Linux Audio

Check our new training course

Loading...
v3.5.6
   1/* SCTP kernel implementation
   2 * (C) Copyright IBM Corp. 2001, 2004
   3 * Copyright (c) 1999-2000 Cisco, Inc.
   4 * Copyright (c) 1999-2001 Motorola, Inc.
   5 * Copyright (c) 2001-2003 Intel Corp.
   6 * Copyright (c) 2001-2002 Nokia, Inc.
   7 * Copyright (c) 2001 La Monte H.P. Yarroll
   8 *
   9 * This file is part of the SCTP kernel implementation
  10 *
  11 * These functions interface with the sockets layer to implement the
  12 * SCTP Extensions for the Sockets API.
  13 *
  14 * Note that the descriptions from the specification are USER level
  15 * functions--this file is the functions which populate the struct proto
  16 * for SCTP which is the BOTTOM of the sockets interface.
  17 *
  18 * This SCTP implementation is free software;
  19 * you can redistribute it and/or modify it under the terms of
  20 * the GNU General Public License as published by
  21 * the Free Software Foundation; either version 2, or (at your option)
  22 * any later version.
  23 *
  24 * This SCTP implementation is distributed in the hope that it
  25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
  26 *                 ************************
  27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  28 * See the GNU General Public License for more details.
  29 *
  30 * You should have received a copy of the GNU General Public License
  31 * along with GNU CC; see the file COPYING.  If not, write to
  32 * the Free Software Foundation, 59 Temple Place - Suite 330,
  33 * Boston, MA 02111-1307, USA.
  34 *
  35 * Please send any bug reports or fixes you make to the
  36 * email address(es):
  37 *    lksctp developers <lksctp-developers@lists.sourceforge.net>
  38 *
  39 * Or submit a bug report through the following website:
  40 *    http://www.sf.net/projects/lksctp
  41 *
  42 * Written or modified by:
  43 *    La Monte H.P. Yarroll <piggy@acm.org>
  44 *    Narasimha Budihal     <narsi@refcode.org>
  45 *    Karl Knutson          <karl@athena.chicago.il.us>
  46 *    Jon Grimm             <jgrimm@us.ibm.com>
  47 *    Xingang Guo           <xingang.guo@intel.com>
  48 *    Daisy Chang           <daisyc@us.ibm.com>
  49 *    Sridhar Samudrala     <samudrala@us.ibm.com>
  50 *    Inaky Perez-Gonzalez  <inaky.gonzalez@intel.com>
  51 *    Ardelle Fan	    <ardelle.fan@intel.com>
  52 *    Ryan Layer	    <rmlayer@us.ibm.com>
  53 *    Anup Pemmaiah         <pemmaiah@cc.usu.edu>
  54 *    Kevin Gao             <kevin.gao@intel.com>
  55 *
  56 * Any bugs reported given to us we will try to fix... any fixes shared will
  57 * be incorporated into the next SCTP release.
  58 */
  59
  60#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  61
  62#include <linux/types.h>
  63#include <linux/kernel.h>
  64#include <linux/wait.h>
  65#include <linux/time.h>
  66#include <linux/ip.h>
  67#include <linux/capability.h>
  68#include <linux/fcntl.h>
  69#include <linux/poll.h>
  70#include <linux/init.h>
  71#include <linux/crypto.h>
  72#include <linux/slab.h>
 
 
  73
  74#include <net/ip.h>
  75#include <net/icmp.h>
  76#include <net/route.h>
  77#include <net/ipv6.h>
  78#include <net/inet_common.h>
  79
  80#include <linux/socket.h> /* for sa_family_t */
  81#include <linux/export.h>
  82#include <net/sock.h>
  83#include <net/sctp/sctp.h>
  84#include <net/sctp/sm.h>
  85
  86/* WARNING:  Please do not remove the SCTP_STATIC attribute to
  87 * any of the functions below as they are used to export functions
  88 * used by a project regression testsuite.
  89 */
  90
  91/* Forward declarations for internal helper functions. */
  92static int sctp_writeable(struct sock *sk);
  93static void sctp_wfree(struct sk_buff *skb);
  94static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
  95				size_t msg_len);
  96static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
  97static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
  98static int sctp_wait_for_accept(struct sock *sk, long timeo);
  99static void sctp_wait_for_close(struct sock *sk, long timeo);
 
 100static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
 101					union sctp_addr *addr, int len);
 102static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
 103static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
 104static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
 105static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
 106static int sctp_send_asconf(struct sctp_association *asoc,
 107			    struct sctp_chunk *chunk);
 108static int sctp_do_bind(struct sock *, union sctp_addr *, int);
 109static int sctp_autobind(struct sock *sk);
 110static void sctp_sock_migrate(struct sock *, struct sock *,
 111			      struct sctp_association *, sctp_socket_type_t);
 112static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
 113
 114extern struct kmem_cache *sctp_bucket_cachep;
 115extern long sysctl_sctp_mem[3];
 116extern int sysctl_sctp_rmem[3];
 117extern int sysctl_sctp_wmem[3];
 118
 119static int sctp_memory_pressure;
 120static atomic_long_t sctp_memory_allocated;
 121struct percpu_counter sctp_sockets_allocated;
 122
 123static void sctp_enter_memory_pressure(struct sock *sk)
 124{
 125	sctp_memory_pressure = 1;
 126}
 127
 128
 129/* Get the sndbuf space available at the time on the association.  */
 130static inline int sctp_wspace(struct sctp_association *asoc)
 131{
 132	int amt;
 133
 134	if (asoc->ep->sndbuf_policy)
 135		amt = asoc->sndbuf_used;
 136	else
 137		amt = sk_wmem_alloc_get(asoc->base.sk);
 138
 139	if (amt >= asoc->base.sk->sk_sndbuf) {
 140		if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
 141			amt = 0;
 142		else {
 143			amt = sk_stream_wspace(asoc->base.sk);
 144			if (amt < 0)
 145				amt = 0;
 146		}
 147	} else {
 148		amt = asoc->base.sk->sk_sndbuf - amt;
 149	}
 150	return amt;
 151}
 152
 153/* Increment the used sndbuf space count of the corresponding association by
 154 * the size of the outgoing data chunk.
 155 * Also, set the skb destructor for sndbuf accounting later.
 156 *
 157 * Since it is always 1-1 between chunk and skb, and also a new skb is always
 158 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
 159 * destructor in the data chunk skb for the purpose of the sndbuf space
 160 * tracking.
 161 */
 162static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
 163{
 164	struct sctp_association *asoc = chunk->asoc;
 165	struct sock *sk = asoc->base.sk;
 166
 167	/* The sndbuf space is tracked per association.  */
 168	sctp_association_hold(asoc);
 169
 170	skb_set_owner_w(chunk->skb, sk);
 171
 172	chunk->skb->destructor = sctp_wfree;
 173	/* Save the chunk pointer in skb for sctp_wfree to use later.  */
 174	*((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
 175
 176	asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
 177				sizeof(struct sk_buff) +
 178				sizeof(struct sctp_chunk);
 179
 180	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
 181	sk->sk_wmem_queued += chunk->skb->truesize;
 182	sk_mem_charge(sk, chunk->skb->truesize);
 183}
 184
 185/* Verify that this is a valid address. */
 186static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
 187				   int len)
 188{
 189	struct sctp_af *af;
 190
 191	/* Verify basic sockaddr. */
 192	af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
 193	if (!af)
 194		return -EINVAL;
 195
 196	/* Is this a valid SCTP address?  */
 197	if (!af->addr_valid(addr, sctp_sk(sk), NULL))
 198		return -EINVAL;
 199
 200	if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
 201		return -EINVAL;
 202
 203	return 0;
 204}
 205
 206/* Look up the association by its id.  If this is not a UDP-style
 207 * socket, the ID field is always ignored.
 208 */
 209struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
 210{
 211	struct sctp_association *asoc = NULL;
 212
 213	/* If this is not a UDP-style socket, assoc id should be ignored. */
 214	if (!sctp_style(sk, UDP)) {
 215		/* Return NULL if the socket state is not ESTABLISHED. It
 216		 * could be a TCP-style listening socket or a socket which
 217		 * hasn't yet called connect() to establish an association.
 218		 */
 219		if (!sctp_sstate(sk, ESTABLISHED))
 220			return NULL;
 221
 222		/* Get the first and the only association from the list. */
 223		if (!list_empty(&sctp_sk(sk)->ep->asocs))
 224			asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
 225					  struct sctp_association, asocs);
 226		return asoc;
 227	}
 228
 229	/* Otherwise this is a UDP-style socket. */
 230	if (!id || (id == (sctp_assoc_t)-1))
 231		return NULL;
 232
 233	spin_lock_bh(&sctp_assocs_id_lock);
 234	asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
 235	spin_unlock_bh(&sctp_assocs_id_lock);
 236
 237	if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
 238		return NULL;
 239
 240	return asoc;
 241}
 242
 243/* Look up the transport from an address and an assoc id. If both address and
 244 * id are specified, the associations matching the address and the id should be
 245 * the same.
 246 */
 247static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
 248					      struct sockaddr_storage *addr,
 249					      sctp_assoc_t id)
 250{
 251	struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
 252	struct sctp_transport *transport;
 253	union sctp_addr *laddr = (union sctp_addr *)addr;
 254
 255	addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
 256					       laddr,
 257					       &transport);
 258
 259	if (!addr_asoc)
 260		return NULL;
 261
 262	id_asoc = sctp_id2assoc(sk, id);
 263	if (id_asoc && (id_asoc != addr_asoc))
 264		return NULL;
 265
 266	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
 267						(union sctp_addr *)addr);
 268
 269	return transport;
 270}
 271
 272/* API 3.1.2 bind() - UDP Style Syntax
 273 * The syntax of bind() is,
 274 *
 275 *   ret = bind(int sd, struct sockaddr *addr, int addrlen);
 276 *
 277 *   sd      - the socket descriptor returned by socket().
 278 *   addr    - the address structure (struct sockaddr_in or struct
 279 *             sockaddr_in6 [RFC 2553]),
 280 *   addr_len - the size of the address structure.
 281 */
 282SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
 283{
 284	int retval = 0;
 285
 286	sctp_lock_sock(sk);
 287
 288	SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
 289			  sk, addr, addr_len);
 290
 291	/* Disallow binding twice. */
 292	if (!sctp_sk(sk)->ep->base.bind_addr.port)
 293		retval = sctp_do_bind(sk, (union sctp_addr *)addr,
 294				      addr_len);
 295	else
 296		retval = -EINVAL;
 297
 298	sctp_release_sock(sk);
 299
 300	return retval;
 301}
 302
 303static long sctp_get_port_local(struct sock *, union sctp_addr *);
 304
 305/* Verify this is a valid sockaddr. */
 306static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
 307					union sctp_addr *addr, int len)
 308{
 309	struct sctp_af *af;
 310
 311	/* Check minimum size.  */
 312	if (len < sizeof (struct sockaddr))
 313		return NULL;
 314
 315	/* V4 mapped address are really of AF_INET family */
 316	if (addr->sa.sa_family == AF_INET6 &&
 317	    ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
 318		if (!opt->pf->af_supported(AF_INET, opt))
 319			return NULL;
 320	} else {
 321		/* Does this PF support this AF? */
 322		if (!opt->pf->af_supported(addr->sa.sa_family, opt))
 323			return NULL;
 324	}
 325
 326	/* If we get this far, af is valid. */
 327	af = sctp_get_af_specific(addr->sa.sa_family);
 328
 329	if (len < af->sockaddr_len)
 330		return NULL;
 331
 332	return af;
 333}
 334
 335/* Bind a local address either to an endpoint or to an association.  */
 336SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
 337{
 
 338	struct sctp_sock *sp = sctp_sk(sk);
 339	struct sctp_endpoint *ep = sp->ep;
 340	struct sctp_bind_addr *bp = &ep->base.bind_addr;
 341	struct sctp_af *af;
 342	unsigned short snum;
 343	int ret = 0;
 344
 345	/* Common sockaddr verification. */
 346	af = sctp_sockaddr_af(sp, addr, len);
 347	if (!af) {
 348		SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
 349				  sk, addr, len);
 350		return -EINVAL;
 351	}
 352
 353	snum = ntohs(addr->v4.sin_port);
 354
 355	SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
 356				 ", port: %d, new port: %d, len: %d)\n",
 357				 sk,
 358				 addr,
 359				 bp->port, snum,
 360				 len);
 361
 362	/* PF specific bind() address verification. */
 363	if (!sp->pf->bind_verify(sp, addr))
 364		return -EADDRNOTAVAIL;
 365
 366	/* We must either be unbound, or bind to the same port.
 367	 * It's OK to allow 0 ports if we are already bound.
 368	 * We'll just inhert an already bound port in this case
 369	 */
 370	if (bp->port) {
 371		if (!snum)
 372			snum = bp->port;
 373		else if (snum != bp->port) {
 374			SCTP_DEBUG_PRINTK("sctp_do_bind:"
 375				  " New port %d does not match existing port "
 376				  "%d.\n", snum, bp->port);
 377			return -EINVAL;
 378		}
 379	}
 380
 381	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 
 382		return -EACCES;
 383
 384	/* See if the address matches any of the addresses we may have
 385	 * already bound before checking against other endpoints.
 386	 */
 387	if (sctp_bind_addr_match(bp, addr, sp))
 388		return -EINVAL;
 389
 390	/* Make sure we are allowed to bind here.
 391	 * The function sctp_get_port_local() does duplicate address
 392	 * detection.
 393	 */
 394	addr->v4.sin_port = htons(snum);
 395	if ((ret = sctp_get_port_local(sk, addr))) {
 396		return -EADDRINUSE;
 397	}
 398
 399	/* Refresh ephemeral port.  */
 400	if (!bp->port)
 401		bp->port = inet_sk(sk)->inet_num;
 402
 403	/* Add the address to the bind address list.
 404	 * Use GFP_ATOMIC since BHs will be disabled.
 405	 */
 406	ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
 407
 408	/* Copy back into socket for getsockname() use. */
 409	if (!ret) {
 410		inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
 411		af->to_sk_saddr(addr, sk);
 412	}
 413
 414	return ret;
 415}
 416
 417 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
 418 *
 419 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
 420 * at any one time.  If a sender, after sending an ASCONF chunk, decides
 421 * it needs to transfer another ASCONF Chunk, it MUST wait until the
 422 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
 423 * subsequent ASCONF. Note this restriction binds each side, so at any
 424 * time two ASCONF may be in-transit on any given association (one sent
 425 * from each endpoint).
 426 */
 427static int sctp_send_asconf(struct sctp_association *asoc,
 428			    struct sctp_chunk *chunk)
 429{
 
 430	int		retval = 0;
 431
 432	/* If there is an outstanding ASCONF chunk, queue it for later
 433	 * transmission.
 434	 */
 435	if (asoc->addip_last_asconf) {
 436		list_add_tail(&chunk->list, &asoc->addip_chunk_list);
 437		goto out;
 438	}
 439
 440	/* Hold the chunk until an ASCONF_ACK is received. */
 441	sctp_chunk_hold(chunk);
 442	retval = sctp_primitive_ASCONF(asoc, chunk);
 443	if (retval)
 444		sctp_chunk_free(chunk);
 445	else
 446		asoc->addip_last_asconf = chunk;
 447
 448out:
 449	return retval;
 450}
 451
 452/* Add a list of addresses as bind addresses to local endpoint or
 453 * association.
 454 *
 455 * Basically run through each address specified in the addrs/addrcnt
 456 * array/length pair, determine if it is IPv6 or IPv4 and call
 457 * sctp_do_bind() on it.
 458 *
 459 * If any of them fails, then the operation will be reversed and the
 460 * ones that were added will be removed.
 461 *
 462 * Only sctp_setsockopt_bindx() is supposed to call this function.
 463 */
 464static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
 465{
 466	int cnt;
 467	int retval = 0;
 468	void *addr_buf;
 469	struct sockaddr *sa_addr;
 470	struct sctp_af *af;
 471
 472	SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
 473			  sk, addrs, addrcnt);
 474
 475	addr_buf = addrs;
 476	for (cnt = 0; cnt < addrcnt; cnt++) {
 477		/* The list may contain either IPv4 or IPv6 address;
 478		 * determine the address length for walking thru the list.
 479		 */
 480		sa_addr = addr_buf;
 481		af = sctp_get_af_specific(sa_addr->sa_family);
 482		if (!af) {
 483			retval = -EINVAL;
 484			goto err_bindx_add;
 485		}
 486
 487		retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
 488				      af->sockaddr_len);
 489
 490		addr_buf += af->sockaddr_len;
 491
 492err_bindx_add:
 493		if (retval < 0) {
 494			/* Failed. Cleanup the ones that have been added */
 495			if (cnt > 0)
 496				sctp_bindx_rem(sk, addrs, cnt);
 497			return retval;
 498		}
 499	}
 500
 501	return retval;
 502}
 503
 504/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
 505 * associations that are part of the endpoint indicating that a list of local
 506 * addresses are added to the endpoint.
 507 *
 508 * If any of the addresses is already in the bind address list of the
 509 * association, we do not send the chunk for that association.  But it will not
 510 * affect other associations.
 511 *
 512 * Only sctp_setsockopt_bindx() is supposed to call this function.
 513 */
 514static int sctp_send_asconf_add_ip(struct sock		*sk,
 515				   struct sockaddr	*addrs,
 516				   int 			addrcnt)
 517{
 
 518	struct sctp_sock		*sp;
 519	struct sctp_endpoint		*ep;
 520	struct sctp_association		*asoc;
 521	struct sctp_bind_addr		*bp;
 522	struct sctp_chunk		*chunk;
 523	struct sctp_sockaddr_entry	*laddr;
 524	union sctp_addr			*addr;
 525	union sctp_addr			saveaddr;
 526	void				*addr_buf;
 527	struct sctp_af			*af;
 528	struct list_head		*p;
 529	int 				i;
 530	int 				retval = 0;
 531
 532	if (!sctp_addip_enable)
 533		return retval;
 534
 535	sp = sctp_sk(sk);
 536	ep = sp->ep;
 537
 538	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
 539			  __func__, sk, addrs, addrcnt);
 540
 541	list_for_each_entry(asoc, &ep->asocs, asocs) {
 542
 543		if (!asoc->peer.asconf_capable)
 544			continue;
 545
 546		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
 547			continue;
 548
 549		if (!sctp_state(asoc, ESTABLISHED))
 550			continue;
 551
 552		/* Check if any address in the packed array of addresses is
 553		 * in the bind address list of the association. If so,
 554		 * do not send the asconf chunk to its peer, but continue with
 555		 * other associations.
 556		 */
 557		addr_buf = addrs;
 558		for (i = 0; i < addrcnt; i++) {
 559			addr = addr_buf;
 560			af = sctp_get_af_specific(addr->v4.sin_family);
 561			if (!af) {
 562				retval = -EINVAL;
 563				goto out;
 564			}
 565
 566			if (sctp_assoc_lookup_laddr(asoc, addr))
 567				break;
 568
 569			addr_buf += af->sockaddr_len;
 570		}
 571		if (i < addrcnt)
 572			continue;
 573
 574		/* Use the first valid address in bind addr list of
 575		 * association as Address Parameter of ASCONF CHUNK.
 576		 */
 577		bp = &asoc->base.bind_addr;
 578		p = bp->address_list.next;
 579		laddr = list_entry(p, struct sctp_sockaddr_entry, list);
 580		chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
 581						   addrcnt, SCTP_PARAM_ADD_IP);
 582		if (!chunk) {
 583			retval = -ENOMEM;
 584			goto out;
 585		}
 586
 587		/* Add the new addresses to the bind address list with
 588		 * use_as_src set to 0.
 589		 */
 590		addr_buf = addrs;
 591		for (i = 0; i < addrcnt; i++) {
 592			addr = addr_buf;
 593			af = sctp_get_af_specific(addr->v4.sin_family);
 594			memcpy(&saveaddr, addr, af->sockaddr_len);
 595			retval = sctp_add_bind_addr(bp, &saveaddr,
 596						    SCTP_ADDR_NEW, GFP_ATOMIC);
 597			addr_buf += af->sockaddr_len;
 598		}
 599		if (asoc->src_out_of_asoc_ok) {
 600			struct sctp_transport *trans;
 601
 602			list_for_each_entry(trans,
 603			    &asoc->peer.transport_addr_list, transports) {
 604				/* Clear the source and route cache */
 605				dst_release(trans->dst);
 606				trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
 607				    2*asoc->pathmtu, 4380));
 608				trans->ssthresh = asoc->peer.i.a_rwnd;
 609				trans->rto = asoc->rto_initial;
 
 610				trans->rtt = trans->srtt = trans->rttvar = 0;
 611				sctp_transport_route(trans, NULL,
 612				    sctp_sk(asoc->base.sk));
 613			}
 614		}
 615		retval = sctp_send_asconf(asoc, chunk);
 616	}
 617
 618out:
 619	return retval;
 620}
 621
 622/* Remove a list of addresses from bind addresses list.  Do not remove the
 623 * last address.
 624 *
 625 * Basically run through each address specified in the addrs/addrcnt
 626 * array/length pair, determine if it is IPv6 or IPv4 and call
 627 * sctp_del_bind() on it.
 628 *
 629 * If any of them fails, then the operation will be reversed and the
 630 * ones that were removed will be added back.
 631 *
 632 * At least one address has to be left; if only one address is
 633 * available, the operation will return -EBUSY.
 634 *
 635 * Only sctp_setsockopt_bindx() is supposed to call this function.
 636 */
 637static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
 638{
 639	struct sctp_sock *sp = sctp_sk(sk);
 640	struct sctp_endpoint *ep = sp->ep;
 641	int cnt;
 642	struct sctp_bind_addr *bp = &ep->base.bind_addr;
 643	int retval = 0;
 644	void *addr_buf;
 645	union sctp_addr *sa_addr;
 646	struct sctp_af *af;
 647
 648	SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
 649			  sk, addrs, addrcnt);
 650
 651	addr_buf = addrs;
 652	for (cnt = 0; cnt < addrcnt; cnt++) {
 653		/* If the bind address list is empty or if there is only one
 654		 * bind address, there is nothing more to be removed (we need
 655		 * at least one address here).
 656		 */
 657		if (list_empty(&bp->address_list) ||
 658		    (sctp_list_single_entry(&bp->address_list))) {
 659			retval = -EBUSY;
 660			goto err_bindx_rem;
 661		}
 662
 663		sa_addr = addr_buf;
 664		af = sctp_get_af_specific(sa_addr->sa.sa_family);
 665		if (!af) {
 666			retval = -EINVAL;
 667			goto err_bindx_rem;
 668		}
 669
 670		if (!af->addr_valid(sa_addr, sp, NULL)) {
 671			retval = -EADDRNOTAVAIL;
 672			goto err_bindx_rem;
 673		}
 674
 675		if (sa_addr->v4.sin_port &&
 676		    sa_addr->v4.sin_port != htons(bp->port)) {
 677			retval = -EINVAL;
 678			goto err_bindx_rem;
 679		}
 680
 681		if (!sa_addr->v4.sin_port)
 682			sa_addr->v4.sin_port = htons(bp->port);
 683
 684		/* FIXME - There is probably a need to check if sk->sk_saddr and
 685		 * sk->sk_rcv_addr are currently set to one of the addresses to
 686		 * be removed. This is something which needs to be looked into
 687		 * when we are fixing the outstanding issues with multi-homing
 688		 * socket routing and failover schemes. Refer to comments in
 689		 * sctp_do_bind(). -daisy
 690		 */
 691		retval = sctp_del_bind_addr(bp, sa_addr);
 692
 693		addr_buf += af->sockaddr_len;
 694err_bindx_rem:
 695		if (retval < 0) {
 696			/* Failed. Add the ones that has been removed back */
 697			if (cnt > 0)
 698				sctp_bindx_add(sk, addrs, cnt);
 699			return retval;
 700		}
 701	}
 702
 703	return retval;
 704}
 705
 706/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
 707 * the associations that are part of the endpoint indicating that a list of
 708 * local addresses are removed from the endpoint.
 709 *
 710 * If any of the addresses is already in the bind address list of the
 711 * association, we do not send the chunk for that association.  But it will not
 712 * affect other associations.
 713 *
 714 * Only sctp_setsockopt_bindx() is supposed to call this function.
 715 */
 716static int sctp_send_asconf_del_ip(struct sock		*sk,
 717				   struct sockaddr	*addrs,
 718				   int			addrcnt)
 719{
 
 720	struct sctp_sock	*sp;
 721	struct sctp_endpoint	*ep;
 722	struct sctp_association	*asoc;
 723	struct sctp_transport	*transport;
 724	struct sctp_bind_addr	*bp;
 725	struct sctp_chunk	*chunk;
 726	union sctp_addr		*laddr;
 727	void			*addr_buf;
 728	struct sctp_af		*af;
 729	struct sctp_sockaddr_entry *saddr;
 730	int 			i;
 731	int 			retval = 0;
 732	int			stored = 0;
 733
 734	chunk = NULL;
 735	if (!sctp_addip_enable)
 736		return retval;
 737
 738	sp = sctp_sk(sk);
 739	ep = sp->ep;
 740
 741	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
 742			  __func__, sk, addrs, addrcnt);
 743
 744	list_for_each_entry(asoc, &ep->asocs, asocs) {
 745
 746		if (!asoc->peer.asconf_capable)
 747			continue;
 748
 749		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
 750			continue;
 751
 752		if (!sctp_state(asoc, ESTABLISHED))
 753			continue;
 754
 755		/* Check if any address in the packed array of addresses is
 756		 * not present in the bind address list of the association.
 757		 * If so, do not send the asconf chunk to its peer, but
 758		 * continue with other associations.
 759		 */
 760		addr_buf = addrs;
 761		for (i = 0; i < addrcnt; i++) {
 762			laddr = addr_buf;
 763			af = sctp_get_af_specific(laddr->v4.sin_family);
 764			if (!af) {
 765				retval = -EINVAL;
 766				goto out;
 767			}
 768
 769			if (!sctp_assoc_lookup_laddr(asoc, laddr))
 770				break;
 771
 772			addr_buf += af->sockaddr_len;
 773		}
 774		if (i < addrcnt)
 775			continue;
 776
 777		/* Find one address in the association's bind address list
 778		 * that is not in the packed array of addresses. This is to
 779		 * make sure that we do not delete all the addresses in the
 780		 * association.
 781		 */
 782		bp = &asoc->base.bind_addr;
 783		laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
 784					       addrcnt, sp);
 785		if ((laddr == NULL) && (addrcnt == 1)) {
 786			if (asoc->asconf_addr_del_pending)
 787				continue;
 788			asoc->asconf_addr_del_pending =
 789			    kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
 790			if (asoc->asconf_addr_del_pending == NULL) {
 791				retval = -ENOMEM;
 792				goto out;
 793			}
 794			asoc->asconf_addr_del_pending->sa.sa_family =
 795				    addrs->sa_family;
 796			asoc->asconf_addr_del_pending->v4.sin_port =
 797				    htons(bp->port);
 798			if (addrs->sa_family == AF_INET) {
 799				struct sockaddr_in *sin;
 800
 801				sin = (struct sockaddr_in *)addrs;
 802				asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
 803			} else if (addrs->sa_family == AF_INET6) {
 804				struct sockaddr_in6 *sin6;
 805
 806				sin6 = (struct sockaddr_in6 *)addrs;
 807				asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
 808			}
 809			SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
 810			    " at %p\n", asoc, asoc->asconf_addr_del_pending,
 811			    asoc->asconf_addr_del_pending);
 
 
 812			asoc->src_out_of_asoc_ok = 1;
 813			stored = 1;
 814			goto skip_mkasconf;
 815		}
 816
 
 
 
 817		/* We do not need RCU protection throughout this loop
 818		 * because this is done under a socket lock from the
 819		 * setsockopt call.
 820		 */
 821		chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
 822						   SCTP_PARAM_DEL_IP);
 823		if (!chunk) {
 824			retval = -ENOMEM;
 825			goto out;
 826		}
 827
 828skip_mkasconf:
 829		/* Reset use_as_src flag for the addresses in the bind address
 830		 * list that are to be deleted.
 831		 */
 832		addr_buf = addrs;
 833		for (i = 0; i < addrcnt; i++) {
 834			laddr = addr_buf;
 835			af = sctp_get_af_specific(laddr->v4.sin_family);
 836			list_for_each_entry(saddr, &bp->address_list, list) {
 837				if (sctp_cmp_addr_exact(&saddr->a, laddr))
 838					saddr->state = SCTP_ADDR_DEL;
 839			}
 840			addr_buf += af->sockaddr_len;
 841		}
 842
 843		/* Update the route and saddr entries for all the transports
 844		 * as some of the addresses in the bind address list are
 845		 * about to be deleted and cannot be used as source addresses.
 846		 */
 847		list_for_each_entry(transport, &asoc->peer.transport_addr_list,
 848					transports) {
 849			dst_release(transport->dst);
 850			sctp_transport_route(transport, NULL,
 851					     sctp_sk(asoc->base.sk));
 852		}
 853
 854		if (stored)
 855			/* We don't need to transmit ASCONF */
 856			continue;
 857		retval = sctp_send_asconf(asoc, chunk);
 858	}
 859out:
 860	return retval;
 861}
 862
 863/* set addr events to assocs in the endpoint.  ep and addr_wq must be locked */
 864int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
 865{
 866	struct sock *sk = sctp_opt2sk(sp);
 867	union sctp_addr *addr;
 868	struct sctp_af *af;
 869
 870	/* It is safe to write port space in caller. */
 871	addr = &addrw->a;
 872	addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
 873	af = sctp_get_af_specific(addr->sa.sa_family);
 874	if (!af)
 875		return -EINVAL;
 876	if (sctp_verify_addr(sk, addr, af->sockaddr_len))
 877		return -EINVAL;
 878
 879	if (addrw->state == SCTP_ADDR_NEW)
 880		return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
 881	else
 882		return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
 883}
 884
 885/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
 886 *
 887 * API 8.1
 888 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
 889 *                int flags);
 890 *
 891 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
 892 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
 893 * or IPv6 addresses.
 894 *
 895 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
 896 * Section 3.1.2 for this usage.
 897 *
 898 * addrs is a pointer to an array of one or more socket addresses. Each
 899 * address is contained in its appropriate structure (i.e. struct
 900 * sockaddr_in or struct sockaddr_in6) the family of the address type
 901 * must be used to distinguish the address length (note that this
 902 * representation is termed a "packed array" of addresses). The caller
 903 * specifies the number of addresses in the array with addrcnt.
 904 *
 905 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
 906 * -1, and sets errno to the appropriate error code.
 907 *
 908 * For SCTP, the port given in each socket address must be the same, or
 909 * sctp_bindx() will fail, setting errno to EINVAL.
 910 *
 911 * The flags parameter is formed from the bitwise OR of zero or more of
 912 * the following currently defined flags:
 913 *
 914 * SCTP_BINDX_ADD_ADDR
 915 *
 916 * SCTP_BINDX_REM_ADDR
 917 *
 918 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
 919 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
 920 * addresses from the association. The two flags are mutually exclusive;
 921 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
 922 * not remove all addresses from an association; sctp_bindx() will
 923 * reject such an attempt with EINVAL.
 924 *
 925 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
 926 * additional addresses with an endpoint after calling bind().  Or use
 927 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
 928 * socket is associated with so that no new association accepted will be
 929 * associated with those addresses. If the endpoint supports dynamic
 930 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
 931 * endpoint to send the appropriate message to the peer to change the
 932 * peers address lists.
 933 *
 934 * Adding and removing addresses from a connected association is
 935 * optional functionality. Implementations that do not support this
 936 * functionality should return EOPNOTSUPP.
 937 *
 938 * Basically do nothing but copying the addresses from user to kernel
 939 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
 940 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
 941 * from userspace.
 942 *
 943 * We don't use copy_from_user() for optimization: we first do the
 944 * sanity checks (buffer size -fast- and access check-healthy
 945 * pointer); if all of those succeed, then we can alloc the memory
 946 * (expensive operation) needed to copy the data to kernel. Then we do
 947 * the copying without checking the user space area
 948 * (__copy_from_user()).
 949 *
 950 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
 951 * it.
 952 *
 953 * sk        The sk of the socket
 954 * addrs     The pointer to the addresses in user land
 955 * addrssize Size of the addrs buffer
 956 * op        Operation to perform (add or remove, see the flags of
 957 *           sctp_bindx)
 958 *
 959 * Returns 0 if ok, <0 errno code on error.
 960 */
 961SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
 962				      struct sockaddr __user *addrs,
 963				      int addrs_size, int op)
 964{
 965	struct sockaddr *kaddrs;
 966	int err;
 967	int addrcnt = 0;
 968	int walk_size = 0;
 969	struct sockaddr *sa_addr;
 970	void *addr_buf;
 971	struct sctp_af *af;
 972
 973	SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
 974			  " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
 975
 976	if (unlikely(addrs_size <= 0))
 977		return -EINVAL;
 978
 979	/* Check the user passed a healthy pointer.  */
 980	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
 981		return -EFAULT;
 982
 983	/* Alloc space for the address array in kernel memory.  */
 984	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
 985	if (unlikely(!kaddrs))
 986		return -ENOMEM;
 987
 988	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
 989		kfree(kaddrs);
 990		return -EFAULT;
 991	}
 992
 993	/* Walk through the addrs buffer and count the number of addresses. */
 994	addr_buf = kaddrs;
 995	while (walk_size < addrs_size) {
 996		if (walk_size + sizeof(sa_family_t) > addrs_size) {
 997			kfree(kaddrs);
 998			return -EINVAL;
 999		}
1000
1001		sa_addr = addr_buf;
1002		af = sctp_get_af_specific(sa_addr->sa_family);
1003
1004		/* If the address family is not supported or if this address
1005		 * causes the address buffer to overflow return EINVAL.
1006		 */
1007		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1008			kfree(kaddrs);
1009			return -EINVAL;
1010		}
1011		addrcnt++;
1012		addr_buf += af->sockaddr_len;
1013		walk_size += af->sockaddr_len;
1014	}
1015
1016	/* Do the work. */
1017	switch (op) {
1018	case SCTP_BINDX_ADD_ADDR:
1019		err = sctp_bindx_add(sk, kaddrs, addrcnt);
1020		if (err)
1021			goto out;
1022		err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1023		break;
1024
1025	case SCTP_BINDX_REM_ADDR:
1026		err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1027		if (err)
1028			goto out;
1029		err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1030		break;
1031
1032	default:
1033		err = -EINVAL;
1034		break;
1035	}
1036
1037out:
1038	kfree(kaddrs);
1039
1040	return err;
1041}
1042
1043/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1044 *
1045 * Common routine for handling connect() and sctp_connectx().
1046 * Connect will come in with just a single address.
1047 */
1048static int __sctp_connect(struct sock* sk,
1049			  struct sockaddr *kaddrs,
1050			  int addrs_size,
1051			  sctp_assoc_t *assoc_id)
1052{
 
1053	struct sctp_sock *sp;
1054	struct sctp_endpoint *ep;
1055	struct sctp_association *asoc = NULL;
1056	struct sctp_association *asoc2;
1057	struct sctp_transport *transport;
1058	union sctp_addr to;
1059	struct sctp_af *af;
1060	sctp_scope_t scope;
1061	long timeo;
1062	int err = 0;
1063	int addrcnt = 0;
1064	int walk_size = 0;
1065	union sctp_addr *sa_addr = NULL;
1066	void *addr_buf;
1067	unsigned short port;
1068	unsigned int f_flags = 0;
1069
1070	sp = sctp_sk(sk);
1071	ep = sp->ep;
1072
1073	/* connect() cannot be done on a socket that is already in ESTABLISHED
1074	 * state - UDP-style peeled off socket or a TCP-style socket that
1075	 * is already connected.
1076	 * It cannot be done even on a TCP-style listening socket.
1077	 */
1078	if (sctp_sstate(sk, ESTABLISHED) ||
1079	    (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1080		err = -EISCONN;
1081		goto out_free;
1082	}
1083
1084	/* Walk through the addrs buffer and count the number of addresses. */
1085	addr_buf = kaddrs;
1086	while (walk_size < addrs_size) {
1087		if (walk_size + sizeof(sa_family_t) > addrs_size) {
1088			err = -EINVAL;
1089			goto out_free;
1090		}
1091
1092		sa_addr = addr_buf;
1093		af = sctp_get_af_specific(sa_addr->sa.sa_family);
1094
1095		/* If the address family is not supported or if this address
1096		 * causes the address buffer to overflow return EINVAL.
1097		 */
1098		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1099			err = -EINVAL;
1100			goto out_free;
1101		}
1102
1103		port = ntohs(sa_addr->v4.sin_port);
1104
1105		/* Save current address so we can work with it */
1106		memcpy(&to, sa_addr, af->sockaddr_len);
1107
1108		err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1109		if (err)
1110			goto out_free;
1111
1112		/* Make sure the destination port is correctly set
1113		 * in all addresses.
1114		 */
1115		if (asoc && asoc->peer.port && asoc->peer.port != port)
 
1116			goto out_free;
1117
1118
1119		/* Check if there already is a matching association on the
1120		 * endpoint (other than the one created here).
1121		 */
1122		asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1123		if (asoc2 && asoc2 != asoc) {
1124			if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1125				err = -EISCONN;
1126			else
1127				err = -EALREADY;
1128			goto out_free;
1129		}
1130
1131		/* If we could not find a matching association on the endpoint,
1132		 * make sure that there is no peeled-off association matching
1133		 * the peer address even on another socket.
1134		 */
1135		if (sctp_endpoint_is_peeled_off(ep, &to)) {
1136			err = -EADDRNOTAVAIL;
1137			goto out_free;
1138		}
1139
1140		if (!asoc) {
1141			/* If a bind() or sctp_bindx() is not called prior to
1142			 * an sctp_connectx() call, the system picks an
1143			 * ephemeral port and will choose an address set
1144			 * equivalent to binding with a wildcard address.
1145			 */
1146			if (!ep->base.bind_addr.port) {
1147				if (sctp_autobind(sk)) {
1148					err = -EAGAIN;
1149					goto out_free;
1150				}
1151			} else {
1152				/*
1153				 * If an unprivileged user inherits a 1-many
1154				 * style socket with open associations on a
1155				 * privileged port, it MAY be permitted to
1156				 * accept new associations, but it SHOULD NOT
1157				 * be permitted to open new associations.
1158				 */
1159				if (ep->base.bind_addr.port < PROT_SOCK &&
1160				    !capable(CAP_NET_BIND_SERVICE)) {
1161					err = -EACCES;
1162					goto out_free;
1163				}
1164			}
1165
1166			scope = sctp_scope(&to);
1167			asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1168			if (!asoc) {
1169				err = -ENOMEM;
1170				goto out_free;
1171			}
1172
1173			err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1174							      GFP_KERNEL);
1175			if (err < 0) {
1176				goto out_free;
1177			}
1178
1179		}
1180
1181		/* Prime the peer's transport structures.  */
1182		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1183						SCTP_UNKNOWN);
1184		if (!transport) {
1185			err = -ENOMEM;
1186			goto out_free;
1187		}
1188
1189		addrcnt++;
1190		addr_buf += af->sockaddr_len;
1191		walk_size += af->sockaddr_len;
1192	}
1193
1194	/* In case the user of sctp_connectx() wants an association
1195	 * id back, assign one now.
1196	 */
1197	if (assoc_id) {
1198		err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1199		if (err < 0)
1200			goto out_free;
1201	}
1202
1203	err = sctp_primitive_ASSOCIATE(asoc, NULL);
1204	if (err < 0) {
1205		goto out_free;
1206	}
1207
1208	/* Initialize sk's dport and daddr for getpeername() */
1209	inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1210	af = sctp_get_af_specific(sa_addr->sa.sa_family);
1211	af->to_sk_daddr(sa_addr, sk);
1212	sk->sk_err = 0;
1213
1214	/* in-kernel sockets don't generally have a file allocated to them
1215	 * if all they do is call sock_create_kern().
1216	 */
1217	if (sk->sk_socket->file)
1218		f_flags = sk->sk_socket->file->f_flags;
1219
1220	timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1221
1222	err = sctp_wait_for_connect(asoc, &timeo);
1223	if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1224		*assoc_id = asoc->assoc_id;
1225
1226	/* Don't free association on exit. */
1227	asoc = NULL;
1228
1229out_free:
 
 
1230
1231	SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1232			  " kaddrs: %p err: %d\n",
1233			  asoc, kaddrs, err);
1234	if (asoc) {
1235		/* sctp_primitive_ASSOCIATE may have added this association
1236		 * To the hash table, try to unhash it, just in case, its a noop
1237		 * if it wasn't hashed so we're safe
1238		 */
1239		sctp_unhash_established(asoc);
1240		sctp_association_free(asoc);
1241	}
1242	return err;
1243}
1244
1245/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1246 *
1247 * API 8.9
1248 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1249 * 			sctp_assoc_t *asoc);
1250 *
1251 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1252 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1253 * or IPv6 addresses.
1254 *
1255 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1256 * Section 3.1.2 for this usage.
1257 *
1258 * addrs is a pointer to an array of one or more socket addresses. Each
1259 * address is contained in its appropriate structure (i.e. struct
1260 * sockaddr_in or struct sockaddr_in6) the family of the address type
1261 * must be used to distengish the address length (note that this
1262 * representation is termed a "packed array" of addresses). The caller
1263 * specifies the number of addresses in the array with addrcnt.
1264 *
1265 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1266 * the association id of the new association.  On failure, sctp_connectx()
1267 * returns -1, and sets errno to the appropriate error code.  The assoc_id
1268 * is not touched by the kernel.
1269 *
1270 * For SCTP, the port given in each socket address must be the same, or
1271 * sctp_connectx() will fail, setting errno to EINVAL.
1272 *
1273 * An application can use sctp_connectx to initiate an association with
1274 * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1275 * allows a caller to specify multiple addresses at which a peer can be
1276 * reached.  The way the SCTP stack uses the list of addresses to set up
1277 * the association is implementation dependent.  This function only
1278 * specifies that the stack will try to make use of all the addresses in
1279 * the list when needed.
1280 *
1281 * Note that the list of addresses passed in is only used for setting up
1282 * the association.  It does not necessarily equal the set of addresses
1283 * the peer uses for the resulting association.  If the caller wants to
1284 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1285 * retrieve them after the association has been set up.
1286 *
1287 * Basically do nothing but copying the addresses from user to kernel
1288 * land and invoking either sctp_connectx(). This is used for tunneling
1289 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1290 *
1291 * We don't use copy_from_user() for optimization: we first do the
1292 * sanity checks (buffer size -fast- and access check-healthy
1293 * pointer); if all of those succeed, then we can alloc the memory
1294 * (expensive operation) needed to copy the data to kernel. Then we do
1295 * the copying without checking the user space area
1296 * (__copy_from_user()).
1297 *
1298 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1299 * it.
1300 *
1301 * sk        The sk of the socket
1302 * addrs     The pointer to the addresses in user land
1303 * addrssize Size of the addrs buffer
1304 *
1305 * Returns >=0 if ok, <0 errno code on error.
1306 */
1307SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1308				      struct sockaddr __user *addrs,
1309				      int addrs_size,
1310				      sctp_assoc_t *assoc_id)
1311{
1312	int err = 0;
1313	struct sockaddr *kaddrs;
1314
1315	SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1316			  __func__, sk, addrs, addrs_size);
1317
1318	if (unlikely(addrs_size <= 0))
1319		return -EINVAL;
1320
1321	/* Check the user passed a healthy pointer.  */
1322	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1323		return -EFAULT;
1324
1325	/* Alloc space for the address array in kernel memory.  */
1326	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1327	if (unlikely(!kaddrs))
1328		return -ENOMEM;
1329
1330	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1331		err = -EFAULT;
1332	} else {
1333		err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1334	}
1335
1336	kfree(kaddrs);
1337
1338	return err;
1339}
1340
1341/*
1342 * This is an older interface.  It's kept for backward compatibility
1343 * to the option that doesn't provide association id.
1344 */
1345SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1346				      struct sockaddr __user *addrs,
1347				      int addrs_size)
1348{
1349	return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1350}
1351
1352/*
1353 * New interface for the API.  The since the API is done with a socket
1354 * option, to make it simple we feed back the association id is as a return
1355 * indication to the call.  Error is always negative and association id is
1356 * always positive.
1357 */
1358SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1359				      struct sockaddr __user *addrs,
1360				      int addrs_size)
1361{
1362	sctp_assoc_t assoc_id = 0;
1363	int err = 0;
1364
1365	err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1366
1367	if (err)
1368		return err;
1369	else
1370		return assoc_id;
1371}
1372
1373/*
1374 * New (hopefully final) interface for the API.
1375 * We use the sctp_getaddrs_old structure so that use-space library
1376 * can avoid any unnecessary allocations.   The only defferent part
1377 * is that we store the actual length of the address buffer into the
1378 * addrs_num structure member.  That way we can re-use the existing
1379 * code.
1380 */
1381SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1382					char __user *optval,
1383					int __user *optlen)
 
 
 
 
 
 
 
 
1384{
1385	struct sctp_getaddrs_old param;
1386	sctp_assoc_t assoc_id = 0;
1387	int err = 0;
1388
1389	if (len < sizeof(param))
1390		return -EINVAL;
 
1391
1392	if (copy_from_user(&param, optval, sizeof(param)))
1393		return -EFAULT;
 
 
1394
1395	err = __sctp_setsockopt_connectx(sk,
1396			(struct sockaddr __user *)param.addrs,
1397			param.addr_num, &assoc_id);
 
 
 
 
 
 
 
 
1398
 
 
 
1399	if (err == 0 || err == -EINPROGRESS) {
1400		if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1401			return -EFAULT;
1402		if (put_user(sizeof(assoc_id), optlen))
1403			return -EFAULT;
1404	}
1405
1406	return err;
1407}
1408
1409/* API 3.1.4 close() - UDP Style Syntax
1410 * Applications use close() to perform graceful shutdown (as described in
1411 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1412 * by a UDP-style socket.
1413 *
1414 * The syntax is
1415 *
1416 *   ret = close(int sd);
1417 *
1418 *   sd      - the socket descriptor of the associations to be closed.
1419 *
1420 * To gracefully shutdown a specific association represented by the
1421 * UDP-style socket, an application should use the sendmsg() call,
1422 * passing no user data, but including the appropriate flag in the
1423 * ancillary data (see Section xxxx).
1424 *
1425 * If sd in the close() call is a branched-off socket representing only
1426 * one association, the shutdown is performed on that association only.
1427 *
1428 * 4.1.6 close() - TCP Style Syntax
1429 *
1430 * Applications use close() to gracefully close down an association.
1431 *
1432 * The syntax is:
1433 *
1434 *    int close(int sd);
1435 *
1436 *      sd      - the socket descriptor of the association to be closed.
1437 *
1438 * After an application calls close() on a socket descriptor, no further
1439 * socket operations will succeed on that descriptor.
1440 *
1441 * API 7.1.4 SO_LINGER
1442 *
1443 * An application using the TCP-style socket can use this option to
1444 * perform the SCTP ABORT primitive.  The linger option structure is:
1445 *
1446 *  struct  linger {
1447 *     int     l_onoff;                // option on/off
1448 *     int     l_linger;               // linger time
1449 * };
1450 *
1451 * To enable the option, set l_onoff to 1.  If the l_linger value is set
1452 * to 0, calling close() is the same as the ABORT primitive.  If the
1453 * value is set to a negative value, the setsockopt() call will return
1454 * an error.  If the value is set to a positive value linger_time, the
1455 * close() can be blocked for at most linger_time ms.  If the graceful
1456 * shutdown phase does not finish during this period, close() will
1457 * return but the graceful shutdown phase continues in the system.
1458 */
1459SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1460{
 
1461	struct sctp_endpoint *ep;
1462	struct sctp_association *asoc;
1463	struct list_head *pos, *temp;
1464	unsigned int data_was_unread;
1465
1466	SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1467
1468	sctp_lock_sock(sk);
1469	sk->sk_shutdown = SHUTDOWN_MASK;
1470	sk->sk_state = SCTP_SS_CLOSING;
1471
1472	ep = sctp_sk(sk)->ep;
1473
1474	/* Clean up any skbs sitting on the receive queue.  */
1475	data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1476	data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1477
1478	/* Walk all associations on an endpoint.  */
1479	list_for_each_safe(pos, temp, &ep->asocs) {
1480		asoc = list_entry(pos, struct sctp_association, asocs);
1481
1482		if (sctp_style(sk, TCP)) {
1483			/* A closed association can still be in the list if
1484			 * it belongs to a TCP-style listening socket that is
1485			 * not yet accepted. If so, free it. If not, send an
1486			 * ABORT or SHUTDOWN based on the linger options.
1487			 */
1488			if (sctp_state(asoc, CLOSED)) {
1489				sctp_unhash_established(asoc);
1490				sctp_association_free(asoc);
1491				continue;
1492			}
1493		}
1494
1495		if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1496		    !skb_queue_empty(&asoc->ulpq.reasm) ||
1497		    (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1498			struct sctp_chunk *chunk;
1499
1500			chunk = sctp_make_abort_user(asoc, NULL, 0);
1501			if (chunk)
1502				sctp_primitive_ABORT(asoc, chunk);
1503		} else
1504			sctp_primitive_SHUTDOWN(asoc, NULL);
1505	}
1506
1507	/* On a TCP-style socket, block for at most linger_time if set. */
1508	if (sctp_style(sk, TCP) && timeout)
1509		sctp_wait_for_close(sk, timeout);
1510
1511	/* This will run the backlog queue.  */
1512	sctp_release_sock(sk);
1513
1514	/* Supposedly, no process has access to the socket, but
1515	 * the net layers still may.
1516	 */
1517	sctp_local_bh_disable();
1518	sctp_bh_lock_sock(sk);
1519
1520	/* Hold the sock, since sk_common_release() will put sock_put()
1521	 * and we have just a little more cleanup.
1522	 */
1523	sock_hold(sk);
1524	sk_common_release(sk);
1525
1526	sctp_bh_unlock_sock(sk);
1527	sctp_local_bh_enable();
1528
1529	sock_put(sk);
1530
1531	SCTP_DBG_OBJCNT_DEC(sock);
1532}
1533
1534/* Handle EPIPE error. */
1535static int sctp_error(struct sock *sk, int flags, int err)
1536{
1537	if (err == -EPIPE)
1538		err = sock_error(sk) ? : -EPIPE;
1539	if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1540		send_sig(SIGPIPE, current, 0);
1541	return err;
1542}
1543
1544/* API 3.1.3 sendmsg() - UDP Style Syntax
1545 *
1546 * An application uses sendmsg() and recvmsg() calls to transmit data to
1547 * and receive data from its peer.
1548 *
1549 *  ssize_t sendmsg(int socket, const struct msghdr *message,
1550 *                  int flags);
1551 *
1552 *  socket  - the socket descriptor of the endpoint.
1553 *  message - pointer to the msghdr structure which contains a single
1554 *            user message and possibly some ancillary data.
1555 *
1556 *            See Section 5 for complete description of the data
1557 *            structures.
1558 *
1559 *  flags   - flags sent or received with the user message, see Section
1560 *            5 for complete description of the flags.
1561 *
1562 * Note:  This function could use a rewrite especially when explicit
1563 * connect support comes in.
1564 */
1565/* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1566
1567SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1568
1569SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1570			     struct msghdr *msg, size_t msg_len)
1571{
 
1572	struct sctp_sock *sp;
1573	struct sctp_endpoint *ep;
1574	struct sctp_association *new_asoc=NULL, *asoc=NULL;
1575	struct sctp_transport *transport, *chunk_tp;
1576	struct sctp_chunk *chunk;
1577	union sctp_addr to;
1578	struct sockaddr *msg_name = NULL;
1579	struct sctp_sndrcvinfo default_sinfo;
1580	struct sctp_sndrcvinfo *sinfo;
1581	struct sctp_initmsg *sinit;
1582	sctp_assoc_t associd = 0;
1583	sctp_cmsgs_t cmsgs = { NULL };
1584	int err;
1585	sctp_scope_t scope;
1586	long timeo;
1587	__u16 sinfo_flags = 0;
1588	struct sctp_datamsg *datamsg;
1589	int msg_flags = msg->msg_flags;
1590
1591	SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1592			  sk, msg, msg_len);
1593
1594	err = 0;
1595	sp = sctp_sk(sk);
1596	ep = sp->ep;
1597
1598	SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
 
1599
1600	/* We cannot send a message over a TCP-style listening socket. */
1601	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1602		err = -EPIPE;
1603		goto out_nounlock;
1604	}
1605
1606	/* Parse out the SCTP CMSGs.  */
1607	err = sctp_msghdr_parse(msg, &cmsgs);
1608
1609	if (err) {
1610		SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1611		goto out_nounlock;
1612	}
1613
1614	/* Fetch the destination address for this packet.  This
1615	 * address only selects the association--it is not necessarily
1616	 * the address we will send to.
1617	 * For a peeled-off socket, msg_name is ignored.
1618	 */
1619	if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1620		int msg_namelen = msg->msg_namelen;
1621
1622		err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1623				       msg_namelen);
1624		if (err)
1625			return err;
1626
1627		if (msg_namelen > sizeof(to))
1628			msg_namelen = sizeof(to);
1629		memcpy(&to, msg->msg_name, msg_namelen);
1630		msg_name = msg->msg_name;
1631	}
1632
1633	sinfo = cmsgs.info;
1634	sinit = cmsgs.init;
1635
1636	/* Did the user specify SNDRCVINFO?  */
1637	if (sinfo) {
1638		sinfo_flags = sinfo->sinfo_flags;
1639		associd = sinfo->sinfo_assoc_id;
1640	}
1641
1642	SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1643			  msg_len, sinfo_flags);
1644
1645	/* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1646	if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1647		err = -EINVAL;
1648		goto out_nounlock;
1649	}
1650
1651	/* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1652	 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1653	 * If SCTP_ABORT is set, the message length could be non zero with
1654	 * the msg_iov set to the user abort reason.
1655	 */
1656	if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1657	    (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1658		err = -EINVAL;
1659		goto out_nounlock;
1660	}
1661
1662	/* If SCTP_ADDR_OVER is set, there must be an address
1663	 * specified in msg_name.
1664	 */
1665	if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1666		err = -EINVAL;
1667		goto out_nounlock;
1668	}
1669
1670	transport = NULL;
1671
1672	SCTP_DEBUG_PRINTK("About to look up association.\n");
1673
1674	sctp_lock_sock(sk);
1675
1676	/* If a msg_name has been specified, assume this is to be used.  */
1677	if (msg_name) {
1678		/* Look for a matching association on the endpoint. */
1679		asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1680		if (!asoc) {
1681			/* If we could not find a matching association on the
1682			 * endpoint, make sure that it is not a TCP-style
1683			 * socket that already has an association or there is
1684			 * no peeled-off association on another socket.
1685			 */
1686			if ((sctp_style(sk, TCP) &&
1687			     sctp_sstate(sk, ESTABLISHED)) ||
1688			    sctp_endpoint_is_peeled_off(ep, &to)) {
1689				err = -EADDRNOTAVAIL;
1690				goto out_unlock;
1691			}
1692		}
1693	} else {
1694		asoc = sctp_id2assoc(sk, associd);
1695		if (!asoc) {
1696			err = -EPIPE;
1697			goto out_unlock;
1698		}
1699	}
1700
1701	if (asoc) {
1702		SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1703
1704		/* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1705		 * socket that has an association in CLOSED state. This can
1706		 * happen when an accepted socket has an association that is
1707		 * already CLOSED.
1708		 */
1709		if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1710			err = -EPIPE;
1711			goto out_unlock;
1712		}
1713
1714		if (sinfo_flags & SCTP_EOF) {
1715			SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1716					  asoc);
1717			sctp_primitive_SHUTDOWN(asoc, NULL);
 
