6_rcv(struct sk_buff *skb)
1131{
1132	return __udp6_lib_rcv(skb, dev_net(skb->dev)->ipv4.udp_table, IPPROTO_UDP);
1133}
1134
1135/*
1136 * Throw away all pending data and cancel the corking. Socket is locked.
1137 */
1138static void udp_v6_flush_pending_frames(struct sock *sk)
1139{
1140	struct udp_sock *up = udp_sk(sk);
1141
1142	if (up->pending == AF_INET)
1143		udp_flush_pending_frames(sk);
1144	else if (up->pending) {
1145		up->len = 0;
1146		WRITE_ONCE(up->pending, 0);
1147		ip6_flush_pending_frames(sk);
1148	}
1149}
1150
1151static int udpv6_pre_connect(struct sock *sk, struct sockaddr *uaddr,
1152			     int addr_len)
1153{
1154	if (addr_len < offsetofend(struct sockaddr, sa_family))
1155		return -EINVAL;
1156	/* The following checks are replicated from __ip6_datagram_connect()
1157	 * and intended to prevent BPF program called below from accessing
1158	 * bytes that are out of the bound specified by user in addr_len.
1159	 */
1160	if (uaddr->sa_family == AF_INET) {
1161		if (ipv6_only_sock(sk))
1162			return -EAFNOSUPPORT;
1163		return udp_pre_connect(sk, uaddr, addr_len);
1164	}
1165
1166	if (addr_len < SIN6_LEN_RFC2133)
1167		return -EINVAL;
1168
1169	return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr, &addr_len);
1170}
1171
1172/**
1173 *	udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
1174 *	@sk:	socket we are sending on
1175 *	@skb:	sk_buff containing the filled-in UDP header
1176 *		(checksum field must be zeroed out)
1177 *	@saddr: source address
1178 *	@daddr: destination address
1179 *	@len:	length of packet
1180 */
1181static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
1182				 const struct in6_addr *saddr,
1183				 const struct in6_addr *daddr, int len)
1184{
1185	unsigned int offset;
1186	struct udphdr *uh = udp_hdr(skb);
1187	struct sk_buff *frags = skb_shinfo(skb)->frag_list;
1188	__wsum csum = 0;
1189
1190	if (!frags) {
1191		/* Only one fragment on the socket.  */
1192		skb->csum_start = skb_transport_header(skb) - skb->head;
1193		skb->csum_offset = offsetof(struct udphdr, check);
1194		uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
1195	} else {
1196		/*
1197		 * HW-checksum won't work as there are two or more
1198		 * fragments on the socket so that all csums of sk_buffs
1199		 * should be together
1200		 */
1201		offset = skb_transport_offset(skb);
1202		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
1203		csum = skb->csum;
1204
1205		skb->ip_summed = CHECKSUM_NONE;
1206
1207		do {
1208			csum = csum_add(csum, frags->csum);
1209		} while ((frags = frags->next));
1210
1211		uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1212					    csum);
1213		if (uh->check == 0)
1214			uh->check = CSUM_MANGLED_0;
1215	}
1216}
1217
1218/*
1219 *	Sending
1220 */
1221
1222static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6,
1223			   struct inet_cork *cork)
1224{
1225	struct sock *sk = skb->sk;
1226	struct udphdr *uh;
1227	int err = 0;
1228	int is_udplite = IS_UDPLITE(sk);
1229	__wsum csum = 0;
1230	int offset = skb_transport_offset(skb);
1231	int len = skb->len - offset;
1232	int datalen = len - sizeof(*uh);
1233
1234	/*
1235	 * Create a UDP header
1236	 */
1237	uh = udp_hdr(skb);
1238	uh->source = fl6->fl6_sport;
1239	uh->dest = fl6->fl6_dport;
1240	uh->len = htons(len);
1241	uh->check = 0;
1242
1243	if (cork->gso_size) {
1244		const int hlen = skb_network_header_len(skb) +
1245				 sizeof(struct udphdr);
1246
1247		if (hlen + cork->gso_size > cork->fragsize) {
1248			kfree_skb(skb);
1249			return -EINVAL;
1250		}
1251		if (datalen > cork->gso_size * UDP_MAX_SEGMENTS) {
1252			kfree_skb(skb);
1253			return -EINVAL;
1254		}
1255		if (udp_get_no_check6_tx(sk)) {
1256			kfree_skb(skb);
1257			return -EINVAL;
1258		}
1259		if (skb->ip_summed != CHECKSUM_PARTIAL || is_udplite ||
1260		    dst_xfrm(skb_dst(skb))) {
1261			kfree_skb(skb);
1262			return -EIO;
1263		}
1264
1265		if (datalen > cork->gso_size) {
1266			skb_shinfo(skb)->gso_size = cork->gso_size;
1267			skb_shinfo(skb)->gso_type = SKB_GSO_UDP_L4;
1268			skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(datalen,
1269								 cork->gso_size);
1270		}
1271		goto csum_partial;
1272	}
1273
1274	if (is_udplite)
1275		csum = udplite_csum(skb);
1276	else if (udp_get_no_check6_tx(sk)) {   /* UDP csum disabled */
1277		skb->ip_summed = CHECKSUM_NONE;
1278		goto send;
1279	} else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
1280csum_partial:
1281		udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
1282		goto send;
1283	} else
1284		csum = udp_csum(skb);
1285
1286	/* add protocol-dependent pseudo-header */
1287	uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1288				    len, fl6->flowi6_proto, csum);
1289	if (uh->check == 0)
1290		uh->check = CSUM_MANGLED_0;
1291
1292send:
1293	err = ip6_send_skb(skb);
1294	if (err) {
1295		if (err == -ENOBUFS && !inet6_test_bit(RECVERR6, sk)) {
1296			UDP6_INC_STATS(sock_net(sk),
1297				       UDP_MIB_SNDBUFERRORS, is_udplite);
1298			err = 0;
1299		}
1300	} else {
1301		UDP6_INC_STATS(sock_net(sk),
1302			       UDP_MIB_OUTDATAGRAMS, is_udplite);
1303	}
1304	return err;
1305}
1306
1307static int udp_v6_push_pending_frames(struct sock *sk)
1308{
1309	struct sk_buff *skb;
1310	struct udp_sock  *up = udp_sk(sk);
 
