net/sctp/socket.c

   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 #include <linux/types.h>
  61 #include <linux/kernel.h>
  62 #include <linux/wait.h>
  63 #include <linux/time.h>
  64 #include <linux/ip.h>
  65 #include <linux/capability.h>
  66 #include <linux/fcntl.h>
  67 #include <linux/poll.h>
  68 #include <linux/init.h>
  69 #include <linux/crypto.h>
  70
  71 #include <net/ip.h>
  72 #include <net/icmp.h>
  73 #include <net/route.h>
  74 #include <net/ipv6.h>
  75 #include <net/inet_common.h>
  76
  77 #include <linux/socket.h> /* for sa_family_t */
  78 #include <net/sock.h>
  79 #include <net/sctp/sctp.h>
  80 #include <net/sctp/sm.h>
  81
  82 /* WARNING:  Please do not remove the SCTP_STATIC attribute to
  83  * any of the functions below as they are used to export functions
  84  * used by a project regression testsuite.
  85  */
  86
  87 /* Forward declarations for internal helper functions. */
  88 static int sctp_writeable(struct sock *sk);
  89 static void sctp_wfree(struct sk_buff *skb);
  90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
  91                                 size_t msg_len);
  92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
  93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
  94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
  95 static void sctp_wait_for_close(struct sock *sk, long timeo);
  96 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
  97                                         union sctp_addr *addr, int len);
  98 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
  99 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
 100 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
 101 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
 102 static int sctp_send_asconf(struct sctp_association *asoc,
 103                             struct sctp_chunk *chunk);
 104 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
 105 static int sctp_autobind(struct sock *sk);
 106 static void sctp_sock_migrate(struct sock *, struct sock *,
 107                               struct sctp_association *, sctp_socket_type_t);
 108 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
 109
 110 extern struct kmem_cache *sctp_bucket_cachep;
 111 extern int sysctl_sctp_mem[3];
 112 extern int sysctl_sctp_rmem[3];
 113 extern int sysctl_sctp_wmem[3];
 114
 115 static int sctp_memory_pressure;
 116 static atomic_t sctp_memory_allocated;
 117 static atomic_t sctp_sockets_allocated;
 118
 119 static void sctp_enter_memory_pressure(struct sock *sk)
 120 {
 121         sctp_memory_pressure = 1;
 122 }
 123
 124
 125 /* Get the sndbuf space available at the time on the association.  */
 126 static inline int sctp_wspace(struct sctp_association *asoc)
 127 {
 128         int amt;
 129
 130         if (asoc->ep->sndbuf_policy)
 131                 amt = asoc->sndbuf_used;
 132         else
 133                 amt = atomic_read(&asoc->base.sk->sk_wmem_alloc);
 134
 135         if (amt >= asoc->base.sk->sk_sndbuf) {
 136                 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
 137                         amt = 0;
 138                 else {
 139                         amt = sk_stream_wspace(asoc->base.sk);
 140                         if (amt < 0)
 141                                 amt = 0;
 142                 }
 143         } else {
 144                 amt = asoc->base.sk->sk_sndbuf - amt;
 145         }
 146         return amt;
 147 }
 148
 149 /* Increment the used sndbuf space count of the corresponding association by
 150  * the size of the outgoing data chunk.
 151  * Also, set the skb destructor for sndbuf accounting later.
 152  *
 153  * Since it is always 1-1 between chunk and skb, and also a new skb is always
 154  * allocated for chunk bundling in sctp_packet_transmit(), we can use the
 155  * destructor in the data chunk skb for the purpose of the sndbuf space
 156  * tracking.
 157  */
 158 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
 159 {
 160         struct sctp_association *asoc = chunk->asoc;
 161         struct sock *sk = asoc->base.sk;
 162
 163         /* The sndbuf space is tracked per association.  */
 164         sctp_association_hold(asoc);
 165
 166         skb_set_owner_w(chunk->skb, sk);
 167
 168         chunk->skb->destructor = sctp_wfree;
 169         /* Save the chunk pointer in skb for sctp_wfree to use later.  */
 170         *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
 171
 172         asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
 173                                 sizeof(struct sk_buff) +
 174                                 sizeof(struct sctp_chunk);
 175
 176         atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
 177         sk->sk_wmem_queued += chunk->skb->truesize;
 178         sk_mem_charge(sk, chunk->skb->truesize);
 179 }
 180
 181 /* Verify that this is a valid address. */
 182 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
 183                                    int len)
 184 {
 185         struct sctp_af *af;
 186
 187         /* Verify basic sockaddr. */
 188         af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
 189         if (!af)
 190                 return -EINVAL;
 191
 192         /* Is this a valid SCTP address?  */
 193         if (!af->addr_valid(addr, sctp_sk(sk), NULL))
 194                 return -EINVAL;
 195
 196         if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
 197                 return -EINVAL;
 198
 199         return 0;
 200 }
 201
 202 /* Look up the association by its id.  If this is not a UDP-style
 203  * socket, the ID field is always ignored.
 204  */
 205 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
 206 {
 207         struct sctp_association *asoc = NULL;
 208
 209         /* If this is not a UDP-style socket, assoc id should be ignored. */
 210         if (!sctp_style(sk, UDP)) {
 211                 /* Return NULL if the socket state is not ESTABLISHED. It
 212                  * could be a TCP-style listening socket or a socket which
 213                  * hasn't yet called connect() to establish an association.
 214                  */
 215                 if (!sctp_sstate(sk, ESTABLISHED))
 216                         return NULL;
 217
 218                 /* Get the first and the only association from the list. */
 219                 if (!list_empty(&sctp_sk(sk)->ep->asocs))
 220                         asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
 221                                           struct sctp_association, asocs);
 222                 return asoc;
 223         }
 224
 225         /* Otherwise this is a UDP-style socket. */
 226         if (!id || (id == (sctp_assoc_t)-1))
 227                 return NULL;
 228
 229         spin_lock_bh(&sctp_assocs_id_lock);
 230         asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
 231         spin_unlock_bh(&sctp_assocs_id_lock);
 232
 233         if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
 234                 return NULL;
 235
 236         return asoc;
 237 }
 238
 239 /* Look up the transport from an address and an assoc id. If both address and
 240  * id are specified, the associations matching the address and the id should be
 241  * the same.
 242  */
 243 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
 244                                               struct sockaddr_storage *addr,
 245                                               sctp_assoc_t id)
 246 {
 247         struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
 248         struct sctp_transport *transport;
 249         union sctp_addr *laddr = (union sctp_addr *)addr;
 250
 251         addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
 252                                                laddr,
 253                                                &transport);
 254
 255         if (!addr_asoc)
 256                 return NULL;
 257
 258         id_asoc = sctp_id2assoc(sk, id);
 259         if (id_asoc && (id_asoc != addr_asoc))
 260                 return NULL;
 261
 262         sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
 263                                                 (union sctp_addr *)addr);
 264
 265         return transport;
 266 }
 267
 268 /* API 3.1.2 bind() - UDP Style Syntax
 269  * The syntax of bind() is,
 270  *
 271  *   ret = bind(int sd, struct sockaddr *addr, int addrlen);
 272  *
 273  *   sd      - the socket descriptor returned by socket().
 274  *   addr    - the address structure (struct sockaddr_in or struct
 275  *             sockaddr_in6 [RFC 2553]),
 276  *   addr_len - the size of the address structure.
 277  */
 278 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
 279 {
 280         int retval = 0;
 281
 282         sctp_lock_sock(sk);
 283
 284         SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
 285                           sk, addr, addr_len);
 286
 287         /* Disallow binding twice. */
 288         if (!sctp_sk(sk)->ep->base.bind_addr.port)
 289                 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
 290                                       addr_len);
 291         else
 292                 retval = -EINVAL;
 293
 294         sctp_release_sock(sk);
 295
 296         return retval;
 297 }
 298
 299 static long sctp_get_port_local(struct sock *, union sctp_addr *);
 300
 301 /* Verify this is a valid sockaddr. */
 302 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
 303                                         union sctp_addr *addr, int len)
 304 {
 305         struct sctp_af *af;
 306
 307         /* Check minimum size.  */
 308         if (len < sizeof (struct sockaddr))
 309                 return NULL;
 310
 311         /* V4 mapped address are really of AF_INET family */
 312         if (addr->sa.sa_family == AF_INET6 &&
 313             ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
 314                 if (!opt->pf->af_supported(AF_INET, opt))
 315                         return NULL;
 316         } else {
 317                 /* Does this PF support this AF? */
 318                 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
 319                         return NULL;
 320         }
 321
 322         /* If we get this far, af is valid. */
 323         af = sctp_get_af_specific(addr->sa.sa_family);
 324
 325         if (len < af->sockaddr_len)
 326                 return NULL;
 327
 328         return af;
 329 }
 330
 331 /* Bind a local address either to an endpoint or to an association.  */
 332 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
 333 {
 334         struct sctp_sock *sp = sctp_sk(sk);
 335         struct sctp_endpoint *ep = sp->ep;
 336         struct sctp_bind_addr *bp = &ep->base.bind_addr;
 337         struct sctp_af *af;
 338         unsigned short snum;
 339         int ret = 0;
 340
 341         /* Common sockaddr verification. */
 342         af = sctp_sockaddr_af(sp, addr, len);
 343         if (!af) {
 344                 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
 345                                   sk, addr, len);
 346                 return -EINVAL;
 347         }
 348
 349         snum = ntohs(addr->v4.sin_port);
 350
 351         SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
 352                                  ", port: %d, new port: %d, len: %d)\n",
 353                                  sk,
 354                                  addr,
 355                                  bp->port, snum,
 356                                  len);
 357
 358         /* PF specific bind() address verification. */
 359         if (!sp->pf->bind_verify(sp, addr))
 360                 return -EADDRNOTAVAIL;
 361
 362         /* We must either be unbound, or bind to the same port.
 363          * It's OK to allow 0 ports if we are already bound.
 364          * We'll just inhert an already bound port in this case
 365          */
 366         if (bp->port) {
 367                 if (!snum)
 368                         snum = bp->port;
 369                 else if (snum != bp->port) {
 370                         SCTP_DEBUG_PRINTK("sctp_do_bind:"
 371                                   " New port %d does not match existing port "
 372                                   "%d.\n", snum, bp->port);
 373                         return -EINVAL;
 374                 }
 375         }
 376
 377         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 378                 return -EACCES;
 379
 380         /* See if the address matches any of the addresses we may have
 381          * already bound before checking against other endpoints.
 382          */
 383         if (sctp_bind_addr_match(bp, addr, sp))
 384                 return -EINVAL;
 385
 386         /* Make sure we are allowed to bind here.
 387          * The function sctp_get_port_local() does duplicate address
 388          * detection.
 389          */
 390         addr->v4.sin_port = htons(snum);
 391         if ((ret = sctp_get_port_local(sk, addr))) {
 392                 return -EADDRINUSE;
 393         }
 394
 395         /* Refresh ephemeral port.  */
 396         if (!bp->port)
 397                 bp->port = inet_sk(sk)->num;
 398
 399         /* Add the address to the bind address list.
 400          * Use GFP_ATOMIC since BHs will be disabled.
 401          */
 402         ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
 403
 404         /* Copy back into socket for getsockname() use. */
 405         if (!ret) {
 406                 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
 407                 af->to_sk_saddr(addr, sk);
 408         }
 409
 410         return ret;
 411 }
 412
 413  /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
 414  *
 415  * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
 416  * at any one time.  If a sender, after sending an ASCONF chunk, decides
 417  * it needs to transfer another ASCONF Chunk, it MUST wait until the
 418  * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
 419  * subsequent ASCONF. Note this restriction binds each side, so at any
 420  * time two ASCONF may be in-transit on any given association (one sent
 421  * from each endpoint).
 422  */
 423 static int sctp_send_asconf(struct sctp_association *asoc,
 424                             struct sctp_chunk *chunk)
 425 {
 426         int             retval = 0;
 427
 428         /* If there is an outstanding ASCONF chunk, queue it for later
 429          * transmission.
 430          */
 431         if (asoc->addip_last_asconf) {
 432                 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
 433                 goto out;
 434         }
 435
 436         /* Hold the chunk until an ASCONF_ACK is received. */
 437         sctp_chunk_hold(chunk);
 438         retval = sctp_primitive_ASCONF(asoc, chunk);
 439         if (retval)
 440                 sctp_chunk_free(chunk);
 441         else
 442                 asoc->addip_last_asconf = chunk;
 443
 444 out:
 445         return retval;
 446 }
 447
 448 /* Add a list of addresses as bind addresses to local endpoint or
 449  * association.
 450  *
 451  * Basically run through each address specified in the addrs/addrcnt
 452  * array/length pair, determine if it is IPv6 or IPv4 and call
 453  * sctp_do_bind() on it.
 454  *
 455  * If any of them fails, then the operation will be reversed and the
 456  * ones that were added will be removed.
 457  *
 458  * Only sctp_setsockopt_bindx() is supposed to call this function.
 459  */
 460 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
 461 {
 462         int cnt;
 463         int retval = 0;
 464         void *addr_buf;
 465         struct sockaddr *sa_addr;
 466         struct sctp_af *af;
 467
 468         SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
 469                           sk, addrs, addrcnt);
 470
 471         addr_buf = addrs;
 472         for (cnt = 0; cnt < addrcnt; cnt++) {
 473                 /* The list may contain either IPv4 or IPv6 address;
 474                  * determine the address length for walking thru the list.
 475                  */
 476                 sa_addr = (struct sockaddr *)addr_buf;
 477                 af = sctp_get_af_specific(sa_addr->sa_family);
 478                 if (!af) {
 479                         retval = -EINVAL;
 480                         goto err_bindx_add;
 481                 }
 482
 483                 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
 484                                       af->sockaddr_len);
 485
 486                 addr_buf += af->sockaddr_len;
 487
 488 err_bindx_add:
 489                 if (retval < 0) {
 490                         /* Failed. Cleanup the ones that have been added */
 491                         if (cnt > 0)
 492                                 sctp_bindx_rem(sk, addrs, cnt);
 493                         return retval;
 494                 }
 495         }
 496
 497         return retval;
 498 }
 499
 500 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
 501  * associations that are part of the endpoint indicating that a list of local
 502  * addresses are added to the endpoint.
 503  *
 504  * If any of the addresses is already in the bind address list of the
 505  * association, we do not send the chunk for that association.  But it will not
 506  * affect other associations.
 507  *
 508  * Only sctp_setsockopt_bindx() is supposed to call this function.
 509  */
 510 static int sctp_send_asconf_add_ip(struct sock          *sk,
 511                                    struct sockaddr      *addrs,
 512                                    int                  addrcnt)
 513 {
 514         struct sctp_sock                *sp;
 515         struct sctp_endpoint            *ep;
 516         struct sctp_association         *asoc;
 517         struct sctp_bind_addr           *bp;
 518         struct sctp_chunk               *chunk;
 519         struct sctp_sockaddr_entry      *laddr;
 520         union sctp_addr                 *addr;
 521         union sctp_addr                 saveaddr;
 522         void                            *addr_buf;
 523         struct sctp_af                  *af;
 524         struct list_head                *p;
 525         int                             i;
 526         int                             retval = 0;
 527
 528         if (!sctp_addip_enable)
 529                 return retval;
 530
 531         sp = sctp_sk(sk);
 532         ep = sp->ep;
 533
 534         SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
 535                           __func__, sk, addrs, addrcnt);
 536
 537         list_for_each_entry(asoc, &ep->asocs, asocs) {
 538
 539                 if (!asoc->peer.asconf_capable)
 540                         continue;
 541
 542                 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
 543                         continue;
 544
 545                 if (!sctp_state(asoc, ESTABLISHED))
 546                         continue;
 547
 548                 /* Check if any address in the packed array of addresses is
 549                  * in the bind address list of the association. If so,
 550                  * do not send the asconf chunk to its peer, but continue with
 551                  * other associations.
 552                  */
 553                 addr_buf = addrs;
 554                 for (i = 0; i < addrcnt; i++) {
 555                         addr = (union sctp_addr *)addr_buf;
 556                         af = sctp_get_af_specific(addr->v4.sin_family);
 557                         if (!af) {
 558                                 retval = -EINVAL;
 559                                 goto out;
 560                         }
 561
 562                         if (sctp_assoc_lookup_laddr(asoc, addr))
 563                                 break;
 564
 565                         addr_buf += af->sockaddr_len;
 566                 }
 567                 if (i < addrcnt)
 568                         continue;
 569
 570                 /* Use the first valid address in bind addr list of
 571                  * association as Address Parameter of ASCONF CHUNK.
 572                  */
 573                 bp = &asoc->base.bind_addr;
 574                 p = bp->address_list.next;
 575                 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
 576                 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
 577                                                    addrcnt, SCTP_PARAM_ADD_IP);
 578                 if (!chunk) {
 579                         retval = -ENOMEM;
 580                         goto out;
 581                 }
 582
 583                 retval = sctp_send_asconf(asoc, chunk);
 584                 if (retval)
 585                         goto out;
 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 = (union sctp_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         }
 600
 601 out:
 602         return retval;
 603 }
 604
 605 /* Remove a list of addresses from bind addresses list.  Do not remove the
 606  * last address.
 607  *
 608  * Basically run through each address specified in the addrs/addrcnt
 609  * array/length pair, determine if it is IPv6 or IPv4 and call
 610  * sctp_del_bind() on it.
 611  *
 612  * If any of them fails, then the operation will be reversed and the
 613  * ones that were removed will be added back.
 614  *
 615  * At least one address has to be left; if only one address is
 616  * available, the operation will return -EBUSY.
 617  *
 618  * Only sctp_setsockopt_bindx() is supposed to call this function.
 619  */
 620 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
 621 {
 622         struct sctp_sock *sp = sctp_sk(sk);
 623         struct sctp_endpoint *ep = sp->ep;
 624         int cnt;
 625         struct sctp_bind_addr *bp = &ep->base.bind_addr;
 626         int retval = 0;
 627         void *addr_buf;
 628         union sctp_addr *sa_addr;
 629         struct sctp_af *af;
 630
 631         SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
 632                           sk, addrs, addrcnt);
 633
 634         addr_buf = addrs;
 635         for (cnt = 0; cnt < addrcnt; cnt++) {
 636                 /* If the bind address list is empty or if there is only one
 637                  * bind address, there is nothing more to be removed (we need
 638                  * at least one address here).
 639                  */
 640                 if (list_empty(&bp->address_list) ||
 641                     (sctp_list_single_entry(&bp->address_list))) {
 642                         retval = -EBUSY;
 643                         goto err_bindx_rem;
 644                 }
 645
 646                 sa_addr = (union sctp_addr *)addr_buf;
 647                 af = sctp_get_af_specific(sa_addr->sa.sa_family);
 648                 if (!af) {
 649                         retval = -EINVAL;
 650                         goto err_bindx_rem;
 651                 }
 652
 653                 if (!af->addr_valid(sa_addr, sp, NULL)) {
 654                         retval = -EADDRNOTAVAIL;
 655                         goto err_bindx_rem;
 656                 }
 657
 658                 if (sa_addr->v4.sin_port != htons(bp->port)) {
 659                         retval = -EINVAL;
 660                         goto err_bindx_rem;
 661                 }
 662
 663                 /* FIXME - There is probably a need to check if sk->sk_saddr and
 664                  * sk->sk_rcv_addr are currently set to one of the addresses to
 665                  * be removed. This is something which needs to be looked into
 666                  * when we are fixing the outstanding issues with multi-homing
 667                  * socket routing and failover schemes. Refer to comments in
 668                  * sctp_do_bind(). -daisy
 669                  */
 670                 retval = sctp_del_bind_addr(bp, sa_addr);
 671
 672                 addr_buf += af->sockaddr_len;
 673 err_bindx_rem:
 674                 if (retval < 0) {
 675                         /* Failed. Add the ones that has been removed back */
 676                         if (cnt > 0)
 677                                 sctp_bindx_add(sk, addrs, cnt);
 678                         return retval;
 679                 }
 680         }
 681
 682         return retval;
 683 }
 684
 685 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
 686  * the associations that are part of the endpoint indicating that a list of
 687  * local addresses are removed from the endpoint.
 688  *
 689  * If any of the addresses is already in the bind address list of the
 690  * association, we do not send the chunk for that association.  But it will not
 691  * affect other associations.
 692  *
 693  * Only sctp_setsockopt_bindx() is supposed to call this function.
 694  */
 695 static int sctp_send_asconf_del_ip(struct sock          *sk,
 696                                    struct sockaddr      *addrs,
 697                                    int                  addrcnt)
 698 {
 699         struct sctp_sock        *sp;
 700         struct sctp_endpoint    *ep;
 701         struct sctp_association *asoc;
 702         struct sctp_transport   *transport;
 703         struct sctp_bind_addr   *bp;
 704         struct sctp_chunk       *chunk;
 705         union sctp_addr         *laddr;
 706         void                    *addr_buf;
 707         struct sctp_af          *af;
 708         struct sctp_sockaddr_entry *saddr;
 709         int                     i;
 710         int                     retval = 0;
 711
 712         if (!sctp_addip_enable)
 713                 return retval;
 714
 715         sp = sctp_sk(sk);
 716         ep = sp->ep;
 717
 718         SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
 719                           __func__, sk, addrs, addrcnt);
 720
 721         list_for_each_entry(asoc, &ep->asocs, asocs) {
 722
 723                 if (!asoc->peer.asconf_capable)
 724                         continue;
 725
 726                 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
 727                         continue;
 728
 729                 if (!sctp_state(asoc, ESTABLISHED))
 730                         continue;
 731
 732                 /* Check if any address in the packed array of addresses is
 733                  * not present in the bind address list of the association.
 734                  * If so, do not send the asconf chunk to its peer, but
 735                  * continue with other associations.
 736                  */
 737                 addr_buf = addrs;
 738                 for (i = 0; i < addrcnt; i++) {
 739                         laddr = (union sctp_addr *)addr_buf;
 740                         af = sctp_get_af_specific(laddr->v4.sin_family);
 741                         if (!af) {
 742                                 retval = -EINVAL;
 743                                 goto out;
 744                         }
 745
 746                         if (!sctp_assoc_lookup_laddr(asoc, laddr))
 747                                 break;
 748
 749                         addr_buf += af->sockaddr_len;
 750                 }
 751                 if (i < addrcnt)
 752                         continue;
 753
 754                 /* Find one address in the association's bind address list
 755                  * that is not in the packed array of addresses. This is to
 756                  * make sure that we do not delete all the addresses in the
 757                  * association.
 758                  */
 759                 bp = &asoc->base.bind_addr;
 760                 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
 761                                                addrcnt, sp);
 762                 if (!laddr)
 763                         continue;
 764
 765                 /* We do not need RCU protection throughout this loop
 766                  * because this is done under a socket lock from the
 767                  * setsockopt call.
 768                  */
 769                 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
 770                                                    SCTP_PARAM_DEL_IP);
 771                 if (!chunk) {
 772                         retval = -ENOMEM;
 773                         goto out;
 774                 }
 775
 776                 /* Reset use_as_src flag for the addresses in the bind address
 777                  * list that are to be deleted.
 778                  */
 779                 addr_buf = addrs;
 780                 for (i = 0; i < addrcnt; i++) {
 781                         laddr = (union sctp_addr *)addr_buf;
 782                         af = sctp_get_af_specific(laddr->v4.sin_family);
 783                         list_for_each_entry(saddr, &bp->address_list, list) {
 784                                 if (sctp_cmp_addr_exact(&saddr->a, laddr))
 785                                         saddr->state = SCTP_ADDR_DEL;
 786                         }
 787                         addr_buf += af->sockaddr_len;
 788                 }
 789
 790                 /* Update the route and saddr entries for all the transports
 791                  * as some of the addresses in the bind address list are
 792                  * about to be deleted and cannot be used as source addresses.
 793                  */
 794                 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
 795                                         transports) {
 796                         dst_release(transport->dst);
 797                         sctp_transport_route(transport, NULL,
 798                                              sctp_sk(asoc->base.sk));
 799                 }
 800
 801                 retval = sctp_send_asconf(asoc, chunk);
 802         }
 803 out:
 804         return retval;
 805 }
 806
 807 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
 808  *
 809  * API 8.1
 810  * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
 811  *                int flags);
 812  *
 813  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
 814  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
 815  * or IPv6 addresses.
 816  *
 817  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
 818  * Section 3.1.2 for this usage.
 819  *
 820  * addrs is a pointer to an array of one or more socket addresses. Each
 821  * address is contained in its appropriate structure (i.e. struct
 822  * sockaddr_in or struct sockaddr_in6) the family of the address type
 823  * must be used to distinguish the address length (note that this
 824  * representation is termed a "packed array" of addresses). The caller
 825  * specifies the number of addresses in the array with addrcnt.
 826  *
 827  * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
 828  * -1, and sets errno to the appropriate error code.
 829  *
 830  * For SCTP, the port given in each socket address must be the same, or
 831  * sctp_bindx() will fail, setting errno to EINVAL.
 832  *
 833  * The flags parameter is formed from the bitwise OR of zero or more of
 834  * the following currently defined flags:
 835  *
 836  * SCTP_BINDX_ADD_ADDR
 837  *
 838  * SCTP_BINDX_REM_ADDR
 839  *
 840  * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
 841  * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
 842  * addresses from the association. The two flags are mutually exclusive;
 843  * if both are given, sctp_bindx() will fail with EINVAL. A caller may
 844  * not remove all addresses from an association; sctp_bindx() will
 845  * reject such an attempt with EINVAL.
 846  *
 847  * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
 848  * additional addresses with an endpoint after calling bind().  Or use
 849  * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
 850  * socket is associated with so that no new association accepted will be
 851  * associated with those addresses. If the endpoint supports dynamic
 852  * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
 853  * endpoint to send the appropriate message to the peer to change the
 854  * peers address lists.
 855  *
 856  * Adding and removing addresses from a connected association is
 857  * optional functionality. Implementations that do not support this
 858  * functionality should return EOPNOTSUPP.
 859  *
 860  * Basically do nothing but copying the addresses from user to kernel
 861  * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
 862  * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
 863  * from userspace.
 864  *
 865  * We don't use copy_from_user() for optimization: we first do the
 866  * sanity checks (buffer size -fast- and access check-healthy
 867  * pointer); if all of those succeed, then we can alloc the memory
 868  * (expensive operation) needed to copy the data to kernel. Then we do
 869  * the copying without checking the user space area
 870  * (__copy_from_user()).
 871  *
 872  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
 873  * it.
 874  *
 875  * sk        The sk of the socket
 876  * addrs     The pointer to the addresses in user land
 877  * addrssize Size of the addrs buffer
 878  * op        Operation to perform (add or remove, see the flags of
 879  *           sctp_bindx)
 880  *
 881  * Returns 0 if ok, <0 errno code on error.