1718			err = 0;
1719			goto out_unlock;
1720		}
1721		if (sinfo_flags & SCTP_ABORT) {
1722
1723			chunk = sctp_make_abort_user(asoc, msg, msg_len);
1724			if (!chunk) {
1725				err = -ENOMEM;
1726				goto out_unlock;
1727			}
1728
1729			SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1730			sctp_primitive_ABORT(asoc, chunk);
 
 
1731			err = 0;
1732			goto out_unlock;
1733		}
1734	}
1735
1736	/* Do we need to create the association?  */
1737	if (!asoc) {
1738		SCTP_DEBUG_PRINTK("There is no association yet.\n");
1739
1740		if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1741			err = -EINVAL;
1742			goto out_unlock;
1743		}
1744
1745		/* Check for invalid stream against the stream counts,
1746		 * either the default or the user specified stream counts.
1747		 */
1748		if (sinfo) {
1749			if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1750				/* Check against the defaults. */
1751				if (sinfo->sinfo_stream >=
1752				    sp->initmsg.sinit_num_ostreams) {
1753					err = -EINVAL;
1754					goto out_unlock;
1755				}
1756			} else {
1757				/* Check against the requested.  */
1758				if (sinfo->sinfo_stream >=
1759				    sinit->sinit_num_ostreams) {
1760					err = -EINVAL;
1761					goto out_unlock;
1762				}
1763			}
1764		}
1765
1766		/*
1767		 * API 3.1.2 bind() - UDP Style Syntax
1768		 * If a bind() or sctp_bindx() is not called prior to a
1769		 * sendmsg() call that initiates a new association, the
1770		 * system picks an ephemeral port and will choose an address
1771		 * set equivalent to binding with a wildcard address.
1772		 */
1773		if (!ep->base.bind_addr.port) {
1774			if (sctp_autobind(sk)) {
1775				err = -EAGAIN;
1776				goto out_unlock;
1777			}
1778		} else {
1779			/*
1780			 * If an unprivileged user inherits a one-to-many
1781			 * style socket with open associations on a privileged
1782			 * port, it MAY be permitted to accept new associations,
1783			 * but it SHOULD NOT be permitted to open new
1784			 * associations.
1785			 */
1786			if (ep->base.bind_addr.port < PROT_SOCK &&
1787			    !capable(CAP_NET_BIND_SERVICE)) {
1788				err = -EACCES;
1789				goto out_unlock;
1790			}
1791		}
1792
1793		scope = sctp_scope(&to);
1794		new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1795		if (!new_asoc) {
1796			err = -ENOMEM;
1797			goto out_unlock;
1798		}
1799		asoc = new_asoc;
1800		err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1801		if (err < 0) {
1802			err = -ENOMEM;
1803			goto out_free;
1804		}
1805
1806		/* If the SCTP_INIT ancillary data is specified, set all
1807		 * the association init values accordingly.
1808		 */
1809		if (sinit) {
1810			if (sinit->sinit_num_ostreams) {
1811				asoc->c.sinit_num_ostreams =
1812					sinit->sinit_num_ostreams;
1813			}
1814			if (sinit->sinit_max_instreams) {
1815				asoc->c.sinit_max_instreams =
1816					sinit->sinit_max_instreams;
1817			}
1818			if (sinit->sinit_max_attempts) {
1819				asoc->max_init_attempts
1820					= sinit->sinit_max_attempts;
1821			}
1822			if (sinit->sinit_max_init_timeo) {
1823				asoc->max_init_timeo =
1824				 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1825			}
1826		}
1827
1828		/* Prime the peer's transport structures.  */
1829		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1830		if (!transport) {
1831			err = -ENOMEM;
1832			goto out_free;
1833		}
1834	}
1835
1836	/* ASSERT: we have a valid association at this point.  */
1837	SCTP_DEBUG_PRINTK("We have a valid association.\n");
1838
1839	if (!sinfo) {
1840		/* If the user didn't specify SNDRCVINFO, make up one with
1841		 * some defaults.
1842		 */
1843		memset(&default_sinfo, 0, sizeof(default_sinfo));
1844		default_sinfo.sinfo_stream = asoc->default_stream;
1845		default_sinfo.sinfo_flags = asoc->default_flags;
1846		default_sinfo.sinfo_ppid = asoc->default_ppid;
1847		default_sinfo.sinfo_context = asoc->default_context;
1848		default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1849		default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1850		sinfo = &default_sinfo;
1851	}
1852
1853	/* API 7.1.7, the sndbuf size per association bounds the
1854	 * maximum size of data that can be sent in a single send call.
1855	 */
1856	if (msg_len > sk->sk_sndbuf) {
1857		err = -EMSGSIZE;
1858		goto out_free;
1859	}
1860
1861	if (asoc->pmtu_pending)
1862		sctp_assoc_pending_pmtu(asoc);
1863
1864	/* If fragmentation is disabled and the message length exceeds the
1865	 * association fragmentation point, return EMSGSIZE.  The I-D
1866	 * does not specify what this error is, but this looks like
1867	 * a great fit.
1868	 */
1869	if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1870		err = -EMSGSIZE;
1871		goto out_free;
1872	}
1873
1874	/* Check for invalid stream. */
1875	if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1876		err = -EINVAL;
1877		goto out_free;
1878	}
1879
1880	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1881	if (!sctp_wspace(asoc)) {
1882		err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1883		if (err)
1884			goto out_free;
1885	}
1886
1887	/* If an address is passed with the sendto/sendmsg call, it is used
1888	 * to override the primary destination address in the TCP model, or
1889	 * when SCTP_ADDR_OVER flag is set in the UDP model.
1890	 */
1891	if ((sctp_style(sk, TCP) && msg_name) ||
1892	    (sinfo_flags & SCTP_ADDR_OVER)) {
1893		chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1894		if (!chunk_tp) {
1895			err = -EINVAL;
1896			goto out_free;
1897		}
1898	} else
1899		chunk_tp = NULL;
1900
1901	/* Auto-connect, if we aren't connected already. */
1902	if (sctp_state(asoc, CLOSED)) {
1903		err = sctp_primitive_ASSOCIATE(asoc, NULL);
1904		if (err < 0)
1905			goto out_free;
1906		SCTP_DEBUG_PRINTK("We associated primitively.\n");
 
1907	}
1908
1909	/* Break the message into multiple chunks of maximum size. */
1910	datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1911	if (!datamsg) {
1912		err = -ENOMEM;
1913		goto out_free;
1914	}
1915
1916	/* Now send the (possibly) fragmented message. */
1917	list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1918		sctp_chunk_hold(chunk);
1919
1920		/* Do accounting for the write space.  */
1921		sctp_set_owner_w(chunk);
1922
1923		chunk->transport = chunk_tp;
1924	}
1925
1926	/* Send it to the lower layers.  Note:  all chunks
1927	 * must either fail or succeed.   The lower layer
1928	 * works that way today.  Keep it that way or this
1929	 * breaks.
1930	 */
1931	err = sctp_primitive_SEND(asoc, datamsg);
1932	/* Did the lower layer accept the chunk? */
1933	if (err)
1934		sctp_datamsg_free(datamsg);
1935	else
1936		sctp_datamsg_put(datamsg);
1937
1938	SCTP_DEBUG_PRINTK("We sent primitively.\n");
1939
1940	if (err)
1941		goto out_free;
1942	else
1943		err = msg_len;
1944
1945	/* If we are already past ASSOCIATE, the lower
1946	 * layers are responsible for association cleanup.
1947	 */
1948	goto out_unlock;
1949
1950out_free:
1951	if (new_asoc) {
1952		sctp_unhash_established(asoc);
1953		sctp_association_free(asoc);
1954	}
1955out_unlock:
1956	sctp_release_sock(sk);
1957
1958out_nounlock:
1959	return sctp_error(sk, msg_flags, err);
1960
1961#if 0
1962do_sock_err:
1963	if (msg_len)
1964		err = msg_len;
1965	else
1966		err = sock_error(sk);
1967	goto out;
1968
1969do_interrupted:
1970	if (msg_len)
1971		err = msg_len;
1972	goto out;
1973#endif /* 0 */
1974}
1975
1976/* This is an extended version of skb_pull() that removes the data from the
1977 * start of a skb even when data is spread across the list of skb's in the
1978 * frag_list. len specifies the total amount of data that needs to be removed.
1979 * when 'len' bytes could be removed from the skb, it returns 0.
1980 * If 'len' exceeds the total skb length,  it returns the no. of bytes that
1981 * could not be removed.
1982 */
1983static int sctp_skb_pull(struct sk_buff *skb, int len)
1984{
1985	struct sk_buff *list;
1986	int skb_len = skb_headlen(skb);
1987	int rlen;
1988
1989	if (len <= skb_len) {
1990		__skb_pull(skb, len);
1991		return 0;
1992	}
1993	len -= skb_len;
1994	__skb_pull(skb, skb_len);
1995
1996	skb_walk_frags(skb, list) {
1997		rlen = sctp_skb_pull(list, len);
1998		skb->len -= (len-rlen);
1999		skb->data_len -= (len-rlen);
2000
2001		if (!rlen)
2002			return 0;
2003
2004		len = rlen;
2005	}
2006
2007	return len;
2008}
2009
2010/* API 3.1.3  recvmsg() - UDP Style Syntax
2011 *
2012 *  ssize_t recvmsg(int socket, struct msghdr *message,
2013 *                    int flags);
2014 *
2015 *  socket  - the socket descriptor of the endpoint.
2016 *  message - pointer to the msghdr structure which contains a single
2017 *            user message and possibly some ancillary data.
2018 *
2019 *            See Section 5 for complete description of the data
2020 *            structures.
2021 *
2022 *  flags   - flags sent or received with the user message, see Section
2023 *            5 for complete description of the flags.
2024 */
2025static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2026
2027SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2028			     struct msghdr *msg, size_t len, int noblock,
2029			     int flags, int *addr_len)
2030{
2031	struct sctp_ulpevent *event = NULL;
2032	struct sctp_sock *sp = sctp_sk(sk);
2033	struct sk_buff *skb;
2034	int copied;
2035	int err = 0;
2036	int skb_len;
2037
2038	SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2039			  "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2040			  "len", len, "knoblauch", noblock,
2041			  "flags", flags, "addr_len", addr_len);
2042
2043	sctp_lock_sock(sk);
2044
2045	if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2046		err = -ENOTCONN;
2047		goto out;
2048	}
2049
2050	skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2051	if (!skb)
2052		goto out;
2053
2054	/* Get the total length of the skb including any skb's in the
2055	 * frag_list.
2056	 */
2057	skb_len = skb->len;
2058
2059	copied = skb_len;
2060	if (copied > len)
2061		copied = len;
2062
2063	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2064
2065	event = sctp_skb2event(skb);
2066
2067	if (err)
2068		goto out_free;
2069
2070	sock_recv_ts_and_drops(msg, sk, skb);
2071	if (sctp_ulpevent_is_notification(event)) {
2072		msg->msg_flags |= MSG_NOTIFICATION;
2073		sp->pf->event_msgname(event, msg->msg_name, addr_len);
2074	} else {
2075		sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2076	}
2077
2078	/* Check if we allow SCTP_SNDRCVINFO. */
2079	if (sp->subscribe.sctp_data_io_event)
2080		sctp_ulpevent_read_sndrcvinfo(event, msg);
2081#if 0
2082	/* FIXME: we should be calling IP/IPv6 layers.  */
2083	if (sk->sk_protinfo.af_inet.cmsg_flags)
2084		ip_cmsg_recv(msg, skb);
2085#endif
2086
2087	err = copied;
2088
2089	/* If skb's length exceeds the user's buffer, update the skb and
2090	 * push it back to the receive_queue so that the next call to
2091	 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2092	 */
2093	if (skb_len > copied) {
2094		msg->msg_flags &= ~MSG_EOR;
2095		if (flags & MSG_PEEK)
2096			goto out_free;
2097		sctp_skb_pull(skb, copied);
2098		skb_queue_head(&sk->sk_receive_queue, skb);
2099
2100		/* When only partial message is copied to the user, increase
2101		 * rwnd by that amount. If all the data in the skb is read,
2102		 * rwnd is updated when the event is freed.
2103		 */
2104		if (!sctp_ulpevent_is_notification(event))
2105			sctp_assoc_rwnd_increase(event->asoc, copied);
2106		goto out;
2107	} else if ((event->msg_flags & MSG_NOTIFICATION) ||
2108		   (event->msg_flags & MSG_EOR))
2109		msg->msg_flags |= MSG_EOR;
2110	else
2111		msg->msg_flags &= ~MSG_EOR;
2112
2113out_free:
2114	if (flags & MSG_PEEK) {
2115		/* Release the skb reference acquired after peeking the skb in
2116		 * sctp_skb_recv_datagram().
2117		 */
2118		kfree_skb(skb);
2119	} else {
2120		/* Free the event which includes releasing the reference to
2121		 * the owner of the skb, freeing the skb and updating the
2122		 * rwnd.
2123		 */
2124		sctp_ulpevent_free(event);
2125	}
2126out:
2127	sctp_release_sock(sk);
2128	return err;
2129}
2130
2131/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2132 *
2133 * This option is a on/off flag.  If enabled no SCTP message
2134 * fragmentation will be performed.  Instead if a message being sent
2135 * exceeds the current PMTU size, the message will NOT be sent and
2136 * instead a error will be indicated to the user.
2137 */
2138static int sctp_setsockopt_disable_fragments(struct sock *sk,
2139					     char __user *optval,
2140					     unsigned int optlen)
2141{
2142	int val;
2143
2144	if (optlen < sizeof(int))
2145		return -EINVAL;
2146
2147	if (get_user(val, (int __user *)optval))
2148		return -EFAULT;
2149
2150	sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2151
2152	return 0;
2153}
2154
2155static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2156				  unsigned int optlen)
2157{
2158	struct sctp_association *asoc;
2159	struct sctp_ulpevent *event;
2160
2161	if (optlen > sizeof(struct sctp_event_subscribe))
2162		return -EINVAL;
2163	if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2164		return -EFAULT;
2165
2166	/*
2167	 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2168	 * if there is no data to be sent or retransmit, the stack will
2169	 * immediately send up this notification.
2170	 */
2171	if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2172				       &sctp_sk(sk)->subscribe)) {
2173		asoc = sctp_id2assoc(sk, 0);
2174
2175		if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2176			event = sctp_ulpevent_make_sender_dry_event(asoc,
2177					GFP_ATOMIC);
2178			if (!event)
2179				return -ENOMEM;
2180
2181			sctp_ulpq_tail_event(&asoc->ulpq, event);
2182		}
2183	}
2184
2185	return 0;
2186}
2187
2188/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2189 *
2190 * This socket option is applicable to the UDP-style socket only.  When
2191 * set it will cause associations that are idle for more than the
2192 * specified number of seconds to automatically close.  An association
2193 * being idle is defined an association that has NOT sent or received
2194 * user data.  The special value of '0' indicates that no automatic
2195 * close of any associations should be performed.  The option expects an
2196 * integer defining the number of seconds of idle time before an
2197 * association is closed.
2198 */
2199static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2200				     unsigned int optlen)
2201{
2202	struct sctp_sock *sp = sctp_sk(sk);
 
2203
2204	/* Applicable to UDP-style socket only */
2205	if (sctp_style(sk, TCP))
2206		return -EOPNOTSUPP;
2207	if (optlen != sizeof(int))
2208		return -EINVAL;
2209	if (copy_from_user(&sp->autoclose, optval, optlen))
2210		return -EFAULT;
2211
 
 
 
2212	return 0;
2213}
2214
2215/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2216 *
2217 * Applications can enable or disable heartbeats for any peer address of
2218 * an association, modify an address's heartbeat interval, force a
2219 * heartbeat to be sent immediately, and adjust the address's maximum
2220 * number of retransmissions sent before an address is considered
2221 * unreachable.  The following structure is used to access and modify an
2222 * address's parameters:
2223 *
2224 *  struct sctp_paddrparams {
2225 *     sctp_assoc_t            spp_assoc_id;
2226 *     struct sockaddr_storage spp_address;
2227 *     uint32_t                spp_hbinterval;
2228 *     uint16_t                spp_pathmaxrxt;
2229 *     uint32_t                spp_pathmtu;
2230 *     uint32_t                spp_sackdelay;
2231 *     uint32_t                spp_flags;
2232 * };
2233 *
2234 *   spp_assoc_id    - (one-to-many style socket) This is filled in the
2235 *                     application, and identifies the association for
2236 *                     this query.
2237 *   spp_address     - This specifies which address is of interest.
2238 *   spp_hbinterval  - This contains the value of the heartbeat interval,
2239 *                     in milliseconds.  If a  value of zero
2240 *                     is present in this field then no changes are to
2241 *                     be made to this parameter.
2242 *   spp_pathmaxrxt  - This contains the maximum number of
2243 *                     retransmissions before this address shall be
2244 *                     considered unreachable. If a  value of zero
2245 *                     is present in this field then no changes are to
2246 *                     be made to this parameter.
2247 *   spp_pathmtu     - When Path MTU discovery is disabled the value
2248 *                     specified here will be the "fixed" path mtu.
2249 *                     Note that if the spp_address field is empty
2250 *                     then all associations on this address will
2251 *                     have this fixed path mtu set upon them.
2252 *
2253 *   spp_sackdelay   - When delayed sack is enabled, this value specifies
2254 *                     the number of milliseconds that sacks will be delayed
2255 *                     for. This value will apply to all addresses of an
2256 *                     association if the spp_address field is empty. Note
2257 *                     also, that if delayed sack is enabled and this
2258 *                     value is set to 0, no change is made to the last
2259 *                     recorded delayed sack timer value.
2260 *
2261 *   spp_flags       - These flags are used to control various features
2262 *                     on an association. The flag field may contain
2263 *                     zero or more of the following options.
2264 *
2265 *                     SPP_HB_ENABLE  - Enable heartbeats on the
2266 *                     specified address. Note that if the address
2267 *                     field is empty all addresses for the association
2268 *                     have heartbeats enabled upon them.
2269 *
2270 *                     SPP_HB_DISABLE - Disable heartbeats on the
2271 *                     speicifed address. Note that if the address
2272 *                     field is empty all addresses for the association
2273 *                     will have their heartbeats disabled. Note also
2274 *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
2275 *                     mutually exclusive, only one of these two should
2276 *                     be specified. Enabling both fields will have
2277 *                     undetermined results.
2278 *
2279 *                     SPP_HB_DEMAND - Request a user initiated heartbeat
2280 *                     to be made immediately.
2281 *
2282 *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
2283 *                     heartbeat delayis to be set to the value of 0
2284 *                     milliseconds.
2285 *
2286 *                     SPP_PMTUD_ENABLE - This field will enable PMTU
2287 *                     discovery upon the specified address. Note that
2288 *                     if the address feild is empty then all addresses
2289 *                     on the association are effected.
2290 *
2291 *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2292 *                     discovery upon the specified address. Note that
2293 *                     if the address feild is empty then all addresses
2294 *                     on the association are effected. Not also that
2295 *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2296 *                     exclusive. Enabling both will have undetermined
2297 *                     results.
2298 *
2299 *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2300 *                     on delayed sack. The time specified in spp_sackdelay
2301 *                     is used to specify the sack delay for this address. Note
2302 *                     that if spp_address is empty then all addresses will
2303 *                     enable delayed sack and take on the sack delay
2304 *                     value specified in spp_sackdelay.
2305 *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2306 *                     off delayed sack. If the spp_address field is blank then
2307 *                     delayed sack is disabled for the entire association. Note
2308 *                     also that this field is mutually exclusive to
2309 *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2310 *                     results.
2311 */
2312static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2313				       struct sctp_transport   *trans,
2314				       struct sctp_association *asoc,
2315				       struct sctp_sock        *sp,
2316				       int                      hb_change,
2317				       int                      pmtud_change,
2318				       int                      sackdelay_change)
2319{
2320	int error;
2321
2322	if (params->spp_flags & SPP_HB_DEMAND && trans) {
2323		error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
 
 
2324		if (error)
2325			return error;
2326	}
2327
2328	/* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2329	 * this field is ignored.  Note also that a value of zero indicates
2330	 * the current setting should be left unchanged.
2331	 */
2332	if (params->spp_flags & SPP_HB_ENABLE) {
2333
2334		/* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2335		 * set.  This lets us use 0 value when this flag
2336		 * is set.
2337		 */
2338		if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2339			params->spp_hbinterval = 0;
2340
2341		if (params->spp_hbinterval ||
2342		    (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2343			if (trans) {
2344				trans->hbinterval =
2345				    msecs_to_jiffies(params->spp_hbinterval);
2346			} else if (asoc) {
2347				asoc->hbinterval =
2348				    msecs_to_jiffies(params->spp_hbinterval);
2349			} else {
2350				sp->hbinterval = params->spp_hbinterval;
2351			}
2352		}
2353	}
2354
2355	if (hb_change) {
2356		if (trans) {
2357			trans->param_flags =
2358				(trans->param_flags & ~SPP_HB) | hb_change;
2359		} else if (asoc) {
2360			asoc->param_flags =
2361				(asoc->param_flags & ~SPP_HB) | hb_change;
2362		} else {
2363			sp->param_flags =
2364				(sp->param_flags & ~SPP_HB) | hb_change;
2365		}
2366	}
2367
2368	/* When Path MTU discovery is disabled the value specified here will
2369	 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2370	 * include the flag SPP_PMTUD_DISABLE for this field to have any
2371	 * effect).
2372	 */
2373	if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2374		if (trans) {
2375			trans->pathmtu = params->spp_pathmtu;
2376			sctp_assoc_sync_pmtu(asoc);
2377		} else if (asoc) {
2378			asoc->pathmtu = params->spp_pathmtu;
2379			sctp_frag_point(asoc, params->spp_pathmtu);
2380		} else {
2381			sp->pathmtu = params->spp_pathmtu;
2382		}
2383	}
2384
2385	if (pmtud_change) {
2386		if (trans) {
2387			int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2388				(params->spp_flags & SPP_PMTUD_ENABLE);
2389			trans->param_flags =
2390				(trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2391			if (update) {
2392				sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2393				sctp_assoc_sync_pmtu(asoc);
2394			}
2395		} else if (asoc) {
2396			asoc->param_flags =
2397				(asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2398		} else {
2399			sp->param_flags =
2400				(sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2401		}
2402	}
2403
2404	/* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2405	 * value of this field is ignored.  Note also that a value of zero
2406	 * indicates the current setting should be left unchanged.
2407	 */
2408	if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2409		if (trans) {
2410			trans->sackdelay =
2411				msecs_to_jiffies(params->spp_sackdelay);
2412		} else if (asoc) {
2413			asoc->sackdelay =
2414				msecs_to_jiffies(params->spp_sackdelay);
2415		} else {
2416			sp->sackdelay = params->spp_sackdelay;
2417		}
2418	}
2419
2420	if (sackdelay_change) {
2421		if (trans) {
2422			trans->param_flags =
2423				(trans->param_flags & ~SPP_SACKDELAY) |
2424				sackdelay_change;
2425		} else if (asoc) {
2426			asoc->param_flags =
2427				(asoc->param_flags & ~SPP_SACKDELAY) |
2428				sackdelay_change;
2429		} else {
2430			sp->param_flags =
2431				(sp->param_flags & ~SPP_SACKDELAY) |
2432				sackdelay_change;
2433		}
2434	}
2435
2436	/* Note that a value of zero indicates the current setting should be
2437	   left unchanged.
2438	 */
2439	if (params->spp_pathmaxrxt) {
2440		if (trans) {
2441			trans->pathmaxrxt = params->spp_pathmaxrxt;
2442		} else if (asoc) {
2443			asoc->pathmaxrxt = params->spp_pathmaxrxt;
2444		} else {
2445			sp->pathmaxrxt = params->spp_pathmaxrxt;
2446		}
2447	}
2448
2449	return 0;
2450}
2451
2452static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2453					    char __user *optval,
2454					    unsigned int optlen)
2455{
2456	struct sctp_paddrparams  params;
2457	struct sctp_transport   *trans = NULL;
2458	struct sctp_association *asoc = NULL;
2459	struct sctp_sock        *sp = sctp_sk(sk);
2460	int error;
2461	int hb_change, pmtud_change, sackdelay_change;
2462
2463	if (optlen != sizeof(struct sctp_paddrparams))
2464		return - EINVAL;
2465
2466	if (copy_from_user(&params, optval, optlen))
2467		return -EFAULT;
2468
2469	/* Validate flags and value parameters. */
2470	hb_change        = params.spp_flags & SPP_HB;
2471	pmtud_change     = params.spp_flags & SPP_PMTUD;
2472	sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2473
2474	if (hb_change        == SPP_HB ||
2475	    pmtud_change     == SPP_PMTUD ||
2476	    sackdelay_change == SPP_SACKDELAY ||
2477	    params.spp_sackdelay > 500 ||
2478	    (params.spp_pathmtu &&
2479	     params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2480		return -EINVAL;
2481
2482	/* If an address other than INADDR_ANY is specified, and
2483	 * no transport is found, then the request is invalid.
2484	 */
2485	if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2486		trans = sctp_addr_id2transport(sk, &params.spp_address,
2487					       params.spp_assoc_id);
2488		if (!trans)
2489			return -EINVAL;
2490	}
2491
2492	/* Get association, if assoc_id != 0 and the socket is a one
2493	 * to many style socket, and an association was not found, then
2494	 * the id was invalid.
2495	 */
2496	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2497	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2498		return -EINVAL;
2499
2500	/* Heartbeat demand can only be sent on a transport or
2501	 * association, but not a socket.
2502	 */
2503	if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2504		return -EINVAL;
2505
2506	/* Process parameters. */
2507	error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2508					    hb_change, pmtud_change,
2509					    sackdelay_change);
2510
2511	if (error)
2512		return error;
2513
2514	/* If changes are for association, also apply parameters to each
2515	 * transport.
2516	 */
2517	if (!trans && asoc) {
2518		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2519				transports) {
2520			sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2521						    hb_change, pmtud_change,
2522						    sackdelay_change);
2523		}
2524	}
2525
2526	return 0;
2527}
2528
 
 
 
 
 
 
 
 
 
 
2529/*
2530 * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
2531 *
2532 * This option will effect the way delayed acks are performed.  This
2533 * option allows you to get or set the delayed ack time, in
2534 * milliseconds.  It also allows changing the delayed ack frequency.
2535 * Changing the frequency to 1 disables the delayed sack algorithm.  If
2536 * the assoc_id is 0, then this sets or gets the endpoints default
2537 * values.  If the assoc_id field is non-zero, then the set or get
2538 * effects the specified association for the one to many model (the
2539 * assoc_id field is ignored by the one to one model).  Note that if
2540 * sack_delay or sack_freq are 0 when setting this option, then the
2541 * current values will remain unchanged.
2542 *
2543 * struct sctp_sack_info {
2544 *     sctp_assoc_t            sack_assoc_id;
2545 *     uint32_t                sack_delay;
2546 *     uint32_t                sack_freq;
2547 * };
2548 *
2549 * sack_assoc_id -  This parameter, indicates which association the user
2550 *    is performing an action upon.  Note that if this field's value is
2551 *    zero then the endpoints default value is changed (effecting future
2552 *    associations only).
2553 *
2554 * sack_delay -  This parameter contains the number of milliseconds that
2555 *    the user is requesting the delayed ACK timer be set to.  Note that
2556 *    this value is defined in the standard to be between 200 and 500
2557 *    milliseconds.
2558 *
2559 * sack_freq -  This parameter contains the number of packets that must
2560 *    be received before a sack is sent without waiting for the delay
2561 *    timer to expire.  The default value for this is 2, setting this
2562 *    value to 1 will disable the delayed sack algorithm.
2563 */
2564
2565static int sctp_setsockopt_delayed_ack(struct sock *sk,
2566				       char __user *optval, unsigned int optlen)
2567{
2568	struct sctp_sack_info    params;
2569	struct sctp_transport   *trans = NULL;
2570	struct sctp_association *asoc = NULL;
2571	struct sctp_sock        *sp = sctp_sk(sk);
2572
2573	if (optlen == sizeof(struct sctp_sack_info)) {
2574		if (copy_from_user(&params, optval, optlen))
2575			return -EFAULT;
2576
2577		if (params.sack_delay == 0 && params.sack_freq == 0)
2578			return 0;
2579	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2580		pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2581		pr_warn("Use struct sctp_sack_info instead\n");
 
 
 
2582		if (copy_from_user(&params, optval, optlen))
2583			return -EFAULT;
2584
2585		if (params.sack_delay == 0)
2586			params.sack_freq = 1;
2587		else
2588			params.sack_freq = 0;
2589	} else
2590		return - EINVAL;
2591
2592	/* Validate value parameter. */
2593	if (params.sack_delay > 500)
2594		return -EINVAL;
2595
2596	/* Get association, if sack_assoc_id != 0 and the socket is a one
2597	 * to many style socket, and an association was not found, then
2598	 * the id was invalid.
2599	 */
2600	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2601	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2602		return -EINVAL;
2603
2604	if (params.sack_delay) {
2605		if (asoc) {
2606			asoc->sackdelay =
2607				msecs_to_jiffies(params.sack_delay);
2608			asoc->param_flags =
2609				(asoc->param_flags & ~SPP_SACKDELAY) |
2610				SPP_SACKDELAY_ENABLE;
2611		} else {
2612			sp->sackdelay = params.sack_delay;
2613			sp->param_flags =
2614				(sp->param_flags & ~SPP_SACKDELAY) |
2615				SPP_SACKDELAY_ENABLE;
2616		}
2617	}
2618
2619	if (params.sack_freq == 1) {
2620		if (asoc) {
2621			asoc->param_flags =
2622				(asoc->param_flags & ~SPP_SACKDELAY) |
2623				SPP_SACKDELAY_DISABLE;
2624		} else {
2625			sp->param_flags =
2626				(sp->param_flags & ~SPP_SACKDELAY) |
2627				SPP_SACKDELAY_DISABLE;
2628		}
2629	} else if (params.sack_freq > 1) {
2630		if (asoc) {
2631			asoc->sackfreq = params.sack_freq;
2632			asoc->param_flags =
2633				(asoc->param_flags & ~SPP_SACKDELAY) |
2634				SPP_SACKDELAY_ENABLE;
2635		} else {
2636			sp->sackfreq = params.sack_freq;
2637			sp->param_flags =
2638				(sp->param_flags & ~SPP_SACKDELAY) |
2639				SPP_SACKDELAY_ENABLE;
2640		}
2641	}
2642
2643	/* If change is for association, also apply to each transport. */
2644	if (asoc) {
2645		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2646				transports) {
2647			if (params.sack_delay) {
2648				trans->sackdelay =
2649					msecs_to_jiffies(params.sack_delay);
2650				trans->param_flags =
2651					(trans->param_flags & ~SPP_SACKDELAY) |
2652					SPP_SACKDELAY_ENABLE;
2653			}
2654			if (params.sack_freq == 1) {
2655				trans->param_flags =
2656					(trans->param_flags & ~SPP_SACKDELAY) |
2657					SPP_SACKDELAY_DISABLE;
2658			} else if (params.sack_freq > 1) {
2659				trans->sackfreq = params.sack_freq;
2660				trans->param_flags =
2661					(trans->param_flags & ~SPP_SACKDELAY) |
2662					SPP_SACKDELAY_ENABLE;
2663			}
2664		}
2665	}
2666
2667	return 0;
2668}
2669
2670/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2671 *
2672 * Applications can specify protocol parameters for the default association
2673 * initialization.  The option name argument to setsockopt() and getsockopt()
2674 * is SCTP_INITMSG.
2675 *
2676 * Setting initialization parameters is effective only on an unconnected
2677 * socket (for UDP-style sockets only future associations are effected
2678 * by the change).  With TCP-style sockets, this option is inherited by
2679 * sockets derived from a listener socket.
2680 */
2681static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2682{
2683	struct sctp_initmsg sinit;
2684	struct sctp_sock *sp = sctp_sk(sk);
2685
2686	if (optlen != sizeof(struct sctp_initmsg))
2687		return -EINVAL;
2688	if (copy_from_user(&sinit, optval, optlen))
2689		return -EFAULT;
2690
2691	if (sinit.sinit_num_ostreams)
2692		sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2693	if (sinit.sinit_max_instreams)
2694		sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2695	if (sinit.sinit_max_attempts)
2696		sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2697	if (sinit.sinit_max_init_timeo)
2698		sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2699
2700	return 0;
2701}
2702
2703/*
2704 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2705 *
2706 *   Applications that wish to use the sendto() system call may wish to
2707 *   specify a default set of parameters that would normally be supplied
2708 *   through the inclusion of ancillary data.  This socket option allows
2709 *   such an application to set the default sctp_sndrcvinfo structure.
2710 *   The application that wishes to use this socket option simply passes
2711 *   in to this call the sctp_sndrcvinfo structure defined in Section
2712 *   5.2.2) The input parameters accepted by this call include
2713 *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2714 *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2715 *   to this call if the caller is using the UDP model.
2716 */
2717static int sctp_setsockopt_default_send_param(struct sock *sk,
2718					      char __user *optval,
2719					      unsigned int optlen)
2720{
2721	struct sctp_sndrcvinfo info;
2722	struct sctp_association *asoc;
2723	struct sctp_sock *sp = sctp_sk(sk);
2724
2725	if (optlen != sizeof(struct sctp_sndrcvinfo))
2726		return -EINVAL;
2727	if (copy_from_user(&info, optval, optlen))
2728		return -EFAULT;
2729
2730	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2731	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2732		return -EINVAL;
2733
2734	if (asoc) {
2735		asoc->default_stream = info.sinfo_stream;
2736		asoc->default_flags = info.sinfo_flags;
2737		asoc->default_ppid = info.sinfo_ppid;
2738		asoc->default_context = info.sinfo_context;
2739		asoc->default_timetolive = info.sinfo_timetolive;
2740	} else {
2741		sp->default_stream = info.sinfo_stream;
2742		sp->default_flags = info.sinfo_flags;
2743		sp->default_ppid = info.sinfo_ppid;
2744		sp->default_context = info.sinfo_context;
2745		sp->default_timetolive = info.sinfo_timetolive;
2746	}
2747
2748	return 0;
2749}
2750
2751/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2752 *
2753 * Requests that the local SCTP stack use the enclosed peer address as
2754 * the association primary.  The enclosed address must be one of the
2755 * association peer's addresses.
2756 */
2757static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2758					unsigned int optlen)
2759{
2760	struct sctp_prim prim;
2761	struct sctp_transport *trans;
2762
2763	if (optlen != sizeof(struct sctp_prim))
2764		return -EINVAL;
2765
2766	if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2767		return -EFAULT;
2768
2769	trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2770	if (!trans)
2771		return -EINVAL;
2772
2773	sctp_assoc_set_primary(trans->asoc, trans);
2774
2775	return 0;
2776}
2777
2778/*
2779 * 7.1.5 SCTP_NODELAY
2780 *
2781 * Turn on/off any Nagle-like algorithm.  This means that packets are
2782 * generally sent as soon as possible and no unnecessary delays are
2783 * introduced, at the cost of more packets in the network.  Expects an
2784 *  integer boolean flag.
2785 */
2786static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2787				   unsigned int optlen)
2788{
2789	int val;
2790
2791	if (optlen < sizeof(int))
2792		return -EINVAL;
2793	if (get_user(val, (int __user *)optval))
2794		return -EFAULT;
2795
2796	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2797	return 0;
2798}
2799
2800/*
2801 *
2802 * 7.1.1 SCTP_RTOINFO
2803 *
2804 * The protocol parameters used to initialize and bound retransmission
2805 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2806 * and modify these parameters.
2807 * All parameters are time values, in milliseconds.  A value of 0, when
2808 * modifying the parameters, indicates that the current value should not
2809 * be changed.
2810 *
2811 */
2812static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2813{
2814	struct sctp_rtoinfo rtoinfo;
2815	struct sctp_association *asoc;
 
 
2816
2817	if (optlen != sizeof (struct sctp_rtoinfo))
2818		return -EINVAL;
2819
2820	if (copy_from_user(&rtoinfo, optval, optlen))
2821		return -EFAULT;
2822
2823	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2824
2825	/* Set the values to the specific association */
2826	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2827		return -EINVAL;
2828
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2829	if (asoc) {
2830		if (rtoinfo.srto_initial != 0)
2831			asoc->rto_initial =
2832				msecs_to_jiffies(rtoinfo.srto_initial);
2833		if (rtoinfo.srto_max != 0)
2834			asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2835		if (rtoinfo.srto_min != 0)
2836			asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2837	} else {
2838		/* If there is no association or the association-id = 0
2839		 * set the values to the endpoint.
2840		 */
2841		struct sctp_sock *sp = sctp_sk(sk);
2842
2843		if (rtoinfo.srto_initial != 0)
2844			sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2845		if (rtoinfo.srto_max != 0)
2846			sp->rtoinfo.srto_max = rtoinfo.srto_max;
2847		if (rtoinfo.srto_min != 0)
2848			sp->rtoinfo.srto_min = rtoinfo.srto_min;
2849	}
2850
2851	return 0;
2852}
2853
2854/*
2855 *
2856 * 7.1.2 SCTP_ASSOCINFO
2857 *
2858 * This option is used to tune the maximum retransmission attempts
2859 * of the association.
2860 * Returns an error if the new association retransmission value is
2861 * greater than the sum of the retransmission value  of the peer.
2862 * See [SCTP] for more information.
2863 *
2864 */
2865static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2866{
2867
2868	struct sctp_assocparams assocparams;
2869	struct sctp_association *asoc;
2870
2871	if (optlen != sizeof(struct sctp_assocparams))
2872		return -EINVAL;
2873	if (copy_from_user(&assocparams, optval, optlen))
2874		return -EFAULT;
2875
2876	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2877
2878	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2879		return -EINVAL;
2880
2881	/* Set the values to the specific association */
2882	if (asoc) {
2883		if (assocparams.sasoc_asocmaxrxt != 0) {
2884			__u32 path_sum = 0;
2885			int   paths = 0;
2886			struct sctp_transport *peer_addr;
2887
2888			list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2889					transports) {
2890				path_sum += peer_addr->pathmaxrxt;
2891				paths++;
2892			}
2893
2894			/* Only validate asocmaxrxt if we have more than
2895			 * one path/transport.  We do this because path
2896			 * retransmissions are only counted when we have more
2897			 * then one path.
2898			 */
2899			if (paths > 1 &&
2900			    assocparams.sasoc_asocmaxrxt > path_sum)
2901				return -EINVAL;
2902
2903			asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2904		}
2905
2906		if (assocparams.sasoc_cookie_life != 0) {
2907			asoc->cookie_life.tv_sec =
2908					assocparams.sasoc_cookie_life / 1000;
2909			asoc->cookie_life.tv_usec =
2910					(assocparams.sasoc_cookie_life % 1000)
2911					* 1000;
2912		}
2913	} else {
2914		/* Set the values to the endpoint */
2915		struct sctp_sock *sp = sctp_sk(sk);
2916
2917		if (assocparams.sasoc_asocmaxrxt != 0)
2918			sp->assocparams.sasoc_asocmaxrxt =
2919						assocparams.sasoc_asocmaxrxt;
2920		if (assocparams.sasoc_cookie_life != 0)
2921			sp->assocparams.sasoc_cookie_life =
2922						assocparams.sasoc_cookie_life;
2923	}
2924	return 0;
2925}
2926
2927/*
2928 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2929 *
2930 * This socket option is a boolean flag which turns on or off mapped V4
2931 * addresses.  If this option is turned on and the socket is type
2932 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2933 * If this option is turned off, then no mapping will be done of V4
2934 * addresses and a user will receive both PF_INET6 and PF_INET type
2935 * addresses on the socket.
2936 */
2937static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2938{
2939	int val;
2940	struct sctp_sock *sp = sctp_sk(sk);
2941
2942	if (optlen < sizeof(int))
2943		return -EINVAL;
2944	if (get_user(val, (int __user *)optval))
2945		return -EFAULT;
2946	if (val)
2947		sp->v4mapped = 1;
2948	else
2949		sp->v4mapped = 0;
2950
2951	return 0;
2952}
2953
2954/*
2955 * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2956 * This option will get or set the maximum size to put in any outgoing
2957 * SCTP DATA chunk.  If a message is larger than this size it will be
2958 * fragmented by SCTP into the specified size.  Note that the underlying
2959 * SCTP implementation may fragment into smaller sized chunks when the
2960 * PMTU of the underlying association is smaller than the value set by
2961 * the user.  The default value for this option is '0' which indicates
2962 * the user is NOT limiting fragmentation and only the PMTU will effect
2963 * SCTP's choice of DATA chunk size.  Note also that values set larger
2964 * than the maximum size of an IP datagram will effectively let SCTP
2965 * control fragmentation (i.e. the same as setting this option to 0).
2966 *
2967 * The following structure is used to access and modify this parameter:
2968 *
2969 * struct sctp_assoc_value {
2970 *   sctp_assoc_t assoc_id;
2971 *   uint32_t assoc_value;
2972 * };
2973 *
2974 * assoc_id:  This parameter is ignored for one-to-one style sockets.
2975 *    For one-to-many style sockets this parameter indicates which
2976 *    association the user is performing an action upon.  Note that if
2977 *    this field's value is zero then the endpoints default value is
2978 *    changed (effecting future associations only).
2979 * assoc_value:  This parameter specifies the maximum size in bytes.
2980 */
2981static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2982{
2983	struct sctp_assoc_value params;
2984	struct sctp_association *asoc;
2985	struct sctp_sock *sp = sctp_sk(sk);
2986	int val;
2987
2988	if (optlen == sizeof(int)) {
2989		pr_warn("Use of int in maxseg socket option deprecated\n");
2990		pr_warn("Use struct sctp_assoc_value instead\n");
 
 
 
2991		if (copy_from_user(&val, optval, optlen))
2992			return -EFAULT;
2993		params.assoc_id = 0;
2994	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2995		if (copy_from_user(&params, optval, optlen))
2996			return -EFAULT;
2997		val = params.assoc_value;
2998	} else
2999		return -EINVAL;
3000
3001	if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3002		return -EINVAL;
3003
3004	asoc = sctp_id2assoc(sk, params.assoc_id);
3005	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3006		return -EINVAL;
3007
3008	if (asoc) {
3009		if (val == 0) {
3010			val = asoc->pathmtu;
3011			val -= sp->pf->af->net_header_len;
3012			val -= sizeof(struct sctphdr) +
3013					sizeof(struct sctp_data_chunk);
3014		}
3015		asoc->user_frag = val;
3016		asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3017	} else {
3018		sp->user_frag = val;
3019	}
3020
3021	return 0;
3022}
3023
3024
3025/*
3026 *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3027 *
3028 *   Requests that the peer mark the enclosed address as the association
3029 *   primary. The enclosed address must be one of the association's
3030 *   locally bound addresses. The following structure is used to make a
3031 *   set primary request:
3032 */
3033static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3034					     unsigned int optlen)
3035{
 
3036	struct sctp_sock	*sp;
3037	struct sctp_association	*asoc = NULL;
3038	struct sctp_setpeerprim	prim;
3039	struct sctp_chunk	*chunk;
3040	struct sctp_af		*af;
3041	int 			err;
3042
3043	sp = sctp_sk(sk);
3044
3045	if (!sctp_addip_enable)
3046		return -EPERM;
3047
3048	if (optlen != sizeof(struct sctp_setpeerprim))
3049		return -EINVAL;
3050
3051	if (copy_from_user(&prim, optval, optlen))
3052		return -EFAULT;
3053
3054	asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3055	if (!asoc)
3056		return -EINVAL;
3057
3058	if (!asoc->peer.asconf_capable)
3059		return -EPERM;
3060
3061	if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3062		return -EPERM;
3063
3064	if (!sctp_state(asoc, ESTABLISHED))
3065		return -ENOTCONN;
3066
3067	af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3068	if (!af)
3069		return -EINVAL;
3070
3071	if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3072		return -EADDRNOTAVAIL;
3073
3074	if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3075		return -EADDRNOTAVAIL;
3076
3077	/* Create an ASCONF chunk with SET_PRIMARY parameter	*/
3078	chunk = sctp_make_asconf_set_prim(asoc,
3079					  (union sctp_addr *)&prim.sspp_addr);
3080	if (!chunk)
3081		return -ENOMEM;
3082
3083	err = sctp_send_asconf(asoc, chunk);
3084
3085	SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3086
3087	return err;
3088}
3089
3090static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3091					    unsigned int optlen)
3092{
3093	struct sctp_setadaptation adaptation;
3094
3095	if (optlen != sizeof(struct sctp_setadaptation))
3096		return -EINVAL;
3097	if (copy_from_user(&adaptation, optval, optlen))
3098		return -EFAULT;
3099
3100	sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3101
3102	return 0;
3103}
3104
3105/*
3106 * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
3107 *
3108 * The context field in the sctp_sndrcvinfo structure is normally only
3109 * used when a failed message is retrieved holding the value that was
3110 * sent down on the actual send call.  This option allows the setting of
3111 * a default context on an association basis that will be received on
3112 * reading messages from the peer.  This is especially helpful in the
3113 * one-2-many model for an application to keep some reference to an
3114 * internal state machine that is processing messages on the
3115 * association.  Note that the setting of this value only effects
3116 * received messages from the peer and does not effect the value that is
3117 * saved with outbound messages.
3118 */
3119static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3120				   unsigned int optlen)
3121{
3122	struct sctp_assoc_value params;
3123	struct sctp_sock *sp;
3124	struct sctp_association *asoc;
3125
3126	if (optlen != sizeof(struct sctp_assoc_value))
3127		return -EINVAL;
3128	if (copy_from_user(&params, optval, optlen))
3129		return -EFAULT;
3130
3131	sp = sctp_sk(sk);
3132
3133	if (params.assoc_id != 0) {
3134		asoc = sctp_id2assoc(sk, params.assoc_id);
3135		if (!asoc)
3136			return -EINVAL;
3137		asoc->default_rcv_context = params.assoc_value;
3138	} else {
3139		sp->default_rcv_context = params.assoc_value;
3140	}
3141
3142	return 0;
3143}
3144
3145/*
3146 * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3147 *
3148 * This options will at a minimum specify if the implementation is doing
3149 * fragmented interleave.  Fragmented interleave, for a one to many
3150 * socket, is when subsequent calls to receive a message may return
3151 * parts of messages from different associations.  Some implementations
3152 * may allow you to turn this value on or off.  If so, when turned off,
3153 * no fragment interleave will occur (which will cause a head of line
3154 * blocking amongst multiple associations sharing the same one to many
3155 * socket).  When this option is turned on, then each receive call may
3156 * come from a different association (thus the user must receive data
3157 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3158 * association each receive belongs to.
3159 *
3160 * This option takes a boolean value.  A non-zero value indicates that
3161 * fragmented interleave is on.  A value of zero indicates that
3162 * fragmented interleave is off.
3163 *
3164 * Note that it is important that an implementation that allows this
3165 * option to be turned on, have it off by default.  Otherwise an unaware
3166 * application using the one to many model may become confused and act
3167 * incorrectly.
3168 */
3169static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3170					       char __user *optval,
3171					       unsigned int optlen)
3172{
3173	int val;
3174
3175	if (optlen != sizeof(int))
3176		return -EINVAL;
3177	if (get_user(val, (int __user *)optval))
3178		return -EFAULT;
3179
3180	sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3181
3182	return 0;
3183}
3184
3185/*
3186 * 8.1.21.  Set or Get the SCTP Partial Delivery Point
3187 *       (SCTP_PARTIAL_DELIVERY_POINT)
3188 *
3189 * This option will set or get the SCTP partial delivery point.  This
3190 * point is the size of a message where the partial delivery API will be
3191 * invoked to help free up rwnd space for the peer.  Setting this to a
3192 * lower value will cause partial deliveries to happen more often.  The
3193 * calls argument is an integer that sets or gets the partial delivery
3194 * point.  Note also that the call will fail if the user attempts to set
3195 * this value larger than the socket receive buffer size.
3196 *
3197 * Note that any single message having a length smaller than or equal to
3198 * the SCTP partial delivery point will be delivered in one single read
3199 * call as long as the user provided buffer is large enough to hold the
3200 * message.
3201 */
3202static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3203						  char __user *optval,
3204						  unsigned int optlen)
3205{
3206	u32 val;
3207
3208	if (optlen != sizeof(u32))
3209		return -EINVAL;
3210	if (get_user(val, (int __user *)optval))
3211		return -EFAULT;
3212
3213	/* Note: We double the receive buffer from what the user sets
3214	 * it to be, also initial rwnd is based on rcvbuf/2.
3215	 */
3216	if (val > (sk->sk_rcvbuf >> 1))
3217		return -EINVAL;
3218
3219	sctp_sk(sk)->pd_point = val;
3220
3221	return 0; /* is this the right error code? */
3222}
3223
3224/*
3225 * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
3226 *
3227 * This option will allow a user to change the maximum burst of packets
3228 * that can be emitted by this association.  Note that the default value
3229 * is 4, and some implementations may restrict this setting so that it
3230 * can only be lowered.
3231 *
3232 * NOTE: This text doesn't seem right.  Do this on a socket basis with
3233 * future associations inheriting the socket value.
3234 */
3235static int sctp_setsockopt_maxburst(struct sock *sk,
3236				    char __user *optval,
3237				    unsigned int optlen)
3238{
3239	struct sctp_assoc_value params;
3240	struct sctp_sock *sp;
3241	struct sctp_association *asoc;
3242	int val;
3243	int assoc_id = 0;
3244
3245	if (optlen == sizeof(int)) {
3246		pr_warn("Use of int in max_burst socket option deprecated\n");
3247		pr_warn("Use struct sctp_assoc_value instead\n");
 
 
 
3248		if (copy_from_user(&val, optval, optlen))
3249			return -EFAULT;
3250	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3251		if (copy_from_user(&params, optval, optlen))
3252			return -EFAULT;
3253		val = params.assoc_value;
3254		assoc_id = params.assoc_id;
3255	} else
3256		return -EINVAL;
3257
3258	sp = sctp_sk(sk);
3259
3260	if (assoc_id != 0) {
3261		asoc = sctp_id2assoc(sk, assoc_id);
3262		if (!asoc)
3263			return -EINVAL;
3264		asoc->max_burst = val;
3265	} else
3266		sp->max_burst = val;
3267
3268	return 0;
3269}
3270
3271/*
3272 * 7.1.18.  Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3273 *
3274 * This set option adds a chunk type that the user is requesting to be
3275 * received only in an authenticated way.  Changes to the list of chunks
3276 * will only effect future associations on the socket.
3277 */
3278static int sctp_setsockopt_auth_chunk(struct sock *sk,
3279				      char __user *optval,
3280				      unsigned int optlen)
3281{
 
3282	struct sctp_authchunk val;
3283
3284	if (!sctp_auth_enable)
3285		return -EACCES;
3286
3287	if (optlen != sizeof(struct sctp_authchunk))
3288		return -EINVAL;
3289	if (copy_from_user(&val, optval, optlen))
3290		return -EFAULT;
3291
3292	switch (val.sauth_chunk) {
3293	case SCTP_CID_INIT:
3294	case SCTP_CID_INIT_ACK:
3295	case SCTP_CID_SHUTDOWN_COMPLETE:
3296	case SCTP_CID_AUTH:
3297		return -EINVAL;
3298	}
3299
3300	/* add this chunk id to the endpoint */
3301	return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3302}
3303
3304/*
3305 * 7.1.19.  Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3306 *
3307 * This option gets or sets the list of HMAC algorithms that the local
3308 * endpoint requires the peer to use.
3309 */
3310static int sctp_setsockopt_hmac_ident(struct sock *sk,
3311				      char __user *optval,
3312				      unsigned int optlen)
3313{
 
3314	struct sctp_hmacalgo *hmacs;
3315	u32 idents;
3316	int err;
3317
3318	if (!sctp_auth_enable)
3319		return -EACCES;
3320
3321	if (optlen < sizeof(struct sctp_hmacalgo))
3322		return -EINVAL;
3323
3324	hmacs= memdup_user(optval, optlen);
3325	if (IS_ERR(hmacs))
3326		return PTR_ERR(hmacs);
3327
3328	idents = hmacs->shmac_num_idents;
3329	if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3330	    (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3331		err = -EINVAL;
3332		goto out;
3333	}
3334
3335	err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3336out:
3337	kfree(hmacs);
3338	return err;
3339}
3340
3341/*
3342 * 7.1.20.  Set a shared key (SCTP_AUTH_KEY)
3343 *
3344 * This option will set a shared secret key which is used to build an
3345 * association shared key.
3346 */
3347static int sctp_setsockopt_auth_key(struct sock *sk,
3348				    char __user *optval,
3349				    unsigned int optlen)
3350{
 
3351	struct sctp_authkey *authkey;
3352	struct sctp_association *asoc;
3353	int ret;
3354
3355	if (!sctp_auth_enable)
3356		return -EACCES;
3357
3358	if (optlen <= sizeof(struct sctp_authkey))
3359		return -EINVAL;
3360
3361	authkey= memdup_user(optval, optlen);
3362	if (IS_ERR(authkey))
3363		return PTR_ERR(authkey);
3364
3365	if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3366		ret = -EINVAL;
3367		goto out;
3368	}
3369
3370	asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3371	if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3372		ret = -EINVAL;
3373		goto out;
3374	}
3375
3376	ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3377out:
3378	kfree(authkey);
3379	return ret;
3380}
3381
3382/*
3383 * 7.1.21.  Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3384 *
3385 * This option will get or set the active shared key to be used to build
3386 * the association shared key.
3387 */
3388static int sctp_setsockopt_active_key(struct sock *sk,
3389				      char __user *optval,
3390				      unsigned int optlen)
3391{
 
3392	struct sctp_authkeyid val;
3393	struct sctp_association *asoc;
3394
3395	if (!sctp_auth_enable)
3396		return -EACCES;
3397
3398	if (optlen != sizeof(struct sctp_authkeyid))
3399		return -EINVAL;
3400	if (copy_from_user(&val, optval, optlen))
3401		return -EFAULT;
3402
3403	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3404	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3405		return -EINVAL;
3406
3407	return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3408					val.scact_keynumber);
3409}
3410
3411/*
3412 * 7.1.22.  Delete a shared key (SCTP_AUTH_DELETE_KEY)
3413 *
3414 * This set option will delete a shared secret key from use.
3415 */
3416static int sctp_setsockopt_del_key(struct sock *sk,
3417				   char __user *optval,
3418				   unsigned int optlen)
3419{
 