1311	int err = 0;
1312
1313	if (up->pending == AF_INET)
1314		return udp_push_pending_frames(sk);
1315
 
 
 
 
 
1316	skb = ip6_finish_skb(sk);
1317	if (!skb)
1318		goto out;
1319
1320	err = udp_v6_send_skb(skb, &inet_sk(sk)->cork.fl.u.ip6,
1321			      &inet_sk(sk)->cork.base);
1322out:
1323	up->len = 0;
1324	WRITE_ONCE(up->pending, 0);
1325	return err;
1326}
1327
1328int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1329{
1330	struct ipv6_txoptions opt_space;
1331	struct udp_sock *up = udp_sk(sk);
1332	struct inet_sock *inet = inet_sk(sk);
1333	struct ipv6_pinfo *np = inet6_sk(sk);
1334	DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1335	struct in6_addr *daddr, *final_p, final;
1336	struct ipv6_txoptions *opt = NULL;
1337	struct ipv6_txoptions *opt_to_free = NULL;
1338	struct ip6_flowlabel *flowlabel = NULL;
1339	struct inet_cork_full cork;
1340	struct flowi6 *fl6 = &cork.fl.u.ip6;
1341	struct dst_entry *dst;
1342	struct ipcm6_cookie ipc6;
1343	int addr_len = msg->msg_namelen;
1344	bool connected = false;
1345	int ulen = len;
1346	int corkreq = udp_test_bit(CORK, sk) || msg->msg_flags & MSG_MORE;
 
 
 
1347	int err;
 