 882  */
 883 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
 884                                       struct sockaddr __user *addrs,
 885                                       int addrs_size, int op)
 886 {
 887         struct sockaddr *kaddrs;
 888         int err;
 889         int addrcnt = 0;
 890         int walk_size = 0;
 891         struct sockaddr *sa_addr;
 892         void *addr_buf;
 893         struct sctp_af *af;
 894
 895         SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
 896                           " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
 897
 898         if (unlikely(addrs_size <= 0))
 899                 return -EINVAL;
 900
 901         /* Check the user passed a healthy pointer.  */
 902         if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
 903                 return -EFAULT;
 904
 905         /* Alloc space for the address array in kernel memory.  */
 906         kaddrs = kmalloc(addrs_size, GFP_KERNEL);
 907         if (unlikely(!kaddrs))
 908                 return -ENOMEM;
 909
 910         if (__copy_from_user(kaddrs, addrs, addrs_size)) {
 911                 kfree(kaddrs);
 912                 return -EFAULT;
 913         }
 914
 915         /* Walk through the addrs buffer and count the number of addresses. */
 916         addr_buf = kaddrs;
 917         while (walk_size < addrs_size) {
 918                 sa_addr = (struct sockaddr *)addr_buf;
 919                 af = sctp_get_af_specific(sa_addr->sa_family);
 920
 921                 /* If the address family is not supported or if this address
 922                  * causes the address buffer to overflow return EINVAL.
 923                  */
 924                 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
 925                         kfree(kaddrs);
 926                         return -EINVAL;
 927                 }
 928                 addrcnt++;
 929                 addr_buf += af->sockaddr_len;
 930                 walk_size += af->sockaddr_len;
 931         }
 932
 933         /* Do the work. */
 934         switch (op) {
 935         case SCTP_BINDX_ADD_ADDR:
 936                 err = sctp_bindx_add(sk, kaddrs, addrcnt);
 937                 if (err)
 938                         goto out;
 939                 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
 940                 break;
 941
 942         case SCTP_BINDX_REM_ADDR:
 943                 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
 944                 if (err)
 945                         goto out;
 946                 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
 947                 break;
 948
 949         default:
 950                 err = -EINVAL;
 951                 break;
 952         }
 953
 954 out:
 955         kfree(kaddrs);
 956
 957         return err;
 958 }
 959
 960 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
 961  *
 962  * Common routine for handling connect() and sctp_connectx().
 963  * Connect will come in with just a single address.
 964  */
 965 static int __sctp_connect(struct sock* sk,
 966                           struct sockaddr *kaddrs,
 967                           int addrs_size,
 968                           sctp_assoc_t *assoc_id)
 969 {
 970         struct sctp_sock *sp;
 971         struct sctp_endpoint *ep;
 972         struct sctp_association *asoc = NULL;
 973         struct sctp_association *asoc2;
 974         struct sctp_transport *transport;
 975         union sctp_addr to;
 976         struct sctp_af *af;
 977         sctp_scope_t scope;
 978         long timeo;
 979         int err = 0;
 980         int addrcnt = 0;
 981         int walk_size = 0;
 982         union sctp_addr *sa_addr = NULL;
 983         void *addr_buf;
 984         unsigned short port;
 985         unsigned int f_flags = 0;
 986
 987         sp = sctp_sk(sk);
 988         ep = sp->ep;
 989
 990         /* connect() cannot be done on a socket that is already in ESTABLISHED
 991          * state - UDP-style peeled off socket or a TCP-style socket that
 992          * is already connected.
 993          * It cannot be done even on a TCP-style listening socket.
 994          */
 995         if (sctp_sstate(sk, ESTABLISHED) ||
 996             (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
 997                 err = -EISCONN;
 998                 goto out_free;
 999         }
1000
1001         /* Walk through the addrs buffer and count the number of addresses. */
1002         addr_buf = kaddrs;
1003         while (walk_size < addrs_size) {
1004                 sa_addr = (union sctp_addr *)addr_buf;
1005                 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1006                 port = ntohs(sa_addr->v4.sin_port);
1007
1008                 /* If the address family is not supported or if this address
1009                  * causes the address buffer to overflow return EINVAL.
1010                  */
1011                 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1012                         err = -EINVAL;
1013                         goto out_free;
1014                 }
1015
1016                 /* Save current address so we can work with it */
1017                 memcpy(&to, sa_addr, af->sockaddr_len);
1018
1019                 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1020                 if (err)
1021                         goto out_free;
1022
1023                 /* Make sure the destination port is correctly set
1024                  * in all addresses.
1025                  */
1026                 if (asoc && asoc->peer.port && asoc->peer.port != port)
1027                         goto out_free;
1028
1029
1030                 /* Check if there already is a matching association on the
1031                  * endpoint (other than the one created here).
1032                  */
1033                 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1034                 if (asoc2 && asoc2 != asoc) {
1035                         if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1036                                 err = -EISCONN;
1037                         else
1038                                 err = -EALREADY;
1039                         goto out_free;
1040                 }
1041
1042                 /* If we could not find a matching association on the endpoint,
1043                  * make sure that there is no peeled-off association matching
1044                  * the peer address even on another socket.
1045                  */
1046                 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1047                         err = -EADDRNOTAVAIL;
1048                         goto out_free;
1049                 }
1050
1051                 if (!asoc) {
1052                         /* If a bind() or sctp_bindx() is not called prior to
1053                          * an sctp_connectx() call, the system picks an
1054                          * ephemeral port and will choose an address set
1055                          * equivalent to binding with a wildcard address.
1056                          */
1057                         if (!ep->base.bind_addr.port) {
1058                                 if (sctp_autobind(sk)) {
1059                                         err = -EAGAIN;
1060                                         goto out_free;
1061                                 }
1062                         } else {
1063                                 /*
1064                                  * If an unprivileged user inherits a 1-many
1065                                  * style socket with open associations on a
1066                                  * privileged port, it MAY be permitted to
1067                                  * accept new associations, but it SHOULD NOT
1068                                  * be permitted to open new associations.
1069                                  */
1070                                 if (ep->base.bind_addr.port < PROT_SOCK &&
1071                                     !capable(CAP_NET_BIND_SERVICE)) {
1072                                         err = -EACCES;
1073                                         goto out_free;
1074                                 }
1075                         }
1076
1077                         scope = sctp_scope(&to);
1078                         asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1079                         if (!asoc) {
1080                                 err = -ENOMEM;
1081                                 goto out_free;
1082                         }
1083                 }
1084
1085                 /* Prime the peer's transport structures.  */
1086                 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1087                                                 SCTP_UNKNOWN);
1088                 if (!transport) {
1089                         err = -ENOMEM;
1090                         goto out_free;
1091                 }
1092
1093                 addrcnt++;
1094                 addr_buf += af->sockaddr_len;
1095                 walk_size += af->sockaddr_len;
1096         }
1097
1098         err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1099         if (err < 0) {
1100                 goto out_free;
1101         }
1102
1103         err = sctp_primitive_ASSOCIATE(asoc, NULL);
1104         if (err < 0) {
1105                 goto out_free;
1106         }
1107
1108         /* Initialize sk's dport and daddr for getpeername() */
1109         inet_sk(sk)->dport = htons(asoc->peer.port);
1110         af = sctp_get_af_specific(sa_addr->sa.sa_family);
1111         af->to_sk_daddr(sa_addr, sk);
1112         sk->sk_err = 0;
1113
1114         /* in-kernel sockets don't generally have a file allocated to them
1115          * if all they do is call sock_create_kern().
1116          */
1117         if (sk->sk_socket->file)
1118                 f_flags = sk->sk_socket->file->f_flags;
1119
1120         timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1121
1122         err = sctp_wait_for_connect(asoc, &timeo);
1123         if (!err && assoc_id)
1124                 *assoc_id = asoc->assoc_id;
1125
1126         /* Don't free association on exit. */
1127         asoc = NULL;
1128
1129 out_free:
1130
1131         SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1132                           " kaddrs: %p err: %d\n",
1133                           asoc, kaddrs, err);
1134         if (asoc)
1135                 sctp_association_free(asoc);
1136         return err;
1137 }
1138
1139 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1140  *
1141  * API 8.9
1142  * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1143  *                      sctp_assoc_t *asoc);
1144  *
1145  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1146  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1147  * or IPv6 addresses.
1148  *
1149  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1150  * Section 3.1.2 for this usage.
1151  *
1152  * addrs is a pointer to an array of one or more socket addresses. Each
1153  * address is contained in its appropriate structure (i.e. struct
1154  * sockaddr_in or struct sockaddr_in6) the family of the address type
1155  * must be used to distengish the address length (note that this
1156  * representation is termed a "packed array" of addresses). The caller
1157  * specifies the number of addresses in the array with addrcnt.
1158  *
1159  * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1160  * the association id of the new association.  On failure, sctp_connectx()
1161  * returns -1, and sets errno to the appropriate error code.  The assoc_id
1162  * is not touched by the kernel.
1163  *
1164  * For SCTP, the port given in each socket address must be the same, or
1165  * sctp_connectx() will fail, setting errno to EINVAL.
1166  *
1167  * An application can use sctp_connectx to initiate an association with
1168  * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1169  * allows a caller to specify multiple addresses at which a peer can be
1170  * reached.  The way the SCTP stack uses the list of addresses to set up
1171  * the association is implementation dependant.  This function only
1172  * specifies that the stack will try to make use of all the addresses in
1173  * the list when needed.
1174  *
1175  * Note that the list of addresses passed in is only used for setting up
1176  * the association.  It does not necessarily equal the set of addresses
1177  * the peer uses for the resulting association.  If the caller wants to
1178  * find out the set of peer addresses, it must use sctp_getpaddrs() to
1179  * retrieve them after the association has been set up.
1180  *
1181  * Basically do nothing but copying the addresses from user to kernel
1182  * land and invoking either sctp_connectx(). This is used for tunneling
1183  * the sctp_connectx() request through sctp_setsockopt() from userspace.
1184  *
1185  * We don't use copy_from_user() for optimization: we first do the
1186  * sanity checks (buffer size -fast- and access check-healthy
1187  * pointer); if all of those succeed, then we can alloc the memory
1188  * (expensive operation) needed to copy the data to kernel. Then we do
1189  * the copying without checking the user space area
1190  * (__copy_from_user()).
1191  *
1192  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1193  * it.
1194  *
1195  * sk        The sk of the socket
1196  * addrs     The pointer to the addresses in user land
1197  * addrssize Size of the addrs buffer
1198  *
1199  * Returns >=0 if ok, <0 errno code on error.
1200  */
1201 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1202                                       struct sockaddr __user *addrs,
1203                                       int addrs_size,
1204                                       sctp_assoc_t *assoc_id)
1205 {
1206         int err = 0;
1207         struct sockaddr *kaddrs;
1208
1209         SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1210                           __func__, sk, addrs, addrs_size);
1211
1212         if (unlikely(addrs_size <= 0))
1213                 return -EINVAL;
1214
1215         /* Check the user passed a healthy pointer.  */
1216         if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1217                 return -EFAULT;
1218
1219         /* Alloc space for the address array in kernel memory.  */
1220         kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1221         if (unlikely(!kaddrs))
1222                 return -ENOMEM;
1223
1224         if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1225                 err = -EFAULT;
1226         } else {
1227                 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1228         }
1229
1230         kfree(kaddrs);
1231
1232         return err;
1233 }
1234
1235 /*
1236  * This is an older interface.  It's kept for backward compatibility
1237  * to the option that doesn't provide association id.
1238  */
1239 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1240                                       struct sockaddr __user *addrs,
1241                                       int addrs_size)
1242 {
1243         return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1244 }
1245
1246 /*
1247  * New interface for the API.  The since the API is done with a socket
1248  * option, to make it simple we feed back the association id is as a return
1249  * indication to the call.  Error is always negative and association id is
1250  * always positive.
1251  */
1252 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1253                                       struct sockaddr __user *addrs,
1254                                       int addrs_size)
1255 {
1256         sctp_assoc_t assoc_id = 0;
1257         int err = 0;
1258
1259         err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1260
1261         if (err)
1262                 return err;
1263         else
1264                 return assoc_id;
1265 }
1266
1267 /* API 3.1.4 close() - UDP Style Syntax
1268  * Applications use close() to perform graceful shutdown (as described in
1269  * Section 10.1 of [SCTP]) on ALL the associations currently represented
1270  * by a UDP-style socket.
1271  *
1272  * The syntax is
1273  *
1274  *   ret = close(int sd);
1275  *
1276  *   sd      - the socket descriptor of the associations to be closed.
1277  *
1278  * To gracefully shutdown a specific association represented by the
1279  * UDP-style socket, an application should use the sendmsg() call,
1280  * passing no user data, but including the appropriate flag in the
1281  * ancillary data (see Section xxxx).
1282  *
1283  * If sd in the close() call is a branched-off socket representing only
1284  * one association, the shutdown is performed on that association only.
1285  *
1286  * 4.1.6 close() - TCP Style Syntax
1287  *
1288  * Applications use close() to gracefully close down an association.
1289  *
1290  * The syntax is:
1291  *
1292  *    int close(int sd);
1293  *
1294  *      sd      - the socket descriptor of the association to be closed.
1295  *
1296  * After an application calls close() on a socket descriptor, no further
1297  * socket operations will succeed on that descriptor.
1298  *
1299  * API 7.1.4 SO_LINGER
1300  *
1301  * An application using the TCP-style socket can use this option to
1302  * perform the SCTP ABORT primitive.  The linger option structure is:
1303  *
1304  *  struct  linger {
1305  *     int     l_onoff;                // option on/off
1306  *     int     l_linger;               // linger time
1307  * };
1308  *
1309  * To enable the option, set l_onoff to 1.  If the l_linger value is set
1310  * to 0, calling close() is the same as the ABORT primitive.  If the
1311  * value is set to a negative value, the setsockopt() call will return
1312  * an error.  If the value is set to a positive value linger_time, the
1313  * close() can be blocked for at most linger_time ms.  If the graceful
1314  * shutdown phase does not finish during this period, close() will
1315  * return but the graceful shutdown phase continues in the system.
1316  */
1317 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1318 {
1319         struct sctp_endpoint *ep;
1320         struct sctp_association *asoc;
1321         struct list_head *pos, *temp;
1322
1323         SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1324
1325         sctp_lock_sock(sk);
1326         sk->sk_shutdown = SHUTDOWN_MASK;
1327
1328         ep = sctp_sk(sk)->ep;
1329
1330         /* Walk all associations on an endpoint.  */
1331         list_for_each_safe(pos, temp, &ep->asocs) {
1332                 asoc = list_entry(pos, struct sctp_association, asocs);
1333
1334                 if (sctp_style(sk, TCP)) {
1335                         /* A closed association can still be in the list if
1336                          * it belongs to a TCP-style listening socket that is
1337                          * not yet accepted. If so, free it. If not, send an
1338                          * ABORT or SHUTDOWN based on the linger options.
1339                          */
1340                         if (sctp_state(asoc, CLOSED)) {
1341                                 sctp_unhash_established(asoc);
1342                                 sctp_association_free(asoc);
1343                                 continue;
1344                         }
1345                 }
1346
1347                 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1348                         struct sctp_chunk *chunk;
1349
1350                         chunk = sctp_make_abort_user(asoc, NULL, 0);
1351                         if (chunk)
1352                                 sctp_primitive_ABORT(asoc, chunk);
1353                 } else
1354                         sctp_primitive_SHUTDOWN(asoc, NULL);
1355         }
1356
1357         /* Clean up any skbs sitting on the receive queue.  */
1358         sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1359         sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1360
1361         /* On a TCP-style socket, block for at most linger_time if set. */
1362         if (sctp_style(sk, TCP) && timeout)
1363                 sctp_wait_for_close(sk, timeout);
1364
1365         /* This will run the backlog queue.  */
1366         sctp_release_sock(sk);
1367
1368         /* Supposedly, no process has access to the socket, but
1369          * the net layers still may.
1370          */
1371         sctp_local_bh_disable();
1372         sctp_bh_lock_sock(sk);
1373
1374         /* Hold the sock, since sk_common_release() will put sock_put()
1375          * and we have just a little more cleanup.
1376          */
1377         sock_hold(sk);
1378         sk_common_release(sk);
1379
1380         sctp_bh_unlock_sock(sk);
1381         sctp_local_bh_enable();
1382
1383         sock_put(sk);
1384
1385         SCTP_DBG_OBJCNT_DEC(sock);
1386 }
1387
1388 /* Handle EPIPE error. */
1389 static int sctp_error(struct sock *sk, int flags, int err)
1390 {
1391         if (err == -EPIPE)
1392                 err = sock_error(sk) ? : -EPIPE;
1393         if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1394                 send_sig(SIGPIPE, current, 0);
1395         return err;
1396 }
1397
1398 /* API 3.1.3 sendmsg() - UDP Style Syntax
1399  *
1400  * An application uses sendmsg() and recvmsg() calls to transmit data to
1401  * and receive data from its peer.
1402  *
1403  *  ssize_t sendmsg(int socket, const struct msghdr *message,
1404  *                  int flags);
1405  *
1406  *  socket  - the socket descriptor of the endpoint.
1407  *  message - pointer to the msghdr structure which contains a single
1408  *            user message and possibly some ancillary data.
1409  *
1410  *            See Section 5 for complete description of the data
1411  *            structures.
1412  *
1413  *  flags   - flags sent or received with the user message, see Section
1414  *            5 for complete description of the flags.
1415  *
1416  * Note:  This function could use a rewrite especially when explicit
1417  * connect support comes in.
1418  */
1419 /* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1420
1421 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1422
1423 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1424                              struct msghdr *msg, size_t msg_len)
1425 {
1426         struct sctp_sock *sp;
1427         struct sctp_endpoint *ep;
1428         struct sctp_association *new_asoc=NULL, *asoc=NULL;
1429         struct sctp_transport *transport, *chunk_tp;
1430         struct sctp_chunk *chunk;
1431         union sctp_addr to;
1432         struct sockaddr *msg_name = NULL;
1433         struct sctp_sndrcvinfo default_sinfo = { 0 };
1434         struct sctp_sndrcvinfo *sinfo;
1435         struct sctp_initmsg *sinit;
1436         sctp_assoc_t associd = 0;
1437         sctp_cmsgs_t cmsgs = { NULL };
1438         int err;
1439         sctp_scope_t scope;
1440         long timeo;
1441         __u16 sinfo_flags = 0;
1442         struct sctp_datamsg *datamsg;
1443         int msg_flags = msg->msg_flags;
1444
1445         SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1446                           sk, msg, msg_len);
1447
1448         err = 0;
1449         sp = sctp_sk(sk);
1450         ep = sp->ep;
1451
1452         SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1453
1454         /* We cannot send a message over a TCP-style listening socket. */
1455         if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1456                 err = -EPIPE;
1457                 goto out_nounlock;
1458         }
1459
1460         /* Parse out the SCTP CMSGs.  */
1461         err = sctp_msghdr_parse(msg, &cmsgs);
1462
1463         if (err) {
1464                 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1465                 goto out_nounlock;
1466         }
1467
1468         /* Fetch the destination address for this packet.  This
1469          * address only selects the association--it is not necessarily
1470          * the address we will send to.
1471          * For a peeled-off socket, msg_name is ignored.
1472          */
1473         if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1474                 int msg_namelen = msg->msg_namelen;
1475
1476                 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1477                                        msg_namelen);
1478                 if (err)
1479                         return err;
1480
1481                 if (msg_namelen > sizeof(to))
1482                         msg_namelen = sizeof(to);
1483                 memcpy(&to, msg->msg_name, msg_namelen);
1484                 msg_name = msg->msg_name;
1485         }
1486
1487         sinfo = cmsgs.info;
1488         sinit = cmsgs.init;
1489
1490         /* Did the user specify SNDRCVINFO?  */
1491         if (sinfo) {
1492                 sinfo_flags = sinfo->sinfo_flags;
1493                 associd = sinfo->sinfo_assoc_id;
1494         }
1495
1496         SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1497                           msg_len, sinfo_flags);
1498
1499         /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1500         if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1501                 err = -EINVAL;
1502                 goto out_nounlock;
1503         }
1504
1505         /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1506          * length messages when SCTP_EOF|SCTP_ABORT is not set.
1507          * If SCTP_ABORT is set, the message length could be non zero with
1508          * the msg_iov set to the user abort reason.
1509          */
1510         if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1511             (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1512                 err = -EINVAL;
1513                 goto out_nounlock;
1514         }
1515
1516         /* If SCTP_ADDR_OVER is set, there must be an address
1517          * specified in msg_name.
1518          */
1519         if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1520                 err = -EINVAL;
1521                 goto out_nounlock;
1522         }
1523
1524         transport = NULL;
1525
1526         SCTP_DEBUG_PRINTK("About to look up association.\n");
1527
1528         sctp_lock_sock(sk);
1529
1530         /* If a msg_name has been specified, assume this is to be used.  */
1531         if (msg_name) {
1532                 /* Look for a matching association on the endpoint. */
1533                 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1534                 if (!asoc) {
1535                         /* If we could not find a matching association on the
1536                          * endpoint, make sure that it is not a TCP-style
1537                          * socket that already has an association or there is
1538                          * no peeled-off association on another socket.
1539                          */
1540                         if ((sctp_style(sk, TCP) &&
1541                              sctp_sstate(sk, ESTABLISHED)) ||
1542                             sctp_endpoint_is_peeled_off(ep, &to)) {
1543                                 err = -EADDRNOTAVAIL;
1544                                 goto out_unlock;
1545                         }
1546                 }
1547         } else {
1548                 asoc = sctp_id2assoc(sk, associd);
1549                 if (!asoc) {
1550                         err = -EPIPE;
1551                         goto out_unlock;
1552                 }
1553         }
1554
1555         if (asoc) {
1556                 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1557
1558                 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1559                  * socket that has an association in CLOSED state. This can
1560                  * happen when an accepted socket has an association that is
1561                  * already CLOSED.
1562                  */
1563                 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1564                         err = -EPIPE;
1565                         goto out_unlock;
1566                 }
1567
1568                 if (sinfo_flags & SCTP_EOF) {
1569                         SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1570                                           asoc);
1571                         sctp_primitive_SHUTDOWN(asoc, NULL);
1572                         err = 0;
1573                         goto out_unlock;
1574                 }
1575                 if (sinfo_flags & SCTP_ABORT) {
1576
1577                         chunk = sctp_make_abort_user(asoc, msg, msg_len);
1578                         if (!chunk) {
1579                                 err = -ENOMEM;
1580                                 goto out_unlock;
1581                         }
1582
1583                         SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1584                         sctp_primitive_ABORT(asoc, chunk);
1585                         err = 0;
1586                         goto out_unlock;
1587                 }
1588         }
1589
1590         /* Do we need to create the association?  */
1591         if (!asoc) {
1592                 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1593
1594                 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1595                         err = -EINVAL;
1596                         goto out_unlock;
1597                 }
1598
1599                 /* Check for invalid stream against the stream counts,
1600                  * either the default or the user specified stream counts.
1601                  */
1602                 if (sinfo) {
1603                         if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1604                                 /* Check against the defaults. */
1605                                 if (sinfo->sinfo_stream >=
1606                                     sp->initmsg.sinit_num_ostreams) {
1607                                         err = -EINVAL;
1608                                         goto out_unlock;
1609                                 }
1610                         } else {
1611                                 /* Check against the requested.  */
1612                                 if (sinfo->sinfo_stream >=
1613                                     sinit->sinit_num_ostreams) {
1614                                         err = -EINVAL;
1615                                         goto out_unlock;
1616                                 }
1617                         }
1618                 }
1619
1620                 /*
1621                  * API 3.1.2 bind() - UDP Style Syntax
1622                  * If a bind() or sctp_bindx() is not called prior to a
1623                  * sendmsg() call that initiates a new association, the
1624                  * system picks an ephemeral port and will choose an address
1625                  * set equivalent to binding with a wildcard address.
1626                  */
1627                 if (!ep->base.bind_addr.port) {
1628                         if (sctp_autobind(sk)) {
1629                                 err = -EAGAIN;
1630                                 goto out_unlock;
1631                         }
1632                 } else {
1633                         /*
1634                          * If an unprivileged user inherits a one-to-many
1635                          * style socket with open associations on a privileged
1636                          * port, it MAY be permitted to accept new associations,
1637                          * but it SHOULD NOT be permitted to open new
1638                          * associations.
1639                          */
1640                         if (ep->base.bind_addr.port < PROT_SOCK &&
1641                             !capable(CAP_NET_BIND_SERVICE)) {
1642                                 err = -EACCES;
1643                                 goto out_unlock;
1644                         }
1645                 }
1646
1647                 scope = sctp_scope(&to);
1648                 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1649                 if (!new_asoc) {
1650                         err = -ENOMEM;
1651                         goto out_unlock;
1652                 }
1653                 asoc = new_asoc;
1654
1655                 /* If the SCTP_INIT ancillary data is specified, set all
1656                  * the association init values accordingly.
1657                  */
1658                 if (sinit) {
1659                         if (sinit->sinit_num_ostreams) {
1660                                 asoc->c.sinit_num_ostreams =
1661                                         sinit->sinit_num_ostreams;
1662                         }
1663                         if (sinit->sinit_max_instreams) {
1664                                 asoc->c.sinit_max_instreams =
1665                                         sinit->sinit_max_instreams;
1666                         }
1667                         if (sinit->sinit_max_attempts) {
1668                                 asoc->max_init_attempts
1669                                         = sinit->sinit_max_attempts;
1670                         }
1671                         if (sinit->sinit_max_init_timeo) {
1672                                 asoc->max_init_timeo =
1673                                  msecs_to_jiffies(sinit->sinit_max_init_timeo);
1674                         }
1675                 }
1676
1677                 /* Prime the peer's transport structures.  */
1678                 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1679                 if (!transport) {
1680                         err = -ENOMEM;
1681                         goto out_free;
1682                 }
1683                 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1684                 if (err < 0) {
1685                         err = -ENOMEM;
1686                         goto out_free;
1687                 }
1688         }
1689
1690         /* ASSERT: we have a valid association at this point.  */
1691         SCTP_DEBUG_PRINTK("We have a valid association.\n");
1692
1693         if (!sinfo) {
1694                 /* If the user didn't specify SNDRCVINFO, make up one with
1695                  * some defaults.
1696                  */
1697                 default_sinfo.sinfo_stream = asoc->default_stream;
1698                 default_sinfo.sinfo_flags = asoc->default_flags;
1699                 default_sinfo.sinfo_ppid = asoc->default_ppid;
1700                 default_sinfo.sinfo_context = asoc->default_context;
1701                 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1702                 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1703                 sinfo = &default_sinfo;
1704         }
1705
1706         /* API 7.1.7, the sndbuf size per association bounds the
1707          * maximum size of data that can be sent in a single send call.