3420	struct sctp_authkeyid val;
3421	struct sctp_association *asoc;
3422
3423	if (!sctp_auth_enable)
3424		return -EACCES;
3425
3426	if (optlen != sizeof(struct sctp_authkeyid))
3427		return -EINVAL;
3428	if (copy_from_user(&val, optval, optlen))
3429		return -EFAULT;
3430
3431	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3432	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3433		return -EINVAL;
3434
3435	return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3436				    val.scact_keynumber);
3437
3438}
3439
3440/*
3441 * 8.1.23 SCTP_AUTO_ASCONF
3442 *
3443 * This option will enable or disable the use of the automatic generation of
3444 * ASCONF chunks to add and delete addresses to an existing association.  Note
3445 * that this option has two caveats namely: a) it only affects sockets that
3446 * are bound to all addresses available to the SCTP stack, and b) the system
3447 * administrator may have an overriding control that turns the ASCONF feature
3448 * off no matter what setting the socket option may have.
3449 * This option expects an integer boolean flag, where a non-zero value turns on
3450 * the option, and a zero value turns off the option.
3451 * Note. In this implementation, socket operation overrides default parameter
3452 * being set by sysctl as well as FreeBSD implementation
3453 */
3454static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3455					unsigned int optlen)
3456{
3457	int val;
3458	struct sctp_sock *sp = sctp_sk(sk);
3459
3460	if (optlen < sizeof(int))
3461		return -EINVAL;
3462	if (get_user(val, (int __user *)optval))
3463		return -EFAULT;
3464	if (!sctp_is_ep_boundall(sk) && val)
3465		return -EINVAL;
3466	if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3467		return 0;
3468
3469	if (val == 0 && sp->do_auto_asconf) {
3470		list_del(&sp->auto_asconf_list);
3471		sp->do_auto_asconf = 0;
3472	} else if (val && !sp->do_auto_asconf) {
3473		list_add_tail(&sp->auto_asconf_list,
3474		    &sctp_auto_asconf_splist);
3475		sp->do_auto_asconf = 1;
3476	}
3477	return 0;
3478}
3479
3480
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3481/* API 6.2 setsockopt(), getsockopt()
3482 *
3483 * Applications use setsockopt() and getsockopt() to set or retrieve
3484 * socket options.  Socket options are used to change the default
3485 * behavior of sockets calls.  They are described in Section 7.
3486 *
3487 * The syntax is:
3488 *
3489 *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
3490 *                    int __user *optlen);
3491 *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3492 *                    int optlen);
3493 *
3494 *   sd      - the socket descript.
3495 *   level   - set to IPPROTO_SCTP for all SCTP options.
3496 *   optname - the option name.
3497 *   optval  - the buffer to store the value of the option.
3498 *   optlen  - the size of the buffer.
3499 */
3500SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3501				char __user *optval, unsigned int optlen)
3502{
3503	int retval = 0;
3504
3505	SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3506			  sk, optname);
3507
3508	/* I can hardly begin to describe how wrong this is.  This is
3509	 * so broken as to be worse than useless.  The API draft
3510	 * REALLY is NOT helpful here...  I am not convinced that the
3511	 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3512	 * are at all well-founded.
3513	 */
3514	if (level != SOL_SCTP) {
3515		struct sctp_af *af = sctp_sk(sk)->pf->af;
3516		retval = af->setsockopt(sk, level, optname, optval, optlen);
3517		goto out_nounlock;
3518	}
3519
3520	sctp_lock_sock(sk);
3521
3522	switch (optname) {
3523	case SCTP_SOCKOPT_BINDX_ADD:
3524		/* 'optlen' is the size of the addresses buffer. */
3525		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3526					       optlen, SCTP_BINDX_ADD_ADDR);
3527		break;
3528
3529	case SCTP_SOCKOPT_BINDX_REM:
3530		/* 'optlen' is the size of the addresses buffer. */
3531		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3532					       optlen, SCTP_BINDX_REM_ADDR);
3533		break;
3534
3535	case SCTP_SOCKOPT_CONNECTX_OLD:
3536		/* 'optlen' is the size of the addresses buffer. */
3537		retval = sctp_setsockopt_connectx_old(sk,
3538					    (struct sockaddr __user *)optval,
3539					    optlen);
3540		break;
3541
3542	case SCTP_SOCKOPT_CONNECTX:
3543		/* 'optlen' is the size of the addresses buffer. */
3544		retval = sctp_setsockopt_connectx(sk,
3545					    (struct sockaddr __user *)optval,
3546					    optlen);
3547		break;
3548
3549	case SCTP_DISABLE_FRAGMENTS:
3550		retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3551		break;
3552
3553	case SCTP_EVENTS:
3554		retval = sctp_setsockopt_events(sk, optval, optlen);
3555		break;
3556
3557	case SCTP_AUTOCLOSE:
3558		retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3559		break;
3560
3561	case SCTP_PEER_ADDR_PARAMS:
3562		retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3563		break;
3564
3565	case SCTP_DELAYED_SACK:
3566		retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3567		break;
3568	case SCTP_PARTIAL_DELIVERY_POINT:
3569		retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3570		break;
3571
3572	case SCTP_INITMSG:
3573		retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3574		break;
3575	case SCTP_DEFAULT_SEND_PARAM:
3576		retval = sctp_setsockopt_default_send_param(sk, optval,
3577							    optlen);
3578		break;
3579	case SCTP_PRIMARY_ADDR:
3580		retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3581		break;
3582	case SCTP_SET_PEER_PRIMARY_ADDR:
3583		retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3584		break;
3585	case SCTP_NODELAY:
3586		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3587		break;
3588	case SCTP_RTOINFO:
3589		retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3590		break;
3591	case SCTP_ASSOCINFO:
3592		retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3593		break;
3594	case SCTP_I_WANT_MAPPED_V4_ADDR:
3595		retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3596		break;
3597	case SCTP_MAXSEG:
3598		retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3599		break;
3600	case SCTP_ADAPTATION_LAYER:
3601		retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3602		break;
3603	case SCTP_CONTEXT:
3604		retval = sctp_setsockopt_context(sk, optval, optlen);
3605		break;
3606	case SCTP_FRAGMENT_INTERLEAVE:
3607		retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3608		break;
3609	case SCTP_MAX_BURST:
3610		retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3611		break;
3612	case SCTP_AUTH_CHUNK:
3613		retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3614		break;
3615	case SCTP_HMAC_IDENT:
3616		retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3617		break;
3618	case SCTP_AUTH_KEY:
3619		retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3620		break;
3621	case SCTP_AUTH_ACTIVE_KEY:
3622		retval = sctp_setsockopt_active_key(sk, optval, optlen);
3623		break;
3624	case SCTP_AUTH_DELETE_KEY:
3625		retval = sctp_setsockopt_del_key(sk, optval, optlen);
3626		break;
3627	case SCTP_AUTO_ASCONF:
3628		retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3629		break;
 
 
 
3630	default:
3631		retval = -ENOPROTOOPT;
3632		break;
3633	}
3634
3635	sctp_release_sock(sk);
3636
3637out_nounlock:
3638	return retval;
3639}
3640
3641/* API 3.1.6 connect() - UDP Style Syntax
3642 *
3643 * An application may use the connect() call in the UDP model to initiate an
3644 * association without sending data.
3645 *
3646 * The syntax is:
3647 *
3648 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3649 *
3650 * sd: the socket descriptor to have a new association added to.
3651 *
3652 * nam: the address structure (either struct sockaddr_in or struct
3653 *    sockaddr_in6 defined in RFC2553 [7]).
3654 *
3655 * len: the size of the address.
3656 */
3657SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3658			     int addr_len)
3659{
3660	int err = 0;
3661	struct sctp_af *af;
3662
3663	sctp_lock_sock(sk);
3664
3665	SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3666			  __func__, sk, addr, addr_len);
3667
3668	/* Validate addr_len before calling common connect/connectx routine. */
3669	af = sctp_get_af_specific(addr->sa_family);
3670	if (!af || addr_len < af->sockaddr_len) {
3671		err = -EINVAL;
3672	} else {
3673		/* Pass correct addr len to common routine (so it knows there
3674		 * is only one address being passed.
3675		 */
3676		err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3677	}
3678
3679	sctp_release_sock(sk);
3680	return err;
3681}
3682
3683/* FIXME: Write comments. */
3684SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3685{
3686	return -EOPNOTSUPP; /* STUB */
3687}
3688
3689/* 4.1.4 accept() - TCP Style Syntax
3690 *
3691 * Applications use accept() call to remove an established SCTP
3692 * association from the accept queue of the endpoint.  A new socket
3693 * descriptor will be returned from accept() to represent the newly
3694 * formed association.
3695 */
3696SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3697{
3698	struct sctp_sock *sp;
3699	struct sctp_endpoint *ep;
3700	struct sock *newsk = NULL;
3701	struct sctp_association *asoc;
3702	long timeo;
3703	int error = 0;
3704
3705	sctp_lock_sock(sk);
3706
3707	sp = sctp_sk(sk);
3708	ep = sp->ep;
3709
3710	if (!sctp_style(sk, TCP)) {
3711		error = -EOPNOTSUPP;
3712		goto out;
3713	}
3714
3715	if (!sctp_sstate(sk, LISTENING)) {
3716		error = -EINVAL;
3717		goto out;
3718	}
3719
3720	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3721
3722	error = sctp_wait_for_accept(sk, timeo);
3723	if (error)
3724		goto out;
3725
3726	/* We treat the list of associations on the endpoint as the accept
3727	 * queue and pick the first association on the list.
3728	 */
3729	asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3730
3731	newsk = sp->pf->create_accept_sk(sk, asoc);
3732	if (!newsk) {
3733		error = -ENOMEM;
3734		goto out;
3735	}
3736
3737	/* Populate the fields of the newsk from the oldsk and migrate the
3738	 * asoc to the newsk.
3739	 */
3740	sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3741
3742out:
3743	sctp_release_sock(sk);
3744	*err = error;
3745	return newsk;
3746}
3747
3748/* The SCTP ioctl handler. */
3749SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3750{
3751	int rc = -ENOTCONN;
3752
3753	sctp_lock_sock(sk);
3754
3755	/*
3756	 * SEQPACKET-style sockets in LISTENING state are valid, for
3757	 * SCTP, so only discard TCP-style sockets in LISTENING state.
3758	 */
3759	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3760		goto out;
3761
3762	switch (cmd) {
3763	case SIOCINQ: {
3764		struct sk_buff *skb;
3765		unsigned int amount = 0;
3766
3767		skb = skb_peek(&sk->sk_receive_queue);
3768		if (skb != NULL) {
3769			/*
3770			 * We will only return the amount of this packet since
3771			 * that is all that will be read.
3772			 */
3773			amount = skb->len;
3774		}
3775		rc = put_user(amount, (int __user *)arg);
3776		break;
3777	}
3778	default:
3779		rc = -ENOIOCTLCMD;
3780		break;
3781	}
3782out:
3783	sctp_release_sock(sk);
3784	return rc;
3785}
3786
3787/* This is the function which gets called during socket creation to
3788 * initialized the SCTP-specific portion of the sock.
3789 * The sock structure should already be zero-filled memory.
3790 */
3791SCTP_STATIC int sctp_init_sock(struct sock *sk)
3792{
3793	struct sctp_endpoint *ep;
3794	struct sctp_sock *sp;
3795
3796	SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3797
3798	sp = sctp_sk(sk);
3799
3800	/* Initialize the SCTP per socket area.  */
3801	switch (sk->sk_type) {
3802	case SOCK_SEQPACKET:
3803		sp->type = SCTP_SOCKET_UDP;
3804		break;
3805	case SOCK_STREAM:
3806		sp->type = SCTP_SOCKET_TCP;
3807		break;
3808	default:
3809		return -ESOCKTNOSUPPORT;
3810	}
3811
3812	/* Initialize default send parameters. These parameters can be
3813	 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3814	 */
3815	sp->default_stream = 0;
3816	sp->default_ppid = 0;
3817	sp->default_flags = 0;
3818	sp->default_context = 0;
3819	sp->default_timetolive = 0;
3820
3821	sp->default_rcv_context = 0;
3822	sp->max_burst = sctp_max_burst;
 
 
3823
3824	/* Initialize default setup parameters. These parameters
3825	 * can be modified with the SCTP_INITMSG socket option or
3826	 * overridden by the SCTP_INIT CMSG.
3827	 */
3828	sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
3829	sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
3830	sp->initmsg.sinit_max_attempts   = sctp_max_retrans_init;
3831	sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3832
3833	/* Initialize default RTO related parameters.  These parameters can
3834	 * be modified for with the SCTP_RTOINFO socket option.
3835	 */
3836	sp->rtoinfo.srto_initial = sctp_rto_initial;
3837	sp->rtoinfo.srto_max     = sctp_rto_max;
3838	sp->rtoinfo.srto_min     = sctp_rto_min;
3839
3840	/* Initialize default association related parameters. These parameters
3841	 * can be modified with the SCTP_ASSOCINFO socket option.
3842	 */
3843	sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3844	sp->assocparams.sasoc_number_peer_destinations = 0;
3845	sp->assocparams.sasoc_peer_rwnd = 0;
3846	sp->assocparams.sasoc_local_rwnd = 0;
3847	sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3848
3849	/* Initialize default event subscriptions. By default, all the
3850	 * options are off.
3851	 */
3852	memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3853
3854	/* Default Peer Address Parameters.  These defaults can
3855	 * be modified via SCTP_PEER_ADDR_PARAMS
3856	 */
3857	sp->hbinterval  = sctp_hb_interval;
3858	sp->pathmaxrxt  = sctp_max_retrans_path;
3859	sp->pathmtu     = 0; // allow default discovery
3860	sp->sackdelay   = sctp_sack_timeout;
3861	sp->sackfreq	= 2;
3862	sp->param_flags = SPP_HB_ENABLE |
3863			  SPP_PMTUD_ENABLE |
3864			  SPP_SACKDELAY_ENABLE;
3865
3866	/* If enabled no SCTP message fragmentation will be performed.
3867	 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3868	 */
3869	sp->disable_fragments = 0;
3870
3871	/* Enable Nagle algorithm by default.  */
3872	sp->nodelay           = 0;
3873
3874	/* Enable by default. */
3875	sp->v4mapped          = 1;
3876
3877	/* Auto-close idle associations after the configured
3878	 * number of seconds.  A value of 0 disables this
3879	 * feature.  Configure through the SCTP_AUTOCLOSE socket option,
3880	 * for UDP-style sockets only.
3881	 */
3882	sp->autoclose         = 0;
3883
3884	/* User specified fragmentation limit. */
3885	sp->user_frag         = 0;
3886
3887	sp->adaptation_ind = 0;
3888
3889	sp->pf = sctp_get_pf_specific(sk->sk_family);
3890
3891	/* Control variables for partial data delivery. */
3892	atomic_set(&sp->pd_mode, 0);
3893	skb_queue_head_init(&sp->pd_lobby);
3894	sp->frag_interleave = 0;
3895
3896	/* Create a per socket endpoint structure.  Even if we
3897	 * change the data structure relationships, this may still
3898	 * be useful for storing pre-connect address information.
3899	 */
3900	ep = sctp_endpoint_new(sk, GFP_KERNEL);
3901	if (!ep)
3902		return -ENOMEM;
3903
3904	sp->ep = ep;
3905	sp->hmac = NULL;
3906
 
 
3907	SCTP_DBG_OBJCNT_INC(sock);
3908
3909	local_bh_disable();
3910	percpu_counter_inc(&sctp_sockets_allocated);
3911	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3912	if (sctp_default_auto_asconf) {
3913		list_add_tail(&sp->auto_asconf_list,
3914		    &sctp_auto_asconf_splist);
3915		sp->do_auto_asconf = 1;
3916	} else
3917		sp->do_auto_asconf = 0;
3918	local_bh_enable();
3919
3920	return 0;
3921}
3922
3923/* Cleanup any SCTP per socket resources.  */
3924SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3925{
3926	struct sctp_sock *sp;
3927
3928	SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3929
3930	/* Release our hold on the endpoint. */
3931	sp = sctp_sk(sk);
 
 
 
 
 
 
3932	if (sp->do_auto_asconf) {
3933		sp->do_auto_asconf = 0;
3934		list_del(&sp->auto_asconf_list);
3935	}
3936	sctp_endpoint_free(sp->ep);
3937	local_bh_disable();
3938	percpu_counter_dec(&sctp_sockets_allocated);
3939	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3940	local_bh_enable();
3941}
3942
 
 
 
 
 
 
 
 
 
 
 
3943/* API 4.1.7 shutdown() - TCP Style Syntax
3944 *     int shutdown(int socket, int how);
3945 *
3946 *     sd      - the socket descriptor of the association to be closed.
3947 *     how     - Specifies the type of shutdown.  The  values  are
3948 *               as follows:
3949 *               SHUT_RD
3950 *                     Disables further receive operations. No SCTP
3951 *                     protocol action is taken.
3952 *               SHUT_WR
3953 *                     Disables further send operations, and initiates
3954 *                     the SCTP shutdown sequence.
3955 *               SHUT_RDWR
3956 *                     Disables further send  and  receive  operations
3957 *                     and initiates the SCTP shutdown sequence.
3958 */
3959SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3960{
 
3961	struct sctp_endpoint *ep;
3962	struct sctp_association *asoc;
3963
3964	if (!sctp_style(sk, TCP))
3965		return;
3966
3967	if (how & SEND_SHUTDOWN) {
3968		ep = sctp_sk(sk)->ep;
3969		if (!list_empty(&ep->asocs)) {
3970			asoc = list_entry(ep->asocs.next,
3971					  struct sctp_association, asocs);
3972			sctp_primitive_SHUTDOWN(asoc, NULL);
3973		}
3974	}
3975}
3976
3977/* 7.2.1 Association Status (SCTP_STATUS)
3978
3979 * Applications can retrieve current status information about an
3980 * association, including association state, peer receiver window size,
3981 * number of unacked data chunks, and number of data chunks pending
3982 * receipt.  This information is read-only.
3983 */
3984static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3985				       char __user *optval,
3986				       int __user *optlen)
3987{
3988	struct sctp_status status;
3989	struct sctp_association *asoc = NULL;
3990	struct sctp_transport *transport;
3991	sctp_assoc_t associd;
3992	int retval = 0;
3993
3994	if (len < sizeof(status)) {
3995		retval = -EINVAL;
3996		goto out;
3997	}
3998
3999	len = sizeof(status);
4000	if (copy_from_user(&status, optval, len)) {
4001		retval = -EFAULT;
4002		goto out;
4003	}
4004
4005	associd = status.sstat_assoc_id;
4006	asoc = sctp_id2assoc(sk, associd);
4007	if (!asoc) {
4008		retval = -EINVAL;
4009		goto out;
4010	}
4011
4012	transport = asoc->peer.primary_path;
4013
4014	status.sstat_assoc_id = sctp_assoc2id(asoc);
4015	status.sstat_state = asoc->state;
4016	status.sstat_rwnd =  asoc->peer.rwnd;
4017	status.sstat_unackdata = asoc->unack_data;
4018
4019	status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4020	status.sstat_instrms = asoc->c.sinit_max_instreams;
4021	status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4022	status.sstat_fragmentation_point = asoc->frag_point;
4023	status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4024	memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4025			transport->af_specific->sockaddr_len);
4026	/* Map ipv4 address into v4-mapped-on-v6 address.  */
4027	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4028		(union sctp_addr *)&status.sstat_primary.spinfo_address);
4029	status.sstat_primary.spinfo_state = transport->state;
4030	status.sstat_primary.spinfo_cwnd = transport->cwnd;
4031	status.sstat_primary.spinfo_srtt = transport->srtt;
4032	status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4033	status.sstat_primary.spinfo_mtu = transport->pathmtu;
4034
4035	if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4036		status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4037
4038	if (put_user(len, optlen)) {
4039		retval = -EFAULT;
4040		goto out;
4041	}
4042
4043	SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4044			  len, status.sstat_state, status.sstat_rwnd,
4045			  status.sstat_assoc_id);
4046
4047	if (copy_to_user(optval, &status, len)) {
4048		retval = -EFAULT;
4049		goto out;
4050	}
4051
4052out:
4053	return retval;
4054}
4055
4056
4057/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4058 *
4059 * Applications can retrieve information about a specific peer address
4060 * of an association, including its reachability state, congestion
4061 * window, and retransmission timer values.  This information is
4062 * read-only.
4063 */
4064static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4065					  char __user *optval,
4066					  int __user *optlen)
4067{
4068	struct sctp_paddrinfo pinfo;
4069	struct sctp_transport *transport;
4070	int retval = 0;
4071
4072	if (len < sizeof(pinfo)) {
4073		retval = -EINVAL;
4074		goto out;
4075	}
4076
4077	len = sizeof(pinfo);
4078	if (copy_from_user(&pinfo, optval, len)) {
4079		retval = -EFAULT;
4080		goto out;
4081	}
4082
4083	transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4084					   pinfo.spinfo_assoc_id);
4085	if (!transport)
4086		return -EINVAL;
4087
4088	pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4089	pinfo.spinfo_state = transport->state;
4090	pinfo.spinfo_cwnd = transport->cwnd;
4091	pinfo.spinfo_srtt = transport->srtt;
4092	pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4093	pinfo.spinfo_mtu = transport->pathmtu;
4094
4095	if (pinfo.spinfo_state == SCTP_UNKNOWN)
4096		pinfo.spinfo_state = SCTP_ACTIVE;
4097
4098	if (put_user(len, optlen)) {
4099		retval = -EFAULT;
4100		goto out;
4101	}
4102
4103	if (copy_to_user(optval, &pinfo, len)) {
4104		retval = -EFAULT;
4105		goto out;
4106	}
4107
4108out:
4109	return retval;
4110}
4111
4112/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4113 *
4114 * This option is a on/off flag.  If enabled no SCTP message
4115 * fragmentation will be performed.  Instead if a message being sent
4116 * exceeds the current PMTU size, the message will NOT be sent and
4117 * instead a error will be indicated to the user.
4118 */
4119static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4120					char __user *optval, int __user *optlen)
4121{
4122	int val;
4123
4124	if (len < sizeof(int))
4125		return -EINVAL;
4126
4127	len = sizeof(int);
4128	val = (sctp_sk(sk)->disable_fragments == 1);
4129	if (put_user(len, optlen))
4130		return -EFAULT;
4131	if (copy_to_user(optval, &val, len))
4132		return -EFAULT;
4133	return 0;
4134}
4135
4136/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4137 *
4138 * This socket option is used to specify various notifications and
4139 * ancillary data the user wishes to receive.
4140 */
4141static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4142				  int __user *optlen)
4143{
4144	if (len <= 0)
4145		return -EINVAL;
4146	if (len > sizeof(struct sctp_event_subscribe))
4147		len = sizeof(struct sctp_event_subscribe);
4148	if (put_user(len, optlen))
4149		return -EFAULT;
4150	if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4151		return -EFAULT;
4152	return 0;
4153}
4154
4155/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4156 *
4157 * This socket option is applicable to the UDP-style socket only.  When
4158 * set it will cause associations that are idle for more than the
4159 * specified number of seconds to automatically close.  An association
4160 * being idle is defined an association that has NOT sent or received
4161 * user data.  The special value of '0' indicates that no automatic
4162 * close of any associations should be performed.  The option expects an
4163 * integer defining the number of seconds of idle time before an
4164 * association is closed.
4165 */
4166static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4167{
4168	/* Applicable to UDP-style socket only */
4169	if (sctp_style(sk, TCP))
4170		return -EOPNOTSUPP;
4171	if (len < sizeof(int))
4172		return -EINVAL;
4173	len = sizeof(int);
4174	if (put_user(len, optlen))
4175		return -EFAULT;
4176	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4177		return -EFAULT;
4178	return 0;
4179}
4180
4181/* Helper routine to branch off an association to a new socket.  */
4182int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4183{
4184	struct sctp_association *asoc = sctp_id2assoc(sk, id);
4185	struct socket *sock;
4186	struct sctp_af *af;
4187	int err = 0;
4188
4189	if (!asoc)
4190		return -EINVAL;
4191
4192	/* An association cannot be branched off from an already peeled-off
4193	 * socket, nor is this supported for tcp style sockets.
4194	 */
4195	if (!sctp_style(sk, UDP))
4196		return -EINVAL;
4197
4198	/* Create a new socket.  */
4199	err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4200	if (err < 0)
4201		return err;
4202
4203	sctp_copy_sock(sock->sk, sk, asoc);
4204
4205	/* Make peeled-off sockets more like 1-1 accepted sockets.
4206	 * Set the daddr and initialize id to something more random
4207	 */
4208	af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4209	af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4210
4211	/* Populate the fields of the newsk from the oldsk and migrate the
4212	 * asoc to the newsk.
4213	 */
4214	sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4215
4216	*sockp = sock;
4217
4218	return err;
4219}
4220EXPORT_SYMBOL(sctp_do_peeloff);
4221
4222static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4223{
4224	sctp_peeloff_arg_t peeloff;
4225	struct socket *newsock;
 
4226	int retval = 0;
4227
4228	if (len < sizeof(sctp_peeloff_arg_t))
4229		return -EINVAL;
4230	len = sizeof(sctp_peeloff_arg_t);
4231	if (copy_from_user(&peeloff, optval, len))
4232		return -EFAULT;
4233
4234	retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4235	if (retval < 0)
4236		goto out;
4237
4238	/* Map the socket to an unused fd that can be returned to the user.  */
4239	retval = sock_map_fd(newsock, 0);
4240	if (retval < 0) {
4241		sock_release(newsock);
4242		goto out;
4243	}
4244
4245	SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4246			  __func__, sk, newsock->sk, retval);
 
 
 
 
 
 
 
4247
4248	/* Return the fd mapped to the new socket.  */
 
 
 
 
 
4249	peeloff.sd = retval;
4250	if (put_user(len, optlen))
 
 
4251		return -EFAULT;
4252	if (copy_to_user(optval, &peeloff, len))
4253		retval = -EFAULT;
4254
4255out:
4256	return retval;
4257}
4258
4259/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4260 *
4261 * Applications can enable or disable heartbeats for any peer address of
4262 * an association, modify an address's heartbeat interval, force a
4263 * heartbeat to be sent immediately, and adjust the address's maximum
4264 * number of retransmissions sent before an address is considered
4265 * unreachable.  The following structure is used to access and modify an
4266 * address's parameters:
4267 *
4268 *  struct sctp_paddrparams {
4269 *     sctp_assoc_t            spp_assoc_id;
4270 *     struct sockaddr_storage spp_address;
4271 *     uint32_t                spp_hbinterval;
4272 *     uint16_t                spp_pathmaxrxt;
4273 *     uint32_t                spp_pathmtu;
4274 *     uint32_t                spp_sackdelay;
4275 *     uint32_t                spp_flags;
4276 * };
4277 *
4278 *   spp_assoc_id    - (one-to-many style socket) This is filled in the
4279 *                     application, and identifies the association for
4280 *                     this query.
4281 *   spp_address     - This specifies which address is of interest.
4282 *   spp_hbinterval  - This contains the value of the heartbeat interval,
4283 *                     in milliseconds.  If a  value of zero
4284 *                     is present in this field then no changes are to
4285 *                     be made to this parameter.
4286 *   spp_pathmaxrxt  - This contains the maximum number of
4287 *                     retransmissions before this address shall be
4288 *                     considered unreachable. If a  value of zero
4289 *                     is present in this field then no changes are to
4290 *                     be made to this parameter.
4291 *   spp_pathmtu     - When Path MTU discovery is disabled the value
4292 *                     specified here will be the "fixed" path mtu.
4293 *                     Note that if the spp_address field is empty
4294 *                     then all associations on this address will
4295 *                     have this fixed path mtu set upon them.
4296 *
4297 *   spp_sackdelay   - When delayed sack is enabled, this value specifies
4298 *                     the number of milliseconds that sacks will be delayed
4299 *                     for. This value will apply to all addresses of an
4300 *                     association if the spp_address field is empty. Note
4301 *                     also, that if delayed sack is enabled and this
4302 *                     value is set to 0, no change is made to the last
4303 *                     recorded delayed sack timer value.
4304 *
4305 *   spp_flags       - These flags are used to control various features
4306 *                     on an association. The flag field may contain
4307 *                     zero or more of the following options.
4308 *
4309 *                     SPP_HB_ENABLE  - Enable heartbeats on the
4310 *                     specified address. Note that if the address
4311 *                     field is empty all addresses for the association
4312 *                     have heartbeats enabled upon them.
4313 *
4314 *                     SPP_HB_DISABLE - Disable heartbeats on the
4315 *                     speicifed address. Note that if the address
4316 *                     field is empty all addresses for the association
4317 *                     will have their heartbeats disabled. Note also
4318 *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
4319 *                     mutually exclusive, only one of these two should
4320 *                     be specified. Enabling both fields will have
4321 *                     undetermined results.
4322 *
4323 *                     SPP_HB_DEMAND - Request a user initiated heartbeat
4324 *                     to be made immediately.
4325 *
4326 *                     SPP_PMTUD_ENABLE - This field will enable PMTU
4327 *                     discovery upon the specified address. Note that
4328 *                     if the address feild is empty then all addresses
4329 *                     on the association are effected.
4330 *
4331 *                     SPP_PMTUD_DISABLE - This field will disable PMTU
4332 *                     discovery upon the specified address. Note that
4333 *                     if the address feild is empty then all addresses
4334 *                     on the association are effected. Not also that
4335 *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4336 *                     exclusive. Enabling both will have undetermined
4337 *                     results.
4338 *
4339 *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
4340 *                     on delayed sack. The time specified in spp_sackdelay
4341 *                     is used to specify the sack delay for this address. Note
4342 *                     that if spp_address is empty then all addresses will
4343 *                     enable delayed sack and take on the sack delay
4344 *                     value specified in spp_sackdelay.
4345 *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
4346 *                     off delayed sack. If the spp_address field is blank then
4347 *                     delayed sack is disabled for the entire association. Note
4348 *                     also that this field is mutually exclusive to
4349 *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
4350 *                     results.
4351 */
4352static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4353					    char __user *optval, int __user *optlen)
4354{
4355	struct sctp_paddrparams  params;
4356	struct sctp_transport   *trans = NULL;
4357	struct sctp_association *asoc = NULL;
4358	struct sctp_sock        *sp = sctp_sk(sk);
4359
4360	if (len < sizeof(struct sctp_paddrparams))
4361		return -EINVAL;
4362	len = sizeof(struct sctp_paddrparams);
4363	if (copy_from_user(&params, optval, len))
4364		return -EFAULT;
4365
4366	/* If an address other than INADDR_ANY is specified, and
4367	 * no transport is found, then the request is invalid.
4368	 */
4369	if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4370		trans = sctp_addr_id2transport(sk, &params.spp_address,
4371					       params.spp_assoc_id);
4372		if (!trans) {
4373			SCTP_DEBUG_PRINTK("Failed no transport\n");
4374			return -EINVAL;
4375		}
4376	}
4377
4378	/* Get association, if assoc_id != 0 and the socket is a one
4379	 * to many style socket, and an association was not found, then
4380	 * the id was invalid.
4381	 */
4382	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4383	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4384		SCTP_DEBUG_PRINTK("Failed no association\n");
4385		return -EINVAL;
4386	}
4387
4388	if (trans) {
4389		/* Fetch transport values. */
4390		params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4391		params.spp_pathmtu    = trans->pathmtu;
4392		params.spp_pathmaxrxt = trans->pathmaxrxt;
4393		params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
4394
4395		/*draft-11 doesn't say what to return in spp_flags*/
4396		params.spp_flags      = trans->param_flags;
4397	} else if (asoc) {
4398		/* Fetch association values. */
4399		params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4400		params.spp_pathmtu    = asoc->pathmtu;
4401		params.spp_pathmaxrxt = asoc->pathmaxrxt;
4402		params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
4403
4404		/*draft-11 doesn't say what to return in spp_flags*/
4405		params.spp_flags      = asoc->param_flags;
4406	} else {
4407		/* Fetch socket values. */
4408		params.spp_hbinterval = sp->hbinterval;
4409		params.spp_pathmtu    = sp->pathmtu;
4410		params.spp_sackdelay  = sp->sackdelay;
4411		params.spp_pathmaxrxt = sp->pathmaxrxt;
4412
4413		/*draft-11 doesn't say what to return in spp_flags*/
4414		params.spp_flags      = sp->param_flags;
4415	}
4416
4417	if (copy_to_user(optval, &params, len))
4418		return -EFAULT;
4419
4420	if (put_user(len, optlen))
4421		return -EFAULT;
4422
4423	return 0;
4424}
4425
4426/*
4427 * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
4428 *
4429 * This option will effect the way delayed acks are performed.  This
4430 * option allows you to get or set the delayed ack time, in
4431 * milliseconds.  It also allows changing the delayed ack frequency.
4432 * Changing the frequency to 1 disables the delayed sack algorithm.  If
4433 * the assoc_id is 0, then this sets or gets the endpoints default
4434 * values.  If the assoc_id field is non-zero, then the set or get
4435 * effects the specified association for the one to many model (the
4436 * assoc_id field is ignored by the one to one model).  Note that if
4437 * sack_delay or sack_freq are 0 when setting this option, then the
4438 * current values will remain unchanged.
4439 *
4440 * struct sctp_sack_info {
4441 *     sctp_assoc_t            sack_assoc_id;
4442 *     uint32_t                sack_delay;
4443 *     uint32_t                sack_freq;
4444 * };
4445 *
4446 * sack_assoc_id -  This parameter, indicates which association the user
4447 *    is performing an action upon.  Note that if this field's value is
4448 *    zero then the endpoints default value is changed (effecting future
4449 *    associations only).
4450 *
4451 * sack_delay -  This parameter contains the number of milliseconds that
4452 *    the user is requesting the delayed ACK timer be set to.  Note that
4453 *    this value is defined in the standard to be between 200 and 500
4454 *    milliseconds.
4455 *
4456 * sack_freq -  This parameter contains the number of packets that must
4457 *    be received before a sack is sent without waiting for the delay
4458 *    timer to expire.  The default value for this is 2, setting this
4459 *    value to 1 will disable the delayed sack algorithm.
4460 */
4461static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4462					    char __user *optval,
4463					    int __user *optlen)
4464{
4465	struct sctp_sack_info    params;
4466	struct sctp_association *asoc = NULL;
4467	struct sctp_sock        *sp = sctp_sk(sk);
4468
4469	if (len >= sizeof(struct sctp_sack_info)) {
4470		len = sizeof(struct sctp_sack_info);
4471
4472		if (copy_from_user(&params, optval, len))
4473			return -EFAULT;
4474	} else if (len == sizeof(struct sctp_assoc_value)) {
4475		pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4476		pr_warn("Use struct sctp_sack_info instead\n");
 
 
 
4477		if (copy_from_user(&params, optval, len))
4478			return -EFAULT;
4479	} else
4480		return - EINVAL;
4481
4482	/* Get association, if sack_assoc_id != 0 and the socket is a one
4483	 * to many style socket, and an association was not found, then
4484	 * the id was invalid.
4485	 */
4486	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4487	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4488		return -EINVAL;
4489
4490	if (asoc) {
4491		/* Fetch association values. */
4492		if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4493			params.sack_delay = jiffies_to_msecs(
4494				asoc->sackdelay);
4495			params.sack_freq = asoc->sackfreq;
4496
4497		} else {
4498			params.sack_delay = 0;
4499			params.sack_freq = 1;
4500		}
4501	} else {
4502		/* Fetch socket values. */
4503		if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4504			params.sack_delay  = sp->sackdelay;
4505			params.sack_freq = sp->sackfreq;
4506		} else {
4507			params.sack_delay  = 0;
4508			params.sack_freq = 1;
4509		}
4510	}
4511
4512	if (copy_to_user(optval, &params, len))
4513		return -EFAULT;
4514
4515	if (put_user(len, optlen))
4516		return -EFAULT;
4517
4518	return 0;
4519}
4520
4521/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4522 *
4523 * Applications can specify protocol parameters for the default association
4524 * initialization.  The option name argument to setsockopt() and getsockopt()
4525 * is SCTP_INITMSG.
4526 *
4527 * Setting initialization parameters is effective only on an unconnected
4528 * socket (for UDP-style sockets only future associations are effected
4529 * by the change).  With TCP-style sockets, this option is inherited by
4530 * sockets derived from a listener socket.
4531 */
4532static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4533{
4534	if (len < sizeof(struct sctp_initmsg))
4535		return -EINVAL;
4536	len = sizeof(struct sctp_initmsg);
4537	if (put_user(len, optlen))
4538		return -EFAULT;
4539	if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4540		return -EFAULT;
4541	return 0;
4542}
4543
4544
4545static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4546				      char __user *optval, int __user *optlen)
4547{
4548	struct sctp_association *asoc;
4549	int cnt = 0;
4550	struct sctp_getaddrs getaddrs;
4551	struct sctp_transport *from;
4552	void __user *to;
4553	union sctp_addr temp;
4554	struct sctp_sock *sp = sctp_sk(sk);
4555	int addrlen;
4556	size_t space_left;
4557	int bytes_copied;
4558
4559	if (len < sizeof(struct sctp_getaddrs))
4560		return -EINVAL;
4561
4562	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4563		return -EFAULT;
4564
4565	/* For UDP-style sockets, id specifies the association to query.  */
4566	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4567	if (!asoc)
4568		return -EINVAL;
4569
4570	to = optval + offsetof(struct sctp_getaddrs,addrs);
4571	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4572
4573	list_for_each_entry(from, &asoc->peer.transport_addr_list,
4574				transports) {
4575		memcpy(&temp, &from->ipaddr, sizeof(temp));
4576		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4577		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4578		if (space_left < addrlen)
4579			return -ENOMEM;
4580		if (copy_to_user(to, &temp, addrlen))
4581			return -EFAULT;
4582		to += addrlen;
4583		cnt++;
4584		space_left -= addrlen;
4585	}
4586
4587	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4588		return -EFAULT;
4589	bytes_copied = ((char __user *)to) - optval;
4590	if (put_user(bytes_copied, optlen))
4591		return -EFAULT;
4592
4593	return 0;
4594}
4595
4596static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4597			    size_t space_left, int *bytes_copied)
4598{
4599	struct sctp_sockaddr_entry *addr;
4600	union sctp_addr temp;
4601	int cnt = 0;
4602	int addrlen;
 
4603
4604	rcu_read_lock();
4605	list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4606		if (!addr->valid)
4607			continue;
4608
4609		if ((PF_INET == sk->sk_family) &&
4610		    (AF_INET6 == addr->a.sa.sa_family))
4611			continue;
4612		if ((PF_INET6 == sk->sk_family) &&
4613		    inet_v6_ipv6only(sk) &&
4614		    (AF_INET == addr->a.sa.sa_family))
4615			continue;
4616		memcpy(&temp, &addr->a, sizeof(temp));
4617		if (!temp.v4.sin_port)
4618			temp.v4.sin_port = htons(port);
4619
4620		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4621								&temp);
4622		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4623		if (space_left < addrlen) {
4624			cnt =  -ENOMEM;
4625			break;
4626		}
4627		memcpy(to, &temp, addrlen);
4628
4629		to += addrlen;
4630		cnt ++;
4631		space_left -= addrlen;
4632		*bytes_copied += addrlen;
4633	}
4634	rcu_read_unlock();
4635
4636	return cnt;
4637}
4638
4639
4640static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4641				       char __user *optval, int __user *optlen)
4642{
4643	struct sctp_bind_addr *bp;
4644	struct sctp_association *asoc;
4645	int cnt = 0;
4646	struct sctp_getaddrs getaddrs;
4647	struct sctp_sockaddr_entry *addr;
4648	void __user *to;
4649	union sctp_addr temp;
4650	struct sctp_sock *sp = sctp_sk(sk);
4651	int addrlen;
4652	int err = 0;
4653	size_t space_left;
4654	int bytes_copied = 0;
4655	void *addrs;
4656	void *buf;
4657
4658	if (len < sizeof(struct sctp_getaddrs))
4659		return -EINVAL;
4660
4661	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4662		return -EFAULT;
4663
4664	/*
4665	 *  For UDP-style sockets, id specifies the association to query.
4666	 *  If the id field is set to the value '0' then the locally bound
4667	 *  addresses are returned without regard to any particular
4668	 *  association.
4669	 */
4670	if (0 == getaddrs.assoc_id) {
4671		bp = &sctp_sk(sk)->ep->base.bind_addr;
4672	} else {
4673		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4674		if (!asoc)
4675			return -EINVAL;
4676		bp = &asoc->base.bind_addr;
4677	}
4678
4679	to = optval + offsetof(struct sctp_getaddrs,addrs);
4680	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4681
4682	addrs = kmalloc(space_left, GFP_KERNEL);
4683	if (!addrs)
4684		return -ENOMEM;
4685
4686	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4687	 * addresses from the global local address list.
4688	 */
4689	if (sctp_list_single_entry(&bp->address_list)) {
4690		addr = list_entry(bp->address_list.next,
4691				  struct sctp_sockaddr_entry, list);
4692		if (sctp_is_any(sk, &addr->a)) {
4693			cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4694						space_left, &bytes_copied);
4695			if (cnt < 0) {
4696				err = cnt;
4697				goto out;
4698			}
4699			goto copy_getaddrs;
4700		}
4701	}
4702
4703	buf = addrs;
4704	/* Protection on the bound address list is not needed since
4705	 * in the socket option context we hold a socket lock and
4706	 * thus the bound address list can't change.
4707	 */
4708	list_for_each_entry(addr, &bp->address_list, list) {
4709		memcpy(&temp, &addr->a, sizeof(temp));
4710		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4711		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4712		if (space_left < addrlen) {
4713			err =  -ENOMEM; /*fixme: right error?*/
4714			goto out;
4715		}
4716		memcpy(buf, &temp, addrlen);
4717		buf += addrlen;
4718		bytes_copied += addrlen;
4719		cnt ++;
4720		space_left -= addrlen;
4721	}
4722
4723copy_getaddrs:
4724	if (copy_to_user(to, addrs, bytes_copied)) {
4725		err = -EFAULT;
4726		goto out;
4727	}
4728	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4729		err = -EFAULT;
4730		goto out;
4731	}
4732	if (put_user(bytes_copied, optlen))
4733		err = -EFAULT;
4734out:
4735	kfree(addrs);
4736	return err;
4737}
4738
4739/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4740 *
4741 * Requests that the local SCTP stack use the enclosed peer address as
4742 * the association primary.  The enclosed address must be one of the
4743 * association peer's addresses.
4744 */
4745static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4746					char __user *optval, int __user *optlen)
4747{
4748	struct sctp_prim prim;
4749	struct sctp_association *asoc;
4750	struct sctp_sock *sp = sctp_sk(sk);
4751
4752	if (len < sizeof(struct sctp_prim))
4753		return -EINVAL;
4754
4755	len = sizeof(struct sctp_prim);
4756
4757	if (copy_from_user(&prim, optval, len))
4758		return -EFAULT;
4759
4760	asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4761	if (!asoc)
4762		return -EINVAL;
4763
4764	if (!asoc->peer.primary_path)
4765		return -ENOTCONN;
4766
4767	memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4768		asoc->peer.primary_path->af_specific->sockaddr_len);
4769
4770	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4771			(union sctp_addr *)&prim.ssp_addr);
4772
4773	if (put_user(len, optlen))
4774		return -EFAULT;
4775	if (copy_to_user(optval, &prim, len))
4776		return -EFAULT;
4777
4778	return 0;
4779}
4780
4781/*
4782 * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4783 *
4784 * Requests that the local endpoint set the specified Adaptation Layer
4785 * Indication parameter for all future INIT and INIT-ACK exchanges.
4786 */
4787static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4788				  char __user *optval, int __user *optlen)
4789{
4790	struct sctp_setadaptation adaptation;
4791
4792	if (len < sizeof(struct sctp_setadaptation))
4793		return -EINVAL;
4794
4795	len = sizeof(struct sctp_setadaptation);
4796
4797	adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4798
4799	if (put_user(len, optlen))
4800		return -EFAULT;
4801	if (copy_to_user(optval, &adaptation, len))
4802		return -EFAULT;
4803
4804	return 0;
4805}
4806
4807/*
4808 *
4809 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4810 *
4811 *   Applications that wish to use the sendto() system call may wish to
4812 *   specify a default set of parameters that would normally be supplied
4813 *   through the inclusion of ancillary data.  This socket option allows
4814 *   such an application to set the default sctp_sndrcvinfo structure.
4815
4816
4817 *   The application that wishes to use this socket option simply passes
4818 *   in to this call the sctp_sndrcvinfo structure defined in Section
4819 *   5.2.2) The input parameters accepted by this call include
4820 *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4821 *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
4822 *   to this call if the caller is using the UDP model.
4823 *
4824 *   For getsockopt, it get the default sctp_sndrcvinfo structure.
4825 */
4826static int sctp_getsockopt_default_send_param(struct sock *sk,
4827					int len, char __user *optval,
4828					int __user *optlen)
4829{
4830	struct sctp_sndrcvinfo info;
4831	struct sctp_association *asoc;
4832	struct sctp_sock *sp = sctp_sk(sk);
4833
4834	if (len < sizeof(struct sctp_sndrcvinfo))
4835		return -EINVAL;
4836
4837	len = sizeof(struct sctp_sndrcvinfo);
4838
4839	if (copy_from_user(&info, optval, len))
4840		return -EFAULT;
4841
4842	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4843	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4844		return -EINVAL;
4845
4846	if (asoc) {
4847		info.sinfo_stream = asoc->default_stream;
4848		info.sinfo_flags = asoc->default_flags;
4849		info.sinfo_ppid = asoc->default_ppid;
4850		info.sinfo_context = asoc->default_context;
4851		info.sinfo_timetolive = asoc->default_timetolive;
4852	} else {
4853		info.sinfo_stream = sp->default_stream;
4854		info.sinfo_flags = sp->default_flags;
4855		info.sinfo_ppid = sp->default_ppid;
4856		info.sinfo_context = sp->default_context;
4857		info.sinfo_timetolive = sp->default_timetolive;
4858	}
4859
4860	if (put_user(len, optlen))
4861		return -EFAULT;
4862	if (copy_to_user(optval, &info, len))
4863		return -EFAULT;
4864
4865	return 0;
4866}
4867
4868/*
4869 *
4870 * 7.1.5 SCTP_NODELAY
4871 *
4872 * Turn on/off any Nagle-like algorithm.  This means that packets are
4873 * generally sent as soon as possible and no unnecessary delays are
4874 * introduced, at the cost of more packets in the network.  Expects an
4875 * integer boolean flag.
4876 */
4877
4878static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4879				   char __user *optval, int __user *optlen)
4880{
4881	int val;
4882
4883	if (len < sizeof(int))
4884		return -EINVAL;
4885
4886	len = sizeof(int);
4887	val = (sctp_sk(sk)->nodelay == 1);
4888	if (put_user(len, optlen))
4889		return -EFAULT;
4890	if (copy_to_user(optval, &val, len))
4891		return -EFAULT;
4892	return 0;
4893}
4894
4895/*
4896 *
4897 * 7.1.1 SCTP_RTOINFO
4898 *
4899 * The protocol parameters used to initialize and bound retransmission
4900 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4901 * and modify these parameters.
4902 * All parameters are time values, in milliseconds.  A value of 0, when
4903 * modifying the parameters, indicates that the current value should not
4904 * be changed.
4905 *
4906 */
4907static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4908				char __user *optval,
4909				int __user *optlen) {
4910	struct sctp_rtoinfo rtoinfo;
4911	struct sctp_association *asoc;
4912
4913	if (len < sizeof (struct sctp_rtoinfo))
4914		return -EINVAL;
4915
4916	len = sizeof(struct sctp_rtoinfo);
4917
4918	if (copy_from_user(&rtoinfo, optval, len))
4919		return -EFAULT;
4920
4921	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4922
4923	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4924		return -EINVAL;
4925
4926	/* Values corresponding to the specific association. */
4927	if (asoc) {
4928		rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4929		rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4930		rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4931	} else {
4932		/* Values corresponding to the endpoint. */
4933		struct sctp_sock *sp = sctp_sk(sk);
4934
4935		rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4936		rtoinfo.srto_max = sp->rtoinfo.srto_max;
4937		rtoinfo.srto_min = sp->rtoinfo.srto_min;
4938	}
4939
4940	if (put_user(len, optlen))
4941		return -EFAULT;
4942
4943	if (copy_to_user(optval, &rtoinfo, len))
4944		return -EFAULT;
4945
4946	return 0;
4947}
4948
4949/*
4950 *
4951 * 7.1.2 SCTP_ASSOCINFO
4952 *
4953 * This option is used to tune the maximum retransmission attempts
4954 * of the association.
4955 * Returns an error if the new association retransmission value is
4956 * greater than the sum of the retransmission value  of the peer.
4957 * See [SCTP] for more information.
4958 *
4959 */
4960static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4961				     char __user *optval,
4962				     int __user *optlen)
4963{
4964
4965	struct sctp_assocparams assocparams;
4966	struct sctp_association *asoc;
4967	struct list_head *pos;
4968	int cnt = 0;
4969
4970	if (len < sizeof (struct sctp_assocparams))
4971		return -EINVAL;
4972
4973	len = sizeof(struct sctp_assocparams);
4974
4975	if (copy_from_user(&assocparams, optval, len))
4976		return -EFAULT;
4977
4978	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4979
4980	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4981		return -EINVAL;
4982
4983	/* Values correspoinding to the specific association */
4984	if (asoc) {
4985		assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4986		assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4987		assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4988		assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4989						* 1000) +
4990						(asoc->cookie_life.tv_usec
4991						/ 1000);
4992
4993		list_for_each(pos, &asoc->peer.transport_addr_list) {
4994			cnt ++;
4995		}
4996
4997		assocparams.sasoc_number_peer_destinations = cnt;
4998	} else {
4999		/* Values corresponding to the endpoint */
5000		struct sctp_sock *sp = sctp_sk(sk);
5001
5002		assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5003		assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5004		assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5005		assocparams.sasoc_cookie_life =
5006					sp->assocparams.sasoc_cookie_life;
5007		assocparams.sasoc_number_peer_destinations =
5008					sp->assocparams.
5009					sasoc_number_peer_destinations;
5010	}
5011
5012	if (put_user(len, optlen))
5013		return -EFAULT;
5014
5015	if (copy_to_user(optval, &assocparams, len))
5016		return -EFAULT;
5017
5018	return 0;
5019}
5020
5021/*
5022 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5023 *
5024 * This socket option is a boolean flag which turns on or off mapped V4
5025 * addresses.  If this option is turned on and the socket is type
5026 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5027 * If this option is turned off, then no mapping will be done of V4
5028 * addresses and a user will receive both PF_INET6 and PF_INET type
5029 * addresses on the socket.
5030 */
5031static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5032				    char __user *optval, int __user *optlen)
5033{
5034	int val;
5035	struct sctp_sock *sp = sctp_sk(sk);
5036
5037	if (len < sizeof(int))
5038		return -EINVAL;
5039
5040	len = sizeof(int);
5041	val = sp->v4mapped;
5042	if (put_user(len, optlen))
5043		return -EFAULT;
5044	if (copy_to_user(optval, &val, len))
5045		return -EFAULT;
5046
5047	return 0;
5048}
5049
5050/*
5051 * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
5052 * (chapter and verse is quoted at sctp_setsockopt_context())
5053 */
5054static int sctp_getsockopt_context(struct sock *sk, int len,
5055				   char __user *optval, int __user *optlen)
5056{
5057	struct sctp_assoc_value params;
5058	struct sctp_sock *sp;
5059	struct sctp_association *asoc;
5060
5061	if (len < sizeof(struct sctp_assoc_value))
5062		return -EINVAL;
5063
5064	len = sizeof(struct sctp_assoc_value);
5065
5066	if (copy_from_user(&params, optval, len))
5067		return -EFAULT;
5068
5069	sp = sctp_sk(sk);
5070
5071	if (params.assoc_id != 0) {
5072		asoc = sctp_id2assoc(sk, params.assoc_id);
5073		if (!asoc)
5074			return -EINVAL;
5075		params.assoc_value = asoc->default_rcv_context;
5076	} else {
5077		params.assoc_value = sp->default_rcv_context;
5078	}
5079
5080	if (put_user(len, optlen))
5081		return -EFAULT;
5082	if (copy_to_user(optval, &params, len))
5083		return -EFAULT;
5084
5085	return 0;
5086}
5087
5088/*
5089 * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5090 * This option will get or set the maximum size to put in any outgoing
5091 * SCTP DATA chunk.  If a message is larger than this size it will be
5092 * fragmented by SCTP into the specified size.  Note that the underlying
5093 * SCTP implementation may fragment into smaller sized chunks when the
5094 * PMTU of the underlying association is smaller than the value set by
5095 * the user.  The default value for this option is '0' which indicates
5096 * the user is NOT limiting fragmentation and only the PMTU will effect
5097 * SCTP's choice of DATA chunk size.  Note also that values set larger
5098 * than the maximum size of an IP datagram will effectively let SCTP
5099 * control fragmentation (i.e. the same as setting this option to 0).
5100 *
5101 * The following structure is used to access and modify this parameter:
5102 *
5103 * struct sctp_assoc_value {
5104 *   sctp_assoc_t assoc_id;
5105 *   uint32_t assoc_value;
5106 * };
5107 *
5108 * assoc_id:  This parameter is ignored for one-to-one style sockets.
5109 *    For one-to-many style sockets this parameter indicates which
5110 *    association the user is performing an action upon.  Note that if
5111 *    this field's value is zero then the endpoints default value is
5112 *    changed (effecting future associations only).
5113 * assoc_value:  This parameter specifies the maximum size in bytes.
5114 */
5115static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5116				  char __user *optval, int __user *optlen)
5117{
5118	struct sctp_assoc_value params;
5119	struct sctp_association *asoc;
5120
5121	if (len == sizeof(int)) {
5122		pr_warn("Use of int in maxseg socket option deprecated\n");
5123		pr_warn("Use struct sctp_assoc_value instead\n");
 
 
 
5124		params.assoc_id = 0;
5125	} else if (len >= sizeof(struct sctp_assoc_value)) {
5126		len = sizeof(struct sctp_assoc_value);
5127		if (copy_from_user(&params, optval, sizeof(params)))
5128			return -EFAULT;
5129	} else
5130		return -EINVAL;
5131
5132	asoc = sctp_id2assoc(sk, params.assoc_id);
5133	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5134		return -EINVAL;
5135
5136	if (asoc)
5137		params.assoc_value = asoc->frag_point;
5138	else
5139		params.assoc_value = sctp_sk(sk)->user_frag;
5140
5141	if (put_user(len, optlen))
5142		return -EFAULT;
5143	if (len == sizeof(int)) {
5144		if (copy_to_user(optval, &params.assoc_value, len))
5145			return -EFAULT;
5146	} else {
5147		if (copy_to_user(optval, &params, len))
5148			return -EFAULT;
5149	}
5150
5151	return 0;
5152}
5153
5154/*
5155 * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5156 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5157 */
5158static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5159					       char __user *optval, int __user *optlen)
5160{
5161	int val;
5162
5163	if (len < sizeof(int))
5164		return -EINVAL;
5165
5166	len = sizeof(int);
5167
5168	val = sctp_sk(sk)->frag_interleave;
5169	if (put_user(len, optlen))
5170		return -EFAULT;
5171	if (copy_to_user(optval, &val, len))
5172		return -EFAULT;
5173
5174	return 0;
5175}
5176
5177/*
5178 * 7.1.25.  Set or Get the sctp partial delivery point
5179 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5180 */
5181static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5182						  char __user *optval,
5183						  int __user *optlen)
5184{
5185	u32 val;
5186
5187	if (len < sizeof(u32))
5188		return -EINVAL;
5189
5190	len = sizeof(u32);
5191
5192	val = sctp_sk(sk)->pd_point;
5193	if (put_user(len, optlen))
5194		return -EFAULT;
5195	if (copy_to_user(optval, &val, len))
5196		return -EFAULT;
5197
5198	return 0;
5199}
5200
5201/*
5202 * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
5203 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5204 */
5205static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5206				    char __user *optval,
5207				    int __user *optlen)
5208{
5209	struct sctp_assoc_value params;
5210	struct sctp_sock *sp;
5211	struct sctp_association *asoc;
5212
5213	if (len == sizeof(int)) {
5214		pr_warn("Use of int in max_burst socket option deprecated\n");
5215		pr_warn("Use struct sctp_assoc_value instead\n");
 
 
 
5216		params.assoc_id = 0;
5217	} else if (len >= sizeof(struct sctp_assoc_value)) {
5218		len = sizeof(struct sctp_assoc_value);
5219		if (copy_from_user(&params, optval, len))
5220			return -EFAULT;
5221	} else
5222		return -EINVAL;
5223
5224	sp = sctp_sk(sk);
5225
5226	if (params.assoc_id != 0) {
5227		asoc = sctp_id2assoc(sk, params.assoc_id);
5228		if (!asoc)
5229			return -EINVAL;
5230		params.assoc_value = asoc->max_burst;
5231	} else
5232		params.assoc_value = sp->max_burst;
5233
5234	if (len == sizeof(int)) {
5235		if (copy_to_user(optval, &params.assoc_value, len))
5236			return -EFAULT;
5237	} else {
5238		if (copy_to_user(optval, &params, len))
5239			return -EFAULT;
5240	}
5241
5242	return 0;
5243
5244}
5245
5246static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5247				    char __user *optval, int __user *optlen)
5248{
 