1348	int is_udplite = IS_UDPLITE(sk);
1349	int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1350
1351	ipcm6_init(&ipc6);
1352	ipc6.gso_size = READ_ONCE(up->gso_size);
1353	ipc6.sockc.tsflags = READ_ONCE(sk->sk_tsflags);
1354	ipc6.sockc.mark = READ_ONCE(sk->sk_mark);
1355
1356	/* destination address check */
1357	if (sin6) {
1358		if (addr_len < offsetof(struct sockaddr, sa_data))
1359			return -EINVAL;
1360
1361		switch (sin6->sin6_family) {
1362		case AF_INET6:
1363			if (addr_len < SIN6_LEN_RFC2133)
1364				return -EINVAL;
1365			daddr = &sin6->sin6_addr;
1366			if (ipv6_addr_any(daddr) &&
1367			    ipv6_addr_v4mapped(&np->saddr))
1368				ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
1369						       daddr);
1370			break;
1371		case AF_INET:
1372			goto do_udp_sendmsg;
1373		case AF_UNSPEC:
1374			msg->msg_name = sin6 = NULL;
1375			msg->msg_namelen = addr_len = 0;
1376			daddr = NULL;
1377			break;
1378		default:
1379			return -EINVAL;
1380		}
1381	} else if (!READ_ONCE(up->pending)) {
1382		if (sk->sk_state != TCP_ESTABLISHED)
1383			return -EDESTADDRREQ;
1384		daddr = &sk->sk_v6_daddr;
1385	} else
1386		daddr = NULL;
1387
1388	if (daddr) {
1389		if (ipv6_addr_v4mapped(daddr)) {
1390			struct sockaddr_in sin;
1391			sin.sin_family = AF_INET;
1392			sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1393			sin.sin_addr.s_addr = daddr->s6_addr32[3];
1394			msg->msg_name = &sin;
1395			msg->msg_namelen = sizeof(sin);
1396do_udp_sendmsg:
1397			err = ipv6_only_sock(sk) ?
1398				-ENETUNREACH : udp_sendmsg(sk, msg, len);
1399			msg->msg_name = sin6;
1400			msg->msg_namelen = addr_len;
1401			return err;
1402		}
1403	}
1404
 
 
 
1405	/* Rough check on arithmetic overflow,
1406	   better check is made in ip6_append_data().
1407	   */
1408	if (len > INT_MAX - sizeof(struct udphdr))
1409		return -EMSGSIZE;
1410
1411	getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1412	if (READ_ONCE(up->pending)) {
1413		if (READ_ONCE(up->pending) == AF_INET)
1414			return udp_sendmsg(sk, msg, len);
1415		/*
1416		 * There are pending frames.
1417		 * The socket lock must be held while it's corked.
1418		 */
1419		lock_sock(sk);
1420		if (likely(up->pending)) {
1421			if (unlikely(up->pending != AF_INET6)) {
1422				release_sock(sk);
1423				return -EAFNOSUPPORT;
1424			}
1425			dst = NULL;
1426			goto do_append_data;
1427		}
1428		release_sock(sk);
1429	}
1430	ulen += sizeof(struct udphdr);
1431
1432	memset(fl6, 0, sizeof(*fl6));
1433
1434	if (sin6) {
1435		if (sin6->sin6_port == 0)
1436			return -EINVAL;
1437
1438		fl6->fl6_dport = sin6->sin6_port;
1439		daddr = &sin6->sin6_addr;
1440
1441		if (inet6_test_bit(SNDFLOW, sk)) {
1442			fl6->flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1443			if (fl6->flowlabel & IPV6_FLOWLABEL_MASK) {
1444				flowlabel = fl6_sock_lookup(sk, fl6->flowlabel);
1445				if (IS_ERR(flowlabel))
1446					return -EINVAL;
1447			}
1448		}
1449
1450		/*
1451		 * Otherwise it will be difficult to maintain
1452		 * sk->sk_dst_cache.
1453		 */
1454		if (sk->sk_state == TCP_ESTABLISHED &&
1455		    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1456			daddr = &sk->sk_v6_daddr;
1457
1458		if (addr_len >= sizeof(struct sockaddr_in6) &&
1459		    sin6->sin6_scope_id &&
1460		    __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1461			fl6->flowi6_oif = sin6->sin6_scope_id;
1462	} else {
1463		if (sk->sk_state != TCP_ESTABLISHED)
1464			return -EDESTADDRREQ;
1465
1466		fl6->fl6_dport = inet->inet_dport;
1467		daddr = &sk->sk_v6_daddr;
1468		fl6->flowlabel = np->flow_label;
1469		connected = true;
1470	}
1471
1472	if (!fl6->flowi6_oif)
1473		fl6->flowi6_oif = READ_ONCE(sk->sk_bound_dev_if);
1474
1475	if (!fl6->flowi6_oif)
1476		fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1477
1478	fl6->flowi6_uid = sk->sk_uid;
1479
1480	if (msg->msg_controllen) {
1481		opt = &opt_space;
1482		memset(opt, 0, sizeof(struct ipv6_txoptions));
1483		opt->tot_len = sizeof(*opt);
1484		ipc6.opt = opt;
1485
1486		err = udp_cmsg_send(sk, msg, &ipc6.gso_size);
1487		if (err > 0) {
1488			err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, fl6,
1489						    &ipc6);
1490			connected = false;
1491		}
1492		if (err < 0) {
1493			fl6_sock_release(flowlabel);
1494			return err;
1495		}
1496		if ((fl6->flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1497			flowlabel = fl6_sock_lookup(sk, fl6->flowlabel);
1498			if (IS_ERR(flowlabel))
1499				return -EINVAL;
1500		}
1501		if (!(opt->opt_nflen|opt->opt_flen))
1502			opt = NULL;
 