1708          */
1709         if (msg_len > sk->sk_sndbuf) {
1710                 err = -EMSGSIZE;
1711                 goto out_free;
1712         }
1713
1714         if (asoc->pmtu_pending)
1715                 sctp_assoc_pending_pmtu(asoc);
1716
1717         /* If fragmentation is disabled and the message length exceeds the
1718          * association fragmentation point, return EMSGSIZE.  The I-D
1719          * does not specify what this error is, but this looks like
1720          * a great fit.
1721          */
1722         if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1723                 err = -EMSGSIZE;
1724                 goto out_free;
1725         }
1726
1727         if (sinfo) {
1728                 /* Check for invalid stream. */
1729                 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1730                         err = -EINVAL;
1731                         goto out_free;
1732                 }
1733         }
1734
1735         timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1736         if (!sctp_wspace(asoc)) {
1737                 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1738                 if (err)
1739                         goto out_free;
1740         }
1741
1742         /* If an address is passed with the sendto/sendmsg call, it is used
1743          * to override the primary destination address in the TCP model, or
1744          * when SCTP_ADDR_OVER flag is set in the UDP model.
1745          */
1746         if ((sctp_style(sk, TCP) && msg_name) ||
1747             (sinfo_flags & SCTP_ADDR_OVER)) {
1748                 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1749                 if (!chunk_tp) {
1750                         err = -EINVAL;
1751                         goto out_free;
1752                 }
1753         } else
1754                 chunk_tp = NULL;
1755
1756         /* Auto-connect, if we aren't connected already. */
1757         if (sctp_state(asoc, CLOSED)) {
1758                 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1759                 if (err < 0)
1760                         goto out_free;
1761                 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1762         }
1763
1764         /* Break the message into multiple chunks of maximum size. */
1765         datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1766         if (!datamsg) {
1767                 err = -ENOMEM;
1768                 goto out_free;
1769         }
1770
1771         /* Now send the (possibly) fragmented message. */
1772         list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1773                 sctp_chunk_hold(chunk);
1774
1775                 /* Do accounting for the write space.  */
1776                 sctp_set_owner_w(chunk);
1777
1778                 chunk->transport = chunk_tp;
1779
1780                 /* Send it to the lower layers.  Note:  all chunks
1781                  * must either fail or succeed.   The lower layer
1782                  * works that way today.  Keep it that way or this
1783                  * breaks.
1784                  */
1785                 err = sctp_primitive_SEND(asoc, chunk);
1786                 /* Did the lower layer accept the chunk? */
1787                 if (err)
1788                         sctp_chunk_free(chunk);
1789                 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1790         }
1791
1792         sctp_datamsg_put(datamsg);
1793         if (err)
1794                 goto out_free;
1795         else
1796                 err = msg_len;
1797
1798         /* If we are already past ASSOCIATE, the lower
1799          * layers are responsible for association cleanup.
1800          */
1801         goto out_unlock;
1802
1803 out_free:
1804         if (new_asoc)
1805                 sctp_association_free(asoc);
1806 out_unlock:
1807         sctp_release_sock(sk);
1808
1809 out_nounlock:
1810         return sctp_error(sk, msg_flags, err);
1811
1812 #if 0
1813 do_sock_err:
1814         if (msg_len)
1815                 err = msg_len;
1816         else
1817                 err = sock_error(sk);
1818         goto out;
1819
1820 do_interrupted:
1821         if (msg_len)
1822                 err = msg_len;
1823         goto out;
1824 #endif /* 0 */
1825 }
1826
1827 /* This is an extended version of skb_pull() that removes the data from the
1828  * start of a skb even when data is spread across the list of skb's in the
1829  * frag_list. len specifies the total amount of data that needs to be removed.
1830  * when 'len' bytes could be removed from the skb, it returns 0.
1831  * If 'len' exceeds the total skb length,  it returns the no. of bytes that
1832  * could not be removed.
1833  */
1834 static int sctp_skb_pull(struct sk_buff *skb, int len)
1835 {
1836         struct sk_buff *list;
1837         int skb_len = skb_headlen(skb);
1838         int rlen;
1839
1840         if (len <= skb_len) {
1841                 __skb_pull(skb, len);
1842                 return 0;
1843         }
1844         len -= skb_len;
1845         __skb_pull(skb, skb_len);
1846
1847         for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1848                 rlen = sctp_skb_pull(list, len);
1849                 skb->len -= (len-rlen);
1850                 skb->data_len -= (len-rlen);
1851
1852                 if (!rlen)
1853                         return 0;
1854
1855                 len = rlen;
1856         }
1857
1858         return len;
1859 }
1860
1861 /* API 3.1.3  recvmsg() - UDP Style Syntax
1862  *
1863  *  ssize_t recvmsg(int socket, struct msghdr *message,
1864  *                    int flags);
1865  *
1866  *  socket  - the socket descriptor of the endpoint.
1867  *  message - pointer to the msghdr structure which contains a single
1868  *            user message and possibly some ancillary data.
1869  *
1870  *            See Section 5 for complete description of the data
1871  *            structures.
1872  *
1873  *  flags   - flags sent or received with the user message, see Section
1874  *            5 for complete description of the flags.
1875  */
1876 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1877
1878 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1879                              struct msghdr *msg, size_t len, int noblock,
1880                              int flags, int *addr_len)
1881 {
1882         struct sctp_ulpevent *event = NULL;
1883         struct sctp_sock *sp = sctp_sk(sk);
1884         struct sk_buff *skb;
1885         int copied;
1886         int err = 0;
1887         int skb_len;
1888
1889         SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1890                           "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1891                           "len", len, "knoblauch", noblock,
1892                           "flags", flags, "addr_len", addr_len);
1893
1894         sctp_lock_sock(sk);
1895
1896         if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1897                 err = -ENOTCONN;
1898                 goto out;
1899         }
1900
1901         skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1902         if (!skb)
1903                 goto out;
1904
1905         /* Get the total length of the skb including any skb's in the
1906          * frag_list.
1907          */
1908         skb_len = skb->len;
1909
1910         copied = skb_len;
1911         if (copied > len)
1912                 copied = len;
1913
1914         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1915
1916         event = sctp_skb2event(skb);
1917
1918         if (err)
1919                 goto out_free;
1920
1921         sock_recv_timestamp(msg, sk, skb);
1922         if (sctp_ulpevent_is_notification(event)) {
1923                 msg->msg_flags |= MSG_NOTIFICATION;
1924                 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1925         } else {
1926                 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1927         }
1928
1929         /* Check if we allow SCTP_SNDRCVINFO. */
1930         if (sp->subscribe.sctp_data_io_event)
1931                 sctp_ulpevent_read_sndrcvinfo(event, msg);
1932 #if 0
1933         /* FIXME: we should be calling IP/IPv6 layers.  */
1934         if (sk->sk_protinfo.af_inet.cmsg_flags)
1935                 ip_cmsg_recv(msg, skb);
1936 #endif
1937
1938         err = copied;
1939
1940         /* If skb's length exceeds the user's buffer, update the skb and
1941          * push it back to the receive_queue so that the next call to
1942          * recvmsg() will return the remaining data. Don't set MSG_EOR.
1943          */
1944         if (skb_len > copied) {
1945                 msg->msg_flags &= ~MSG_EOR;
1946                 if (flags & MSG_PEEK)
1947                         goto out_free;
1948                 sctp_skb_pull(skb, copied);
1949                 skb_queue_head(&sk->sk_receive_queue, skb);
1950
1951                 /* When only partial message is copied to the user, increase
1952                  * rwnd by that amount. If all the data in the skb is read,
1953                  * rwnd is updated when the event is freed.
1954                  */
1955                 if (!sctp_ulpevent_is_notification(event))
1956                         sctp_assoc_rwnd_increase(event->asoc, copied);
1957                 goto out;
1958         } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1959                    (event->msg_flags & MSG_EOR))
1960                 msg->msg_flags |= MSG_EOR;
1961         else
1962                 msg->msg_flags &= ~MSG_EOR;
1963
1964 out_free:
1965         if (flags & MSG_PEEK) {
1966                 /* Release the skb reference acquired after peeking the skb in
1967                  * sctp_skb_recv_datagram().
1968                  */
1969                 kfree_skb(skb);
1970         } else {
1971                 /* Free the event which includes releasing the reference to
1972                  * the owner of the skb, freeing the skb and updating the
1973                  * rwnd.
1974                  */
1975                 sctp_ulpevent_free(event);
1976         }
1977 out:
1978         sctp_release_sock(sk);
1979         return err;
1980 }
1981
1982 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1983  *
1984  * This option is a on/off flag.  If enabled no SCTP message
1985  * fragmentation will be performed.  Instead if a message being sent
1986  * exceeds the current PMTU size, the message will NOT be sent and
1987  * instead a error will be indicated to the user.
1988  */
1989 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1990                                             char __user *optval, int optlen)
1991 {
1992         int val;
1993
1994         if (optlen < sizeof(int))
1995                 return -EINVAL;
1996
1997         if (get_user(val, (int __user *)optval))
1998                 return -EFAULT;
1999
2000         sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2001
2002         return 0;
2003 }
2004
2005 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2006                                         int optlen)
2007 {
2008         if (optlen > sizeof(struct sctp_event_subscribe))
2009                 return -EINVAL;
2010         if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2011                 return -EFAULT;
2012         return 0;
2013 }
2014
2015 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2016  *
2017  * This socket option is applicable to the UDP-style socket only.  When
2018  * set it will cause associations that are idle for more than the
2019  * specified number of seconds to automatically close.  An association
2020  * being idle is defined an association that has NOT sent or received
2021  * user data.  The special value of '0' indicates that no automatic
2022  * close of any associations should be performed.  The option expects an
2023  * integer defining the number of seconds of idle time before an
2024  * association is closed.
2025  */
2026 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2027                                             int optlen)
2028 {
2029         struct sctp_sock *sp = sctp_sk(sk);
2030
2031         /* Applicable to UDP-style socket only */
2032         if (sctp_style(sk, TCP))
2033                 return -EOPNOTSUPP;
2034         if (optlen != sizeof(int))
2035                 return -EINVAL;
2036         if (copy_from_user(&sp->autoclose, optval, optlen))
2037                 return -EFAULT;
2038
2039         return 0;
2040 }
2041
2042 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2043  *
2044  * Applications can enable or disable heartbeats for any peer address of
2045  * an association, modify an address's heartbeat interval, force a
2046  * heartbeat to be sent immediately, and adjust the address's maximum
2047  * number of retransmissions sent before an address is considered
2048  * unreachable.  The following structure is used to access and modify an
2049  * address's parameters:
2050  *
2051  *  struct sctp_paddrparams {
2052  *     sctp_assoc_t            spp_assoc_id;
2053  *     struct sockaddr_storage spp_address;
2054  *     uint32_t                spp_hbinterval;
2055  *     uint16_t                spp_pathmaxrxt;
2056  *     uint32_t                spp_pathmtu;
2057  *     uint32_t                spp_sackdelay;
2058  *     uint32_t                spp_flags;
2059  * };
2060  *
2061  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
2062  *                     application, and identifies the association for
2063  *                     this query.
2064  *   spp_address     - This specifies which address is of interest.
2065  *   spp_hbinterval  - This contains the value of the heartbeat interval,
2066  *                     in milliseconds.  If a  value of zero
2067  *                     is present in this field then no changes are to
2068  *                     be made to this parameter.
2069  *   spp_pathmaxrxt  - This contains the maximum number of
2070  *                     retransmissions before this address shall be
2071  *                     considered unreachable. If a  value of zero
2072  *                     is present in this field then no changes are to
2073  *                     be made to this parameter.
2074  *   spp_pathmtu     - When Path MTU discovery is disabled the value
2075  *                     specified here will be the "fixed" path mtu.
2076  *                     Note that if the spp_address field is empty
2077  *                     then all associations on this address will
2078  *                     have this fixed path mtu set upon them.
2079  *
2080  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
2081  *                     the number of milliseconds that sacks will be delayed
2082  *                     for. This value will apply to all addresses of an
2083  *                     association if the spp_address field is empty. Note
2084  *                     also, that if delayed sack is enabled and this
2085  *                     value is set to 0, no change is made to the last
2086  *                     recorded delayed sack timer value.
2087  *
2088  *   spp_flags       - These flags are used to control various features
2089  *                     on an association. The flag field may contain
2090  *                     zero or more of the following options.
2091  *
2092  *                     SPP_HB_ENABLE  - Enable heartbeats on the
2093  *                     specified address. Note that if the address
2094  *                     field is empty all addresses for the association
2095  *                     have heartbeats enabled upon them.
2096  *
2097  *                     SPP_HB_DISABLE - Disable heartbeats on the
2098  *                     speicifed address. Note that if the address
2099  *                     field is empty all addresses for the association
2100  *                     will have their heartbeats disabled. Note also
2101  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
2102  *                     mutually exclusive, only one of these two should
2103  *                     be specified. Enabling both fields will have
2104  *                     undetermined results.
2105  *
2106  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
2107  *                     to be made immediately.
2108  *
2109  *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
2110  *                     heartbeat delayis to be set to the value of 0
2111  *                     milliseconds.
2112  *
2113  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
2114  *                     discovery upon the specified address. Note that
2115  *                     if the address feild is empty then all addresses
2116  *                     on the association are effected.
2117  *
2118  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2119  *                     discovery upon the specified address. Note that
2120  *                     if the address feild is empty then all addresses
2121  *                     on the association are effected. Not also that
2122  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2123  *                     exclusive. Enabling both will have undetermined
2124  *                     results.
2125  *
2126  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2127  *                     on delayed sack. The time specified in spp_sackdelay
2128  *                     is used to specify the sack delay for this address. Note
2129  *                     that if spp_address is empty then all addresses will
2130  *                     enable delayed sack and take on the sack delay
2131  *                     value specified in spp_sackdelay.
2132  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2133  *                     off delayed sack. If the spp_address field is blank then
2134  *                     delayed sack is disabled for the entire association. Note
2135  *                     also that this field is mutually exclusive to
2136  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2137  *                     results.
2138  */
2139 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2140                                        struct sctp_transport   *trans,
2141                                        struct sctp_association *asoc,
2142                                        struct sctp_sock        *sp,
2143                                        int                      hb_change,
2144                                        int                      pmtud_change,
2145                                        int                      sackdelay_change)
2146 {
2147         int error;
2148
2149         if (params->spp_flags & SPP_HB_DEMAND && trans) {
2150                 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2151                 if (error)
2152                         return error;
2153         }
2154
2155         /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2156          * this field is ignored.  Note also that a value of zero indicates
2157          * the current setting should be left unchanged.
2158          */
2159         if (params->spp_flags & SPP_HB_ENABLE) {
2160
2161                 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2162                  * set.  This lets us use 0 value when this flag
2163                  * is set.
2164                  */
2165                 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2166                         params->spp_hbinterval = 0;
2167
2168                 if (params->spp_hbinterval ||
2169                     (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2170                         if (trans) {
2171                                 trans->hbinterval =
2172                                     msecs_to_jiffies(params->spp_hbinterval);
2173                         } else if (asoc) {
2174                                 asoc->hbinterval =
2175                                     msecs_to_jiffies(params->spp_hbinterval);
2176                         } else {
2177                                 sp->hbinterval = params->spp_hbinterval;
2178                         }
2179                 }
2180         }
2181
2182         if (hb_change) {
2183                 if (trans) {
2184                         trans->param_flags =
2185                                 (trans->param_flags & ~SPP_HB) | hb_change;
2186                 } else if (asoc) {
2187                         asoc->param_flags =
2188                                 (asoc->param_flags & ~SPP_HB) | hb_change;
2189                 } else {
2190                         sp->param_flags =
2191                                 (sp->param_flags & ~SPP_HB) | hb_change;
2192                 }
2193         }
2194
2195         /* When Path MTU discovery is disabled the value specified here will
2196          * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2197          * include the flag SPP_PMTUD_DISABLE for this field to have any
2198          * effect).
2199          */
2200         if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2201                 if (trans) {
2202                         trans->pathmtu = params->spp_pathmtu;
2203                         sctp_assoc_sync_pmtu(asoc);
2204                 } else if (asoc) {
2205                         asoc->pathmtu = params->spp_pathmtu;
2206                         sctp_frag_point(sp, params->spp_pathmtu);
2207                 } else {
2208                         sp->pathmtu = params->spp_pathmtu;
2209                 }
2210         }
2211
2212         if (pmtud_change) {
2213                 if (trans) {
2214                         int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2215                                 (params->spp_flags & SPP_PMTUD_ENABLE);
2216                         trans->param_flags =
2217                                 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2218                         if (update) {
2219                                 sctp_transport_pmtu(trans);
2220                                 sctp_assoc_sync_pmtu(asoc);
2221                         }
2222                 } else if (asoc) {
2223                         asoc->param_flags =
2224                                 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2225                 } else {
2226                         sp->param_flags =
2227                                 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2228                 }
2229         }
2230
2231         /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2232          * value of this field is ignored.  Note also that a value of zero
2233          * indicates the current setting should be left unchanged.
2234          */
2235         if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2236                 if (trans) {
2237                         trans->sackdelay =
2238                                 msecs_to_jiffies(params->spp_sackdelay);
2239                 } else if (asoc) {
2240                         asoc->sackdelay =
2241                                 msecs_to_jiffies(params->spp_sackdelay);
2242                 } else {
2243                         sp->sackdelay = params->spp_sackdelay;
2244                 }
2245         }
2246
2247         if (sackdelay_change) {
2248                 if (trans) {
2249                         trans->param_flags =
2250                                 (trans->param_flags & ~SPP_SACKDELAY) |
2251                                 sackdelay_change;
2252                 } else if (asoc) {
2253                         asoc->param_flags =
2254                                 (asoc->param_flags & ~SPP_SACKDELAY) |
2255                                 sackdelay_change;
2256                 } else {
2257                         sp->param_flags =
2258                                 (sp->param_flags & ~SPP_SACKDELAY) |
2259                                 sackdelay_change;
2260                 }
2261         }
2262
2263         /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2264          * of this field is ignored.  Note also that a value of zero
2265          * indicates the current setting should be left unchanged.
2266          */
2267         if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2268                 if (trans) {
2269                         trans->pathmaxrxt = params->spp_pathmaxrxt;
2270                 } else if (asoc) {
2271                         asoc->pathmaxrxt = params->spp_pathmaxrxt;
2272                 } else {
2273                         sp->pathmaxrxt = params->spp_pathmaxrxt;
2274                 }
2275         }
2276
2277         return 0;
2278 }
2279
2280 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2281                                             char __user *optval, int optlen)
2282 {
2283         struct sctp_paddrparams  params;
2284         struct sctp_transport   *trans = NULL;
2285         struct sctp_association *asoc = NULL;
2286         struct sctp_sock        *sp = sctp_sk(sk);
2287         int error;
2288         int hb_change, pmtud_change, sackdelay_change;
2289
2290         if (optlen != sizeof(struct sctp_paddrparams))
2291                 return - EINVAL;
2292
2293         if (copy_from_user(&params, optval, optlen))
2294                 return -EFAULT;
2295
2296         /* Validate flags and value parameters. */
2297         hb_change        = params.spp_flags & SPP_HB;
2298         pmtud_change     = params.spp_flags & SPP_PMTUD;
2299         sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2300
2301         if (hb_change        == SPP_HB ||
2302             pmtud_change     == SPP_PMTUD ||
2303             sackdelay_change == SPP_SACKDELAY ||
2304             params.spp_sackdelay > 500 ||
2305             (params.spp_pathmtu
2306             && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2307                 return -EINVAL;
2308
2309         /* If an address other than INADDR_ANY is specified, and
2310          * no transport is found, then the request is invalid.
2311          */
2312         if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2313                 trans = sctp_addr_id2transport(sk, &params.spp_address,
2314                                                params.spp_assoc_id);
2315                 if (!trans)
2316                         return -EINVAL;
2317         }
2318
2319         /* Get association, if assoc_id != 0 and the socket is a one
2320          * to many style socket, and an association was not found, then
2321          * the id was invalid.
2322          */
2323         asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2324         if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2325                 return -EINVAL;
2326
2327         /* Heartbeat demand can only be sent on a transport or
2328          * association, but not a socket.
2329          */
2330         if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2331                 return -EINVAL;
2332
2333         /* Process parameters. */
2334         error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2335                                             hb_change, pmtud_change,
2336                                             sackdelay_change);
2337
2338         if (error)
2339                 return error;
2340
2341         /* If changes are for association, also apply parameters to each
2342          * transport.
2343          */
2344         if (!trans && asoc) {
2345                 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2346                                 transports) {
2347                         sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2348                                                     hb_change, pmtud_change,
2349                                                     sackdelay_change);
2350                 }
2351         }
2352
2353         return 0;
2354 }
2355
2356 /*
2357  * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
2358  *
2359  * This option will effect the way delayed acks are performed.  This
2360  * option allows you to get or set the delayed ack time, in
2361  * milliseconds.  It also allows changing the delayed ack frequency.
2362  * Changing the frequency to 1 disables the delayed sack algorithm.  If
2363  * the assoc_id is 0, then this sets or gets the endpoints default
2364  * values.  If the assoc_id field is non-zero, then the set or get
2365  * effects the specified association for the one to many model (the
2366  * assoc_id field is ignored by the one to one model).  Note that if
2367  * sack_delay or sack_freq are 0 when setting this option, then the
2368  * current values will remain unchanged.
2369  *
2370  * struct sctp_sack_info {
2371  *     sctp_assoc_t            sack_assoc_id;
2372  *     uint32_t                sack_delay;
2373  *     uint32_t                sack_freq;
2374  * };
2375  *
2376  * sack_assoc_id -  This parameter, indicates which association the user
2377  *    is performing an action upon.  Note that if this field's value is
2378  *    zero then the endpoints default value is changed (effecting future
2379  *    associations only).
2380  *
2381  * sack_delay -  This parameter contains the number of milliseconds that
2382  *    the user is requesting the delayed ACK timer be set to.  Note that
2383  *    this value is defined in the standard to be between 200 and 500
2384  *    milliseconds.
2385  *
2386  * sack_freq -  This parameter contains the number of packets that must
2387  *    be received before a sack is sent without waiting for the delay
2388  *    timer to expire.  The default value for this is 2, setting this
2389  *    value to 1 will disable the delayed sack algorithm.
2390  */
2391
2392 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2393                                             char __user *optval, int optlen)
2394 {
2395         struct sctp_sack_info    params;
2396         struct sctp_transport   *trans = NULL;
2397         struct sctp_association *asoc = NULL;
2398         struct sctp_sock        *sp = sctp_sk(sk);
2399
2400         if (optlen == sizeof(struct sctp_sack_info)) {
2401                 if (copy_from_user(&params, optval, optlen))
2402                         return -EFAULT;
2403
2404                 if (params.sack_delay == 0 && params.sack_freq == 0)
2405                         return 0;
2406         } else if (optlen == sizeof(struct sctp_assoc_value)) {
2407                 printk(KERN_WARNING "SCTP: Use of struct sctp_sack_info "
2408                        "in delayed_ack socket option deprecated\n");
2409                 printk(KERN_WARNING "SCTP: struct sctp_sack_info instead\n");
2410                 if (copy_from_user(&params, optval, optlen))
2411                         return -EFAULT;
2412
2413                 if (params.sack_delay == 0)
2414                         params.sack_freq = 1;
2415                 else
2416                         params.sack_freq = 0;
2417         } else
2418                 return - EINVAL;
2419
2420         /* Validate value parameter. */
2421         if (params.sack_delay > 500)
2422                 return -EINVAL;
2423
2424         /* Get association, if sack_assoc_id != 0 and the socket is a one
2425          * to many style socket, and an association was not found, then
2426          * the id was invalid.
2427          */
2428         asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2429         if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2430                 return -EINVAL;
2431
2432         if (params.sack_delay) {
2433                 if (asoc) {
2434                         asoc->sackdelay =
2435                                 msecs_to_jiffies(params.sack_delay);
2436                         asoc->param_flags =
2437                                 (asoc->param_flags & ~SPP_SACKDELAY) |
2438                                 SPP_SACKDELAY_ENABLE;
2439                 } else {
2440                         sp->sackdelay = params.sack_delay;
2441                         sp->param_flags =
2442                                 (sp->param_flags & ~SPP_SACKDELAY) |
2443                                 SPP_SACKDELAY_ENABLE;
2444                 }
2445         }
2446
2447         if (params.sack_freq == 1) {
2448                 if (asoc) {
2449                         asoc->param_flags =
2450                                 (asoc->param_flags & ~SPP_SACKDELAY) |
2451                                 SPP_SACKDELAY_DISABLE;
2452                 } else {
2453                         sp->param_flags =
2454                                 (sp->param_flags & ~SPP_SACKDELAY) |
2455                                 SPP_SACKDELAY_DISABLE;
2456                 }
2457         } else if (params.sack_freq > 1) {
2458                 if (asoc) {
2459                         asoc->sackfreq = params.sack_freq;
2460                         asoc->param_flags =
2461                                 (asoc->param_flags & ~SPP_SACKDELAY) |
2462                                 SPP_SACKDELAY_ENABLE;
2463                 } else {
2464                         sp->sackfreq = params.sack_freq;
2465                         sp->param_flags =
2466                                 (sp->param_flags & ~SPP_SACKDELAY) |
2467                                 SPP_SACKDELAY_ENABLE;
2468                 }
2469         }
2470
2471         /* If change is for association, also apply to each transport. */
2472         if (asoc) {
2473                 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2474                                 transports) {
2475                         if (params.sack_delay) {
2476                                 trans->sackdelay =
2477                                         msecs_to_jiffies(params.sack_delay);
2478                                 trans->param_flags =
2479                                         (trans->param_flags & ~SPP_SACKDELAY) |
2480                                         SPP_SACKDELAY_ENABLE;
2481                         }
2482                         if (params.sack_freq == 1) {
2483                                 trans->param_flags =
2484                                         (trans->param_flags & ~SPP_SACKDELAY) |
2485                                         SPP_SACKDELAY_DISABLE;
2486                         } else if (params.sack_freq > 1) {
2487                                 trans->sackfreq = params.sack_freq;
2488                                 trans->param_flags =
2489                                         (trans->param_flags & ~SPP_SACKDELAY) |
2490                                         SPP_SACKDELAY_ENABLE;
2491                         }
2492                 }
2493         }
2494
2495         return 0;
2496 }
2497
2498 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2499  *
2500  * Applications can specify protocol parameters for the default association
2501  * initialization.  The option name argument to setsockopt() and getsockopt()
2502  * is SCTP_INITMSG.
2503  *
2504  * Setting initialization parameters is effective only on an unconnected
2505  * socket (for UDP-style sockets only future associations are effected
2506  * by the change).  With TCP-style sockets, this option is inherited by
2507  * sockets derived from a listener socket.