5249	struct sctp_hmacalgo  __user *p = (void __user *)optval;
5250	struct sctp_hmac_algo_param *hmacs;
5251	__u16 data_len = 0;
5252	u32 num_idents;
5253
5254	if (!sctp_auth_enable)
5255		return -EACCES;
5256
5257	hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5258	data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5259
5260	if (len < sizeof(struct sctp_hmacalgo) + data_len)
5261		return -EINVAL;
5262
5263	len = sizeof(struct sctp_hmacalgo) + data_len;
5264	num_idents = data_len / sizeof(u16);
5265
5266	if (put_user(len, optlen))
5267		return -EFAULT;
5268	if (put_user(num_idents, &p->shmac_num_idents))
5269		return -EFAULT;
5270	if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5271		return -EFAULT;
5272	return 0;
5273}
5274
5275static int sctp_getsockopt_active_key(struct sock *sk, int len,
5276				    char __user *optval, int __user *optlen)
5277{
 
5278	struct sctp_authkeyid val;
5279	struct sctp_association *asoc;
5280
5281	if (!sctp_auth_enable)
5282		return -EACCES;
5283
5284	if (len < sizeof(struct sctp_authkeyid))
5285		return -EINVAL;
5286	if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5287		return -EFAULT;
5288
5289	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5290	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5291		return -EINVAL;
5292
5293	if (asoc)
5294		val.scact_keynumber = asoc->active_key_id;
5295	else
5296		val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5297
5298	len = sizeof(struct sctp_authkeyid);
5299	if (put_user(len, optlen))
5300		return -EFAULT;
5301	if (copy_to_user(optval, &val, len))
5302		return -EFAULT;
5303
5304	return 0;
5305}
5306
5307static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5308				    char __user *optval, int __user *optlen)
5309{
 
5310	struct sctp_authchunks __user *p = (void __user *)optval;
5311	struct sctp_authchunks val;
5312	struct sctp_association *asoc;
5313	struct sctp_chunks_param *ch;
5314	u32    num_chunks = 0;
5315	char __user *to;
5316
5317	if (!sctp_auth_enable)
5318		return -EACCES;
5319
5320	if (len < sizeof(struct sctp_authchunks))
5321		return -EINVAL;
5322
5323	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5324		return -EFAULT;
5325
5326	to = p->gauth_chunks;
5327	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5328	if (!asoc)
5329		return -EINVAL;
5330
5331	ch = asoc->peer.peer_chunks;
5332	if (!ch)
5333		goto num;
5334
5335	/* See if the user provided enough room for all the data */
5336	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5337	if (len < num_chunks)
5338		return -EINVAL;
5339
5340	if (copy_to_user(to, ch->chunks, num_chunks))
5341		return -EFAULT;
5342num:
5343	len = sizeof(struct sctp_authchunks) + num_chunks;
5344	if (put_user(len, optlen)) return -EFAULT;
 
5345	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5346		return -EFAULT;
5347	return 0;
5348}
5349
5350static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5351				    char __user *optval, int __user *optlen)
5352{
 
5353	struct sctp_authchunks __user *p = (void __user *)optval;
5354	struct sctp_authchunks val;
5355	struct sctp_association *asoc;
5356	struct sctp_chunks_param *ch;
5357	u32    num_chunks = 0;
5358	char __user *to;
5359
5360	if (!sctp_auth_enable)
5361		return -EACCES;
5362
5363	if (len < sizeof(struct sctp_authchunks))
5364		return -EINVAL;
5365
5366	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5367		return -EFAULT;
5368
5369	to = p->gauth_chunks;
5370	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5371	if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5372		return -EINVAL;
5373
5374	if (asoc)
5375		ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5376	else
5377		ch = sctp_sk(sk)->ep->auth_chunk_list;
5378
5379	if (!ch)
5380		goto num;
5381
5382	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5383	if (len < sizeof(struct sctp_authchunks) + num_chunks)
5384		return -EINVAL;
5385
5386	if (copy_to_user(to, ch->chunks, num_chunks))
5387		return -EFAULT;
5388num:
5389	len = sizeof(struct sctp_authchunks) + num_chunks;
5390	if (put_user(len, optlen))
5391		return -EFAULT;
5392	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5393		return -EFAULT;
5394
5395	return 0;
5396}
5397
5398/*
5399 * 8.2.5.  Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5400 * This option gets the current number of associations that are attached
5401 * to a one-to-many style socket.  The option value is an uint32_t.
5402 */
5403static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5404				    char __user *optval, int __user *optlen)
5405{
5406	struct sctp_sock *sp = sctp_sk(sk);
5407	struct sctp_association *asoc;
5408	u32 val = 0;
5409
5410	if (sctp_style(sk, TCP))
5411		return -EOPNOTSUPP;
5412
5413	if (len < sizeof(u32))
5414		return -EINVAL;
5415
5416	len = sizeof(u32);
5417
5418	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5419		val++;
5420	}
5421
5422	if (put_user(len, optlen))
5423		return -EFAULT;
5424	if (copy_to_user(optval, &val, len))
5425		return -EFAULT;
5426
5427	return 0;
5428}
5429
5430/*
5431 * 8.1.23 SCTP_AUTO_ASCONF
5432 * See the corresponding setsockopt entry as description
5433 */
5434static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5435				   char __user *optval, int __user *optlen)
5436{
5437	int val = 0;
5438
5439	if (len < sizeof(int))
5440		return -EINVAL;
5441
5442	len = sizeof(int);
5443	if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5444		val = 1;
5445	if (put_user(len, optlen))
5446		return -EFAULT;
5447	if (copy_to_user(optval, &val, len))
5448		return -EFAULT;
5449	return 0;
5450}
5451
5452/*
5453 * 8.2.6. Get the Current Identifiers of Associations
5454 *        (SCTP_GET_ASSOC_ID_LIST)
5455 *
5456 * This option gets the current list of SCTP association identifiers of
5457 * the SCTP associations handled by a one-to-many style socket.
5458 */
5459static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5460				    char __user *optval, int __user *optlen)
5461{
5462	struct sctp_sock *sp = sctp_sk(sk);
5463	struct sctp_association *asoc;
5464	struct sctp_assoc_ids *ids;
5465	u32 num = 0;
5466
5467	if (sctp_style(sk, TCP))
5468		return -EOPNOTSUPP;
5469
5470	if (len < sizeof(struct sctp_assoc_ids))
5471		return -EINVAL;
5472
5473	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5474		num++;
5475	}
5476
5477	if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5478		return -EINVAL;
5479
5480	len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5481
5482	ids = kmalloc(len, GFP_KERNEL);
5483	if (unlikely(!ids))
5484		return -ENOMEM;
5485
5486	ids->gaids_number_of_ids = num;
5487	num = 0;
5488	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5489		ids->gaids_assoc_id[num++] = asoc->assoc_id;
5490	}
5491
5492	if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5493		kfree(ids);
5494		return -EFAULT;
5495	}
5496
5497	kfree(ids);
5498	return 0;
5499}
5500
5501SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5502				char __user *optval, int __user *optlen)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5503{
5504	int retval = 0;
5505	int len;
5506
5507	SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5508			  sk, optname);
5509
5510	/* I can hardly begin to describe how wrong this is.  This is
5511	 * so broken as to be worse than useless.  The API draft
5512	 * REALLY is NOT helpful here...  I am not convinced that the
5513	 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5514	 * are at all well-founded.
5515	 */
5516	if (level != SOL_SCTP) {
5517		struct sctp_af *af = sctp_sk(sk)->pf->af;
5518
5519		retval = af->getsockopt(sk, level, optname, optval, optlen);
5520		return retval;
5521	}
5522
5523	if (get_user(len, optlen))
5524		return -EFAULT;
5525
5526	sctp_lock_sock(sk);
5527
5528	switch (optname) {
5529	case SCTP_STATUS:
5530		retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5531		break;
5532	case SCTP_DISABLE_FRAGMENTS:
5533		retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5534							   optlen);
5535		break;
5536	case SCTP_EVENTS:
5537		retval = sctp_getsockopt_events(sk, len, optval, optlen);
5538		break;
5539	case SCTP_AUTOCLOSE:
5540		retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5541		break;
5542	case SCTP_SOCKOPT_PEELOFF:
5543		retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5544		break;
5545	case SCTP_PEER_ADDR_PARAMS:
5546		retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5547							  optlen);
5548		break;
5549	case SCTP_DELAYED_SACK:
5550		retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5551							  optlen);
5552		break;
5553	case SCTP_INITMSG:
5554		retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5555		break;
5556	case SCTP_GET_PEER_ADDRS:
5557		retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5558						    optlen);
5559		break;
5560	case SCTP_GET_LOCAL_ADDRS:
5561		retval = sctp_getsockopt_local_addrs(sk, len, optval,
5562						     optlen);
5563		break;
5564	case SCTP_SOCKOPT_CONNECTX3:
5565		retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5566		break;
5567	case SCTP_DEFAULT_SEND_PARAM:
5568		retval = sctp_getsockopt_default_send_param(sk, len,
5569							    optval, optlen);
5570		break;
5571	case SCTP_PRIMARY_ADDR:
5572		retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5573		break;
5574	case SCTP_NODELAY:
5575		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5576		break;
5577	case SCTP_RTOINFO:
5578		retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5579		break;
5580	case SCTP_ASSOCINFO:
5581		retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5582		break;
5583	case SCTP_I_WANT_MAPPED_V4_ADDR:
5584		retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5585		break;
5586	case SCTP_MAXSEG:
5587		retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5588		break;
5589	case SCTP_GET_PEER_ADDR_INFO:
5590		retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5591							optlen);
5592		break;
5593	case SCTP_ADAPTATION_LAYER:
5594		retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5595							optlen);
5596		break;
5597	case SCTP_CONTEXT:
5598		retval = sctp_getsockopt_context(sk, len, optval, optlen);
5599		break;
5600	case SCTP_FRAGMENT_INTERLEAVE:
5601		retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5602							     optlen);
5603		break;
5604	case SCTP_PARTIAL_DELIVERY_POINT:
5605		retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5606								optlen);
5607		break;
5608	case SCTP_MAX_BURST:
5609		retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5610		break;
5611	case SCTP_AUTH_KEY:
5612	case SCTP_AUTH_CHUNK:
5613	case SCTP_AUTH_DELETE_KEY:
5614		retval = -EOPNOTSUPP;
5615		break;
5616	case SCTP_HMAC_IDENT:
5617		retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5618		break;
5619	case SCTP_AUTH_ACTIVE_KEY:
5620		retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5621		break;
5622	case SCTP_PEER_AUTH_CHUNKS:
5623		retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5624							optlen);
5625		break;
5626	case SCTP_LOCAL_AUTH_CHUNKS:
5627		retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5628							optlen);
5629		break;
5630	case SCTP_GET_ASSOC_NUMBER:
5631		retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5632		break;
5633	case SCTP_GET_ASSOC_ID_LIST:
5634		retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5635		break;
5636	case SCTP_AUTO_ASCONF:
5637		retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5638		break;
 
 
 
 
 
 
5639	default:
5640		retval = -ENOPROTOOPT;
5641		break;
5642	}
5643
5644	sctp_release_sock(sk);
5645	return retval;
5646}
5647
5648static void sctp_hash(struct sock *sk)
5649{
5650	/* STUB */
5651}
5652
5653static void sctp_unhash(struct sock *sk)
5654{
5655	/* STUB */
5656}
5657
5658/* Check if port is acceptable.  Possibly find first available port.
5659 *
5660 * The port hash table (contained in the 'global' SCTP protocol storage
5661 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5662 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5663 * list (the list number is the port number hashed out, so as you
5664 * would expect from a hash function, all the ports in a given list have
5665 * such a number that hashes out to the same list number; you were
5666 * expecting that, right?); so each list has a set of ports, with a
5667 * link to the socket (struct sock) that uses it, the port number and
5668 * a fastreuse flag (FIXME: NPI ipg).
5669 */
5670static struct sctp_bind_bucket *sctp_bucket_create(
5671	struct sctp_bind_hashbucket *head, unsigned short snum);
5672
5673static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5674{
5675	struct sctp_bind_hashbucket *head; /* hash list */
5676	struct sctp_bind_bucket *pp; /* hash list port iterator */
5677	struct hlist_node *node;
5678	unsigned short snum;
5679	int ret;
5680
5681	snum = ntohs(addr->v4.sin_port);
5682
5683	SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5684	sctp_local_bh_disable();
 
5685
5686	if (snum == 0) {
5687		/* Search for an available port. */
5688		int low, high, remaining, index;
5689		unsigned int rover;
5690
5691		inet_get_local_port_range(&low, &high);
5692		remaining = (high - low) + 1;
5693		rover = net_random() % remaining + low;
5694
5695		do {
5696			rover++;
5697			if ((rover < low) || (rover > high))
5698				rover = low;
5699			if (inet_is_reserved_local_port(rover))
5700				continue;
5701			index = sctp_phashfn(rover);
5702			head = &sctp_port_hashtable[index];
5703			sctp_spin_lock(&head->lock);
5704			sctp_for_each_hentry(pp, node, &head->chain)
5705				if (pp->port == rover)
 
5706					goto next;
5707			break;
5708		next:
5709			sctp_spin_unlock(&head->lock);
5710		} while (--remaining > 0);
5711
5712		/* Exhausted local port range during search? */
5713		ret = 1;
5714		if (remaining <= 0)
5715			goto fail;
5716
5717		/* OK, here is the one we will use.  HEAD (the port
5718		 * hash table list entry) is non-NULL and we hold it's
5719		 * mutex.
5720		 */
5721		snum = rover;
5722	} else {
5723		/* We are given an specific port number; we verify
5724		 * that it is not being used. If it is used, we will
5725		 * exahust the search in the hash list corresponding
5726		 * to the port number (snum) - we detect that with the
5727		 * port iterator, pp being NULL.
5728		 */
5729		head = &sctp_port_hashtable[sctp_phashfn(snum)];
5730		sctp_spin_lock(&head->lock);
5731		sctp_for_each_hentry(pp, node, &head->chain) {
5732			if (pp->port == snum)
5733				goto pp_found;
5734		}
5735	}
5736	pp = NULL;
5737	goto pp_not_found;
5738pp_found:
5739	if (!hlist_empty(&pp->owner)) {
5740		/* We had a port hash table hit - there is an
5741		 * available port (pp != NULL) and it is being
5742		 * used by other socket (pp->owner not empty); that other
5743		 * socket is going to be sk2.
5744		 */
5745		int reuse = sk->sk_reuse;
5746		struct sock *sk2;
5747
5748		SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
 
5749		if (pp->fastreuse && sk->sk_reuse &&
5750			sk->sk_state != SCTP_SS_LISTENING)
5751			goto success;
5752
5753		/* Run through the list of sockets bound to the port
5754		 * (pp->port) [via the pointers bind_next and
5755		 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5756		 * we get the endpoint they describe and run through
5757		 * the endpoint's list of IP (v4 or v6) addresses,
5758		 * comparing each of the addresses with the address of
5759		 * the socket sk. If we find a match, then that means
5760		 * that this port/socket (sk) combination are already
5761		 * in an endpoint.
5762		 */
5763		sk_for_each_bound(sk2, node, &pp->owner) {
5764			struct sctp_endpoint *ep2;
5765			ep2 = sctp_sk(sk2)->ep;
5766
5767			if (sk == sk2 ||
5768			    (reuse && sk2->sk_reuse &&
5769			     sk2->sk_state != SCTP_SS_LISTENING))
5770				continue;
5771
5772			if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5773						 sctp_sk(sk2), sctp_sk(sk))) {
5774				ret = (long)sk2;
5775				goto fail_unlock;
5776			}
5777		}
5778		SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
 
5779	}
5780pp_not_found:
5781	/* If there was a hash table miss, create a new port.  */
5782	ret = 1;
5783	if (!pp && !(pp = sctp_bucket_create(head, snum)))
5784		goto fail_unlock;
5785
5786	/* In either case (hit or miss), make sure fastreuse is 1 only
5787	 * if sk->sk_reuse is too (that is, if the caller requested
5788	 * SO_REUSEADDR on this socket -sk-).
5789	 */
5790	if (hlist_empty(&pp->owner)) {
5791		if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5792			pp->fastreuse = 1;
5793		else
5794			pp->fastreuse = 0;
5795	} else if (pp->fastreuse &&
5796		(!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5797		pp->fastreuse = 0;
5798
5799	/* We are set, so fill up all the data in the hash table
5800	 * entry, tie the socket list information with the rest of the
5801	 * sockets FIXME: Blurry, NPI (ipg).
5802	 */
5803success:
5804	if (!sctp_sk(sk)->bind_hash) {
5805		inet_sk(sk)->inet_num = snum;
5806		sk_add_bind_node(sk, &pp->owner);
5807		sctp_sk(sk)->bind_hash = pp;
5808	}
5809	ret = 0;
5810
5811fail_unlock:
5812	sctp_spin_unlock(&head->lock);
5813
5814fail:
5815	sctp_local_bh_enable();
5816	return ret;
5817}
5818
5819/* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
5820 * port is requested.
5821 */
5822static int sctp_get_port(struct sock *sk, unsigned short snum)
5823{
5824	long ret;
5825	union sctp_addr addr;
5826	struct sctp_af *af = sctp_sk(sk)->pf->af;
5827
5828	/* Set up a dummy address struct from the sk. */
5829	af->from_sk(&addr, sk);
5830	addr.v4.sin_port = htons(snum);
5831
5832	/* Note: sk->sk_num gets filled in if ephemeral port request. */
5833	ret = sctp_get_port_local(sk, &addr);
5834
5835	return ret ? 1 : 0;
5836}
5837
5838/*
5839 *  Move a socket to LISTENING state.
5840 */
5841SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5842{
5843	struct sctp_sock *sp = sctp_sk(sk);
5844	struct sctp_endpoint *ep = sp->ep;
5845	struct crypto_hash *tfm = NULL;
 
5846
5847	/* Allocate HMAC for generating cookie. */
5848	if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5849		tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
 
5850		if (IS_ERR(tfm)) {
5851			net_info_ratelimited("failed to load transform for %s: %ld\n",
5852					     sctp_hmac_alg, PTR_ERR(tfm));
5853			return -ENOSYS;
5854		}
5855		sctp_sk(sk)->hmac = tfm;
5856	}
5857
5858	/*
5859	 * If a bind() or sctp_bindx() is not called prior to a listen()
5860	 * call that allows new associations to be accepted, the system
5861	 * picks an ephemeral port and will choose an address set equivalent
5862	 * to binding with a wildcard address.
5863	 *
5864	 * This is not currently spelled out in the SCTP sockets
5865	 * extensions draft, but follows the practice as seen in TCP
5866	 * sockets.
5867	 *
5868	 */
5869	sk->sk_state = SCTP_SS_LISTENING;
5870	if (!ep->base.bind_addr.port) {
5871		if (sctp_autobind(sk))
5872			return -EAGAIN;
5873	} else {
5874		if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5875			sk->sk_state = SCTP_SS_CLOSED;
5876			return -EADDRINUSE;
5877		}
5878	}
5879
5880	sk->sk_max_ack_backlog = backlog;
5881	sctp_hash_endpoint(ep);
5882	return 0;
5883}
5884
5885/*
5886 * 4.1.3 / 5.1.3 listen()
5887 *
5888 *   By default, new associations are not accepted for UDP style sockets.
5889 *   An application uses listen() to mark a socket as being able to
5890 *   accept new associations.
5891 *
5892 *   On TCP style sockets, applications use listen() to ready the SCTP
5893 *   endpoint for accepting inbound associations.
5894 *
5895 *   On both types of endpoints a backlog of '0' disables listening.
5896 *
5897 *  Move a socket to LISTENING state.
5898 */
5899int sctp_inet_listen(struct socket *sock, int backlog)
5900{
5901	struct sock *sk = sock->sk;
5902	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5903	int err = -EINVAL;
5904
5905	if (unlikely(backlog < 0))
5906		return err;
5907
5908	sctp_lock_sock(sk);
5909
5910	/* Peeled-off sockets are not allowed to listen().  */
5911	if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5912		goto out;
5913
5914	if (sock->state != SS_UNCONNECTED)
5915		goto out;
5916
5917	/* If backlog is zero, disable listening. */
5918	if (!backlog) {
5919		if (sctp_sstate(sk, CLOSED))
5920			goto out;
5921
5922		err = 0;
5923		sctp_unhash_endpoint(ep);
5924		sk->sk_state = SCTP_SS_CLOSED;
5925		if (sk->sk_reuse)
5926			sctp_sk(sk)->bind_hash->fastreuse = 1;
5927		goto out;
5928	}
5929
5930	/* If we are already listening, just update the backlog */
5931	if (sctp_sstate(sk, LISTENING))
5932		sk->sk_max_ack_backlog = backlog;
5933	else {
5934		err = sctp_listen_start(sk, backlog);
5935		if (err)
5936			goto out;
5937	}
5938
5939	err = 0;
5940out:
5941	sctp_release_sock(sk);
5942	return err;
5943}
5944
5945/*
5946 * This function is done by modeling the current datagram_poll() and the
5947 * tcp_poll().  Note that, based on these implementations, we don't
5948 * lock the socket in this function, even though it seems that,
5949 * ideally, locking or some other mechanisms can be used to ensure
5950 * the integrity of the counters (sndbuf and wmem_alloc) used
5951 * in this place.  We assume that we don't need locks either until proven
5952 * otherwise.
5953 *
5954 * Another thing to note is that we include the Async I/O support
5955 * here, again, by modeling the current TCP/UDP code.  We don't have
5956 * a good way to test with it yet.
5957 */
5958unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5959{
5960	struct sock *sk = sock->sk;
5961	struct sctp_sock *sp = sctp_sk(sk);
5962	unsigned int mask;
5963
5964	poll_wait(file, sk_sleep(sk), wait);
5965
5966	/* A TCP-style listening socket becomes readable when the accept queue
5967	 * is not empty.
5968	 */
5969	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5970		return (!list_empty(&sp->ep->asocs)) ?
5971			(POLLIN | POLLRDNORM) : 0;
5972
5973	mask = 0;
5974
5975	/* Is there any exceptional events?  */
5976	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5977		mask |= POLLERR;
 
5978	if (sk->sk_shutdown & RCV_SHUTDOWN)
5979		mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5980	if (sk->sk_shutdown == SHUTDOWN_MASK)
5981		mask |= POLLHUP;
5982
5983	/* Is it readable?  Reconsider this code with TCP-style support.  */
5984	if (!skb_queue_empty(&sk->sk_receive_queue))
5985		mask |= POLLIN | POLLRDNORM;
5986
5987	/* The association is either gone or not ready.  */
5988	if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5989		return mask;
5990
5991	/* Is it writable?  */
5992	if (sctp_writeable(sk)) {
5993		mask |= POLLOUT | POLLWRNORM;
5994	} else {
5995		set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5996		/*
5997		 * Since the socket is not locked, the buffer
5998		 * might be made available after the writeable check and
5999		 * before the bit is set.  This could cause a lost I/O
6000		 * signal.  tcp_poll() has a race breaker for this race
6001		 * condition.  Based on their implementation, we put
6002		 * in the following code to cover it as well.
6003		 */
6004		if (sctp_writeable(sk))
6005			mask |= POLLOUT | POLLWRNORM;
6006	}
6007	return mask;
6008}
6009
6010/********************************************************************
6011 * 2nd Level Abstractions
6012 ********************************************************************/
6013
6014static struct sctp_bind_bucket *sctp_bucket_create(
6015	struct sctp_bind_hashbucket *head, unsigned short snum)
6016{
6017	struct sctp_bind_bucket *pp;
6018
6019	pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6020	if (pp) {
6021		SCTP_DBG_OBJCNT_INC(bind_bucket);
6022		pp->port = snum;
6023		pp->fastreuse = 0;
6024		INIT_HLIST_HEAD(&pp->owner);
 
6025		hlist_add_head(&pp->node, &head->chain);
6026	}
6027	return pp;
6028}
6029
6030/* Caller must hold hashbucket lock for this tb with local BH disabled */
6031static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6032{
6033	if (pp && hlist_empty(&pp->owner)) {
6034		__hlist_del(&pp->node);
6035		kmem_cache_free(sctp_bucket_cachep, pp);
6036		SCTP_DBG_OBJCNT_DEC(bind_bucket);
6037	}
6038}
6039
6040/* Release this socket's reference to a local port.  */
6041static inline void __sctp_put_port(struct sock *sk)
6042{
6043	struct sctp_bind_hashbucket *head =
6044		&sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
 
6045	struct sctp_bind_bucket *pp;
6046
6047	sctp_spin_lock(&head->lock);
6048	pp = sctp_sk(sk)->bind_hash;
6049	__sk_del_bind_node(sk);
6050	sctp_sk(sk)->bind_hash = NULL;
6051	inet_sk(sk)->inet_num = 0;
6052	sctp_bucket_destroy(pp);
6053	sctp_spin_unlock(&head->lock);
6054}
6055
6056void sctp_put_port(struct sock *sk)
6057{
6058	sctp_local_bh_disable();
6059	__sctp_put_port(sk);
6060	sctp_local_bh_enable();
6061}
6062
6063/*
6064 * The system picks an ephemeral port and choose an address set equivalent
6065 * to binding with a wildcard address.
6066 * One of those addresses will be the primary address for the association.
6067 * This automatically enables the multihoming capability of SCTP.
6068 */
6069static int sctp_autobind(struct sock *sk)
6070{
6071	union sctp_addr autoaddr;
6072	struct sctp_af *af;
6073	__be16 port;
6074
6075	/* Initialize a local sockaddr structure to INADDR_ANY. */
6076	af = sctp_sk(sk)->pf->af;
6077
6078	port = htons(inet_sk(sk)->inet_num);
6079	af->inaddr_any(&autoaddr, port);
6080
6081	return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6082}
6083
6084/* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
6085 *
6086 * From RFC 2292
6087 * 4.2 The cmsghdr Structure *
6088 *
6089 * When ancillary data is sent or received, any number of ancillary data
6090 * objects can be specified by the msg_control and msg_controllen members of
6091 * the msghdr structure, because each object is preceded by
6092 * a cmsghdr structure defining the object's length (the cmsg_len member).
6093 * Historically Berkeley-derived implementations have passed only one object
6094 * at a time, but this API allows multiple objects to be
6095 * passed in a single call to sendmsg() or recvmsg(). The following example
6096 * shows two ancillary data objects in a control buffer.
6097 *
6098 *   |<--------------------------- msg_controllen -------------------------->|
6099 *   |                                                                       |
6100 *
6101 *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
6102 *
6103 *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6104 *   |                                   |                                   |
6105 *
6106 *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
6107 *
6108 *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
6109 *   |                                |  |                                |  |
6110 *
6111 *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6112 *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
6113 *
6114 *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
6115 *
6116 *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6117 *    ^
6118 *    |
6119 *
6120 * msg_control
6121 * points here
6122 */
6123SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6124				  sctp_cmsgs_t *cmsgs)
6125{
6126	struct cmsghdr *cmsg;
6127	struct msghdr *my_msg = (struct msghdr *)msg;
6128
6129	for (cmsg = CMSG_FIRSTHDR(msg);
6130	     cmsg != NULL;
6131	     cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6132		if (!CMSG_OK(my_msg, cmsg))
6133			return -EINVAL;
6134
6135		/* Should we parse this header or ignore?  */
6136		if (cmsg->cmsg_level != IPPROTO_SCTP)
6137			continue;
6138
6139		/* Strictly check lengths following example in SCM code.  */
6140		switch (cmsg->cmsg_type) {
6141		case SCTP_INIT:
6142			/* SCTP Socket API Extension
6143			 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6144			 *
6145			 * This cmsghdr structure provides information for
6146			 * initializing new SCTP associations with sendmsg().
6147			 * The SCTP_INITMSG socket option uses this same data
6148			 * structure.  This structure is not used for
6149			 * recvmsg().
6150			 *
6151			 * cmsg_level    cmsg_type      cmsg_data[]
6152			 * ------------  ------------   ----------------------
6153			 * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
6154			 */
6155			if (cmsg->cmsg_len !=
6156			    CMSG_LEN(sizeof(struct sctp_initmsg)))
6157				return -EINVAL;
6158			cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6159			break;
6160
6161		case SCTP_SNDRCV:
6162			/* SCTP Socket API Extension
6163			 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6164			 *
6165			 * This cmsghdr structure specifies SCTP options for
6166			 * sendmsg() and describes SCTP header information
6167			 * about a received message through recvmsg().
6168			 *
6169			 * cmsg_level    cmsg_type      cmsg_data[]
6170			 * ------------  ------------   ----------------------
6171			 * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
6172			 */
6173			if (cmsg->cmsg_len !=
6174			    CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6175				return -EINVAL;
6176
6177			cmsgs->info =
6178				(struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6179
6180			/* Minimally, validate the sinfo_flags. */
6181			if (cmsgs->info->sinfo_flags &
6182			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6183			      SCTP_ABORT | SCTP_EOF))
6184				return -EINVAL;
6185			break;
6186
6187		default:
6188			return -EINVAL;
6189		}
6190	}
6191	return 0;
6192}
6193
6194/*
6195 * Wait for a packet..
6196 * Note: This function is the same function as in core/datagram.c
6197 * with a few modifications to make lksctp work.
6198 */
6199static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6200{
6201	int error;
6202	DEFINE_WAIT(wait);
6203
6204	prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6205
6206	/* Socket errors? */
6207	error = sock_error(sk);
6208	if (error)
6209		goto out;
6210
6211	if (!skb_queue_empty(&sk->sk_receive_queue))
6212		goto ready;
6213
6214	/* Socket shut down?  */
6215	if (sk->sk_shutdown & RCV_SHUTDOWN)
6216		goto out;
6217
6218	/* Sequenced packets can come disconnected.  If so we report the
6219	 * problem.
6220	 */
6221	error = -ENOTCONN;
6222
6223	/* Is there a good reason to think that we may receive some data?  */
6224	if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6225		goto out;
6226
6227	/* Handle signals.  */
6228	if (signal_pending(current))
6229		goto interrupted;
6230
6231	/* Let another process have a go.  Since we are going to sleep
6232	 * anyway.  Note: This may cause odd behaviors if the message
6233	 * does not fit in the user's buffer, but this seems to be the
6234	 * only way to honor MSG_DONTWAIT realistically.
6235	 */
6236	sctp_release_sock(sk);
6237	*timeo_p = schedule_timeout(*timeo_p);
6238	sctp_lock_sock(sk);
6239
6240ready:
6241	finish_wait(sk_sleep(sk), &wait);
6242	return 0;
6243
6244interrupted:
6245	error = sock_intr_errno(*timeo_p);
6246
6247out:
6248	finish_wait(sk_sleep(sk), &wait);
6249	*err = error;
6250	return error;
6251}
6252
6253/* Receive a datagram.
6254 * Note: This is pretty much the same routine as in core/datagram.c
6255 * with a few changes to make lksctp work.
6256 */
6257static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6258					      int noblock, int *err)
6259{
6260	int error;
6261	struct sk_buff *skb;
6262	long timeo;
6263
6264	timeo = sock_rcvtimeo(sk, noblock);
6265
6266	SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6267			  timeo, MAX_SCHEDULE_TIMEOUT);
6268
6269	do {
6270		/* Again only user level code calls this function,
6271		 * so nothing interrupt level
6272		 * will suddenly eat the receive_queue.
6273		 *
6274		 *  Look at current nfs client by the way...
6275		 *  However, this function was correct in any case. 8)
6276		 */
6277		if (flags & MSG_PEEK) {
6278			spin_lock_bh(&sk->sk_receive_queue.lock);
6279			skb = skb_peek(&sk->sk_receive_queue);
6280			if (skb)
6281				atomic_inc(&skb->users);
6282			spin_unlock_bh(&sk->sk_receive_queue.lock);
6283		} else {
6284			skb = skb_dequeue(&sk->sk_receive_queue);
6285		}
6286
6287		if (skb)
6288			return skb;
6289
6290		/* Caller is allowed not to check sk->sk_err before calling. */
6291		error = sock_error(sk);
6292		if (error)
6293			goto no_packet;
6294
6295		if (sk->sk_shutdown & RCV_SHUTDOWN)
6296			break;
6297
6298		/* User doesn't want to wait.  */
6299		error = -EAGAIN;
6300		if (!timeo)
6301			goto no_packet;
6302	} while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6303
6304	return NULL;
6305
6306no_packet:
6307	*err = error;
6308	return NULL;
6309}
6310
6311/* If sndbuf has changed, wake up per association sndbuf waiters.  */
6312static void __sctp_write_space(struct sctp_association *asoc)
6313{
6314	struct sock *sk = asoc->base.sk;
6315	struct socket *sock = sk->sk_socket;
6316
6317	if ((sctp_wspace(asoc) > 0) && sock) {
6318		if (waitqueue_active(&asoc->wait))
6319			wake_up_interruptible(&asoc->wait);
6320
6321		if (sctp_writeable(sk)) {
6322			wait_queue_head_t *wq = sk_sleep(sk);
6323
6324			if (wq && waitqueue_active(wq))
6325				wake_up_interruptible(wq);
6326
6327			/* Note that we try to include the Async I/O support
6328			 * here by modeling from the current TCP/UDP code.
6329			 * We have not tested with it yet.
6330			 */
6331			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6332				sock_wake_async(sock,
6333						SOCK_WAKE_SPACE, POLL_OUT);
6334		}
6335	}
6336}
6337
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6338/* Do accounting for the sndbuf space.
6339 * Decrement the used sndbuf space of the corresponding association by the
6340 * data size which was just transmitted(freed).
6341 */
6342static void sctp_wfree(struct sk_buff *skb)
6343{
6344	struct sctp_association *asoc;
6345	struct sctp_chunk *chunk;
6346	struct sock *sk;
6347
6348	/* Get the saved chunk pointer.  */
6349	chunk = *((struct sctp_chunk **)(skb->cb));
6350	asoc = chunk->asoc;
6351	sk = asoc->base.sk;
6352	asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6353				sizeof(struct sk_buff) +
6354				sizeof(struct sctp_chunk);
6355
6356	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6357
6358	/*
6359	 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6360	 */
6361	sk->sk_wmem_queued   -= skb->truesize;
6362	sk_mem_uncharge(sk, skb->truesize);
6363
6364	sock_wfree(skb);
6365	__sctp_write_space(asoc);
6366
6367	sctp_association_put(asoc);
6368}
6369
6370/* Do accounting for the receive space on the socket.
6371 * Accounting for the association is done in ulpevent.c
6372 * We set this as a destructor for the cloned data skbs so that
6373 * accounting is done at the correct time.
6374 */
6375void sctp_sock_rfree(struct sk_buff *skb)
6376{
6377	struct sock *sk = skb->sk;
6378	struct sctp_ulpevent *event = sctp_skb2event(skb);
6379
6380	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6381
6382	/*
6383	 * Mimic the behavior of sock_rfree
6384	 */
6385	sk_mem_uncharge(sk, event->rmem_len);
6386}
6387
6388
6389/* Helper function to wait for space in the sndbuf.  */
6390static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6391				size_t msg_len)
6392{
6393	struct sock *sk = asoc->base.sk;
6394	int err = 0;
6395	long current_timeo = *timeo_p;
6396	DEFINE_WAIT(wait);
6397
6398	SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6399			  asoc, (long)(*timeo_p), msg_len);
6400
6401	/* Increment the association's refcnt.  */
6402	sctp_association_hold(asoc);
6403
6404	/* Wait on the association specific sndbuf space. */
6405	for (;;) {
6406		prepare_to_wait_exclusive(&asoc->wait, &wait,
6407					  TASK_INTERRUPTIBLE);
6408		if (!*timeo_p)
6409			goto do_nonblock;
6410		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6411		    asoc->base.dead)
6412			goto do_error;
6413		if (signal_pending(current))
6414			goto do_interrupted;
6415		if (msg_len <= sctp_wspace(asoc))
6416			break;
6417
6418		/* Let another process have a go.  Since we are going
6419		 * to sleep anyway.
6420		 */
6421		sctp_release_sock(sk);
6422		current_timeo = schedule_timeout(current_timeo);
6423		BUG_ON(sk != asoc->base.sk);
6424		sctp_lock_sock(sk);
6425
6426		*timeo_p = current_timeo;
6427	}
6428
6429out:
6430	finish_wait(&asoc->wait, &wait);
6431
6432	/* Release the association's refcnt.  */
6433	sctp_association_put(asoc);
6434
6435	return err;
6436
6437do_error:
6438	err = -EPIPE;
6439	goto out;
6440
6441do_interrupted:
6442	err = sock_intr_errno(*timeo_p);
6443	goto out;
6444
6445do_nonblock:
6446	err = -EAGAIN;
6447	goto out;
6448}
6449
6450void sctp_data_ready(struct sock *sk, int len)
6451{
6452	struct socket_wq *wq;
6453
6454	rcu_read_lock();
6455	wq = rcu_dereference(sk->sk_wq);
6456	if (wq_has_sleeper(wq))
6457		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6458						POLLRDNORM | POLLRDBAND);
6459	sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6460	rcu_read_unlock();
6461}
6462
6463/* If socket sndbuf has changed, wake up all per association waiters.  */
6464void sctp_write_space(struct sock *sk)
6465{
6466	struct sctp_association *asoc;
6467
6468	/* Wake up the tasks in each wait queue.  */
6469	list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6470		__sctp_write_space(asoc);
6471	}
6472}
6473
6474/* Is there any sndbuf space available on the socket?
6475 *
6476 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6477 * associations on the same socket.  For a UDP-style socket with
6478 * multiple associations, it is possible for it to be "unwriteable"
6479 * prematurely.  I assume that this is acceptable because
6480 * a premature "unwriteable" is better than an accidental "writeable" which
6481 * would cause an unwanted block under certain circumstances.  For the 1-1
6482 * UDP-style sockets or TCP-style sockets, this code should work.
6483 *  - Daisy
6484 */
6485static int sctp_writeable(struct sock *sk)
6486{
6487	int amt = 0;
6488
6489	amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6490	if (amt < 0)
6491		amt = 0;
6492	return amt;
6493}
6494
6495/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6496 * returns immediately with EINPROGRESS.
6497 */
6498static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6499{
6500	struct sock *sk = asoc->base.sk;
6501	int err = 0;
6502	long current_timeo = *timeo_p;
6503	DEFINE_WAIT(wait);
6504
6505	SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6506			  (long)(*timeo_p));
6507
6508	/* Increment the association's refcnt.  */
6509	sctp_association_hold(asoc);
6510
6511	for (;;) {
6512		prepare_to_wait_exclusive(&asoc->wait, &wait,
6513					  TASK_INTERRUPTIBLE);
6514		if (!*timeo_p)
6515			goto do_nonblock;
6516		if (sk->sk_shutdown & RCV_SHUTDOWN)
6517			break;
6518		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6519		    asoc->base.dead)
6520			goto do_error;
6521		if (signal_pending(current))
6522			goto do_interrupted;
6523
6524		if (sctp_state(asoc, ESTABLISHED))
6525			break;
6526
6527		/* Let another process have a go.  Since we are going
6528		 * to sleep anyway.
6529		 */
6530		sctp_release_sock(sk);
6531		current_timeo = schedule_timeout(current_timeo);
6532		sctp_lock_sock(sk);
6533
6534		*timeo_p = current_timeo;
6535	}
6536
6537out:
6538	finish_wait(&asoc->wait, &wait);
6539
6540	/* Release the association's refcnt.  */
6541	sctp_association_put(asoc);
6542
6543	return err;
6544
6545do_error:
6546	if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6547		err = -ETIMEDOUT;
6548	else
6549		err = -ECONNREFUSED;
6550	goto out;
6551
6552do_interrupted:
6553	err = sock_intr_errno(*timeo_p);
6554	goto out;
6555
6556do_nonblock:
6557	err = -EINPROGRESS;
6558	goto out;
6559}
6560
6561static int sctp_wait_for_accept(struct sock *sk, long timeo)
6562{
6563	struct sctp_endpoint *ep;
6564	int err = 0;
6565	DEFINE_WAIT(wait);
6566
6567	ep = sctp_sk(sk)->ep;
6568
6569
6570	for (;;) {
6571		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6572					  TASK_INTERRUPTIBLE);
6573
6574		if (list_empty(&ep->asocs)) {
6575			sctp_release_sock(sk);
6576			timeo = schedule_timeout(timeo);
6577			sctp_lock_sock(sk);
6578		}
6579
6580		err = -EINVAL;
6581		if (!sctp_sstate(sk, LISTENING))
6582			break;
6583
6584		err = 0;
6585		if (!list_empty(&ep->asocs))
6586			break;
6587
6588		err = sock_intr_errno(timeo);
6589		if (signal_pending(current))
6590			break;
6591
6592		err = -EAGAIN;
6593		if (!timeo)
6594			break;
6595	}
6596
6597	finish_wait(sk_sleep(sk), &wait);
6598
6599	return err;
6600}
6601
6602static void sctp_wait_for_close(struct sock *sk, long timeout)
6603{
6604	DEFINE_WAIT(wait);
6605
6606	do {
6607		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6608		if (list_empty(&sctp_sk(sk)->ep->asocs))
6609			break;
6610		sctp_release_sock(sk);
6611		timeout = schedule_timeout(timeout);
6612		sctp_lock_sock(sk);
6613	} while (!signal_pending(current) && timeout);
6614
6615	finish_wait(sk_sleep(sk), &wait);
6616}
6617
6618static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6619{
6620	struct sk_buff *frag;
6621
6622	if (!skb->data_len)
6623		goto done;
6624
6625	/* Don't forget the fragments. */
6626	skb_walk_frags(skb, frag)
6627		sctp_skb_set_owner_r_frag(frag, sk);
6628
6629done:
6630	sctp_skb_set_owner_r(skb, sk);
6631}
6632
6633void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6634		    struct sctp_association *asoc)
6635{
6636	struct inet_sock *inet = inet_sk(sk);
6637	struct inet_sock *newinet;
6638
6639	newsk->sk_type = sk->sk_type;
6640	newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6641	newsk->sk_flags = sk->sk_flags;
6642	newsk->sk_no_check = sk->sk_no_check;
6643	newsk->sk_reuse = sk->sk_reuse;
6644
6645	newsk->sk_shutdown = sk->sk_shutdown;
6646	newsk->sk_destruct = inet_sock_destruct;
6647	newsk->sk_family = sk->sk_family;
6648	newsk->sk_protocol = IPPROTO_SCTP;
6649	newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6650	newsk->sk_sndbuf = sk->sk_sndbuf;
6651	newsk->sk_rcvbuf = sk->sk_rcvbuf;
6652	newsk->sk_lingertime = sk->sk_lingertime;
6653	newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6654	newsk->sk_sndtimeo = sk->sk_sndtimeo;
6655
6656	newinet = inet_sk(newsk);
6657
6658	/* Initialize sk's sport, dport, rcv_saddr and daddr for
6659	 * getsockname() and getpeername()
6660	 */
6661	newinet->inet_sport = inet->inet_sport;
6662	newinet->inet_saddr = inet->inet_saddr;
6663	newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6664	newinet->inet_dport = htons(asoc->peer.port);
6665	newinet->pmtudisc = inet->pmtudisc;
6666	newinet->inet_id = asoc->next_tsn ^ jiffies;
6667
6668	newinet->uc_ttl = inet->uc_ttl;
6669	newinet->mc_loop = 1;
6670	newinet->mc_ttl = 1;
6671	newinet->mc_index = 0;
6672	newinet->mc_list = NULL;
6673}
6674
6675/* Populate the fields of the newsk from the oldsk and migrate the assoc
6676 * and its messages to the newsk.
6677 */
6678static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6679			      struct sctp_association *assoc,
6680			      sctp_socket_type_t type)
6681{
6682	struct sctp_sock *oldsp = sctp_sk(oldsk);
6683	struct sctp_sock *newsp = sctp_sk(newsk);
6684	struct sctp_bind_bucket *pp; /* hash list port iterator */
6685	struct sctp_endpoint *newep = newsp->ep;
6686	struct sk_buff *skb, *tmp;
6687	struct sctp_ulpevent *event;
6688	struct sctp_bind_hashbucket *head;
6689	struct list_head tmplist;
6690
6691	/* Migrate socket buffer sizes and all the socket level options to the
6692	 * new socket.
6693	 */
6694	newsk->sk_sndbuf = oldsk->sk_sndbuf;
6695	newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6696	/* Brute force copy old sctp opt. */
6697	if (oldsp->do_auto_asconf) {
6698		memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6699		inet_sk_copy_descendant(newsk, oldsk);
6700		memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6701	} else
6702		inet_sk_copy_descendant(newsk, oldsk);
6703
6704	/* Restore the ep value that was overwritten with the above structure
6705	 * copy.
6706	 */
6707	newsp->ep = newep;
6708	newsp->hmac = NULL;
6709
6710	/* Hook this new socket in to the bind_hash list. */
6711	head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6712	sctp_local_bh_disable();
6713	sctp_spin_lock(&head->lock);
 
6714	pp = sctp_sk(oldsk)->bind_hash;
6715	sk_add_bind_node(newsk, &pp->owner);
6716	sctp_sk(newsk)->bind_hash = pp;
6717	inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6718	sctp_spin_unlock(&head->lock);
6719	sctp_local_bh_enable();
6720
6721	/* Copy the bind_addr list from the original endpoint to the new
6722	 * endpoint so that we can handle restarts properly
6723	 */
6724	sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6725				&oldsp->ep->base.bind_addr, GFP_KERNEL);
6726
6727	/* Move any messages in the old socket's receive queue that are for the
6728	 * peeled off association to the new socket's receive queue.
6729	 */
6730	sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6731		event = sctp_skb2event(skb);
6732		if (event->asoc == assoc) {
6733			__skb_unlink(skb, &oldsk->sk_receive_queue);
6734			__skb_queue_tail(&newsk->sk_receive_queue, skb);
6735			sctp_skb_set_owner_r_frag(skb, newsk);
6736		}
6737	}
6738
6739	/* Clean up any messages pending delivery due to partial
6740	 * delivery.   Three cases:
6741	 * 1) No partial deliver;  no work.
6742	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6743	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6744	 */
6745	skb_queue_head_init(&newsp->pd_lobby);
6746	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6747
6748	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6749		struct sk_buff_head *queue;
6750
6751		/* Decide which queue to move pd_lobby skbs to. */
6752		if (assoc->ulpq.pd_mode) {
6753			queue = &newsp->pd_lobby;
6754		} else
6755			queue = &newsk->sk_receive_queue;
6756
6757		/* Walk through the pd_lobby, looking for skbs that
6758		 * need moved to the new socket.
6759		 */
6760		sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6761			event = sctp_skb2event(skb);
6762			if (event->asoc == assoc) {
6763				__skb_unlink(skb, &oldsp->pd_lobby);
6764				__skb_queue_tail(queue, skb);
6765				sctp_skb_set_owner_r_frag(skb, newsk);
6766			}
6767		}
6768
6769		/* Clear up any skbs waiting for the partial
6770		 * delivery to finish.
6771		 */
6772		if (assoc->ulpq.pd_mode)
6773			sctp_clear_pd(oldsk, NULL);
6774
6775	}
6776
6777	sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6778		sctp_skb_set_owner_r_frag(skb, newsk);
6779
6780	sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6781		sctp_skb_set_owner_r_frag(skb, newsk);
6782
6783	/* Set the type of socket to indicate that it is peeled off from the
6784	 * original UDP-style socket or created with the accept() call on a
6785	 * TCP-style socket..
6786	 */
6787	newsp->type = type;
6788
6789	/* Mark the new socket "in-use" by the user so that any packets
6790	 * that may arrive on the association after we've moved it are
6791	 * queued to the backlog.  This prevents a potential race between
6792	 * backlog processing on the old socket and new-packet processing
6793	 * on the new socket.
6794	 *
6795	 * The caller has just allocated newsk so we can guarantee that other
6796	 * paths won't try to lock it and then oldsk.
6797	 */
6798	lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6799	sctp_assoc_migrate(assoc, newsk);
6800
6801	/* If the association on the newsk is already closed before accept()
6802	 * is called, set RCV_SHUTDOWN flag.
6803	 */
6804	if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6805		newsk->sk_shutdown |= RCV_SHUTDOWN;
6806
6807	newsk->sk_state = SCTP_SS_ESTABLISHED;
6808	sctp_release_sock(newsk);
6809}
6810
6811
6812/* This proto struct describes the ULP interface for SCTP.  */
6813struct proto sctp_prot = {
6814	.name        =	"SCTP",
6815	.owner       =	THIS_MODULE,
6816	.close       =	sctp_close,
6817	.connect     =	sctp_connect,
6818	.disconnect  =	sctp_disconnect,
6819	.accept      =	sctp_accept,
6820	.ioctl       =	sctp_ioctl,
6821	.init        =	sctp_init_sock,
6822	.destroy     =	sctp_destroy_sock,
6823	.shutdown    =	sctp_shutdown,
6824	.setsockopt  =	sctp_setsockopt,
6825	.getsockopt  =	sctp_getsockopt,
6826	.sendmsg     =	sctp_sendmsg,
6827	.recvmsg     =	sctp_recvmsg,
6828	.bind        =	sctp_bind,
6829	.backlog_rcv =	sctp_backlog_rcv,
6830	.hash        =	sctp_hash,
6831	.unhash      =	sctp_unhash,
6832	.get_port    =	sctp_get_port,
6833	.obj_size    =  sizeof(struct sctp_sock),
6834	.sysctl_mem  =  sysctl_sctp_mem,
6835	.sysctl_rmem =  sysctl_sctp_rmem,
6836	.sysctl_wmem =  sysctl_sctp_wmem,
6837	.memory_pressure = &sctp_memory_pressure,
6838	.enter_memory_pressure = sctp_enter_memory_pressure,
6839	.memory_allocated = &sctp_memory_allocated,
6840	.sockets_allocated = &sctp_sockets_allocated,
6841};
6842
6843#if IS_ENABLED(CONFIG_IPV6)
6844
6845struct proto sctpv6_prot = {
6846	.name		= "SCTPv6",
6847	.owner		= THIS_MODULE,
6848	.close		= sctp_close,
6849	.connect	= sctp_connect,
6850	.disconnect	= sctp_disconnect,
6851	.accept		= sctp_accept,
6852	.ioctl		= sctp_ioctl,
6853	.init		= sctp_init_sock,
6854	.destroy	= sctp_destroy_sock,
6855	.shutdown	= sctp_shutdown,
6856	.setsockopt	= sctp_setsockopt,
6857	.getsockopt	= sctp_getsockopt,
6858	.sendmsg	= sctp_sendmsg,
6859	.recvmsg	= sctp_recvmsg,
6860	.bind		= sctp_bind,
6861	.backlog_rcv	= sctp_backlog_rcv,
6862	.hash		= sctp_hash,
6863	.unhash		= sctp_unhash,
6864	.get_port	= sctp_get_port,
6865	.obj_size	= sizeof(struct sctp6_sock),
6866	.sysctl_mem	= sysctl_sctp_mem,
6867	.sysctl_rmem	= sysctl_sctp_rmem,
6868	.sysctl_wmem	= sysctl_sctp_wmem,
6869	.memory_pressure = &sctp_memory_pressure,
6870	.enter_memory_pressure = sctp_enter_memory_pressure,
6871	.memory_allocated = &sctp_memory_allocated,
6872	.sockets_allocated = &sctp_sockets_allocated,
6873};
6874#endif /* IS_ENABLED(CONFIG_IPV6) */
v3.15
   1/* SCTP kernel implementation
   2 * (C) Copyright IBM Corp. 2001, 2004
   3 * Copyright (c) 1999-2000 Cisco, Inc.
   4 * Copyright (c) 1999-2001 Motorola, Inc.
   5 * Copyright (c) 2001-2003 Intel Corp.
   6 * Copyright (c) 2001-2002 Nokia, Inc.
   7 * Copyright (c) 2001 La Monte H.P. Yarroll
   8 *
   9 * This file is part of the SCTP kernel implementation
  10 *
  11 * These functions interface with the sockets layer to implement the
  12 * SCTP Extensions for the Sockets API.
  13 *
  14 * Note that the descriptions from the specification are USER level
  15 * functions--this file is the functions which populate the struct proto
  16 * for SCTP which is the BOTTOM of the sockets interface.
  17 *
  18 * This SCTP implementation is free software;
  19 * you can redistribute it and/or modify it under the terms of
  20 * the GNU General Public License as published by
  21 * the Free Software Foundation; either version 2, or (at your option)
  22 * any later version.
  23 *
  24 * This SCTP implementation is distributed in the hope that it
  25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
  26 *                 ************************
  27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  28 * See the GNU General Public License for more details.
  29 *
  30 * You should have received a copy of the GNU General Public License
  31 * along with GNU CC; see the file COPYING.  If not, see
  32 * <http://www.gnu.org/licenses/>.
 