1503	}
1504	if (!opt) {
1505		opt = txopt_get(np);
1506		opt_to_free = opt;
1507	}
1508	if (flowlabel)
1509		opt = fl6_merge_options(&opt_space, flowlabel, opt);
1510	opt = ipv6_fixup_options(&opt_space, opt);
1511	ipc6.opt = opt;
1512
1513	fl6->flowi6_proto = sk->sk_protocol;
1514	fl6->flowi6_mark = ipc6.sockc.mark;
1515	fl6->daddr = *daddr;
1516	if (ipv6_addr_any(&fl6->saddr) && !ipv6_addr_any(&np->saddr))
1517		fl6->saddr = np->saddr;
1518	fl6->fl6_sport = inet->inet_sport;
1519
1520	if (cgroup_bpf_enabled(CGROUP_UDP6_SENDMSG) && !connected) {
1521		err = BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk,
1522					   (struct sockaddr *)sin6,
1523					   &addr_len,
1524					   &fl6->saddr);
1525		if (err)
1526			goto out_no_dst;
1527		if (sin6) {
1528			if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
1529				/* BPF program rewrote IPv6-only by IPv4-mapped
1530				 * IPv6. It's currently unsupported.
1531				 */
1532				err = -ENOTSUPP;
1533				goto out_no_dst;
1534			}
1535			if (sin6->sin6_port == 0) {
1536				/* BPF program set invalid port. Reject it. */
1537				err = -EINVAL;
1538				goto out_no_dst;
1539			}
1540			fl6->fl6_dport = sin6->sin6_port;
1541			fl6->daddr = sin6->sin6_addr;
1542		}
1543	}
1544
1545	if (ipv6_addr_any(&fl6->daddr))
1546		fl6->daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1547
1548	final_p = fl6_update_dst(fl6, opt, &final);
1549	if (final_p)
1550		connected = false;
1551
1552	if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr)) {
1553		fl6->flowi6_oif = READ_ONCE(np->mcast_oif);
1554		connected = false;
1555	} else if (!fl6->flowi6_oif)
1556		fl6->flowi6_oif = READ_ONCE(np->ucast_oif);
1557
1558	security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6));
1559
1560	if (ipc6.tclass < 0)
1561		ipc6.tclass = np->tclass;
 
 
 
1562
1563	fl6->flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6->flowlabel);
1564
1565	dst = ip6_sk_dst_lookup_flow(sk, fl6, final_p, connected);
1566	if (IS_ERR(dst)) {
1567		err = PTR_ERR(dst);
1568		dst = NULL;
1569		goto out;
1570	}
1571
1572	if (ipc6.hlimit < 0)
1573		ipc6.hlimit = ip6_sk_dst_hoplimit(np, fl6, dst);
 
 
 
1574
1575	if (msg->msg_flags&MSG_CONFIRM)
1576		goto do_confirm;
1577back_from_confirm:
1578
1579	/* Lockless fast path for the non-corking case */
1580	if (!corkreq) {
1581		struct sk_buff *skb;
1582
1583		skb = ip6_make_skb(sk, getfrag, msg, ulen,
1584				   sizeof(struct udphdr), &ipc6,
1585				   dst_rt6_info(dst),
1586				   msg->msg_flags, &cork);
1587		err = PTR_ERR(skb);
1588		if (!IS_ERR_OR_NULL(skb))
1589			err = udp_v6_send_skb(skb, fl6, &cork.base);
1590		/* ip6_make_skb steals dst reference */
1591		goto out_no_dst;
1592	}
1593
1594	lock_sock(sk);
1595	if (unlikely(up->pending)) {
1596		/* The socket is already corked while preparing it. */
1597		/* ... which is an evident application bug. --ANK */
1598		release_sock(sk);
1599
1600		net_dbg_ratelimited("udp cork app bug 2\n");
1601		err = -EINVAL;
1602		goto out;
1603	}
1604
1605	WRITE_ONCE(up->pending, AF_INET6);
1606
1607do_append_data:
1608	if (ipc6.dontfrag < 0)
1609		ipc6.dontfrag = inet6_test_bit(DONTFRAG, sk);
1610	up->len += ulen;
1611	err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr),
1612			      &ipc6, fl6, dst_rt6_info(dst),
1613			      corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
 