2508  */
2509 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2510 {
2511         struct sctp_initmsg sinit;
2512         struct sctp_sock *sp = sctp_sk(sk);
2513
2514         if (optlen != sizeof(struct sctp_initmsg))
2515                 return -EINVAL;
2516         if (copy_from_user(&sinit, optval, optlen))
2517                 return -EFAULT;
2518
2519         if (sinit.sinit_num_ostreams)
2520                 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2521         if (sinit.sinit_max_instreams)
2522                 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2523         if (sinit.sinit_max_attempts)
2524                 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2525         if (sinit.sinit_max_init_timeo)
2526                 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2527
2528         return 0;
2529 }
2530
2531 /*
2532  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2533  *
2534  *   Applications that wish to use the sendto() system call may wish to
2535  *   specify a default set of parameters that would normally be supplied
2536  *   through the inclusion of ancillary data.  This socket option allows
2537  *   such an application to set the default sctp_sndrcvinfo structure.
2538  *   The application that wishes to use this socket option simply passes
2539  *   in to this call the sctp_sndrcvinfo structure defined in Section
2540  *   5.2.2) The input parameters accepted by this call include
2541  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2542  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2543  *   to this call if the caller is using the UDP model.
2544  */
2545 static int sctp_setsockopt_default_send_param(struct sock *sk,
2546                                                 char __user *optval, int optlen)
2547 {
2548         struct sctp_sndrcvinfo info;
2549         struct sctp_association *asoc;
2550         struct sctp_sock *sp = sctp_sk(sk);
2551
2552         if (optlen != sizeof(struct sctp_sndrcvinfo))
2553                 return -EINVAL;
2554         if (copy_from_user(&info, optval, optlen))
2555                 return -EFAULT;
2556
2557         asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2558         if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2559                 return -EINVAL;
2560
2561         if (asoc) {
2562                 asoc->default_stream = info.sinfo_stream;
2563                 asoc->default_flags = info.sinfo_flags;
2564                 asoc->default_ppid = info.sinfo_ppid;
2565                 asoc->default_context = info.sinfo_context;
2566                 asoc->default_timetolive = info.sinfo_timetolive;
2567         } else {
2568                 sp->default_stream = info.sinfo_stream;
2569                 sp->default_flags = info.sinfo_flags;
2570                 sp->default_ppid = info.sinfo_ppid;
2571                 sp->default_context = info.sinfo_context;
2572                 sp->default_timetolive = info.sinfo_timetolive;
2573         }
2574
2575         return 0;
2576 }
2577
2578 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2579  *
2580  * Requests that the local SCTP stack use the enclosed peer address as
2581  * the association primary.  The enclosed address must be one of the
2582  * association peer's addresses.
2583  */
2584 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2585                                         int optlen)
2586 {
2587         struct sctp_prim prim;
2588         struct sctp_transport *trans;
2589
2590         if (optlen != sizeof(struct sctp_prim))
2591                 return -EINVAL;
2592
2593         if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2594                 return -EFAULT;
2595
2596         trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2597         if (!trans)
2598                 return -EINVAL;
2599
2600         sctp_assoc_set_primary(trans->asoc, trans);
2601
2602         return 0;
2603 }
2604
2605 /*
2606  * 7.1.5 SCTP_NODELAY
2607  *
2608  * Turn on/off any Nagle-like algorithm.  This means that packets are
2609  * generally sent as soon as possible and no unnecessary delays are
2610  * introduced, at the cost of more packets in the network.  Expects an
2611  *  integer boolean flag.
2612  */
2613 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2614                                         int optlen)
2615 {
2616         int val;
2617
2618         if (optlen < sizeof(int))
2619                 return -EINVAL;
2620         if (get_user(val, (int __user *)optval))
2621                 return -EFAULT;
2622
2623         sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2624         return 0;
2625 }
2626
2627 /*
2628  *
2629  * 7.1.1 SCTP_RTOINFO
2630  *
2631  * The protocol parameters used to initialize and bound retransmission
2632  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2633  * and modify these parameters.
2634  * All parameters are time values, in milliseconds.  A value of 0, when
2635  * modifying the parameters, indicates that the current value should not
2636  * be changed.
2637  *
2638  */
2639 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2640         struct sctp_rtoinfo rtoinfo;
2641         struct sctp_association *asoc;
2642
2643         if (optlen != sizeof (struct sctp_rtoinfo))
2644                 return -EINVAL;
2645
2646         if (copy_from_user(&rtoinfo, optval, optlen))
2647                 return -EFAULT;
2648
2649         asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2650
2651         /* Set the values to the specific association */
2652         if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2653                 return -EINVAL;
2654
2655         if (asoc) {
2656                 if (rtoinfo.srto_initial != 0)
2657                         asoc->rto_initial =
2658                                 msecs_to_jiffies(rtoinfo.srto_initial);
2659                 if (rtoinfo.srto_max != 0)
2660                         asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2661                 if (rtoinfo.srto_min != 0)
2662                         asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2663         } else {
2664                 /* If there is no association or the association-id = 0
2665                  * set the values to the endpoint.
2666                  */
2667                 struct sctp_sock *sp = sctp_sk(sk);
2668
2669                 if (rtoinfo.srto_initial != 0)
2670                         sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2671                 if (rtoinfo.srto_max != 0)
2672                         sp->rtoinfo.srto_max = rtoinfo.srto_max;
2673                 if (rtoinfo.srto_min != 0)
2674                         sp->rtoinfo.srto_min = rtoinfo.srto_min;
2675         }
2676
2677         return 0;
2678 }
2679
2680 /*
2681  *
2682  * 7.1.2 SCTP_ASSOCINFO
2683  *
2684  * This option is used to tune the maximum retransmission attempts
2685  * of the association.
2686  * Returns an error if the new association retransmission value is
2687  * greater than the sum of the retransmission value  of the peer.
2688  * See [SCTP] for more information.
2689  *
2690  */
2691 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2692 {
2693
2694         struct sctp_assocparams assocparams;
2695         struct sctp_association *asoc;
2696
2697         if (optlen != sizeof(struct sctp_assocparams))
2698                 return -EINVAL;
2699         if (copy_from_user(&assocparams, optval, optlen))
2700                 return -EFAULT;
2701
2702         asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2703
2704         if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2705                 return -EINVAL;
2706
2707         /* Set the values to the specific association */
2708         if (asoc) {
2709                 if (assocparams.sasoc_asocmaxrxt != 0) {
2710                         __u32 path_sum = 0;
2711                         int   paths = 0;
2712                         struct sctp_transport *peer_addr;
2713
2714                         list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2715                                         transports) {
2716                                 path_sum += peer_addr->pathmaxrxt;
2717                                 paths++;
2718                         }
2719
2720                         /* Only validate asocmaxrxt if we have more then
2721                          * one path/transport.  We do this because path
2722                          * retransmissions are only counted when we have more
2723                          * then one path.
2724                          */
2725                         if (paths > 1 &&
2726                             assocparams.sasoc_asocmaxrxt > path_sum)
2727                                 return -EINVAL;
2728
2729                         asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2730                 }
2731
2732                 if (assocparams.sasoc_cookie_life != 0) {
2733                         asoc->cookie_life.tv_sec =
2734                                         assocparams.sasoc_cookie_life / 1000;
2735                         asoc->cookie_life.tv_usec =
2736                                         (assocparams.sasoc_cookie_life % 1000)
2737                                         * 1000;
2738                 }
2739         } else {
2740                 /* Set the values to the endpoint */
2741                 struct sctp_sock *sp = sctp_sk(sk);
2742
2743                 if (assocparams.sasoc_asocmaxrxt != 0)
2744                         sp->assocparams.sasoc_asocmaxrxt =
2745                                                 assocparams.sasoc_asocmaxrxt;
2746                 if (assocparams.sasoc_cookie_life != 0)
2747                         sp->assocparams.sasoc_cookie_life =
2748                                                 assocparams.sasoc_cookie_life;
2749         }
2750         return 0;
2751 }
2752
2753 /*
2754  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2755  *
2756  * This socket option is a boolean flag which turns on or off mapped V4
2757  * addresses.  If this option is turned on and the socket is type
2758  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2759  * If this option is turned off, then no mapping will be done of V4
2760  * addresses and a user will receive both PF_INET6 and PF_INET type
2761  * addresses on the socket.
2762  */
2763 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2764 {
2765         int val;
2766         struct sctp_sock *sp = sctp_sk(sk);
2767
2768         if (optlen < sizeof(int))
2769                 return -EINVAL;
2770         if (get_user(val, (int __user *)optval))
2771                 return -EFAULT;
2772         if (val)
2773                 sp->v4mapped = 1;
2774         else
2775                 sp->v4mapped = 0;
2776
2777         return 0;
2778 }
2779
2780 /*
2781  * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2782  *
2783  * This socket option specifies the maximum size to put in any outgoing
2784  * SCTP chunk.  If a message is larger than this size it will be
2785  * fragmented by SCTP into the specified size.  Note that the underlying
2786  * SCTP implementation may fragment into smaller sized chunks when the
2787  * PMTU of the underlying association is smaller than the value set by
2788  * the user.
2789  */
2790 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2791 {
2792         struct sctp_association *asoc;
2793         struct sctp_sock *sp = sctp_sk(sk);
2794         int val;
2795
2796         if (optlen < sizeof(int))
2797                 return -EINVAL;
2798         if (get_user(val, (int __user *)optval))
2799                 return -EFAULT;
2800         if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2801                 return -EINVAL;
2802         sp->user_frag = val;
2803
2804         /* Update the frag_point of the existing associations. */
2805         list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
2806                 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2807         }
2808
2809         return 0;
2810 }
2811
2812
2813 /*
2814  *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2815  *
2816  *   Requests that the peer mark the enclosed address as the association
2817  *   primary. The enclosed address must be one of the association's
2818  *   locally bound addresses. The following structure is used to make a
2819  *   set primary request:
2820  */
2821 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2822                                              int optlen)
2823 {
2824         struct sctp_sock        *sp;
2825         struct sctp_endpoint    *ep;
2826         struct sctp_association *asoc = NULL;
2827         struct sctp_setpeerprim prim;
2828         struct sctp_chunk       *chunk;
2829         int                     err;
2830
2831         sp = sctp_sk(sk);
2832         ep = sp->ep;
2833
2834         if (!sctp_addip_enable)
2835                 return -EPERM;
2836
2837         if (optlen != sizeof(struct sctp_setpeerprim))
2838                 return -EINVAL;
2839
2840         if (copy_from_user(&prim, optval, optlen))
2841                 return -EFAULT;
2842
2843         asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2844         if (!asoc)
2845                 return -EINVAL;
2846
2847         if (!asoc->peer.asconf_capable)
2848                 return -EPERM;
2849
2850         if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2851                 return -EPERM;
2852
2853         if (!sctp_state(asoc, ESTABLISHED))
2854                 return -ENOTCONN;
2855
2856         if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2857                 return -EADDRNOTAVAIL;
2858
2859         /* Create an ASCONF chunk with SET_PRIMARY parameter    */
2860         chunk = sctp_make_asconf_set_prim(asoc,
2861                                           (union sctp_addr *)&prim.sspp_addr);
2862         if (!chunk)
2863                 return -ENOMEM;
2864
2865         err = sctp_send_asconf(asoc, chunk);
2866
2867         SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2868
2869         return err;
2870 }
2871
2872 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2873                                           int optlen)
2874 {
2875         struct sctp_setadaptation adaptation;
2876
2877         if (optlen != sizeof(struct sctp_setadaptation))
2878                 return -EINVAL;
2879         if (copy_from_user(&adaptation, optval, optlen))
2880                 return -EFAULT;
2881
2882         sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2883
2884         return 0;
2885 }
2886
2887 /*
2888  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
2889  *
2890  * The context field in the sctp_sndrcvinfo structure is normally only
2891  * used when a failed message is retrieved holding the value that was
2892  * sent down on the actual send call.  This option allows the setting of
2893  * a default context on an association basis that will be received on
2894  * reading messages from the peer.  This is especially helpful in the
2895  * one-2-many model for an application to keep some reference to an
2896  * internal state machine that is processing messages on the
2897  * association.  Note that the setting of this value only effects
2898  * received messages from the peer and does not effect the value that is
2899  * saved with outbound messages.
2900  */
2901 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2902                                    int optlen)
2903 {
2904         struct sctp_assoc_value params;
2905         struct sctp_sock *sp;
2906         struct sctp_association *asoc;
2907
2908         if (optlen != sizeof(struct sctp_assoc_value))
2909                 return -EINVAL;
2910         if (copy_from_user(&params, optval, optlen))
2911                 return -EFAULT;
2912
2913         sp = sctp_sk(sk);
2914
2915         if (params.assoc_id != 0) {
2916                 asoc = sctp_id2assoc(sk, params.assoc_id);
2917                 if (!asoc)
2918                         return -EINVAL;
2919                 asoc->default_rcv_context = params.assoc_value;
2920         } else {
2921                 sp->default_rcv_context = params.assoc_value;
2922         }
2923
2924         return 0;
2925 }
2926
2927 /*
2928  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2929  *
2930  * This options will at a minimum specify if the implementation is doing
2931  * fragmented interleave.  Fragmented interleave, for a one to many
2932  * socket, is when subsequent calls to receive a message may return
2933  * parts of messages from different associations.  Some implementations
2934  * may allow you to turn this value on or off.  If so, when turned off,
2935  * no fragment interleave will occur (which will cause a head of line
2936  * blocking amongst multiple associations sharing the same one to many
2937  * socket).  When this option is turned on, then each receive call may
2938  * come from a different association (thus the user must receive data
2939  * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2940  * association each receive belongs to.
2941  *
2942  * This option takes a boolean value.  A non-zero value indicates that
2943  * fragmented interleave is on.  A value of zero indicates that
2944  * fragmented interleave is off.
2945  *
2946  * Note that it is important that an implementation that allows this
2947  * option to be turned on, have it off by default.  Otherwise an unaware
2948  * application using the one to many model may become confused and act
2949  * incorrectly.
2950  */
2951 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2952                                                char __user *optval,
2953                                                int optlen)
2954 {
2955         int val;
2956
2957         if (optlen != sizeof(int))
2958                 return -EINVAL;
2959         if (get_user(val, (int __user *)optval))
2960                 return -EFAULT;
2961
2962         sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2963
2964         return 0;
2965 }
2966
2967 /*
2968  * 7.1.25.  Set or Get the sctp partial delivery point
2969  *       (SCTP_PARTIAL_DELIVERY_POINT)
2970  * This option will set or get the SCTP partial delivery point.  This
2971  * point is the size of a message where the partial delivery API will be
2972  * invoked to help free up rwnd space for the peer.  Setting this to a
2973  * lower value will cause partial delivery's to happen more often.  The
2974  * calls argument is an integer that sets or gets the partial delivery
2975  * point.
2976  */
2977 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2978                                                   char __user *optval,
2979                                                   int optlen)
2980 {
2981         u32 val;
2982
2983         if (optlen != sizeof(u32))
2984                 return -EINVAL;
2985         if (get_user(val, (int __user *)optval))
2986                 return -EFAULT;
2987
2988         sctp_sk(sk)->pd_point = val;
2989
2990         return 0; /* is this the right error code? */
2991 }
2992
2993 /*
2994  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
2995  *
2996  * This option will allow a user to change the maximum burst of packets
2997  * that can be emitted by this association.  Note that the default value
2998  * is 4, and some implementations may restrict this setting so that it
2999  * can only be lowered.
3000  *
3001  * NOTE: This text doesn't seem right.  Do this on a socket basis with
3002  * future associations inheriting the socket value.
3003  */
3004 static int sctp_setsockopt_maxburst(struct sock *sk,
3005                                     char __user *optval,
3006                                     int optlen)
3007 {
3008         struct sctp_assoc_value params;
3009         struct sctp_sock *sp;
3010         struct sctp_association *asoc;
3011         int val;
3012         int assoc_id = 0;
3013
3014         if (optlen < sizeof(int))
3015                 return -EINVAL;
3016
3017         if (optlen == sizeof(int)) {
3018                 printk(KERN_WARNING
3019                    "SCTP: Use of int in max_burst socket option deprecated\n");
3020                 printk(KERN_WARNING
3021                    "SCTP: Use struct sctp_assoc_value instead\n");
3022                 if (copy_from_user(&val, optval, optlen))
3023                         return -EFAULT;
3024         } else if (optlen == sizeof(struct sctp_assoc_value)) {
3025                 if (copy_from_user(&params, optval, optlen))
3026                         return -EFAULT;
3027                 val = params.assoc_value;
3028                 assoc_id = params.assoc_id;
3029         } else
3030                 return -EINVAL;
3031
3032         sp = sctp_sk(sk);
3033
3034         if (assoc_id != 0) {
3035                 asoc = sctp_id2assoc(sk, assoc_id);
3036                 if (!asoc)
3037                         return -EINVAL;
3038                 asoc->max_burst = val;
3039         } else
3040                 sp->max_burst = val;
3041
3042         return 0;
3043 }
3044
3045 /*
3046  * 7.1.18.  Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3047  *
3048  * This set option adds a chunk type that the user is requesting to be
3049  * received only in an authenticated way.  Changes to the list of chunks
3050  * will only effect future associations on the socket.
3051  */
3052 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3053                                     char __user *optval,
3054                                     int optlen)
3055 {
3056         struct sctp_authchunk val;
3057
3058         if (!sctp_auth_enable)
3059                 return -EACCES;
3060
3061         if (optlen != sizeof(struct sctp_authchunk))
3062                 return -EINVAL;
3063         if (copy_from_user(&val, optval, optlen))
3064                 return -EFAULT;
3065
3066         switch (val.sauth_chunk) {
3067                 case SCTP_CID_INIT:
3068                 case SCTP_CID_INIT_ACK:
3069                 case SCTP_CID_SHUTDOWN_COMPLETE:
3070                 case SCTP_CID_AUTH:
3071                         return -EINVAL;
3072         }
3073
3074         /* add this chunk id to the endpoint */
3075         return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3076 }
3077
3078 /*
3079  * 7.1.19.  Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3080  *
3081  * This option gets or sets the list of HMAC algorithms that the local
3082  * endpoint requires the peer to use.
3083  */
3084 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3085                                     char __user *optval,
3086                                     int optlen)
3087 {
3088         struct sctp_hmacalgo *hmacs;
3089         u32 idents;
3090         int err;
3091
3092         if (!sctp_auth_enable)
3093                 return -EACCES;
3094
3095         if (optlen < sizeof(struct sctp_hmacalgo))
3096                 return -EINVAL;
3097
3098         hmacs = kmalloc(optlen, GFP_KERNEL);
3099         if (!hmacs)
3100                 return -ENOMEM;
3101
3102         if (copy_from_user(hmacs, optval, optlen)) {
3103                 err = -EFAULT;
3104                 goto out;
3105         }
3106
3107         idents = hmacs->shmac_num_idents;
3108         if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3109             (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3110                 err = -EINVAL;
3111                 goto out;
3112         }
3113
3114         err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3115 out:
3116         kfree(hmacs);
3117         return err;
3118 }
3119
3120 /*
3121  * 7.1.20.  Set a shared key (SCTP_AUTH_KEY)
3122  *
3123  * This option will set a shared secret key which is used to build an
3124  * association shared key.
3125  */
3126 static int sctp_setsockopt_auth_key(struct sock *sk,
3127                                     char __user *optval,
3128                                     int optlen)
3129 {
3130         struct sctp_authkey *authkey;
3131         struct sctp_association *asoc;
3132         int ret;
3133
3134         if (!sctp_auth_enable)
3135                 return -EACCES;
3136
3137         if (optlen <= sizeof(struct sctp_authkey))
3138                 return -EINVAL;
3139
3140         authkey = kmalloc(optlen, GFP_KERNEL);
3141         if (!authkey)
3142                 return -ENOMEM;
3143
3144         if (copy_from_user(authkey, optval, optlen)) {
3145                 ret = -EFAULT;
3146                 goto out;
3147         }
3148
3149         if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3150                 ret = -EINVAL;
3151                 goto out;
3152         }
3153
3154         asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3155         if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3156                 ret = -EINVAL;
3157                 goto out;
3158         }
3159
3160         ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3161 out:
3162         kfree(authkey);
3163         return ret;
3164 }
3165
3166 /*
3167  * 7.1.21.  Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3168  *
3169  * This option will get or set the active shared key to be used to build
3170  * the association shared key.
3171  */
3172 static int sctp_setsockopt_active_key(struct sock *sk,
3173                                         char __user *optval,
3174                                         int optlen)
3175 {
3176         struct sctp_authkeyid val;
3177         struct sctp_association *asoc;
3178
3179         if (!sctp_auth_enable)
3180                 return -EACCES;
3181
3182         if (optlen != sizeof(struct sctp_authkeyid))
3183                 return -EINVAL;
3184         if (copy_from_user(&val, optval, optlen))
3185                 return -EFAULT;
3186
3187         asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3188         if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3189                 return -EINVAL;
3190
3191         return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3192                                         val.scact_keynumber);
3193 }
3194
3195 /*
3196  * 7.1.22.  Delete a shared key (SCTP_AUTH_DELETE_KEY)
3197  *
3198  * This set option will delete a shared secret key from use.
3199  */
3200 static int sctp_setsockopt_del_key(struct sock *sk,
3201                                         char __user *optval,
3202                                         int optlen)
3203 {
3204         struct sctp_authkeyid val;
3205         struct sctp_association *asoc;
3206
3207         if (!sctp_auth_enable)
3208                 return -EACCES;
3209
3210         if (optlen != sizeof(struct sctp_authkeyid))
3211                 return -EINVAL;
3212         if (copy_from_user(&val, optval, optlen))
3213                 return -EFAULT;
3214
3215         asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3216         if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3217                 return -EINVAL;
3218
3219         return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3220                                     val.scact_keynumber);
3221
3222 }
3223
3224
3225 /* API 6.2 setsockopt(), getsockopt()
3226  *
3227  * Applications use setsockopt() and getsockopt() to set or retrieve
3228  * socket options.  Socket options are used to change the default
3229  * behavior of sockets calls.  They are described in Section 7.
3230  *
3231  * The syntax is:
3232  *
3233  *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
3234  *                    int __user *optlen);
3235  *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3236  *                    int optlen);
3237  *
3238  *   sd      - the socket descript.
3239  *   level   - set to IPPROTO_SCTP for all SCTP options.
3240  *   optname - the option name.
3241  *   optval  - the buffer to store the value of the option.
3242  *   optlen  - the size of the buffer.
3243  */
3244 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3245                                 char __user *optval, int optlen)
3246 {
3247         int retval = 0;
3248
3249         SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3250                           sk, optname);
3251
3252         /* I can hardly begin to describe how wrong this is.  This is
3253          * so broken as to be worse than useless.  The API draft
3254          * REALLY is NOT helpful here...  I am not convinced that the
3255          * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3256          * are at all well-founded.
3257          */
3258         if (level != SOL_SCTP) {
3259                 struct sctp_af *af = sctp_sk(sk)->pf->af;
3260                 retval = af->setsockopt(sk, level, optname, optval, optlen);
3261                 goto out_nounlock;
3262         }
3263
3264         sctp_lock_sock(sk);
3265
3266         switch (optname) {
3267         case SCTP_SOCKOPT_BINDX_ADD:
3268                 /* 'optlen' is the size of the addresses buffer. */
3269                 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3270                                                optlen, SCTP_BINDX_ADD_ADDR);
3271                 break;
3272
3273         case SCTP_SOCKOPT_BINDX_REM:
3274                 /* 'optlen' is the size of the addresses buffer. */
3275                 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3276                                                optlen, SCTP_BINDX_REM_ADDR);
3277                 break;
3278
3279         case SCTP_SOCKOPT_CONNECTX_OLD:
3280                 /* 'optlen' is the size of the addresses buffer. */
3281                 retval = sctp_setsockopt_connectx_old(sk,
3282                                             (struct sockaddr __user *)optval,
3283                                             optlen);
3284                 break;
3285
3286         case SCTP_SOCKOPT_CONNECTX:
3287                 /* 'optlen' is the size of the addresses buffer. */
3288                 retval = sctp_setsockopt_connectx(sk,
3289                                             (struct sockaddr __user *)optval,
3290                                             optlen);
3291                 break;
3292
3293         case SCTP_DISABLE_FRAGMENTS:
3294                 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3295                 break;
3296
3297         case SCTP_EVENTS:
3298                 retval = sctp_setsockopt_events(sk, optval, optlen);
3299                 break;
3300
3301         case SCTP_AUTOCLOSE:
3302                 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3303                 break;
3304
3305         case SCTP_PEER_ADDR_PARAMS:
3306                 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3307                 break;
3308
3309         case SCTP_DELAYED_ACK:
3310                 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3311                 break;
3312         case SCTP_PARTIAL_DELIVERY_POINT:
3313                 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3314                 break;
3315
3316         case SCTP_INITMSG:
3317                 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3318                 break;
3319         case SCTP_DEFAULT_SEND_PARAM:
3320                 retval = sctp_setsockopt_default_send_param(sk, optval,
3321                                                             optlen);
3322                 break;
3323         case SCTP_PRIMARY_ADDR:
3324                 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3325                 break;
3326         case SCTP_SET_PEER_PRIMARY_ADDR:
3327                 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3328                 break;
3329         case SCTP_NODELAY:
3330                 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3331                 break;
3332         case SCTP_RTOINFO:
3333                 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3334                 break;
3335         case SCTP_ASSOCINFO:
3336                 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3337                 break;
3338         case SCTP_I_WANT_MAPPED_V4_ADDR:
3339                 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3340                 break;
3341         case SCTP_MAXSEG:
3342                 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3343                 break;
3344         case SCTP_ADAPTATION_LAYER:
3345                 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3346                 break;
3347         case SCTP_CONTEXT:
3348                 retval = sctp_setsockopt_context(sk, optval, optlen);
3349                 break;
3350         case SCTP_FRAGMENT_INTERLEAVE:
3351                 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3352                 break;
3353         case SCTP_MAX_BURST:
3354                 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3355                 break;
3356         case SCTP_AUTH_CHUNK:
3357                 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3358                 break;
3359         case SCTP_HMAC_IDENT:
3360                 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3361                 break;
3362         case SCTP_AUTH_KEY:
3363                 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3364                 break;
3365         case SCTP_AUTH_ACTIVE_KEY:
3366                 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3367                 break;
3368         case SCTP_AUTH_DELETE_KEY:
3369                 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3370                 break;
3371         default:
3372                 retval = -ENOPROTOOPT;
3373                 break;
3374         }
3375
3376         sctp_release_sock(sk);
3377
3378 out_nounlock:
3379         return retval;
3380 }
3381
3382 /* API 3.1.6 connect() - UDP Style Syntax
3383  *
3384  * An application may use the connect() call in the UDP model to initiate an
3385  * association without sending data.
3386  *
3387  * The syntax is:
3388  *
3389  * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3390  *
3391  * sd: the socket descriptor to have a new association added to.
3392  *
3393  * nam: the address structure (either struct sockaddr_in or struct
3394  *    sockaddr_in6 defined in RFC2553 [7]).
3395  *
3396  * len: the size of the address.