  33 *
  34 * Please send any bug reports or fixes you make to the
  35 * email address(es):
  36 *    lksctp developers <linux-sctp@vger.kernel.org>
 
 
 
  37 *
  38 * Written or modified by:
  39 *    La Monte H.P. Yarroll <piggy@acm.org>
  40 *    Narasimha Budihal     <narsi@refcode.org>
  41 *    Karl Knutson          <karl@athena.chicago.il.us>
  42 *    Jon Grimm             <jgrimm@us.ibm.com>
  43 *    Xingang Guo           <xingang.guo@intel.com>
  44 *    Daisy Chang           <daisyc@us.ibm.com>
  45 *    Sridhar Samudrala     <samudrala@us.ibm.com>
  46 *    Inaky Perez-Gonzalez  <inaky.gonzalez@intel.com>
  47 *    Ardelle Fan	    <ardelle.fan@intel.com>
  48 *    Ryan Layer	    <rmlayer@us.ibm.com>
  49 *    Anup Pemmaiah         <pemmaiah@cc.usu.edu>
  50 *    Kevin Gao             <kevin.gao@intel.com>
 
 
 
  51 */
  52
  53#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  54
  55#include <linux/types.h>
  56#include <linux/kernel.h>
  57#include <linux/wait.h>
  58#include <linux/time.h>
  59#include <linux/ip.h>
  60#include <linux/capability.h>
  61#include <linux/fcntl.h>
  62#include <linux/poll.h>
  63#include <linux/init.h>
  64#include <linux/crypto.h>
  65#include <linux/slab.h>
  66#include <linux/file.h>
  67#include <linux/compat.h>
  68
  69#include <net/ip.h>
  70#include <net/icmp.h>
  71#include <net/route.h>
  72#include <net/ipv6.h>
  73#include <net/inet_common.h>
  74
  75#include <linux/socket.h> /* for sa_family_t */
  76#include <linux/export.h>
  77#include <net/sock.h>
  78#include <net/sctp/sctp.h>
  79#include <net/sctp/sm.h>
  80
 
 
 
 
 
  81/* Forward declarations for internal helper functions. */
  82static int sctp_writeable(struct sock *sk);
  83static void sctp_wfree(struct sk_buff *skb);
  84static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
  85				size_t msg_len);
  86static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
  87static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
  88static int sctp_wait_for_accept(struct sock *sk, long timeo);
  89static void sctp_wait_for_close(struct sock *sk, long timeo);
  90static void sctp_destruct_sock(struct sock *sk);
  91static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
  92					union sctp_addr *addr, int len);
  93static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
  94static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
  95static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
  96static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
  97static int sctp_send_asconf(struct sctp_association *asoc,
  98			    struct sctp_chunk *chunk);
  99static int sctp_do_bind(struct sock *, union sctp_addr *, int);
 100static int sctp_autobind(struct sock *sk);
 101static void sctp_sock_migrate(struct sock *, struct sock *,
 102			      struct sctp_association *, sctp_socket_type_t);
 
 103
 104extern struct kmem_cache *sctp_bucket_cachep;
 105extern long sysctl_sctp_mem[3];
 106extern int sysctl_sctp_rmem[3];
 107extern int sysctl_sctp_wmem[3];
 108
 109static int sctp_memory_pressure;
 110static atomic_long_t sctp_memory_allocated;
 111struct percpu_counter sctp_sockets_allocated;
 112
 113static void sctp_enter_memory_pressure(struct sock *sk)
 114{
 115	sctp_memory_pressure = 1;
 116}
 117
 118
 119/* Get the sndbuf space available at the time on the association.  */
 120static inline int sctp_wspace(struct sctp_association *asoc)
 121{
 122	int amt;
 123
 124	if (asoc->ep->sndbuf_policy)
 125		amt = asoc->sndbuf_used;
 126	else
 127		amt = sk_wmem_alloc_get(asoc->base.sk);
 128
 129	if (amt >= asoc->base.sk->sk_sndbuf) {
 130		if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
 131			amt = 0;
 132		else {
 133			amt = sk_stream_wspace(asoc->base.sk);
 134			if (amt < 0)
 135				amt = 0;
 136		}
 137	} else {
 138		amt = asoc->base.sk->sk_sndbuf - amt;
 139	}
 140	return amt;
 141}
 142
 143/* Increment the used sndbuf space count of the corresponding association by
 144 * the size of the outgoing data chunk.
 145 * Also, set the skb destructor for sndbuf accounting later.
 146 *
 147 * Since it is always 1-1 between chunk and skb, and also a new skb is always
 148 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
 149 * destructor in the data chunk skb for the purpose of the sndbuf space
 150 * tracking.
 151 */
 152static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
 153{
 154	struct sctp_association *asoc = chunk->asoc;
 155	struct sock *sk = asoc->base.sk;
 156
 157	/* The sndbuf space is tracked per association.  */
 158	sctp_association_hold(asoc);
 159
 160	skb_set_owner_w(chunk->skb, sk);
 161
 162	chunk->skb->destructor = sctp_wfree;
 163	/* Save the chunk pointer in skb for sctp_wfree to use later.  */
 164	*((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
 165
 166	asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
 167				sizeof(struct sk_buff) +
 168				sizeof(struct sctp_chunk);
 169
 170	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
 171	sk->sk_wmem_queued += chunk->skb->truesize;
 172	sk_mem_charge(sk, chunk->skb->truesize);
 173}
 174
 175/* Verify that this is a valid address. */
 176static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
 177				   int len)
 178{
 179	struct sctp_af *af;
 180
 181	/* Verify basic sockaddr. */
 182	af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
 183	if (!af)
 184		return -EINVAL;
 185
 186	/* Is this a valid SCTP address?  */
 187	if (!af->addr_valid(addr, sctp_sk(sk), NULL))
 188		return -EINVAL;
 189
 190	if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
 191		return -EINVAL;
 192
 193	return 0;
 194}
 195
 196/* Look up the association by its id.  If this is not a UDP-style
 197 * socket, the ID field is always ignored.
 198 */
 199struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
 200{
 201	struct sctp_association *asoc = NULL;
 202
 203	/* If this is not a UDP-style socket, assoc id should be ignored. */
 204	if (!sctp_style(sk, UDP)) {
 205		/* Return NULL if the socket state is not ESTABLISHED. It
 206		 * could be a TCP-style listening socket or a socket which
 207		 * hasn't yet called connect() to establish an association.
 208		 */
 209		if (!sctp_sstate(sk, ESTABLISHED))
 210			return NULL;
 211
 212		/* Get the first and the only association from the list. */
 213		if (!list_empty(&sctp_sk(sk)->ep->asocs))
 214			asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
 215					  struct sctp_association, asocs);
 216		return asoc;
 217	}
 218
 219	/* Otherwise this is a UDP-style socket. */
 220	if (!id || (id == (sctp_assoc_t)-1))
 221		return NULL;
 222
 223	spin_lock_bh(&sctp_assocs_id_lock);
 224	asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
 225	spin_unlock_bh(&sctp_assocs_id_lock);
 226
 227	if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
 228		return NULL;
 229
 230	return asoc;
 231}
 232
 233/* Look up the transport from an address and an assoc id. If both address and
 234 * id are specified, the associations matching the address and the id should be
 235 * the same.
 236 */
 237static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
 238					      struct sockaddr_storage *addr,
 239					      sctp_assoc_t id)
 240{
 241	struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
 242	struct sctp_transport *transport;
 243	union sctp_addr *laddr = (union sctp_addr *)addr;
 244
 245	addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
 246					       laddr,
 247					       &transport);
 248
 249	if (!addr_asoc)
 250		return NULL;
 251
 252	id_asoc = sctp_id2assoc(sk, id);
 253	if (id_asoc && (id_asoc != addr_asoc))
 254		return NULL;
 255
 256	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
 257						(union sctp_addr *)addr);
 258
 259	return transport;
 260}
 261
 262/* API 3.1.2 bind() - UDP Style Syntax
 263 * The syntax of bind() is,
 264 *
 265 *   ret = bind(int sd, struct sockaddr *addr, int addrlen);
 266 *
 267 *   sd      - the socket descriptor returned by socket().
 268 *   addr    - the address structure (struct sockaddr_in or struct
 269 *             sockaddr_in6 [RFC 2553]),
 270 *   addr_len - the size of the address structure.
 271 */
 272static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
 273{
 274	int retval = 0;
 275
 276	lock_sock(sk);
 277
 278	pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
 279		 addr, addr_len);
 280
 281	/* Disallow binding twice. */
 282	if (!sctp_sk(sk)->ep->base.bind_addr.port)
 283		retval = sctp_do_bind(sk, (union sctp_addr *)addr,
 284				      addr_len);
 285	else
 286		retval = -EINVAL;
 287
 288	release_sock(sk);
 289
 290	return retval;
 291}
 292
 293static long sctp_get_port_local(struct sock *, union sctp_addr *);
 294
 295/* Verify this is a valid sockaddr. */
 296static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
 297					union sctp_addr *addr, int len)
 298{
 299	struct sctp_af *af;
 300
 301	/* Check minimum size.  */
 302	if (len < sizeof (struct sockaddr))
 303		return NULL;
 304
 305	/* V4 mapped address are really of AF_INET family */
 306	if (addr->sa.sa_family == AF_INET6 &&
 307	    ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
 308		if (!opt->pf->af_supported(AF_INET, opt))
 309			return NULL;
 310	} else {
 311		/* Does this PF support this AF? */
 312		if (!opt->pf->af_supported(addr->sa.sa_family, opt))
 313			return NULL;
 314	}
 315
 316	/* If we get this far, af is valid. */
 317	af = sctp_get_af_specific(addr->sa.sa_family);
 318
 319	if (len < af->sockaddr_len)
 320		return NULL;
 321
 322	return af;
 323}
 324
 325/* Bind a local address either to an endpoint or to an association.  */
 326static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
 327{
 328	struct net *net = sock_net(sk);
 329	struct sctp_sock *sp = sctp_sk(sk);
 330	struct sctp_endpoint *ep = sp->ep;
 331	struct sctp_bind_addr *bp = &ep->base.bind_addr;
 332	struct sctp_af *af;
 333	unsigned short snum;
 334	int ret = 0;
 335
 336	/* Common sockaddr verification. */
 337	af = sctp_sockaddr_af(sp, addr, len);
 338	if (!af) {
 339		pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
 340			 __func__, sk, addr, len);
 341		return -EINVAL;
 342	}
 343
 344	snum = ntohs(addr->v4.sin_port);
 345
 346	pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
 347		 __func__, sk, &addr->sa, bp->port, snum, len);
 
 
 
 
 348
 349	/* PF specific bind() address verification. */
 350	if (!sp->pf->bind_verify(sp, addr))
 351		return -EADDRNOTAVAIL;
 352
 353	/* We must either be unbound, or bind to the same port.
 354	 * It's OK to allow 0 ports if we are already bound.
 355	 * We'll just inhert an already bound port in this case
 356	 */
 357	if (bp->port) {
 358		if (!snum)
 359			snum = bp->port;
 360		else if (snum != bp->port) {
 361			pr_debug("%s: new port %d doesn't match existing port "
 362				 "%d\n", __func__, snum, bp->port);
 
 363			return -EINVAL;
 364		}
 365	}
 366
 367	if (snum && snum < PROT_SOCK &&
 368	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 369		return -EACCES;
 370
 371	/* See if the address matches any of the addresses we may have
 372	 * already bound before checking against other endpoints.
 373	 */
 374	if (sctp_bind_addr_match(bp, addr, sp))
 375		return -EINVAL;
 376
 377	/* Make sure we are allowed to bind here.
 378	 * The function sctp_get_port_local() does duplicate address
 379	 * detection.
 380	 */
 381	addr->v4.sin_port = htons(snum);
 382	if ((ret = sctp_get_port_local(sk, addr))) {
 383		return -EADDRINUSE;
 384	}
 385
 386	/* Refresh ephemeral port.  */
 387	if (!bp->port)
 388		bp->port = inet_sk(sk)->inet_num;
 389
 390	/* Add the address to the bind address list.
 391	 * Use GFP_ATOMIC since BHs will be disabled.
 392	 */
 393	ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
 394
 395	/* Copy back into socket for getsockname() use. */
 396	if (!ret) {
 397		inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
 398		af->to_sk_saddr(addr, sk);
 399	}
 400
 401	return ret;
 402}
 403
 404 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
 405 *
 406 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
 407 * at any one time.  If a sender, after sending an ASCONF chunk, decides
 408 * it needs to transfer another ASCONF Chunk, it MUST wait until the
 409 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
 410 * subsequent ASCONF. Note this restriction binds each side, so at any
 411 * time two ASCONF may be in-transit on any given association (one sent
 412 * from each endpoint).
 413 */
 414static int sctp_send_asconf(struct sctp_association *asoc,
 415			    struct sctp_chunk *chunk)
 416{
 417	struct net 	*net = sock_net(asoc->base.sk);
 418	int		retval = 0;
 419
 420	/* If there is an outstanding ASCONF chunk, queue it for later
 421	 * transmission.
 422	 */
 423	if (asoc->addip_last_asconf) {
 424		list_add_tail(&chunk->list, &asoc->addip_chunk_list);
 425		goto out;
 426	}
 427
 428	/* Hold the chunk until an ASCONF_ACK is received. */
 429	sctp_chunk_hold(chunk);
 430	retval = sctp_primitive_ASCONF(net, asoc, chunk);
 431	if (retval)
 432		sctp_chunk_free(chunk);
 433	else
 434		asoc->addip_last_asconf = chunk;
 435
 436out:
 437	return retval;
 438}
 439
 440/* Add a list of addresses as bind addresses to local endpoint or
 441 * association.
 442 *
 443 * Basically run through each address specified in the addrs/addrcnt
 444 * array/length pair, determine if it is IPv6 or IPv4 and call
 445 * sctp_do_bind() on it.
 446 *
 447 * If any of them fails, then the operation will be reversed and the
 448 * ones that were added will be removed.
 449 *
 450 * Only sctp_setsockopt_bindx() is supposed to call this function.
 451 */
 452static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
 453{
 454	int cnt;
 455	int retval = 0;
 456	void *addr_buf;
 457	struct sockaddr *sa_addr;
 458	struct sctp_af *af;
 459
 460	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
 461		 addrs, addrcnt);
 462
 463	addr_buf = addrs;
 464	for (cnt = 0; cnt < addrcnt; cnt++) {
 465		/* The list may contain either IPv4 or IPv6 address;
 466		 * determine the address length for walking thru the list.
 467		 */
 468		sa_addr = addr_buf;
 469		af = sctp_get_af_specific(sa_addr->sa_family);
 470		if (!af) {
 471			retval = -EINVAL;
 472			goto err_bindx_add;
 473		}
 474
 475		retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
 476				      af->sockaddr_len);
 477
 478		addr_buf += af->sockaddr_len;
 479
 480err_bindx_add:
 481		if (retval < 0) {
 482			/* Failed. Cleanup the ones that have been added */
 483			if (cnt > 0)
 484				sctp_bindx_rem(sk, addrs, cnt);
 485			return retval;
 486		}
 487	}
 488
 489	return retval;
 490}
 491
 492/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
 493 * associations that are part of the endpoint indicating that a list of local
 494 * addresses are added to the endpoint.
 495 *
 496 * If any of the addresses is already in the bind address list of the
 497 * association, we do not send the chunk for that association.  But it will not
 498 * affect other associations.
 499 *
 500 * Only sctp_setsockopt_bindx() is supposed to call this function.
 501 */
 502static int sctp_send_asconf_add_ip(struct sock		*sk,
 503				   struct sockaddr	*addrs,
 504				   int 			addrcnt)
 505{
 506	struct net *net = sock_net(sk);
 507	struct sctp_sock		*sp;
 508	struct sctp_endpoint		*ep;
 509	struct sctp_association		*asoc;
 510	struct sctp_bind_addr		*bp;
 511	struct sctp_chunk		*chunk;
 512	struct sctp_sockaddr_entry	*laddr;
 513	union sctp_addr			*addr;
 514	union sctp_addr			saveaddr;
 515	void				*addr_buf;
 516	struct sctp_af			*af;
 517	struct list_head		*p;
 518	int 				i;
 519	int 				retval = 0;
 520
 521	if (!net->sctp.addip_enable)
 522		return retval;
 523
 524	sp = sctp_sk(sk);
 525	ep = sp->ep;
 526
 527	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
 528		 __func__, sk, addrs, addrcnt);
 529
 530	list_for_each_entry(asoc, &ep->asocs, asocs) {
 
 531		if (!asoc->peer.asconf_capable)
 532			continue;
 533
 534		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
 535			continue;
 536
 537		if (!sctp_state(asoc, ESTABLISHED))
 538			continue;
 539
 540		/* Check if any address in the packed array of addresses is
 541		 * in the bind address list of the association. If so,
 542		 * do not send the asconf chunk to its peer, but continue with
 543		 * other associations.
 544		 */
 545		addr_buf = addrs;
 546		for (i = 0; i < addrcnt; i++) {
 547			addr = addr_buf;
 548			af = sctp_get_af_specific(addr->v4.sin_family);
 549			if (!af) {
 550				retval = -EINVAL;
 551				goto out;
 552			}
 553
 554			if (sctp_assoc_lookup_laddr(asoc, addr))
 555				break;
 556
 557			addr_buf += af->sockaddr_len;
 558		}
 559		if (i < addrcnt)
 560			continue;
 561
 562		/* Use the first valid address in bind addr list of
 563		 * association as Address Parameter of ASCONF CHUNK.
 564		 */
 565		bp = &asoc->base.bind_addr;
 566		p = bp->address_list.next;
 567		laddr = list_entry(p, struct sctp_sockaddr_entry, list);
 568		chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
 569						   addrcnt, SCTP_PARAM_ADD_IP);
 570		if (!chunk) {
 571			retval = -ENOMEM;
 572			goto out;
 573		}
 574
 575		/* Add the new addresses to the bind address list with
 576		 * use_as_src set to 0.
 577		 */
 578		addr_buf = addrs;
 579		for (i = 0; i < addrcnt; i++) {
 580			addr = addr_buf;
 581			af = sctp_get_af_specific(addr->v4.sin_family);
 582			memcpy(&saveaddr, addr, af->sockaddr_len);
 583			retval = sctp_add_bind_addr(bp, &saveaddr,
 584						    SCTP_ADDR_NEW, GFP_ATOMIC);
 585			addr_buf += af->sockaddr_len;
 586		}
 587		if (asoc->src_out_of_asoc_ok) {
 588			struct sctp_transport *trans;
 589
 590			list_for_each_entry(trans,
 591			    &asoc->peer.transport_addr_list, transports) {
 592				/* Clear the source and route cache */
 593				dst_release(trans->dst);
 594				trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
 595				    2*asoc->pathmtu, 4380));
 596				trans->ssthresh = asoc->peer.i.a_rwnd;
 597				trans->rto = asoc->rto_initial;
 598				sctp_max_rto(asoc, trans);
 599				trans->rtt = trans->srtt = trans->rttvar = 0;
 600				sctp_transport_route(trans, NULL,
 601				    sctp_sk(asoc->base.sk));
 602			}
 603		}
 604		retval = sctp_send_asconf(asoc, chunk);
 605	}
 606
 607out:
 608	return retval;
 609}
 610
 611/* Remove a list of addresses from bind addresses list.  Do not remove the
 612 * last address.
 613 *
 614 * Basically run through each address specified in the addrs/addrcnt
 615 * array/length pair, determine if it is IPv6 or IPv4 and call
 616 * sctp_del_bind() on it.
 617 *
 618 * If any of them fails, then the operation will be reversed and the
 619 * ones that were removed will be added back.
 620 *
 621 * At least one address has to be left; if only one address is
 622 * available, the operation will return -EBUSY.
 623 *
 624 * Only sctp_setsockopt_bindx() is supposed to call this function.
 625 */
 626static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
 627{
 628	struct sctp_sock *sp = sctp_sk(sk);
 629	struct sctp_endpoint *ep = sp->ep;
 630	int cnt;
 631	struct sctp_bind_addr *bp = &ep->base.bind_addr;
 632	int retval = 0;
 633	void *addr_buf;
 634	union sctp_addr *sa_addr;
 635	struct sctp_af *af;
 636
 637	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
 638		 __func__, sk, addrs, addrcnt);
 639
 640	addr_buf = addrs;
 641	for (cnt = 0; cnt < addrcnt; cnt++) {
 642		/* If the bind address list is empty or if there is only one
 643		 * bind address, there is nothing more to be removed (we need
 644		 * at least one address here).
 645		 */
 646		if (list_empty(&bp->address_list) ||
 647		    (sctp_list_single_entry(&bp->address_list))) {
 648			retval = -EBUSY;
 649			goto err_bindx_rem;
 650		}
 651
 652		sa_addr = addr_buf;
 653		af = sctp_get_af_specific(sa_addr->sa.sa_family);
 654		if (!af) {
 655			retval = -EINVAL;
 656			goto err_bindx_rem;
 657		}
 658
 659		if (!af->addr_valid(sa_addr, sp, NULL)) {
 660			retval = -EADDRNOTAVAIL;
 661			goto err_bindx_rem;
 662		}
 663
 664		if (sa_addr->v4.sin_port &&
 665		    sa_addr->v4.sin_port != htons(bp->port)) {
 666			retval = -EINVAL;
 667			goto err_bindx_rem;
 668		}
 669
 670		if (!sa_addr->v4.sin_port)
 671			sa_addr->v4.sin_port = htons(bp->port);
 672
 673		/* FIXME - There is probably a need to check if sk->sk_saddr and
 674		 * sk->sk_rcv_addr are currently set to one of the addresses to
 675		 * be removed. This is something which needs to be looked into
 676		 * when we are fixing the outstanding issues with multi-homing
 677		 * socket routing and failover schemes. Refer to comments in
 678		 * sctp_do_bind(). -daisy
 679		 */
 680		retval = sctp_del_bind_addr(bp, sa_addr);
 681
 682		addr_buf += af->sockaddr_len;
 683err_bindx_rem:
 684		if (retval < 0) {
 685			/* Failed. Add the ones that has been removed back */
 686			if (cnt > 0)
 687				sctp_bindx_add(sk, addrs, cnt);
 688			return retval;
 689		}
 690	}
 691
 692	return retval;
 693}
 694
 695/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
 696 * the associations that are part of the endpoint indicating that a list of
 697 * local addresses are removed from the endpoint.
 698 *
 699 * If any of the addresses is already in the bind address list of the
 700 * association, we do not send the chunk for that association.  But it will not
 701 * affect other associations.
 702 *
 703 * Only sctp_setsockopt_bindx() is supposed to call this function.
 704 */
 705static int sctp_send_asconf_del_ip(struct sock		*sk,
 706				   struct sockaddr	*addrs,
 707				   int			addrcnt)
 708{
 709	struct net *net = sock_net(sk);
 710	struct sctp_sock	*sp;
 711	struct sctp_endpoint	*ep;
 712	struct sctp_association	*asoc;
 713	struct sctp_transport	*transport;
 714	struct sctp_bind_addr	*bp;
 715	struct sctp_chunk	*chunk;
 716	union sctp_addr		*laddr;
 717	void			*addr_buf;
 718	struct sctp_af		*af;
 719	struct sctp_sockaddr_entry *saddr;
 720	int 			i;
 721	int 			retval = 0;
 722	int			stored = 0;
 723
 724	chunk = NULL;
 725	if (!net->sctp.addip_enable)
 726		return retval;
 727
 728	sp = sctp_sk(sk);
 729	ep = sp->ep;
 730
 731	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
 732		 __func__, sk, addrs, addrcnt);
 733
 734	list_for_each_entry(asoc, &ep->asocs, asocs) {
 735
 736		if (!asoc->peer.asconf_capable)
 737			continue;
 738
 739		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
 740			continue;
 741
 742		if (!sctp_state(asoc, ESTABLISHED))
 743			continue;
 744
 745		/* Check if any address in the packed array of addresses is
 746		 * not present in the bind address list of the association.
 747		 * If so, do not send the asconf chunk to its peer, but
 748		 * continue with other associations.
 749		 */
 750		addr_buf = addrs;
 751		for (i = 0; i < addrcnt; i++) {
 752			laddr = addr_buf;
 753			af = sctp_get_af_specific(laddr->v4.sin_family);
 754			if (!af) {
 755				retval = -EINVAL;
 756				goto out;
 757			}
 758
 759			if (!sctp_assoc_lookup_laddr(asoc, laddr))
 760				break;
 761
 762			addr_buf += af->sockaddr_len;
 763		}
 764		if (i < addrcnt)
 765			continue;
 766
 767		/* Find one address in the association's bind address list
 768		 * that is not in the packed array of addresses. This is to
 769		 * make sure that we do not delete all the addresses in the
 770		 * association.
 771		 */
 772		bp = &asoc->base.bind_addr;
 773		laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
 774					       addrcnt, sp);
 775		if ((laddr == NULL) && (addrcnt == 1)) {
 776			if (asoc->asconf_addr_del_pending)
 777				continue;
 778			asoc->asconf_addr_del_pending =
 779			    kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
 780			if (asoc->asconf_addr_del_pending == NULL) {
 781				retval = -ENOMEM;
 782				goto out;
 783			}
 784			asoc->asconf_addr_del_pending->sa.sa_family =
 785				    addrs->sa_family;
 786			asoc->asconf_addr_del_pending->v4.sin_port =
 787				    htons(bp->port);
 788			if (addrs->sa_family == AF_INET) {
 789				struct sockaddr_in *sin;
 790
 791				sin = (struct sockaddr_in *)addrs;
 792				asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
 793			} else if (addrs->sa_family == AF_INET6) {
 794				struct sockaddr_in6 *sin6;
 795
 796				sin6 = (struct sockaddr_in6 *)addrs;
 797				asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
 798			}
 799
 800			pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
 801				 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
 802				 asoc->asconf_addr_del_pending);
 803
 804			asoc->src_out_of_asoc_ok = 1;
 805			stored = 1;
 806			goto skip_mkasconf;
 807		}
 808
 809		if (laddr == NULL)
 810			return -EINVAL;
 811
 812		/* We do not need RCU protection throughout this loop
 813		 * because this is done under a socket lock from the
 814		 * setsockopt call.
 815		 */
 816		chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
 817						   SCTP_PARAM_DEL_IP);
 818		if (!chunk) {
 819			retval = -ENOMEM;
 820			goto out;
 821		}
 822
 823skip_mkasconf:
 824		/* Reset use_as_src flag for the addresses in the bind address
 825		 * list that are to be deleted.
 826		 */
 827		addr_buf = addrs;
 828		for (i = 0; i < addrcnt; i++) {
 829			laddr = addr_buf;
 830			af = sctp_get_af_specific(laddr->v4.sin_family);
 831			list_for_each_entry(saddr, &bp->address_list, list) {
 832				if (sctp_cmp_addr_exact(&saddr->a, laddr))
 833					saddr->state = SCTP_ADDR_DEL;
 834			}
 835			addr_buf += af->sockaddr_len;
 836		}
 837
 838		/* Update the route and saddr entries for all the transports
 839		 * as some of the addresses in the bind address list are
 840		 * about to be deleted and cannot be used as source addresses.
 841		 */
 842		list_for_each_entry(transport, &asoc->peer.transport_addr_list,
 843					transports) {
 844			dst_release(transport->dst);
 845			sctp_transport_route(transport, NULL,
 846					     sctp_sk(asoc->base.sk));
 847		}
 848
 849		if (stored)
 850			/* We don't need to transmit ASCONF */
 851			continue;
 852		retval = sctp_send_asconf(asoc, chunk);
 853	}
 854out:
 855	return retval;
 856}
 857
 858/* set addr events to assocs in the endpoint.  ep and addr_wq must be locked */
 859int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
 860{
 861	struct sock *sk = sctp_opt2sk(sp);
 862	union sctp_addr *addr;
 863	struct sctp_af *af;
 864
 865	/* It is safe to write port space in caller. */
 866	addr = &addrw->a;
 867	addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
 868	af = sctp_get_af_specific(addr->sa.sa_family);
 869	if (!af)
 870		return -EINVAL;
 871	if (sctp_verify_addr(sk, addr, af->sockaddr_len))
 872		return -EINVAL;
 873
 874	if (addrw->state == SCTP_ADDR_NEW)
 875		return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
 876	else
 877		return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
 878}
 879
 880/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
 881 *
 882 * API 8.1
 883 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
 884 *                int flags);
 885 *
 886 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
 887 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
 888 * or IPv6 addresses.
 889 *
 890 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
 891 * Section 3.1.2 for this usage.
 892 *
 893 * addrs is a pointer to an array of one or more socket addresses. Each
 894 * address is contained in its appropriate structure (i.e. struct
 895 * sockaddr_in or struct sockaddr_in6) the family of the address type
 896 * must be used to distinguish the address length (note that this
 897 * representation is termed a "packed array" of addresses). The caller
 898 * specifies the number of addresses in the array with addrcnt.
 899 *
 900 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
 901 * -1, and sets errno to the appropriate error code.
 902 *
 903 * For SCTP, the port given in each socket address must be the same, or
 904 * sctp_bindx() will fail, setting errno to EINVAL.
 905 *
 906 * The flags parameter is formed from the bitwise OR of zero or more of
 907 * the following currently defined flags:
 908 *
 909 * SCTP_BINDX_ADD_ADDR
 910 *
 911 * SCTP_BINDX_REM_ADDR
 912 *
 913 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
 914 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
 915 * addresses from the association. The two flags are mutually exclusive;
 916 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
 917 * not remove all addresses from an association; sctp_bindx() will
 918 * reject such an attempt with EINVAL.
 919 *
 920 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
 921 * additional addresses with an endpoint after calling bind().  Or use
 922 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
 923 * socket is associated with so that no new association accepted will be
 924 * associated with those addresses. If the endpoint supports dynamic
 925 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
 926 * endpoint to send the appropriate message to the peer to change the
 927 * peers address lists.
 928 *
 929 * Adding and removing addresses from a connected association is
 930 * optional functionality. Implementations that do not support this
 931 * functionality should return EOPNOTSUPP.
 932 *
 933 * Basically do nothing but copying the addresses from user to kernel
 934 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
 935 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
 936 * from userspace.
 937 *
 938 * We don't use copy_from_user() for optimization: we first do the
 939 * sanity checks (buffer size -fast- and access check-healthy
 940 * pointer); if all of those succeed, then we can alloc the memory
 941 * (expensive operation) needed to copy the data to kernel. Then we do
 942 * the copying without checking the user space area
 943 * (__copy_from_user()).
 944 *
 945 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
 946 * it.
 947 *
 948 * sk        The sk of the socket
 949 * addrs     The pointer to the addresses in user land
 950 * addrssize Size of the addrs buffer
 951 * op        Operation to perform (add or remove, see the flags of
 952 *           sctp_bindx)
 953 *
 954 * Returns 0 if ok, <0 errno code on error.
 955 */
 956static int sctp_setsockopt_bindx(struct sock *sk,
 957				 struct sockaddr __user *addrs,
 958				 int addrs_size, int op)
 959{
 960	struct sockaddr *kaddrs;
 961	int err;
 962	int addrcnt = 0;
 963	int walk_size = 0;
 964	struct sockaddr *sa_addr;
 965	void *addr_buf;
 966	struct sctp_af *af;
 967
 968	pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
 969		 __func__, sk, addrs, addrs_size, op);
 970
 971	if (unlikely(addrs_size <= 0))
 972		return -EINVAL;
 973
 974	/* Check the user passed a healthy pointer.  */
 975	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
 976		return -EFAULT;
 977
 978	/* Alloc space for the address array in kernel memory.  */
 979	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
 980	if (unlikely(!kaddrs))
 981		return -ENOMEM;
 982
 983	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
 984		kfree(kaddrs);
 985		return -EFAULT;
 986	}
 987
 988	/* Walk through the addrs buffer and count the number of addresses. */
 989	addr_buf = kaddrs;
 990	while (walk_size < addrs_size) {
 991		if (walk_size + sizeof(sa_family_t) > addrs_size) {
 992			kfree(kaddrs);
 993			return -EINVAL;
 994		}
 995
 996		sa_addr = addr_buf;
 997		af = sctp_get_af_specific(sa_addr->sa_family);
 998
 999		/* If the address family is not supported or if this address
1000		 * causes the address buffer to overflow return EINVAL.
1001		 */
1002		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1003			kfree(kaddrs);
1004			return -EINVAL;
1005		}
1006		addrcnt++;
1007		addr_buf += af->sockaddr_len;
1008		walk_size += af->sockaddr_len;
1009	}
1010
1011	/* Do the work. */
1012	switch (op) {
1013	case SCTP_BINDX_ADD_ADDR:
1014		err = sctp_bindx_add(sk, kaddrs, addrcnt);
1015		if (err)
1016			goto out;
1017		err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1018		break;
1019
1020	case SCTP_BINDX_REM_ADDR:
1021		err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1022		if (err)
1023			goto out;
1024		err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1025		break;
1026
1027	default:
1028		err = -EINVAL;
1029		break;
1030	}
1031
1032out:
1033	kfree(kaddrs);
1034
1035	return err;
1036}
1037
1038/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1039 *
1040 * Common routine for handling connect() and sctp_connectx().
1041 * Connect will come in with just a single address.
1042 */
1043static int __sctp_connect(struct sock *sk,
1044			  struct sockaddr *kaddrs,
1045			  int addrs_size,
1046			  sctp_assoc_t *assoc_id)
1047{
1048	struct net *net = sock_net(sk);
1049	struct sctp_sock *sp;
1050	struct sctp_endpoint *ep;
1051	struct sctp_association *asoc = NULL;
1052	struct sctp_association *asoc2;
1053	struct sctp_transport *transport;
1054	union sctp_addr to;
1055	struct sctp_af *af;
1056	sctp_scope_t scope;
1057	long timeo;
1058	int err = 0;
1059	int addrcnt = 0;
1060	int walk_size = 0;
1061	union sctp_addr *sa_addr = NULL;
1062	void *addr_buf;
1063	unsigned short port;
1064	unsigned int f_flags = 0;
1065
1066	sp = sctp_sk(sk);
1067	ep = sp->ep;
1068
1069	/* connect() cannot be done on a socket that is already in ESTABLISHED
1070	 * state - UDP-style peeled off socket or a TCP-style socket that
1071	 * is already connected.
1072	 * It cannot be done even on a TCP-style listening socket.
1073	 */
1074	if (sctp_sstate(sk, ESTABLISHED) ||
1075	    (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1076		err = -EISCONN;
1077		goto out_free;
1078	}
1079
1080	/* Walk through the addrs buffer and count the number of addresses. */
1081	addr_buf = kaddrs;
1082	while (walk_size < addrs_size) {
1083		if (walk_size + sizeof(sa_family_t) > addrs_size) {
1084			err = -EINVAL;
1085			goto out_free;
1086		}
1087
1088		sa_addr = addr_buf;
1089		af = sctp_get_af_specific(sa_addr->sa.sa_family);
1090
1091		/* If the address family is not supported or if this address
1092		 * causes the address buffer to overflow return EINVAL.
1093		 */
1094		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1095			err = -EINVAL;
1096			goto out_free;
1097		}
1098
1099		port = ntohs(sa_addr->v4.sin_port);
1100
1101		/* Save current address so we can work with it */
1102		memcpy(&to, sa_addr, af->sockaddr_len);
1103
1104		err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1105		if (err)
1106			goto out_free;
1107
1108		/* Make sure the destination port is correctly set
1109		 * in all addresses.
1110		 */
1111		if (asoc && asoc->peer.port && asoc->peer.port != port) {
1112			err = -EINVAL;
1113			goto out_free;
1114		}
1115
1116		/* Check if there already is a matching association on the
1117		 * endpoint (other than the one created here).
1118		 */
1119		asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1120		if (asoc2 && asoc2 != asoc) {
1121			if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1122				err = -EISCONN;
1123			else
1124				err = -EALREADY;
1125			goto out_free;
1126		}
1127
1128		/* If we could not find a matching association on the endpoint,
1129		 * make sure that there is no peeled-off association matching
1130		 * the peer address even on another socket.
1131		 */
1132		if (sctp_endpoint_is_peeled_off(ep, &to)) {
1133			err = -EADDRNOTAVAIL;
1134			goto out_free;
1135		}
1136
1137		if (!asoc) {
1138			/* If a bind() or sctp_bindx() is not called prior to
1139			 * an sctp_connectx() call, the system picks an
1140			 * ephemeral port and will choose an address set
1141			 * equivalent to binding with a wildcard address.
1142			 */
1143			if (!ep->base.bind_addr.port) {
1144				if (sctp_autobind(sk)) {
1145					err = -EAGAIN;
1146					goto out_free;
1147				}
1148			} else {
1149				/*
1150				 * If an unprivileged user inherits a 1-many
1151				 * style socket with open associations on a
1152				 * privileged port, it MAY be permitted to
1153				 * accept new associations, but it SHOULD NOT
1154				 * be permitted to open new associations.
1155				 */
1156				if (ep->base.bind_addr.port < PROT_SOCK &&
1157				    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1158					err = -EACCES;
1159					goto out_free;
1160				}
1161			}
1162
1163			scope = sctp_scope(&to);
1164			asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1165			if (!asoc) {
1166				err = -ENOMEM;
1167				goto out_free;
1168			}
1169
1170			err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1171							      GFP_KERNEL);
1172			if (err < 0) {
1173				goto out_free;
1174			}
1175
1176		}
1177
1178		/* Prime the peer's transport structures.  */
1179		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1180						SCTP_UNKNOWN);
1181		if (!transport) {
1182			err = -ENOMEM;
1183			goto out_free;
1184		}
1185
1186		addrcnt++;
1187		addr_buf += af->sockaddr_len;
1188		walk_size += af->sockaddr_len;
1189	}
1190
1191	/* In case the user of sctp_connectx() wants an association
1192	 * id back, assign one now.
1193	 */
1194	if (assoc_id) {
1195		err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1196		if (err < 0)
1197			goto out_free;
1198	}
1199
1200	err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1201	if (err < 0) {
1202		goto out_free;
1203	}
1204
1205	/* Initialize sk's dport and daddr for getpeername() */
1206	inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1207	af = sctp_get_af_specific(sa_addr->sa.sa_family);
1208	af->to_sk_daddr(sa_addr, sk);
1209	sk->sk_err = 0;
1210
1211	/* in-kernel sockets don't generally have a file allocated to them
1212	 * if all they do is call sock_create_kern().
1213	 */
1214	if (sk->sk_socket->file)
1215		f_flags = sk->sk_socket->file->f_flags;
1216
1217	timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1218
1219	err = sctp_wait_for_connect(asoc, &timeo);
1220	if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1221		*assoc_id = asoc->assoc_id;
1222
1223	/* Don't free association on exit. */
1224	asoc = NULL;
1225
1226out_free:
1227	pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1228		 __func__, asoc, kaddrs, err);
1229
 
 
 
1230	if (asoc) {
1231		/* sctp_primitive_ASSOCIATE may have added this association
1232		 * To the hash table, try to unhash it, just in case, its a noop
1233		 * if it wasn't hashed so we're safe
1234		 */
1235		sctp_unhash_established(asoc);
1236		sctp_association_free(asoc);
1237	}
1238	return err;
1239}
1240
1241/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1242 *
1243 * API 8.9
1244 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1245 * 			sctp_assoc_t *asoc);
1246 *
1247 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1248 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1249 * or IPv6 addresses.
1250 *
1251 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1252 * Section 3.1.2 for this usage.
1253 *
1254 * addrs is a pointer to an array of one or more socket addresses. Each
1255 * address is contained in its appropriate structure (i.e. struct
1256 * sockaddr_in or struct sockaddr_in6) the family of the address type
1257 * must be used to distengish the address length (note that this
1258 * representation is termed a "packed array" of addresses). The caller
1259 * specifies the number of addresses in the array with addrcnt.
1260 *
1261 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1262 * the association id of the new association.  On failure, sctp_connectx()
1263 * returns -1, and sets errno to the appropriate error code.  The assoc_id
1264 * is not touched by the kernel.
1265 *
1266 * For SCTP, the port given in each socket address must be the same, or
1267 * sctp_connectx() will fail, setting errno to EINVAL.
1268 *
1269 * An application can use sctp_connectx to initiate an association with
1270 * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1271 * allows a caller to specify multiple addresses at which a peer can be
1272 * reached.  The way the SCTP stack uses the list of addresses to set up
1273 * the association is implementation dependent.  This function only
1274 * specifies that the stack will try to make use of all the addresses in
1275 * the list when needed.
1276 *
1277 * Note that the list of addresses passed in is only used for setting up
1278 * the association.  It does not necessarily equal the set of addresses
1279 * the peer uses for the resulting association.  If the caller wants to
1280 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1281 * retrieve them after the association has been set up.
1282 *
1283 * Basically do nothing but copying the addresses from user to kernel
1284 * land and invoking either sctp_connectx(). This is used for tunneling
1285 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1286 *
1287 * We don't use copy_from_user() for optimization: we first do the
1288 * sanity checks (buffer size -fast- and access check-healthy
1289 * pointer); if all of those succeed, then we can alloc the memory
1290 * (expensive operation) needed to copy the data to kernel. Then we do
1291 * the copying without checking the user space area
1292 * (__copy_from_user()).
1293 *
1294 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1295 * it.
1296 *
1297 * sk        The sk of the socket
1298 * addrs     The pointer to the addresses in user land
1299 * addrssize Size of the addrs buffer
1300 *
1301 * Returns >=0 if ok, <0 errno code on error.
1302 */
1303static int __sctp_setsockopt_connectx(struct sock *sk,
1304				      struct sockaddr __user *addrs,
1305				      int addrs_size,
1306				      sctp_assoc_t *assoc_id)
1307{
1308	int err = 0;
1309	struct sockaddr *kaddrs;
1310
1311	pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1312		 __func__, sk, addrs, addrs_size);
1313
1314	if (unlikely(addrs_size <= 0))
1315		return -EINVAL;
1316
1317	/* Check the user passed a healthy pointer.  */
1318	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1319		return -EFAULT;
1320
1321	/* Alloc space for the address array in kernel memory.  */
1322	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1323	if (unlikely(!kaddrs))
1324		return -ENOMEM;
1325
1326	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1327		err = -EFAULT;
1328	} else {
1329		err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1330	}
1331
1332	kfree(kaddrs);
1333
1334	return err;
1335}
1336
1337/*
1338 * This is an older interface.  It's kept for backward compatibility
1339 * to the option that doesn't provide association id.
1340 */
1341static int sctp_setsockopt_connectx_old(struct sock *sk,
1342					struct sockaddr __user *addrs,
1343					int addrs_size)
1344{
1345	return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1346}
1347
1348/*
1349 * New interface for the API.  The since the API is done with a socket
1350 * option, to make it simple we feed back the association id is as a return
1351 * indication to the call.  Error is always negative and association id is
1352 * always positive.
1353 */
1354static int sctp_setsockopt_connectx(struct sock *sk,
1355				    struct sockaddr __user *addrs,
1356				    int addrs_size)
1357{
1358	sctp_assoc_t assoc_id = 0;
1359	int err = 0;
1360
1361	err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1362
1363	if (err)
1364		return err;
1365	else
1366		return assoc_id;
1367}
1368
1369/*
1370 * New (hopefully final) interface for the API.
1371 * We use the sctp_getaddrs_old structure so that use-space library
1372 * can avoid any unnecessary allocations. The only different part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1375 * code.
1376 */
1377#ifdef CONFIG_COMPAT
1378struct compat_sctp_getaddrs_old {
1379	sctp_assoc_t	assoc_id;
1380	s32		addr_num;
1381	compat_uptr_t	addrs;		/* struct sockaddr * */
1382};
1383#endif
1384
1385static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1386				     char __user *optval,
1387				     int __user *optlen)
1388{
1389	struct sctp_getaddrs_old param;
1390	sctp_assoc_t assoc_id = 0;
1391	int err = 0;
1392
1393#ifdef CONFIG_COMPAT
1394	if (is_compat_task()) {
1395		struct compat_sctp_getaddrs_old param32;
1396
1397		if (len < sizeof(param32))
1398			return -EINVAL;
1399		if (copy_from_user(&param32, optval, sizeof(param32)))
1400			return -EFAULT;
1401
1402		param.assoc_id = param32.assoc_id;
1403		param.addr_num = param32.addr_num;
1404		param.addrs = compat_ptr(param32.addrs);
1405	} else
1406#endif
1407	{
1408		if (len < sizeof(param))
1409			return -EINVAL;
1410		if (copy_from_user(&param, optval, sizeof(param)))
1411			return -EFAULT;
1412	}
1413
1414	err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1415					 param.addrs, param.addr_num,
1416					 &assoc_id);
1417	if (err == 0 || err == -EINPROGRESS) {
1418		if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1419			return -EFAULT;
1420		if (put_user(sizeof(assoc_id), optlen))
1421			return -EFAULT;
1422	}
1423
1424	return err;
1425}
1426
1427/* API 3.1.4 close() - UDP Style Syntax
1428 * Applications use close() to perform graceful shutdown (as described in
1429 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1430 * by a UDP-style socket.
1431 *
1432 * The syntax is
1433 *
1434 *   ret = close(int sd);
1435 *
1436 *   sd      - the socket descriptor of the associations to be closed.
1437 *
1438 * To gracefully shutdown a specific association represented by the
1439 * UDP-style socket, an application should use the sendmsg() call,
1440 * passing no user data, but including the appropriate flag in the
1441 * ancillary data (see Section xxxx).
1442 *
1443 * If sd in the close() call is a branched-off socket representing only
1444 * one association, the shutdown is performed on that association only.
1445 *
1446 * 4.1.6 close() - TCP Style Syntax
1447 *
1448 * Applications use close() to gracefully close down an association.
1449 *
1450 * The syntax is:
1451 *
1452 *    int close(int sd);
1453 *
1454 *      sd      - the socket descriptor of the association to be closed.
1455 *
1456 * After an application calls close() on a socket descriptor, no further
1457 * socket operations will succeed on that descriptor.
1458 *
1459 * API 7.1.4 SO_LINGER
1460 *
1461 * An application using the TCP-style socket can use this option to
1462 * perform the SCTP ABORT primitive.  The linger option structure is:
1463 *
1464 *  struct  linger {
1465 *     int     l_onoff;                // option on/off
1466 *     int     l_linger;               // linger time
1467 * };
1468 *
1469 * To enable the option, set l_onoff to 1.  If the l_linger value is set
1470 * to 0, calling close() is the same as the ABORT primitive.  If the
1471 * value is set to a negative value, the setsockopt() call will return
1472 * an error.  If the value is set to a positive value linger_time, the
1473 * close() can be blocked for at most linger_time ms.  If the graceful
1474 * shutdown phase does not finish during this period, close() will
1475 * return but the graceful shutdown phase continues in the system.
1476 */
1477static void sctp_close(struct sock *sk, long timeout)
1478{
1479	struct net *net = sock_net(sk);
1480	struct sctp_endpoint *ep;
1481	struct sctp_association *asoc;
1482	struct list_head *pos, *temp;
1483	unsigned int data_was_unread;
1484
1485	pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1486
1487	lock_sock(sk);
1488	sk->sk_shutdown = SHUTDOWN_MASK;
1489	sk->sk_state = SCTP_SS_CLOSING;
1490
1491	ep = sctp_sk(sk)->ep;
1492
1493	/* Clean up any skbs sitting on the receive queue.  */
1494	data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1495	data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1496
1497	/* Walk all associations on an endpoint.  */
1498	list_for_each_safe(pos, temp, &ep->asocs) {
1499		asoc = list_entry(pos, struct sctp_association, asocs);
1500
1501		if (sctp_style(sk, TCP)) {
1502			/* A closed association can still be in the list if
1503			 * it belongs to a TCP-style listening socket that is
1504			 * not yet accepted. If so, free it. If not, send an
1505			 * ABORT or SHUTDOWN based on the linger options.
1506			 */
1507			if (sctp_state(asoc, CLOSED)) {
1508				sctp_unhash_established(asoc);
1509				sctp_association_free(asoc);
1510				continue;
1511			}
1512		}
1513
1514		if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1515		    !skb_queue_empty(&asoc->ulpq.reasm) ||
1516		    (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1517			struct sctp_chunk *chunk;
1518
1519			chunk = sctp_make_abort_user(asoc, NULL, 0);
1520			if (chunk)
1521				sctp_primitive_ABORT(net, asoc, chunk);
1522		} else
1523			sctp_primitive_SHUTDOWN(net, asoc, NULL);
1524	}
1525
1526	/* On a TCP-style socket, block for at most linger_time if set. */
1527	if (sctp_style(sk, TCP) && timeout)
1528		sctp_wait_for_close(sk, timeout);
1529
1530	/* This will run the backlog queue.  */
1531	release_sock(sk);
1532
1533	/* Supposedly, no process has access to the socket, but
1534	 * the net layers still may.
1535	 */
1536	local_bh_disable();
1537	bh_lock_sock(sk);
1538
1539	/* Hold the sock, since sk_common_release() will put sock_put()
1540	 * and we have just a little more cleanup.
1541	 */
1542	sock_hold(sk);
1543	sk_common_release(sk);
1544
1545	bh_unlock_sock(sk);
1546	local_bh_enable();
1547
1548	sock_put(sk);
1549
1550	SCTP_DBG_OBJCNT_DEC(sock);
1551}
1552
1553/* Handle EPIPE error. */
1554static int sctp_error(struct sock *sk, int flags, int err)
1555{
1556	if (err == -EPIPE)
1557		err = sock_error(sk) ? : -EPIPE;
1558	if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1559		send_sig(SIGPIPE, current, 0);
1560	return err;
1561}
1562
1563/* API 3.1.3 sendmsg() - UDP Style Syntax
1564 *
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1567 *
1568 *  ssize_t sendmsg(int socket, const struct msghdr *message,
1569 *                  int flags);
1570 *
1571 *  socket  - the socket descriptor of the endpoint.
1572 *  message - pointer to the msghdr structure which contains a single
1573 *            user message and possibly some ancillary data.
1574 *
1575 *            See Section 5 for complete description of the data
1576 *            structures.
1577 *
1578 *  flags   - flags sent or received with the user message, see Section
1579 *            5 for complete description of the flags.
1580 *
1581 * Note:  This function could use a rewrite especially when explicit
1582 * connect support comes in.
1583 */
1584/* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1585
1586static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1587
1588static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1589			struct msghdr *msg, size_t msg_len)
1590{
1591	struct net *net = sock_net(sk);
1592	struct sctp_sock *sp;
1593	struct sctp_endpoint *ep;
1594	struct sctp_association *new_asoc = NULL, *asoc = NULL;
1595	struct sctp_transport *transport, *chunk_tp;
1596	struct sctp_chunk *chunk;
1597	union sctp_addr to;
1598	struct sockaddr *msg_name = NULL;
1599	struct sctp_sndrcvinfo default_sinfo;
1600	struct sctp_sndrcvinfo *sinfo;
1601	struct sctp_initmsg *sinit;
1602	sctp_assoc_t associd = 0;
1603	sctp_cmsgs_t cmsgs = { NULL };
1604	int err;
1605	sctp_scope_t scope;
1606	long timeo;
1607	__u16 sinfo_flags = 0;
1608	struct sctp_datamsg *datamsg;
1609	int msg_flags = msg->msg_flags;
1610
 
 
 
1611	err = 0;
1612	sp = sctp_sk(sk);
1613	ep = sp->ep;
1614
1615	pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1616		 msg, msg_len, ep);
1617
1618	/* We cannot send a message over a TCP-style listening socket. */
1619	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1620		err = -EPIPE;
1621		goto out_nounlock;
1622	}
1623
1624	/* Parse out the SCTP CMSGs.  */
1625	err = sctp_msghdr_parse(msg, &cmsgs);
 
1626	if (err) {
1627		pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1628		goto out_nounlock;
1629	}
1630
1631	/* Fetch the destination address for this packet.  This
1632	 * address only selects the association--it is not necessarily
1633	 * the address we will send to.
1634	 * For a peeled-off socket, msg_name is ignored.
1635	 */
1636	if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1637		int msg_namelen = msg->msg_namelen;
1638
1639		err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1640				       msg_namelen);
1641		if (err)
1642			return err;
1643
1644		if (msg_namelen > sizeof(to))
1645			msg_namelen = sizeof(to);
1646		memcpy(&to, msg->msg_name, msg_namelen);
1647		msg_name = msg->msg_name;
1648	}
1649
1650	sinfo = cmsgs.info;
1651	sinit = cmsgs.init;
1652
1653	/* Did the user specify SNDRCVINFO?  */
1654	if (sinfo) {
1655		sinfo_flags = sinfo->sinfo_flags;
1656		associd = sinfo->sinfo_assoc_id;
1657	}
1658
1659	pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1660		 msg_len, sinfo_flags);
1661
1662	/* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1663	if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1664		err = -EINVAL;
1665		goto out_nounlock;
1666	}
1667
1668	/* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1669	 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1670	 * If SCTP_ABORT is set, the message length could be non zero with
1671	 * the msg_iov set to the user abort reason.
1672	 */
1673	if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1674	    (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1675		err = -EINVAL;
1676		goto out_nounlock;
1677	}
1678
1679	/* If SCTP_ADDR_OVER is set, there must be an address
1680	 * specified in msg_name.
1681	 */
1682	if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1683		err = -EINVAL;
1684		goto out_nounlock;
1685	}
1686
1687	transport = NULL;
1688
1689	pr_debug("%s: about to look up association\n", __func__);
1690
1691	lock_sock(sk);
1692
1693	/* If a msg_name has been specified, assume this is to be used.  */
1694	if (msg_name) {
1695		/* Look for a matching association on the endpoint. */
1696		asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1697		if (!asoc) {
1698			/* If we could not find a matching association on the
1699			 * endpoint, make sure that it is not a TCP-style
1700			 * socket that already has an association or there is
1701			 * no peeled-off association on another socket.
1702			 */
1703			if ((sctp_style(sk, TCP) &&
1704			     sctp_sstate(sk, ESTABLISHED)) ||
1705			    sctp_endpoint_is_peeled_off(ep, &to)) {
1706				err = -EADDRNOTAVAIL;
1707				goto out_unlock;
1708			}
1709		}
1710	} else {
1711		asoc = sctp_id2assoc(sk, associd);
1712		if (!asoc) {
1713			err = -EPIPE;
1714			goto out_unlock;
1715		}
1716	}
1717
1718	if (asoc) {
1719		pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1720
1721		/* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1722		 * socket that has an association in CLOSED state. This can
1723		 * happen when an accepted socket has an association that is
1724		 * already CLOSED.
1725		 */
1726		if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1727			err = -EPIPE;
1728			goto out_unlock;
1729		}
1730
1731		if (sinfo_flags & SCTP_EOF) {
1732			pr_debug("%s: shutting down association:%p\n",
1733				 __func__, asoc);
1734
1735			sctp_primitive_SHUTDOWN(net, asoc, NULL);
1736			err = 0;
1737			goto out_unlock;
1738		}
1739		if (sinfo_flags & SCTP_ABORT) {
1740
1741			chunk = sctp_make_abort_user(asoc, msg, msg_len);
1742			if (!chunk) {
1743				err = -ENOMEM;
1744				goto out_unlock;
1745			}
1746
1747			pr_debug("%s: aborting association:%p\n",
1748				 __func__, asoc);
1749
1750			sctp_primitive_ABORT(net, asoc, chunk);
1751			err = 0;
1752			goto out_unlock;
1753		}
1754	}
1755
1756	/* Do we need to create the association?  */
1757	if (!asoc) {
1758		pr_debug("%s: there is no association yet\n", __func__);
1759
1760		if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1761			err = -EINVAL;
1762			goto out_unlock;
1763		}
1764
1765		/* Check for invalid stream against the stream counts,
1766		 * either the default or the user specified stream counts.
1767		 */
1768		if (sinfo) {
1769			if (!sinit || !sinit->sinit_num_ostreams) {
1770				/* Check against the defaults. */
1771				if (sinfo->sinfo_stream >=
1772				    sp->initmsg.sinit_num_ostreams) {
1773					err = -EINVAL;
1774					goto out_unlock;
1775				}
1776			} else {
1777				/* Check against the requested.  */
1778				if (sinfo->sinfo_stream >=
1779				    sinit->sinit_num_ostreams) {
1780					err = -EINVAL;
1781					goto out_unlock;
1782				}
1783			}
1784		}
1785
1786		/*
1787		 * API 3.1.2 bind() - UDP Style Syntax
1788		 * If a bind() or sctp_bindx() is not called prior to a
1789		 * sendmsg() call that initiates a new association, the
1790		 * system picks an ephemeral port and will choose an address
1791		 * set equivalent to binding with a wildcard address.
1792		 */
1793		if (!ep->base.bind_addr.port) {
1794			if (sctp_autobind(sk)) {
1795				err = -EAGAIN;
1796				goto out_unlock;
1797			}
1798		} else {
1799			/*
1800			 * If an unprivileged user inherits a one-to-many
1801			 * style socket with open associations on a privileged
1802			 * port, it MAY be permitted to accept new associations,
1803			 * but it SHOULD NOT be permitted to open new
1804			 * associations.
1805			 */
1806			if (ep->base.bind_addr.port < PROT_SOCK &&
1807			    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1808				err = -EACCES;
1809				goto out_unlock;
1810			}
1811		}
1812
1813		scope = sctp_scope(&to);
1814		new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1815		if (!new_asoc) {
1816			err = -ENOMEM;
1817			goto out_unlock;
1818		}
1819		asoc = new_asoc;
1820		err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1821		if (err < 0) {
1822			err = -ENOMEM;
1823			goto out_free;
1824		}
1825
1826		/* If the SCTP_INIT ancillary data is specified, set all
1827		 * the association init values accordingly.
1828		 */
1829		if (sinit) {
1830			if (sinit->sinit_num_ostreams) {
1831				asoc->c.sinit_num_ostreams =
1832					sinit->sinit_num_ostreams;
1833			}
1834			if (sinit->sinit_max_instreams) {
1835				asoc->c.sinit_max_instreams =
1836					sinit->sinit_max_instreams;
1837			}
1838			if (sinit->sinit_max_attempts) {
1839				asoc->max_init_attempts
1840					= sinit->sinit_max_attempts;
1841			}
1842			if (sinit->sinit_max_init_timeo) {
1843				asoc->max_init_timeo =
1844				 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1845			}
1846		}
1847
1848		/* Prime the peer's transport structures.  */
1849		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1850		if (!transport) {
1851			err = -ENOMEM;
1852			goto out_free;
1853		}
1854	}
1855
1856	/* ASSERT: we have a valid association at this point.  */
1857	pr_debug("%s: we have a valid association\n", __func__);
1858
1859	if (!sinfo) {
1860		/* If the user didn't specify SNDRCVINFO, make up one with
1861		 * some defaults.
1862		 */
1863		memset(&default_sinfo, 0, sizeof(default_sinfo));
1864		default_sinfo.sinfo_stream = asoc->default_stream;
1865		default_sinfo.sinfo_flags = asoc->default_flags;
1866		default_sinfo.sinfo_ppid = asoc->default_ppid;
1867		default_sinfo.sinfo_context = asoc->default_context;
1868		default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1869		default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1870		sinfo = &default_sinfo;
1871	}
1872
1873	/* API 7.1.7, the sndbuf size per association bounds the
1874	 * maximum size of data that can be sent in a single send call.
1875	 */
1876	if (msg_len > sk->sk_sndbuf) {
1877		err = -EMSGSIZE;
1878		goto out_free;
1879	}
1880
1881	if (asoc->pmtu_pending)
1882		sctp_assoc_pending_pmtu(sk, asoc);
1883
1884	/* If fragmentation is disabled and the message length exceeds the
1885	 * association fragmentation point, return EMSGSIZE.  The I-D
1886	 * does not specify what this error is, but this looks like
1887	 * a great fit.
1888	 */
1889	if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1890		err = -EMSGSIZE;
1891		goto out_free;
1892	}
1893
1894	/* Check for invalid stream. */
1895	if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1896		err = -EINVAL;
1897		goto out_free;
1898	}
1899
1900	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1901	if (!sctp_wspace(asoc)) {
1902		err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1903		if (err)
1904			goto out_free;
1905	}
1906
1907	/* If an address is passed with the sendto/sendmsg call, it is used
1908	 * to override the primary destination address in the TCP model, or
1909	 * when SCTP_ADDR_OVER flag is set in the UDP model.
1910	 */
1911	if ((sctp_style(sk, TCP) && msg_name) ||
1912	    (sinfo_flags & SCTP_ADDR_OVER)) {
1913		chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1914		if (!chunk_tp) {
1915			err = -EINVAL;
1916			goto out_free;
1917		}
1918	} else
1919		chunk_tp = NULL;
1920
1921	/* Auto-connect, if we aren't connected already. */
1922	if (sctp_state(asoc, CLOSED)) {
1923		err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1924		if (err < 0)
1925			goto out_free;
1926
1927		pr_debug("%s: we associated primitively\n", __func__);
1928	}
1929
1930	/* Break the message into multiple chunks of maximum size. */
1931	datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1932	if (IS_ERR(datamsg)) {
1933		err = PTR_ERR(datamsg);
1934		goto out_free;
1935	}
1936
1937	/* Now send the (possibly) fragmented message. */
1938	list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1939		sctp_chunk_hold(chunk);
1940
1941		/* Do accounting for the write space.  */
1942		sctp_set_owner_w(chunk);
1943
1944		chunk->transport = chunk_tp;
1945	}
1946
1947	/* Send it to the lower layers.  Note:  all chunks
1948	 * must either fail or succeed.   The lower layer
1949	 * works that way today.  Keep it that way or this
1950	 * breaks.
1951	 */
1952	err = sctp_primitive_SEND(net, asoc, datamsg);
1953	/* Did the lower layer accept the chunk? */
1954	if (err) {
1955		sctp_datamsg_free(datamsg);
1956		goto out_free;
1957	}
1958
1959	pr_debug("%s: we sent primitively\n", __func__);
1960
1961	sctp_datamsg_put(datamsg);
1962	err = msg_len;
 