1614	if (err)
1615		udp_v6_flush_pending_frames(sk);
1616	else if (!corkreq)
1617		err = udp_v6_push_pending_frames(sk);
1618	else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1619		WRITE_ONCE(up->pending, 0);
1620
1621	if (err > 0)
1622		err = inet6_test_bit(RECVERR6, sk) ? net_xmit_errno(err) : 0;
1623	release_sock(sk);
1624
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1625out:
1626	dst_release(dst);
1627out_no_dst:
1628	fl6_sock_release(flowlabel);
1629	txopt_put(opt_to_free);
1630	if (!err)
1631		return len;
1632	/*
1633	 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1634	 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1635	 * we don't have a good statistic (IpOutDiscards but it can be too many
1636	 * things).  We could add another new stat but at least for now that
1637	 * seems like overkill.
1638	 */
1639	if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1640		UDP6_INC_STATS(sock_net(sk),
1641			       UDP_MIB_SNDBUFERRORS, is_udplite);
1642	}
1643	return err;
1644
1645do_confirm:
1646	if (msg->msg_flags & MSG_PROBE)
1647		dst_confirm_neigh(dst, &fl6->daddr);
1648	if (!(msg->msg_flags&MSG_PROBE) || len)
1649		goto back_from_confirm;
1650	err = 0;
1651	goto out;
1652}
1653EXPORT_SYMBOL(udpv6_sendmsg);
1654
1655static void udpv6_splice_eof(struct socket *sock)
1656{
1657	struct sock *sk = sock->sk;
1658	struct udp_sock *up = udp_sk(sk);
1659
1660	if (!READ_ONCE(up->pending) || udp_test_bit(CORK, sk))
1661		return;
1662
1663	lock_sock(sk);
1664	if (up->pending && !udp_test_bit(CORK, sk))
1665		udp_v6_push_pending_frames(sk);
1666	release_sock(sk);
1667}
1668
1669void udpv6_destroy_sock(struct sock *sk)
1670{
1671	struct udp_sock *up = udp_sk(sk);
1672	lock_sock(sk);
1673
1674	/* protects from races with udp_abort() */
1675	sock_set_flag(sk, SOCK_DEAD);
1676	udp_v6_flush_pending_frames(sk);
1677	release_sock(sk);
1678
1679	if (static_branch_unlikely(&udpv6_encap_needed_key)) {
1680		if (up->encap_type) {
1681			void (*encap_destroy)(struct sock *sk);
1682			encap_destroy = READ_ONCE(up->encap_destroy);
1683			if (encap_destroy)
1684				encap_destroy(sk);
1685		}
1686		if (udp_test_bit(ENCAP_ENABLED, sk)) {
1687			static_branch_dec(&udpv6_encap_needed_key);
1688			udp_encap_disable();
1689		}
1690	}
 
 
1691}
1692
1693/*
1694 *	Socket option code for UDP
1695 */
1696int udpv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1697		     unsigned int optlen)
1698{
1699	if (level == SOL_UDP  ||  level == SOL_UDPLITE || level == SOL_SOCKET)
1700		return udp_lib_setsockopt(sk, level, optname,
1701					  optval, optlen,
1702					  udp_v6_push_pending_frames);
1703	return ipv6_setsockopt(sk, level, optname, optval, optlen);
1704}
1705
 
 
 
 
 
 
 
 
 
 
 
1706int udpv6_getsockopt(struct sock *sk, int level, int optname,
1707		     char __user *optval, int __user *optlen)
1708{
1709	if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1710		return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1711	return ipv6_getsockopt(sk, level, optname, optval, optlen);
1712}
1713
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1714
1715/* ------------------------------------------------------------------------ */
1716#ifdef CONFIG_PROC_FS
1717int udp6_seq_show(struct seq_file *seq, void *v)
1718{
1719	if (v == SEQ_START_TOKEN) {
1720		seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1721	} else {
1722		int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1723		const struct inet_sock *inet = inet_sk((const struct sock *)v);
1724		__u16 srcp = ntohs(inet->inet_sport);
1725		__u16 destp = ntohs(inet->inet_dport);
1726		__ip6_dgram_sock_seq_show(seq, v, srcp, destp,
1727					  udp_rqueue_get(v), bucket);
1728	}
1729	return 0;
1730}
1731
1732const struct seq_operations udp6_seq_ops = {
1733	.start		= udp_seq_start,
1734	.next		= udp_seq_next,
1735	.stop		= udp_seq_stop,
1736	.show		= udp6_seq_show,
 
1737};
1738EXPORT_SYMBOL(udp6_seq_ops);
1739
1740static struct udp_seq_afinfo udp6_seq_afinfo = {
 
1741	.family		= AF_INET6,
1742	.udp_table	= NULL,
 
 
 
 
1743};
1744
1745int __net_init udp6_proc_init(struct net *net)
1746{
1747	if (!proc_create_net_data("udp6", 0444, net->proc_net, &udp6_seq_ops,
1748			sizeof(struct udp_iter_state), &udp6_seq_afinfo))
1749		return -ENOMEM;
1750	return 0;
1751}
1752
1753void udp6_proc_exit(struct net *net)
1754{
1755	remove_proc_entry("udp6", net->proc_net);
1756}
1757#endif /* CONFIG_PROC_FS */
1758
 
 
 
 
 
 
 
 
 
 
 
1759/* ------------------------------------------------------------------------ */
1760
1761struct proto udpv6_prot = {
1762	.name			= "UDPv6",
1763	.owner			= THIS_MODULE,
1764	.close			= udp_lib_close,
1765	.pre_connect		= udpv6_pre_connect,
1766	.connect		= ip6_datagram_connect,
1767	.disconnect		= udp_disconnect,
1768	.ioctl			= udp_ioctl,
1769	.init			= udpv6_init_sock,
1770	.destroy		= udpv6_destroy_sock,
1771	.setsockopt		= udpv6_setsockopt,
1772	.getsockopt		= udpv6_getsockopt,
1773	.sendmsg		= udpv6_sendmsg,
1774	.recvmsg		= udpv6_recvmsg,
1775	.splice_eof		= udpv6_splice_eof,
1776	.release_cb		= ip6_datagram_release_cb,
1777	.hash			= udp_lib_hash,
1778	.unhash			= udp_lib_unhash,
1779	.rehash			= udp_v6_rehash,
1780	.get_port		= udp_v6_get_port,
1781	.put_port		= udp_lib_unhash,
1782#ifdef CONFIG_BPF_SYSCALL
1783	.psock_update_sk_prot	= udp_bpf_update_proto,
 
 
 
 
 
1784#endif
1785
1786	.memory_allocated	= &udp_memory_allocated,
1787	.per_cpu_fw_alloc	= &udp_memory_per_cpu_fw_alloc,
1788
1789	.sysctl_mem		= sysctl_udp_mem,
1790	.sysctl_wmem_offset     = offsetof(struct net, ipv4.sysctl_udp_wmem_min),
1791	.sysctl_rmem_offset     = offsetof(struct net, ipv4.sysctl_udp_rmem_min),
1792	.obj_size		= sizeof(struct udp6_sock),
1793	.ipv6_pinfo_offset = offsetof(struct udp6_sock, inet6),
1794	.h.udp_table		= NULL,
1795	.diag_destroy		= udp_abort,
1796};
1797
1798static struct inet_protosw udpv6_protosw = {
1799	.type =      SOCK_DGRAM,
1800	.protocol =  IPPROTO_UDP,
1801	.prot =      &udpv6_prot,
1802	.ops =       &inet6_dgram_ops,
1803	.flags =     INET_PROTOSW_PERMANENT,
1804};
1805
1806int __init udpv6_init(void)
1807{
1808	int ret;
1809
1810	net_hotdata.udpv6_protocol = (struct inet6_protocol) {
1811		.handler     = udpv6_rcv,
1812		.err_handler = udpv6_err,
1813		.flags	     = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1814	};
1815	ret = inet6_add_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP);
1816	if (ret)
1817		goto out;
1818
1819	ret = inet6_register_protosw(&udpv6_protosw);
1820	if (ret)
1821		goto out_udpv6_protocol;
1822out:
1823	return ret;
1824
1825out_udpv6_protocol:
1826	inet6_del_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP);
1827	goto out;
1828}
1829
1830void udpv6_exit(void)
1831{
1832	inet6_unregister_protosw(&udpv6_protosw);
1833	inet6_del_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP);
1834}