3397  */
3398 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3399                              int addr_len)
3400 {
3401         int err = 0;
3402         struct sctp_af *af;
3403
3404         sctp_lock_sock(sk);
3405
3406         SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3407                           __func__, sk, addr, addr_len);
3408
3409         /* Validate addr_len before calling common connect/connectx routine. */
3410         af = sctp_get_af_specific(addr->sa_family);
3411         if (!af || addr_len < af->sockaddr_len) {
3412                 err = -EINVAL;
3413         } else {
3414                 /* Pass correct addr len to common routine (so it knows there
3415                  * is only one address being passed.
3416                  */
3417                 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3418         }
3419
3420         sctp_release_sock(sk);
3421         return err;
3422 }
3423
3424 /* FIXME: Write comments. */
3425 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3426 {
3427         return -EOPNOTSUPP; /* STUB */
3428 }
3429
3430 /* 4.1.4 accept() - TCP Style Syntax
3431  *
3432  * Applications use accept() call to remove an established SCTP
3433  * association from the accept queue of the endpoint.  A new socket
3434  * descriptor will be returned from accept() to represent the newly
3435  * formed association.
3436  */
3437 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3438 {
3439         struct sctp_sock *sp;
3440         struct sctp_endpoint *ep;
3441         struct sock *newsk = NULL;
3442         struct sctp_association *asoc;
3443         long timeo;
3444         int error = 0;
3445
3446         sctp_lock_sock(sk);
3447
3448         sp = sctp_sk(sk);
3449         ep = sp->ep;
3450
3451         if (!sctp_style(sk, TCP)) {
3452                 error = -EOPNOTSUPP;
3453                 goto out;
3454         }
3455
3456         if (!sctp_sstate(sk, LISTENING)) {
3457                 error = -EINVAL;
3458                 goto out;
3459         }
3460
3461         timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3462
3463         error = sctp_wait_for_accept(sk, timeo);
3464         if (error)
3465                 goto out;
3466
3467         /* We treat the list of associations on the endpoint as the accept
3468          * queue and pick the first association on the list.
3469          */
3470         asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3471
3472         newsk = sp->pf->create_accept_sk(sk, asoc);
3473         if (!newsk) {
3474                 error = -ENOMEM;
3475                 goto out;
3476         }
3477
3478         /* Populate the fields of the newsk from the oldsk and migrate the
3479          * asoc to the newsk.
3480          */
3481         sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3482
3483 out:
3484         sctp_release_sock(sk);
3485         *err = error;
3486         return newsk;
3487 }
3488
3489 /* The SCTP ioctl handler. */
3490 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3491 {
3492         return -ENOIOCTLCMD;
3493 }
3494
3495 /* This is the function which gets called during socket creation to
3496  * initialized the SCTP-specific portion of the sock.
3497  * The sock structure should already be zero-filled memory.
3498  */
3499 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3500 {
3501         struct sctp_endpoint *ep;
3502         struct sctp_sock *sp;
3503
3504         SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3505
3506         sp = sctp_sk(sk);
3507
3508         /* Initialize the SCTP per socket area.  */
3509         switch (sk->sk_type) {
3510         case SOCK_SEQPACKET:
3511                 sp->type = SCTP_SOCKET_UDP;
3512                 break;
3513         case SOCK_STREAM:
3514                 sp->type = SCTP_SOCKET_TCP;
3515                 break;
3516         default:
3517                 return -ESOCKTNOSUPPORT;
3518         }
3519
3520         /* Initialize default send parameters. These parameters can be
3521          * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3522          */
3523         sp->default_stream = 0;
3524         sp->default_ppid = 0;
3525         sp->default_flags = 0;
3526         sp->default_context = 0;
3527         sp->default_timetolive = 0;
3528
3529         sp->default_rcv_context = 0;
3530         sp->max_burst = sctp_max_burst;
3531
3532         /* Initialize default setup parameters. These parameters
3533          * can be modified with the SCTP_INITMSG socket option or
3534          * overridden by the SCTP_INIT CMSG.
3535          */
3536         sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
3537         sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
3538         sp->initmsg.sinit_max_attempts   = sctp_max_retrans_init;
3539         sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3540
3541         /* Initialize default RTO related parameters.  These parameters can
3542          * be modified for with the SCTP_RTOINFO socket option.
3543          */
3544         sp->rtoinfo.srto_initial = sctp_rto_initial;
3545         sp->rtoinfo.srto_max     = sctp_rto_max;
3546         sp->rtoinfo.srto_min     = sctp_rto_min;
3547
3548         /* Initialize default association related parameters. These parameters
3549          * can be modified with the SCTP_ASSOCINFO socket option.
3550          */
3551         sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3552         sp->assocparams.sasoc_number_peer_destinations = 0;
3553         sp->assocparams.sasoc_peer_rwnd = 0;
3554         sp->assocparams.sasoc_local_rwnd = 0;
3555         sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3556
3557         /* Initialize default event subscriptions. By default, all the
3558          * options are off.
3559          */
3560         memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3561
3562         /* Default Peer Address Parameters.  These defaults can
3563          * be modified via SCTP_PEER_ADDR_PARAMS
3564          */
3565         sp->hbinterval  = sctp_hb_interval;
3566         sp->pathmaxrxt  = sctp_max_retrans_path;
3567         sp->pathmtu     = 0; // allow default discovery
3568         sp->sackdelay   = sctp_sack_timeout;
3569         sp->sackfreq    = 2;
3570         sp->param_flags = SPP_HB_ENABLE |
3571                           SPP_PMTUD_ENABLE |
3572                           SPP_SACKDELAY_ENABLE;
3573
3574         /* If enabled no SCTP message fragmentation will be performed.
3575          * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3576          */
3577         sp->disable_fragments = 0;
3578
3579         /* Enable Nagle algorithm by default.  */
3580         sp->nodelay           = 0;
3581
3582         /* Enable by default. */
3583         sp->v4mapped          = 1;
3584
3585         /* Auto-close idle associations after the configured
3586          * number of seconds.  A value of 0 disables this
3587          * feature.  Configure through the SCTP_AUTOCLOSE socket option,
3588          * for UDP-style sockets only.
3589          */
3590         sp->autoclose         = 0;
3591
3592         /* User specified fragmentation limit. */
3593         sp->user_frag         = 0;
3594
3595         sp->adaptation_ind = 0;
3596
3597         sp->pf = sctp_get_pf_specific(sk->sk_family);
3598
3599         /* Control variables for partial data delivery. */
3600         atomic_set(&sp->pd_mode, 0);
3601         skb_queue_head_init(&sp->pd_lobby);
3602         sp->frag_interleave = 0;
3603
3604         /* Create a per socket endpoint structure.  Even if we
3605          * change the data structure relationships, this may still
3606          * be useful for storing pre-connect address information.
3607          */
3608         ep = sctp_endpoint_new(sk, GFP_KERNEL);
3609         if (!ep)
3610                 return -ENOMEM;
3611
3612         sp->ep = ep;
3613         sp->hmac = NULL;
3614
3615         SCTP_DBG_OBJCNT_INC(sock);
3616         atomic_inc(&sctp_sockets_allocated);
3617         return 0;
3618 }
3619
3620 /* Cleanup any SCTP per socket resources.  */
3621 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3622 {
3623         struct sctp_endpoint *ep;
3624
3625         SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3626
3627         /* Release our hold on the endpoint. */
3628         ep = sctp_sk(sk)->ep;
3629         sctp_endpoint_free(ep);
3630         atomic_dec(&sctp_sockets_allocated);
3631 }
3632
3633 /* API 4.1.7 shutdown() - TCP Style Syntax
3634  *     int shutdown(int socket, int how);
3635  *
3636  *     sd      - the socket descriptor of the association to be closed.
3637  *     how     - Specifies the type of shutdown.  The  values  are
3638  *               as follows:
3639  *               SHUT_RD
3640  *                     Disables further receive operations. No SCTP
3641  *                     protocol action is taken.
3642  *               SHUT_WR
3643  *                     Disables further send operations, and initiates
3644  *                     the SCTP shutdown sequence.
3645  *               SHUT_RDWR
3646  *                     Disables further send  and  receive  operations
3647  *                     and initiates the SCTP shutdown sequence.
3648  */
3649 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3650 {
3651         struct sctp_endpoint *ep;
3652         struct sctp_association *asoc;
3653
3654         if (!sctp_style(sk, TCP))
3655                 return;
3656
3657         if (how & SEND_SHUTDOWN) {
3658                 ep = sctp_sk(sk)->ep;
3659                 if (!list_empty(&ep->asocs)) {
3660                         asoc = list_entry(ep->asocs.next,
3661                                           struct sctp_association, asocs);
3662                         sctp_primitive_SHUTDOWN(asoc, NULL);
3663                 }
3664         }
3665 }
3666
3667 /* 7.2.1 Association Status (SCTP_STATUS)
3668
3669  * Applications can retrieve current status information about an
3670  * association, including association state, peer receiver window size,
3671  * number of unacked data chunks, and number of data chunks pending
3672  * receipt.  This information is read-only.
3673  */
3674 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3675                                        char __user *optval,
3676                                        int __user *optlen)
3677 {
3678         struct sctp_status status;
3679         struct sctp_association *asoc = NULL;
3680         struct sctp_transport *transport;
3681         sctp_assoc_t associd;
3682         int retval = 0;
3683
3684         if (len < sizeof(status)) {
3685                 retval = -EINVAL;
3686                 goto out;
3687         }
3688
3689         len = sizeof(status);
3690         if (copy_from_user(&status, optval, len)) {
3691                 retval = -EFAULT;
3692                 goto out;
3693         }
3694
3695         associd = status.sstat_assoc_id;
3696         asoc = sctp_id2assoc(sk, associd);
3697         if (!asoc) {
3698                 retval = -EINVAL;
3699                 goto out;
3700         }
3701
3702         transport = asoc->peer.primary_path;
3703
3704         status.sstat_assoc_id = sctp_assoc2id(asoc);
3705         status.sstat_state = asoc->state;
3706         status.sstat_rwnd =  asoc->peer.rwnd;
3707         status.sstat_unackdata = asoc->unack_data;
3708
3709         status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3710         status.sstat_instrms = asoc->c.sinit_max_instreams;
3711         status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3712         status.sstat_fragmentation_point = asoc->frag_point;
3713         status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3714         memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3715                         transport->af_specific->sockaddr_len);
3716         /* Map ipv4 address into v4-mapped-on-v6 address.  */
3717         sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3718                 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3719         status.sstat_primary.spinfo_state = transport->state;
3720         status.sstat_primary.spinfo_cwnd = transport->cwnd;
3721         status.sstat_primary.spinfo_srtt = transport->srtt;
3722         status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3723         status.sstat_primary.spinfo_mtu = transport->pathmtu;
3724
3725         if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3726                 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3727
3728         if (put_user(len, optlen)) {
3729                 retval = -EFAULT;
3730                 goto out;
3731         }
3732
3733         SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3734                           len, status.sstat_state, status.sstat_rwnd,
3735                           status.sstat_assoc_id);
3736
3737         if (copy_to_user(optval, &status, len)) {
3738                 retval = -EFAULT;
3739                 goto out;
3740         }
3741
3742 out:
3743         return (retval);
3744 }
3745
3746
3747 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3748  *
3749  * Applications can retrieve information about a specific peer address
3750  * of an association, including its reachability state, congestion
3751  * window, and retransmission timer values.  This information is
3752  * read-only.
3753  */
3754 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3755                                           char __user *optval,
3756                                           int __user *optlen)
3757 {
3758         struct sctp_paddrinfo pinfo;
3759         struct sctp_transport *transport;
3760         int retval = 0;
3761
3762         if (len < sizeof(pinfo)) {
3763                 retval = -EINVAL;
3764                 goto out;
3765         }
3766
3767         len = sizeof(pinfo);
3768         if (copy_from_user(&pinfo, optval, len)) {
3769                 retval = -EFAULT;
3770                 goto out;
3771         }
3772
3773         transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3774                                            pinfo.spinfo_assoc_id);
3775         if (!transport)
3776                 return -EINVAL;
3777
3778         pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3779         pinfo.spinfo_state = transport->state;
3780         pinfo.spinfo_cwnd = transport->cwnd;
3781         pinfo.spinfo_srtt = transport->srtt;
3782         pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3783         pinfo.spinfo_mtu = transport->pathmtu;
3784
3785         if (pinfo.spinfo_state == SCTP_UNKNOWN)
3786                 pinfo.spinfo_state = SCTP_ACTIVE;
3787
3788         if (put_user(len, optlen)) {
3789                 retval = -EFAULT;
3790                 goto out;
3791         }
3792
3793         if (copy_to_user(optval, &pinfo, len)) {
3794                 retval = -EFAULT;
3795                 goto out;
3796         }
3797
3798 out:
3799         return (retval);
3800 }
3801
3802 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3803  *
3804  * This option is a on/off flag.  If enabled no SCTP message
3805  * fragmentation will be performed.  Instead if a message being sent
3806  * exceeds the current PMTU size, the message will NOT be sent and
3807  * instead a error will be indicated to the user.
3808  */
3809 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3810                                         char __user *optval, int __user *optlen)
3811 {
3812         int val;
3813
3814         if (len < sizeof(int))
3815                 return -EINVAL;
3816
3817         len = sizeof(int);
3818         val = (sctp_sk(sk)->disable_fragments == 1);
3819         if (put_user(len, optlen))
3820                 return -EFAULT;
3821         if (copy_to_user(optval, &val, len))
3822                 return -EFAULT;
3823         return 0;
3824 }
3825
3826 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3827  *
3828  * This socket option is used to specify various notifications and
3829  * ancillary data the user wishes to receive.
3830  */
3831 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3832                                   int __user *optlen)
3833 {
3834         if (len < sizeof(struct sctp_event_subscribe))
3835                 return -EINVAL;
3836         len = sizeof(struct sctp_event_subscribe);
3837         if (put_user(len, optlen))
3838                 return -EFAULT;
3839         if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3840                 return -EFAULT;
3841         return 0;
3842 }
3843
3844 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3845  *
3846  * This socket option is applicable to the UDP-style socket only.  When
3847  * set it will cause associations that are idle for more than the
3848  * specified number of seconds to automatically close.  An association
3849  * being idle is defined an association that has NOT sent or received
3850  * user data.  The special value of '0' indicates that no automatic
3851  * close of any associations should be performed.  The option expects an
3852  * integer defining the number of seconds of idle time before an
3853  * association is closed.
3854  */
3855 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3856 {
3857         /* Applicable to UDP-style socket only */
3858         if (sctp_style(sk, TCP))
3859                 return -EOPNOTSUPP;
3860         if (len < sizeof(int))
3861                 return -EINVAL;
3862         len = sizeof(int);
3863         if (put_user(len, optlen))
3864                 return -EFAULT;
3865         if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3866                 return -EFAULT;
3867         return 0;
3868 }
3869
3870 /* Helper routine to branch off an association to a new socket.  */
3871 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3872                                 struct socket **sockp)
3873 {
3874         struct sock *sk = asoc->base.sk;
3875         struct socket *sock;
3876         struct inet_sock *inetsk;
3877         struct sctp_af *af;
3878         int err = 0;
3879
3880         /* An association cannot be branched off from an already peeled-off
3881          * socket, nor is this supported for tcp style sockets.
3882          */
3883         if (!sctp_style(sk, UDP))
3884                 return -EINVAL;
3885
3886         /* Create a new socket.  */
3887         err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3888         if (err < 0)
3889                 return err;
3890
3891         /* Populate the fields of the newsk from the oldsk and migrate the
3892          * asoc to the newsk.
3893          */
3894         sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3895
3896         /* Make peeled-off sockets more like 1-1 accepted sockets.
3897          * Set the daddr and initialize id to something more random
3898          */
3899         af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3900         af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3901         inetsk = inet_sk(sock->sk);
3902         inetsk->id = asoc->next_tsn ^ jiffies;
3903
3904         *sockp = sock;
3905
3906         return err;
3907 }
3908
3909 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3910 {
3911         sctp_peeloff_arg_t peeloff;
3912         struct socket *newsock;
3913         int retval = 0;
3914         struct sctp_association *asoc;
3915
3916         if (len < sizeof(sctp_peeloff_arg_t))
3917                 return -EINVAL;
3918         len = sizeof(sctp_peeloff_arg_t);
3919         if (copy_from_user(&peeloff, optval, len))
3920                 return -EFAULT;
3921
3922         asoc = sctp_id2assoc(sk, peeloff.associd);
3923         if (!asoc) {
3924                 retval = -EINVAL;
3925                 goto out;
3926         }
3927
3928         SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
3929
3930         retval = sctp_do_peeloff(asoc, &newsock);
3931         if (retval < 0)
3932                 goto out;
3933
3934         /* Map the socket to an unused fd that can be returned to the user.  */
3935         retval = sock_map_fd(newsock, 0);
3936         if (retval < 0) {
3937                 sock_release(newsock);
3938                 goto out;
3939         }
3940
3941         SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3942                           __func__, sk, asoc, newsock->sk, retval);
3943
3944         /* Return the fd mapped to the new socket.  */
3945         peeloff.sd = retval;
3946         if (put_user(len, optlen))
3947                 return -EFAULT;
3948         if (copy_to_user(optval, &peeloff, len))
3949                 retval = -EFAULT;
3950
3951 out:
3952         return retval;
3953 }
3954
3955 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3956  *
3957  * Applications can enable or disable heartbeats for any peer address of
3958  * an association, modify an address's heartbeat interval, force a
3959  * heartbeat to be sent immediately, and adjust the address's maximum
3960  * number of retransmissions sent before an address is considered
3961  * unreachable.  The following structure is used to access and modify an
3962  * address's parameters:
3963  *
3964  *  struct sctp_paddrparams {
3965  *     sctp_assoc_t            spp_assoc_id;
3966  *     struct sockaddr_storage spp_address;
3967  *     uint32_t                spp_hbinterval;
3968  *     uint16_t                spp_pathmaxrxt;
3969  *     uint32_t                spp_pathmtu;
3970  *     uint32_t                spp_sackdelay;
3971  *     uint32_t                spp_flags;
3972  * };
3973  *
3974  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
3975  *                     application, and identifies the association for
3976  *                     this query.
3977  *   spp_address     - This specifies which address is of interest.
3978  *   spp_hbinterval  - This contains the value of the heartbeat interval,
3979  *                     in milliseconds.  If a  value of zero
3980  *                     is present in this field then no changes are to
3981  *                     be made to this parameter.
3982  *   spp_pathmaxrxt  - This contains the maximum number of
3983  *                     retransmissions before this address shall be
3984  *                     considered unreachable. If a  value of zero
3985  *                     is present in this field then no changes are to
3986  *                     be made to this parameter.
3987  *   spp_pathmtu     - When Path MTU discovery is disabled the value
3988  *                     specified here will be the "fixed" path mtu.
3989  *                     Note that if the spp_address field is empty
3990  *                     then all associations on this address will
3991  *                     have this fixed path mtu set upon them.
3992  *
3993  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
3994  *                     the number of milliseconds that sacks will be delayed
3995  *                     for. This value will apply to all addresses of an
3996  *                     association if the spp_address field is empty. Note
3997  *                     also, that if delayed sack is enabled and this
3998  *                     value is set to 0, no change is made to the last
3999  *                     recorded delayed sack timer value.
4000  *
4001  *   spp_flags       - These flags are used to control various features
4002  *                     on an association. The flag field may contain
4003  *                     zero or more of the following options.
4004  *
4005  *                     SPP_HB_ENABLE  - Enable heartbeats on the
4006  *                     specified address. Note that if the address
4007  *                     field is empty all addresses for the association
4008  *                     have heartbeats enabled upon them.
4009  *
4010  *                     SPP_HB_DISABLE - Disable heartbeats on the
4011  *                     speicifed address. Note that if the address
4012  *                     field is empty all addresses for the association
4013  *                     will have their heartbeats disabled. Note also
4014  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
4015  *                     mutually exclusive, only one of these two should
4016  *                     be specified. Enabling both fields will have
4017  *                     undetermined results.
4018  *
4019  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
4020  *                     to be made immediately.
4021  *
4022  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
4023  *                     discovery upon the specified address. Note that
4024  *                     if the address feild is empty then all addresses
4025  *                     on the association are effected.
4026  *
4027  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
4028  *                     discovery upon the specified address. Note that
4029  *                     if the address feild is empty then all addresses
4030  *                     on the association are effected. Not also that
4031  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4032  *                     exclusive. Enabling both will have undetermined
4033  *                     results.
4034  *
4035  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
4036  *                     on delayed sack. The time specified in spp_sackdelay
4037  *                     is used to specify the sack delay for this address. Note
4038  *                     that if spp_address is empty then all addresses will
4039  *                     enable delayed sack and take on the sack delay
4040  *                     value specified in spp_sackdelay.
4041  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
4042  *                     off delayed sack. If the spp_address field is blank then
4043  *                     delayed sack is disabled for the entire association. Note
4044  *                     also that this field is mutually exclusive to
4045  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
4046  *                     results.
4047  */
4048 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4049                                             char __user *optval, int __user *optlen)
4050 {
4051         struct sctp_paddrparams  params;
4052         struct sctp_transport   *trans = NULL;
4053         struct sctp_association *asoc = NULL;
4054         struct sctp_sock        *sp = sctp_sk(sk);
4055
4056         if (len < sizeof(struct sctp_paddrparams))
4057                 return -EINVAL;
4058         len = sizeof(struct sctp_paddrparams);
4059         if (copy_from_user(&params, optval, len))
4060                 return -EFAULT;
4061
4062         /* If an address other than INADDR_ANY is specified, and
4063          * no transport is found, then the request is invalid.
4064          */
4065         if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4066                 trans = sctp_addr_id2transport(sk, &params.spp_address,
4067                                                params.spp_assoc_id);
4068                 if (!trans) {
4069                         SCTP_DEBUG_PRINTK("Failed no transport\n");
4070                         return -EINVAL;
4071                 }
4072         }
4073
4074         /* Get association, if assoc_id != 0 and the socket is a one
4075          * to many style socket, and an association was not found, then
4076          * the id was invalid.
4077          */
4078         asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4079         if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4080                 SCTP_DEBUG_PRINTK("Failed no association\n");
4081                 return -EINVAL;
4082         }
4083
4084         if (trans) {
4085                 /* Fetch transport values. */
4086                 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4087                 params.spp_pathmtu    = trans->pathmtu;
4088                 params.spp_pathmaxrxt = trans->pathmaxrxt;
4089                 params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
4090
4091                 /*draft-11 doesn't say what to return in spp_flags*/
4092                 params.spp_flags      = trans->param_flags;
4093         } else if (asoc) {
4094                 /* Fetch association values. */
4095                 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4096                 params.spp_pathmtu    = asoc->pathmtu;
4097                 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4098                 params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
4099
4100                 /*draft-11 doesn't say what to return in spp_flags*/
4101                 params.spp_flags      = asoc->param_flags;
4102         } else {
4103                 /* Fetch socket values. */
4104                 params.spp_hbinterval = sp->hbinterval;
4105                 params.spp_pathmtu    = sp->pathmtu;
4106                 params.spp_sackdelay  = sp->sackdelay;
4107                 params.spp_pathmaxrxt = sp->pathmaxrxt;
4108
4109                 /*draft-11 doesn't say what to return in spp_flags*/
4110                 params.spp_flags      = sp->param_flags;
4111         }
4112
4113         if (copy_to_user(optval, &params, len))
4114                 return -EFAULT;
4115
4116         if (put_user(len, optlen))
4117                 return -EFAULT;
4118
4119         return 0;
4120 }
4121
4122 /*
4123  * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
4124  *
4125  * This option will effect the way delayed acks are performed.  This
4126  * option allows you to get or set the delayed ack time, in
4127  * milliseconds.  It also allows changing the delayed ack frequency.
4128  * Changing the frequency to 1 disables the delayed sack algorithm.  If
4129  * the assoc_id is 0, then this sets or gets the endpoints default
4130  * values.  If the assoc_id field is non-zero, then the set or get
4131  * effects the specified association for the one to many model (the
4132  * assoc_id field is ignored by the one to one model).  Note that if
4133  * sack_delay or sack_freq are 0 when setting this option, then the
4134  * current values will remain unchanged.
4135  *
4136  * struct sctp_sack_info {
4137  *     sctp_assoc_t            sack_assoc_id;
4138  *     uint32_t                sack_delay;
4139  *     uint32_t                sack_freq;
4140  * };
4141  *
4142  * sack_assoc_id -  This parameter, indicates which association the user
4143  *    is performing an action upon.  Note that if this field's value is
4144  *    zero then the endpoints default value is changed (effecting future
4145  *    associations only).
4146  *
4147  * sack_delay -  This parameter contains the number of milliseconds that
4148  *    the user is requesting the delayed ACK timer be set to.  Note that
4149  *    this value is defined in the standard to be between 200 and 500
4150  *    milliseconds.
4151  *
4152  * sack_freq -  This parameter contains the number of packets that must
4153  *    be received before a sack is sent without waiting for the delay
4154  *    timer to expire.  The default value for this is 2, setting this
4155  *    value to 1 will disable the delayed sack algorithm.
4156  */
4157 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4158                                             char __user *optval,
4159                                             int __user *optlen)
4160 {
4161         struct sctp_sack_info    params;
4162         struct sctp_association *asoc = NULL;
4163         struct sctp_sock        *sp = sctp_sk(sk);
4164
4165         if (len >= sizeof(struct sctp_sack_info)) {
4166                 len = sizeof(struct sctp_sack_info);
4167
4168                 if (copy_from_user(&params, optval, len))
4169                         return -EFAULT;
4170         } else if (len == sizeof(struct sctp_assoc_value)) {
4171                 printk(KERN_WARNING "SCTP: Use of struct sctp_sack_info "
4172                        "in delayed_ack socket option deprecated\n");
4173                 printk(KERN_WARNING "SCTP: struct sctp_sack_info instead\n");
4174                 if (copy_from_user(&params, optval, len))
4175                         return -EFAULT;
4176         } else
4177                 return - EINVAL;
4178
4179         /* Get association, if sack_assoc_id != 0 and the socket is a one
4180          * to many style socket, and an association was not found, then
4181          * the id was invalid.
4182          */
4183         asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4184         if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4185                 return -EINVAL;
4186
4187         if (asoc) {
4188                 /* Fetch association values. */
4189                 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4190                         params.sack_delay = jiffies_to_msecs(
4191                                 asoc->sackdelay);
4192                         params.sack_freq = asoc->sackfreq;
4193
4194                 } else {
4195                         params.sack_delay = 0;
4196                         params.sack_freq = 1;
4197                 }
4198         } else {
4199                 /* Fetch socket values. */
4200                 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4201                         params.sack_delay  = sp->sackdelay;
4202                         params.sack_freq = sp->sackfreq;
4203                 } else {
4204                         params.sack_delay  = 0;
4205                         params.sack_freq = 1;
4206                 }
4207         }
4208
4209         if (copy_to_user(optval, &params, len))
4210                 return -EFAULT;
4211
4212         if (put_user(len, optlen))
4213                 return -EFAULT;
4214
4215         return 0;
4216 }
4217
4218 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4219  *
4220  * Applications can specify protocol parameters for the default association
4221  * initialization.  The option name argument to setsockopt() and getsockopt()
4222  * is SCTP_INITMSG.