 
1963
1964	/* If we are already past ASSOCIATE, the lower
1965	 * layers are responsible for association cleanup.
1966	 */
1967	goto out_unlock;
1968
1969out_free:
1970	if (new_asoc) {
1971		sctp_unhash_established(asoc);
1972		sctp_association_free(asoc);
1973	}
1974out_unlock:
1975	release_sock(sk);
1976
1977out_nounlock:
1978	return sctp_error(sk, msg_flags, err);
1979
1980#if 0
1981do_sock_err:
1982	if (msg_len)
1983		err = msg_len;
1984	else
1985		err = sock_error(sk);
1986	goto out;
1987
1988do_interrupted:
1989	if (msg_len)
1990		err = msg_len;
1991	goto out;
1992#endif /* 0 */
1993}
1994
1995/* This is an extended version of skb_pull() that removes the data from the
1996 * start of a skb even when data is spread across the list of skb's in the
1997 * frag_list. len specifies the total amount of data that needs to be removed.
1998 * when 'len' bytes could be removed from the skb, it returns 0.
1999 * If 'len' exceeds the total skb length,  it returns the no. of bytes that
2000 * could not be removed.
2001 */
2002static int sctp_skb_pull(struct sk_buff *skb, int len)
2003{
2004	struct sk_buff *list;
2005	int skb_len = skb_headlen(skb);
2006	int rlen;
2007
2008	if (len <= skb_len) {
2009		__skb_pull(skb, len);
2010		return 0;
2011	}
2012	len -= skb_len;
2013	__skb_pull(skb, skb_len);
2014
2015	skb_walk_frags(skb, list) {
2016		rlen = sctp_skb_pull(list, len);
2017		skb->len -= (len-rlen);
2018		skb->data_len -= (len-rlen);
2019
2020		if (!rlen)
2021			return 0;
2022
2023		len = rlen;
2024	}
2025
2026	return len;
2027}
2028
2029/* API 3.1.3  recvmsg() - UDP Style Syntax
2030 *
2031 *  ssize_t recvmsg(int socket, struct msghdr *message,
2032 *                    int flags);
2033 *
2034 *  socket  - the socket descriptor of the endpoint.
2035 *  message - pointer to the msghdr structure which contains a single
2036 *            user message and possibly some ancillary data.
2037 *
2038 *            See Section 5 for complete description of the data
2039 *            structures.
2040 *
2041 *  flags   - flags sent or received with the user message, see Section
2042 *            5 for complete description of the flags.
2043 */
2044static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2045
2046static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2047			struct msghdr *msg, size_t len, int noblock,
2048			int flags, int *addr_len)
2049{
2050	struct sctp_ulpevent *event = NULL;
2051	struct sctp_sock *sp = sctp_sk(sk);
2052	struct sk_buff *skb;
2053	int copied;
2054	int err = 0;
2055	int skb_len;
2056
2057	pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2058		 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2059		 addr_len);
 
2060
2061	lock_sock(sk);
2062
2063	if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2064		err = -ENOTCONN;
2065		goto out;
2066	}
2067
2068	skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2069	if (!skb)
2070		goto out;
2071
2072	/* Get the total length of the skb including any skb's in the
2073	 * frag_list.
2074	 */
2075	skb_len = skb->len;
2076
2077	copied = skb_len;
2078	if (copied > len)
2079		copied = len;
2080
2081	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2082
2083	event = sctp_skb2event(skb);
2084
2085	if (err)
2086		goto out_free;
2087
2088	sock_recv_ts_and_drops(msg, sk, skb);
2089	if (sctp_ulpevent_is_notification(event)) {
2090		msg->msg_flags |= MSG_NOTIFICATION;
2091		sp->pf->event_msgname(event, msg->msg_name, addr_len);
2092	} else {
2093		sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2094	}
2095
2096	/* Check if we allow SCTP_SNDRCVINFO. */
2097	if (sp->subscribe.sctp_data_io_event)
2098		sctp_ulpevent_read_sndrcvinfo(event, msg);
2099#if 0
2100	/* FIXME: we should be calling IP/IPv6 layers.  */
2101	if (sk->sk_protinfo.af_inet.cmsg_flags)
2102		ip_cmsg_recv(msg, skb);
2103#endif
2104
2105	err = copied;
2106
2107	/* If skb's length exceeds the user's buffer, update the skb and
2108	 * push it back to the receive_queue so that the next call to
2109	 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2110	 */
2111	if (skb_len > copied) {
2112		msg->msg_flags &= ~MSG_EOR;
2113		if (flags & MSG_PEEK)
2114			goto out_free;
2115		sctp_skb_pull(skb, copied);
2116		skb_queue_head(&sk->sk_receive_queue, skb);
2117
2118		/* When only partial message is copied to the user, increase
2119		 * rwnd by that amount. If all the data in the skb is read,
2120		 * rwnd is updated when the event is freed.
2121		 */
2122		if (!sctp_ulpevent_is_notification(event))
2123			sctp_assoc_rwnd_increase(event->asoc, copied);
2124		goto out;
2125	} else if ((event->msg_flags & MSG_NOTIFICATION) ||
2126		   (event->msg_flags & MSG_EOR))
2127		msg->msg_flags |= MSG_EOR;
2128	else
2129		msg->msg_flags &= ~MSG_EOR;
2130
2131out_free:
2132	if (flags & MSG_PEEK) {
2133		/* Release the skb reference acquired after peeking the skb in
2134		 * sctp_skb_recv_datagram().
2135		 */
2136		kfree_skb(skb);
2137	} else {
2138		/* Free the event which includes releasing the reference to
2139		 * the owner of the skb, freeing the skb and updating the
2140		 * rwnd.
2141		 */
2142		sctp_ulpevent_free(event);
2143	}
2144out:
2145	release_sock(sk);
2146	return err;
2147}
2148
2149/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2150 *
2151 * This option is a on/off flag.  If enabled no SCTP message
2152 * fragmentation will be performed.  Instead if a message being sent
2153 * exceeds the current PMTU size, the message will NOT be sent and
2154 * instead a error will be indicated to the user.
2155 */
2156static int sctp_setsockopt_disable_fragments(struct sock *sk,
2157					     char __user *optval,
2158					     unsigned int optlen)
2159{
2160	int val;
2161
2162	if (optlen < sizeof(int))
2163		return -EINVAL;
2164
2165	if (get_user(val, (int __user *)optval))
2166		return -EFAULT;
2167
2168	sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2169
2170	return 0;
2171}
2172
2173static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2174				  unsigned int optlen)
2175{
2176	struct sctp_association *asoc;
2177	struct sctp_ulpevent *event;
2178
2179	if (optlen > sizeof(struct sctp_event_subscribe))
2180		return -EINVAL;
2181	if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2182		return -EFAULT;
2183
2184	/*
2185	 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2186	 * if there is no data to be sent or retransmit, the stack will
2187	 * immediately send up this notification.
2188	 */
2189	if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2190				       &sctp_sk(sk)->subscribe)) {
2191		asoc = sctp_id2assoc(sk, 0);
2192
2193		if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2194			event = sctp_ulpevent_make_sender_dry_event(asoc,
2195					GFP_ATOMIC);
2196			if (!event)
2197				return -ENOMEM;
2198
2199			sctp_ulpq_tail_event(&asoc->ulpq, event);
2200		}
2201	}
2202
2203	return 0;
2204}
2205
2206/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2207 *
2208 * This socket option is applicable to the UDP-style socket only.  When
2209 * set it will cause associations that are idle for more than the
2210 * specified number of seconds to automatically close.  An association
2211 * being idle is defined an association that has NOT sent or received
2212 * user data.  The special value of '0' indicates that no automatic
2213 * close of any associations should be performed.  The option expects an
2214 * integer defining the number of seconds of idle time before an
2215 * association is closed.
2216 */
2217static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2218				     unsigned int optlen)
2219{
2220	struct sctp_sock *sp = sctp_sk(sk);
2221	struct net *net = sock_net(sk);
2222
2223	/* Applicable to UDP-style socket only */
2224	if (sctp_style(sk, TCP))
2225		return -EOPNOTSUPP;
2226	if (optlen != sizeof(int))
2227		return -EINVAL;
2228	if (copy_from_user(&sp->autoclose, optval, optlen))
2229		return -EFAULT;
2230
2231	if (sp->autoclose > net->sctp.max_autoclose)
2232		sp->autoclose = net->sctp.max_autoclose;
2233
2234	return 0;
2235}
2236
2237/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2238 *
2239 * Applications can enable or disable heartbeats for any peer address of
2240 * an association, modify an address's heartbeat interval, force a
2241 * heartbeat to be sent immediately, and adjust the address's maximum
2242 * number of retransmissions sent before an address is considered
2243 * unreachable.  The following structure is used to access and modify an
2244 * address's parameters:
2245 *
2246 *  struct sctp_paddrparams {
2247 *     sctp_assoc_t            spp_assoc_id;
2248 *     struct sockaddr_storage spp_address;
2249 *     uint32_t                spp_hbinterval;
2250 *     uint16_t                spp_pathmaxrxt;
2251 *     uint32_t                spp_pathmtu;
2252 *     uint32_t                spp_sackdelay;
2253 *     uint32_t                spp_flags;
2254 * };
2255 *
2256 *   spp_assoc_id    - (one-to-many style socket) This is filled in the
2257 *                     application, and identifies the association for
2258 *                     this query.
2259 *   spp_address     - This specifies which address is of interest.
2260 *   spp_hbinterval  - This contains the value of the heartbeat interval,
2261 *                     in milliseconds.  If a  value of zero
2262 *                     is present in this field then no changes are to
2263 *                     be made to this parameter.
2264 *   spp_pathmaxrxt  - This contains the maximum number of
2265 *                     retransmissions before this address shall be
2266 *                     considered unreachable. If a  value of zero
2267 *                     is present in this field then no changes are to
2268 *                     be made to this parameter.
2269 *   spp_pathmtu     - When Path MTU discovery is disabled the value
2270 *                     specified here will be the "fixed" path mtu.
2271 *                     Note that if the spp_address field is empty
2272 *                     then all associations on this address will
2273 *                     have this fixed path mtu set upon them.
2274 *
2275 *   spp_sackdelay   - When delayed sack is enabled, this value specifies
2276 *                     the number of milliseconds that sacks will be delayed
2277 *                     for. This value will apply to all addresses of an
2278 *                     association if the spp_address field is empty. Note
2279 *                     also, that if delayed sack is enabled and this
2280 *                     value is set to 0, no change is made to the last
2281 *                     recorded delayed sack timer value.
2282 *
2283 *   spp_flags       - These flags are used to control various features
2284 *                     on an association. The flag field may contain
2285 *                     zero or more of the following options.
2286 *
2287 *                     SPP_HB_ENABLE  - Enable heartbeats on the
2288 *                     specified address. Note that if the address
2289 *                     field is empty all addresses for the association
2290 *                     have heartbeats enabled upon them.
2291 *
2292 *                     SPP_HB_DISABLE - Disable heartbeats on the
2293 *                     speicifed address. Note that if the address
2294 *                     field is empty all addresses for the association
2295 *                     will have their heartbeats disabled. Note also
2296 *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
2297 *                     mutually exclusive, only one of these two should
2298 *                     be specified. Enabling both fields will have
2299 *                     undetermined results.
2300 *
2301 *                     SPP_HB_DEMAND - Request a user initiated heartbeat
2302 *                     to be made immediately.
2303 *
2304 *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
2305 *                     heartbeat delayis to be set to the value of 0
2306 *                     milliseconds.
2307 *
2308 *                     SPP_PMTUD_ENABLE - This field will enable PMTU
2309 *                     discovery upon the specified address. Note that
2310 *                     if the address feild is empty then all addresses
2311 *                     on the association are effected.
2312 *
2313 *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2314 *                     discovery upon the specified address. Note that
2315 *                     if the address feild is empty then all addresses
2316 *                     on the association are effected. Not also that
2317 *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2318 *                     exclusive. Enabling both will have undetermined
2319 *                     results.
2320 *
2321 *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2322 *                     on delayed sack. The time specified in spp_sackdelay
2323 *                     is used to specify the sack delay for this address. Note
2324 *                     that if spp_address is empty then all addresses will
2325 *                     enable delayed sack and take on the sack delay
2326 *                     value specified in spp_sackdelay.
2327 *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2328 *                     off delayed sack. If the spp_address field is blank then
2329 *                     delayed sack is disabled for the entire association. Note
2330 *                     also that this field is mutually exclusive to
2331 *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2332 *                     results.
2333 */
2334static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2335				       struct sctp_transport   *trans,
2336				       struct sctp_association *asoc,
2337				       struct sctp_sock        *sp,
2338				       int                      hb_change,
2339				       int                      pmtud_change,
2340				       int                      sackdelay_change)
2341{
2342	int error;
2343
2344	if (params->spp_flags & SPP_HB_DEMAND && trans) {
2345		struct net *net = sock_net(trans->asoc->base.sk);
2346
2347		error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2348		if (error)
2349			return error;
2350	}
2351
2352	/* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2353	 * this field is ignored.  Note also that a value of zero indicates
2354	 * the current setting should be left unchanged.
2355	 */
2356	if (params->spp_flags & SPP_HB_ENABLE) {
2357
2358		/* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2359		 * set.  This lets us use 0 value when this flag
2360		 * is set.
2361		 */
2362		if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2363			params->spp_hbinterval = 0;
2364
2365		if (params->spp_hbinterval ||
2366		    (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2367			if (trans) {
2368				trans->hbinterval =
2369				    msecs_to_jiffies(params->spp_hbinterval);
2370			} else if (asoc) {
2371				asoc->hbinterval =
2372				    msecs_to_jiffies(params->spp_hbinterval);
2373			} else {
2374				sp->hbinterval = params->spp_hbinterval;
2375			}
2376		}
2377	}
2378
2379	if (hb_change) {
2380		if (trans) {
2381			trans->param_flags =
2382				(trans->param_flags & ~SPP_HB) | hb_change;
2383		} else if (asoc) {
2384			asoc->param_flags =
2385				(asoc->param_flags & ~SPP_HB) | hb_change;
2386		} else {
2387			sp->param_flags =
2388				(sp->param_flags & ~SPP_HB) | hb_change;
2389		}
2390	}
2391
2392	/* When Path MTU discovery is disabled the value specified here will
2393	 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2394	 * include the flag SPP_PMTUD_DISABLE for this field to have any
2395	 * effect).
2396	 */
2397	if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2398		if (trans) {
2399			trans->pathmtu = params->spp_pathmtu;
2400			sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2401		} else if (asoc) {
2402			asoc->pathmtu = params->spp_pathmtu;
2403			sctp_frag_point(asoc, params->spp_pathmtu);
2404		} else {
2405			sp->pathmtu = params->spp_pathmtu;
2406		}
2407	}
2408
2409	if (pmtud_change) {
2410		if (trans) {
2411			int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2412				(params->spp_flags & SPP_PMTUD_ENABLE);
2413			trans->param_flags =
2414				(trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2415			if (update) {
2416				sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2417				sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2418			}
2419		} else if (asoc) {
2420			asoc->param_flags =
2421				(asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2422		} else {
2423			sp->param_flags =
2424				(sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2425		}
2426	}
2427
2428	/* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2429	 * value of this field is ignored.  Note also that a value of zero
2430	 * indicates the current setting should be left unchanged.
2431	 */
2432	if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2433		if (trans) {
2434			trans->sackdelay =
2435				msecs_to_jiffies(params->spp_sackdelay);
2436		} else if (asoc) {
2437			asoc->sackdelay =
2438				msecs_to_jiffies(params->spp_sackdelay);
2439		} else {
2440			sp->sackdelay = params->spp_sackdelay;
2441		}
2442	}
2443
2444	if (sackdelay_change) {
2445		if (trans) {
2446			trans->param_flags =
2447				(trans->param_flags & ~SPP_SACKDELAY) |
2448				sackdelay_change;
2449		} else if (asoc) {
2450			asoc->param_flags =
2451				(asoc->param_flags & ~SPP_SACKDELAY) |
2452				sackdelay_change;
2453		} else {
2454			sp->param_flags =
2455				(sp->param_flags & ~SPP_SACKDELAY) |
2456				sackdelay_change;
2457		}
2458	}
2459
2460	/* Note that a value of zero indicates the current setting should be
2461	   left unchanged.
2462	 */
2463	if (params->spp_pathmaxrxt) {
2464		if (trans) {
2465			trans->pathmaxrxt = params->spp_pathmaxrxt;
2466		} else if (asoc) {
2467			asoc->pathmaxrxt = params->spp_pathmaxrxt;
2468		} else {
2469			sp->pathmaxrxt = params->spp_pathmaxrxt;
2470		}
2471	}
2472
2473	return 0;
2474}
2475
2476static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2477					    char __user *optval,
2478					    unsigned int optlen)
2479{
2480	struct sctp_paddrparams  params;
2481	struct sctp_transport   *trans = NULL;
2482	struct sctp_association *asoc = NULL;
2483	struct sctp_sock        *sp = sctp_sk(sk);
2484	int error;
2485	int hb_change, pmtud_change, sackdelay_change;
2486
2487	if (optlen != sizeof(struct sctp_paddrparams))
2488		return -EINVAL;
2489
2490	if (copy_from_user(&params, optval, optlen))
2491		return -EFAULT;
2492
2493	/* Validate flags and value parameters. */
2494	hb_change        = params.spp_flags & SPP_HB;
2495	pmtud_change     = params.spp_flags & SPP_PMTUD;
2496	sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2497
2498	if (hb_change        == SPP_HB ||
2499	    pmtud_change     == SPP_PMTUD ||
2500	    sackdelay_change == SPP_SACKDELAY ||
2501	    params.spp_sackdelay > 500 ||
2502	    (params.spp_pathmtu &&
2503	     params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2504		return -EINVAL;
2505
2506	/* If an address other than INADDR_ANY is specified, and
2507	 * no transport is found, then the request is invalid.
2508	 */
2509	if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2510		trans = sctp_addr_id2transport(sk, &params.spp_address,
2511					       params.spp_assoc_id);
2512		if (!trans)
2513			return -EINVAL;
2514	}
2515
2516	/* Get association, if assoc_id != 0 and the socket is a one
2517	 * to many style socket, and an association was not found, then
2518	 * the id was invalid.
2519	 */
2520	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2521	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2522		return -EINVAL;
2523
2524	/* Heartbeat demand can only be sent on a transport or
2525	 * association, but not a socket.
2526	 */
2527	if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2528		return -EINVAL;
2529
2530	/* Process parameters. */
2531	error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2532					    hb_change, pmtud_change,
2533					    sackdelay_change);
2534
2535	if (error)
2536		return error;
2537
2538	/* If changes are for association, also apply parameters to each
2539	 * transport.
2540	 */
2541	if (!trans && asoc) {
2542		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2543				transports) {
2544			sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2545						    hb_change, pmtud_change,
2546						    sackdelay_change);
2547		}
2548	}
2549
2550	return 0;
2551}
2552
2553static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2554{
2555	return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2556}
2557
2558static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2559{
2560	return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2561}
2562
2563/*
2564 * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
2565 *
2566 * This option will effect the way delayed acks are performed.  This
2567 * option allows you to get or set the delayed ack time, in
2568 * milliseconds.  It also allows changing the delayed ack frequency.
2569 * Changing the frequency to 1 disables the delayed sack algorithm.  If
2570 * the assoc_id is 0, then this sets or gets the endpoints default
2571 * values.  If the assoc_id field is non-zero, then the set or get
2572 * effects the specified association for the one to many model (the
2573 * assoc_id field is ignored by the one to one model).  Note that if
2574 * sack_delay or sack_freq are 0 when setting this option, then the
2575 * current values will remain unchanged.
2576 *
2577 * struct sctp_sack_info {
2578 *     sctp_assoc_t            sack_assoc_id;
2579 *     uint32_t                sack_delay;
2580 *     uint32_t                sack_freq;
2581 * };
2582 *
2583 * sack_assoc_id -  This parameter, indicates which association the user
2584 *    is performing an action upon.  Note that if this field's value is
2585 *    zero then the endpoints default value is changed (effecting future
2586 *    associations only).
2587 *
2588 * sack_delay -  This parameter contains the number of milliseconds that
2589 *    the user is requesting the delayed ACK timer be set to.  Note that
2590 *    this value is defined in the standard to be between 200 and 500
2591 *    milliseconds.
2592 *
2593 * sack_freq -  This parameter contains the number of packets that must
2594 *    be received before a sack is sent without waiting for the delay
2595 *    timer to expire.  The default value for this is 2, setting this
2596 *    value to 1 will disable the delayed sack algorithm.
2597 */
2598
2599static int sctp_setsockopt_delayed_ack(struct sock *sk,
2600				       char __user *optval, unsigned int optlen)
2601{
2602	struct sctp_sack_info    params;
2603	struct sctp_transport   *trans = NULL;
2604	struct sctp_association *asoc = NULL;
2605	struct sctp_sock        *sp = sctp_sk(sk);
2606
2607	if (optlen == sizeof(struct sctp_sack_info)) {
2608		if (copy_from_user(&params, optval, optlen))
2609			return -EFAULT;
2610
2611		if (params.sack_delay == 0 && params.sack_freq == 0)
2612			return 0;
2613	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2614		pr_warn_ratelimited(DEPRECATED
2615				    "%s (pid %d) "
2616				    "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2617				    "Use struct sctp_sack_info instead\n",
2618				    current->comm, task_pid_nr(current));
2619		if (copy_from_user(&params, optval, optlen))
2620			return -EFAULT;
2621
2622		if (params.sack_delay == 0)
2623			params.sack_freq = 1;
2624		else
2625			params.sack_freq = 0;
2626	} else
2627		return -EINVAL;
2628
2629	/* Validate value parameter. */
2630	if (params.sack_delay > 500)
2631		return -EINVAL;
2632
2633	/* Get association, if sack_assoc_id != 0 and the socket is a one
2634	 * to many style socket, and an association was not found, then
2635	 * the id was invalid.
2636	 */
2637	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2638	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2639		return -EINVAL;
2640
2641	if (params.sack_delay) {
2642		if (asoc) {
2643			asoc->sackdelay =
2644				msecs_to_jiffies(params.sack_delay);
2645			asoc->param_flags =
2646				sctp_spp_sackdelay_enable(asoc->param_flags);
 
2647		} else {
2648			sp->sackdelay = params.sack_delay;
2649			sp->param_flags =
2650				sctp_spp_sackdelay_enable(sp->param_flags);
 
2651		}
2652	}
2653
2654	if (params.sack_freq == 1) {
2655		if (asoc) {
2656			asoc->param_flags =
2657				sctp_spp_sackdelay_disable(asoc->param_flags);
 
2658		} else {
2659			sp->param_flags =
2660				sctp_spp_sackdelay_disable(sp->param_flags);
 
2661		}
2662	} else if (params.sack_freq > 1) {
2663		if (asoc) {
2664			asoc->sackfreq = params.sack_freq;
2665			asoc->param_flags =
2666				sctp_spp_sackdelay_enable(asoc->param_flags);
 
2667		} else {
2668			sp->sackfreq = params.sack_freq;
2669			sp->param_flags =
2670				sctp_spp_sackdelay_enable(sp->param_flags);
 
2671		}
2672	}
2673
2674	/* If change is for association, also apply to each transport. */
2675	if (asoc) {
2676		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2677				transports) {
2678			if (params.sack_delay) {
2679				trans->sackdelay =
2680					msecs_to_jiffies(params.sack_delay);
2681				trans->param_flags =
2682					sctp_spp_sackdelay_enable(trans->param_flags);
 
2683			}
2684			if (params.sack_freq == 1) {
2685				trans->param_flags =
2686					sctp_spp_sackdelay_disable(trans->param_flags);
 
2687			} else if (params.sack_freq > 1) {
2688				trans->sackfreq = params.sack_freq;
2689				trans->param_flags =
2690					sctp_spp_sackdelay_enable(trans->param_flags);
 
2691			}
2692		}
2693	}
2694
2695	return 0;
2696}
2697
2698/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2699 *
2700 * Applications can specify protocol parameters for the default association
2701 * initialization.  The option name argument to setsockopt() and getsockopt()
2702 * is SCTP_INITMSG.
2703 *
2704 * Setting initialization parameters is effective only on an unconnected
2705 * socket (for UDP-style sockets only future associations are effected
2706 * by the change).  With TCP-style sockets, this option is inherited by
2707 * sockets derived from a listener socket.
2708 */
2709static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2710{
2711	struct sctp_initmsg sinit;
2712	struct sctp_sock *sp = sctp_sk(sk);
2713
2714	if (optlen != sizeof(struct sctp_initmsg))
2715		return -EINVAL;
2716	if (copy_from_user(&sinit, optval, optlen))
2717		return -EFAULT;
2718
2719	if (sinit.sinit_num_ostreams)
2720		sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2721	if (sinit.sinit_max_instreams)
2722		sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2723	if (sinit.sinit_max_attempts)
2724		sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2725	if (sinit.sinit_max_init_timeo)
2726		sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2727
2728	return 0;
2729}
2730
2731/*
2732 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2733 *
2734 *   Applications that wish to use the sendto() system call may wish to
2735 *   specify a default set of parameters that would normally be supplied
2736 *   through the inclusion of ancillary data.  This socket option allows
2737 *   such an application to set the default sctp_sndrcvinfo structure.
2738 *   The application that wishes to use this socket option simply passes
2739 *   in to this call the sctp_sndrcvinfo structure defined in Section
2740 *   5.2.2) The input parameters accepted by this call include
2741 *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2742 *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2743 *   to this call if the caller is using the UDP model.
2744 */
2745static int sctp_setsockopt_default_send_param(struct sock *sk,
2746					      char __user *optval,
2747					      unsigned int optlen)
2748{
2749	struct sctp_sndrcvinfo info;
2750	struct sctp_association *asoc;
2751	struct sctp_sock *sp = sctp_sk(sk);
2752
2753	if (optlen != sizeof(struct sctp_sndrcvinfo))
2754		return -EINVAL;
2755	if (copy_from_user(&info, optval, optlen))
2756		return -EFAULT;
2757
2758	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2759	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2760		return -EINVAL;
2761
2762	if (asoc) {
2763		asoc->default_stream = info.sinfo_stream;
2764		asoc->default_flags = info.sinfo_flags;
2765		asoc->default_ppid = info.sinfo_ppid;
2766		asoc->default_context = info.sinfo_context;
2767		asoc->default_timetolive = info.sinfo_timetolive;
2768	} else {
2769		sp->default_stream = info.sinfo_stream;
2770		sp->default_flags = info.sinfo_flags;
2771		sp->default_ppid = info.sinfo_ppid;
2772		sp->default_context = info.sinfo_context;
2773		sp->default_timetolive = info.sinfo_timetolive;
2774	}
2775
2776	return 0;
2777}
2778
2779/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2780 *
2781 * Requests that the local SCTP stack use the enclosed peer address as
2782 * the association primary.  The enclosed address must be one of the
2783 * association peer's addresses.
2784 */
2785static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2786					unsigned int optlen)
2787{
2788	struct sctp_prim prim;
2789	struct sctp_transport *trans;
2790
2791	if (optlen != sizeof(struct sctp_prim))
2792		return -EINVAL;
2793
2794	if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2795		return -EFAULT;
2796
2797	trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2798	if (!trans)
2799		return -EINVAL;
2800
2801	sctp_assoc_set_primary(trans->asoc, trans);
2802
2803	return 0;
2804}
2805
2806/*
2807 * 7.1.5 SCTP_NODELAY
2808 *
2809 * Turn on/off any Nagle-like algorithm.  This means that packets are
2810 * generally sent as soon as possible and no unnecessary delays are
2811 * introduced, at the cost of more packets in the network.  Expects an
2812 *  integer boolean flag.
2813 */
2814static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2815				   unsigned int optlen)
2816{
2817	int val;
2818
2819	if (optlen < sizeof(int))
2820		return -EINVAL;
2821	if (get_user(val, (int __user *)optval))
2822		return -EFAULT;
2823
2824	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2825	return 0;
2826}
2827
2828/*
2829 *
2830 * 7.1.1 SCTP_RTOINFO
2831 *
2832 * The protocol parameters used to initialize and bound retransmission
2833 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2834 * and modify these parameters.
2835 * All parameters are time values, in milliseconds.  A value of 0, when
2836 * modifying the parameters, indicates that the current value should not
2837 * be changed.
2838 *
2839 */
2840static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2841{
2842	struct sctp_rtoinfo rtoinfo;
2843	struct sctp_association *asoc;
2844	unsigned long rto_min, rto_max;
2845	struct sctp_sock *sp = sctp_sk(sk);
2846
2847	if (optlen != sizeof (struct sctp_rtoinfo))
2848		return -EINVAL;
2849
2850	if (copy_from_user(&rtoinfo, optval, optlen))
2851		return -EFAULT;
2852
2853	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2854
2855	/* Set the values to the specific association */
2856	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2857		return -EINVAL;
2858
2859	rto_max = rtoinfo.srto_max;
2860	rto_min = rtoinfo.srto_min;
2861
2862	if (rto_max)
2863		rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2864	else
2865		rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2866
2867	if (rto_min)
2868		rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2869	else
2870		rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2871
2872	if (rto_min > rto_max)
2873		return -EINVAL;
2874
2875	if (asoc) {
2876		if (rtoinfo.srto_initial != 0)
2877			asoc->rto_initial =
2878				msecs_to_jiffies(rtoinfo.srto_initial);
2879		asoc->rto_max = rto_max;
2880		asoc->rto_min = rto_min;
 
 
2881	} else {
2882		/* If there is no association or the association-id = 0
2883		 * set the values to the endpoint.
2884		 */
 
 
2885		if (rtoinfo.srto_initial != 0)
2886			sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2887		sp->rtoinfo.srto_max = rto_max;
2888		sp->rtoinfo.srto_min = rto_min;
 
 
2889	}
2890
2891	return 0;
2892}
2893
2894/*
2895 *
2896 * 7.1.2 SCTP_ASSOCINFO
2897 *
2898 * This option is used to tune the maximum retransmission attempts
2899 * of the association.
2900 * Returns an error if the new association retransmission value is
2901 * greater than the sum of the retransmission value  of the peer.
2902 * See [SCTP] for more information.
2903 *
2904 */
2905static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2906{
2907
2908	struct sctp_assocparams assocparams;
2909	struct sctp_association *asoc;
2910
2911	if (optlen != sizeof(struct sctp_assocparams))
2912		return -EINVAL;
2913	if (copy_from_user(&assocparams, optval, optlen))
2914		return -EFAULT;
2915
2916	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2917
2918	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2919		return -EINVAL;
2920
2921	/* Set the values to the specific association */
2922	if (asoc) {
2923		if (assocparams.sasoc_asocmaxrxt != 0) {
2924			__u32 path_sum = 0;
2925			int   paths = 0;
2926			struct sctp_transport *peer_addr;
2927
2928			list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2929					transports) {
2930				path_sum += peer_addr->pathmaxrxt;
2931				paths++;
2932			}
2933
2934			/* Only validate asocmaxrxt if we have more than
2935			 * one path/transport.  We do this because path
2936			 * retransmissions are only counted when we have more
2937			 * then one path.
2938			 */
2939			if (paths > 1 &&
2940			    assocparams.sasoc_asocmaxrxt > path_sum)
2941				return -EINVAL;
2942
2943			asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2944		}
2945
2946		if (assocparams.sasoc_cookie_life != 0)
2947			asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
 
 
 
 
 
2948	} else {
2949		/* Set the values to the endpoint */
2950		struct sctp_sock *sp = sctp_sk(sk);
2951
2952		if (assocparams.sasoc_asocmaxrxt != 0)
2953			sp->assocparams.sasoc_asocmaxrxt =
2954						assocparams.sasoc_asocmaxrxt;
2955		if (assocparams.sasoc_cookie_life != 0)
2956			sp->assocparams.sasoc_cookie_life =
2957						assocparams.sasoc_cookie_life;
2958	}
2959	return 0;
2960}
2961
2962/*
2963 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2964 *
2965 * This socket option is a boolean flag which turns on or off mapped V4
2966 * addresses.  If this option is turned on and the socket is type
2967 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2968 * If this option is turned off, then no mapping will be done of V4
2969 * addresses and a user will receive both PF_INET6 and PF_INET type
2970 * addresses on the socket.
2971 */
2972static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2973{
2974	int val;
2975	struct sctp_sock *sp = sctp_sk(sk);
2976
2977	if (optlen < sizeof(int))
2978		return -EINVAL;
2979	if (get_user(val, (int __user *)optval))
2980		return -EFAULT;
2981	if (val)
2982		sp->v4mapped = 1;
2983	else
2984		sp->v4mapped = 0;
2985
2986	return 0;
2987}
2988
2989/*
2990 * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2991 * This option will get or set the maximum size to put in any outgoing
2992 * SCTP DATA chunk.  If a message is larger than this size it will be
2993 * fragmented by SCTP into the specified size.  Note that the underlying
2994 * SCTP implementation may fragment into smaller sized chunks when the
2995 * PMTU of the underlying association is smaller than the value set by
2996 * the user.  The default value for this option is '0' which indicates
2997 * the user is NOT limiting fragmentation and only the PMTU will effect
2998 * SCTP's choice of DATA chunk size.  Note also that values set larger
2999 * than the maximum size of an IP datagram will effectively let SCTP
3000 * control fragmentation (i.e. the same as setting this option to 0).
3001 *
3002 * The following structure is used to access and modify this parameter:
3003 *
3004 * struct sctp_assoc_value {
3005 *   sctp_assoc_t assoc_id;
3006 *   uint32_t assoc_value;
3007 * };
3008 *
3009 * assoc_id:  This parameter is ignored for one-to-one style sockets.
3010 *    For one-to-many style sockets this parameter indicates which
3011 *    association the user is performing an action upon.  Note that if
3012 *    this field's value is zero then the endpoints default value is
3013 *    changed (effecting future associations only).
3014 * assoc_value:  This parameter specifies the maximum size in bytes.
3015 */
3016static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3017{
3018	struct sctp_assoc_value params;
3019	struct sctp_association *asoc;
3020	struct sctp_sock *sp = sctp_sk(sk);
3021	int val;
3022
3023	if (optlen == sizeof(int)) {
3024		pr_warn_ratelimited(DEPRECATED
3025				    "%s (pid %d) "
3026				    "Use of int in maxseg socket option.\n"
3027				    "Use struct sctp_assoc_value instead\n",
3028				    current->comm, task_pid_nr(current));
3029		if (copy_from_user(&val, optval, optlen))
3030			return -EFAULT;
3031		params.assoc_id = 0;
3032	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3033		if (copy_from_user(&params, optval, optlen))
3034			return -EFAULT;
3035		val = params.assoc_value;
3036	} else
3037		return -EINVAL;
3038
3039	if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3040		return -EINVAL;
3041
3042	asoc = sctp_id2assoc(sk, params.assoc_id);
3043	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3044		return -EINVAL;
3045
3046	if (asoc) {
3047		if (val == 0) {
3048			val = asoc->pathmtu;
3049			val -= sp->pf->af->net_header_len;
3050			val -= sizeof(struct sctphdr) +
3051					sizeof(struct sctp_data_chunk);
3052		}
3053		asoc->user_frag = val;
3054		asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3055	} else {
3056		sp->user_frag = val;
3057	}
3058
3059	return 0;
3060}
3061
3062
3063/*
3064 *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3065 *
3066 *   Requests that the peer mark the enclosed address as the association
3067 *   primary. The enclosed address must be one of the association's
3068 *   locally bound addresses. The following structure is used to make a
3069 *   set primary request:
3070 */
3071static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3072					     unsigned int optlen)
3073{
3074	struct net *net = sock_net(sk);
3075	struct sctp_sock	*sp;
3076	struct sctp_association	*asoc = NULL;
3077	struct sctp_setpeerprim	prim;
3078	struct sctp_chunk	*chunk;
3079	struct sctp_af		*af;
3080	int 			err;
3081
3082	sp = sctp_sk(sk);
3083
3084	if (!net->sctp.addip_enable)
3085		return -EPERM;
3086
3087	if (optlen != sizeof(struct sctp_setpeerprim))
3088		return -EINVAL;
3089
3090	if (copy_from_user(&prim, optval, optlen))
3091		return -EFAULT;
3092
3093	asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3094	if (!asoc)
3095		return -EINVAL;
3096
3097	if (!asoc->peer.asconf_capable)
3098		return -EPERM;
3099
3100	if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3101		return -EPERM;
3102
3103	if (!sctp_state(asoc, ESTABLISHED))
3104		return -ENOTCONN;
3105
3106	af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3107	if (!af)
3108		return -EINVAL;
3109
3110	if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3111		return -EADDRNOTAVAIL;
3112
3113	if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3114		return -EADDRNOTAVAIL;
3115
3116	/* Create an ASCONF chunk with SET_PRIMARY parameter	*/
3117	chunk = sctp_make_asconf_set_prim(asoc,
3118					  (union sctp_addr *)&prim.sspp_addr);
3119	if (!chunk)
3120		return -ENOMEM;
3121
3122	err = sctp_send_asconf(asoc, chunk);
3123
3124	pr_debug("%s: we set peer primary addr primitively\n", __func__);
3125
3126	return err;
3127}
3128
3129static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3130					    unsigned int optlen)
3131{
3132	struct sctp_setadaptation adaptation;
3133
3134	if (optlen != sizeof(struct sctp_setadaptation))
3135		return -EINVAL;
3136	if (copy_from_user(&adaptation, optval, optlen))
3137		return -EFAULT;
3138
3139	sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3140
3141	return 0;
3142}
3143
3144/*
3145 * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
3146 *
3147 * The context field in the sctp_sndrcvinfo structure is normally only
3148 * used when a failed message is retrieved holding the value that was
3149 * sent down on the actual send call.  This option allows the setting of
3150 * a default context on an association basis that will be received on
3151 * reading messages from the peer.  This is especially helpful in the
3152 * one-2-many model for an application to keep some reference to an
3153 * internal state machine that is processing messages on the
3154 * association.  Note that the setting of this value only effects
3155 * received messages from the peer and does not effect the value that is
3156 * saved with outbound messages.
3157 */
3158static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3159				   unsigned int optlen)
3160{
3161	struct sctp_assoc_value params;
3162	struct sctp_sock *sp;
3163	struct sctp_association *asoc;
3164
3165	if (optlen != sizeof(struct sctp_assoc_value))
3166		return -EINVAL;
3167	if (copy_from_user(&params, optval, optlen))
3168		return -EFAULT;
3169
3170	sp = sctp_sk(sk);
3171
3172	if (params.assoc_id != 0) {
3173		asoc = sctp_id2assoc(sk, params.assoc_id);
3174		if (!asoc)
3175			return -EINVAL;
3176		asoc->default_rcv_context = params.assoc_value;
3177	} else {
3178		sp->default_rcv_context = params.assoc_value;
3179	}
3180
3181	return 0;
3182}
3183
3184/*
3185 * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3186 *
3187 * This options will at a minimum specify if the implementation is doing
3188 * fragmented interleave.  Fragmented interleave, for a one to many
3189 * socket, is when subsequent calls to receive a message may return
3190 * parts of messages from different associations.  Some implementations
3191 * may allow you to turn this value on or off.  If so, when turned off,
3192 * no fragment interleave will occur (which will cause a head of line
3193 * blocking amongst multiple associations sharing the same one to many
3194 * socket).  When this option is turned on, then each receive call may
3195 * come from a different association (thus the user must receive data
3196 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3197 * association each receive belongs to.
3198 *
3199 * This option takes a boolean value.  A non-zero value indicates that
3200 * fragmented interleave is on.  A value of zero indicates that
3201 * fragmented interleave is off.
3202 *
3203 * Note that it is important that an implementation that allows this
3204 * option to be turned on, have it off by default.  Otherwise an unaware
3205 * application using the one to many model may become confused and act
3206 * incorrectly.
3207 */
3208static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3209					       char __user *optval,
3210					       unsigned int optlen)
3211{
3212	int val;
3213
3214	if (optlen != sizeof(int))
3215		return -EINVAL;
3216	if (get_user(val, (int __user *)optval))
3217		return -EFAULT;
3218
3219	sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3220
3221	return 0;
3222}
3223
3224/*
3225 * 8.1.21.  Set or Get the SCTP Partial Delivery Point
3226 *       (SCTP_PARTIAL_DELIVERY_POINT)
3227 *
3228 * This option will set or get the SCTP partial delivery point.  This
3229 * point is the size of a message where the partial delivery API will be
3230 * invoked to help free up rwnd space for the peer.  Setting this to a
3231 * lower value will cause partial deliveries to happen more often.  The
3232 * calls argument is an integer that sets or gets the partial delivery
3233 * point.  Note also that the call will fail if the user attempts to set
3234 * this value larger than the socket receive buffer size.
3235 *
3236 * Note that any single message having a length smaller than or equal to
3237 * the SCTP partial delivery point will be delivered in one single read
3238 * call as long as the user provided buffer is large enough to hold the
3239 * message.
3240 */
3241static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3242						  char __user *optval,
3243						  unsigned int optlen)
3244{
3245	u32 val;
3246
3247	if (optlen != sizeof(u32))
3248		return -EINVAL;
3249	if (get_user(val, (int __user *)optval))
3250		return -EFAULT;
3251
3252	/* Note: We double the receive buffer from what the user sets
3253	 * it to be, also initial rwnd is based on rcvbuf/2.
3254	 */
3255	if (val > (sk->sk_rcvbuf >> 1))
3256		return -EINVAL;
3257
3258	sctp_sk(sk)->pd_point = val;
3259
3260	return 0; /* is this the right error code? */
3261}
3262
3263/*
3264 * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
3265 *
3266 * This option will allow a user to change the maximum burst of packets
3267 * that can be emitted by this association.  Note that the default value
3268 * is 4, and some implementations may restrict this setting so that it
3269 * can only be lowered.
3270 *
3271 * NOTE: This text doesn't seem right.  Do this on a socket basis with
3272 * future associations inheriting the socket value.
3273 */
3274static int sctp_setsockopt_maxburst(struct sock *sk,
3275				    char __user *optval,
3276				    unsigned int optlen)
3277{
3278	struct sctp_assoc_value params;
3279	struct sctp_sock *sp;
3280	struct sctp_association *asoc;
3281	int val;
3282	int assoc_id = 0;
3283
3284	if (optlen == sizeof(int)) {
3285		pr_warn_ratelimited(DEPRECATED
3286				    "%s (pid %d) "
3287				    "Use of int in max_burst socket option deprecated.\n"
3288				    "Use struct sctp_assoc_value instead\n",
3289				    current->comm, task_pid_nr(current));
3290		if (copy_from_user(&val, optval, optlen))
3291			return -EFAULT;
3292	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3293		if (copy_from_user(&params, optval, optlen))
3294			return -EFAULT;
3295		val = params.assoc_value;
3296		assoc_id = params.assoc_id;
3297	} else
3298		return -EINVAL;
3299
3300	sp = sctp_sk(sk);
3301
3302	if (assoc_id != 0) {
3303		asoc = sctp_id2assoc(sk, assoc_id);
3304		if (!asoc)
3305			return -EINVAL;
3306		asoc->max_burst = val;
3307	} else
3308		sp->max_burst = val;
3309
3310	return 0;
3311}
3312
3313/*
3314 * 7.1.18.  Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3315 *
3316 * This set option adds a chunk type that the user is requesting to be
3317 * received only in an authenticated way.  Changes to the list of chunks
3318 * will only effect future associations on the socket.
3319 */
3320static int sctp_setsockopt_auth_chunk(struct sock *sk,
3321				      char __user *optval,
3322				      unsigned int optlen)
3323{
3324	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3325	struct sctp_authchunk val;
3326
3327	if (!ep->auth_enable)
3328		return -EACCES;
3329
3330	if (optlen != sizeof(struct sctp_authchunk))
3331		return -EINVAL;
3332	if (copy_from_user(&val, optval, optlen))
3333		return -EFAULT;
3334
3335	switch (val.sauth_chunk) {
3336	case SCTP_CID_INIT:
3337	case SCTP_CID_INIT_ACK:
3338	case SCTP_CID_SHUTDOWN_COMPLETE:
3339	case SCTP_CID_AUTH:
3340		return -EINVAL;
3341	}
3342
3343	/* add this chunk id to the endpoint */
3344	return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3345}
3346
3347/*
3348 * 7.1.19.  Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3349 *
3350 * This option gets or sets the list of HMAC algorithms that the local
3351 * endpoint requires the peer to use.
3352 */
3353static int sctp_setsockopt_hmac_ident(struct sock *sk,
3354				      char __user *optval,
3355				      unsigned int optlen)
3356{
3357	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3358	struct sctp_hmacalgo *hmacs;
3359	u32 idents;
3360	int err;
3361
3362	if (!ep->auth_enable)
3363		return -EACCES;
3364
3365	if (optlen < sizeof(struct sctp_hmacalgo))
3366		return -EINVAL;
3367
3368	hmacs = memdup_user(optval, optlen);
3369	if (IS_ERR(hmacs))
3370		return PTR_ERR(hmacs);
3371
3372	idents = hmacs->shmac_num_idents;
3373	if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3374	    (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3375		err = -EINVAL;
3376		goto out;
3377	}
3378
3379	err = sctp_auth_ep_set_hmacs(ep, hmacs);
3380out:
3381	kfree(hmacs);
3382	return err;
3383}
3384
3385/*
3386 * 7.1.20.  Set a shared key (SCTP_AUTH_KEY)
3387 *
3388 * This option will set a shared secret key which is used to build an
3389 * association shared key.
3390 */
3391static int sctp_setsockopt_auth_key(struct sock *sk,
3392				    char __user *optval,
3393				    unsigned int optlen)
3394{
3395	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3396	struct sctp_authkey *authkey;
3397	struct sctp_association *asoc;
3398	int ret;
3399
3400	if (!ep->auth_enable)
3401		return -EACCES;
3402
3403	if (optlen <= sizeof(struct sctp_authkey))
3404		return -EINVAL;
3405
3406	authkey = memdup_user(optval, optlen);
3407	if (IS_ERR(authkey))
3408		return PTR_ERR(authkey);
3409
3410	if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3411		ret = -EINVAL;
3412		goto out;
3413	}
3414
3415	asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3416	if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3417		ret = -EINVAL;
3418		goto out;
3419	}
3420
3421	ret = sctp_auth_set_key(ep, asoc, authkey);
3422out:
3423	kzfree(authkey);
3424	return ret;
3425}
3426
3427/*
3428 * 7.1.21.  Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3429 *
3430 * This option will get or set the active shared key to be used to build
3431 * the association shared key.
3432 */
3433static int sctp_setsockopt_active_key(struct sock *sk,
3434				      char __user *optval,
3435				      unsigned int optlen)
3436{
3437	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3438	struct sctp_authkeyid val;
3439	struct sctp_association *asoc;
3440
3441	if (!ep->auth_enable)
3442		return -EACCES;
3443
3444	if (optlen != sizeof(struct sctp_authkeyid))
3445		return -EINVAL;
3446	if (copy_from_user(&val, optval, optlen))
3447		return -EFAULT;
3448
3449	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3450	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3451		return -EINVAL;
3452
3453	return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
 
3454}
3455
3456/*
3457 * 7.1.22.  Delete a shared key (SCTP_AUTH_DELETE_KEY)
3458 *
3459 * This set option will delete a shared secret key from use.
3460 */
3461static int sctp_setsockopt_del_key(struct sock *sk,
3462				   char __user *optval,
3463				   unsigned int optlen)
3464{
3465	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3466	struct sctp_authkeyid val;
3467	struct sctp_association *asoc;
3468
3469	if (!ep->auth_enable)
3470		return -EACCES;
3471
3472	if (optlen != sizeof(struct sctp_authkeyid))
3473		return -EINVAL;
3474	if (copy_from_user(&val, optval, optlen))
3475		return -EFAULT;
3476
3477	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3478	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3479		return -EINVAL;
3480
3481	return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
 
3482
3483}
3484
3485/*
3486 * 8.1.23 SCTP_AUTO_ASCONF
3487 *
3488 * This option will enable or disable the use of the automatic generation of
3489 * ASCONF chunks to add and delete addresses to an existing association.  Note
3490 * that this option has two caveats namely: a) it only affects sockets that
3491 * are bound to all addresses available to the SCTP stack, and b) the system
3492 * administrator may have an overriding control that turns the ASCONF feature
3493 * off no matter what setting the socket option may have.
3494 * This option expects an integer boolean flag, where a non-zero value turns on
3495 * the option, and a zero value turns off the option.
3496 * Note. In this implementation, socket operation overrides default parameter
3497 * being set by sysctl as well as FreeBSD implementation
3498 */
3499static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3500					unsigned int optlen)
3501{
3502	int val;
3503	struct sctp_sock *sp = sctp_sk(sk);
3504
3505	if (optlen < sizeof(int))
3506		return -EINVAL;
3507	if (get_user(val, (int __user *)optval))
3508		return -EFAULT;
3509	if (!sctp_is_ep_boundall(sk) && val)
3510		return -EINVAL;
3511	if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3512		return 0;
3513
3514	if (val == 0 && sp->do_auto_asconf) {
3515		list_del(&sp->auto_asconf_list);
3516		sp->do_auto_asconf = 0;
3517	} else if (val && !sp->do_auto_asconf) {
3518		list_add_tail(&sp->auto_asconf_list,
3519		    &sock_net(sk)->sctp.auto_asconf_splist);
3520		sp->do_auto_asconf = 1;
3521	}
3522	return 0;
3523}
3524
3525
3526/*
3527 * SCTP_PEER_ADDR_THLDS
3528 *
3529 * This option allows us to alter the partially failed threshold for one or all
3530 * transports in an association.  See Section 6.1 of:
3531 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3532 */
3533static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3534					    char __user *optval,
3535					    unsigned int optlen)
3536{
3537	struct sctp_paddrthlds val;
3538	struct sctp_transport *trans;
3539	struct sctp_association *asoc;
3540
3541	if (optlen < sizeof(struct sctp_paddrthlds))
3542		return -EINVAL;
3543	if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3544			   sizeof(struct sctp_paddrthlds)))
3545		return -EFAULT;
3546
3547
3548	if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3549		asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3550		if (!asoc)
3551			return -ENOENT;
3552		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3553				    transports) {
3554			if (val.spt_pathmaxrxt)
3555				trans->pathmaxrxt = val.spt_pathmaxrxt;
3556			trans->pf_retrans = val.spt_pathpfthld;
3557		}
3558
3559		if (val.spt_pathmaxrxt)
3560			asoc->pathmaxrxt = val.spt_pathmaxrxt;
3561		asoc->pf_retrans = val.spt_pathpfthld;
3562	} else {
3563		trans = sctp_addr_id2transport(sk, &val.spt_address,
3564					       val.spt_assoc_id);
3565		if (!trans)
3566			return -ENOENT;
3567
3568		if (val.spt_pathmaxrxt)
3569			trans->pathmaxrxt = val.spt_pathmaxrxt;
3570		trans->pf_retrans = val.spt_pathpfthld;
3571	}
3572
3573	return 0;
3574}
3575
3576/* API 6.2 setsockopt(), getsockopt()
3577 *
3578 * Applications use setsockopt() and getsockopt() to set or retrieve
3579 * socket options.  Socket options are used to change the default
3580 * behavior of sockets calls.  They are described in Section 7.
3581 *
3582 * The syntax is:
3583 *
3584 *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
3585 *                    int __user *optlen);
3586 *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3587 *                    int optlen);
3588 *
3589 *   sd      - the socket descript.
3590 *   level   - set to IPPROTO_SCTP for all SCTP options.
3591 *   optname - the option name.
3592 *   optval  - the buffer to store the value of the option.
3593 *   optlen  - the size of the buffer.
3594 */
3595static int sctp_setsockopt(struct sock *sk, int level, int optname,
3596			   char __user *optval, unsigned int optlen)
3597{
3598	int retval = 0;
3599
3600	pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
 
3601
3602	/* I can hardly begin to describe how wrong this is.  This is
3603	 * so broken as to be worse than useless.  The API draft
3604	 * REALLY is NOT helpful here...  I am not convinced that the
3605	 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3606	 * are at all well-founded.
3607	 */
3608	if (level != SOL_SCTP) {
3609		struct sctp_af *af = sctp_sk(sk)->pf->af;
3610		retval = af->setsockopt(sk, level, optname, optval, optlen);
3611		goto out_nounlock;
3612	}
3613
3614	lock_sock(sk);
3615
3616	switch (optname) {
3617	case SCTP_SOCKOPT_BINDX_ADD:
3618		/* 'optlen' is the size of the addresses buffer. */
3619		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3620					       optlen, SCTP_BINDX_ADD_ADDR);
3621		break;
3622
3623	case SCTP_SOCKOPT_BINDX_REM:
3624		/* 'optlen' is the size of the addresses buffer. */
3625		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3626					       optlen, SCTP_BINDX_REM_ADDR);
3627		break;
3628
3629	case SCTP_SOCKOPT_CONNECTX_OLD:
3630		/* 'optlen' is the size of the addresses buffer. */
3631		retval = sctp_setsockopt_connectx_old(sk,
3632					    (struct sockaddr __user *)optval,
3633					    optlen);
3634		break;
3635
3636	case SCTP_SOCKOPT_CONNECTX:
3637		/* 'optlen' is the size of the addresses buffer. */
3638		retval = sctp_setsockopt_connectx(sk,
3639					    (struct sockaddr __user *)optval,
3640					    optlen);
3641		break;
3642
3643	case SCTP_DISABLE_FRAGMENTS:
3644		retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3645		break;
3646
3647	case SCTP_EVENTS:
3648		retval = sctp_setsockopt_events(sk, optval, optlen);
3649		break;
3650
3651	case SCTP_AUTOCLOSE:
3652		retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3653		break;
3654
3655	case SCTP_PEER_ADDR_PARAMS:
3656		retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3657		break;
3658
3659	case SCTP_DELAYED_SACK:
3660		retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3661		break;
3662	case SCTP_PARTIAL_DELIVERY_POINT:
3663		retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3664		break;
3665
3666	case SCTP_INITMSG:
3667		retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3668		break;
3669	case SCTP_DEFAULT_SEND_PARAM:
3670		retval = sctp_setsockopt_default_send_param(sk, optval,
3671							    optlen);
3672		break;
3673	case SCTP_PRIMARY_ADDR:
3674		retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3675		break;
3676	case SCTP_SET_PEER_PRIMARY_ADDR:
3677		retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3678		break;
3679	case SCTP_NODELAY:
3680		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3681		break;
3682	case SCTP_RTOINFO:
3683		retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3684		break;
3685	case SCTP_ASSOCINFO:
3686		retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3687		break;
3688	case SCTP_I_WANT_MAPPED_V4_ADDR:
3689		retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3690		break;
3691	case SCTP_MAXSEG:
3692		retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3693		break;
3694	case SCTP_ADAPTATION_LAYER:
3695		retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3696		break;
3697	case SCTP_CONTEXT:
3698		retval = sctp_setsockopt_context(sk, optval, optlen);
3699		break;
3700	case SCTP_FRAGMENT_INTERLEAVE:
3701		retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3702		break;
3703	case SCTP_MAX_BURST:
3704		retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3705		break;
3706	case SCTP_AUTH_CHUNK:
3707		retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3708		break;
3709	case SCTP_HMAC_IDENT:
3710		retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3711		break;
3712	case SCTP_AUTH_KEY:
3713		retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3714		break;
3715	case SCTP_AUTH_ACTIVE_KEY:
3716		retval = sctp_setsockopt_active_key(sk, optval, optlen);
3717		break;
3718	case SCTP_AUTH_DELETE_KEY:
3719		retval = sctp_setsockopt_del_key(sk, optval, optlen);
3720		break;
3721	case SCTP_AUTO_ASCONF:
3722		retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3723		break;
3724	case SCTP_PEER_ADDR_THLDS:
3725		retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3726		break;
3727	default:
3728		retval = -ENOPROTOOPT;
3729		break;
3730	}
3731
3732	release_sock(sk);
3733
3734out_nounlock:
3735	return retval;
3736}
3737
3738/* API 3.1.6 connect() - UDP Style Syntax
3739 *
3740 * An application may use the connect() call in the UDP model to initiate an
3741 * association without sending data.
3742 *
3743 * The syntax is:
3744 *
3745 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3746 *
3747 * sd: the socket descriptor to have a new association added to.
3748 *
3749 * nam: the address structure (either struct sockaddr_in or struct
3750 *    sockaddr_in6 defined in RFC2553 [7]).
3751 *
3752 * len: the size of the address.
3753 */
3754static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3755			int addr_len)
3756{
3757	int err = 0;
3758	struct sctp_af *af;
3759
3760	lock_sock(sk);
3761
3762	pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3763		 addr, addr_len);
3764
3765	/* Validate addr_len before calling common connect/connectx routine. */
3766	af = sctp_get_af_specific(addr->sa_family);
3767	if (!af || addr_len < af->sockaddr_len) {
3768		err = -EINVAL;
3769	} else {
3770		/* Pass correct addr len to common routine (so it knows there
3771		 * is only one address being passed.
3772		 */
3773		err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3774	}
3775
3776	release_sock(sk);
3777	return err;
3778}
3779
3780/* FIXME: Write comments. */
3781static int sctp_disconnect(struct sock *sk, int flags)
3782{
3783	return -EOPNOTSUPP; /* STUB */
3784}
3785
3786/* 4.1.4 accept() - TCP Style Syntax
3787 *
3788 * Applications use accept() call to remove an established SCTP
3789 * association from the accept queue of the endpoint.  A new socket
3790 * descriptor will be returned from accept() to represent the newly
3791 * formed association.
3792 */
3793static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3794{
3795	struct sctp_sock *sp;
3796	struct sctp_endpoint *ep;
3797	struct sock *newsk = NULL;
3798	struct sctp_association *asoc;
3799	long timeo;
3800	int error = 0;
3801
3802	lock_sock(sk);
3803
3804	sp = sctp_sk(sk);
3805	ep = sp->ep;
3806
3807	if (!sctp_style(sk, TCP)) {
3808		error = -EOPNOTSUPP;
3809		goto out;
3810	}
3811
3812	if (!sctp_sstate(sk, LISTENING)) {
3813		error = -EINVAL;
3814		goto out;
3815	}
3816
3817	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3818
3819	error = sctp_wait_for_accept(sk, timeo);
3820	if (error)
3821		goto out;
3822
3823	/* We treat the list of associations on the endpoint as the accept
3824	 * queue and pick the first association on the list.
3825	 */
3826	asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3827
3828	newsk = sp->pf->create_accept_sk(sk, asoc);
3829	if (!newsk) {
3830		error = -ENOMEM;
3831		goto out;
3832	}
3833
3834	/* Populate the fields of the newsk from the oldsk and migrate the
3835	 * asoc to the newsk.
3836	 */
3837	sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3838
3839out:
3840	release_sock(sk);
3841	*err = error;
3842	return newsk;
3843}
3844
3845/* The SCTP ioctl handler. */
3846static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3847{
3848	int rc = -ENOTCONN;
3849
3850	lock_sock(sk);
3851
3852	/*
3853	 * SEQPACKET-style sockets in LISTENING state are valid, for
3854	 * SCTP, so only discard TCP-style sockets in LISTENING state.
3855	 */
3856	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3857		goto out;
3858
3859	switch (cmd) {
3860	case SIOCINQ: {
3861		struct sk_buff *skb;
3862		unsigned int amount = 0;
3863
3864		skb = skb_peek(&sk->sk_receive_queue);
3865		if (skb != NULL) {
3866			/*
3867			 * We will only return the amount of this packet since
3868			 * that is all that will be read.
3869			 */
3870			amount = skb->len;
3871		}
3872		rc = put_user(amount, (int __user *)arg);
3873		break;
3874	}
3875	default:
3876		rc = -ENOIOCTLCMD;
3877		break;
3878	}
3879out:
3880	release_sock(sk);
3881	return rc;
3882}
3883
3884/* This is the function which gets called during socket creation to
3885 * initialized the SCTP-specific portion of the sock.
3886 * The sock structure should already be zero-filled memory.
3887 */
3888static int sctp_init_sock(struct sock *sk)
3889{
3890	struct net *net = sock_net(sk);
3891	struct sctp_sock *sp;
3892
3893	pr_debug("%s: sk:%p\n", __func__, sk);
3894
3895	sp = sctp_sk(sk);
3896
3897	/* Initialize the SCTP per socket area.  */
3898	switch (sk->sk_type) {
3899	case SOCK_SEQPACKET:
3900		sp->type = SCTP_SOCKET_UDP;
3901		break;
3902	case SOCK_STREAM:
3903		sp->type = SCTP_SOCKET_TCP;
3904		break;
3905	default:
3906		return -ESOCKTNOSUPPORT;
3907	}
3908
3909	/* Initialize default send parameters. These parameters can be
3910	 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3911	 */
3912	sp->default_stream = 0;
3913	sp->default_ppid = 0;
3914	sp->default_flags = 0;
3915	sp->default_context = 0;
3916	sp->default_timetolive = 0;
3917
3918	sp->default_rcv_context = 0;
3919	sp->max_burst = net->sctp.max_burst;
3920
3921	sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3922
3923	/* Initialize default setup parameters. These parameters
3924	 * can be modified with the SCTP_INITMSG socket option or
3925	 * overridden by the SCTP_INIT CMSG.
3926	 */
3927	sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
3928	sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
3929	sp->initmsg.sinit_max_attempts   = net->sctp.max_retrans_init;
3930	sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3931
3932	/* Initialize default RTO related parameters.  These parameters can
3933	 * be modified for with the SCTP_RTOINFO socket option.
3934	 */
3935	sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3936	sp->rtoinfo.srto_max     = net->sctp.rto_max;
3937	sp->rtoinfo.srto_min     = net->sctp.rto_min;
3938
3939	/* Initialize default association related parameters. These parameters
3940	 * can be modified with the SCTP_ASSOCINFO socket option.
3941	 */
3942	sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3943	sp->assocparams.sasoc_number_peer_destinations = 0;
3944	sp->assocparams.sasoc_peer_rwnd = 0;
3945	sp->assocparams.sasoc_local_rwnd = 0;
3946	sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3947
3948	/* Initialize default event subscriptions. By default, all the
3949	 * options are off.
3950	 */
3951	memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3952
3953	/* Default Peer Address Parameters.  These defaults can
3954	 * be modified via SCTP_PEER_ADDR_PARAMS
3955	 */
3956	sp->hbinterval  = net->sctp.hb_interval;
3957	sp->pathmaxrxt  = net->sctp.max_retrans_path;
3958	sp->pathmtu     = 0; /* allow default discovery */
3959	sp->sackdelay   = net->sctp.sack_timeout;
3960	sp->sackfreq	= 2;
3961	sp->param_flags = SPP_HB_ENABLE |
3962			  SPP_PMTUD_ENABLE |
3963			  SPP_SACKDELAY_ENABLE;
3964
3965	/* If enabled no SCTP message fragmentation will be performed.
3966	 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3967	 */
3968	sp->disable_fragments = 0;
3969
3970	/* Enable Nagle algorithm by default.  */
3971	sp->nodelay           = 0;
3972
3973	/* Enable by default. */
3974	sp->v4mapped          = 1;
3975
3976	/* Auto-close idle associations after the configured
3977	 * number of seconds.  A value of 0 disables this
3978	 * feature.  Configure through the SCTP_AUTOCLOSE socket option,
3979	 * for UDP-style sockets only.
3980	 */
3981	sp->autoclose         = 0;
3982
3983	/* User specified fragmentation limit. */
3984	sp->user_frag         = 0;
3985
3986	sp->adaptation_ind = 0;
3987
3988	sp->pf = sctp_get_pf_specific(sk->sk_family);
3989
3990	/* Control variables for partial data delivery. */
3991	atomic_set(&sp->pd_mode, 0);
3992	skb_queue_head_init(&sp->pd_lobby);
3993	sp->frag_interleave = 0;
3994
3995	/* Create a per socket endpoint structure.  Even if we
3996	 * change the data structure relationships, this may still
3997	 * be useful for storing pre-connect address information.
3998	 */
3999	sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4000	if (!sp->ep)
4001		return -ENOMEM;
4002
 
4003	sp->hmac = NULL;
4004
4005	sk->sk_destruct = sctp_destruct_sock;
4006
4007	SCTP_DBG_OBJCNT_INC(sock);
4008
4009	local_bh_disable();
4010	percpu_counter_inc(&sctp_sockets_allocated);
4011	sock_prot_inuse_add(net, sk->sk_prot, 1);
4012	if (net->sctp.default_auto_asconf) {
4013		list_add_tail(&sp->auto_asconf_list,
4014		    &net->sctp.auto_asconf_splist);
4015		sp->do_auto_asconf = 1;
4016	} else
4017		sp->do_auto_asconf = 0;
4018	local_bh_enable();
4019
4020	return 0;
4021}
4022
4023/* Cleanup any SCTP per socket resources.  */
4024static void sctp_destroy_sock(struct sock *sk)
4025{
4026	struct sctp_sock *sp;
4027
4028	pr_debug("%s: sk:%p\n", __func__, sk);
4029
4030	/* Release our hold on the endpoint. */
4031	sp = sctp_sk(sk);
4032	/* This could happen during socket init, thus we bail out
4033	 * early, since the rest of the below is not setup either.
4034	 */
4035	if (sp->ep == NULL)
4036		return;
4037
4038	if (sp->do_auto_asconf) {
4039		sp->do_auto_asconf = 0;
4040		list_del(&sp->auto_asconf_list);
4041	}
4042	sctp_endpoint_free(sp->ep);
4043	local_bh_disable();
4044	percpu_counter_dec(&sctp_sockets_allocated);
4045	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4046	local_bh_enable();
4047}
4048
4049/* Triggered when there are no references on the socket anymore */
4050static void sctp_destruct_sock(struct sock *sk)
4051{
4052	struct sctp_sock *sp = sctp_sk(sk);
4053
4054	/* Free up the HMAC transform. */
4055	crypto_free_hash(sp->hmac);
4056
4057	inet_sock_destruct(sk);
4058}
4059
4060/* API 4.1.7 shutdown() - TCP Style Syntax
4061 *     int shutdown(int socket, int how);
4062 *
4063 *     sd      - the socket descriptor of the association to be closed.
4064 *     how     - Specifies the type of shutdown.  The  values  are
4065 *               as follows:
4066 *               SHUT_RD
4067 *                     Disables further receive operations. No SCTP
4068 *                     protocol action is taken.
4069 *               SHUT_WR
4070 *                     Disables further send operations, and initiates
4071 *                     the SCTP shutdown sequence.
4072 *               SHUT_RDWR
4073 *                     Disables further send  and  receive  operations
4074 *                     and initiates the SCTP shutdown sequence.
4075 */
4076static void sctp_shutdown(struct sock *sk, int how)
4077{
4078	struct net *net = sock_net(sk);
4079	struct sctp_endpoint *ep;
4080	struct sctp_association *asoc;
4081
4082	if (!sctp_style(sk, TCP))
4083		return;
4084
4085	if (how & SEND_SHUTDOWN) {
4086		ep = sctp_sk(sk)->ep;
4087		if (!list_empty(&ep->asocs)) {
4088			asoc = list_entry(ep->asocs.next,
4089					  struct sctp_association, asocs);
4090			sctp_primitive_SHUTDOWN(net, asoc, NULL);
4091		}
4092	}
4093}
4094
4095/* 7.2.1 Association Status (SCTP_STATUS)
4096
4097 * Applications can retrieve current status information about an
4098 * association, including association state, peer receiver window size,
4099 * number of unacked data chunks, and number of data chunks pending
4100 * receipt.  This information is read-only.
4101 */
4102static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4103				       char __user *optval,
4104				       int __user *optlen)
4105{
4106	struct sctp_status status;
4107	struct sctp_association *asoc = NULL;
4108	struct sctp_transport *transport;
4109	sctp_assoc_t associd;
4110	int retval = 0;
4111
4112	if (len < sizeof(status)) {
4113		retval = -EINVAL;
4114		goto out;
4115	}
4116
4117	len = sizeof(status);
4118	if (copy_from_user(&status, optval, len)) {
4119		retval = -EFAULT;
4120		goto out;
4121	}
4122
4123	associd = status.sstat_assoc_id;
4124	asoc = sctp_id2assoc(sk, associd);
4125	if (!asoc) {
4126		retval = -EINVAL;
4127		goto out;
4128	}
4129
4130	transport = asoc->peer.primary_path;
4131
4132	status.sstat_assoc_id = sctp_assoc2id(asoc);
4133	status.sstat_state = asoc->state;
4134	status.sstat_rwnd =  asoc->peer.rwnd;
4135	status.sstat_unackdata = asoc->unack_data;
4136
4137	status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4138	status.sstat_instrms = asoc->c.sinit_max_instreams;
4139	status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4140	status.sstat_fragmentation_point = asoc->frag_point;
4141	status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4142	memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4143			transport->af_specific->sockaddr_len);
4144	/* Map ipv4 address into v4-mapped-on-v6 address.  */
4145	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4146		(union sctp_addr *)&status.sstat_primary.spinfo_address);
4147	status.sstat_primary.spinfo_state = transport->state;
4148	status.sstat_primary.spinfo_cwnd = transport->cwnd;
4149	status.sstat_primary.spinfo_srtt = transport->srtt;
4150	status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4151	status.sstat_primary.spinfo_mtu = transport->pathmtu;
4152
4153	if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4154		status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4155
4156	if (put_user(len, optlen)) {
4157		retval = -EFAULT;
4158		goto out;
4159	}
4160
4161	pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4162		 __func__, len, status.sstat_state, status.sstat_rwnd,
4163		 status.sstat_assoc_id);
4164
4165	if (copy_to_user(optval, &status, len)) {
4166		retval = -EFAULT;
4167		goto out;
4168	}
4169
4170out:
4171	return retval;
4172}
4173
4174
4175/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4176 *
4177 * Applications can retrieve information about a specific peer address
4178 * of an association, including its reachability state, congestion
4179 * window, and retransmission timer values.  This information is
4180 * read-only.
4181 */
4182static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4183					  char __user *optval,
4184					  int __user *optlen)
4185{
4186	struct sctp_paddrinfo pinfo;
4187	struct sctp_transport *transport;
4188	int retval = 0;
4189
4190	if (len < sizeof(pinfo)) {
4191		retval = -EINVAL;
4192		goto out;
4193	}
4194
4195	len = sizeof(pinfo);
4196	if (copy_from_user(&pinfo, optval, len)) {
4197		retval = -EFAULT;
4198		goto out;
4199	}
4200
4201	transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4202					   pinfo.spinfo_assoc_id);
4203	if (!transport)
4204		return -EINVAL;
4205
4206	pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4207	pinfo.spinfo_state = transport->state;
4208	pinfo.spinfo_cwnd = transport->cwnd;
4209	pinfo.spinfo_srtt = transport->srtt;
4210	pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4211	pinfo.spinfo_mtu = transport->pathmtu;
4212
4213	if (pinfo.spinfo_state == SCTP_UNKNOWN)
4214		pinfo.spinfo_state = SCTP_ACTIVE;
4215
4216	if (put_user(len, optlen)) {
4217		retval = -EFAULT;
4218		goto out;
4219	}
4220
4221	if (copy_to_user(optval, &pinfo, len)) {
4222		retval = -EFAULT;
4223		goto out;
4224	}
4225
4226out:
4227	return retval;
4228}
4229
4230/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4231 *
4232 * This option is a on/off flag.  If enabled no SCTP message
4233 * fragmentation will be performed.  Instead if a message being sent
4234 * exceeds the current PMTU size, the message will NOT be sent and
4235 * instead a error will be indicated to the user.
4236 */
4237static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4238					char __user *optval, int __user *optlen)
4239{
4240	int val;
4241
4242	if (len < sizeof(int))
4243		return -EINVAL;
4244
4245	len = sizeof(int);
4246	val = (sctp_sk(sk)->disable_fragments == 1);
4247	if (put_user(len, optlen))
4248		return -EFAULT;
4249	if (copy_to_user(optval, &val, len))
4250		return -EFAULT;
4251	return 0;
4252}
4253
4254/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4255 *
4256 * This socket option is used to specify various notifications and
4257 * ancillary data the user wishes to receive.
4258 */
4259static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4260				  int __user *optlen)
4261{
4262	if (len <= 0)
4263		return -EINVAL;
4264	if (len > sizeof(struct sctp_event_subscribe))
4265		len = sizeof(struct sctp_event_subscribe);
4266	if (put_user(len, optlen))
4267		return -EFAULT;
4268	if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4269		return -EFAULT;
4270	return 0;
4271}
4272
4273/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4274 *
4275 * This socket option is applicable to the UDP-style socket only.  When
4276 * set it will cause associations that are idle for more than the
4277 * specified number of seconds to automatically close.  An association
4278 * being idle is defined an association that has NOT sent or received
4279 * user data.  The special value of '0' indicates that no automatic
4280 * close of any associations should be performed.  The option expects an
4281 * integer defining the number of seconds of idle time before an
4282 * association is closed.
4283 */
4284static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4285{
4286	/* Applicable to UDP-style socket only */
4287	if (sctp_style(sk, TCP))
4288		return -EOPNOTSUPP;
4289	if (len < sizeof(int))
4290		return -EINVAL;
4291	len = sizeof(int);
4292	if (put_user(len, optlen))
4293		return -EFAULT;
4294	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4295		return -EFAULT;
4296	return 0;
4297}
4298
4299/* Helper routine to branch off an association to a new socket.  */
4300int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4301{
4302	struct sctp_association *asoc = sctp_id2assoc(sk, id);
4303	struct socket *sock;
4304	struct sctp_af *af;
4305	int err = 0;
4306
4307	if (!asoc)
4308		return -EINVAL;
4309
4310	/* An association cannot be branched off from an already peeled-off
4311	 * socket, nor is this supported for tcp style sockets.
4312	 */
4313	if (!sctp_style(sk, UDP))
4314		return -EINVAL;
4315
4316	/* Create a new socket.  */
4317	err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4318	if (err < 0)
4319		return err;
4320
4321	sctp_copy_sock(sock->sk, sk, asoc);
4322
4323	/* Make peeled-off sockets more like 1-1 accepted sockets.
4324	 * Set the daddr and initialize id to something more random
4325	 */
4326	af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4327	af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4328
4329	/* Populate the fields of the newsk from the oldsk and migrate the
4330	 * asoc to the newsk.
4331	 */
4332	sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4333
4334	*sockp = sock;
4335
4336	return err;
4337}
4338EXPORT_SYMBOL(sctp_do_peeloff);
4339
4340static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4341{
4342	sctp_peeloff_arg_t peeloff;
4343	struct socket *newsock;
4344	struct file *newfile;
4345	int retval = 0;
4346
4347	if (len < sizeof(sctp_peeloff_arg_t))
4348		return -EINVAL;
4349	len = sizeof(sctp_peeloff_arg_t);
4350	if (copy_from_user(&peeloff, optval, len))
4351		return -EFAULT;
4352
4353	retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4354	if (retval < 0)
4355		goto out;
4356
4357	/* Map the socket to an unused fd that can be returned to the user.  */
4358	retval = get_unused_fd_flags(0);
4359	if (retval < 0) {
4360		sock_release(newsock);
4361		goto out;
4362	}
4363
4364	newfile = sock_alloc_file(newsock, 0, NULL);
4365	if (unlikely(IS_ERR(newfile))) {
4366		put_unused_fd(retval);
4367		sock_release(newsock);
4368		return PTR_ERR(newfile);
4369	}
4370
4371	pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4372		 retval);
4373
4374	/* Return the fd mapped to the new socket.  */
4375	if (put_user(len, optlen)) {
4376		fput(newfile);
4377		put_unused_fd(retval);
4378		return -EFAULT;
4379	}
4380	peeloff.sd = retval;
4381	if (copy_to_user(optval, &peeloff, len)) {
4382		fput(newfile);
4383		put_unused_fd(retval);
4384		return -EFAULT;
4385	}
4386	fd_install(retval, newfile);
 
4387out:
4388	return retval;
4389}
4390
4391/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4392 *
4393 * Applications can enable or disable heartbeats for any peer address of
4394 * an association, modify an address's heartbeat interval, force a
4395 * heartbeat to be sent immediately, and adjust the address's maximum
4396 * number of retransmissions sent before an address is considered
4397 * unreachable.  The following structure is used to access and modify an
4398 * address's parameters:
4399 *
4400 *  struct sctp_paddrparams {
4401 *     sctp_assoc_t            spp_assoc_id;
4402 *     struct sockaddr_storage spp_address;
4403 *     uint32_t                spp_hbinterval;
4404 *     uint16_t                spp_pathmaxrxt;
4405 *     uint32_t                spp_pathmtu;
4406 *     uint32_t                spp_sackdelay;
4407 *     uint32_t                spp_flags;
4408 * };
4409 *
4410 *   spp_assoc_id    - (one-to-many style socket) This is filled in the
4411 *                     application, and identifies the association for
4412 *                     this query.
4413 *   spp_address     - This specifies which address is of interest.
4414 *   spp_hbinterval  - This contains the value of the heartbeat interval,
4415 *                     in milliseconds.  If a  value of zero
4416 *                     is present in this field then no changes are to
4417 *                     be made to this parameter.
4418 *   spp_pathmaxrxt  - This contains the maximum number of
4419 *                     retransmissions before this address shall be
4420 *                     considered unreachable. If a  value of zero
4421 *                     is present in this field then no changes are to
4422 *                     be made to this parameter.
4423 *   spp_pathmtu     - When Path MTU discovery is disabled the value
4424 *                     specified here will be the "fixed" path mtu.
4425 *                     Note that if the spp_address field is empty
4426 *                     then all associations on this address will
4427 *                     have this fixed path mtu set upon them.
4428 *
4429 *   spp_sackdelay   - When delayed sack is enabled, this value specifies
4430 *                     the number of milliseconds that sacks will be delayed
4431 *                     for. This value will apply to all addresses of an
4432 *                     association if the spp_address field is empty. Note
4433 *                     also, that if delayed sack is enabled and this
4434 *                     value is set to 0, no change is made to the last
4435 *                     recorded delayed sack timer value.
4436 *
4437 *   spp_flags       - These flags are used to control various features
4438 *                     on an association. The flag field may contain
4439 *                     zero or more of the following options.
4440 *
4441 *                     SPP_HB_ENABLE  - Enable heartbeats on the
4442 *                     specified address. Note that if the address
4443 *                     field is empty all addresses for the association
4444 *                     have heartbeats enabled upon them.
4445 *
4446 *                     SPP_HB_DISABLE - Disable heartbeats on the
4447 *                     speicifed address. Note that if the address
4448 *                     field is empty all addresses for the association
4449 *                     will have their heartbeats disabled. Note also
4450 *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
4451 *                     mutually exclusive, only one of these two should
4452 *                     be specified. Enabling both fields will have
4453 *                     undetermined results.
4454 *
4455 *                     SPP_HB_DEMAND - Request a user initiated heartbeat
4456 *                     to be made immediately.
4457 *
4458 *                     SPP_PMTUD_ENABLE - This field will enable PMTU
4459 *                     discovery upon the specified address. Note that
4460 *                     if the address feild is empty then all addresses
4461 *                     on the association are effected.
4462 *
4463 *                     SPP_PMTUD_DISABLE - This field will disable PMTU
4464 *                     discovery upon the specified address. Note that
4465 *                     if the address feild is empty then all addresses
4466 *                     on the association are effected. Not also that
4467 *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4468 *                     exclusive. Enabling both will have undetermined
4469 *                     results.
4470 *
4471 *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
4472 *                     on delayed sack. The time specified in spp_sackdelay
4473 *                     is used to specify the sack delay for this address. Note
4474 *                     that if spp_address is empty then all addresses will
4475 *                     enable delayed sack and take on the sack delay
4476 *                     value specified in spp_sackdelay.
4477 *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
4478 *                     off delayed sack. If the spp_address field is blank then
4479 *                     delayed sack is disabled for the entire association. Note
4480 *                     also that this field is mutually exclusive to
4481 *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
4482 *                     results.
4483 */
4484static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4485					    char __user *optval, int __user *optlen)
4486{
4487	struct sctp_paddrparams  params;
4488	struct sctp_transport   *trans = NULL;
4489	struct sctp_association *asoc = NULL;
4490	struct sctp_sock        *sp = sctp_sk(sk);
4491
4492	if (len < sizeof(struct sctp_paddrparams))
4493		return -EINVAL;
4494	len = sizeof(struct sctp_paddrparams);
4495	if (copy_from_user(&params, optval, len))
4496		return -EFAULT;
4497
4498	/* If an address other than INADDR_ANY is specified, and
4499	 * no transport is found, then the request is invalid.
4500	 */
4501	if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4502		trans = sctp_addr_id2transport(sk, &params.spp_address,
4503					       params.spp_assoc_id);
4504		if (!trans) {
4505			pr_debug("%s: failed no transport\n", __func__);
4506			return -EINVAL;
4507		}
4508	}
4509
4510	/* Get association, if assoc_id != 0 and the socket is a one
4511	 * to many style socket, and an association was not found, then
4512	 * the id was invalid.
4513	 */
4514	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4515	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4516		pr_debug("%s: failed no association\n", __func__);
4517		return -EINVAL;
4518	}
4519
4520	if (trans) {
4521		/* Fetch transport values. */
4522		params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4523		params.spp_pathmtu    = trans->pathmtu;
4524		params.spp_pathmaxrxt = trans->pathmaxrxt;
4525		params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
4526
4527		/*draft-11 doesn't say what to return in spp_flags*/
4528		params.spp_flags      = trans->param_flags;
4529	} else if (asoc) {
4530		/* Fetch association values. */
4531		params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4532		params.spp_pathmtu    = asoc->pathmtu;
4533		params.spp_pathmaxrxt = asoc->pathmaxrxt;
4534		params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
4535
4536		/*draft-11 doesn't say what to return in spp_flags*/
4537		params.spp_flags      = asoc->param_flags;
4538	} else {
4539		/* Fetch socket values. */
4540		params.spp_hbinterval = sp->hbinterval;
4541		params.spp_pathmtu    = sp->pathmtu;
4542		params.spp_sackdelay  = sp->sackdelay;
4543		params.spp_pathmaxrxt = sp->pathmaxrxt;
4544
4545		/*draft-11 doesn't say what to return in spp_flags*/
4546		params.spp_flags      = sp->param_flags;
4547	}
4548
4549	if (copy_to_user(optval, &params, len))
4550		return -EFAULT;
4551
4552	if (put_user(len, optlen))
4553		return -EFAULT;
4554
4555	return 0;
4556}
4557
4558/*
4559 * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
4560 *
4561 * This option will effect the way delayed acks are performed.  This
4562 * option allows you to get or set the delayed ack time, in
4563 * milliseconds.  It also allows changing the delayed ack frequency.
4564 * Changing the frequency to 1 disables the delayed sack algorithm.  If
4565 * the assoc_id is 0, then this sets or gets the endpoints default
4566 * values.  If the assoc_id field is non-zero, then the set or get
4567 * effects the specified association for the one to many model (the
4568 * assoc_id field is ignored by the one to one model).  Note that if
4569 * sack_delay or sack_freq are 0 when setting this option, then the
4570 * current values will remain unchanged.
4571 *
4572 * struct sctp_sack_info {
4573 *     sctp_assoc_t            sack_assoc_id;
4574 *     uint32_t                sack_delay;
4575 *     uint32_t                sack_freq;
4576 * };
4577 *
4578 * sack_assoc_id -  This parameter, indicates which association the user
4579 *    is performing an action upon.  Note that if this field's value is
4580 *    zero then the endpoints default value is changed (effecting future
4581 *    associations only).
4582 *
4583 * sack_delay -  This parameter contains the number of milliseconds that
4584 *    the user is requesting the delayed ACK timer be set to.  Note that
4585 *    this value is defined in the standard to be between 200 and 500
4586 *    milliseconds.
4587 *
4588 * sack_freq -  This parameter contains the number of packets that must
4589 *    be received before a sack is sent without waiting for the delay
4590 *    timer to expire.  The default value for this is 2, setting this
4591 *    value to 1 will disable the delayed sack algorithm.
4592 */
4593static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4594					    char __user *optval,
4595					    int __user *optlen)
4596{
4597	struct sctp_sack_info    params;
4598	struct sctp_association *asoc = NULL;
4599	struct sctp_sock        *sp = sctp_sk(sk);
4600
4601	if (len >= sizeof(struct sctp_sack_info)) {
4602		len = sizeof(struct sctp_sack_info);
4603
4604		if (copy_from_user(&params, optval, len))
4605			return -EFAULT;
4606	} else if (len == sizeof(struct sctp_assoc_value)) {
4607		pr_warn_ratelimited(DEPRECATED
4608				    "%s (pid %d) "
4609				    "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4610				    "Use struct sctp_sack_info instead\n",
4611				    current->comm, task_pid_nr(current));
4612		if (copy_from_user(&params, optval, len))
4613			return -EFAULT;
4614	} else
4615		return -EINVAL;
4616
4617	/* Get association, if sack_assoc_id != 0 and the socket is a one
4618	 * to many style socket, and an association was not found, then
4619	 * the id was invalid.
4620	 */
4621	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4622	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4623		return -EINVAL;
4624
4625	if (asoc) {
4626		/* Fetch association values. */
4627		if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4628			params.sack_delay = jiffies_to_msecs(
4629				asoc->sackdelay);
4630			params.sack_freq = asoc->sackfreq;
4631
4632		} else {
4633			params.sack_delay = 0;
4634			params.sack_freq = 1;
4635		}
4636	} else {
4637		/* Fetch socket values. */
4638		if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4639			params.sack_delay  = sp->sackdelay;
4640			params.sack_freq = sp->sackfreq;
4641		} else {
4642			params.sack_delay  = 0;
4643			params.sack_freq = 1;
4644		}
4645	}
4646
4647	if (copy_to_user(optval, &params, len))
4648		return -EFAULT;
4649
4650	if (put_user(len, optlen))
4651		return -EFAULT;
4652
4653	return 0;
4654}
4655
4656/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4657 *
4658 * Applications can specify protocol parameters for the default association
4659 * initialization.  The option name argument to setsockopt() and getsockopt()
4660 * is SCTP_INITMSG.
4661 *
4662 * Setting initialization parameters is effective only on an unconnected
4663 * socket (for UDP-style sockets only future associations are effected
4664 * by the change).  With TCP-style sockets, this option is inherited by
4665 * sockets derived from a listener socket.
4666 */
4667static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4668{
4669	if (len < sizeof(struct sctp_initmsg))
4670		return -EINVAL;
4671	len = sizeof(struct sctp_initmsg);
4672	if (put_user(len, optlen))
4673		return -EFAULT;
4674	if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4675		return -EFAULT;
4676	return 0;
4677}
4678
4679
4680static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4681				      char __user *optval, int __user *optlen)
4682{
4683	struct sctp_association *asoc;
4684	int cnt = 0;
4685	struct sctp_getaddrs getaddrs;
4686	struct sctp_transport *from;
4687	void __user *to;
4688	union sctp_addr temp;
4689	struct sctp_sock *sp = sctp_sk(sk);
4690	int addrlen;
4691	size_t space_left;
4692	int bytes_copied;
4693
4694	if (len < sizeof(struct sctp_getaddrs))
4695		return -EINVAL;
4696
4697	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4698		return -EFAULT;
4699
4700	/* For UDP-style sockets, id specifies the association to query.  */
4701	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4702	if (!asoc)
4703		return -EINVAL;
4704
4705	to = optval + offsetof(struct sctp_getaddrs, addrs);
4706	space_left = len - offsetof(struct sctp_getaddrs, addrs);
4707
4708	list_for_each_entry(from, &asoc->peer.transport_addr_list,
4709				transports) {
4710		memcpy(&temp, &from->ipaddr, sizeof(temp));
4711		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4712		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4713		if (space_left < addrlen)
4714			return -ENOMEM;
4715		if (copy_to_user(to, &temp, addrlen))
4716			return -EFAULT;
4717		to += addrlen;
4718		cnt++;
4719		space_left -= addrlen;
4720	}
4721
4722	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4723		return -EFAULT;
4724	bytes_copied = ((char __user *)to) - optval;
4725	if (put_user(bytes_copied, optlen))
4726		return -EFAULT;
4727
4728	return 0;
4729}
4730
4731static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4732			    size_t space_left, int *bytes_copied)
4733{
4734	struct sctp_sockaddr_entry *addr;
4735	union sctp_addr temp;
4736	int cnt = 0;
4737	int addrlen;
4738	struct net *net = sock_net(sk);
4739
4740	rcu_read_lock();
4741	list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4742		if (!addr->valid)
4743			continue;
4744
4745		if ((PF_INET == sk->sk_family) &&
4746		    (AF_INET6 == addr->a.sa.sa_family))
4747			continue;
4748		if ((PF_INET6 == sk->sk_family) &&
4749		    inet_v6_ipv6only(sk) &&
4750		    (AF_INET == addr->a.sa.sa_family))
4751			continue;
4752		memcpy(&temp, &addr->a, sizeof(temp));
4753		if (!temp.v4.sin_port)
4754			temp.v4.sin_port = htons(port);
4755
4756		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4757								&temp);
4758		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4759		if (space_left < addrlen) {
4760			cnt =  -ENOMEM;
4761			break;
4762		}
4763		memcpy(to, &temp, addrlen);
4764
4765		to += addrlen;
4766		cnt++;
4767		space_left -= addrlen;
4768		*bytes_copied += addrlen;
4769	}
4770	rcu_read_unlock();
4771
4772	return cnt;
4773}
4774
4775
4776static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4777				       char __user *optval, int __user *optlen)
4778{
4779	struct sctp_bind_addr *bp;
4780	struct sctp_association *asoc;
4781	int cnt = 0;
4782	struct sctp_getaddrs getaddrs;
4783	struct sctp_sockaddr_entry *addr;
4784	void __user *to;
4785	union sctp_addr temp;
4786	struct sctp_sock *sp = sctp_sk(sk);
4787	int addrlen;
4788	int err = 0;
4789	size_t space_left;
4790	int bytes_copied = 0;
4791	void *addrs;
4792	void *buf;
4793
4794	if (len < sizeof(struct sctp_getaddrs))
4795		return -EINVAL;
4796
4797	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4798		return -EFAULT;
4799
4800	/*
4801	 *  For UDP-style sockets, id specifies the association to query.
4802	 *  If the id field is set to the value '0' then the locally bound
4803	 *  addresses are returned without regard to any particular
4804	 *  association.
4805	 */
4806	if (0 == getaddrs.assoc_id) {
4807		bp = &sctp_sk(sk)->ep->base.bind_addr;
4808	} else {
4809		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4810		if (!asoc)
4811			return -EINVAL;
4812		bp = &asoc->base.bind_addr;
4813	}
4814
4815	to = optval + offsetof(struct sctp_getaddrs, addrs);
4816	space_left = len - offsetof(struct sctp_getaddrs, addrs);
4817
4818	addrs = kmalloc(space_left, GFP_KERNEL);
4819	if (!addrs)
4820		return -ENOMEM;
4821
4822	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4823	 * addresses from the global local address list.
4824	 */
4825	if (sctp_list_single_entry(&bp->address_list)) {
4826		addr = list_entry(bp->address_list.next,
4827				  struct sctp_sockaddr_entry, list);
4828		if (sctp_is_any(sk, &addr->a)) {
4829			cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4830						space_left, &bytes_copied);
4831			if (cnt < 0) {
4832				err = cnt;
4833				goto out;
4834			}
4835			goto copy_getaddrs;
4836		}
4837	}
4838
4839	buf = addrs;
4840	/* Protection on the bound address list is not needed since
4841	 * in the socket option context we hold a socket lock and
4842	 * thus the bound address list can't change.
4843	 */
4844	list_for_each_entry(addr, &bp->address_list, list) {
4845		memcpy(&temp, &addr->a, sizeof(temp));
4846		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4847		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4848		if (space_left < addrlen) {
4849			err =  -ENOMEM; /*fixme: right error?*/
4850			goto out;
4851		}
4852		memcpy(buf, &temp, addrlen);
4853		buf += addrlen;
4854		bytes_copied += addrlen;
4855		cnt++;
4856		space_left -= addrlen;
4857	}
4858
4859copy_getaddrs:
4860	if (copy_to_user(to, addrs, bytes_copied)) {
4861		err = -EFAULT;
4862		goto out;
4863	}
4864	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4865		err = -EFAULT;
4866		goto out;
4867	}
4868	if (put_user(bytes_copied, optlen))
4869		err = -EFAULT;
4870out:
4871	kfree(addrs);
4872	return err;
4873}
4874
4875/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4876 *
4877 * Requests that the local SCTP stack use the enclosed peer address as
4878 * the association primary.  The enclosed address must be one of the
4879 * association peer's addresses.
4880 */
4881static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4882					char __user *optval, int __user *optlen)
4883{
4884	struct sctp_prim prim;
4885	struct sctp_association *asoc;
4886	struct sctp_sock *sp = sctp_sk(sk);
4887
4888	if (len < sizeof(struct sctp_prim))
4889		return -EINVAL;
4890
4891	len = sizeof(struct sctp_prim);
4892
4893	if (copy_from_user(&prim, optval, len))
4894		return -EFAULT;
4895
4896	asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4897	if (!asoc)
4898		return -EINVAL;
4899
4900	if (!asoc->peer.primary_path)
4901		return -ENOTCONN;
4902
4903	memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4904		asoc->peer.primary_path->af_specific->sockaddr_len);
4905
4906	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4907			(union sctp_addr *)&prim.ssp_addr);
4908
4909	if (put_user(len, optlen))
4910		return -EFAULT;
4911	if (copy_to_user(optval, &prim, len))
4912		return -EFAULT;
4913
4914	return 0;
4915}
4916
4917/*
4918 * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4919 *
4920 * Requests that the local endpoint set the specified Adaptation Layer
4921 * Indication parameter for all future INIT and INIT-ACK exchanges.
4922 */
4923static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4924				  char __user *optval, int __user *optlen)
4925{
4926	struct sctp_setadaptation adaptation;
4927
4928	if (len < sizeof(struct sctp_setadaptation))
4929		return -EINVAL;
4930
4931	len = sizeof(struct sctp_setadaptation);
4932
4933	adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4934
4935	if (put_user(len, optlen))
4936		return -EFAULT;
4937	if (copy_to_user(optval, &adaptation, len))
4938		return -EFAULT;
4939
4940	return 0;
4941}
4942
4943/*
4944 *
4945 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4946 *
4947 *   Applications that wish to use the sendto() system call may wish to
4948 *   specify a default set of parameters that would normally be supplied
4949 *   through the inclusion of ancillary data.  This socket option allows
4950 *   such an application to set the default sctp_sndrcvinfo structure.
4951
4952
4953 *   The application that wishes to use this socket option simply passes
4954 *   in to this call the sctp_sndrcvinfo structure defined in Section
4955 *   5.2.2) The input parameters accepted by this call include
4956 *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4957 *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
4958 *   to this call if the caller is using the UDP model.
4959 *
4960 *   For getsockopt, it get the default sctp_sndrcvinfo structure.
4961 */
4962static int sctp_getsockopt_default_send_param(struct sock *sk,
4963					int len, char __user *optval,
4964					int __user *optlen)
4965{
4966	struct sctp_sndrcvinfo info;
4967	struct sctp_association *asoc;
4968	struct sctp_sock *sp = sctp_sk(sk);
4969
4970	if (len < sizeof(struct sctp_sndrcvinfo))
4971		return -EINVAL;
4972
4973	len = sizeof(struct sctp_sndrcvinfo);
4974
4975	if (copy_from_user(&info, optval, len))
4976		return -EFAULT;
4977
4978	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4979	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4980		return -EINVAL;
4981
4982	if (asoc) {
4983		info.sinfo_stream = asoc->default_stream;
4984		info.sinfo_flags = asoc->default_flags;
4985		info.sinfo_ppid = asoc->default_ppid;
4986		info.sinfo_context = asoc->default_context;
4987		info.sinfo_timetolive = asoc->default_timetolive;
4988	} else {
4989		info.sinfo_stream = sp->default_stream;
4990		info.sinfo_flags = sp->default_flags;
4991		info.sinfo_ppid = sp->default_ppid;
4992		info.sinfo_context = sp->default_context;
4993		info.sinfo_timetolive = sp->default_timetolive;
4994	}
4995
4996	if (put_user(len, optlen))
4997		return -EFAULT;
4998	if (copy_to_user(optval, &info, len))
4999		return -EFAULT;
5000
5001	return 0;
5002}
5003
5004/*
5005 *
5006 * 7.1.5 SCTP_NODELAY
5007 *
5008 * Turn on/off any Nagle-like algorithm.  This means that packets are
5009 * generally sent as soon as possible and no unnecessary delays are
5010 * introduced, at the cost of more packets in the network.  Expects an
5011 * integer boolean flag.
5012 */
5013
5014static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5015				   char __user *optval, int __user *optlen)
5016{
5017	int val;
5018
5019	if (len < sizeof(int))
5020		return -EINVAL;
5021
5022	len = sizeof(int);
5023	val = (sctp_sk(sk)->nodelay == 1);
5024	if (put_user(len, optlen))
5025		return -EFAULT;
5026	if (copy_to_user(optval, &val, len))
5027		return -EFAULT;
5028	return 0;
5029}
5030
5031/*
5032 *
5033 * 7.1.1 SCTP_RTOINFO
5034 *
5035 * The protocol parameters used to initialize and bound retransmission
5036 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5037 * and modify these parameters.
5038 * All parameters are time values, in milliseconds.  A value of 0, when
5039 * modifying the parameters, indicates that the current value should not
5040 * be changed.
5041 *
5042 */
5043static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5044				char __user *optval,
5045				int __user *optlen) {
5046	struct sctp_rtoinfo rtoinfo;
5047	struct sctp_association *asoc;
5048
5049	if (len < sizeof (struct sctp_rtoinfo))
5050		return -EINVAL;
5051
5052	len = sizeof(struct sctp_rtoinfo);
5053
5054	if (copy_from_user(&rtoinfo, optval, len))
5055		return -EFAULT;
5056
5057	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5058
5059	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5060		return -EINVAL;
5061
5062	/* Values corresponding to the specific association. */
5063	if (asoc) {
5064		rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5065		rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5066		rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5067	} else {
5068		/* Values corresponding to the endpoint. */
5069		struct sctp_sock *sp = sctp_sk(sk);
5070
5071		rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5072		rtoinfo.srto_max = sp->rtoinfo.srto_max;
5073		rtoinfo.srto_min = sp->rtoinfo.srto_min;
5074	}
5075
5076	if (put_user(len, optlen))
5077		return -EFAULT;
5078
5079	if (copy_to_user(optval, &rtoinfo, len))
5080		return -EFAULT;
5081
5082	return 0;
5083}
5084
5085/*
5086 *
5087 * 7.1.2 SCTP_ASSOCINFO
5088 *
5089 * This option is used to tune the maximum retransmission attempts
5090 * of the association.
5091 * Returns an error if the new association retransmission value is
5092 * greater than the sum of the retransmission value  of the peer.
5093 * See [SCTP] for more information.
5094 *
5095 */
5096static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5097				     char __user *optval,
5098				     int __user *optlen)
5099{
5100
5101	struct sctp_assocparams assocparams;
5102	struct sctp_association *asoc;
5103	struct list_head *pos;
5104	int cnt = 0;
5105
5106	if (len < sizeof (struct sctp_assocparams))
5107		return -EINVAL;
5108
5109	len = sizeof(struct sctp_assocparams);
5110
5111	if (copy_from_user(&assocparams, optval, len))
5112		return -EFAULT;
5113
5114	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5115
5116	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5117		return -EINVAL;
5118
5119	/* Values correspoinding to the specific association */
5120	if (asoc) {
5121		assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5122		assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5123		assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5124		assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
 
 
 
5125
5126		list_for_each(pos, &asoc->peer.transport_addr_list) {
5127			cnt++;
5128		}
5129
5130		assocparams.sasoc_number_peer_destinations = cnt;
5131	} else {
5132		/* Values corresponding to the endpoint */
5133		struct sctp_sock *sp = sctp_sk(sk);
5134
5135		assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5136		assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5137		assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5138		assocparams.sasoc_cookie_life =
5139					sp->assocparams.sasoc_cookie_life;
5140		assocparams.sasoc_number_peer_destinations =
5141					sp->assocparams.
5142					sasoc_number_peer_destinations;
5143	}
5144
5145	if (put_user(len, optlen))
5146		return -EFAULT;
5147
5148	if (copy_to_user(optval, &assocparams, len))
5149		return -EFAULT;
5150
5151	return 0;
5152}
5153
5154/*
5155 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5156 *
5157 * This socket option is a boolean flag which turns on or off mapped V4
5158 * addresses.  If this option is turned on and the socket is type
5159 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5160 * If this option is turned off, then no mapping will be done of V4
5161 * addresses and a user will receive both PF_INET6 and PF_INET type
5162 * addresses on the socket.
5163 */
5164static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5165				    char __user *optval, int __user *optlen)
5166{
5167	int val;
5168	struct sctp_sock *sp = sctp_sk(sk);
5169
5170	if (len < sizeof(int))
5171		return -EINVAL;
5172
5173	len = sizeof(int);
5174	val = sp->v4mapped;
5175	if (put_user(len, optlen))
5176		return -EFAULT;
5177	if (copy_to_user(optval, &val, len))
5178		return -EFAULT;
5179
5180	return 0;
5181}
5182
5183/*
5184 * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
5185 * (chapter and verse is quoted at sctp_setsockopt_context())
5186 */
5187static int sctp_getsockopt_context(struct sock *sk, int len,
5188				   char __user *optval, int __user *optlen)
5189{
5190	struct sctp_assoc_value params;
5191	struct sctp_sock *sp;
5192	struct sctp_association *asoc;
5193
5194	if (len < sizeof(struct sctp_assoc_value))
5195		return -EINVAL;
5196
5197	len = sizeof(struct sctp_assoc_value);
5198
5199	if (copy_from_user(&params, optval, len))
5200		return -EFAULT;
5201
5202	sp = sctp_sk(sk);
5203
5204	if (params.assoc_id != 0) {
5205		asoc = sctp_id2assoc(sk, params.assoc_id);
5206		if (!asoc)
5207			return -EINVAL;
5208		params.assoc_value = asoc->default_rcv_context;
5209	} else {
5210		params.assoc_value = sp->default_rcv_context;
5211	}
5212
5213	if (put_user(len, optlen))
5214		return -EFAULT;
5215	if (copy_to_user(optval, &params, len))
5216		return -EFAULT;
5217
5218	return 0;
5219}
5220
5221/*
5222 * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5223 * This option will get or set the maximum size to put in any outgoing
5224 * SCTP DATA chunk.  If a message is larger than this size it will be
5225 * fragmented by SCTP into the specified size.  Note that the underlying
5226 * SCTP implementation may fragment into smaller sized chunks when the
5227 * PMTU of the underlying association is smaller than the value set by
5228 * the user.  The default value for this option is '0' which indicates
5229 * the user is NOT limiting fragmentation and only the PMTU will effect
5230 * SCTP's choice of DATA chunk size.  Note also that values set larger
5231 * than the maximum size of an IP datagram will effectively let SCTP
5232 * control fragmentation (i.e. the same as setting this option to 0).
5233 *
5234 * The following structure is used to access and modify this parameter:
5235 *
5236 * struct sctp_assoc_value {
5237 *   sctp_assoc_t assoc_id;
5238 *   uint32_t assoc_value;
5239 * };
5240 *
5241 * assoc_id:  This parameter is ignored for one-to-one style sockets.
5242 *    For one-to-many style sockets this parameter indicates which
5243 *    association the user is performing an action upon.  Note that if
5244 *    this field's value is zero then the endpoints default value is
5245 *    changed (effecting future associations only).
5246 * assoc_value:  This parameter specifies the maximum size in bytes.
5247 */
5248static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5249				  char __user *optval, int __user *optlen)
5250{
5251	struct sctp_assoc_value params;
5252	struct sctp_association *asoc;
5253
5254	if (len == sizeof(int)) {
5255		pr_warn_ratelimited(DEPRECATED
5256				    "%s (pid %d) "
5257				    "Use of int in maxseg socket option.\n"
5258				    "Use struct sctp_assoc_value instead\n",
5259				    current->comm, task_pid_nr(current));
5260		params.assoc_id = 0;
5261	} else if (len >= sizeof(struct sctp_assoc_value)) {
5262		len = sizeof(struct sctp_assoc_value);
5263		if (copy_from_user(&params, optval, sizeof(params)))
5264			return -EFAULT;
5265	} else
5266		return -EINVAL;
5267
5268	asoc = sctp_id2assoc(sk, params.assoc_id);
5269	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5270		return -EINVAL;
5271
5272	if (asoc)
5273		params.assoc_value = asoc->frag_point;
5274	else
5275		params.assoc_value = sctp_sk(sk)->user_frag;
5276
5277	if (put_user(len, optlen))
5278		return -EFAULT;
5279	if (len == sizeof(int)) {
5280		if (copy_to_user(optval, &params.assoc_value, len))
5281			return -EFAULT;
5282	} else {
5283		if (copy_to_user(optval, &params, len))
5284			return -EFAULT;
5285	}
5286
5287	return 0;
5288}
5289
5290/*
5291 * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5292 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5293 */
5294static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5295					       char __user *optval, int __user *optlen)
5296{
5297	int val;
5298
5299	if (len < sizeof(int))
5300		return -EINVAL;
5301
5302	len = sizeof(int);
5303
5304	val = sctp_sk(sk)->frag_interleave;
5305	if (put_user(len, optlen))
5306		return -EFAULT;
5307	if (copy_to_user(optval, &val, len))
5308		return -EFAULT;
5309
5310	return 0;
5311}
5312
5313/*
5314 * 7.1.25.  Set or Get the sctp partial delivery point
5315 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5316 */
5317static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5318						  char __user *optval,
5319						  int __user *optlen)
5320{
5321	u32 val;
5322
5323	if (len < sizeof(u32))
5324		return -EINVAL;
5325
5326	len = sizeof(u32);
5327
5328	val = sctp_sk(sk)->pd_point;
5329	if (put_user(len, optlen))
5330		return -EFAULT;
5331	if (copy_to_user(optval, &val, len))
5332		return -EFAULT;
5333
5334	return 0;
5335}
5336
5337/*
5338 * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
5339 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5340 */
5341static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5342				    char __user *optval,
5343				    int __user *optlen)
5344{
5345	struct sctp_assoc_value params;
5346	struct sctp_sock *sp;
5347	struct sctp_association *asoc;
5348
5349	if (len == sizeof(int)) {
5350		pr_warn_ratelimited(DEPRECATED
5351				    "%s (pid %d) "
5352				    "Use of int in max_burst socket option.\n"
5353				    "Use struct sctp_assoc_value instead\n",
5354				    current->comm, task_pid_nr(current));
5355		params.assoc_id = 0;
5356	} else if (len >= sizeof(struct sctp_assoc_value)) {
5357		len = sizeof(struct sctp_assoc_value);
5358		if (copy_from_user(&params, optval, len))
5359			return -EFAULT;
5360	} else
5361		return -EINVAL;
5362
5363	sp = sctp_sk(sk);
5364
5365	if (params.assoc_id != 0) {
5366		asoc = sctp_id2assoc(sk, params.assoc_id);
5367		if (!asoc)
5368			return -EINVAL;
5369		params.assoc_value = asoc->max_burst;
5370	} else
5371		params.assoc_value = sp->max_burst;
5372
5373	if (len == sizeof(int)) {
5374		if (copy_to_user(optval, &params.assoc_value, len))
5375			return -EFAULT;
5376	} else {
5377		if (copy_to_user(optval, &params, len))
5378			return -EFAULT;
5379	}
5380
5381	return 0;
5382
5383}
5384
5385static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5386				    char __user *optval, int __user *optlen)
5387{
5388	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5389	struct sctp_hmacalgo  __user *p = (void __user *)optval;
5390	struct sctp_hmac_algo_param *hmacs;
5391	__u16 data_len = 0;
5392	u32 num_idents;
5393
5394	if (!ep->auth_enable)
5395		return -EACCES;
5396
5397	hmacs = ep->auth_hmacs_list;
5398	data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5399
5400	if (len < sizeof(struct sctp_hmacalgo) + data_len)
5401		return -EINVAL;
5402
5403	len = sizeof(struct sctp_hmacalgo) + data_len;
5404	num_idents = data_len / sizeof(u16);
5405
5406	if (put_user(len, optlen))
5407		return -EFAULT;
5408	if (put_user(num_idents, &p->shmac_num_idents))
5409		return -EFAULT;
5410	if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5411		return -EFAULT;
5412	return 0;
5413}
5414
5415static int sctp_getsockopt_active_key(struct sock *sk, int len,
5416				    char __user *optval, int __user *optlen)
5417{
5418	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5419	struct sctp_authkeyid val;
5420	struct sctp_association *asoc;
5421
5422	if (!ep->auth_enable)
5423		return -EACCES;
5424
5425	if (len < sizeof(struct sctp_authkeyid))
5426		return -EINVAL;
5427	if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5428		return -EFAULT;
5429
5430	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5431	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5432		return -EINVAL;
5433
5434	if (asoc)
5435		val.scact_keynumber = asoc->active_key_id;
5436	else
5437		val.scact_keynumber = ep->active_key_id;
5438
5439	len = sizeof(struct sctp_authkeyid);
5440	if (put_user(len, optlen))
5441		return -EFAULT;
5442	if (copy_to_user(optval, &val, len))
5443		return -EFAULT;
5444
5445	return 0;
5446}
5447
5448static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5449				    char __user *optval, int __user *optlen)
5450{
5451	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5452	struct sctp_authchunks __user *p = (void __user *)optval;
5453	struct sctp_authchunks val;
5454	struct sctp_association *asoc;
5455	struct sctp_chunks_param *ch;
5456	u32    num_chunks = 0;
5457	char __user *to;
5458
5459	if (!ep->auth_enable)
5460		return -EACCES;
5461
5462	if (len < sizeof(struct sctp_authchunks))
5463		return -EINVAL;
5464
5465	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5466		return -EFAULT;
5467
5468	to = p->gauth_chunks;
5469	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5470	if (!asoc)
5471		return -EINVAL;
5472
5473	ch = asoc->peer.peer_chunks;
5474	if (!ch)
5475		goto num;
5476
5477	/* See if the user provided enough room for all the data */
5478	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5479	if (len < num_chunks)
5480		return -EINVAL;
5481
5482	if (copy_to_user(to, ch->chunks, num_chunks))
5483		return -EFAULT;
5484num:
5485	len = sizeof(struct sctp_authchunks) + num_chunks;
5486	if (put_user(len, optlen))
5487		return -EFAULT;
5488	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5489		return -EFAULT;
5490	return 0;
5491}
5492
5493static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5494				    char __user *optval, int __user *optlen)
5495{
5496	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5497	struct sctp_authchunks __user *p = (void __user *)optval;
5498	struct sctp_authchunks val;
5499	struct sctp_association *asoc;
5500	struct sctp_chunks_param *ch;
5501	u32    num_chunks = 0;
5502	char __user *to;
5503
5504	if (!ep->auth_enable)
5505		return -EACCES;
5506
5507	if (len < sizeof(struct sctp_authchunks))
5508		return -EINVAL;
5509
5510	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5511		return -EFAULT;
5512
5513	to = p->gauth_chunks;
5514	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5515	if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5516		return -EINVAL;
5517
5518	if (asoc)
5519		ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5520	else
5521		ch = ep->auth_chunk_list;
5522
5523	if (!ch)
5524		goto num;
5525
5526	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5527	if (len < sizeof(struct sctp_authchunks) + num_chunks)
5528		return -EINVAL;
5529
5530	if (copy_to_user(to, ch->chunks, num_chunks))
5531		return -EFAULT;
5532num:
5533	len = sizeof(struct sctp_authchunks) + num_chunks;
5534	if (put_user(len, optlen))
5535		return -EFAULT;
5536	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5537		return -EFAULT;
5538
5539	return 0;
5540}
5541
5542/*
5543 * 8.2.5.  Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5544 * This option gets the current number of associations that are attached
5545 * to a one-to-many style socket.  The option value is an uint32_t.
5546 */
5547static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5548				    char __user *optval, int __user *optlen)
5549{
5550	struct sctp_sock *sp = sctp_sk(sk);
5551	struct sctp_association *asoc;
5552	u32 val = 0;
5553
5554	if (sctp_style(sk, TCP))
5555		return -EOPNOTSUPP;
5556
5557	if (len < sizeof(u32))
5558		return -EINVAL;
5559
5560	len = sizeof(u32);
5561
5562	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5563		val++;
5564	}
5565
5566	if (put_user(len, optlen))
5567		return -EFAULT;
5568	if (copy_to_user(optval, &val, len))
5569		return -EFAULT;
5570
5571	return 0;
5572}
5573
5574/*
5575 * 8.1.23 SCTP_AUTO_ASCONF
5576 * See the corresponding setsockopt entry as description
5577 */
5578static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5579				   char __user *optval, int __user *optlen)
5580{
5581	int val = 0;
5582
5583	if (len < sizeof(int))
5584		return -EINVAL;
5585
5586	len = sizeof(int);
5587	if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5588		val = 1;
5589	if (put_user(len, optlen))
5590		return -EFAULT;
5591	if (copy_to_user(optval, &val, len))
5592		return -EFAULT;
5593	return 0;
5594}
5595
5596/*
5597 * 8.2.6. Get the Current Identifiers of Associations
5598 *        (SCTP_GET_ASSOC_ID_LIST)
5599 *
5600 * This option gets the current list of SCTP association identifiers of
5601 * the SCTP associations handled by a one-to-many style socket.
5602 */
5603static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5604				    char __user *optval, int __user *optlen)
5605{
5606	struct sctp_sock *sp = sctp_sk(sk);
5607	struct sctp_association *asoc;
5608	struct sctp_assoc_ids *ids;
5609	u32 num = 0;
5610
5611	if (sctp_style(sk, TCP))
5612		return -EOPNOTSUPP;
5613
5614	if (len < sizeof(struct sctp_assoc_ids))
5615		return -EINVAL;
5616
5617	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5618		num++;
5619	}
5620
5621	if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5622		return -EINVAL;
5623
5624	len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5625
5626	ids = kmalloc(len, GFP_KERNEL);
5627	if (unlikely(!ids))
5628		return -ENOMEM;
5629
5630	ids->gaids_number_of_ids = num;
5631	num = 0;
5632	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5633		ids->gaids_assoc_id[num++] = asoc->assoc_id;
5634	}
5635
5636	if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5637		kfree(ids);
5638		return -EFAULT;
5639	}
5640
5641	kfree(ids);
5642	return 0;
5643}
5644
5645/*
5646 * SCTP_PEER_ADDR_THLDS
5647 *
5648 * This option allows us to fetch the partially failed threshold for one or all
5649 * transports in an association.  See Section 6.1 of:
5650 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5651 */
5652static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5653					    char __user *optval,
5654					    int len,
5655					    int __user *optlen)
5656{
5657	struct sctp_paddrthlds val;
5658	struct sctp_transport *trans;
5659	struct sctp_association *asoc;
5660
5661	if (len < sizeof(struct sctp_paddrthlds))
5662		return -EINVAL;
5663	len = sizeof(struct sctp_paddrthlds);
5664	if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5665		return -EFAULT;
5666
5667	if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5668		asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5669		if (!asoc)
5670			return -ENOENT;
5671
5672		val.spt_pathpfthld = asoc->pf_retrans;
5673		val.spt_pathmaxrxt = asoc->pathmaxrxt;
5674	} else {
5675		trans = sctp_addr_id2transport(sk, &val.spt_address,
5676					       val.spt_assoc_id);
5677		if (!trans)
5678			return -ENOENT;
5679
5680		val.spt_pathmaxrxt = trans->pathmaxrxt;
5681		val.spt_pathpfthld = trans->pf_retrans;
5682	}
5683
5684	if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5685		return -EFAULT;
5686
5687	return 0;
5688}
5689
5690/*
5691 * SCTP_GET_ASSOC_STATS
5692 *
5693 * This option retrieves local per endpoint statistics. It is modeled
5694 * after OpenSolaris' implementation
5695 */
5696static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5697				       char __user *optval,
5698				       int __user *optlen)
5699{
5700	struct sctp_assoc_stats sas;
5701	struct sctp_association *asoc = NULL;
5702
5703	/* User must provide at least the assoc id */
5704	if (len < sizeof(sctp_assoc_t))
5705		return -EINVAL;
5706
5707	/* Allow the struct to grow and fill in as much as possible */
5708	len = min_t(size_t, len, sizeof(sas));
5709
5710	if (copy_from_user(&sas, optval, len))
5711		return -EFAULT;
5712
5713	asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5714	if (!asoc)
5715		return -EINVAL;
5716
5717	sas.sas_rtxchunks = asoc->stats.rtxchunks;
5718	sas.sas_gapcnt = asoc->stats.gapcnt;
5719	sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5720	sas.sas_osacks = asoc->stats.osacks;
5721	sas.sas_isacks = asoc->stats.isacks;
5722	sas.sas_octrlchunks = asoc->stats.octrlchunks;
5723	sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5724	sas.sas_oodchunks = asoc->stats.oodchunks;
5725	sas.sas_iodchunks = asoc->stats.iodchunks;
5726	sas.sas_ouodchunks = asoc->stats.ouodchunks;
5727	sas.sas_iuodchunks = asoc->stats.iuodchunks;
5728	sas.sas_idupchunks = asoc->stats.idupchunks;
5729	sas.sas_opackets = asoc->stats.opackets;
5730	sas.sas_ipackets = asoc->stats.ipackets;
5731
5732	/* New high max rto observed, will return 0 if not a single
5733	 * RTO update took place. obs_rto_ipaddr will be bogus
5734	 * in such a case
5735	 */
5736	sas.sas_maxrto = asoc->stats.max_obs_rto;
5737	memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5738		sizeof(struct sockaddr_storage));
5739
5740	/* Mark beginning of a new observation period */
5741	asoc->stats.max_obs_rto = asoc->rto_min;
5742
5743	if (put_user(len, optlen))
5744		return -EFAULT;
5745
5746	pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5747
5748	if (copy_to_user(optval, &sas, len))
5749		return -EFAULT;
5750
5751	return 0;
5752}
5753
5754static int sctp_getsockopt(struct sock *sk, int level, int optname,
5755			   char __user *optval, int __user *optlen)
5756{
5757	int retval = 0;
5758	int len;
5759
5760	pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
 
5761
5762	/* I can hardly begin to describe how wrong this is.  This is
5763	 * so broken as to be worse than useless.  The API draft
5764	 * REALLY is NOT helpful here...  I am not convinced that the
5765	 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5766	 * are at all well-founded.
5767	 */
5768	if (level != SOL_SCTP) {
5769		struct sctp_af *af = sctp_sk(sk)->pf->af;
5770
5771		retval = af->getsockopt(sk, level, optname, optval, optlen);
5772		return retval;
5773	}
5774
5775	if (get_user(len, optlen))
5776		return -EFAULT;
5777
5778	lock_sock(sk);
5779
5780	switch (optname) {
5781	case SCTP_STATUS:
5782		retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5783		break;
5784	case SCTP_DISABLE_FRAGMENTS:
5785		retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5786							   optlen);
5787		break;
5788	case SCTP_EVENTS:
5789		retval = sctp_getsockopt_events(sk, len, optval, optlen);
5790		break;
5791	case SCTP_AUTOCLOSE:
5792		retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5793		break;
5794	case SCTP_SOCKOPT_PEELOFF:
5795		retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5796		break;
5797	case SCTP_PEER_ADDR_PARAMS:
5798		retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5799							  optlen);
5800		break;
5801	case SCTP_DELAYED_SACK:
5802		retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5803							  optlen);
5804		break;
5805	case SCTP_INITMSG:
5806		retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5807		break;
5808	case SCTP_GET_PEER_ADDRS:
5809		retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5810						    optlen);
5811		break;
5812	case SCTP_GET_LOCAL_ADDRS:
5813		retval = sctp_getsockopt_local_addrs(sk, len, optval,
5814						     optlen);
5815		break;
5816	case SCTP_SOCKOPT_CONNECTX3:
5817		retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5818		break;
5819	case SCTP_DEFAULT_SEND_PARAM:
5820		retval = sctp_getsockopt_default_send_param(sk, len,
5821							    optval, optlen);
5822		break;
5823	case SCTP_PRIMARY_ADDR:
5824		retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5825		break;
5826	case SCTP_NODELAY:
5827		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5828		break;
5829	case SCTP_RTOINFO:
5830		retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5831		break;
5832	case SCTP_ASSOCINFO:
5833		retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5834		break;
5835	case SCTP_I_WANT_MAPPED_V4_ADDR:
5836		retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5837		break;
5838	case SCTP_MAXSEG:
5839		retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5840		break;
5841	case SCTP_GET_PEER_ADDR_INFO:
5842		retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5843							optlen);
5844		break;
5845	case SCTP_ADAPTATION_LAYER:
5846		retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5847							optlen);
5848		break;
5849	case SCTP_CONTEXT:
5850		retval = sctp_getsockopt_context(sk, len, optval, optlen);
5851		break;
5852	case SCTP_FRAGMENT_INTERLEAVE:
5853		retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5854							     optlen);
5855		break;
5856	case SCTP_PARTIAL_DELIVERY_POINT:
5857		retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5858								optlen);
5859		break;
5860	case SCTP_MAX_BURST:
5861		retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5862		break;
5863	case SCTP_AUTH_KEY:
5864	case SCTP_AUTH_CHUNK:
5865	case SCTP_AUTH_DELETE_KEY:
5866		retval = -EOPNOTSUPP;
5867		break;
5868	case SCTP_HMAC_IDENT:
5869		retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5870		break;
5871	case SCTP_AUTH_ACTIVE_KEY:
5872		retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5873		break;
5874	case SCTP_PEER_AUTH_CHUNKS:
5875		retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5876							optlen);
5877		break;
5878	case SCTP_LOCAL_AUTH_CHUNKS:
5879		retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5880							optlen);
5881		break;
5882	case SCTP_GET_ASSOC_NUMBER:
5883		retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5884		break;
5885	case SCTP_GET_ASSOC_ID_LIST:
5886		retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5887		break;
5888	case SCTP_AUTO_ASCONF:
5889		retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5890		break;
5891	case SCTP_PEER_ADDR_THLDS:
5892		retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5893		break;
5894	case SCTP_GET_ASSOC_STATS:
5895		retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5896		break;
5897	default:
5898		retval = -ENOPROTOOPT;
5899		break;
5900	}
5901
5902	release_sock(sk);
5903	return retval;
5904}
5905
5906static void sctp_hash(struct sock *sk)
5907{
5908	/* STUB */
5909}
5910
5911static void sctp_unhash(struct sock *sk)
5912{
5913	/* STUB */
5914}
5915
5916/* Check if port is acceptable.  Possibly find first available port.
5917 *
5918 * The port hash table (contained in the 'global' SCTP protocol storage
5919 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5920 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5921 * list (the list number is the port number hashed out, so as you
5922 * would expect from a hash function, all the ports in a given list have
5923 * such a number that hashes out to the same list number; you were
5924 * expecting that, right?); so each list has a set of ports, with a
5925 * link to the socket (struct sock) that uses it, the port number and
5926 * a fastreuse flag (FIXME: NPI ipg).
5927 */
5928static struct sctp_bind_bucket *sctp_bucket_create(
5929	struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5930
5931static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5932{
5933	struct sctp_bind_hashbucket *head; /* hash list */
5934	struct sctp_bind_bucket *pp;
 
5935	unsigned short snum;
5936	int ret;
5937
5938	snum = ntohs(addr->v4.sin_port);
5939
5940	pr_debug("%s: begins, snum:%d\n", __func__, snum);
5941
5942	local_bh_disable();
5943
5944	if (snum == 0) {
5945		/* Search for an available port. */
5946		int low, high, remaining, index;
5947		unsigned int rover;
5948
5949		inet_get_local_port_range(sock_net(sk), &low, &high);
5950		remaining = (high - low) + 1;
5951		rover = prandom_u32() % remaining + low;
5952
5953		do {
5954			rover++;
5955			if ((rover < low) || (rover > high))
5956				rover = low;
5957			if (inet_is_reserved_local_port(rover))
5958				continue;
5959			index = sctp_phashfn(sock_net(sk), rover);
5960			head = &sctp_port_hashtable[index];
5961			spin_lock(&head->lock);
5962			sctp_for_each_hentry(pp, &head->chain)
5963				if ((pp->port == rover) &&
5964				    net_eq(sock_net(sk), pp->net))
5965					goto next;
5966			break;
5967		next:
5968			spin_unlock(&head->lock);
5969		} while (--remaining > 0);
5970
5971		/* Exhausted local port range during search? */
5972		ret = 1;
5973		if (remaining <= 0)
5974			goto fail;
5975
5976		/* OK, here is the one we will use.  HEAD (the port
5977		 * hash table list entry) is non-NULL and we hold it's
5978		 * mutex.
5979		 */
5980		snum = rover;
5981	} else {
5982		/* We are given an specific port number; we verify
5983		 * that it is not being used. If it is used, we will
5984		 * exahust the search in the hash list corresponding
5985		 * to the port number (snum) - we detect that with the
5986		 * port iterator, pp being NULL.
5987		 */
5988		head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5989		spin_lock(&head->lock);
5990		sctp_for_each_hentry(pp, &head->chain) {
5991			if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5992				goto pp_found;
5993		}
5994	}
5995	pp = NULL;
5996	goto pp_not_found;
5997pp_found:
5998	if (!hlist_empty(&pp->owner)) {
5999		/* We had a port hash table hit - there is an
6000		 * available port (pp != NULL) and it is being
6001		 * used by other socket (pp->owner not empty); that other
6002		 * socket is going to be sk2.
6003		 */
6004		int reuse = sk->sk_reuse;
6005		struct sock *sk2;
6006
6007		pr_debug("%s: found a possible match\n", __func__);
6008
6009		if (pp->fastreuse && sk->sk_reuse &&
6010			sk->sk_state != SCTP_SS_LISTENING)
6011			goto success;
6012
6013		/* Run through the list of sockets bound to the port
6014		 * (pp->port) [via the pointers bind_next and
6015		 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6016		 * we get the endpoint they describe and run through
6017		 * the endpoint's list of IP (v4 or v6) addresses,
6018		 * comparing each of the addresses with the address of
6019		 * the socket sk. If we find a match, then that means
6020		 * that this port/socket (sk) combination are already
6021		 * in an endpoint.
6022		 */
6023		sk_for_each_bound(sk2, &pp->owner) {
6024			struct sctp_endpoint *ep2;
6025			ep2 = sctp_sk(sk2)->ep;
6026
6027			if (sk == sk2 ||
6028			    (reuse && sk2->sk_reuse &&
6029			     sk2->sk_state != SCTP_SS_LISTENING))
6030				continue;
6031
6032			if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6033						 sctp_sk(sk2), sctp_sk(sk))) {
6034				ret = (long)sk2;
6035				goto fail_unlock;
6036			}
6037		}
6038
6039		pr_debug("%s: found a match\n", __func__);
6040	}
6041pp_not_found:
6042	/* If there was a hash table miss, create a new port.  */
6043	ret = 1;
6044	if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6045		goto fail_unlock;
6046
6047	/* In either case (hit or miss), make sure fastreuse is 1 only
6048	 * if sk->sk_reuse is too (that is, if the caller requested
6049	 * SO_REUSEADDR on this socket -sk-).
6050	 */
6051	if (hlist_empty(&pp->owner)) {
6052		if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6053			pp->fastreuse = 1;
6054		else
6055			pp->fastreuse = 0;
6056	} else if (pp->fastreuse &&
6057		(!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6058		pp->fastreuse = 0;
6059
6060	/* We are set, so fill up all the data in the hash table
6061	 * entry, tie the socket list information with the rest of the
6062	 * sockets FIXME: Blurry, NPI (ipg).
6063	 */
6064success:
6065	if (!sctp_sk(sk)->bind_hash) {
6066		inet_sk(sk)->inet_num = snum;
6067		sk_add_bind_node(sk, &pp->owner);
6068		sctp_sk(sk)->bind_hash = pp;
6069	}
6070	ret = 0;
6071
6072fail_unlock:
6073	spin_unlock(&head->lock);
6074
6075fail:
6076	local_bh_enable();
6077	return ret;
6078}
6079
6080/* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
6081 * port is requested.
6082 */
6083static int sctp_get_port(struct sock *sk, unsigned short snum)
6084{
 
6085	union sctp_addr addr;
6086	struct sctp_af *af = sctp_sk(sk)->pf->af;
6087
6088	/* Set up a dummy address struct from the sk. */
6089	af->from_sk(&addr, sk);
6090	addr.v4.sin_port = htons(snum);
6091
6092	/* Note: sk->sk_num gets filled in if ephemeral port request. */
6093	return !!sctp_get_port_local(sk, &addr);
 
 
6094}
6095
6096/*
6097 *  Move a socket to LISTENING state.
6098 */
6099static int sctp_listen_start(struct sock *sk, int backlog)
6100{
6101	struct sctp_sock *sp = sctp_sk(sk);
6102	struct sctp_endpoint *ep = sp->ep;
6103	struct crypto_hash *tfm = NULL;
6104	char alg[32];
6105
6106	/* Allocate HMAC for generating cookie. */
6107	if (!sp->hmac && sp->sctp_hmac_alg) {
6108		sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6109		tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6110		if (IS_ERR(tfm)) {
6111			net_info_ratelimited("failed to load transform for %s: %ld\n",
6112					     sp->sctp_hmac_alg, PTR_ERR(tfm));
6113			return -ENOSYS;
6114		}
6115		sctp_sk(sk)->hmac = tfm;
6116	}
6117
6118	/*
6119	 * If a bind() or sctp_bindx() is not called prior to a listen()
6120	 * call that allows new associations to be accepted, the system
6121	 * picks an ephemeral port and will choose an address set equivalent
6122	 * to binding with a wildcard address.
6123	 *
6124	 * This is not currently spelled out in the SCTP sockets
6125	 * extensions draft, but follows the practice as seen in TCP
6126	 * sockets.
6127	 *
6128	 */
6129	sk->sk_state = SCTP_SS_LISTENING;
6130	if (!ep->base.bind_addr.port) {
6131		if (sctp_autobind(sk))
6132			return -EAGAIN;
6133	} else {
6134		if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6135			sk->sk_state = SCTP_SS_CLOSED;
6136			return -EADDRINUSE;
6137		}
6138	}
6139
6140	sk->sk_max_ack_backlog = backlog;
6141	sctp_hash_endpoint(ep);
6142	return 0;
6143}
6144
6145/*
6146 * 4.1.3 / 5.1.3 listen()
6147 *
6148 *   By default, new associations are not accepted for UDP style sockets.
6149 *   An application uses listen() to mark a socket as being able to
6150 *   accept new associations.
6151 *
6152 *   On TCP style sockets, applications use listen() to ready the SCTP
6153 *   endpoint for accepting inbound associations.
6154 *
6155 *   On both types of endpoints a backlog of '0' disables listening.
6156 *
6157 *  Move a socket to LISTENING state.
6158 */
6159int sctp_inet_listen(struct socket *sock, int backlog)
6160{
6161	struct sock *sk = sock->sk;
6162	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6163	int err = -EINVAL;
6164
6165	if (unlikely(backlog < 0))
6166		return err;
6167
6168	lock_sock(sk);
6169
6170	/* Peeled-off sockets are not allowed to listen().  */
6171	if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6172		goto out;
6173
6174	if (sock->state != SS_UNCONNECTED)
6175		goto out;
6176
6177	/* If backlog is zero, disable listening. */
6178	if (!backlog) {
6179		if (sctp_sstate(sk, CLOSED))
6180			goto out;
6181
6182		err = 0;
6183		sctp_unhash_endpoint(ep);
6184		sk->sk_state = SCTP_SS_CLOSED;
6185		if (sk->sk_reuse)
6186			sctp_sk(sk)->bind_hash->fastreuse = 1;
6187		goto out;
6188	}
6189
6190	/* If we are already listening, just update the backlog */
6191	if (sctp_sstate(sk, LISTENING))
6192		sk->sk_max_ack_backlog = backlog;
6193	else {
6194		err = sctp_listen_start(sk, backlog);
6195		if (err)
6196			goto out;
6197	}
6198
6199	err = 0;
6200out:
6201	release_sock(sk);
6202	return err;
6203}
6204
6205/*
6206 * This function is done by modeling the current datagram_poll() and the
6207 * tcp_poll().  Note that, based on these implementations, we don't
6208 * lock the socket in this function, even though it seems that,
6209 * ideally, locking or some other mechanisms can be used to ensure
6210 * the integrity of the counters (sndbuf and wmem_alloc) used
6211 * in this place.  We assume that we don't need locks either until proven
6212 * otherwise.
6213 *
6214 * Another thing to note is that we include the Async I/O support
6215 * here, again, by modeling the current TCP/UDP code.  We don't have
6216 * a good way to test with it yet.
6217 */
6218unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6219{
6220	struct sock *sk = sock->sk;
6221	struct sctp_sock *sp = sctp_sk(sk);
6222	unsigned int mask;
6223
6224	poll_wait(file, sk_sleep(sk), wait);
6225
6226	/* A TCP-style listening socket becomes readable when the accept queue
6227	 * is not empty.
6228	 */
6229	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6230		return (!list_empty(&sp->ep->asocs)) ?
6231			(POLLIN | POLLRDNORM) : 0;
6232
6233	mask = 0;
6234
6235	/* Is there any exceptional events?  */
6236	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6237		mask |= POLLERR |
6238			(sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6239	if (sk->sk_shutdown & RCV_SHUTDOWN)
6240		mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6241	if (sk->sk_shutdown == SHUTDOWN_MASK)
6242		mask |= POLLHUP;
6243
6244	/* Is it readable?  Reconsider this code with TCP-style support.  */
6245	if (!skb_queue_empty(&sk->sk_receive_queue))
6246		mask |= POLLIN | POLLRDNORM;
6247
6248	/* The association is either gone or not ready.  */
6249	if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6250		return mask;
6251
6252	/* Is it writable?  */
6253	if (sctp_writeable(sk)) {
6254		mask |= POLLOUT | POLLWRNORM;
6255	} else {
6256		set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6257		/*
6258		 * Since the socket is not locked, the buffer
6259		 * might be made available after the writeable check and
6260		 * before the bit is set.  This could cause a lost I/O
6261		 * signal.  tcp_poll() has a race breaker for this race
6262		 * condition.  Based on their implementation, we put
6263		 * in the following code to cover it as well.
6264		 */
6265		if (sctp_writeable(sk))
6266			mask |= POLLOUT | POLLWRNORM;
6267	}
6268	return mask;
6269}
6270
6271/********************************************************************
6272 * 2nd Level Abstractions
6273 ********************************************************************/
6274
6275static struct sctp_bind_bucket *sctp_bucket_create(
6276	struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6277{
6278	struct sctp_bind_bucket *pp;
6279
6280	pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6281	if (pp) {
6282		SCTP_DBG_OBJCNT_INC(bind_bucket);
6283		pp->port = snum;
6284		pp->fastreuse = 0;
6285		INIT_HLIST_HEAD(&pp->owner);
6286		pp->net = net;
6287		hlist_add_head(&pp->node, &head->chain);
6288	}
6289	return pp;
6290}
6291
6292/* Caller must hold hashbucket lock for this tb with local BH disabled */
6293static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6294{
6295	if (pp && hlist_empty(&pp->owner)) {
6296		__hlist_del(&pp->node);
6297		kmem_cache_free(sctp_bucket_cachep, pp);
6298		SCTP_DBG_OBJCNT_DEC(bind_bucket);
6299	}
6300}
6301
6302/* Release this socket's reference to a local port.  */
6303static inline void __sctp_put_port(struct sock *sk)
6304{
6305	struct sctp_bind_hashbucket *head =
6306		&sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6307						  inet_sk(sk)->inet_num)];
6308	struct sctp_bind_bucket *pp;
6309
6310	spin_lock(&head->lock);
6311	pp = sctp_sk(sk)->bind_hash;
6312	__sk_del_bind_node(sk);
6313	sctp_sk(sk)->bind_hash = NULL;
6314	inet_sk(sk)->inet_num = 0;
6315	sctp_bucket_destroy(pp);
6316	spin_unlock(&head->lock);
6317}
6318
6319void sctp_put_port(struct sock *sk)
6320{
6321	local_bh_disable();
6322	__sctp_put_port(sk);
6323	local_bh_enable();
6324}
6325
6326/*
6327 * The system picks an ephemeral port and choose an address set equivalent
6328 * to binding with a wildcard address.
6329 * One of those addresses will be the primary address for the association.
6330 * This automatically enables the multihoming capability of SCTP.
6331 */
6332static int sctp_autobind(struct sock *sk)
6333{
6334	union sctp_addr autoaddr;
6335	struct sctp_af *af;
6336	__be16 port;
6337
6338	/* Initialize a local sockaddr structure to INADDR_ANY. */
6339	af = sctp_sk(sk)->pf->af;
6340
6341	port = htons(inet_sk(sk)->inet_num);
6342	af->inaddr_any(&autoaddr, port);
6343
6344	return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6345}
6346
6347/* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
6348 *
6349 * From RFC 2292
6350 * 4.2 The cmsghdr Structure *
6351 *
6352 * When ancillary data is sent or received, any number of ancillary data
6353 * objects can be specified by the msg_control and msg_controllen members of
6354 * the msghdr structure, because each object is preceded by
6355 * a cmsghdr structure defining the object's length (the cmsg_len member).
6356 * Historically Berkeley-derived implementations have passed only one object
6357 * at a time, but this API allows multiple objects to be
6358 * passed in a single call to sendmsg() or recvmsg(). The following example
6359 * shows two ancillary data objects in a control buffer.
6360 *
6361 *   |<--------------------------- msg_controllen -------------------------->|
6362 *   |                                                                       |
6363 *
6364 *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
6365 *
6366 *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6367 *   |                                   |                                   |
6368 *
6369 *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
6370 *
6371 *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
6372 *   |                                |  |                                |  |
6373 *
6374 *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6375 *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
6376 *
6377 *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
6378 *
6379 *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6380 *    ^
6381 *    |
6382 *
6383 * msg_control
6384 * points here
6385 */
6386static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
 
6387{
6388	struct cmsghdr *cmsg;
6389	struct msghdr *my_msg = (struct msghdr *)msg;
6390
6391	for (cmsg = CMSG_FIRSTHDR(msg);
6392	     cmsg != NULL;
6393	     cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6394		if (!CMSG_OK(my_msg, cmsg))
6395			return -EINVAL;
6396
6397		/* Should we parse this header or ignore?  */
6398		if (cmsg->cmsg_level != IPPROTO_SCTP)
6399			continue;
6400
6401		/* Strictly check lengths following example in SCM code.  */
6402		switch (cmsg->cmsg_type) {
6403		case SCTP_INIT:
6404			/* SCTP Socket API Extension
6405			 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6406			 *
6407			 * This cmsghdr structure provides information for
6408			 * initializing new SCTP associations with sendmsg().
6409			 * The SCTP_INITMSG socket option uses this same data
6410			 * structure.  This structure is not used for
6411			 * recvmsg().
6412			 *
6413			 * cmsg_level    cmsg_type      cmsg_data[]
6414			 * ------------  ------------   ----------------------
6415			 * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
6416			 */
6417			if (cmsg->cmsg_len !=
6418			    CMSG_LEN(sizeof(struct sctp_initmsg)))
6419				return -EINVAL;
6420			cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6421			break;
6422
6423		case SCTP_SNDRCV:
6424			/* SCTP Socket API Extension
6425			 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6426			 *
6427			 * This cmsghdr structure specifies SCTP options for
6428			 * sendmsg() and describes SCTP header information
6429			 * about a received message through recvmsg().
6430			 *
6431			 * cmsg_level    cmsg_type      cmsg_data[]
6432			 * ------------  ------------   ----------------------
6433			 * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
6434			 */
6435			if (cmsg->cmsg_len !=
6436			    CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6437				return -EINVAL;
6438
6439			cmsgs->info =
6440				(struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6441
6442			/* Minimally, validate the sinfo_flags. */
6443			if (cmsgs->info->sinfo_flags &
6444			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6445			      SCTP_ABORT | SCTP_EOF))
6446				return -EINVAL;
6447			break;
6448
6449		default:
6450			return -EINVAL;
6451		}
6452	}
6453	return 0;
6454}
6455
6456/*
6457 * Wait for a packet..
6458 * Note: This function is the same function as in core/datagram.c
6459 * with a few modifications to make lksctp work.
6460 */
6461static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6462{
6463	int error;
6464	DEFINE_WAIT(wait);
6465
6466	prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6467
6468	/* Socket errors? */
6469	error = sock_error(sk);
6470	if (error)
6471		goto out;
6472
6473	if (!skb_queue_empty(&sk->sk_receive_queue))
6474		goto ready;
6475
6476	/* Socket shut down?  */
6477	if (sk->sk_shutdown & RCV_SHUTDOWN)
6478		goto out;
6479
6480	/* Sequenced packets can come disconnected.  If so we report the
6481	 * problem.
6482	 */
6483	error = -ENOTCONN;
6484
6485	/* Is there a good reason to think that we may receive some data?  */
6486	if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6487		goto out;
6488
6489	/* Handle signals.  */
6490	if (signal_pending(current))
6491		goto interrupted;
6492
6493	/* Let another process have a go.  Since we are going to sleep
6494	 * anyway.  Note: This may cause odd behaviors if the message
6495	 * does not fit in the user's buffer, but this seems to be the
6496	 * only way to honor MSG_DONTWAIT realistically.
6497	 */
6498	release_sock(sk);
6499	*timeo_p = schedule_timeout(*timeo_p);
6500	lock_sock(sk);
6501
6502ready:
6503	finish_wait(sk_sleep(sk), &wait);
6504	return 0;
6505
6506interrupted:
6507	error = sock_intr_errno(*timeo_p);
6508
6509out:
6510	finish_wait(sk_sleep(sk), &wait);
6511	*err = error;
6512	return error;
6513}
6514
6515/* Receive a datagram.
6516 * Note: This is pretty much the same routine as in core/datagram.c
6517 * with a few changes to make lksctp work.
6518 */
6519static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6520					      int noblock, int *err)
6521{
6522	int error;
6523	struct sk_buff *skb;
6524	long timeo;
6525
6526	timeo = sock_rcvtimeo(sk, noblock);
6527
6528	pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6529		 MAX_SCHEDULE_TIMEOUT);
6530
6531	do {
6532		/* Again only user level code calls this function,
6533		 * so nothing interrupt level
6534		 * will suddenly eat the receive_queue.
6535		 *
6536		 *  Look at current nfs client by the way...
6537		 *  However, this function was correct in any case. 8)
6538		 */
6539		if (flags & MSG_PEEK) {
6540			spin_lock_bh(&sk->sk_receive_queue.lock);
6541			skb = skb_peek(&sk->sk_receive_queue);
6542			if (skb)
6543				atomic_inc(&skb->users);
6544			spin_unlock_bh(&sk->sk_receive_queue.lock);
6545		} else {
6546			skb = skb_dequeue(&sk->sk_receive_queue);
6547		}
6548
6549		if (skb)
6550			return skb;
6551
6552		/* Caller is allowed not to check sk->sk_err before calling. */
6553		error = sock_error(sk);
6554		if (error)
6555			goto no_packet;
6556
6557		if (sk->sk_shutdown & RCV_SHUTDOWN)
6558			break;
6559
6560		/* User doesn't want to wait.  */
6561		error = -EAGAIN;
6562		if (!timeo)
6563			goto no_packet;
6564	} while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6565
6566	return NULL;
6567
6568no_packet:
6569	*err = error;
6570	return NULL;
6571}
6572
6573/* If sndbuf has changed, wake up per association sndbuf waiters.  */
6574static void __sctp_write_space(struct sctp_association *asoc)
6575{
6576	struct sock *sk = asoc->base.sk;
6577	struct socket *sock = sk->sk_socket;
6578
6579	if ((sctp_wspace(asoc) > 0) && sock) {
6580		if (waitqueue_active(&asoc->wait))
6581			wake_up_interruptible(&asoc->wait);
6582
6583		if (sctp_writeable(sk)) {
6584			wait_queue_head_t *wq = sk_sleep(sk);
6585
6586			if (wq && waitqueue_active(wq))
6587				wake_up_interruptible(wq);
6588
6589			/* Note that we try to include the Async I/O support
6590			 * here by modeling from the current TCP/UDP code.
6591			 * We have not tested with it yet.
6592			 */
6593			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6594				sock_wake_async(sock,
6595						SOCK_WAKE_SPACE, POLL_OUT);
6596		}
6597	}
6598}
6599
6600static void sctp_wake_up_waiters(struct sock *sk,
6601				 struct sctp_association *asoc)
6602{
6603	struct sctp_association *tmp = asoc;
6604
6605	/* We do accounting for the sndbuf space per association,
6606	 * so we only need to wake our own association.
6607	 */
6608	if (asoc->ep->sndbuf_policy)
6609		return __sctp_write_space(asoc);
6610
6611	/* If association goes down and is just flushing its
6612	 * outq, then just normally notify others.
6613	 */
6614	if (asoc->base.dead)
6615		return sctp_write_space(sk);
6616
6617	/* Accounting for the sndbuf space is per socket, so we
6618	 * need to wake up others, try to be fair and in case of
6619	 * other associations, let them have a go first instead
6620	 * of just doing a sctp_write_space() call.
6621	 *
6622	 * Note that we reach sctp_wake_up_waiters() only when
6623	 * associations free up queued chunks, thus we are under
6624	 * lock and the list of associations on a socket is
6625	 * guaranteed not to change.
6626	 */
6627	for (tmp = list_next_entry(tmp, asocs); 1;
6628	     tmp = list_next_entry(tmp, asocs)) {
6629		/* Manually skip the head element. */
6630		if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6631			continue;
6632		/* Wake up association. */
6633		__sctp_write_space(tmp);
6634		/* We've reached the end. */
6635		if (tmp == asoc)
6636			break;
6637	}
6638}
6639
6640/* Do accounting for the sndbuf space.
6641 * Decrement the used sndbuf space of the corresponding association by the
6642 * data size which was just transmitted(freed).
6643 */
6644static void sctp_wfree(struct sk_buff *skb)
6645{
6646	struct sctp_association *asoc;
6647	struct sctp_chunk *chunk;
6648	struct sock *sk;
6649
6650	/* Get the saved chunk pointer.  */
6651	chunk = *((struct sctp_chunk **)(skb->cb));
6652	asoc = chunk->asoc;
6653	sk = asoc->base.sk;
6654	asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6655				sizeof(struct sk_buff) +
6656				sizeof(struct sctp_chunk);
6657
6658	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6659
6660	/*
6661	 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6662	 */
6663	sk->sk_wmem_queued   -= skb->truesize;
6664	sk_mem_uncharge(sk, skb->truesize);
6665
6666	sock_wfree(skb);
6667	sctp_wake_up_waiters(sk, asoc);
6668
6669	sctp_association_put(asoc);
6670}
6671
6672/* Do accounting for the receive space on the socket.
6673 * Accounting for the association is done in ulpevent.c
6674 * We set this as a destructor for the cloned data skbs so that
6675 * accounting is done at the correct time.
6676 */
6677void sctp_sock_rfree(struct sk_buff *skb)
6678{
6679	struct sock *sk = skb->sk;
6680	struct sctp_ulpevent *event = sctp_skb2event(skb);
6681
6682	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6683
6684	/*
6685	 * Mimic the behavior of sock_rfree
6686	 */
6687	sk_mem_uncharge(sk, event->rmem_len);
6688}
6689
6690
6691/* Helper function to wait for space in the sndbuf.  */
6692static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6693				size_t msg_len)
6694{
6695	struct sock *sk = asoc->base.sk;
6696	int err = 0;
6697	long current_timeo = *timeo_p;
6698	DEFINE_WAIT(wait);
6699
6700	pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6701		 *timeo_p, msg_len);
6702
6703	/* Increment the association's refcnt.  */
6704	sctp_association_hold(asoc);
6705
6706	/* Wait on the association specific sndbuf space. */
6707	for (;;) {
6708		prepare_to_wait_exclusive(&asoc->wait, &wait,
6709					  TASK_INTERRUPTIBLE);
6710		if (!*timeo_p)
6711			goto do_nonblock;
6712		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6713		    asoc->base.dead)
6714			goto do_error;
6715		if (signal_pending(current))
6716			goto do_interrupted;
6717		if (msg_len <= sctp_wspace(asoc))
6718			break;
6719
6720		/* Let another process have a go.  Since we are going
6721		 * to sleep anyway.
6722		 */
6723		release_sock(sk);
6724		current_timeo = schedule_timeout(current_timeo);
6725		BUG_ON(sk != asoc->base.sk);
6726		lock_sock(sk);
6727
6728		*timeo_p = current_timeo;
6729	}
6730
6731out:
6732	finish_wait(&asoc->wait, &wait);
6733
6734	/* Release the association's refcnt.  */
6735	sctp_association_put(asoc);
6736
6737	return err;
6738
6739do_error:
6740	err = -EPIPE;
6741	goto out;
6742
6743do_interrupted:
6744	err = sock_intr_errno(*timeo_p);
6745	goto out;
6746
6747do_nonblock:
6748	err = -EAGAIN;
6749	goto out;
6750}
6751
6752void sctp_data_ready(struct sock *sk)
6753{
6754	struct socket_wq *wq;
6755
6756	rcu_read_lock();
6757	wq = rcu_dereference(sk->sk_wq);
6758	if (wq_has_sleeper(wq))
6759		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6760						POLLRDNORM | POLLRDBAND);
6761	sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6762	rcu_read_unlock();
6763}
6764
6765/* If socket sndbuf has changed, wake up all per association waiters.  */
6766void sctp_write_space(struct sock *sk)
6767{
6768	struct sctp_association *asoc;
6769
6770	/* Wake up the tasks in each wait queue.  */
6771	list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6772		__sctp_write_space(asoc);
6773	}
6774}
6775
6776/* Is there any sndbuf space available on the socket?
6777 *
6778 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6779 * associations on the same socket.  For a UDP-style socket with
6780 * multiple associations, it is possible for it to be "unwriteable"
6781 * prematurely.  I assume that this is acceptable because
6782 * a premature "unwriteable" is better than an accidental "writeable" which
6783 * would cause an unwanted block under certain circumstances.  For the 1-1
6784 * UDP-style sockets or TCP-style sockets, this code should work.
6785 *  - Daisy
6786 */
6787static int sctp_writeable(struct sock *sk)
6788{
6789	int amt = 0;
6790
6791	amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6792	if (amt < 0)
6793		amt = 0;
6794	return amt;
6795}
6796
6797/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6798 * returns immediately with EINPROGRESS.
6799 */
6800static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6801{
6802	struct sock *sk = asoc->base.sk;
6803	int err = 0;
6804	long current_timeo = *timeo_p;
6805	DEFINE_WAIT(wait);
6806
6807	pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
 
6808
6809	/* Increment the association's refcnt.  */
6810	sctp_association_hold(asoc);
6811
6812	for (;;) {
6813		prepare_to_wait_exclusive(&asoc->wait, &wait,
6814					  TASK_INTERRUPTIBLE);
6815		if (!*timeo_p)
6816			goto do_nonblock;
6817		if (sk->sk_shutdown & RCV_SHUTDOWN)
6818			break;
6819		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6820		    asoc->base.dead)
6821			goto do_error;
6822		if (signal_pending(current))
6823			goto do_interrupted;
6824
6825		if (sctp_state(asoc, ESTABLISHED))
6826			break;
6827
6828		/* Let another process have a go.  Since we are going
6829		 * to sleep anyway.
6830		 */
6831		release_sock(sk);
6832		current_timeo = schedule_timeout(current_timeo);
6833		lock_sock(sk);
6834
6835		*timeo_p = current_timeo;
6836	}
6837
6838out:
6839	finish_wait(&asoc->wait, &wait);
6840
6841	/* Release the association's refcnt.  */
6842	sctp_association_put(asoc);
6843
6844	return err;
6845
6846do_error:
6847	if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6848		err = -ETIMEDOUT;
6849	else
6850		err = -ECONNREFUSED;
6851	goto out;
6852
6853do_interrupted:
6854	err = sock_intr_errno(*timeo_p);
6855	goto out;
6856
6857do_nonblock:
6858	err = -EINPROGRESS;
6859	goto out;
6860}
6861
6862static int sctp_wait_for_accept(struct sock *sk, long timeo)
6863{
6864	struct sctp_endpoint *ep;
6865	int err = 0;
6866	DEFINE_WAIT(wait);
6867
6868	ep = sctp_sk(sk)->ep;
6869
6870
6871	for (;;) {
6872		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6873					  TASK_INTERRUPTIBLE);
6874
6875		if (list_empty(&ep->asocs)) {
6876			release_sock(sk);
6877			timeo = schedule_timeout(timeo);
6878			lock_sock(sk);
6879		}
6880
6881		err = -EINVAL;
6882		if (!sctp_sstate(sk, LISTENING))
6883			break;
6884
6885		err = 0;
6886		if (!list_empty(&ep->asocs))
6887			break;
6888
6889		err = sock_intr_errno(timeo);
6890		if (signal_pending(current))
6891			break;
6892
6893		err = -EAGAIN;
6894		if (!timeo)
6895			break;
6896	}
6897
6898	finish_wait(sk_sleep(sk), &wait);
6899
6900	return err;
6901}
6902
6903static void sctp_wait_for_close(struct sock *sk, long timeout)
6904{
6905	DEFINE_WAIT(wait);
6906
6907	do {
6908		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6909		if (list_empty(&sctp_sk(sk)->ep->asocs))
6910			break;
6911		release_sock(sk);
6912		timeout = schedule_timeout(timeout);
6913		lock_sock(sk);
6914	} while (!signal_pending(current) && timeout);
6915
6916	finish_wait(sk_sleep(sk), &wait);
6917}
6918
6919static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6920{
6921	struct sk_buff *frag;
6922
6923	if (!skb->data_len)
6924		goto done;
6925
6926	/* Don't forget the fragments. */
6927	skb_walk_frags(skb, frag)
6928		sctp_skb_set_owner_r_frag(frag, sk);
6929
6930done:
6931	sctp_skb_set_owner_r(skb, sk);
6932}
6933
6934void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6935		    struct sctp_association *asoc)
6936{
6937	struct inet_sock *inet = inet_sk(sk);
6938	struct inet_sock *newinet;
6939
6940	newsk->sk_type = sk->sk_type;
6941	newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6942	newsk->sk_flags = sk->sk_flags;
6943	newsk->sk_no_check = sk->sk_no_check;
6944	newsk->sk_reuse = sk->sk_reuse;
6945
6946	newsk->sk_shutdown = sk->sk_shutdown;
6947	newsk->sk_destruct = sctp_destruct_sock;
6948	newsk->sk_family = sk->sk_family;
6949	newsk->sk_protocol = IPPROTO_SCTP;
6950	newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6951	newsk->sk_sndbuf = sk->sk_sndbuf;
6952	newsk->sk_rcvbuf = sk->sk_rcvbuf;
6953	newsk->sk_lingertime = sk->sk_lingertime;
6954	newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6955	newsk->sk_sndtimeo = sk->sk_sndtimeo;
6956
6957	newinet = inet_sk(newsk);
6958
6959	/* Initialize sk's sport, dport, rcv_saddr and daddr for
6960	 * getsockname() and getpeername()
6961	 */
6962	newinet->inet_sport = inet->inet_sport;
6963	newinet->inet_saddr = inet->inet_saddr;
6964	newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6965	newinet->inet_dport = htons(asoc->peer.port);
6966	newinet->pmtudisc = inet->pmtudisc;
6967	newinet->inet_id = asoc->next_tsn ^ jiffies;
6968
6969	newinet->uc_ttl = inet->uc_ttl;
6970	newinet->mc_loop = 1;
6971	newinet->mc_ttl = 1;
6972	newinet->mc_index = 0;
6973	newinet->mc_list = NULL;
6974}
6975
6976/* Populate the fields of the newsk from the oldsk and migrate the assoc
6977 * and its messages to the newsk.
6978 */
6979static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6980			      struct sctp_association *assoc,
6981			      sctp_socket_type_t type)
6982{
6983	struct sctp_sock *oldsp = sctp_sk(oldsk);
6984	struct sctp_sock *newsp = sctp_sk(newsk);
6985	struct sctp_bind_bucket *pp; /* hash list port iterator */
6986	struct sctp_endpoint *newep = newsp->ep;
6987	struct sk_buff *skb, *tmp;
6988	struct sctp_ulpevent *event;
6989	struct sctp_bind_hashbucket *head;
6990	struct list_head tmplist;
6991
6992	/* Migrate socket buffer sizes and all the socket level options to the
6993	 * new socket.
6994	 */
6995	newsk->sk_sndbuf = oldsk->sk_sndbuf;
6996	newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6997	/* Brute force copy old sctp opt. */
6998	if (oldsp->do_auto_asconf) {
6999		memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
7000		inet_sk_copy_descendant(newsk, oldsk);
7001		memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
7002	} else
7003		inet_sk_copy_descendant(newsk, oldsk);
7004
7005	/* Restore the ep value that was overwritten with the above structure
7006	 * copy.
7007	 */
7008	newsp->ep = newep;
7009	newsp->hmac = NULL;
7010
7011	/* Hook this new socket in to the bind_hash list. */
7012	head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7013						 inet_sk(oldsk)->inet_num)];
7014	local_bh_disable();
7015	spin_lock(&head->lock);
7016	pp = sctp_sk(oldsk)->bind_hash;
7017	sk_add_bind_node(newsk, &pp->owner);
7018	sctp_sk(newsk)->bind_hash = pp;
7019	inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7020	spin_unlock(&head->lock);
7021	local_bh_enable();
7022
7023	/* Copy the bind_addr list from the original endpoint to the new
7024	 * endpoint so that we can handle restarts properly
7025	 */
7026	sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7027				&oldsp->ep->base.bind_addr, GFP_KERNEL);
7028
7029	/* Move any messages in the old socket's receive queue that are for the
7030	 * peeled off association to the new socket's receive queue.
7031	 */
7032	sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7033		event = sctp_skb2event(skb);
7034		if (event->asoc == assoc) {
7035			__skb_unlink(skb, &oldsk->sk_receive_queue);
7036			__skb_queue_tail(&newsk->sk_receive_queue, skb);
7037			sctp_skb_set_owner_r_frag(skb, newsk);
7038		}
7039	}
7040
7041	/* Clean up any messages pending delivery due to partial
7042	 * delivery.   Three cases:
7043	 * 1) No partial deliver;  no work.
7044	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7045	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7046	 */
7047	skb_queue_head_init(&newsp->pd_lobby);
7048	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7049
7050	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7051		struct sk_buff_head *queue;
7052
7053		/* Decide which queue to move pd_lobby skbs to. */
7054		if (assoc->ulpq.pd_mode) {
7055			queue = &newsp->pd_lobby;
7056		} else
7057			queue = &newsk->sk_receive_queue;
7058
7059		/* Walk through the pd_lobby, looking for skbs that
7060		 * need moved to the new socket.
7061		 */
7062		sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7063			event = sctp_skb2event(skb);
7064			if (event->asoc == assoc) {
7065				__skb_unlink(skb, &oldsp->pd_lobby);
7066				__skb_queue_tail(queue, skb);
7067				sctp_skb_set_owner_r_frag(skb, newsk);
7068			}
7069		}
7070
7071		/* Clear up any skbs waiting for the partial
7072		 * delivery to finish.
7073		 */
7074		if (assoc->ulpq.pd_mode)
7075			sctp_clear_pd(oldsk, NULL);
7076
7077	}
7078
7079	sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7080		sctp_skb_set_owner_r_frag(skb, newsk);
7081
7082	sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7083		sctp_skb_set_owner_r_frag(skb, newsk);
7084
7085	/* Set the type of socket to indicate that it is peeled off from the
7086	 * original UDP-style socket or created with the accept() call on a
7087	 * TCP-style socket..
7088	 */
7089	newsp->type = type;
7090
7091	/* Mark the new socket "in-use" by the user so that any packets
7092	 * that may arrive on the association after we've moved it are
7093	 * queued to the backlog.  This prevents a potential race between
7094	 * backlog processing on the old socket and new-packet processing
7095	 * on the new socket.
7096	 *
7097	 * The caller has just allocated newsk so we can guarantee that other
7098	 * paths won't try to lock it and then oldsk.
7099	 */
7100	lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7101	sctp_assoc_migrate(assoc, newsk);
7102
7103	/* If the association on the newsk is already closed before accept()
7104	 * is called, set RCV_SHUTDOWN flag.
7105	 */
7106	if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7107		newsk->sk_shutdown |= RCV_SHUTDOWN;
7108
7109	newsk->sk_state = SCTP_SS_ESTABLISHED;
7110	release_sock(newsk);
7111}
7112
7113
7114/* This proto struct describes the ULP interface for SCTP.  */
7115struct proto sctp_prot = {
7116	.name        =	"SCTP",
7117	.owner       =	THIS_MODULE,
7118	.close       =	sctp_close,
7119	.connect     =	sctp_connect,
7120	.disconnect  =	sctp_disconnect,
7121	.accept      =	sctp_accept,
7122	.ioctl       =	sctp_ioctl,
7123	.init        =	sctp_init_sock,
7124	.destroy     =	sctp_destroy_sock,
7125	.shutdown    =	sctp_shutdown,
7126	.setsockopt  =	sctp_setsockopt,
7127	.getsockopt  =	sctp_getsockopt,
7128	.sendmsg     =	sctp_sendmsg,
7129	.recvmsg     =	sctp_recvmsg,
7130	.bind        =	sctp_bind,
7131	.backlog_rcv =	sctp_backlog_rcv,
7132	.hash        =	sctp_hash,
7133	.unhash      =	sctp_unhash,
7134	.get_port    =	sctp_get_port,
7135	.obj_size    =  sizeof(struct sctp_sock),
7136	.sysctl_mem  =  sysctl_sctp_mem,
7137	.sysctl_rmem =  sysctl_sctp_rmem,
7138	.sysctl_wmem =  sysctl_sctp_wmem,
7139	.memory_pressure = &sctp_memory_pressure,
7140	.enter_memory_pressure = sctp_enter_memory_pressure,
7141	.memory_allocated = &sctp_memory_allocated,
7142	.sockets_allocated = &sctp_sockets_allocated,
7143};
7144
7145#if IS_ENABLED(CONFIG_IPV6)
7146
7147struct proto sctpv6_prot = {
7148	.name		= "SCTPv6",
7149	.owner		= THIS_MODULE,
7150	.close		= sctp_close,
7151	.connect	= sctp_connect,
7152	.disconnect	= sctp_disconnect,
7153	.accept		= sctp_accept,
7154	.ioctl		= sctp_ioctl,
7155	.init		= sctp_init_sock,
7156	.destroy	= sctp_destroy_sock,
7157	.shutdown	= sctp_shutdown,
7158	.setsockopt	= sctp_setsockopt,
7159	.getsockopt	= sctp_getsockopt,
7160	.sendmsg	= sctp_sendmsg,
7161	.recvmsg	= sctp_recvmsg,
7162	.bind		= sctp_bind,
7163	.backlog_rcv	= sctp_backlog_rcv,
7164	.hash		= sctp_hash,
7165	.unhash		= sctp_unhash,
7166	.get_port	= sctp_get_port,
7167	.obj_size	= sizeof(struct sctp6_sock),
7168	.sysctl_mem	= sysctl_sctp_mem,
7169	.sysctl_rmem	= sysctl_sctp_rmem,
7170	.sysctl_wmem	= sysctl_sctp_wmem,
7171	.memory_pressure = &sctp_memory_pressure,
7172	.enter_memory_pressure = sctp_enter_memory_pressure,
7173	.memory_allocated = &sctp_memory_allocated,
7174	.sockets_allocated = &sctp_sockets_allocated,
7175};
7176#endif /* IS_ENABLED(CONFIG_IPV6) */