4223  *
4224  * Setting initialization parameters is effective only on an unconnected
4225  * socket (for UDP-style sockets only future associations are effected
4226  * by the change).  With TCP-style sockets, this option is inherited by
4227  * sockets derived from a listener socket.
4228  */
4229 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4230 {
4231         if (len < sizeof(struct sctp_initmsg))
4232                 return -EINVAL;
4233         len = sizeof(struct sctp_initmsg);
4234         if (put_user(len, optlen))
4235                 return -EFAULT;
4236         if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4237                 return -EFAULT;
4238         return 0;
4239 }
4240
4241 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4242                                               char __user *optval,
4243                                               int __user *optlen)
4244 {
4245         sctp_assoc_t id;
4246         struct sctp_association *asoc;
4247         struct list_head *pos;
4248         int cnt = 0;
4249
4250         if (len < sizeof(sctp_assoc_t))
4251                 return -EINVAL;
4252
4253         if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4254                 return -EFAULT;
4255
4256         printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_NUM_OLD "
4257                             "socket option deprecated\n");
4258         /* For UDP-style sockets, id specifies the association to query.  */
4259         asoc = sctp_id2assoc(sk, id);
4260         if (!asoc)
4261                 return -EINVAL;
4262
4263         list_for_each(pos, &asoc->peer.transport_addr_list) {
4264                 cnt ++;
4265         }
4266
4267         return cnt;
4268 }
4269
4270 /*
4271  * Old API for getting list of peer addresses. Does not work for 32-bit
4272  * programs running on a 64-bit kernel
4273  */
4274 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4275                                           char __user *optval,
4276                                           int __user *optlen)
4277 {
4278         struct sctp_association *asoc;
4279         int cnt = 0;
4280         struct sctp_getaddrs_old getaddrs;
4281         struct sctp_transport *from;
4282         void __user *to;
4283         union sctp_addr temp;
4284         struct sctp_sock *sp = sctp_sk(sk);
4285         int addrlen;
4286
4287         if (len < sizeof(struct sctp_getaddrs_old))
4288                 return -EINVAL;
4289
4290         len = sizeof(struct sctp_getaddrs_old);
4291
4292         if (copy_from_user(&getaddrs, optval, len))
4293                 return -EFAULT;
4294
4295         if (getaddrs.addr_num <= 0) return -EINVAL;
4296
4297         printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_OLD "
4298                             "socket option deprecated\n");
4299
4300         /* For UDP-style sockets, id specifies the association to query.  */
4301         asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4302         if (!asoc)
4303                 return -EINVAL;
4304
4305         to = (void __user *)getaddrs.addrs;
4306         list_for_each_entry(from, &asoc->peer.transport_addr_list,
4307                                 transports) {
4308                 memcpy(&temp, &from->ipaddr, sizeof(temp));
4309                 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4310                 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4311                 if (copy_to_user(to, &temp, addrlen))
4312                         return -EFAULT;
4313                 to += addrlen ;
4314                 cnt ++;
4315                 if (cnt >= getaddrs.addr_num) break;
4316         }
4317         getaddrs.addr_num = cnt;
4318         if (put_user(len, optlen))
4319                 return -EFAULT;
4320         if (copy_to_user(optval, &getaddrs, len))
4321                 return -EFAULT;
4322
4323         return 0;
4324 }
4325
4326 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4327                                       char __user *optval, int __user *optlen)
4328 {
4329         struct sctp_association *asoc;
4330         int cnt = 0;
4331         struct sctp_getaddrs getaddrs;
4332         struct sctp_transport *from;
4333         void __user *to;
4334         union sctp_addr temp;
4335         struct sctp_sock *sp = sctp_sk(sk);
4336         int addrlen;
4337         size_t space_left;
4338         int bytes_copied;
4339
4340         if (len < sizeof(struct sctp_getaddrs))
4341                 return -EINVAL;
4342
4343         if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4344                 return -EFAULT;
4345
4346         /* For UDP-style sockets, id specifies the association to query.  */
4347         asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4348         if (!asoc)
4349                 return -EINVAL;
4350
4351         to = optval + offsetof(struct sctp_getaddrs,addrs);
4352         space_left = len - offsetof(struct sctp_getaddrs,addrs);
4353
4354         list_for_each_entry(from, &asoc->peer.transport_addr_list,
4355                                 transports) {
4356                 memcpy(&temp, &from->ipaddr, sizeof(temp));
4357                 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4358                 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4359                 if (space_left < addrlen)
4360                         return -ENOMEM;
4361                 if (copy_to_user(to, &temp, addrlen))
4362                         return -EFAULT;
4363                 to += addrlen;
4364                 cnt++;
4365                 space_left -= addrlen;
4366         }
4367
4368         if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4369                 return -EFAULT;
4370         bytes_copied = ((char __user *)to) - optval;
4371         if (put_user(bytes_copied, optlen))
4372                 return -EFAULT;
4373
4374         return 0;
4375 }
4376
4377 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4378                                                char __user *optval,
4379                                                int __user *optlen)
4380 {
4381         sctp_assoc_t id;
4382         struct sctp_bind_addr *bp;
4383         struct sctp_association *asoc;
4384         struct sctp_sockaddr_entry *addr;
4385         int cnt = 0;
4386
4387         if (len < sizeof(sctp_assoc_t))
4388                 return -EINVAL;
4389
4390         if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4391                 return -EFAULT;
4392
4393         printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_NUM_OLD "
4394                             "socket option deprecated\n");
4395
4396         /*
4397          *  For UDP-style sockets, id specifies the association to query.
4398          *  If the id field is set to the value '0' then the locally bound
4399          *  addresses are returned without regard to any particular
4400          *  association.
4401          */
4402         if (0 == id) {
4403                 bp = &sctp_sk(sk)->ep->base.bind_addr;
4404         } else {
4405                 asoc = sctp_id2assoc(sk, id);
4406                 if (!asoc)
4407                         return -EINVAL;
4408                 bp = &asoc->base.bind_addr;
4409         }
4410
4411         /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4412          * addresses from the global local address list.
4413          */
4414         if (sctp_list_single_entry(&bp->address_list)) {
4415                 addr = list_entry(bp->address_list.next,
4416                                   struct sctp_sockaddr_entry, list);
4417                 if (sctp_is_any(sk, &addr->a)) {
4418                         rcu_read_lock();
4419                         list_for_each_entry_rcu(addr,
4420                                                 &sctp_local_addr_list, list) {
4421                                 if (!addr->valid)
4422                                         continue;
4423
4424                                 if ((PF_INET == sk->sk_family) &&
4425                                     (AF_INET6 == addr->a.sa.sa_family))
4426                                         continue;
4427
4428                                 if ((PF_INET6 == sk->sk_family) &&
4429                                     inet_v6_ipv6only(sk) &&
4430                                     (AF_INET == addr->a.sa.sa_family))
4431                                         continue;
4432
4433                                 cnt++;
4434                         }
4435                         rcu_read_unlock();
4436                 } else {
4437                         cnt = 1;
4438                 }
4439                 goto done;
4440         }
4441
4442         /* Protection on the bound address list is not needed,
4443          * since in the socket option context we hold the socket lock,
4444          * so there is no way that the bound address list can change.
4445          */
4446         list_for_each_entry(addr, &bp->address_list, list) {
4447                 cnt ++;
4448         }
4449 done:
4450         return cnt;
4451 }
4452
4453 /* Helper function that copies local addresses to user and returns the number
4454  * of addresses copied.
4455  */
4456 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4457                                         int max_addrs, void *to,
4458                                         int *bytes_copied)
4459 {
4460         struct sctp_sockaddr_entry *addr;
4461         union sctp_addr temp;
4462         int cnt = 0;
4463         int addrlen;
4464
4465         rcu_read_lock();
4466         list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4467                 if (!addr->valid)
4468                         continue;
4469
4470                 if ((PF_INET == sk->sk_family) &&
4471                     (AF_INET6 == addr->a.sa.sa_family))
4472                         continue;
4473                 if ((PF_INET6 == sk->sk_family) &&
4474                     inet_v6_ipv6only(sk) &&
4475                     (AF_INET == addr->a.sa.sa_family))
4476                         continue;
4477                 memcpy(&temp, &addr->a, sizeof(temp));
4478                 if (!temp.v4.sin_port)
4479                         temp.v4.sin_port = htons(port);
4480
4481                 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4482                                                                 &temp);
4483                 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4484                 memcpy(to, &temp, addrlen);
4485
4486                 to += addrlen;
4487                 *bytes_copied += addrlen;
4488                 cnt ++;
4489                 if (cnt >= max_addrs) break;
4490         }
4491         rcu_read_unlock();
4492
4493         return cnt;
4494 }
4495
4496 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4497                             size_t space_left, int *bytes_copied)
4498 {
4499         struct sctp_sockaddr_entry *addr;
4500         union sctp_addr temp;
4501         int cnt = 0;
4502         int addrlen;
4503
4504         rcu_read_lock();
4505         list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4506                 if (!addr->valid)
4507                         continue;
4508
4509                 if ((PF_INET == sk->sk_family) &&
4510                     (AF_INET6 == addr->a.sa.sa_family))
4511                         continue;
4512                 if ((PF_INET6 == sk->sk_family) &&
4513                     inet_v6_ipv6only(sk) &&
4514                     (AF_INET == addr->a.sa.sa_family))
4515                         continue;
4516                 memcpy(&temp, &addr->a, sizeof(temp));
4517                 if (!temp.v4.sin_port)
4518                         temp.v4.sin_port = htons(port);
4519
4520                 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4521                                                                 &temp);
4522                 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4523                 if (space_left < addrlen) {
4524                         cnt =  -ENOMEM;
4525                         break;
4526                 }
4527                 memcpy(to, &temp, addrlen);
4528
4529                 to += addrlen;
4530                 cnt ++;
4531                 space_left -= addrlen;
4532                 *bytes_copied += addrlen;
4533         }
4534         rcu_read_unlock();
4535
4536         return cnt;
4537 }
4538
4539 /* Old API for getting list of local addresses. Does not work for 32-bit
4540  * programs running on a 64-bit kernel
4541  */
4542 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4543                                            char __user *optval, int __user *optlen)
4544 {
4545         struct sctp_bind_addr *bp;
4546         struct sctp_association *asoc;
4547         int cnt = 0;
4548         struct sctp_getaddrs_old getaddrs;
4549         struct sctp_sockaddr_entry *addr;
4550         void __user *to;
4551         union sctp_addr temp;
4552         struct sctp_sock *sp = sctp_sk(sk);
4553         int addrlen;
4554         int err = 0;
4555         void *addrs;
4556         void *buf;
4557         int bytes_copied = 0;
4558
4559         if (len < sizeof(struct sctp_getaddrs_old))
4560                 return -EINVAL;
4561
4562         len = sizeof(struct sctp_getaddrs_old);
4563         if (copy_from_user(&getaddrs, optval, len))
4564                 return -EFAULT;
4565
4566         if (getaddrs.addr_num <= 0 ||
4567             getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4568                 return -EINVAL;
4569
4570         printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_OLD "
4571                             "socket option deprecated\n");
4572
4573         /*
4574          *  For UDP-style sockets, id specifies the association to query.
4575          *  If the id field is set to the value '0' then the locally bound
4576          *  addresses are returned without regard to any particular
4577          *  association.
4578          */
4579         if (0 == getaddrs.assoc_id) {
4580                 bp = &sctp_sk(sk)->ep->base.bind_addr;
4581         } else {
4582                 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4583                 if (!asoc)
4584                         return -EINVAL;
4585                 bp = &asoc->base.bind_addr;
4586         }
4587
4588         to = getaddrs.addrs;
4589
4590         /* Allocate space for a local instance of packed array to hold all
4591          * the data.  We store addresses here first and then put write them
4592          * to the user in one shot.
4593          */
4594         addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4595                         GFP_KERNEL);
4596         if (!addrs)
4597                 return -ENOMEM;
4598
4599         /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4600          * addresses from the global local address list.
4601          */
4602         if (sctp_list_single_entry(&bp->address_list)) {
4603                 addr = list_entry(bp->address_list.next,
4604                                   struct sctp_sockaddr_entry, list);
4605                 if (sctp_is_any(sk, &addr->a)) {
4606                         cnt = sctp_copy_laddrs_old(sk, bp->port,
4607                                                    getaddrs.addr_num,
4608                                                    addrs, &bytes_copied);
4609                         goto copy_getaddrs;
4610                 }
4611         }
4612
4613         buf = addrs;
4614         /* Protection on the bound address list is not needed since
4615          * in the socket option context we hold a socket lock and
4616          * thus the bound address list can't change.
4617          */
4618         list_for_each_entry(addr, &bp->address_list, list) {
4619                 memcpy(&temp, &addr->a, sizeof(temp));
4620                 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4621                 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4622                 memcpy(buf, &temp, addrlen);
4623                 buf += addrlen;
4624                 bytes_copied += addrlen;
4625                 cnt ++;
4626                 if (cnt >= getaddrs.addr_num) break;
4627         }
4628
4629 copy_getaddrs:
4630         /* copy the entire address list into the user provided space */
4631         if (copy_to_user(to, addrs, bytes_copied)) {
4632                 err = -EFAULT;
4633                 goto error;
4634         }
4635
4636         /* copy the leading structure back to user */
4637         getaddrs.addr_num = cnt;
4638         if (copy_to_user(optval, &getaddrs, len))
4639                 err = -EFAULT;
4640
4641 error:
4642         kfree(addrs);
4643         return err;
4644 }
4645
4646 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4647                                        char __user *optval, int __user *optlen)
4648 {
4649         struct sctp_bind_addr *bp;
4650         struct sctp_association *asoc;
4651         int cnt = 0;
4652         struct sctp_getaddrs getaddrs;
4653         struct sctp_sockaddr_entry *addr;
4654         void __user *to;
4655         union sctp_addr temp;
4656         struct sctp_sock *sp = sctp_sk(sk);
4657         int addrlen;
4658         int err = 0;
4659         size_t space_left;
4660         int bytes_copied = 0;
4661         void *addrs;
4662         void *buf;
4663
4664         if (len < sizeof(struct sctp_getaddrs))
4665                 return -EINVAL;
4666
4667         if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4668                 return -EFAULT;
4669
4670         /*
4671          *  For UDP-style sockets, id specifies the association to query.
4672          *  If the id field is set to the value '0' then the locally bound
4673          *  addresses are returned without regard to any particular
4674          *  association.
4675          */
4676         if (0 == getaddrs.assoc_id) {
4677                 bp = &sctp_sk(sk)->ep->base.bind_addr;
4678         } else {
4679                 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4680                 if (!asoc)
4681                         return -EINVAL;
4682                 bp = &asoc->base.bind_addr;
4683         }
4684
4685         to = optval + offsetof(struct sctp_getaddrs,addrs);
4686         space_left = len - offsetof(struct sctp_getaddrs,addrs);
4687
4688         addrs = kmalloc(space_left, GFP_KERNEL);
4689         if (!addrs)
4690                 return -ENOMEM;
4691
4692         /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4693          * addresses from the global local address list.
4694          */
4695         if (sctp_list_single_entry(&bp->address_list)) {
4696                 addr = list_entry(bp->address_list.next,
4697                                   struct sctp_sockaddr_entry, list);
4698                 if (sctp_is_any(sk, &addr->a)) {
4699                         cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4700                                                 space_left, &bytes_copied);
4701                         if (cnt < 0) {
4702                                 err = cnt;
4703                                 goto out;
4704                         }
4705                         goto copy_getaddrs;
4706                 }
4707         }
4708
4709         buf = addrs;
4710         /* Protection on the bound address list is not needed since
4711          * in the socket option context we hold a socket lock and
4712          * thus the bound address list can't change.
4713          */
4714         list_for_each_entry(addr, &bp->address_list, list) {
4715                 memcpy(&temp, &addr->a, sizeof(temp));
4716                 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4717                 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4718                 if (space_left < addrlen) {
4719                         err =  -ENOMEM; /*fixme: right error?*/
4720                         goto out;
4721                 }
4722                 memcpy(buf, &temp, addrlen);
4723                 buf += addrlen;
4724                 bytes_copied += addrlen;
4725                 cnt ++;
4726                 space_left -= addrlen;
4727         }
4728
4729 copy_getaddrs:
4730         if (copy_to_user(to, addrs, bytes_copied)) {
4731                 err = -EFAULT;
4732                 goto out;
4733         }
4734         if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4735                 err = -EFAULT;
4736                 goto out;
4737         }
4738         if (put_user(bytes_copied, optlen))
4739                 err = -EFAULT;
4740 out:
4741         kfree(addrs);
4742         return err;
4743 }
4744
4745 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4746  *
4747  * Requests that the local SCTP stack use the enclosed peer address as
4748  * the association primary.  The enclosed address must be one of the
4749  * association peer's addresses.
4750  */
4751 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4752                                         char __user *optval, int __user *optlen)
4753 {
4754         struct sctp_prim prim;
4755         struct sctp_association *asoc;
4756         struct sctp_sock *sp = sctp_sk(sk);
4757
4758         if (len < sizeof(struct sctp_prim))
4759                 return -EINVAL;
4760
4761         len = sizeof(struct sctp_prim);
4762
4763         if (copy_from_user(&prim, optval, len))
4764                 return -EFAULT;
4765
4766         asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4767         if (!asoc)
4768                 return -EINVAL;
4769
4770         if (!asoc->peer.primary_path)
4771                 return -ENOTCONN;
4772
4773         memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4774                 asoc->peer.primary_path->af_specific->sockaddr_len);
4775
4776         sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4777                         (union sctp_addr *)&prim.ssp_addr);
4778
4779         if (put_user(len, optlen))
4780                 return -EFAULT;
4781         if (copy_to_user(optval, &prim, len))
4782                 return -EFAULT;
4783
4784         return 0;
4785 }
4786
4787 /*
4788  * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4789  *
4790  * Requests that the local endpoint set the specified Adaptation Layer
4791  * Indication parameter for all future INIT and INIT-ACK exchanges.
4792  */
4793 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4794                                   char __user *optval, int __user *optlen)
4795 {
4796         struct sctp_setadaptation adaptation;
4797
4798         if (len < sizeof(struct sctp_setadaptation))
4799                 return -EINVAL;
4800
4801         len = sizeof(struct sctp_setadaptation);
4802
4803         adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4804
4805         if (put_user(len, optlen))
4806                 return -EFAULT;
4807         if (copy_to_user(optval, &adaptation, len))
4808                 return -EFAULT;
4809
4810         return 0;
4811 }
4812
4813 /*
4814  *
4815  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4816  *
4817  *   Applications that wish to use the sendto() system call may wish to
4818  *   specify a default set of parameters that would normally be supplied
4819  *   through the inclusion of ancillary data.  This socket option allows
4820  *   such an application to set the default sctp_sndrcvinfo structure.
4821
4822
4823  *   The application that wishes to use this socket option simply passes
4824  *   in to this call the sctp_sndrcvinfo structure defined in Section
4825  *   5.2.2) The input parameters accepted by this call include
4826  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4827  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
4828  *   to this call if the caller is using the UDP model.
4829  *
4830  *   For getsockopt, it get the default sctp_sndrcvinfo structure.
4831  */
4832 static int sctp_getsockopt_default_send_param(struct sock *sk,
4833                                         int len, char __user *optval,
4834                                         int __user *optlen)
4835 {
4836         struct sctp_sndrcvinfo info;
4837         struct sctp_association *asoc;
4838         struct sctp_sock *sp = sctp_sk(sk);
4839
4840         if (len < sizeof(struct sctp_sndrcvinfo))
4841                 return -EINVAL;
4842
4843         len = sizeof(struct sctp_sndrcvinfo);
4844
4845         if (copy_from_user(&info, optval, len))
4846                 return -EFAULT;
4847
4848         asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4849         if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4850                 return -EINVAL;
4851
4852         if (asoc) {
4853                 info.sinfo_stream = asoc->default_stream;
4854                 info.sinfo_flags = asoc->default_flags;
4855                 info.sinfo_ppid = asoc->default_ppid;
4856                 info.sinfo_context = asoc->default_context;
4857                 info.sinfo_timetolive = asoc->default_timetolive;
4858         } else {
4859                 info.sinfo_stream = sp->default_stream;
4860                 info.sinfo_flags = sp->default_flags;
4861                 info.sinfo_ppid = sp->default_ppid;
4862                 info.sinfo_context = sp->default_context;
4863                 info.sinfo_timetolive = sp->default_timetolive;
4864         }
4865
4866         if (put_user(len, optlen))
4867                 return -EFAULT;
4868         if (copy_to_user(optval, &info, len))
4869                 return -EFAULT;
4870
4871         return 0;
4872 }
4873
4874 /*
4875  *
4876  * 7.1.5 SCTP_NODELAY
4877  *
4878  * Turn on/off any Nagle-like algorithm.  This means that packets are
4879  * generally sent as soon as possible and no unnecessary delays are
4880  * introduced, at the cost of more packets in the network.  Expects an
4881  * integer boolean flag.
4882  */
4883
4884 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4885                                    char __user *optval, int __user *optlen)
4886 {
4887         int val;
4888
4889         if (len < sizeof(int))
4890                 return -EINVAL;
4891
4892         len = sizeof(int);
4893         val = (sctp_sk(sk)->nodelay == 1);
4894         if (put_user(len, optlen))
4895                 return -EFAULT;
4896         if (copy_to_user(optval, &val, len))
4897                 return -EFAULT;
4898         return 0;
4899 }
4900
4901 /*
4902  *
4903  * 7.1.1 SCTP_RTOINFO
4904  *
4905  * The protocol parameters used to initialize and bound retransmission
4906  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4907  * and modify these parameters.
4908  * All parameters are time values, in milliseconds.  A value of 0, when
4909  * modifying the parameters, indicates that the current value should not
4910  * be changed.
4911  *
4912  */
4913 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4914                                 char __user *optval,
4915                                 int __user *optlen) {
4916         struct sctp_rtoinfo rtoinfo;
4917         struct sctp_association *asoc;
4918
4919         if (len < sizeof (struct sctp_rtoinfo))
4920                 return -EINVAL;
4921
4922         len = sizeof(struct sctp_rtoinfo);
4923
4924         if (copy_from_user(&rtoinfo, optval, len))
4925                 return -EFAULT;
4926
4927         asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4928
4929         if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4930                 return -EINVAL;
4931
4932         /* Values corresponding to the specific association. */
4933         if (asoc) {
4934                 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4935                 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4936                 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4937         } else {
4938                 /* Values corresponding to the endpoint. */
4939                 struct sctp_sock *sp = sctp_sk(sk);
4940
4941                 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4942                 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4943                 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4944         }
4945
4946         if (put_user(len, optlen))
4947                 return -EFAULT;
4948
4949         if (copy_to_user(optval, &rtoinfo, len))
4950                 return -EFAULT;
4951
4952         return 0;
4953 }
4954
4955 /*
4956  *
4957  * 7.1.2 SCTP_ASSOCINFO
4958  *
4959  * This option is used to tune the maximum retransmission attempts
4960  * of the association.
4961  * Returns an error if the new association retransmission value is
4962  * greater than the sum of the retransmission value  of the peer.
4963  * See [SCTP] for more information.
4964  *
4965  */
4966 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4967                                      char __user *optval,
4968                                      int __user *optlen)
4969 {
4970
4971         struct sctp_assocparams assocparams;
4972         struct sctp_association *asoc;
4973         struct list_head *pos;
4974         int cnt = 0;
4975
4976         if (len < sizeof (struct sctp_assocparams))
4977                 return -EINVAL;
4978
4979         len = sizeof(struct sctp_assocparams);
4980
4981         if (copy_from_user(&assocparams, optval, len))
4982                 return -EFAULT;
4983
4984         asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4985
4986         if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4987                 return -EINVAL;
4988
4989         /* Values correspoinding to the specific association */
4990         if (asoc) {
4991                 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4992                 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4993                 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4994                 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4995                                                 * 1000) +
4996                                                 (asoc->cookie_life.tv_usec
4997                                                 / 1000);
4998
4999                 list_for_each(pos, &asoc->peer.transport_addr_list) {
5000                         cnt ++;
5001                 }
5002
5003                 assocparams.sasoc_number_peer_destinations = cnt;
5004         } else {
5005                 /* Values corresponding to the endpoint */
5006                 struct sctp_sock *sp = sctp_sk(sk);
5007
5008                 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5009                 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5010                 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5011                 assocparams.sasoc_cookie_life =
5012                                         sp->assocparams.sasoc_cookie_life;
5013                 assocparams.sasoc_number_peer_destinations =
5014                                         sp->assocparams.
5015                                         sasoc_number_peer_destinations;
5016         }
5017
5018         if (put_user(len, optlen))
5019                 return -EFAULT;
5020
5021         if (copy_to_user(optval, &assocparams, len))
5022                 return -EFAULT;
5023
5024         return 0;
5025 }
5026
5027 /*
5028  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5029  *
5030  * This socket option is a boolean flag which turns on or off mapped V4
5031  * addresses.  If this option is turned on and the socket is type
5032  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5033  * If this option is turned off, then no mapping will be done of V4
5034  * addresses and a user will receive both PF_INET6 and PF_INET type
5035  * addresses on the socket.
5036  */
5037 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5038                                     char __user *optval, int __user *optlen)
5039 {
5040         int val;
5041         struct sctp_sock *sp = sctp_sk(sk);
5042
5043         if (len < sizeof(int))
5044                 return -EINVAL;
5045
5046         len = sizeof(int);
5047         val = sp->v4mapped;
5048         if (put_user(len, optlen))
5049                 return -EFAULT;
5050         if (copy_to_user(optval, &val, len))
5051                 return -EFAULT;
5052
5053         return 0;
5054 }
5055
5056 /*
5057  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
5058  * (chapter and verse is quoted at sctp_setsockopt_context())
5059  */
5060 static int sctp_getsockopt_context(struct sock *sk, int len,
5061                                    char __user *optval, int __user *optlen)
5062 {
5063         struct sctp_assoc_value params;
5064         struct sctp_sock *sp;
5065         struct sctp_association *asoc;
5066
5067         if (len < sizeof(struct sctp_assoc_value))
5068                 return -EINVAL;
5069
5070         len = sizeof(struct sctp_assoc_value);
5071
5072         if (copy_from_user(&params, optval, len))
5073                 return -EFAULT;
5074
5075         sp = sctp_sk(sk);
5076
5077         if (params.assoc_id != 0) {
5078                 asoc = sctp_id2assoc(sk, params.assoc_id);
5079                 if (!asoc)
5080                         return -EINVAL;
5081                 params.assoc_value = asoc->default_rcv_context;
5082         } else {
5083                 params.assoc_value = sp->default_rcv_context;
5084         }
5085
5086         if (put_user(len, optlen))
5087                 return -EFAULT;
5088         if (copy_to_user(optval, &params, len))
5089                 return -EFAULT;
5090
5091         return 0;
5092 }
5093
5094 /*
5095  * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
5096  *
5097  * This socket option specifies the maximum size to put in any outgoing
5098  * SCTP chunk.  If a message is larger than this size it will be
5099  * fragmented by SCTP into the specified size.  Note that the underlying
5100  * SCTP implementation may fragment into smaller sized chunks when the
5101  * PMTU of the underlying association is smaller than the value set by
5102  * the user.
5103  */
5104 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5105                                   char __user *optval, int __user *optlen)
5106 {
5107         int val;
5108
5109         if (len < sizeof(int))
5110                 return -EINVAL;
5111
5112         len = sizeof(int);
5113
5114         val = sctp_sk(sk)->user_frag;
5115         if (put_user(len, optlen))
5116                 return -EFAULT;
5117         if (copy_to_user(optval, &val, len))
5118                 return -EFAULT;
5119
5120         return 0;
5121 }
5122
5123 /*
5124  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5125  * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5126  */
5127 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5128                                                char __user *optval, int __user *optlen)
5129 {
5130         int val;
5131
5132         if (len < sizeof(int))
5133                 return -EINVAL;
5134
5135         len = sizeof(int);
5136
5137         val = sctp_sk(sk)->frag_interleave;
5138         if (put_user(len, optlen))
5139                 return -EFAULT;
5140         if (copy_to_user(optval, &val, len))
5141                 return -EFAULT;
5142
5143         return 0;
5144 }
5145
5146 /*
5147  * 7.1.25.  Set or Get the sctp partial delivery point
5148  * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5149  */
5150 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5151                                                   char __user *optval,
5152                                                   int __user *optlen)
5153 {
5154         u32 val;
5155
5156         if (len < sizeof(u32))
5157                 return -EINVAL;
5158
5159         len = sizeof(u32);
5160
5161         val = sctp_sk(sk)->pd_point;
5162         if (put_user(len, optlen))
5163                 return -EFAULT;
5164         if (copy_to_user(optval, &val, len))
5165                 return -EFAULT;
5166
5167         return -ENOTSUPP;
5168 }
5169
5170 /*
5171  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
5172  * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5173  */
5174 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5175                                     char __user *optval,
5176                                     int __user *optlen)
5177 {
5178         struct sctp_assoc_value params;
5179         struct sctp_sock *sp;
5180         struct sctp_association *asoc;
5181
5182         if (len < sizeof(int))
5183                 return -EINVAL;
5184
5185         if (len == sizeof(int)) {
5186                 printk(KERN_WARNING
5187                    "SCTP: Use of int in max_burst socket option deprecated\n");
5188                 printk(KERN_WARNING
5189                    "SCTP: Use struct sctp_assoc_value instead\n");
5190                 params.assoc_id = 0;
5191         } else if (len == sizeof (struct sctp_assoc_value)) {
5192                 if (copy_from_user(&params, optval, len))
5193                         return -EFAULT;
5194         } else
5195                 return -EINVAL;
5196
5197         sp = sctp_sk(sk);
5198
5199         if (params.assoc_id != 0) {
5200                 asoc = sctp_id2assoc(sk, params.assoc_id);
5201                 if (!asoc)
5202                         return -EINVAL;
5203                 params.assoc_value = asoc->max_burst;
5204         } else
5205                 params.assoc_value = sp->max_burst;
5206
5207         if (len == sizeof(int)) {
5208                 if (copy_to_user(optval, &params.assoc_value, len))
5209                         return -EFAULT;
5210         } else {
5211                 if (copy_to_user(optval, &params, len))
5212                         return -EFAULT;
5213         }
5214
5215         return 0;
5216
5217 }
5218
5219 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5220                                     char __user *optval, int __user *optlen)
5221 {
5222         struct sctp_hmacalgo  __user *p = (void __user *)optval;
5223         struct sctp_hmac_algo_param *hmacs;
5224         __u16 data_len = 0;
5225         u32 num_idents;
5226
5227         if (!sctp_auth_enable)
5228                 return -EACCES;
5229
5230         hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5231         data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5232
5233         if (len < sizeof(struct sctp_hmacalgo) + data_len)
5234                 return -EINVAL;
5235
5236         len = sizeof(struct sctp_hmacalgo) + data_len;
5237         num_idents = data_len / sizeof(u16);
5238
5239         if (put_user(len, optlen))
5240                 return -EFAULT;
5241         if (put_user(num_idents, &p->shmac_num_idents))
5242                 return -EFAULT;
5243         if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5244                 return -EFAULT;
5245         return 0;
5246 }
5247
5248 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5249                                     char __user *optval, int __user *optlen)
5250 {
5251         struct sctp_authkeyid val;
5252         struct sctp_association *asoc;
5253
5254         if (!sctp_auth_enable)
5255                 return -EACCES;
5256
5257         if (len < sizeof(struct sctp_authkeyid))
5258                 return -EINVAL;
5259         if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5260                 return -EFAULT;
5261
5262         asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5263         if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5264                 return -EINVAL;
5265
5266         if (asoc)
5267                 val.scact_keynumber = asoc->active_key_id;
5268         else
5269                 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5270
5271         len = sizeof(struct sctp_authkeyid);
5272         if (put_user(len, optlen))
5273                 return -EFAULT;
5274         if (copy_to_user(optval, &val, len))
5275                 return -EFAULT;
5276
5277         return 0;
5278 }
5279
5280 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5281                                     char __user *optval, int __user *optlen)
5282 {
5283         struct sctp_authchunks __user *p = (void __user *)optval;
5284         struct sctp_authchunks val;
5285         struct sctp_association *asoc;
5286         struct sctp_chunks_param *ch;
5287         u32    num_chunks = 0;
5288         char __user *to;
5289
5290         if (!sctp_auth_enable)
5291                 return -EACCES;
5292
5293         if (len < sizeof(struct sctp_authchunks))
5294                 return -EINVAL;
5295
5296         if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5297                 return -EFAULT;
5298
5299         to = p->gauth_chunks;
5300         asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5301         if (!asoc)
5302                 return -EINVAL;
5303
5304         ch = asoc->peer.peer_chunks;
5305         if (!ch)
5306                 goto num;
5307
5308         /* See if the user provided enough room for all the data */
5309         num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5310         if (len < num_chunks)
5311                 return -EINVAL;
5312
5313         if (copy_to_user(to, ch->chunks, num_chunks))
5314                 return -EFAULT;
5315 num:
5316         len = sizeof(struct sctp_authchunks) + num_chunks;
5317         if (put_user(len, optlen)) return -EFAULT;
5318         if (put_user(num_chunks, &p->gauth_number_of_chunks))
5319                 return -EFAULT;
5320         return 0;
5321 }
5322
5323 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5324                                     char __user *optval, int __user *optlen)
5325 {
5326         struct sctp_authchunks __user *p = (void __user *)optval;
5327         struct sctp_authchunks val;
5328         struct sctp_association *asoc;
5329         struct sctp_chunks_param *ch;
5330         u32    num_chunks = 0;
5331         char __user *to;
5332
5333         if (!sctp_auth_enable)
5334                 return -EACCES;
5335
5336         if (len < sizeof(struct sctp_authchunks))
5337                 return -EINVAL;
5338
5339         if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5340                 return -EFAULT;
5341
5342         to = p->gauth_chunks;
5343         asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5344         if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5345                 return -EINVAL;
5346
5347         if (asoc)
5348                 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5349         else
5350                 ch = sctp_sk(sk)->ep->auth_chunk_list;
5351
5352         if (!ch)
5353                 goto num;
5354
5355         num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5356         if (len < sizeof(struct sctp_authchunks) + num_chunks)
5357                 return -EINVAL;
5358
5359         if (copy_to_user(to, ch->chunks, num_chunks))
5360                 return -EFAULT;
5361 num:
5362         len = sizeof(struct sctp_authchunks) + num_chunks;
5363         if (put_user(len, optlen))
5364                 return -EFAULT;
5365         if (put_user(num_chunks, &p->gauth_number_of_chunks))
5366                 return -EFAULT;
5367
5368         return 0;
5369 }
5370
5371 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5372                                 char __user *optval, int __user *optlen)
5373 {
5374         int retval = 0;
5375         int len;
5376
5377         SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5378                           sk, optname);
5379
5380         /* I can hardly begin to describe how wrong this is.  This is
5381          * so broken as to be worse than useless.  The API draft
5382          * REALLY is NOT helpful here...  I am not convinced that the
5383          * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5384          * are at all well-founded.
5385          */
5386         if (level != SOL_SCTP) {
5387                 struct sctp_af *af = sctp_sk(sk)->pf->af;
5388
5389                 retval = af->getsockopt(sk, level, optname, optval, optlen);
5390                 return retval;
5391         }
5392
5393         if (get_user(len, optlen))
5394                 return -EFAULT;
5395
5396         sctp_lock_sock(sk);
5397
5398         switch (optname) {
5399         case SCTP_STATUS:
5400                 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5401                 break;
5402         case SCTP_DISABLE_FRAGMENTS:
5403                 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5404                                                            optlen);
5405                 break;
5406         case SCTP_EVENTS:
5407                 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5408                 break;
5409         case SCTP_AUTOCLOSE:
5410                 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5411                 break;
5412         case SCTP_SOCKOPT_PEELOFF:
5413                 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5414                 break;
5415         case SCTP_PEER_ADDR_PARAMS:
5416                 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5417                                                           optlen);
5418                 break;
5419         case SCTP_DELAYED_ACK:
5420                 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5421                                                           optlen);
5422                 break;
5423         case SCTP_INITMSG:
5424                 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5425                 break;
5426         case SCTP_GET_PEER_ADDRS_NUM_OLD:
5427                 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5428                                                             optlen);
5429                 break;
5430         case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5431                 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5432                                                              optlen);
5433                 break;
5434         case SCTP_GET_PEER_ADDRS_OLD:
5435                 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5436                                                         optlen);
5437                 break;
5438         case SCTP_GET_LOCAL_ADDRS_OLD:
5439                 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5440                                                          optlen);
5441                 break;
5442         case SCTP_GET_PEER_ADDRS:
5443                 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5444                                                     optlen);
5445                 break;
5446         case SCTP_GET_LOCAL_ADDRS:
5447                 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5448                                                      optlen);
5449                 break;
5450         case SCTP_DEFAULT_SEND_PARAM:
5451                 retval = sctp_getsockopt_default_send_param(sk, len,
5452                                                             optval, optlen);
5453                 break;
5454         case SCTP_PRIMARY_ADDR:
5455                 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5456                 break;
5457         case SCTP_NODELAY:
5458                 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5459                 break;
5460         case SCTP_RTOINFO:
5461                 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5462                 break;
5463         case SCTP_ASSOCINFO:
5464                 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5465                 break;
5466         case SCTP_I_WANT_MAPPED_V4_ADDR:
5467                 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5468                 break;
5469         case SCTP_MAXSEG:
5470                 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5471                 break;
5472         case SCTP_GET_PEER_ADDR_INFO:
5473                 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5474                                                         optlen);
5475                 break;
5476         case SCTP_ADAPTATION_LAYER:
5477                 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5478                                                         optlen);
5479                 break;
5480         case SCTP_CONTEXT:
5481                 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5482                 break;
5483         case SCTP_FRAGMENT_INTERLEAVE:
5484                 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5485                                                              optlen);
5486                 break;
5487         case SCTP_PARTIAL_DELIVERY_POINT:
5488                 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5489                                                                 optlen);
5490                 break;
5491         case SCTP_MAX_BURST:
5492                 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5493                 break;
5494         case SCTP_AUTH_KEY:
5495         case SCTP_AUTH_CHUNK:
5496         case SCTP_AUTH_DELETE_KEY:
5497                 retval = -EOPNOTSUPP;
5498                 break;
5499         case SCTP_HMAC_IDENT:
5500                 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5501                 break;
5502         case SCTP_AUTH_ACTIVE_KEY:
5503                 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5504                 break;
5505         case SCTP_PEER_AUTH_CHUNKS:
5506                 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5507                                                         optlen);
5508                 break;
5509         case SCTP_LOCAL_AUTH_CHUNKS:
5510                 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5511                                                         optlen);
5512                 break;
5513         default:
5514                 retval = -ENOPROTOOPT;
5515                 break;
5516         }
5517
5518         sctp_release_sock(sk);
5519         return retval;
5520 }
5521
5522 static void sctp_hash(struct sock *sk)
5523 {
5524         /* STUB */
5525 }
5526
5527 static void sctp_unhash(struct sock *sk)
5528 {
5529         /* STUB */
5530 }
5531
5532 /* Check if port is acceptable.  Possibly find first available port.
5533  *
5534  * The port hash table (contained in the 'global' SCTP protocol storage
5535  * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5536  * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5537  * list (the list number is the port number hashed out, so as you
5538  * would expect from a hash function, all the ports in a given list have
5539  * such a number that hashes out to the same list number; you were
5540  * expecting that, right?); so each list has a set of ports, with a
5541  * link to the socket (struct sock) that uses it, the port number and
5542  * a fastreuse flag (FIXME: NPI ipg).
5543  */
5544 static struct sctp_bind_bucket *sctp_bucket_create(
5545         struct sctp_bind_hashbucket *head, unsigned short snum);
5546
5547 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5548 {
5549         struct sctp_bind_hashbucket *head; /* hash list */
5550         struct sctp_bind_bucket *pp; /* hash list port iterator */
5551         struct hlist_node *node;
5552         unsigned short snum;
5553         int ret;
5554
5555         snum = ntohs(addr->v4.sin_port);
5556
5557         SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5558         sctp_local_bh_disable();
5559
5560         if (snum == 0) {
5561                 /* Search for an available port. */
5562                 int low, high, remaining, index;
5563                 unsigned int rover;
5564
5565                 inet_get_local_port_range(&low, &high);
5566                 remaining = (high - low) + 1;
5567                 rover = net_random() % remaining + low;
5568
5569                 do {
5570                         rover++;
5571                         if ((rover < low) || (rover > high))
5572                                 rover = low;
5573                         index = sctp_phashfn(rover);
5574                         head = &sctp_port_hashtable[index];
5575                         sctp_spin_lock(&head->lock);
5576                         sctp_for_each_hentry(pp, node, &head->chain)
5577                                 if (pp->port == rover)
5578                                         goto next;
5579                         break;
5580                 next:
5581                         sctp_spin_unlock(&head->lock);
5582                 } while (--remaining > 0);
5583
5584                 /* Exhausted local port range during search? */
5585                 ret = 1;
5586                 if (remaining <= 0)
5587                         goto fail;
5588
5589                 /* OK, here is the one we will use.  HEAD (the port
5590                  * hash table list entry) is non-NULL and we hold it's
5591                  * mutex.
5592                  */
5593                 snum = rover;
5594         } else {
5595                 /* We are given an specific port number; we verify
5596                  * that it is not being used. If it is used, we will
5597                  * exahust the search in the hash list corresponding
5598                  * to the port number (snum) - we detect that with the
5599                  * port iterator, pp being NULL.
5600                  */
5601                 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5602                 sctp_spin_lock(&head->lock);
5603                 sctp_for_each_hentry(pp, node, &head->chain) {
5604                         if (pp->port == snum)
5605                                 goto pp_found;
5606                 }
5607         }
5608         pp = NULL;
5609         goto pp_not_found;
5610 pp_found:
5611         if (!hlist_empty(&pp->owner)) {
5612                 /* We had a port hash table hit - there is an
5613                  * available port (pp != NULL) and it is being
5614                  * used by other socket (pp->owner not empty); that other
5615                  * socket is going to be sk2.
5616                  */
5617                 int reuse = sk->sk_reuse;
5618                 struct sock *sk2;
5619                 struct hlist_node *node;
5620
5621                 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5622                 if (pp->fastreuse && sk->sk_reuse &&
5623                         sk->sk_state != SCTP_SS_LISTENING)
5624                         goto success;
5625
5626                 /* Run through the list of sockets bound to the port
5627                  * (pp->port) [via the pointers bind_next and
5628                  * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5629                  * we get the endpoint they describe and run through
5630                  * the endpoint's list of IP (v4 or v6) addresses,
5631                  * comparing each of the addresses with the address of
5632                  * the socket sk. If we find a match, then that means
5633                  * that this port/socket (sk) combination are already
5634                  * in an endpoint.
5635                  */
5636                 sk_for_each_bound(sk2, node, &pp->owner) {
5637                         struct sctp_endpoint *ep2;
5638                         ep2 = sctp_sk(sk2)->ep;
5639
5640                         if (sk == sk2 ||
5641                             (reuse && sk2->sk_reuse &&
5642                              sk2->sk_state != SCTP_SS_LISTENING))
5643                                 continue;
5644
5645                         if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5646                                                  sctp_sk(sk2), sctp_sk(sk))) {
5647                                 ret = (long)sk2;
5648                                 goto fail_unlock;
5649                         }
5650                 }
5651                 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5652         }
5653 pp_not_found:
5654         /* If there was a hash table miss, create a new port.  */
5655         ret = 1;
5656         if (!pp && !(pp = sctp_bucket_create(head, snum)))
5657                 goto fail_unlock;
5658
5659         /* In either case (hit or miss), make sure fastreuse is 1 only
5660          * if sk->sk_reuse is too (that is, if the caller requested
5661          * SO_REUSEADDR on this socket -sk-).
5662          */
5663         if (hlist_empty(&pp->owner)) {
5664                 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5665                         pp->fastreuse = 1;
5666                 else
5667                         pp->fastreuse = 0;
5668         } else if (pp->fastreuse &&
5669                 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5670                 pp->fastreuse = 0;
5671
5672         /* We are set, so fill up all the data in the hash table
5673          * entry, tie the socket list information with the rest of the
5674          * sockets FIXME: Blurry, NPI (ipg).
5675          */
5676 success:
5677         if (!sctp_sk(sk)->bind_hash) {
5678                 inet_sk(sk)->num = snum;
5679                 sk_add_bind_node(sk, &pp->owner);
5680                 sctp_sk(sk)->bind_hash = pp;
5681         }
5682         ret = 0;
5683
5684 fail_unlock:
5685         sctp_spin_unlock(&head->lock);
5686
5687 fail:
5688         sctp_local_bh_enable();
5689         return ret;
5690 }
5691
5692 /* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
5693  * port is requested.
5694  */
5695 static int sctp_get_port(struct sock *sk, unsigned short snum)
5696 {
5697         long ret;
5698         union sctp_addr addr;
5699         struct sctp_af *af = sctp_sk(sk)->pf->af;
5700
5701         /* Set up a dummy address struct from the sk. */
5702         af->from_sk(&addr, sk);
5703         addr.v4.sin_port = htons(snum);
5704
5705         /* Note: sk->sk_num gets filled in if ephemeral port request. */
5706         ret = sctp_get_port_local(sk, &addr);
5707
5708         return (ret ? 1 : 0);
5709 }
5710
5711 /*
5712  * 3.1.3 listen() - UDP Style Syntax
5713  *
5714  *   By default, new associations are not accepted for UDP style sockets.
5715  *   An application uses listen() to mark a socket as being able to
5716  *   accept new associations.
5717  */
5718 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5719 {
5720         struct sctp_sock *sp = sctp_sk(sk);
5721         struct sctp_endpoint *ep = sp->ep;
5722
5723         /* Only UDP style sockets that are not peeled off are allowed to
5724          * listen().
5725          */
5726         if (!sctp_style(sk, UDP))
5727                 return -EINVAL;
5728
5729         /* If backlog is zero, disable listening. */
5730         if (!backlog) {
5731                 if (sctp_sstate(sk, CLOSED))
5732                         return 0;
5733
5734                 sctp_unhash_endpoint(ep);
5735                 sk->sk_state = SCTP_SS_CLOSED;
5736                 return 0;
5737         }
5738
5739         /* Return if we are already listening. */
5740         if (sctp_sstate(sk, LISTENING))
5741                 return 0;
5742
5743         /*
5744          * If a bind() or sctp_bindx() is not called prior to a listen()
5745          * call that allows new associations to be accepted, the system
5746          * picks an ephemeral port and will choose an address set equivalent
5747          * to binding with a wildcard address.
5748          *
5749          * This is not currently spelled out in the SCTP sockets
5750          * extensions draft, but follows the practice as seen in TCP
5751          * sockets.
5752          *
5753          * Additionally, turn off fastreuse flag since we are not listening
5754          */
5755         sk->sk_state = SCTP_SS_LISTENING;
5756         if (!ep->base.bind_addr.port) {
5757                 if (sctp_autobind(sk))
5758                         return -EAGAIN;
5759         } else {
5760                 if (sctp_get_port(sk, inet_sk(sk)->num)) {
5761                         sk->sk_state = SCTP_SS_CLOSED;
5762                         return -EADDRINUSE;
5763                 }
5764                 sctp_sk(sk)->bind_hash->fastreuse = 0;
5765         }
5766
5767         sctp_hash_endpoint(ep);
5768         return 0;
5769 }
5770
5771 /*
5772  * 4.1.3 listen() - TCP Style Syntax
5773  *
5774  *   Applications uses listen() to ready the SCTP endpoint for accepting
5775  *   inbound associations.
5776  */
5777 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5778 {
5779         struct sctp_sock *sp = sctp_sk(sk);
5780         struct sctp_endpoint *ep = sp->ep;
5781
5782         /* If backlog is zero, disable listening. */
5783         if (!backlog) {
5784                 if (sctp_sstate(sk, CLOSED))
5785                         return 0;
5786
5787                 sctp_unhash_endpoint(ep);
5788                 sk->sk_state = SCTP_SS_CLOSED;
5789                 return 0;
5790         }
5791
5792         if (sctp_sstate(sk, LISTENING))
5793                 return 0;
5794
5795         /*
5796          * If a bind() or sctp_bindx() is not called prior to a listen()
5797          * call that allows new associations to be accepted, the system
5798          * picks an ephemeral port and will choose an address set equivalent
5799          * to binding with a wildcard address.
5800          *
5801          * This is not currently spelled out in the SCTP sockets
5802          * extensions draft, but follows the practice as seen in TCP
5803          * sockets.
5804          */
5805         sk->sk_state = SCTP_SS_LISTENING;
5806         if (!ep->base.bind_addr.port) {
5807                 if (sctp_autobind(sk))
5808                         return -EAGAIN;
5809         } else
5810                 sctp_sk(sk)->bind_hash->fastreuse = 0;
5811
5812         sk->sk_max_ack_backlog = backlog;
5813         sctp_hash_endpoint(ep);
5814         return 0;
5815 }
5816
5817 /*
5818  *  Move a socket to LISTENING state.
5819  */
5820 int sctp_inet_listen(struct socket *sock, int backlog)
5821 {
5822         struct sock *sk = sock->sk;
5823         struct crypto_hash *tfm = NULL;
5824         int err = -EINVAL;
5825
5826         if (unlikely(backlog < 0))
5827                 goto out;
5828
5829         sctp_lock_sock(sk);
5830
5831         if (sock->state != SS_UNCONNECTED)
5832                 goto out;
5833
5834         /* Allocate HMAC for generating cookie. */
5835         if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5836                 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5837                 if (IS_ERR(tfm)) {
5838                         if (net_ratelimit()) {
5839                                 printk(KERN_INFO
5840                                        "SCTP: failed to load transform for %s: %ld\n",
5841                                         sctp_hmac_alg, PTR_ERR(tfm));
5842                         }
5843                         err = -ENOSYS;
5844                         goto out;
5845                 }
5846         }
5847
5848         switch (sock->type) {
5849         case SOCK_SEQPACKET:
5850                 err = sctp_seqpacket_listen(sk, backlog);
5851                 break;
5852         case SOCK_STREAM:
5853                 err = sctp_stream_listen(sk, backlog);
5854                 break;
5855         default:
5856                 break;
5857         }
5858
5859         if (err)
5860                 goto cleanup;
5861
5862         /* Store away the transform reference. */
5863         if (!sctp_sk(sk)->hmac)
5864                 sctp_sk(sk)->hmac = tfm;
5865 out:
5866         sctp_release_sock(sk);
5867         return err;
5868 cleanup:
5869         crypto_free_hash(tfm);
5870         goto out;
5871 }
5872
5873 /*
5874  * This function is done by modeling the current datagram_poll() and the
5875  * tcp_poll().  Note that, based on these implementations, we don't
5876  * lock the socket in this function, even though it seems that,
5877  * ideally, locking or some other mechanisms can be used to ensure
5878  * the integrity of the counters (sndbuf and wmem_alloc) used
5879  * in this place.  We assume that we don't need locks either until proven
5880  * otherwise.
5881  *
5882  * Another thing to note is that we include the Async I/O support
5883  * here, again, by modeling the current TCP/UDP code.  We don't have
5884  * a good way to test with it yet.
5885  */
5886 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5887 {
5888         struct sock *sk = sock->sk;
5889         struct sctp_sock *sp = sctp_sk(sk);
5890         unsigned int mask;
5891
5892         poll_wait(file, sk->sk_sleep, wait);
5893
5894         /* A TCP-style listening socket becomes readable when the accept queue
5895          * is not empty.
5896          */
5897         if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5898                 return (!list_empty(&sp->ep->asocs)) ?
5899                         (POLLIN | POLLRDNORM) : 0;
5900
5901         mask = 0;
5902
5903         /* Is there any exceptional events?  */
5904         if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5905                 mask |= POLLERR;
5906         if (sk->sk_shutdown & RCV_SHUTDOWN)
5907                 mask |= POLLRDHUP;
5908         if (sk->sk_shutdown == SHUTDOWN_MASK)
5909                 mask |= POLLHUP;
5910
5911         /* Is it readable?  Reconsider this code with TCP-style support.  */
5912         if (!skb_queue_empty(&sk->sk_receive_queue) ||
5913             (sk->sk_shutdown & RCV_SHUTDOWN))
5914                 mask |= POLLIN | POLLRDNORM;
5915
5916         /* The association is either gone or not ready.  */
5917         if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5918                 return mask;
5919
5920         /* Is it writable?  */
5921         if (sctp_writeable(sk)) {
5922                 mask |= POLLOUT | POLLWRNORM;
5923         } else {
5924                 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5925                 /*
5926                  * Since the socket is not locked, the buffer
5927                  * might be made available after the writeable check and
5928                  * before the bit is set.  This could cause a lost I/O
5929                  * signal.  tcp_poll() has a race breaker for this race
5930                  * condition.  Based on their implementation, we put
5931                  * in the following code to cover it as well.
5932                  */
5933                 if (sctp_writeable(sk))
5934                         mask |= POLLOUT | POLLWRNORM;
5935         }
5936         return mask;
5937 }
5938
5939 /********************************************************************
5940  * 2nd Level Abstractions
5941  ********************************************************************/
5942
5943 static struct sctp_bind_bucket *sctp_bucket_create(
5944         struct sctp_bind_hashbucket *head, unsigned short snum)
5945 {
5946         struct sctp_bind_bucket *pp;
5947
5948         pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5949         if (pp) {
5950                 SCTP_DBG_OBJCNT_INC(bind_bucket);
5951                 pp->port = snum;
5952                 pp->fastreuse = 0;
5953                 INIT_HLIST_HEAD(&pp->owner);
5954                 hlist_add_head(&pp->node, &head->chain);
5955         }
5956         return pp;
5957 }
5958
5959 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5960 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5961 {
5962         if (pp && hlist_empty(&pp->owner)) {
5963                 __hlist_del(&pp->node);
5964                 kmem_cache_free(sctp_bucket_cachep, pp);
5965                 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5966         }
5967 }
5968
5969 /* Release this socket's reference to a local port.  */
5970 static inline void __sctp_put_port(struct sock *sk)
5971 {
5972         struct sctp_bind_hashbucket *head =
5973                 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5974         struct sctp_bind_bucket *pp;
5975
5976         sctp_spin_lock(&head->lock);
5977         pp = sctp_sk(sk)->bind_hash;
5978         __sk_del_bind_node(sk);
5979         sctp_sk(sk)->bind_hash = NULL;
5980         inet_sk(sk)->num = 0;
5981         sctp_bucket_destroy(pp);
5982         sctp_spin_unlock(&head->lock);
5983 }
5984
5985 void sctp_put_port(struct sock *sk)
5986 {
5987         sctp_local_bh_disable();
5988         __sctp_put_port(sk);
5989         sctp_local_bh_enable();
5990 }
5991
5992 /*
5993  * The system picks an ephemeral port and choose an address set equivalent
5994  * to binding with a wildcard address.
5995  * One of those addresses will be the primary address for the association.
5996  * This automatically enables the multihoming capability of SCTP.
5997  */
5998 static int sctp_autobind(struct sock *sk)
5999 {
6000         union sctp_addr autoaddr;
6001         struct sctp_af *af;
6002         __be16 port;
6003
6004         /* Initialize a local sockaddr structure to INADDR_ANY. */
6005         af = sctp_sk(sk)->pf->af;
6006
6007         port = htons(inet_sk(sk)->num);
6008         af->inaddr_any(&autoaddr, port);
6009
6010         return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6011 }
6012
6013 /* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
6014  *
6015  * From RFC 2292
6016  * 4.2 The cmsghdr Structure *
6017  *
6018  * When ancillary data is sent or received, any number of ancillary data
6019  * objects can be specified by the msg_control and msg_controllen members of
6020  * the msghdr structure, because each object is preceded by
6021  * a cmsghdr structure defining the object's length (the cmsg_len member).
6022  * Historically Berkeley-derived implementations have passed only one object
6023  * at a time, but this API allows multiple objects to be
6024  * passed in a single call to sendmsg() or recvmsg(). The following example
6025  * shows two ancillary data objects in a control buffer.
6026  *
6027  *   |<--------------------------- msg_controllen -------------------------->|
6028  *   |                                                                       |
6029  *
6030  *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
6031  *
6032  *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6033  *   |                                   |                                   |
6034  *
6035  *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
6036  *
6037  *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
6038  *   |                                |  |                                |  |
6039  *
6040  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6041  *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
6042  *
6043  *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
6044  *
6045  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6046  *    ^
6047  *    |
6048  *
6049  * msg_control
6050  * points here
6051  */
6052 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6053                                   sctp_cmsgs_t *cmsgs)
6054 {
6055         struct cmsghdr *cmsg;
6056         struct msghdr *my_msg = (struct msghdr *)msg;
6057
6058         for (cmsg = CMSG_FIRSTHDR(msg);
6059              cmsg != NULL;
6060              cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6061                 if (!CMSG_OK(my_msg, cmsg))
6062                         return -EINVAL;
6063
6064                 /* Should we parse this header or ignore?  */
6065                 if (cmsg->cmsg_level != IPPROTO_SCTP)
6066                         continue;
6067
6068                 /* Strictly check lengths following example in SCM code.  */
6069                 switch (cmsg->cmsg_type) {
6070                 case SCTP_INIT:
6071                         /* SCTP Socket API Extension
6072                          * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6073                          *
6074                          * This cmsghdr structure provides information for
6075                          * initializing new SCTP associations with sendmsg().
6076                          * The SCTP_INITMSG socket option uses this same data
6077                          * structure.  This structure is not used for
6078                          * recvmsg().
6079                          *
6080                          * cmsg_level    cmsg_type      cmsg_data[]
6081                          * ------------  ------------   ----------------------
6082                          * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
6083                          */
6084                         if (cmsg->cmsg_len !=
6085                             CMSG_LEN(sizeof(struct sctp_initmsg)))
6086                                 return -EINVAL;
6087                         cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6088                         break;
6089
6090                 case SCTP_SNDRCV:
6091                         /* SCTP Socket API Extension
6092                          * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6093                          *
6094                          * This cmsghdr structure specifies SCTP options for
6095                          * sendmsg() and describes SCTP header information
6096                          * about a received message through recvmsg().
6097                          *
6098                          * cmsg_level    cmsg_type      cmsg_data[]
6099                          * ------------  ------------   ----------------------
6100                          * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
6101                          */
6102                         if (cmsg->cmsg_len !=
6103                             CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6104                                 return -EINVAL;
6105
6106                         cmsgs->info =
6107                                 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6108
6109                         /* Minimally, validate the sinfo_flags. */
6110                         if (cmsgs->info->sinfo_flags &
6111                             ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6112                               SCTP_ABORT | SCTP_EOF))
6113                                 return -EINVAL;
6114                         break;
6115
6116                 default:
6117                         return -EINVAL;
6118                 }
6119         }
6120         return 0;
6121 }
6122
6123 /*
6124  * Wait for a packet..
6125  * Note: This function is the same function as in core/datagram.c
6126  * with a few modifications to make lksctp work.
6127  */
6128 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6129 {
6130         int error;
6131         DEFINE_WAIT(wait);
6132
6133         prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6134
6135         /* Socket errors? */
6136         error = sock_error(sk);
6137         if (error)
6138                 goto out;
6139
6140         if (!skb_queue_empty(&sk->sk_receive_queue))
6141                 goto ready;
6142
6143         /* Socket shut down?  */
6144         if (sk->sk_shutdown & RCV_SHUTDOWN)
6145                 goto out;
6146
6147         /* Sequenced packets can come disconnected.  If so we report the
6148          * problem.
6149          */
6150         error = -ENOTCONN;
6151
6152         /* Is there a good reason to think that we may receive some data?  */
6153         if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6154                 goto out;
6155
6156         /* Handle signals.  */
6157         if (signal_pending(current))
6158                 goto interrupted;
6159
6160         /* Let another process have a go.  Since we are going to sleep
6161          * anyway.  Note: This may cause odd behaviors if the message
6162          * does not fit in the user's buffer, but this seems to be the
6163          * only way to honor MSG_DONTWAIT realistically.
6164          */
6165         sctp_release_sock(sk);
6166         *timeo_p = schedule_timeout(*timeo_p);
6167         sctp_lock_sock(sk);
6168
6169 ready:
6170         finish_wait(sk->sk_sleep, &wait);
6171         return 0;
6172
6173 interrupted:
6174         error = sock_intr_errno(*timeo_p);
6175
6176 out:
6177         finish_wait(sk->sk_sleep, &wait);
6178         *err = error;
6179         return error;
6180 }
6181
6182 /* Receive a datagram.
6183  * Note: This is pretty much the same routine as in core/datagram.c
6184  * with a few changes to make lksctp work.
6185  */
6186 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6187                                               int noblock, int *err)
6188 {
6189         int error;
6190         struct sk_buff *skb;
6191         long timeo;
6192
6193         timeo = sock_rcvtimeo(sk, noblock);
6194
6195         SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6196                           timeo, MAX_SCHEDULE_TIMEOUT);
6197
6198         do {
6199                 /* Again only user level code calls this function,
6200                  * so nothing interrupt level
6201                  * will suddenly eat the receive_queue.
6202                  *
6203                  *  Look at current nfs client by the way...
6204                  *  However, this function was corrent in any case. 8)
6205                  */
6206                 if (flags & MSG_PEEK) {
6207                         spin_lock_bh(&sk->sk_receive_queue.lock);
6208                         skb = skb_peek(&sk->sk_receive_queue);
6209                         if (skb)
6210                                 atomic_inc(&skb->users);
6211                         spin_unlock_bh(&sk->sk_receive_queue.lock);
6212                 } else {
6213                         skb = skb_dequeue(&sk->sk_receive_queue);
6214                 }
6215
6216                 if (skb)
6217                         return skb;
6218
6219                 /* Caller is allowed not to check sk->sk_err before calling. */
6220                 error = sock_error(sk);
6221                 if (error)
6222                         goto no_packet;
6223
6224                 if (sk->sk_shutdown & RCV_SHUTDOWN)
6225                         break;
6226
6227                 /* User doesn't want to wait.  */
6228                 error = -EAGAIN;
6229                 if (!timeo)
6230                         goto no_packet;
6231         } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6232
6233         return NULL;
6234
6235 no_packet:
6236         *err = error;
6237         return NULL;
6238 }
6239
6240 /* If sndbuf has changed, wake up per association sndbuf waiters.  */
6241 static void __sctp_write_space(struct sctp_association *asoc)
6242 {
6243         struct sock *sk = asoc->base.sk;
6244         struct socket *sock = sk->sk_socket;
6245
6246         if ((sctp_wspace(asoc) > 0) && sock) {
6247                 if (waitqueue_active(&asoc->wait))
6248                         wake_up_interruptible(&asoc->wait);
6249
6250                 if (sctp_writeable(sk)) {
6251                         if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6252                                 wake_up_interruptible(sk->sk_sleep);
6253
6254                         /* Note that we try to include the Async I/O support
6255                          * here by modeling from the current TCP/UDP code.
6256                          * We have not tested with it yet.
6257                          */
6258                         if (sock->fasync_list &&
6259                             !(sk->sk_shutdown & SEND_SHUTDOWN))
6260                                 sock_wake_async(sock,
6261                                                 SOCK_WAKE_SPACE, POLL_OUT);
6262                 }
6263         }
6264 }
6265
6266 /* Do accounting for the sndbuf space.
6267  * Decrement the used sndbuf space of the corresponding association by the
6268  * data size which was just transmitted(freed).
6269  */
6270 static void sctp_wfree(struct sk_buff *skb)
6271 {
6272         struct sctp_association *asoc;
6273         struct sctp_chunk *chunk;
6274         struct sock *sk;
6275
6276         /* Get the saved chunk pointer.  */
6277         chunk = *((struct sctp_chunk **)(skb->cb));
6278         asoc = chunk->asoc;
6279         sk = asoc->base.sk;
6280         asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6281                                 sizeof(struct sk_buff) +
6282                                 sizeof(struct sctp_chunk);
6283
6284         atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6285
6286         /*
6287          * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6288          */
6289         sk->sk_wmem_queued   -= skb->truesize;
6290         sk_mem_uncharge(sk, skb->truesize);
6291
6292         sock_wfree(skb);
6293         __sctp_write_space(asoc);
6294
6295         sctp_association_put(asoc);
6296 }
6297
6298 /* Do accounting for the receive space on the socket.
6299  * Accounting for the association is done in ulpevent.c
6300  * We set this as a destructor for the cloned data skbs so that
6301  * accounting is done at the correct time.
6302  */
6303 void sctp_sock_rfree(struct sk_buff *skb)
6304 {
6305         struct sock *sk = skb->sk;
6306         struct sctp_ulpevent *event = sctp_skb2event(skb);
6307
6308         atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6309
6310         /*
6311          * Mimic the behavior of sock_rfree
6312          */
6313         sk_mem_uncharge(sk, event->rmem_len);
6314 }
6315
6316
6317 /* Helper function to wait for space in the sndbuf.  */
6318 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6319                                 size_t msg_len)
6320 {
6321         struct sock *sk = asoc->base.sk;
6322         int err = 0;
6323         long current_timeo = *timeo_p;
6324         DEFINE_WAIT(wait);
6325
6326         SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6327                           asoc, (long)(*timeo_p), msg_len);
6328
6329         /* Increment the association's refcnt.  */
6330         sctp_association_hold(asoc);
6331
6332         /* Wait on the association specific sndbuf space. */
6333         for (;;) {
6334                 prepare_to_wait_exclusive(&asoc->wait, &wait,
6335                                           TASK_INTERRUPTIBLE);
6336                 if (!*timeo_p)
6337                         goto do_nonblock;
6338                 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6339                     asoc->base.dead)
6340                         goto do_error;
6341                 if (signal_pending(current))
6342                         goto do_interrupted;
6343                 if (msg_len <= sctp_wspace(asoc))
6344                         break;
6345
6346                 /* Let another process have a go.  Since we are going
6347                  * to sleep anyway.
6348                  */
6349                 sctp_release_sock(sk);
6350                 current_timeo = schedule_timeout(current_timeo);
6351                 BUG_ON(sk != asoc->base.sk);
6352                 sctp_lock_sock(sk);
6353
6354                 *timeo_p = current_timeo;
6355         }
6356
6357 out:
6358         finish_wait(&asoc->wait, &wait);
6359
6360         /* Release the association's refcnt.  */
6361         sctp_association_put(asoc);
6362
6363         return err;
6364
6365 do_error:
6366         err = -EPIPE;
6367         goto out;
6368
6369 do_interrupted:
6370         err = sock_intr_errno(*timeo_p);
6371         goto out;
6372
6373 do_nonblock:
6374         err = -EAGAIN;
6375         goto out;
6376 }
6377
6378 /* If socket sndbuf has changed, wake up all per association waiters.  */
6379 void sctp_write_space(struct sock *sk)
6380 {
6381         struct sctp_association *asoc;
6382
6383         /* Wake up the tasks in each wait queue.  */
6384         list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6385                 __sctp_write_space(asoc);
6386         }
6387 }
6388
6389 /* Is there any sndbuf space available on the socket?
6390  *
6391  * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6392  * associations on the same socket.  For a UDP-style socket with
6393  * multiple associations, it is possible for it to be "unwriteable"
6394  * prematurely.  I assume that this is acceptable because
6395  * a premature "unwriteable" is better than an accidental "writeable" which
6396  * would cause an unwanted block under certain circumstances.  For the 1-1
6397  * UDP-style sockets or TCP-style sockets, this code should work.
6398  *  - Daisy
6399  */
6400 static int sctp_writeable(struct sock *sk)
6401 {
6402         int amt = 0;
6403
6404         amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
6405         if (amt < 0)
6406                 amt = 0;
6407         return amt;
6408 }
6409
6410 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6411  * returns immediately with EINPROGRESS.
6412  */
6413 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6414 {
6415         struct sock *sk = asoc->base.sk;
6416         int err = 0;
6417         long current_timeo = *timeo_p;
6418         DEFINE_WAIT(wait);
6419
6420         SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6421                           (long)(*timeo_p));
6422
6423         /* Increment the association's refcnt.  */
6424         sctp_association_hold(asoc);
6425
6426         for (;;) {
6427                 prepare_to_wait_exclusive(&asoc->wait, &wait,
6428                                           TASK_INTERRUPTIBLE);
6429                 if (!*timeo_p)
6430                         goto do_nonblock;
6431                 if (sk->sk_shutdown & RCV_SHUTDOWN)
6432                         break;
6433                 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6434                     asoc->base.dead)
6435                         goto do_error;
6436                 if (signal_pending(current))
6437                         goto do_interrupted;
6438
6439                 if (sctp_state(asoc, ESTABLISHED))
6440                         break;
6441
6442                 /* Let another process have a go.  Since we are going
6443                  * to sleep anyway.
6444                  */
6445                 sctp_release_sock(sk);
6446                 current_timeo = schedule_timeout(current_timeo);
6447                 sctp_lock_sock(sk);
6448
6449                 *timeo_p = current_timeo;
6450         }
6451
6452 out:
6453         finish_wait(&asoc->wait, &wait);
6454
6455         /* Release the association's refcnt.  */
6456         sctp_association_put(asoc);
6457
6458         return err;
6459
6460 do_error:
6461         if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6462                 err = -ETIMEDOUT;
6463         else
6464                 err = -ECONNREFUSED;
6465         goto out;
6466
6467 do_interrupted:
6468         err = sock_intr_errno(*timeo_p);
6469         goto out;
6470
6471 do_nonblock:
6472         err = -EINPROGRESS;
6473         goto out;
6474 }
6475
6476 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6477 {
6478         struct sctp_endpoint *ep;
6479         int err = 0;
6480         DEFINE_WAIT(wait);
6481
6482         ep = sctp_sk(sk)->ep;
6483
6484
6485         for (;;) {
6486                 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6487                                           TASK_INTERRUPTIBLE);
6488
6489                 if (list_empty(&ep->asocs)) {
6490                         sctp_release_sock(sk);
6491                         timeo = schedule_timeout(timeo);
6492                         sctp_lock_sock(sk);
6493                 }
6494
6495                 err = -EINVAL;
6496                 if (!sctp_sstate(sk, LISTENING))
6497                         break;
6498
6499                 err = 0;
6500                 if (!list_empty(&ep->asocs))
6501                         break;
6502
6503                 err = sock_intr_errno(timeo);
6504                 if (signal_pending(current))
6505                         break;
6506
6507                 err = -EAGAIN;
6508                 if (!timeo)
6509                         break;
6510         }
6511
6512         finish_wait(sk->sk_sleep, &wait);
6513
6514         return err;
6515 }
6516
6517 static void sctp_wait_for_close(struct sock *sk, long timeout)
6518 {
6519         DEFINE_WAIT(wait);
6520
6521         do {
6522                 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6523                 if (list_empty(&sctp_sk(sk)->ep->asocs))
6524                         break;
6525                 sctp_release_sock(sk);
6526                 timeout = schedule_timeout(timeout);
6527                 sctp_lock_sock(sk);
6528         } while (!signal_pending(current) && timeout);
6529
6530         finish_wait(sk->sk_sleep, &wait);
6531 }
6532
6533 static void sctp_sock_rfree_frag(struct sk_buff *skb)
6534 {
6535         struct sk_buff *frag;
6536
6537         if (!skb->data_len)
6538                 goto done;
6539
6540         /* Don't forget the fragments. */
6541         for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6542                 sctp_sock_rfree_frag(frag);
6543
6544 done:
6545         sctp_sock_rfree(skb);
6546 }
6547
6548 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6549 {
6550         struct sk_buff *frag;
6551
6552         if (!skb->data_len)
6553                 goto done;
6554
6555         /* Don't forget the fragments. */
6556         for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6557                 sctp_skb_set_owner_r_frag(frag, sk);
6558
6559 done:
6560         sctp_skb_set_owner_r(skb, sk);
6561 }
6562
6563 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6564  * and its messages to the newsk.
6565  */
6566 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6567                               struct sctp_association *assoc,
6568                               sctp_socket_type_t type)
6569 {
6570         struct sctp_sock *oldsp = sctp_sk(oldsk);
6571         struct sctp_sock *newsp = sctp_sk(newsk);
6572         struct sctp_bind_bucket *pp; /* hash list port iterator */
6573         struct sctp_endpoint *newep = newsp->ep;
6574         struct sk_buff *skb, *tmp;
6575         struct sctp_ulpevent *event;
6576         struct sctp_bind_hashbucket *head;
6577
6578         /* Migrate socket buffer sizes and all the socket level options to the
6579          * new socket.
6580          */
6581         newsk->sk_sndbuf = oldsk->sk_sndbuf;
6582         newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6583         /* Brute force copy old sctp opt. */
6584         inet_sk_copy_descendant(newsk, oldsk);
6585
6586         /* Restore the ep value that was overwritten with the above structure
6587          * copy.
6588          */
6589         newsp->ep = newep;
6590         newsp->hmac = NULL;
6591
6592         /* Hook this new socket in to the bind_hash list. */
6593         head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6594         sctp_local_bh_disable();
6595         sctp_spin_lock(&head->lock);
6596         pp = sctp_sk(oldsk)->bind_hash;
6597         sk_add_bind_node(newsk, &pp->owner);
6598         sctp_sk(newsk)->bind_hash = pp;
6599         inet_sk(newsk)->num = inet_sk(oldsk)->num;
6600         sctp_spin_unlock(&head->lock);
6601         sctp_local_bh_enable();
6602
6603         /* Copy the bind_addr list from the original endpoint to the new
6604          * endpoint so that we can handle restarts properly
6605          */
6606         sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6607                                 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6608
6609         /* Move any messages in the old socket's receive queue that are for the
6610          * peeled off association to the new socket's receive queue.
6611          */
6612         sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6613                 event = sctp_skb2event(skb);
6614                 if (event->asoc == assoc) {
6615                         sctp_sock_rfree_frag(skb);
6616                         __skb_unlink(skb, &oldsk->sk_receive_queue);
6617                         __skb_queue_tail(&newsk->sk_receive_queue, skb);
6618                         sctp_skb_set_owner_r_frag(skb, newsk);
6619                 }
6620         }
6621
6622         /* Clean up any messages pending delivery due to partial
6623          * delivery.   Three cases:
6624          * 1) No partial deliver;  no work.
6625          * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6626          * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6627          */
6628         skb_queue_head_init(&newsp->pd_lobby);
6629         atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6630
6631         if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6632                 struct sk_buff_head *queue;
6633
6634                 /* Decide which queue to move pd_lobby skbs to. */
6635                 if (assoc->ulpq.pd_mode) {
6636                         queue = &newsp->pd_lobby;
6637                 } else
6638                         queue = &newsk->sk_receive_queue;
6639
6640                 /* Walk through the pd_lobby, looking for skbs that
6641                  * need moved to the new socket.
6642                  */
6643                 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6644                         event = sctp_skb2event(skb);
6645                         if (event->asoc == assoc) {
6646                                 sctp_sock_rfree_frag(skb);
6647                                 __skb_unlink(skb, &oldsp->pd_lobby);
6648                                 __skb_queue_tail(queue, skb);
6649                                 sctp_skb_set_owner_r_frag(skb, newsk);
6650                         }
6651                 }
6652
6653                 /* Clear up any skbs waiting for the partial
6654                  * delivery to finish.
6655                  */
6656                 if (assoc->ulpq.pd_mode)
6657                         sctp_clear_pd(oldsk, NULL);
6658
6659         }
6660
6661         sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6662                 sctp_sock_rfree_frag(skb);
6663                 sctp_skb_set_owner_r_frag(skb, newsk);
6664         }
6665
6666         sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6667                 sctp_sock_rfree_frag(skb);
6668                 sctp_skb_set_owner_r_frag(skb, newsk);
6669         }
6670
6671         /* Set the type of socket to indicate that it is peeled off from the
6672          * original UDP-style socket or created with the accept() call on a
6673          * TCP-style socket..
6674          */
6675         newsp->type = type;
6676
6677         /* Mark the new socket "in-use" by the user so that any packets
6678          * that may arrive on the association after we've moved it are
6679          * queued to the backlog.  This prevents a potential race between
6680          * backlog processing on the old socket and new-packet processing
6681          * on the new socket.
6682          *
6683          * The caller has just allocated newsk so we can guarantee that other
6684          * paths won't try to lock it and then oldsk.
6685          */
6686         lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6687         sctp_assoc_migrate(assoc, newsk);
6688
6689         /* If the association on the newsk is already closed before accept()
6690          * is called, set RCV_SHUTDOWN flag.
6691          */
6692         if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6693                 newsk->sk_shutdown |= RCV_SHUTDOWN;
6694
6695         newsk->sk_state = SCTP_SS_ESTABLISHED;
6696         sctp_release_sock(newsk);
6697 }
6698
6699
6700 /* This proto struct describes the ULP interface for SCTP.  */
6701 struct proto sctp_prot = {
6702         .name        =  "SCTP",
6703         .owner       =  THIS_MODULE,
6704         .close       =  sctp_close,
6705         .connect     =  sctp_connect,
6706         .disconnect  =  sctp_disconnect,
6707         .accept      =  sctp_accept,
6708         .ioctl       =  sctp_ioctl,
6709         .init        =  sctp_init_sock,
6710         .destroy     =  sctp_destroy_sock,
6711         .shutdown    =  sctp_shutdown,
6712         .setsockopt  =  sctp_setsockopt,
6713         .getsockopt  =  sctp_getsockopt,
6714         .sendmsg     =  sctp_sendmsg,
6715         .recvmsg     =  sctp_recvmsg,
6716         .bind        =  sctp_bind,
6717         .backlog_rcv =  sctp_backlog_rcv,
6718         .hash        =  sctp_hash,
6719         .unhash      =  sctp_unhash,
6720         .get_port    =  sctp_get_port,
6721         .obj_size    =  sizeof(struct sctp_sock),
6722         .sysctl_mem  =  sysctl_sctp_mem,
6723         .sysctl_rmem =  sysctl_sctp_rmem,
6724         .sysctl_wmem =  sysctl_sctp_wmem,
6725         .memory_pressure = &sctp_memory_pressure,
6726         .enter_memory_pressure = sctp_enter_memory_pressure,
6727         .memory_allocated = &sctp_memory_allocated,
6728         .sockets_allocated = &sctp_sockets_allocated,
6729 };
6730
6731 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6732
6733 struct proto sctpv6_prot = {
6734         .name           = "SCTPv6",
6735         .owner          = THIS_MODULE,
6736         .close          = sctp_close,
6737         .connect        = sctp_connect,
6738         .disconnect     = sctp_disconnect,
6739         .accept         = sctp_accept,
6740         .ioctl          = sctp_ioctl,
6741         .init           = sctp_init_sock,
6742         .destroy        = sctp_destroy_sock,
6743         .shutdown       = sctp_shutdown,
6744         .setsockopt     = sctp_setsockopt,
6745         .getsockopt     = sctp_getsockopt,
6746         .sendmsg        = sctp_sendmsg,
6747         .recvmsg        = sctp_recvmsg,
6748         .bind           = sctp_bind,
6749         .backlog_rcv    = sctp_backlog_rcv,
6750         .hash           = sctp_hash,
6751         .unhash         = sctp_unhash,
6752         .get_port       = sctp_get_port,
6753         .obj_size       = sizeof(struct sctp6_sock),
6754         .sysctl_mem     = sysctl_sctp_mem,
6755         .sysctl_rmem    = sysctl_sctp_rmem,
6756         .sysctl_wmem    = sysctl_sctp_wmem,
6757         .memory_pressure = &sctp_memory_pressure,
6758         .enter_memory_pressure = sctp_enter_memory_pressure,
6759         .memory_allocated = &sctp_memory_allocated,
6760         .sockets_allocated = &sctp_sockets_allocated,
6761 };
6762 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */