1 /* SCTP kernel implementation 2 * (C) Copyright IBM Corp. 2001, 2004 3 * Copyright (c) 1999-2000 Cisco, Inc. 4 * Copyright (c) 1999-2001 Motorola, Inc. 5 * Copyright (c) 2001-2003 Intel Corp. 6 * Copyright (c) 2001-2002 Nokia, Inc. 7 * Copyright (c) 2001 La Monte H.P. Yarroll 8 * 9 * This file is part of the SCTP kernel implementation 10 * 11 * These functions interface with the sockets layer to implement the 12 * SCTP Extensions for the Sockets API. 13 * 14 * Note that the descriptions from the specification are USER level 15 * functions--this file is the functions which populate the struct proto 16 * for SCTP which is the BOTTOM of the sockets interface. 17 * 18 * This SCTP implementation is free software; 19 * you can redistribute it and/or modify it under the terms of 20 * the GNU General Public License as published by 21 * the Free Software Foundation; either version 2, or (at your option) 22 * any later version. 23 * 24 * This SCTP implementation is distributed in the hope that it 25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied 26 * ************************ 27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. 28 * See the GNU General Public License for more details. 29 * 30 * You should have received a copy of the GNU General Public License 31 * along with GNU CC; see the file COPYING. If not, see 32 * <http://www.gnu.org/licenses/>. 33 * 34 * Please send any bug reports or fixes you make to the 35 * email address(es): 36 * lksctp developers <linux-sctp@vger.kernel.org> 37 * 38 * Written or modified by: 39 * La Monte H.P. Yarroll <piggy@acm.org> 40 * Narasimha Budihal <narsi@refcode.org> 41 * Karl Knutson <karl@athena.chicago.il.us> 42 * Jon Grimm <jgrimm@us.ibm.com> 43 * Xingang Guo <xingang.guo@intel.com> 44 * Daisy Chang <daisyc@us.ibm.com> 45 * Sridhar Samudrala <samudrala@us.ibm.com> 46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com> 47 * Ardelle Fan <ardelle.fan@intel.com> 48 * Ryan Layer <rmlayer@us.ibm.com> 49 * Anup Pemmaiah <pemmaiah@cc.usu.edu> 50 * Kevin Gao <kevin.gao@intel.com> 51 */ 52 53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 54 55 #include <linux/types.h> 56 #include <linux/kernel.h> 57 #include <linux/wait.h> 58 #include <linux/time.h> 59 #include <linux/ip.h> 60 #include <linux/capability.h> 61 #include <linux/fcntl.h> 62 #include <linux/poll.h> 63 #include <linux/init.h> 64 #include <linux/crypto.h> 65 #include <linux/slab.h> 66 #include <linux/file.h> 67 #include <linux/compat.h> 68 69 #include <net/ip.h> 70 #include <net/icmp.h> 71 #include <net/route.h> 72 #include <net/ipv6.h> 73 #include <net/inet_common.h> 74 #include <net/busy_poll.h> 75 76 #include <linux/socket.h> /* for sa_family_t */ 77 #include <linux/export.h> 78 #include <net/sock.h> 79 #include <net/sctp/sctp.h> 80 #include <net/sctp/sm.h> 81 82 /* Forward declarations for internal helper functions. */ 83 static int sctp_writeable(struct sock *sk); 84 static void sctp_wfree(struct sk_buff *skb); 85 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p, 86 size_t msg_len); 87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p); 88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p); 89 static int sctp_wait_for_accept(struct sock *sk, long timeo); 90 static void sctp_wait_for_close(struct sock *sk, long timeo); 91 static void sctp_destruct_sock(struct sock *sk); 92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt, 93 union sctp_addr *addr, int len); 94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int); 95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int); 96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int); 97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int); 98 static int sctp_send_asconf(struct sctp_association *asoc, 99 struct sctp_chunk *chunk); 100 static int sctp_do_bind(struct sock *, union sctp_addr *, int); 101 static int sctp_autobind(struct sock *sk); 102 static void sctp_sock_migrate(struct sock *, struct sock *, 103 struct sctp_association *, sctp_socket_type_t); 104 105 static int sctp_memory_pressure; 106 static atomic_long_t sctp_memory_allocated; 107 struct percpu_counter sctp_sockets_allocated; 108 109 static void sctp_enter_memory_pressure(struct sock *sk) 110 { 111 sctp_memory_pressure = 1; 112 } 113 114 115 /* Get the sndbuf space available at the time on the association. */ 116 static inline int sctp_wspace(struct sctp_association *asoc) 117 { 118 int amt; 119 120 if (asoc->ep->sndbuf_policy) 121 amt = asoc->sndbuf_used; 122 else 123 amt = sk_wmem_alloc_get(asoc->base.sk); 124 125 if (amt >= asoc->base.sk->sk_sndbuf) { 126 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK) 127 amt = 0; 128 else { 129 amt = sk_stream_wspace(asoc->base.sk); 130 if (amt < 0) 131 amt = 0; 132 } 133 } else { 134 amt = asoc->base.sk->sk_sndbuf - amt; 135 } 136 return amt; 137 } 138 139 /* Increment the used sndbuf space count of the corresponding association by 140 * the size of the outgoing data chunk. 141 * Also, set the skb destructor for sndbuf accounting later. 142 * 143 * Since it is always 1-1 between chunk and skb, and also a new skb is always 144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the 145 * destructor in the data chunk skb for the purpose of the sndbuf space 146 * tracking. 147 */ 148 static inline void sctp_set_owner_w(struct sctp_chunk *chunk) 149 { 150 struct sctp_association *asoc = chunk->asoc; 151 struct sock *sk = asoc->base.sk; 152 153 /* The sndbuf space is tracked per association. */ 154 sctp_association_hold(asoc); 155 156 skb_set_owner_w(chunk->skb, sk); 157 158 chunk->skb->destructor = sctp_wfree; 159 /* Save the chunk pointer in skb for sctp_wfree to use later. */ 160 skb_shinfo(chunk->skb)->destructor_arg = chunk; 161 162 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) + 163 sizeof(struct sk_buff) + 164 sizeof(struct sctp_chunk); 165 166 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc); 167 sk->sk_wmem_queued += chunk->skb->truesize; 168 sk_mem_charge(sk, chunk->skb->truesize); 169 } 170 171 /* Verify that this is a valid address. */ 172 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr, 173 int len) 174 { 175 struct sctp_af *af; 176 177 /* Verify basic sockaddr. */ 178 af = sctp_sockaddr_af(sctp_sk(sk), addr, len); 179 if (!af) 180 return -EINVAL; 181 182 /* Is this a valid SCTP address? */ 183 if (!af->addr_valid(addr, sctp_sk(sk), NULL)) 184 return -EINVAL; 185 186 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr))) 187 return -EINVAL; 188 189 return 0; 190 } 191 192 /* Look up the association by its id. If this is not a UDP-style 193 * socket, the ID field is always ignored. 194 */ 195 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id) 196 { 197 struct sctp_association *asoc = NULL; 198 199 /* If this is not a UDP-style socket, assoc id should be ignored. */ 200 if (!sctp_style(sk, UDP)) { 201 /* Return NULL if the socket state is not ESTABLISHED. It 202 * could be a TCP-style listening socket or a socket which 203 * hasn't yet called connect() to establish an association. 204 */ 205 if (!sctp_sstate(sk, ESTABLISHED)) 206 return NULL; 207 208 /* Get the first and the only association from the list. */ 209 if (!list_empty(&sctp_sk(sk)->ep->asocs)) 210 asoc = list_entry(sctp_sk(sk)->ep->asocs.next, 211 struct sctp_association, asocs); 212 return asoc; 213 } 214 215 /* Otherwise this is a UDP-style socket. */ 216 if (!id || (id == (sctp_assoc_t)-1)) 217 return NULL; 218 219 spin_lock_bh(&sctp_assocs_id_lock); 220 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id); 221 spin_unlock_bh(&sctp_assocs_id_lock); 222 223 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead) 224 return NULL; 225 226 return asoc; 227 } 228 229 /* Look up the transport from an address and an assoc id. If both address and 230 * id are specified, the associations matching the address and the id should be 231 * the same. 232 */ 233 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk, 234 struct sockaddr_storage *addr, 235 sctp_assoc_t id) 236 { 237 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL; 238 struct sctp_transport *transport; 239 union sctp_addr *laddr = (union sctp_addr *)addr; 240 241 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep, 242 laddr, 243 &transport); 244 245 if (!addr_asoc) 246 return NULL; 247 248 id_asoc = sctp_id2assoc(sk, id); 249 if (id_asoc && (id_asoc != addr_asoc)) 250 return NULL; 251 252 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk), 253 (union sctp_addr *)addr); 254 255 return transport; 256 } 257 258 /* API 3.1.2 bind() - UDP Style Syntax 259 * The syntax of bind() is, 260 * 261 * ret = bind(int sd, struct sockaddr *addr, int addrlen); 262 * 263 * sd - the socket descriptor returned by socket(). 264 * addr - the address structure (struct sockaddr_in or struct 265 * sockaddr_in6 [RFC 2553]), 266 * addr_len - the size of the address structure. 267 */ 268 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len) 269 { 270 int retval = 0; 271 272 lock_sock(sk); 273 274 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk, 275 addr, addr_len); 276 277 /* Disallow binding twice. */ 278 if (!sctp_sk(sk)->ep->base.bind_addr.port) 279 retval = sctp_do_bind(sk, (union sctp_addr *)addr, 280 addr_len); 281 else 282 retval = -EINVAL; 283 284 release_sock(sk); 285 286 return retval; 287 } 288 289 static long sctp_get_port_local(struct sock *, union sctp_addr *); 290 291 /* Verify this is a valid sockaddr. */ 292 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt, 293 union sctp_addr *addr, int len) 294 { 295 struct sctp_af *af; 296 297 /* Check minimum size. */ 298 if (len < sizeof (struct sockaddr)) 299 return NULL; 300 301 /* V4 mapped address are really of AF_INET family */ 302 if (addr->sa.sa_family == AF_INET6 && 303 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) { 304 if (!opt->pf->af_supported(AF_INET, opt)) 305 return NULL; 306 } else { 307 /* Does this PF support this AF? */ 308 if (!opt->pf->af_supported(addr->sa.sa_family, opt)) 309 return NULL; 310 } 311 312 /* If we get this far, af is valid. */ 313 af = sctp_get_af_specific(addr->sa.sa_family); 314 315 if (len < af->sockaddr_len) 316 return NULL; 317 318 return af; 319 } 320 321 /* Bind a local address either to an endpoint or to an association. */ 322 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len) 323 { 324 struct net *net = sock_net(sk); 325 struct sctp_sock *sp = sctp_sk(sk); 326 struct sctp_endpoint *ep = sp->ep; 327 struct sctp_bind_addr *bp = &ep->base.bind_addr; 328 struct sctp_af *af; 329 unsigned short snum; 330 int ret = 0; 331 332 /* Common sockaddr verification. */ 333 af = sctp_sockaddr_af(sp, addr, len); 334 if (!af) { 335 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n", 336 __func__, sk, addr, len); 337 return -EINVAL; 338 } 339 340 snum = ntohs(addr->v4.sin_port); 341 342 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n", 343 __func__, sk, &addr->sa, bp->port, snum, len); 344 345 /* PF specific bind() address verification. */ 346 if (!sp->pf->bind_verify(sp, addr)) 347 return -EADDRNOTAVAIL; 348 349 /* We must either be unbound, or bind to the same port. 350 * It's OK to allow 0 ports if we are already bound. 351 * We'll just inhert an already bound port in this case 352 */ 353 if (bp->port) { 354 if (!snum) 355 snum = bp->port; 356 else if (snum != bp->port) { 357 pr_debug("%s: new port %d doesn't match existing port " 358 "%d\n", __func__, snum, bp->port); 359 return -EINVAL; 360 } 361 } 362 363 if (snum && snum < PROT_SOCK && 364 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 365 return -EACCES; 366 367 /* See if the address matches any of the addresses we may have 368 * already bound before checking against other endpoints. 369 */ 370 if (sctp_bind_addr_match(bp, addr, sp)) 371 return -EINVAL; 372 373 /* Make sure we are allowed to bind here. 374 * The function sctp_get_port_local() does duplicate address 375 * detection. 376 */ 377 addr->v4.sin_port = htons(snum); 378 if ((ret = sctp_get_port_local(sk, addr))) { 379 return -EADDRINUSE; 380 } 381 382 /* Refresh ephemeral port. */ 383 if (!bp->port) 384 bp->port = inet_sk(sk)->inet_num; 385 386 /* Add the address to the bind address list. 387 * Use GFP_ATOMIC since BHs will be disabled. 388 */ 389 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC); 390 391 /* Copy back into socket for getsockname() use. */ 392 if (!ret) { 393 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num); 394 sp->pf->to_sk_saddr(addr, sk); 395 } 396 397 return ret; 398 } 399 400 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks 401 * 402 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged 403 * at any one time. If a sender, after sending an ASCONF chunk, decides 404 * it needs to transfer another ASCONF Chunk, it MUST wait until the 405 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a 406 * subsequent ASCONF. Note this restriction binds each side, so at any 407 * time two ASCONF may be in-transit on any given association (one sent 408 * from each endpoint). 409 */ 410 static int sctp_send_asconf(struct sctp_association *asoc, 411 struct sctp_chunk *chunk) 412 { 413 struct net *net = sock_net(asoc->base.sk); 414 int retval = 0; 415 416 /* If there is an outstanding ASCONF chunk, queue it for later 417 * transmission. 418 */ 419 if (asoc->addip_last_asconf) { 420 list_add_tail(&chunk->list, &asoc->addip_chunk_list); 421 goto out; 422 } 423 424 /* Hold the chunk until an ASCONF_ACK is received. */ 425 sctp_chunk_hold(chunk); 426 retval = sctp_primitive_ASCONF(net, asoc, chunk); 427 if (retval) 428 sctp_chunk_free(chunk); 429 else 430 asoc->addip_last_asconf = chunk; 431 432 out: 433 return retval; 434 } 435 436 /* Add a list of addresses as bind addresses to local endpoint or 437 * association. 438 * 439 * Basically run through each address specified in the addrs/addrcnt 440 * array/length pair, determine if it is IPv6 or IPv4 and call 441 * sctp_do_bind() on it. 442 * 443 * If any of them fails, then the operation will be reversed and the 444 * ones that were added will be removed. 445 * 446 * Only sctp_setsockopt_bindx() is supposed to call this function. 447 */ 448 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt) 449 { 450 int cnt; 451 int retval = 0; 452 void *addr_buf; 453 struct sockaddr *sa_addr; 454 struct sctp_af *af; 455 456 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk, 457 addrs, addrcnt); 458 459 addr_buf = addrs; 460 for (cnt = 0; cnt < addrcnt; cnt++) { 461 /* The list may contain either IPv4 or IPv6 address; 462 * determine the address length for walking thru the list. 463 */ 464 sa_addr = addr_buf; 465 af = sctp_get_af_specific(sa_addr->sa_family); 466 if (!af) { 467 retval = -EINVAL; 468 goto err_bindx_add; 469 } 470 471 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr, 472 af->sockaddr_len); 473 474 addr_buf += af->sockaddr_len; 475 476 err_bindx_add: 477 if (retval < 0) { 478 /* Failed. Cleanup the ones that have been added */ 479 if (cnt > 0) 480 sctp_bindx_rem(sk, addrs, cnt); 481 return retval; 482 } 483 } 484 485 return retval; 486 } 487 488 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the 489 * associations that are part of the endpoint indicating that a list of local 490 * addresses are added to the endpoint. 491 * 492 * If any of the addresses is already in the bind address list of the 493 * association, we do not send the chunk for that association. But it will not 494 * affect other associations. 495 * 496 * Only sctp_setsockopt_bindx() is supposed to call this function. 497 */ 498 static int sctp_send_asconf_add_ip(struct sock *sk, 499 struct sockaddr *addrs, 500 int addrcnt) 501 { 502 struct net *net = sock_net(sk); 503 struct sctp_sock *sp; 504 struct sctp_endpoint *ep; 505 struct sctp_association *asoc; 506 struct sctp_bind_addr *bp; 507 struct sctp_chunk *chunk; 508 struct sctp_sockaddr_entry *laddr; 509 union sctp_addr *addr; 510 union sctp_addr saveaddr; 511 void *addr_buf; 512 struct sctp_af *af; 513 struct list_head *p; 514 int i; 515 int retval = 0; 516 517 if (!net->sctp.addip_enable) 518 return retval; 519 520 sp = sctp_sk(sk); 521 ep = sp->ep; 522 523 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", 524 __func__, sk, addrs, addrcnt); 525 526 list_for_each_entry(asoc, &ep->asocs, asocs) { 527 if (!asoc->peer.asconf_capable) 528 continue; 529 530 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP) 531 continue; 532 533 if (!sctp_state(asoc, ESTABLISHED)) 534 continue; 535 536 /* Check if any address in the packed array of addresses is 537 * in the bind address list of the association. If so, 538 * do not send the asconf chunk to its peer, but continue with 539 * other associations. 540 */ 541 addr_buf = addrs; 542 for (i = 0; i < addrcnt; i++) { 543 addr = addr_buf; 544 af = sctp_get_af_specific(addr->v4.sin_family); 545 if (!af) { 546 retval = -EINVAL; 547 goto out; 548 } 549 550 if (sctp_assoc_lookup_laddr(asoc, addr)) 551 break; 552 553 addr_buf += af->sockaddr_len; 554 } 555 if (i < addrcnt) 556 continue; 557 558 /* Use the first valid address in bind addr list of 559 * association as Address Parameter of ASCONF CHUNK. 560 */ 561 bp = &asoc->base.bind_addr; 562 p = bp->address_list.next; 563 laddr = list_entry(p, struct sctp_sockaddr_entry, list); 564 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs, 565 addrcnt, SCTP_PARAM_ADD_IP); 566 if (!chunk) { 567 retval = -ENOMEM; 568 goto out; 569 } 570 571 /* Add the new addresses to the bind address list with 572 * use_as_src set to 0. 573 */ 574 addr_buf = addrs; 575 for (i = 0; i < addrcnt; i++) { 576 addr = addr_buf; 577 af = sctp_get_af_specific(addr->v4.sin_family); 578 memcpy(&saveaddr, addr, af->sockaddr_len); 579 retval = sctp_add_bind_addr(bp, &saveaddr, 580 SCTP_ADDR_NEW, GFP_ATOMIC); 581 addr_buf += af->sockaddr_len; 582 } 583 if (asoc->src_out_of_asoc_ok) { 584 struct sctp_transport *trans; 585 586 list_for_each_entry(trans, 587 &asoc->peer.transport_addr_list, transports) { 588 /* Clear the source and route cache */ 589 dst_release(trans->dst); 590 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32, 591 2*asoc->pathmtu, 4380)); 592 trans->ssthresh = asoc->peer.i.a_rwnd; 593 trans->rto = asoc->rto_initial; 594 sctp_max_rto(asoc, trans); 595 trans->rtt = trans->srtt = trans->rttvar = 0; 596 sctp_transport_route(trans, NULL, 597 sctp_sk(asoc->base.sk)); 598 } 599 } 600 retval = sctp_send_asconf(asoc, chunk); 601 } 602 603 out: 604 return retval; 605 } 606 607 /* Remove a list of addresses from bind addresses list. Do not remove the 608 * last address. 609 * 610 * Basically run through each address specified in the addrs/addrcnt 611 * array/length pair, determine if it is IPv6 or IPv4 and call 612 * sctp_del_bind() on it. 613 * 614 * If any of them fails, then the operation will be reversed and the 615 * ones that were removed will be added back. 616 * 617 * At least one address has to be left; if only one address is 618 * available, the operation will return -EBUSY. 619 * 620 * Only sctp_setsockopt_bindx() is supposed to call this function. 621 */ 622 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt) 623 { 624 struct sctp_sock *sp = sctp_sk(sk); 625 struct sctp_endpoint *ep = sp->ep; 626 int cnt; 627 struct sctp_bind_addr *bp = &ep->base.bind_addr; 628 int retval = 0; 629 void *addr_buf; 630 union sctp_addr *sa_addr; 631 struct sctp_af *af; 632 633 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", 634 __func__, sk, addrs, addrcnt); 635 636 addr_buf = addrs; 637 for (cnt = 0; cnt < addrcnt; cnt++) { 638 /* If the bind address list is empty or if there is only one 639 * bind address, there is nothing more to be removed (we need 640 * at least one address here). 641 */ 642 if (list_empty(&bp->address_list) || 643 (sctp_list_single_entry(&bp->address_list))) { 644 retval = -EBUSY; 645 goto err_bindx_rem; 646 } 647 648 sa_addr = addr_buf; 649 af = sctp_get_af_specific(sa_addr->sa.sa_family); 650 if (!af) { 651 retval = -EINVAL; 652 goto err_bindx_rem; 653 } 654 655 if (!af->addr_valid(sa_addr, sp, NULL)) { 656 retval = -EADDRNOTAVAIL; 657 goto err_bindx_rem; 658 } 659 660 if (sa_addr->v4.sin_port && 661 sa_addr->v4.sin_port != htons(bp->port)) { 662 retval = -EINVAL; 663 goto err_bindx_rem; 664 } 665 666 if (!sa_addr->v4.sin_port) 667 sa_addr->v4.sin_port = htons(bp->port); 668 669 /* FIXME - There is probably a need to check if sk->sk_saddr and 670 * sk->sk_rcv_addr are currently set to one of the addresses to 671 * be removed. This is something which needs to be looked into 672 * when we are fixing the outstanding issues with multi-homing 673 * socket routing and failover schemes. Refer to comments in 674 * sctp_do_bind(). -daisy 675 */ 676 retval = sctp_del_bind_addr(bp, sa_addr); 677 678 addr_buf += af->sockaddr_len; 679 err_bindx_rem: 680 if (retval < 0) { 681 /* Failed. Add the ones that has been removed back */ 682 if (cnt > 0) 683 sctp_bindx_add(sk, addrs, cnt); 684 return retval; 685 } 686 } 687 688 return retval; 689 } 690 691 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of 692 * the associations that are part of the endpoint indicating that a list of 693 * local addresses are removed from the endpoint. 694 * 695 * If any of the addresses is already in the bind address list of the 696 * association, we do not send the chunk for that association. But it will not 697 * affect other associations. 698 * 699 * Only sctp_setsockopt_bindx() is supposed to call this function. 700 */ 701 static int sctp_send_asconf_del_ip(struct sock *sk, 702 struct sockaddr *addrs, 703 int addrcnt) 704 { 705 struct net *net = sock_net(sk); 706 struct sctp_sock *sp; 707 struct sctp_endpoint *ep; 708 struct sctp_association *asoc; 709 struct sctp_transport *transport; 710 struct sctp_bind_addr *bp; 711 struct sctp_chunk *chunk; 712 union sctp_addr *laddr; 713 void *addr_buf; 714 struct sctp_af *af; 715 struct sctp_sockaddr_entry *saddr; 716 int i; 717 int retval = 0; 718 int stored = 0; 719 720 chunk = NULL; 721 if (!net->sctp.addip_enable) 722 return retval; 723 724 sp = sctp_sk(sk); 725 ep = sp->ep; 726 727 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", 728 __func__, sk, addrs, addrcnt); 729 730 list_for_each_entry(asoc, &ep->asocs, asocs) { 731 732 if (!asoc->peer.asconf_capable) 733 continue; 734 735 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP) 736 continue; 737 738 if (!sctp_state(asoc, ESTABLISHED)) 739 continue; 740 741 /* Check if any address in the packed array of addresses is 742 * not present in the bind address list of the association. 743 * If so, do not send the asconf chunk to its peer, but 744 * continue with other associations. 745 */ 746 addr_buf = addrs; 747 for (i = 0; i < addrcnt; i++) { 748 laddr = addr_buf; 749 af = sctp_get_af_specific(laddr->v4.sin_family); 750 if (!af) { 751 retval = -EINVAL; 752 goto out; 753 } 754 755 if (!sctp_assoc_lookup_laddr(asoc, laddr)) 756 break; 757 758 addr_buf += af->sockaddr_len; 759 } 760 if (i < addrcnt) 761 continue; 762 763 /* Find one address in the association's bind address list 764 * that is not in the packed array of addresses. This is to 765 * make sure that we do not delete all the addresses in the 766 * association. 767 */ 768 bp = &asoc->base.bind_addr; 769 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs, 770 addrcnt, sp); 771 if ((laddr == NULL) && (addrcnt == 1)) { 772 if (asoc->asconf_addr_del_pending) 773 continue; 774 asoc->asconf_addr_del_pending = 775 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC); 776 if (asoc->asconf_addr_del_pending == NULL) { 777 retval = -ENOMEM; 778 goto out; 779 } 780 asoc->asconf_addr_del_pending->sa.sa_family = 781 addrs->sa_family; 782 asoc->asconf_addr_del_pending->v4.sin_port = 783 htons(bp->port); 784 if (addrs->sa_family == AF_INET) { 785 struct sockaddr_in *sin; 786 787 sin = (struct sockaddr_in *)addrs; 788 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr; 789 } else if (addrs->sa_family == AF_INET6) { 790 struct sockaddr_in6 *sin6; 791 792 sin6 = (struct sockaddr_in6 *)addrs; 793 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr; 794 } 795 796 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n", 797 __func__, asoc, &asoc->asconf_addr_del_pending->sa, 798 asoc->asconf_addr_del_pending); 799 800 asoc->src_out_of_asoc_ok = 1; 801 stored = 1; 802 goto skip_mkasconf; 803 } 804 805 if (laddr == NULL) 806 return -EINVAL; 807 808 /* We do not need RCU protection throughout this loop 809 * because this is done under a socket lock from the 810 * setsockopt call. 811 */ 812 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt, 813 SCTP_PARAM_DEL_IP); 814 if (!chunk) { 815 retval = -ENOMEM; 816 goto out; 817 } 818 819 skip_mkasconf: 820 /* Reset use_as_src flag for the addresses in the bind address 821 * list that are to be deleted. 822 */ 823 addr_buf = addrs; 824 for (i = 0; i < addrcnt; i++) { 825 laddr = addr_buf; 826 af = sctp_get_af_specific(laddr->v4.sin_family); 827 list_for_each_entry(saddr, &bp->address_list, list) { 828 if (sctp_cmp_addr_exact(&saddr->a, laddr)) 829 saddr->state = SCTP_ADDR_DEL; 830 } 831 addr_buf += af->sockaddr_len; 832 } 833 834 /* Update the route and saddr entries for all the transports 835 * as some of the addresses in the bind address list are 836 * about to be deleted and cannot be used as source addresses. 837 */ 838 list_for_each_entry(transport, &asoc->peer.transport_addr_list, 839 transports) { 840 dst_release(transport->dst); 841 sctp_transport_route(transport, NULL, 842 sctp_sk(asoc->base.sk)); 843 } 844 845 if (stored) 846 /* We don't need to transmit ASCONF */ 847 continue; 848 retval = sctp_send_asconf(asoc, chunk); 849 } 850 out: 851 return retval; 852 } 853 854 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */ 855 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw) 856 { 857 struct sock *sk = sctp_opt2sk(sp); 858 union sctp_addr *addr; 859 struct sctp_af *af; 860 861 /* It is safe to write port space in caller. */ 862 addr = &addrw->a; 863 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port); 864 af = sctp_get_af_specific(addr->sa.sa_family); 865 if (!af) 866 return -EINVAL; 867 if (sctp_verify_addr(sk, addr, af->sockaddr_len)) 868 return -EINVAL; 869 870 if (addrw->state == SCTP_ADDR_NEW) 871 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1); 872 else 873 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1); 874 } 875 876 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt() 877 * 878 * API 8.1 879 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt, 880 * int flags); 881 * 882 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses. 883 * If the sd is an IPv6 socket, the addresses passed can either be IPv4 884 * or IPv6 addresses. 885 * 886 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see 887 * Section 3.1.2 for this usage. 888 * 889 * addrs is a pointer to an array of one or more socket addresses. Each 890 * address is contained in its appropriate structure (i.e. struct 891 * sockaddr_in or struct sockaddr_in6) the family of the address type 892 * must be used to distinguish the address length (note that this 893 * representation is termed a "packed array" of addresses). The caller 894 * specifies the number of addresses in the array with addrcnt. 895 * 896 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns 897 * -1, and sets errno to the appropriate error code. 898 * 899 * For SCTP, the port given in each socket address must be the same, or 900 * sctp_bindx() will fail, setting errno to EINVAL. 901 * 902 * The flags parameter is formed from the bitwise OR of zero or more of 903 * the following currently defined flags: 904 * 905 * SCTP_BINDX_ADD_ADDR 906 * 907 * SCTP_BINDX_REM_ADDR 908 * 909 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the 910 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given 911 * addresses from the association. The two flags are mutually exclusive; 912 * if both are given, sctp_bindx() will fail with EINVAL. A caller may 913 * not remove all addresses from an association; sctp_bindx() will 914 * reject such an attempt with EINVAL. 915 * 916 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate 917 * additional addresses with an endpoint after calling bind(). Or use 918 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening 919 * socket is associated with so that no new association accepted will be 920 * associated with those addresses. If the endpoint supports dynamic 921 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a 922 * endpoint to send the appropriate message to the peer to change the 923 * peers address lists. 924 * 925 * Adding and removing addresses from a connected association is 926 * optional functionality. Implementations that do not support this 927 * functionality should return EOPNOTSUPP. 928 * 929 * Basically do nothing but copying the addresses from user to kernel 930 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk. 931 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt() 932 * from userspace. 933 * 934 * We don't use copy_from_user() for optimization: we first do the 935 * sanity checks (buffer size -fast- and access check-healthy 936 * pointer); if all of those succeed, then we can alloc the memory 937 * (expensive operation) needed to copy the data to kernel. Then we do 938 * the copying without checking the user space area 939 * (__copy_from_user()). 940 * 941 * On exit there is no need to do sockfd_put(), sys_setsockopt() does 942 * it. 943 * 944 * sk The sk of the socket 945 * addrs The pointer to the addresses in user land 946 * addrssize Size of the addrs buffer 947 * op Operation to perform (add or remove, see the flags of 948 * sctp_bindx) 949 * 950 * Returns 0 if ok, <0 errno code on error. 951 */ 952 static int sctp_setsockopt_bindx(struct sock *sk, 953 struct sockaddr __user *addrs, 954 int addrs_size, int op) 955 { 956 struct sockaddr *kaddrs; 957 int err; 958 int addrcnt = 0; 959 int walk_size = 0; 960 struct sockaddr *sa_addr; 961 void *addr_buf; 962 struct sctp_af *af; 963 964 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n", 965 __func__, sk, addrs, addrs_size, op); 966 967 if (unlikely(addrs_size <= 0)) 968 return -EINVAL; 969 970 /* Check the user passed a healthy pointer. */ 971 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size))) 972 return -EFAULT; 973 974 /* Alloc space for the address array in kernel memory. */ 975 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN); 976 if (unlikely(!kaddrs)) 977 return -ENOMEM; 978 979 if (__copy_from_user(kaddrs, addrs, addrs_size)) { 980 kfree(kaddrs); 981 return -EFAULT; 982 } 983 984 /* Walk through the addrs buffer and count the number of addresses. */ 985 addr_buf = kaddrs; 986 while (walk_size < addrs_size) { 987 if (walk_size + sizeof(sa_family_t) > addrs_size) { 988 kfree(kaddrs); 989 return -EINVAL; 990 } 991 992 sa_addr = addr_buf; 993 af = sctp_get_af_specific(sa_addr->sa_family); 994 995 /* If the address family is not supported or if this address 996 * causes the address buffer to overflow return EINVAL. 997 */ 998 if (!af || (walk_size + af->sockaddr_len) > addrs_size) { 999 kfree(kaddrs); 1000 return -EINVAL; 1001 } 1002 addrcnt++; 1003 addr_buf += af->sockaddr_len; 1004 walk_size += af->sockaddr_len; 1005 } 1006 1007 /* Do the work. */ 1008 switch (op) { 1009 case SCTP_BINDX_ADD_ADDR: 1010 err = sctp_bindx_add(sk, kaddrs, addrcnt); 1011 if (err) 1012 goto out; 1013 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt); 1014 break; 1015 1016 case SCTP_BINDX_REM_ADDR: 1017 err = sctp_bindx_rem(sk, kaddrs, addrcnt); 1018 if (err) 1019 goto out; 1020 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt); 1021 break; 1022 1023 default: 1024 err = -EINVAL; 1025 break; 1026 } 1027 1028 out: 1029 kfree(kaddrs); 1030 1031 return err; 1032 } 1033 1034 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size) 1035 * 1036 * Common routine for handling connect() and sctp_connectx(). 1037 * Connect will come in with just a single address. 1038 */ 1039 static int __sctp_connect(struct sock *sk, 1040 struct sockaddr *kaddrs, 1041 int addrs_size, 1042 sctp_assoc_t *assoc_id) 1043 { 1044 struct net *net = sock_net(sk); 1045 struct sctp_sock *sp; 1046 struct sctp_endpoint *ep; 1047 struct sctp_association *asoc = NULL; 1048 struct sctp_association *asoc2; 1049 struct sctp_transport *transport; 1050 union sctp_addr to; 1051 sctp_scope_t scope; 1052 long timeo; 1053 int err = 0; 1054 int addrcnt = 0; 1055 int walk_size = 0; 1056 union sctp_addr *sa_addr = NULL; 1057 void *addr_buf; 1058 unsigned short port; 1059 unsigned int f_flags = 0; 1060 1061 sp = sctp_sk(sk); 1062 ep = sp->ep; 1063 1064 /* connect() cannot be done on a socket that is already in ESTABLISHED 1065 * state - UDP-style peeled off socket or a TCP-style socket that 1066 * is already connected. 1067 * It cannot be done even on a TCP-style listening socket. 1068 */ 1069 if (sctp_sstate(sk, ESTABLISHED) || 1070 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) { 1071 err = -EISCONN; 1072 goto out_free; 1073 } 1074 1075 /* Walk through the addrs buffer and count the number of addresses. */ 1076 addr_buf = kaddrs; 1077 while (walk_size < addrs_size) { 1078 struct sctp_af *af; 1079 1080 if (walk_size + sizeof(sa_family_t) > addrs_size) { 1081 err = -EINVAL; 1082 goto out_free; 1083 } 1084 1085 sa_addr = addr_buf; 1086 af = sctp_get_af_specific(sa_addr->sa.sa_family); 1087 1088 /* If the address family is not supported or if this address 1089 * causes the address buffer to overflow return EINVAL. 1090 */ 1091 if (!af || (walk_size + af->sockaddr_len) > addrs_size) { 1092 err = -EINVAL; 1093 goto out_free; 1094 } 1095 1096 port = ntohs(sa_addr->v4.sin_port); 1097 1098 /* Save current address so we can work with it */ 1099 memcpy(&to, sa_addr, af->sockaddr_len); 1100 1101 err = sctp_verify_addr(sk, &to, af->sockaddr_len); 1102 if (err) 1103 goto out_free; 1104 1105 /* Make sure the destination port is correctly set 1106 * in all addresses. 1107 */ 1108 if (asoc && asoc->peer.port && asoc->peer.port != port) { 1109 err = -EINVAL; 1110 goto out_free; 1111 } 1112 1113 /* Check if there already is a matching association on the 1114 * endpoint (other than the one created here). 1115 */ 1116 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport); 1117 if (asoc2 && asoc2 != asoc) { 1118 if (asoc2->state >= SCTP_STATE_ESTABLISHED) 1119 err = -EISCONN; 1120 else 1121 err = -EALREADY; 1122 goto out_free; 1123 } 1124 1125 /* If we could not find a matching association on the endpoint, 1126 * make sure that there is no peeled-off association matching 1127 * the peer address even on another socket. 1128 */ 1129 if (sctp_endpoint_is_peeled_off(ep, &to)) { 1130 err = -EADDRNOTAVAIL; 1131 goto out_free; 1132 } 1133 1134 if (!asoc) { 1135 /* If a bind() or sctp_bindx() is not called prior to 1136 * an sctp_connectx() call, the system picks an 1137 * ephemeral port and will choose an address set 1138 * equivalent to binding with a wildcard address. 1139 */ 1140 if (!ep->base.bind_addr.port) { 1141 if (sctp_autobind(sk)) { 1142 err = -EAGAIN; 1143 goto out_free; 1144 } 1145 } else { 1146 /* 1147 * If an unprivileged user inherits a 1-many 1148 * style socket with open associations on a 1149 * privileged port, it MAY be permitted to 1150 * accept new associations, but it SHOULD NOT 1151 * be permitted to open new associations. 1152 */ 1153 if (ep->base.bind_addr.port < PROT_SOCK && 1154 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) { 1155 err = -EACCES; 1156 goto out_free; 1157 } 1158 } 1159 1160 scope = sctp_scope(&to); 1161 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL); 1162 if (!asoc) { 1163 err = -ENOMEM; 1164 goto out_free; 1165 } 1166 1167 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, 1168 GFP_KERNEL); 1169 if (err < 0) { 1170 goto out_free; 1171 } 1172 1173 } 1174 1175 /* Prime the peer's transport structures. */ 1176 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, 1177 SCTP_UNKNOWN); 1178 if (!transport) { 1179 err = -ENOMEM; 1180 goto out_free; 1181 } 1182 1183 addrcnt++; 1184 addr_buf += af->sockaddr_len; 1185 walk_size += af->sockaddr_len; 1186 } 1187 1188 /* In case the user of sctp_connectx() wants an association 1189 * id back, assign one now. 1190 */ 1191 if (assoc_id) { 1192 err = sctp_assoc_set_id(asoc, GFP_KERNEL); 1193 if (err < 0) 1194 goto out_free; 1195 } 1196 1197 err = sctp_primitive_ASSOCIATE(net, asoc, NULL); 1198 if (err < 0) { 1199 goto out_free; 1200 } 1201 1202 /* Initialize sk's dport and daddr for getpeername() */ 1203 inet_sk(sk)->inet_dport = htons(asoc->peer.port); 1204 sp->pf->to_sk_daddr(sa_addr, sk); 1205 sk->sk_err = 0; 1206 1207 /* in-kernel sockets don't generally have a file allocated to them 1208 * if all they do is call sock_create_kern(). 1209 */ 1210 if (sk->sk_socket->file) 1211 f_flags = sk->sk_socket->file->f_flags; 1212 1213 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK); 1214 1215 err = sctp_wait_for_connect(asoc, &timeo); 1216 if ((err == 0 || err == -EINPROGRESS) && assoc_id) 1217 *assoc_id = asoc->assoc_id; 1218 1219 /* Don't free association on exit. */ 1220 asoc = NULL; 1221 1222 out_free: 1223 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n", 1224 __func__, asoc, kaddrs, err); 1225 1226 if (asoc) { 1227 /* sctp_primitive_ASSOCIATE may have added this association 1228 * To the hash table, try to unhash it, just in case, its a noop 1229 * if it wasn't hashed so we're safe 1230 */ 1231 sctp_unhash_established(asoc); 1232 sctp_association_free(asoc); 1233 } 1234 return err; 1235 } 1236 1237 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt() 1238 * 1239 * API 8.9 1240 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt, 1241 * sctp_assoc_t *asoc); 1242 * 1243 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses. 1244 * If the sd is an IPv6 socket, the addresses passed can either be IPv4 1245 * or IPv6 addresses. 1246 * 1247 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see 1248 * Section 3.1.2 for this usage. 1249 * 1250 * addrs is a pointer to an array of one or more socket addresses. Each 1251 * address is contained in its appropriate structure (i.e. struct 1252 * sockaddr_in or struct sockaddr_in6) the family of the address type 1253 * must be used to distengish the address length (note that this 1254 * representation is termed a "packed array" of addresses). The caller 1255 * specifies the number of addresses in the array with addrcnt. 1256 * 1257 * On success, sctp_connectx() returns 0. It also sets the assoc_id to 1258 * the association id of the new association. On failure, sctp_connectx() 1259 * returns -1, and sets errno to the appropriate error code. The assoc_id 1260 * is not touched by the kernel. 1261 * 1262 * For SCTP, the port given in each socket address must be the same, or 1263 * sctp_connectx() will fail, setting errno to EINVAL. 1264 * 1265 * An application can use sctp_connectx to initiate an association with 1266 * an endpoint that is multi-homed. Much like sctp_bindx() this call 1267 * allows a caller to specify multiple addresses at which a peer can be 1268 * reached. The way the SCTP stack uses the list of addresses to set up 1269 * the association is implementation dependent. This function only 1270 * specifies that the stack will try to make use of all the addresses in 1271 * the list when needed. 1272 * 1273 * Note that the list of addresses passed in is only used for setting up 1274 * the association. It does not necessarily equal the set of addresses 1275 * the peer uses for the resulting association. If the caller wants to 1276 * find out the set of peer addresses, it must use sctp_getpaddrs() to 1277 * retrieve them after the association has been set up. 1278 * 1279 * Basically do nothing but copying the addresses from user to kernel 1280 * land and invoking either sctp_connectx(). This is used for tunneling 1281 * the sctp_connectx() request through sctp_setsockopt() from userspace. 1282 * 1283 * We don't use copy_from_user() for optimization: we first do the 1284 * sanity checks (buffer size -fast- and access check-healthy 1285 * pointer); if all of those succeed, then we can alloc the memory 1286 * (expensive operation) needed to copy the data to kernel. Then we do 1287 * the copying without checking the user space area 1288 * (__copy_from_user()). 1289 * 1290 * On exit there is no need to do sockfd_put(), sys_setsockopt() does 1291 * it. 1292 * 1293 * sk The sk of the socket 1294 * addrs The pointer to the addresses in user land 1295 * addrssize Size of the addrs buffer 1296 * 1297 * Returns >=0 if ok, <0 errno code on error. 1298 */ 1299 static int __sctp_setsockopt_connectx(struct sock *sk, 1300 struct sockaddr __user *addrs, 1301 int addrs_size, 1302 sctp_assoc_t *assoc_id) 1303 { 1304 int err = 0; 1305 struct sockaddr *kaddrs; 1306 1307 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n", 1308 __func__, sk, addrs, addrs_size); 1309 1310 if (unlikely(addrs_size <= 0)) 1311 return -EINVAL; 1312 1313 /* Check the user passed a healthy pointer. */ 1314 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size))) 1315 return -EFAULT; 1316 1317 /* Alloc space for the address array in kernel memory. */ 1318 kaddrs = kmalloc(addrs_size, GFP_KERNEL); 1319 if (unlikely(!kaddrs)) 1320 return -ENOMEM; 1321 1322 if (__copy_from_user(kaddrs, addrs, addrs_size)) { 1323 err = -EFAULT; 1324 } else { 1325 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id); 1326 } 1327 1328 kfree(kaddrs); 1329 1330 return err; 1331 } 1332 1333 /* 1334 * This is an older interface. It's kept for backward compatibility 1335 * to the option that doesn't provide association id. 1336 */ 1337 static int sctp_setsockopt_connectx_old(struct sock *sk, 1338 struct sockaddr __user *addrs, 1339 int addrs_size) 1340 { 1341 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL); 1342 } 1343 1344 /* 1345 * New interface for the API. The since the API is done with a socket 1346 * option, to make it simple we feed back the association id is as a return 1347 * indication to the call. Error is always negative and association id is 1348 * always positive. 1349 */ 1350 static int sctp_setsockopt_connectx(struct sock *sk, 1351 struct sockaddr __user *addrs, 1352 int addrs_size) 1353 { 1354 sctp_assoc_t assoc_id = 0; 1355 int err = 0; 1356 1357 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id); 1358 1359 if (err) 1360 return err; 1361 else 1362 return assoc_id; 1363 } 1364 1365 /* 1366 * New (hopefully final) interface for the API. 1367 * We use the sctp_getaddrs_old structure so that use-space library 1368 * can avoid any unnecessary allocations. The only different part 1369 * is that we store the actual length of the address buffer into the 1370 * addrs_num structure member. That way we can re-use the existing 1371 * code. 1372 */ 1373 #ifdef CONFIG_COMPAT 1374 struct compat_sctp_getaddrs_old { 1375 sctp_assoc_t assoc_id; 1376 s32 addr_num; 1377 compat_uptr_t addrs; /* struct sockaddr * */ 1378 }; 1379 #endif 1380 1381 static int sctp_getsockopt_connectx3(struct sock *sk, int len, 1382 char __user *optval, 1383 int __user *optlen) 1384 { 1385 struct sctp_getaddrs_old param; 1386 sctp_assoc_t assoc_id = 0; 1387 int err = 0; 1388 1389 #ifdef CONFIG_COMPAT 1390 if (is_compat_task()) { 1391 struct compat_sctp_getaddrs_old param32; 1392 1393 if (len < sizeof(param32)) 1394 return -EINVAL; 1395 if (copy_from_user(¶m32, optval, sizeof(param32))) 1396 return -EFAULT; 1397 1398 param.assoc_id = param32.assoc_id; 1399 param.addr_num = param32.addr_num; 1400 param.addrs = compat_ptr(param32.addrs); 1401 } else 1402 #endif 1403 { 1404 if (len < sizeof(param)) 1405 return -EINVAL; 1406 if (copy_from_user(¶m, optval, sizeof(param))) 1407 return -EFAULT; 1408 } 1409 1410 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *) 1411 param.addrs, param.addr_num, 1412 &assoc_id); 1413 if (err == 0 || err == -EINPROGRESS) { 1414 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id))) 1415 return -EFAULT; 1416 if (put_user(sizeof(assoc_id), optlen)) 1417 return -EFAULT; 1418 } 1419 1420 return err; 1421 } 1422 1423 /* API 3.1.4 close() - UDP Style Syntax 1424 * Applications use close() to perform graceful shutdown (as described in 1425 * Section 10.1 of [SCTP]) on ALL the associations currently represented 1426 * by a UDP-style socket. 1427 * 1428 * The syntax is 1429 * 1430 * ret = close(int sd); 1431 * 1432 * sd - the socket descriptor of the associations to be closed. 1433 * 1434 * To gracefully shutdown a specific association represented by the 1435 * UDP-style socket, an application should use the sendmsg() call, 1436 * passing no user data, but including the appropriate flag in the 1437 * ancillary data (see Section xxxx). 1438 * 1439 * If sd in the close() call is a branched-off socket representing only 1440 * one association, the shutdown is performed on that association only. 1441 * 1442 * 4.1.6 close() - TCP Style Syntax 1443 * 1444 * Applications use close() to gracefully close down an association. 1445 * 1446 * The syntax is: 1447 * 1448 * int close(int sd); 1449 * 1450 * sd - the socket descriptor of the association to be closed. 1451 * 1452 * After an application calls close() on a socket descriptor, no further 1453 * socket operations will succeed on that descriptor. 1454 * 1455 * API 7.1.4 SO_LINGER 1456 * 1457 * An application using the TCP-style socket can use this option to 1458 * perform the SCTP ABORT primitive. The linger option structure is: 1459 * 1460 * struct linger { 1461 * int l_onoff; // option on/off 1462 * int l_linger; // linger time 1463 * }; 1464 * 1465 * To enable the option, set l_onoff to 1. If the l_linger value is set 1466 * to 0, calling close() is the same as the ABORT primitive. If the 1467 * value is set to a negative value, the setsockopt() call will return 1468 * an error. If the value is set to a positive value linger_time, the 1469 * close() can be blocked for at most linger_time ms. If the graceful 1470 * shutdown phase does not finish during this period, close() will 1471 * return but the graceful shutdown phase continues in the system. 1472 */ 1473 static void sctp_close(struct sock *sk, long timeout) 1474 { 1475 struct net *net = sock_net(sk); 1476 struct sctp_endpoint *ep; 1477 struct sctp_association *asoc; 1478 struct list_head *pos, *temp; 1479 unsigned int data_was_unread; 1480 1481 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout); 1482 1483 lock_sock(sk); 1484 sk->sk_shutdown = SHUTDOWN_MASK; 1485 sk->sk_state = SCTP_SS_CLOSING; 1486 1487 ep = sctp_sk(sk)->ep; 1488 1489 /* Clean up any skbs sitting on the receive queue. */ 1490 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue); 1491 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby); 1492 1493 /* Walk all associations on an endpoint. */ 1494 list_for_each_safe(pos, temp, &ep->asocs) { 1495 asoc = list_entry(pos, struct sctp_association, asocs); 1496 1497 if (sctp_style(sk, TCP)) { 1498 /* A closed association can still be in the list if 1499 * it belongs to a TCP-style listening socket that is 1500 * not yet accepted. If so, free it. If not, send an 1501 * ABORT or SHUTDOWN based on the linger options. 1502 */ 1503 if (sctp_state(asoc, CLOSED)) { 1504 sctp_unhash_established(asoc); 1505 sctp_association_free(asoc); 1506 continue; 1507 } 1508 } 1509 1510 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) || 1511 !skb_queue_empty(&asoc->ulpq.reasm) || 1512 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) { 1513 struct sctp_chunk *chunk; 1514 1515 chunk = sctp_make_abort_user(asoc, NULL, 0); 1516 if (chunk) 1517 sctp_primitive_ABORT(net, asoc, chunk); 1518 } else 1519 sctp_primitive_SHUTDOWN(net, asoc, NULL); 1520 } 1521 1522 /* On a TCP-style socket, block for at most linger_time if set. */ 1523 if (sctp_style(sk, TCP) && timeout) 1524 sctp_wait_for_close(sk, timeout); 1525 1526 /* This will run the backlog queue. */ 1527 release_sock(sk); 1528 1529 /* Supposedly, no process has access to the socket, but 1530 * the net layers still may. 1531 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock 1532 * held and that should be grabbed before socket lock. 1533 */ 1534 spin_lock_bh(&net->sctp.addr_wq_lock); 1535 bh_lock_sock(sk); 1536 1537 /* Hold the sock, since sk_common_release() will put sock_put() 1538 * and we have just a little more cleanup. 1539 */ 1540 sock_hold(sk); 1541 sk_common_release(sk); 1542 1543 bh_unlock_sock(sk); 1544 spin_unlock_bh(&net->sctp.addr_wq_lock); 1545 1546 sock_put(sk); 1547 1548 SCTP_DBG_OBJCNT_DEC(sock); 1549 } 1550 1551 /* Handle EPIPE error. */ 1552 static int sctp_error(struct sock *sk, int flags, int err) 1553 { 1554 if (err == -EPIPE) 1555 err = sock_error(sk) ? : -EPIPE; 1556 if (err == -EPIPE && !(flags & MSG_NOSIGNAL)) 1557 send_sig(SIGPIPE, current, 0); 1558 return err; 1559 } 1560 1561 /* API 3.1.3 sendmsg() - UDP Style Syntax 1562 * 1563 * An application uses sendmsg() and recvmsg() calls to transmit data to 1564 * and receive data from its peer. 1565 * 1566 * ssize_t sendmsg(int socket, const struct msghdr *message, 1567 * int flags); 1568 * 1569 * socket - the socket descriptor of the endpoint. 1570 * message - pointer to the msghdr structure which contains a single 1571 * user message and possibly some ancillary data. 1572 * 1573 * See Section 5 for complete description of the data 1574 * structures. 1575 * 1576 * flags - flags sent or received with the user message, see Section 1577 * 5 for complete description of the flags. 1578 * 1579 * Note: This function could use a rewrite especially when explicit 1580 * connect support comes in. 1581 */ 1582 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */ 1583 1584 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *); 1585 1586 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len) 1587 { 1588 struct net *net = sock_net(sk); 1589 struct sctp_sock *sp; 1590 struct sctp_endpoint *ep; 1591 struct sctp_association *new_asoc = NULL, *asoc = NULL; 1592 struct sctp_transport *transport, *chunk_tp; 1593 struct sctp_chunk *chunk; 1594 union sctp_addr to; 1595 struct sockaddr *msg_name = NULL; 1596 struct sctp_sndrcvinfo default_sinfo; 1597 struct sctp_sndrcvinfo *sinfo; 1598 struct sctp_initmsg *sinit; 1599 sctp_assoc_t associd = 0; 1600 sctp_cmsgs_t cmsgs = { NULL }; 1601 sctp_scope_t scope; 1602 bool fill_sinfo_ttl = false, wait_connect = false; 1603 struct sctp_datamsg *datamsg; 1604 int msg_flags = msg->msg_flags; 1605 __u16 sinfo_flags = 0; 1606 long timeo; 1607 int err; 1608 1609 err = 0; 1610 sp = sctp_sk(sk); 1611 ep = sp->ep; 1612 1613 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk, 1614 msg, msg_len, ep); 1615 1616 /* We cannot send a message over a TCP-style listening socket. */ 1617 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) { 1618 err = -EPIPE; 1619 goto out_nounlock; 1620 } 1621 1622 /* Parse out the SCTP CMSGs. */ 1623 err = sctp_msghdr_parse(msg, &cmsgs); 1624 if (err) { 1625 pr_debug("%s: msghdr parse err:%x\n", __func__, err); 1626 goto out_nounlock; 1627 } 1628 1629 /* Fetch the destination address for this packet. This 1630 * address only selects the association--it is not necessarily 1631 * the address we will send to. 1632 * For a peeled-off socket, msg_name is ignored. 1633 */ 1634 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) { 1635 int msg_namelen = msg->msg_namelen; 1636 1637 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name, 1638 msg_namelen); 1639 if (err) 1640 return err; 1641 1642 if (msg_namelen > sizeof(to)) 1643 msg_namelen = sizeof(to); 1644 memcpy(&to, msg->msg_name, msg_namelen); 1645 msg_name = msg->msg_name; 1646 } 1647 1648 sinit = cmsgs.init; 1649 if (cmsgs.sinfo != NULL) { 1650 memset(&default_sinfo, 0, sizeof(default_sinfo)); 1651 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid; 1652 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags; 1653 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid; 1654 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context; 1655 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id; 1656 1657 sinfo = &default_sinfo; 1658 fill_sinfo_ttl = true; 1659 } else { 1660 sinfo = cmsgs.srinfo; 1661 } 1662 /* Did the user specify SNDINFO/SNDRCVINFO? */ 1663 if (sinfo) { 1664 sinfo_flags = sinfo->sinfo_flags; 1665 associd = sinfo->sinfo_assoc_id; 1666 } 1667 1668 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__, 1669 msg_len, sinfo_flags); 1670 1671 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */ 1672 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) { 1673 err = -EINVAL; 1674 goto out_nounlock; 1675 } 1676 1677 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero 1678 * length messages when SCTP_EOF|SCTP_ABORT is not set. 1679 * If SCTP_ABORT is set, the message length could be non zero with 1680 * the msg_iov set to the user abort reason. 1681 */ 1682 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) || 1683 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) { 1684 err = -EINVAL; 1685 goto out_nounlock; 1686 } 1687 1688 /* If SCTP_ADDR_OVER is set, there must be an address 1689 * specified in msg_name. 1690 */ 1691 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) { 1692 err = -EINVAL; 1693 goto out_nounlock; 1694 } 1695 1696 transport = NULL; 1697 1698 pr_debug("%s: about to look up association\n", __func__); 1699 1700 lock_sock(sk); 1701 1702 /* If a msg_name has been specified, assume this is to be used. */ 1703 if (msg_name) { 1704 /* Look for a matching association on the endpoint. */ 1705 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport); 1706 if (!asoc) { 1707 /* If we could not find a matching association on the 1708 * endpoint, make sure that it is not a TCP-style 1709 * socket that already has an association or there is 1710 * no peeled-off association on another socket. 1711 */ 1712 if ((sctp_style(sk, TCP) && 1713 sctp_sstate(sk, ESTABLISHED)) || 1714 sctp_endpoint_is_peeled_off(ep, &to)) { 1715 err = -EADDRNOTAVAIL; 1716 goto out_unlock; 1717 } 1718 } 1719 } else { 1720 asoc = sctp_id2assoc(sk, associd); 1721 if (!asoc) { 1722 err = -EPIPE; 1723 goto out_unlock; 1724 } 1725 } 1726 1727 if (asoc) { 1728 pr_debug("%s: just looked up association:%p\n", __func__, asoc); 1729 1730 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED 1731 * socket that has an association in CLOSED state. This can 1732 * happen when an accepted socket has an association that is 1733 * already CLOSED. 1734 */ 1735 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) { 1736 err = -EPIPE; 1737 goto out_unlock; 1738 } 1739 1740 if (sinfo_flags & SCTP_EOF) { 1741 pr_debug("%s: shutting down association:%p\n", 1742 __func__, asoc); 1743 1744 sctp_primitive_SHUTDOWN(net, asoc, NULL); 1745 err = 0; 1746 goto out_unlock; 1747 } 1748 if (sinfo_flags & SCTP_ABORT) { 1749 1750 chunk = sctp_make_abort_user(asoc, msg, msg_len); 1751 if (!chunk) { 1752 err = -ENOMEM; 1753 goto out_unlock; 1754 } 1755 1756 pr_debug("%s: aborting association:%p\n", 1757 __func__, asoc); 1758 1759 sctp_primitive_ABORT(net, asoc, chunk); 1760 err = 0; 1761 goto out_unlock; 1762 } 1763 } 1764 1765 /* Do we need to create the association? */ 1766 if (!asoc) { 1767 pr_debug("%s: there is no association yet\n", __func__); 1768 1769 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) { 1770 err = -EINVAL; 1771 goto out_unlock; 1772 } 1773 1774 /* Check for invalid stream against the stream counts, 1775 * either the default or the user specified stream counts. 1776 */ 1777 if (sinfo) { 1778 if (!sinit || !sinit->sinit_num_ostreams) { 1779 /* Check against the defaults. */ 1780 if (sinfo->sinfo_stream >= 1781 sp->initmsg.sinit_num_ostreams) { 1782 err = -EINVAL; 1783 goto out_unlock; 1784 } 1785 } else { 1786 /* Check against the requested. */ 1787 if (sinfo->sinfo_stream >= 1788 sinit->sinit_num_ostreams) { 1789 err = -EINVAL; 1790 goto out_unlock; 1791 } 1792 } 1793 } 1794 1795 /* 1796 * API 3.1.2 bind() - UDP Style Syntax 1797 * If a bind() or sctp_bindx() is not called prior to a 1798 * sendmsg() call that initiates a new association, the 1799 * system picks an ephemeral port and will choose an address 1800 * set equivalent to binding with a wildcard address. 1801 */ 1802 if (!ep->base.bind_addr.port) { 1803 if (sctp_autobind(sk)) { 1804 err = -EAGAIN; 1805 goto out_unlock; 1806 } 1807 } else { 1808 /* 1809 * If an unprivileged user inherits a one-to-many 1810 * style socket with open associations on a privileged 1811 * port, it MAY be permitted to accept new associations, 1812 * but it SHOULD NOT be permitted to open new 1813 * associations. 1814 */ 1815 if (ep->base.bind_addr.port < PROT_SOCK && 1816 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) { 1817 err = -EACCES; 1818 goto out_unlock; 1819 } 1820 } 1821 1822 scope = sctp_scope(&to); 1823 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL); 1824 if (!new_asoc) { 1825 err = -ENOMEM; 1826 goto out_unlock; 1827 } 1828 asoc = new_asoc; 1829 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL); 1830 if (err < 0) { 1831 err = -ENOMEM; 1832 goto out_free; 1833 } 1834 1835 /* If the SCTP_INIT ancillary data is specified, set all 1836 * the association init values accordingly. 1837 */ 1838 if (sinit) { 1839 if (sinit->sinit_num_ostreams) { 1840 asoc->c.sinit_num_ostreams = 1841 sinit->sinit_num_ostreams; 1842 } 1843 if (sinit->sinit_max_instreams) { 1844 asoc->c.sinit_max_instreams = 1845 sinit->sinit_max_instreams; 1846 } 1847 if (sinit->sinit_max_attempts) { 1848 asoc->max_init_attempts 1849 = sinit->sinit_max_attempts; 1850 } 1851 if (sinit->sinit_max_init_timeo) { 1852 asoc->max_init_timeo = 1853 msecs_to_jiffies(sinit->sinit_max_init_timeo); 1854 } 1855 } 1856 1857 /* Prime the peer's transport structures. */ 1858 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN); 1859 if (!transport) { 1860 err = -ENOMEM; 1861 goto out_free; 1862 } 1863 } 1864 1865 /* ASSERT: we have a valid association at this point. */ 1866 pr_debug("%s: we have a valid association\n", __func__); 1867 1868 if (!sinfo) { 1869 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up 1870 * one with some defaults. 1871 */ 1872 memset(&default_sinfo, 0, sizeof(default_sinfo)); 1873 default_sinfo.sinfo_stream = asoc->default_stream; 1874 default_sinfo.sinfo_flags = asoc->default_flags; 1875 default_sinfo.sinfo_ppid = asoc->default_ppid; 1876 default_sinfo.sinfo_context = asoc->default_context; 1877 default_sinfo.sinfo_timetolive = asoc->default_timetolive; 1878 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc); 1879 1880 sinfo = &default_sinfo; 1881 } else if (fill_sinfo_ttl) { 1882 /* In case SNDINFO was specified, we still need to fill 1883 * it with a default ttl from the assoc here. 1884 */ 1885 sinfo->sinfo_timetolive = asoc->default_timetolive; 1886 } 1887 1888 /* API 7.1.7, the sndbuf size per association bounds the 1889 * maximum size of data that can be sent in a single send call. 1890 */ 1891 if (msg_len > sk->sk_sndbuf) { 1892 err = -EMSGSIZE; 1893 goto out_free; 1894 } 1895 1896 if (asoc->pmtu_pending) 1897 sctp_assoc_pending_pmtu(sk, asoc); 1898 1899 /* If fragmentation is disabled and the message length exceeds the 1900 * association fragmentation point, return EMSGSIZE. The I-D 1901 * does not specify what this error is, but this looks like 1902 * a great fit. 1903 */ 1904 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) { 1905 err = -EMSGSIZE; 1906 goto out_free; 1907 } 1908 1909 /* Check for invalid stream. */ 1910 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) { 1911 err = -EINVAL; 1912 goto out_free; 1913 } 1914 1915 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); 1916 if (!sctp_wspace(asoc)) { 1917 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len); 1918 if (err) 1919 goto out_free; 1920 } 1921 1922 /* If an address is passed with the sendto/sendmsg call, it is used 1923 * to override the primary destination address in the TCP model, or 1924 * when SCTP_ADDR_OVER flag is set in the UDP model. 1925 */ 1926 if ((sctp_style(sk, TCP) && msg_name) || 1927 (sinfo_flags & SCTP_ADDR_OVER)) { 1928 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to); 1929 if (!chunk_tp) { 1930 err = -EINVAL; 1931 goto out_free; 1932 } 1933 } else 1934 chunk_tp = NULL; 1935 1936 /* Auto-connect, if we aren't connected already. */ 1937 if (sctp_state(asoc, CLOSED)) { 1938 err = sctp_primitive_ASSOCIATE(net, asoc, NULL); 1939 if (err < 0) 1940 goto out_free; 1941 1942 wait_connect = true; 1943 pr_debug("%s: we associated primitively\n", __func__); 1944 } 1945 1946 /* Break the message into multiple chunks of maximum size. */ 1947 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter); 1948 if (IS_ERR(datamsg)) { 1949 err = PTR_ERR(datamsg); 1950 goto out_free; 1951 } 1952 1953 /* Now send the (possibly) fragmented message. */ 1954 list_for_each_entry(chunk, &datamsg->chunks, frag_list) { 1955 /* Do accounting for the write space. */ 1956 sctp_set_owner_w(chunk); 1957 1958 chunk->transport = chunk_tp; 1959 } 1960 1961 /* Send it to the lower layers. Note: all chunks 1962 * must either fail or succeed. The lower layer 1963 * works that way today. Keep it that way or this 1964 * breaks. 1965 */ 1966 err = sctp_primitive_SEND(net, asoc, datamsg); 1967 sctp_datamsg_put(datamsg); 1968 /* Did the lower layer accept the chunk? */ 1969 if (err) 1970 goto out_free; 1971 1972 pr_debug("%s: we sent primitively\n", __func__); 1973 1974 err = msg_len; 1975 1976 if (unlikely(wait_connect)) { 1977 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT); 1978 sctp_wait_for_connect(asoc, &timeo); 1979 } 1980 1981 /* If we are already past ASSOCIATE, the lower 1982 * layers are responsible for association cleanup. 1983 */ 1984 goto out_unlock; 1985 1986 out_free: 1987 if (new_asoc) { 1988 sctp_unhash_established(asoc); 1989 sctp_association_free(asoc); 1990 } 1991 out_unlock: 1992 release_sock(sk); 1993 1994 out_nounlock: 1995 return sctp_error(sk, msg_flags, err); 1996 1997 #if 0 1998 do_sock_err: 1999 if (msg_len) 2000 err = msg_len; 2001 else 2002 err = sock_error(sk); 2003 goto out; 2004 2005 do_interrupted: 2006 if (msg_len) 2007 err = msg_len; 2008 goto out; 2009 #endif /* 0 */ 2010 } 2011 2012 /* This is an extended version of skb_pull() that removes the data from the 2013 * start of a skb even when data is spread across the list of skb's in the 2014 * frag_list. len specifies the total amount of data that needs to be removed. 2015 * when 'len' bytes could be removed from the skb, it returns 0. 2016 * If 'len' exceeds the total skb length, it returns the no. of bytes that 2017 * could not be removed. 2018 */ 2019 static int sctp_skb_pull(struct sk_buff *skb, int len) 2020 { 2021 struct sk_buff *list; 2022 int skb_len = skb_headlen(skb); 2023 int rlen; 2024 2025 if (len <= skb_len) { 2026 __skb_pull(skb, len); 2027 return 0; 2028 } 2029 len -= skb_len; 2030 __skb_pull(skb, skb_len); 2031 2032 skb_walk_frags(skb, list) { 2033 rlen = sctp_skb_pull(list, len); 2034 skb->len -= (len-rlen); 2035 skb->data_len -= (len-rlen); 2036 2037 if (!rlen) 2038 return 0; 2039 2040 len = rlen; 2041 } 2042 2043 return len; 2044 } 2045 2046 /* API 3.1.3 recvmsg() - UDP Style Syntax 2047 * 2048 * ssize_t recvmsg(int socket, struct msghdr *message, 2049 * int flags); 2050 * 2051 * socket - the socket descriptor of the endpoint. 2052 * message - pointer to the msghdr structure which contains a single 2053 * user message and possibly some ancillary data. 2054 * 2055 * See Section 5 for complete description of the data 2056 * structures. 2057 * 2058 * flags - flags sent or received with the user message, see Section 2059 * 5 for complete description of the flags. 2060 */ 2061 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 2062 int noblock, int flags, int *addr_len) 2063 { 2064 struct sctp_ulpevent *event = NULL; 2065 struct sctp_sock *sp = sctp_sk(sk); 2066 struct sk_buff *skb; 2067 int copied; 2068 int err = 0; 2069 int skb_len; 2070 2071 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, " 2072 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags, 2073 addr_len); 2074 2075 lock_sock(sk); 2076 2077 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) { 2078 err = -ENOTCONN; 2079 goto out; 2080 } 2081 2082 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err); 2083 if (!skb) 2084 goto out; 2085 2086 /* Get the total length of the skb including any skb's in the 2087 * frag_list. 2088 */ 2089 skb_len = skb->len; 2090 2091 copied = skb_len; 2092 if (copied > len) 2093 copied = len; 2094 2095 err = skb_copy_datagram_msg(skb, 0, msg, copied); 2096 2097 event = sctp_skb2event(skb); 2098 2099 if (err) 2100 goto out_free; 2101 2102 sock_recv_ts_and_drops(msg, sk, skb); 2103 if (sctp_ulpevent_is_notification(event)) { 2104 msg->msg_flags |= MSG_NOTIFICATION; 2105 sp->pf->event_msgname(event, msg->msg_name, addr_len); 2106 } else { 2107 sp->pf->skb_msgname(skb, msg->msg_name, addr_len); 2108 } 2109 2110 /* Check if we allow SCTP_NXTINFO. */ 2111 if (sp->recvnxtinfo) 2112 sctp_ulpevent_read_nxtinfo(event, msg, sk); 2113 /* Check if we allow SCTP_RCVINFO. */ 2114 if (sp->recvrcvinfo) 2115 sctp_ulpevent_read_rcvinfo(event, msg); 2116 /* Check if we allow SCTP_SNDRCVINFO. */ 2117 if (sp->subscribe.sctp_data_io_event) 2118 sctp_ulpevent_read_sndrcvinfo(event, msg); 2119 2120 err = copied; 2121 2122 /* If skb's length exceeds the user's buffer, update the skb and 2123 * push it back to the receive_queue so that the next call to 2124 * recvmsg() will return the remaining data. Don't set MSG_EOR. 2125 */ 2126 if (skb_len > copied) { 2127 msg->msg_flags &= ~MSG_EOR; 2128 if (flags & MSG_PEEK) 2129 goto out_free; 2130 sctp_skb_pull(skb, copied); 2131 skb_queue_head(&sk->sk_receive_queue, skb); 2132 2133 /* When only partial message is copied to the user, increase 2134 * rwnd by that amount. If all the data in the skb is read, 2135 * rwnd is updated when the event is freed. 2136 */ 2137 if (!sctp_ulpevent_is_notification(event)) 2138 sctp_assoc_rwnd_increase(event->asoc, copied); 2139 goto out; 2140 } else if ((event->msg_flags & MSG_NOTIFICATION) || 2141 (event->msg_flags & MSG_EOR)) 2142 msg->msg_flags |= MSG_EOR; 2143 else 2144 msg->msg_flags &= ~MSG_EOR; 2145 2146 out_free: 2147 if (flags & MSG_PEEK) { 2148 /* Release the skb reference acquired after peeking the skb in 2149 * sctp_skb_recv_datagram(). 2150 */ 2151 kfree_skb(skb); 2152 } else { 2153 /* Free the event which includes releasing the reference to 2154 * the owner of the skb, freeing the skb and updating the 2155 * rwnd. 2156 */ 2157 sctp_ulpevent_free(event); 2158 } 2159 out: 2160 release_sock(sk); 2161 return err; 2162 } 2163 2164 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS) 2165 * 2166 * This option is a on/off flag. If enabled no SCTP message 2167 * fragmentation will be performed. Instead if a message being sent 2168 * exceeds the current PMTU size, the message will NOT be sent and 2169 * instead a error will be indicated to the user. 2170 */ 2171 static int sctp_setsockopt_disable_fragments(struct sock *sk, 2172 char __user *optval, 2173 unsigned int optlen) 2174 { 2175 int val; 2176 2177 if (optlen < sizeof(int)) 2178 return -EINVAL; 2179 2180 if (get_user(val, (int __user *)optval)) 2181 return -EFAULT; 2182 2183 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1; 2184 2185 return 0; 2186 } 2187 2188 static int sctp_setsockopt_events(struct sock *sk, char __user *optval, 2189 unsigned int optlen) 2190 { 2191 struct sctp_association *asoc; 2192 struct sctp_ulpevent *event; 2193 2194 if (optlen > sizeof(struct sctp_event_subscribe)) 2195 return -EINVAL; 2196 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen)) 2197 return -EFAULT; 2198 2199 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT, 2200 * if there is no data to be sent or retransmit, the stack will 2201 * immediately send up this notification. 2202 */ 2203 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT, 2204 &sctp_sk(sk)->subscribe)) { 2205 asoc = sctp_id2assoc(sk, 0); 2206 2207 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) { 2208 event = sctp_ulpevent_make_sender_dry_event(asoc, 2209 GFP_ATOMIC); 2210 if (!event) 2211 return -ENOMEM; 2212 2213 sctp_ulpq_tail_event(&asoc->ulpq, event); 2214 } 2215 } 2216 2217 return 0; 2218 } 2219 2220 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE) 2221 * 2222 * This socket option is applicable to the UDP-style socket only. When 2223 * set it will cause associations that are idle for more than the 2224 * specified number of seconds to automatically close. An association 2225 * being idle is defined an association that has NOT sent or received 2226 * user data. The special value of '0' indicates that no automatic 2227 * close of any associations should be performed. The option expects an 2228 * integer defining the number of seconds of idle time before an 2229 * association is closed. 2230 */ 2231 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval, 2232 unsigned int optlen) 2233 { 2234 struct sctp_sock *sp = sctp_sk(sk); 2235 struct net *net = sock_net(sk); 2236 2237 /* Applicable to UDP-style socket only */ 2238 if (sctp_style(sk, TCP)) 2239 return -EOPNOTSUPP; 2240 if (optlen != sizeof(int)) 2241 return -EINVAL; 2242 if (copy_from_user(&sp->autoclose, optval, optlen)) 2243 return -EFAULT; 2244 2245 if (sp->autoclose > net->sctp.max_autoclose) 2246 sp->autoclose = net->sctp.max_autoclose; 2247 2248 return 0; 2249 } 2250 2251 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS) 2252 * 2253 * Applications can enable or disable heartbeats for any peer address of 2254 * an association, modify an address's heartbeat interval, force a 2255 * heartbeat to be sent immediately, and adjust the address's maximum 2256 * number of retransmissions sent before an address is considered 2257 * unreachable. The following structure is used to access and modify an 2258 * address's parameters: 2259 * 2260 * struct sctp_paddrparams { 2261 * sctp_assoc_t spp_assoc_id; 2262 * struct sockaddr_storage spp_address; 2263 * uint32_t spp_hbinterval; 2264 * uint16_t spp_pathmaxrxt; 2265 * uint32_t spp_pathmtu; 2266 * uint32_t spp_sackdelay; 2267 * uint32_t spp_flags; 2268 * }; 2269 * 2270 * spp_assoc_id - (one-to-many style socket) This is filled in the 2271 * application, and identifies the association for 2272 * this query. 2273 * spp_address - This specifies which address is of interest. 2274 * spp_hbinterval - This contains the value of the heartbeat interval, 2275 * in milliseconds. If a value of zero 2276 * is present in this field then no changes are to 2277 * be made to this parameter. 2278 * spp_pathmaxrxt - This contains the maximum number of 2279 * retransmissions before this address shall be 2280 * considered unreachable. If a value of zero 2281 * is present in this field then no changes are to 2282 * be made to this parameter. 2283 * spp_pathmtu - When Path MTU discovery is disabled the value 2284 * specified here will be the "fixed" path mtu. 2285 * Note that if the spp_address field is empty 2286 * then all associations on this address will 2287 * have this fixed path mtu set upon them. 2288 * 2289 * spp_sackdelay - When delayed sack is enabled, this value specifies 2290 * the number of milliseconds that sacks will be delayed 2291 * for. This value will apply to all addresses of an 2292 * association if the spp_address field is empty. Note 2293 * also, that if delayed sack is enabled and this 2294 * value is set to 0, no change is made to the last 2295 * recorded delayed sack timer value. 2296 * 2297 * spp_flags - These flags are used to control various features 2298 * on an association. The flag field may contain 2299 * zero or more of the following options. 2300 * 2301 * SPP_HB_ENABLE - Enable heartbeats on the 2302 * specified address. Note that if the address 2303 * field is empty all addresses for the association 2304 * have heartbeats enabled upon them. 2305 * 2306 * SPP_HB_DISABLE - Disable heartbeats on the 2307 * speicifed address. Note that if the address 2308 * field is empty all addresses for the association 2309 * will have their heartbeats disabled. Note also 2310 * that SPP_HB_ENABLE and SPP_HB_DISABLE are 2311 * mutually exclusive, only one of these two should 2312 * be specified. Enabling both fields will have 2313 * undetermined results. 2314 * 2315 * SPP_HB_DEMAND - Request a user initiated heartbeat 2316 * to be made immediately. 2317 * 2318 * SPP_HB_TIME_IS_ZERO - Specify's that the time for 2319 * heartbeat delayis to be set to the value of 0 2320 * milliseconds. 2321 * 2322 * SPP_PMTUD_ENABLE - This field will enable PMTU 2323 * discovery upon the specified address. Note that 2324 * if the address feild is empty then all addresses 2325 * on the association are effected. 2326 * 2327 * SPP_PMTUD_DISABLE - This field will disable PMTU 2328 * discovery upon the specified address. Note that 2329 * if the address feild is empty then all addresses 2330 * on the association are effected. Not also that 2331 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually 2332 * exclusive. Enabling both will have undetermined 2333 * results. 2334 * 2335 * SPP_SACKDELAY_ENABLE - Setting this flag turns 2336 * on delayed sack. The time specified in spp_sackdelay 2337 * is used to specify the sack delay for this address. Note 2338 * that if spp_address is empty then all addresses will 2339 * enable delayed sack and take on the sack delay 2340 * value specified in spp_sackdelay. 2341 * SPP_SACKDELAY_DISABLE - Setting this flag turns 2342 * off delayed sack. If the spp_address field is blank then 2343 * delayed sack is disabled for the entire association. Note 2344 * also that this field is mutually exclusive to 2345 * SPP_SACKDELAY_ENABLE, setting both will have undefined 2346 * results. 2347 */ 2348 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params, 2349 struct sctp_transport *trans, 2350 struct sctp_association *asoc, 2351 struct sctp_sock *sp, 2352 int hb_change, 2353 int pmtud_change, 2354 int sackdelay_change) 2355 { 2356 int error; 2357 2358 if (params->spp_flags & SPP_HB_DEMAND && trans) { 2359 struct net *net = sock_net(trans->asoc->base.sk); 2360 2361 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans); 2362 if (error) 2363 return error; 2364 } 2365 2366 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of 2367 * this field is ignored. Note also that a value of zero indicates 2368 * the current setting should be left unchanged. 2369 */ 2370 if (params->spp_flags & SPP_HB_ENABLE) { 2371 2372 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is 2373 * set. This lets us use 0 value when this flag 2374 * is set. 2375 */ 2376 if (params->spp_flags & SPP_HB_TIME_IS_ZERO) 2377 params->spp_hbinterval = 0; 2378 2379 if (params->spp_hbinterval || 2380 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) { 2381 if (trans) { 2382 trans->hbinterval = 2383 msecs_to_jiffies(params->spp_hbinterval); 2384 } else if (asoc) { 2385 asoc->hbinterval = 2386 msecs_to_jiffies(params->spp_hbinterval); 2387 } else { 2388 sp->hbinterval = params->spp_hbinterval; 2389 } 2390 } 2391 } 2392 2393 if (hb_change) { 2394 if (trans) { 2395 trans->param_flags = 2396 (trans->param_flags & ~SPP_HB) | hb_change; 2397 } else if (asoc) { 2398 asoc->param_flags = 2399 (asoc->param_flags & ~SPP_HB) | hb_change; 2400 } else { 2401 sp->param_flags = 2402 (sp->param_flags & ~SPP_HB) | hb_change; 2403 } 2404 } 2405 2406 /* When Path MTU discovery is disabled the value specified here will 2407 * be the "fixed" path mtu (i.e. the value of the spp_flags field must 2408 * include the flag SPP_PMTUD_DISABLE for this field to have any 2409 * effect). 2410 */ 2411 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) { 2412 if (trans) { 2413 trans->pathmtu = params->spp_pathmtu; 2414 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc); 2415 } else if (asoc) { 2416 asoc->pathmtu = params->spp_pathmtu; 2417 sctp_frag_point(asoc, params->spp_pathmtu); 2418 } else { 2419 sp->pathmtu = params->spp_pathmtu; 2420 } 2421 } 2422 2423 if (pmtud_change) { 2424 if (trans) { 2425 int update = (trans->param_flags & SPP_PMTUD_DISABLE) && 2426 (params->spp_flags & SPP_PMTUD_ENABLE); 2427 trans->param_flags = 2428 (trans->param_flags & ~SPP_PMTUD) | pmtud_change; 2429 if (update) { 2430 sctp_transport_pmtu(trans, sctp_opt2sk(sp)); 2431 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc); 2432 } 2433 } else if (asoc) { 2434 asoc->param_flags = 2435 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change; 2436 } else { 2437 sp->param_flags = 2438 (sp->param_flags & ~SPP_PMTUD) | pmtud_change; 2439 } 2440 } 2441 2442 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the 2443 * value of this field is ignored. Note also that a value of zero 2444 * indicates the current setting should be left unchanged. 2445 */ 2446 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) { 2447 if (trans) { 2448 trans->sackdelay = 2449 msecs_to_jiffies(params->spp_sackdelay); 2450 } else if (asoc) { 2451 asoc->sackdelay = 2452 msecs_to_jiffies(params->spp_sackdelay); 2453 } else { 2454 sp->sackdelay = params->spp_sackdelay; 2455 } 2456 } 2457 2458 if (sackdelay_change) { 2459 if (trans) { 2460 trans->param_flags = 2461 (trans->param_flags & ~SPP_SACKDELAY) | 2462 sackdelay_change; 2463 } else if (asoc) { 2464 asoc->param_flags = 2465 (asoc->param_flags & ~SPP_SACKDELAY) | 2466 sackdelay_change; 2467 } else { 2468 sp->param_flags = 2469 (sp->param_flags & ~SPP_SACKDELAY) | 2470 sackdelay_change; 2471 } 2472 } 2473 2474 /* Note that a value of zero indicates the current setting should be 2475 left unchanged. 2476 */ 2477 if (params->spp_pathmaxrxt) { 2478 if (trans) { 2479 trans->pathmaxrxt = params->spp_pathmaxrxt; 2480 } else if (asoc) { 2481 asoc->pathmaxrxt = params->spp_pathmaxrxt; 2482 } else { 2483 sp->pathmaxrxt = params->spp_pathmaxrxt; 2484 } 2485 } 2486 2487 return 0; 2488 } 2489 2490 static int sctp_setsockopt_peer_addr_params(struct sock *sk, 2491 char __user *optval, 2492 unsigned int optlen) 2493 { 2494 struct sctp_paddrparams params; 2495 struct sctp_transport *trans = NULL; 2496 struct sctp_association *asoc = NULL; 2497 struct sctp_sock *sp = sctp_sk(sk); 2498 int error; 2499 int hb_change, pmtud_change, sackdelay_change; 2500 2501 if (optlen != sizeof(struct sctp_paddrparams)) 2502 return -EINVAL; 2503 2504 if (copy_from_user(¶ms, optval, optlen)) 2505 return -EFAULT; 2506 2507 /* Validate flags and value parameters. */ 2508 hb_change = params.spp_flags & SPP_HB; 2509 pmtud_change = params.spp_flags & SPP_PMTUD; 2510 sackdelay_change = params.spp_flags & SPP_SACKDELAY; 2511 2512 if (hb_change == SPP_HB || 2513 pmtud_change == SPP_PMTUD || 2514 sackdelay_change == SPP_SACKDELAY || 2515 params.spp_sackdelay > 500 || 2516 (params.spp_pathmtu && 2517 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT)) 2518 return -EINVAL; 2519 2520 /* If an address other than INADDR_ANY is specified, and 2521 * no transport is found, then the request is invalid. 2522 */ 2523 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) { 2524 trans = sctp_addr_id2transport(sk, ¶ms.spp_address, 2525 params.spp_assoc_id); 2526 if (!trans) 2527 return -EINVAL; 2528 } 2529 2530 /* Get association, if assoc_id != 0 and the socket is a one 2531 * to many style socket, and an association was not found, then 2532 * the id was invalid. 2533 */ 2534 asoc = sctp_id2assoc(sk, params.spp_assoc_id); 2535 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) 2536 return -EINVAL; 2537 2538 /* Heartbeat demand can only be sent on a transport or 2539 * association, but not a socket. 2540 */ 2541 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc) 2542 return -EINVAL; 2543 2544 /* Process parameters. */ 2545 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp, 2546 hb_change, pmtud_change, 2547 sackdelay_change); 2548 2549 if (error) 2550 return error; 2551 2552 /* If changes are for association, also apply parameters to each 2553 * transport. 2554 */ 2555 if (!trans && asoc) { 2556 list_for_each_entry(trans, &asoc->peer.transport_addr_list, 2557 transports) { 2558 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp, 2559 hb_change, pmtud_change, 2560 sackdelay_change); 2561 } 2562 } 2563 2564 return 0; 2565 } 2566 2567 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags) 2568 { 2569 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE; 2570 } 2571 2572 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags) 2573 { 2574 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE; 2575 } 2576 2577 /* 2578 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK) 2579 * 2580 * This option will effect the way delayed acks are performed. This 2581 * option allows you to get or set the delayed ack time, in 2582 * milliseconds. It also allows changing the delayed ack frequency. 2583 * Changing the frequency to 1 disables the delayed sack algorithm. If 2584 * the assoc_id is 0, then this sets or gets the endpoints default 2585 * values. If the assoc_id field is non-zero, then the set or get 2586 * effects the specified association for the one to many model (the 2587 * assoc_id field is ignored by the one to one model). Note that if 2588 * sack_delay or sack_freq are 0 when setting this option, then the 2589 * current values will remain unchanged. 2590 * 2591 * struct sctp_sack_info { 2592 * sctp_assoc_t sack_assoc_id; 2593 * uint32_t sack_delay; 2594 * uint32_t sack_freq; 2595 * }; 2596 * 2597 * sack_assoc_id - This parameter, indicates which association the user 2598 * is performing an action upon. Note that if this field's value is 2599 * zero then the endpoints default value is changed (effecting future 2600 * associations only). 2601 * 2602 * sack_delay - This parameter contains the number of milliseconds that 2603 * the user is requesting the delayed ACK timer be set to. Note that 2604 * this value is defined in the standard to be between 200 and 500 2605 * milliseconds. 2606 * 2607 * sack_freq - This parameter contains the number of packets that must 2608 * be received before a sack is sent without waiting for the delay 2609 * timer to expire. The default value for this is 2, setting this 2610 * value to 1 will disable the delayed sack algorithm. 2611 */ 2612 2613 static int sctp_setsockopt_delayed_ack(struct sock *sk, 2614 char __user *optval, unsigned int optlen) 2615 { 2616 struct sctp_sack_info params; 2617 struct sctp_transport *trans = NULL; 2618 struct sctp_association *asoc = NULL; 2619 struct sctp_sock *sp = sctp_sk(sk); 2620 2621 if (optlen == sizeof(struct sctp_sack_info)) { 2622 if (copy_from_user(¶ms, optval, optlen)) 2623 return -EFAULT; 2624 2625 if (params.sack_delay == 0 && params.sack_freq == 0) 2626 return 0; 2627 } else if (optlen == sizeof(struct sctp_assoc_value)) { 2628 pr_warn_ratelimited(DEPRECATED 2629 "%s (pid %d) " 2630 "Use of struct sctp_assoc_value in delayed_ack socket option.\n" 2631 "Use struct sctp_sack_info instead\n", 2632 current->comm, task_pid_nr(current)); 2633 if (copy_from_user(¶ms, optval, optlen)) 2634 return -EFAULT; 2635 2636 if (params.sack_delay == 0) 2637 params.sack_freq = 1; 2638 else 2639 params.sack_freq = 0; 2640 } else 2641 return -EINVAL; 2642 2643 /* Validate value parameter. */ 2644 if (params.sack_delay > 500) 2645 return -EINVAL; 2646 2647 /* Get association, if sack_assoc_id != 0 and the socket is a one 2648 * to many style socket, and an association was not found, then 2649 * the id was invalid. 2650 */ 2651 asoc = sctp_id2assoc(sk, params.sack_assoc_id); 2652 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP)) 2653 return -EINVAL; 2654 2655 if (params.sack_delay) { 2656 if (asoc) { 2657 asoc->sackdelay = 2658 msecs_to_jiffies(params.sack_delay); 2659 asoc->param_flags = 2660 sctp_spp_sackdelay_enable(asoc->param_flags); 2661 } else { 2662 sp->sackdelay = params.sack_delay; 2663 sp->param_flags = 2664 sctp_spp_sackdelay_enable(sp->param_flags); 2665 } 2666 } 2667 2668 if (params.sack_freq == 1) { 2669 if (asoc) { 2670 asoc->param_flags = 2671 sctp_spp_sackdelay_disable(asoc->param_flags); 2672 } else { 2673 sp->param_flags = 2674 sctp_spp_sackdelay_disable(sp->param_flags); 2675 } 2676 } else if (params.sack_freq > 1) { 2677 if (asoc) { 2678 asoc->sackfreq = params.sack_freq; 2679 asoc->param_flags = 2680 sctp_spp_sackdelay_enable(asoc->param_flags); 2681 } else { 2682 sp->sackfreq = params.sack_freq; 2683 sp->param_flags = 2684 sctp_spp_sackdelay_enable(sp->param_flags); 2685 } 2686 } 2687 2688 /* If change is for association, also apply to each transport. */ 2689 if (asoc) { 2690 list_for_each_entry(trans, &asoc->peer.transport_addr_list, 2691 transports) { 2692 if (params.sack_delay) { 2693 trans->sackdelay = 2694 msecs_to_jiffies(params.sack_delay); 2695 trans->param_flags = 2696 sctp_spp_sackdelay_enable(trans->param_flags); 2697 } 2698 if (params.sack_freq == 1) { 2699 trans->param_flags = 2700 sctp_spp_sackdelay_disable(trans->param_flags); 2701 } else if (params.sack_freq > 1) { 2702 trans->sackfreq = params.sack_freq; 2703 trans->param_flags = 2704 sctp_spp_sackdelay_enable(trans->param_flags); 2705 } 2706 } 2707 } 2708 2709 return 0; 2710 } 2711 2712 /* 7.1.3 Initialization Parameters (SCTP_INITMSG) 2713 * 2714 * Applications can specify protocol parameters for the default association 2715 * initialization. The option name argument to setsockopt() and getsockopt() 2716 * is SCTP_INITMSG. 2717 * 2718 * Setting initialization parameters is effective only on an unconnected 2719 * socket (for UDP-style sockets only future associations are effected 2720 * by the change). With TCP-style sockets, this option is inherited by 2721 * sockets derived from a listener socket. 2722 */ 2723 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen) 2724 { 2725 struct sctp_initmsg sinit; 2726 struct sctp_sock *sp = sctp_sk(sk); 2727 2728 if (optlen != sizeof(struct sctp_initmsg)) 2729 return -EINVAL; 2730 if (copy_from_user(&sinit, optval, optlen)) 2731 return -EFAULT; 2732 2733 if (sinit.sinit_num_ostreams) 2734 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams; 2735 if (sinit.sinit_max_instreams) 2736 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams; 2737 if (sinit.sinit_max_attempts) 2738 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts; 2739 if (sinit.sinit_max_init_timeo) 2740 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo; 2741 2742 return 0; 2743 } 2744 2745 /* 2746 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM) 2747 * 2748 * Applications that wish to use the sendto() system call may wish to 2749 * specify a default set of parameters that would normally be supplied 2750 * through the inclusion of ancillary data. This socket option allows 2751 * such an application to set the default sctp_sndrcvinfo structure. 2752 * The application that wishes to use this socket option simply passes 2753 * in to this call the sctp_sndrcvinfo structure defined in Section 2754 * 5.2.2) The input parameters accepted by this call include 2755 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context, 2756 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in 2757 * to this call if the caller is using the UDP model. 2758 */ 2759 static int sctp_setsockopt_default_send_param(struct sock *sk, 2760 char __user *optval, 2761 unsigned int optlen) 2762 { 2763 struct sctp_sock *sp = sctp_sk(sk); 2764 struct sctp_association *asoc; 2765 struct sctp_sndrcvinfo info; 2766 2767 if (optlen != sizeof(info)) 2768 return -EINVAL; 2769 if (copy_from_user(&info, optval, optlen)) 2770 return -EFAULT; 2771 if (info.sinfo_flags & 2772 ~(SCTP_UNORDERED | SCTP_ADDR_OVER | 2773 SCTP_ABORT | SCTP_EOF)) 2774 return -EINVAL; 2775 2776 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id); 2777 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP)) 2778 return -EINVAL; 2779 if (asoc) { 2780 asoc->default_stream = info.sinfo_stream; 2781 asoc->default_flags = info.sinfo_flags; 2782 asoc->default_ppid = info.sinfo_ppid; 2783 asoc->default_context = info.sinfo_context; 2784 asoc->default_timetolive = info.sinfo_timetolive; 2785 } else { 2786 sp->default_stream = info.sinfo_stream; 2787 sp->default_flags = info.sinfo_flags; 2788 sp->default_ppid = info.sinfo_ppid; 2789 sp->default_context = info.sinfo_context; 2790 sp->default_timetolive = info.sinfo_timetolive; 2791 } 2792 2793 return 0; 2794 } 2795 2796 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters 2797 * (SCTP_DEFAULT_SNDINFO) 2798 */ 2799 static int sctp_setsockopt_default_sndinfo(struct sock *sk, 2800 char __user *optval, 2801 unsigned int optlen) 2802 { 2803 struct sctp_sock *sp = sctp_sk(sk); 2804 struct sctp_association *asoc; 2805 struct sctp_sndinfo info; 2806 2807 if (optlen != sizeof(info)) 2808 return -EINVAL; 2809 if (copy_from_user(&info, optval, optlen)) 2810 return -EFAULT; 2811 if (info.snd_flags & 2812 ~(SCTP_UNORDERED | SCTP_ADDR_OVER | 2813 SCTP_ABORT | SCTP_EOF)) 2814 return -EINVAL; 2815 2816 asoc = sctp_id2assoc(sk, info.snd_assoc_id); 2817 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP)) 2818 return -EINVAL; 2819 if (asoc) { 2820 asoc->default_stream = info.snd_sid; 2821 asoc->default_flags = info.snd_flags; 2822 asoc->default_ppid = info.snd_ppid; 2823 asoc->default_context = info.snd_context; 2824 } else { 2825 sp->default_stream = info.snd_sid; 2826 sp->default_flags = info.snd_flags; 2827 sp->default_ppid = info.snd_ppid; 2828 sp->default_context = info.snd_context; 2829 } 2830 2831 return 0; 2832 } 2833 2834 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR) 2835 * 2836 * Requests that the local SCTP stack use the enclosed peer address as 2837 * the association primary. The enclosed address must be one of the 2838 * association peer's addresses. 2839 */ 2840 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval, 2841 unsigned int optlen) 2842 { 2843 struct sctp_prim prim; 2844 struct sctp_transport *trans; 2845 2846 if (optlen != sizeof(struct sctp_prim)) 2847 return -EINVAL; 2848 2849 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim))) 2850 return -EFAULT; 2851 2852 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id); 2853 if (!trans) 2854 return -EINVAL; 2855 2856 sctp_assoc_set_primary(trans->asoc, trans); 2857 2858 return 0; 2859 } 2860 2861 /* 2862 * 7.1.5 SCTP_NODELAY 2863 * 2864 * Turn on/off any Nagle-like algorithm. This means that packets are 2865 * generally sent as soon as possible and no unnecessary delays are 2866 * introduced, at the cost of more packets in the network. Expects an 2867 * integer boolean flag. 2868 */ 2869 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval, 2870 unsigned int optlen) 2871 { 2872 int val; 2873 2874 if (optlen < sizeof(int)) 2875 return -EINVAL; 2876 if (get_user(val, (int __user *)optval)) 2877 return -EFAULT; 2878 2879 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1; 2880 return 0; 2881 } 2882 2883 /* 2884 * 2885 * 7.1.1 SCTP_RTOINFO 2886 * 2887 * The protocol parameters used to initialize and bound retransmission 2888 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access 2889 * and modify these parameters. 2890 * All parameters are time values, in milliseconds. A value of 0, when 2891 * modifying the parameters, indicates that the current value should not 2892 * be changed. 2893 * 2894 */ 2895 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen) 2896 { 2897 struct sctp_rtoinfo rtoinfo; 2898 struct sctp_association *asoc; 2899 unsigned long rto_min, rto_max; 2900 struct sctp_sock *sp = sctp_sk(sk); 2901 2902 if (optlen != sizeof (struct sctp_rtoinfo)) 2903 return -EINVAL; 2904 2905 if (copy_from_user(&rtoinfo, optval, optlen)) 2906 return -EFAULT; 2907 2908 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id); 2909 2910 /* Set the values to the specific association */ 2911 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP)) 2912 return -EINVAL; 2913 2914 rto_max = rtoinfo.srto_max; 2915 rto_min = rtoinfo.srto_min; 2916 2917 if (rto_max) 2918 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max; 2919 else 2920 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max; 2921 2922 if (rto_min) 2923 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min; 2924 else 2925 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min; 2926 2927 if (rto_min > rto_max) 2928 return -EINVAL; 2929 2930 if (asoc) { 2931 if (rtoinfo.srto_initial != 0) 2932 asoc->rto_initial = 2933 msecs_to_jiffies(rtoinfo.srto_initial); 2934 asoc->rto_max = rto_max; 2935 asoc->rto_min = rto_min; 2936 } else { 2937 /* If there is no association or the association-id = 0 2938 * set the values to the endpoint. 2939 */ 2940 if (rtoinfo.srto_initial != 0) 2941 sp->rtoinfo.srto_initial = rtoinfo.srto_initial; 2942 sp->rtoinfo.srto_max = rto_max; 2943 sp->rtoinfo.srto_min = rto_min; 2944 } 2945 2946 return 0; 2947 } 2948 2949 /* 2950 * 2951 * 7.1.2 SCTP_ASSOCINFO 2952 * 2953 * This option is used to tune the maximum retransmission attempts 2954 * of the association. 2955 * Returns an error if the new association retransmission value is 2956 * greater than the sum of the retransmission value of the peer. 2957 * See [SCTP] for more information. 2958 * 2959 */ 2960 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen) 2961 { 2962 2963 struct sctp_assocparams assocparams; 2964 struct sctp_association *asoc; 2965 2966 if (optlen != sizeof(struct sctp_assocparams)) 2967 return -EINVAL; 2968 if (copy_from_user(&assocparams, optval, optlen)) 2969 return -EFAULT; 2970 2971 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id); 2972 2973 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP)) 2974 return -EINVAL; 2975 2976 /* Set the values to the specific association */ 2977 if (asoc) { 2978 if (assocparams.sasoc_asocmaxrxt != 0) { 2979 __u32 path_sum = 0; 2980 int paths = 0; 2981 struct sctp_transport *peer_addr; 2982 2983 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list, 2984 transports) { 2985 path_sum += peer_addr->pathmaxrxt; 2986 paths++; 2987 } 2988 2989 /* Only validate asocmaxrxt if we have more than 2990 * one path/transport. We do this because path 2991 * retransmissions are only counted when we have more 2992 * then one path. 2993 */ 2994 if (paths > 1 && 2995 assocparams.sasoc_asocmaxrxt > path_sum) 2996 return -EINVAL; 2997 2998 asoc->max_retrans = assocparams.sasoc_asocmaxrxt; 2999 } 3000 3001 if (assocparams.sasoc_cookie_life != 0) 3002 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life); 3003 } else { 3004 /* Set the values to the endpoint */ 3005 struct sctp_sock *sp = sctp_sk(sk); 3006 3007 if (assocparams.sasoc_asocmaxrxt != 0) 3008 sp->assocparams.sasoc_asocmaxrxt = 3009 assocparams.sasoc_asocmaxrxt; 3010 if (assocparams.sasoc_cookie_life != 0) 3011 sp->assocparams.sasoc_cookie_life = 3012 assocparams.sasoc_cookie_life; 3013 } 3014 return 0; 3015 } 3016 3017 /* 3018 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR) 3019 * 3020 * This socket option is a boolean flag which turns on or off mapped V4 3021 * addresses. If this option is turned on and the socket is type 3022 * PF_INET6, then IPv4 addresses will be mapped to V6 representation. 3023 * If this option is turned off, then no mapping will be done of V4 3024 * addresses and a user will receive both PF_INET6 and PF_INET type 3025 * addresses on the socket. 3026 */ 3027 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen) 3028 { 3029 int val; 3030 struct sctp_sock *sp = sctp_sk(sk); 3031 3032 if (optlen < sizeof(int)) 3033 return -EINVAL; 3034 if (get_user(val, (int __user *)optval)) 3035 return -EFAULT; 3036 if (val) 3037 sp->v4mapped = 1; 3038 else 3039 sp->v4mapped = 0; 3040 3041 return 0; 3042 } 3043 3044 /* 3045 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG) 3046 * This option will get or set the maximum size to put in any outgoing 3047 * SCTP DATA chunk. If a message is larger than this size it will be 3048 * fragmented by SCTP into the specified size. Note that the underlying 3049 * SCTP implementation may fragment into smaller sized chunks when the 3050 * PMTU of the underlying association is smaller than the value set by 3051 * the user. The default value for this option is '0' which indicates 3052 * the user is NOT limiting fragmentation and only the PMTU will effect 3053 * SCTP's choice of DATA chunk size. Note also that values set larger 3054 * than the maximum size of an IP datagram will effectively let SCTP 3055 * control fragmentation (i.e. the same as setting this option to 0). 3056 * 3057 * The following structure is used to access and modify this parameter: 3058 * 3059 * struct sctp_assoc_value { 3060 * sctp_assoc_t assoc_id; 3061 * uint32_t assoc_value; 3062 * }; 3063 * 3064 * assoc_id: This parameter is ignored for one-to-one style sockets. 3065 * For one-to-many style sockets this parameter indicates which 3066 * association the user is performing an action upon. Note that if 3067 * this field's value is zero then the endpoints default value is 3068 * changed (effecting future associations only). 3069 * assoc_value: This parameter specifies the maximum size in bytes. 3070 */ 3071 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen) 3072 { 3073 struct sctp_assoc_value params; 3074 struct sctp_association *asoc; 3075 struct sctp_sock *sp = sctp_sk(sk); 3076 int val; 3077 3078 if (optlen == sizeof(int)) { 3079 pr_warn_ratelimited(DEPRECATED 3080 "%s (pid %d) " 3081 "Use of int in maxseg socket option.\n" 3082 "Use struct sctp_assoc_value instead\n", 3083 current->comm, task_pid_nr(current)); 3084 if (copy_from_user(&val, optval, optlen)) 3085 return -EFAULT; 3086 params.assoc_id = 0; 3087 } else if (optlen == sizeof(struct sctp_assoc_value)) { 3088 if (copy_from_user(¶ms, optval, optlen)) 3089 return -EFAULT; 3090 val = params.assoc_value; 3091 } else 3092 return -EINVAL; 3093 3094 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN))) 3095 return -EINVAL; 3096 3097 asoc = sctp_id2assoc(sk, params.assoc_id); 3098 if (!asoc && params.assoc_id && sctp_style(sk, UDP)) 3099 return -EINVAL; 3100 3101 if (asoc) { 3102 if (val == 0) { 3103 val = asoc->pathmtu; 3104 val -= sp->pf->af->net_header_len; 3105 val -= sizeof(struct sctphdr) + 3106 sizeof(struct sctp_data_chunk); 3107 } 3108 asoc->user_frag = val; 3109 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu); 3110 } else { 3111 sp->user_frag = val; 3112 } 3113 3114 return 0; 3115 } 3116 3117 3118 /* 3119 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR) 3120 * 3121 * Requests that the peer mark the enclosed address as the association 3122 * primary. The enclosed address must be one of the association's 3123 * locally bound addresses. The following structure is used to make a 3124 * set primary request: 3125 */ 3126 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval, 3127 unsigned int optlen) 3128 { 3129 struct net *net = sock_net(sk); 3130 struct sctp_sock *sp; 3131 struct sctp_association *asoc = NULL; 3132 struct sctp_setpeerprim prim; 3133 struct sctp_chunk *chunk; 3134 struct sctp_af *af; 3135 int err; 3136 3137 sp = sctp_sk(sk); 3138 3139 if (!net->sctp.addip_enable) 3140 return -EPERM; 3141 3142 if (optlen != sizeof(struct sctp_setpeerprim)) 3143 return -EINVAL; 3144 3145 if (copy_from_user(&prim, optval, optlen)) 3146 return -EFAULT; 3147 3148 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id); 3149 if (!asoc) 3150 return -EINVAL; 3151 3152 if (!asoc->peer.asconf_capable) 3153 return -EPERM; 3154 3155 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY) 3156 return -EPERM; 3157 3158 if (!sctp_state(asoc, ESTABLISHED)) 3159 return -ENOTCONN; 3160 3161 af = sctp_get_af_specific(prim.sspp_addr.ss_family); 3162 if (!af) 3163 return -EINVAL; 3164 3165 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL)) 3166 return -EADDRNOTAVAIL; 3167 3168 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr)) 3169 return -EADDRNOTAVAIL; 3170 3171 /* Create an ASCONF chunk with SET_PRIMARY parameter */ 3172 chunk = sctp_make_asconf_set_prim(asoc, 3173 (union sctp_addr *)&prim.sspp_addr); 3174 if (!chunk) 3175 return -ENOMEM; 3176 3177 err = sctp_send_asconf(asoc, chunk); 3178 3179 pr_debug("%s: we set peer primary addr primitively\n", __func__); 3180 3181 return err; 3182 } 3183 3184 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval, 3185 unsigned int optlen) 3186 { 3187 struct sctp_setadaptation adaptation; 3188 3189 if (optlen != sizeof(struct sctp_setadaptation)) 3190 return -EINVAL; 3191 if (copy_from_user(&adaptation, optval, optlen)) 3192 return -EFAULT; 3193 3194 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind; 3195 3196 return 0; 3197 } 3198 3199 /* 3200 * 7.1.29. Set or Get the default context (SCTP_CONTEXT) 3201 * 3202 * The context field in the sctp_sndrcvinfo structure is normally only 3203 * used when a failed message is retrieved holding the value that was 3204 * sent down on the actual send call. This option allows the setting of 3205 * a default context on an association basis that will be received on 3206 * reading messages from the peer. This is especially helpful in the 3207 * one-2-many model for an application to keep some reference to an 3208 * internal state machine that is processing messages on the 3209 * association. Note that the setting of this value only effects 3210 * received messages from the peer and does not effect the value that is 3211 * saved with outbound messages. 3212 */ 3213 static int sctp_setsockopt_context(struct sock *sk, char __user *optval, 3214 unsigned int optlen) 3215 { 3216 struct sctp_assoc_value params; 3217 struct sctp_sock *sp; 3218 struct sctp_association *asoc; 3219 3220 if (optlen != sizeof(struct sctp_assoc_value)) 3221 return -EINVAL; 3222 if (copy_from_user(¶ms, optval, optlen)) 3223 return -EFAULT; 3224 3225 sp = sctp_sk(sk); 3226 3227 if (params.assoc_id != 0) { 3228 asoc = sctp_id2assoc(sk, params.assoc_id); 3229 if (!asoc) 3230 return -EINVAL; 3231 asoc->default_rcv_context = params.assoc_value; 3232 } else { 3233 sp->default_rcv_context = params.assoc_value; 3234 } 3235 3236 return 0; 3237 } 3238 3239 /* 3240 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE) 3241 * 3242 * This options will at a minimum specify if the implementation is doing 3243 * fragmented interleave. Fragmented interleave, for a one to many 3244 * socket, is when subsequent calls to receive a message may return 3245 * parts of messages from different associations. Some implementations 3246 * may allow you to turn this value on or off. If so, when turned off, 3247 * no fragment interleave will occur (which will cause a head of line 3248 * blocking amongst multiple associations sharing the same one to many 3249 * socket). When this option is turned on, then each receive call may 3250 * come from a different association (thus the user must receive data 3251 * with the extended calls (e.g. sctp_recvmsg) to keep track of which 3252 * association each receive belongs to. 3253 * 3254 * This option takes a boolean value. A non-zero value indicates that 3255 * fragmented interleave is on. A value of zero indicates that 3256 * fragmented interleave is off. 3257 * 3258 * Note that it is important that an implementation that allows this 3259 * option to be turned on, have it off by default. Otherwise an unaware 3260 * application using the one to many model may become confused and act 3261 * incorrectly. 3262 */ 3263 static int sctp_setsockopt_fragment_interleave(struct sock *sk, 3264 char __user *optval, 3265 unsigned int optlen) 3266 { 3267 int val; 3268 3269 if (optlen != sizeof(int)) 3270 return -EINVAL; 3271 if (get_user(val, (int __user *)optval)) 3272 return -EFAULT; 3273 3274 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1; 3275 3276 return 0; 3277 } 3278 3279 /* 3280 * 8.1.21. Set or Get the SCTP Partial Delivery Point 3281 * (SCTP_PARTIAL_DELIVERY_POINT) 3282 * 3283 * This option will set or get the SCTP partial delivery point. This 3284 * point is the size of a message where the partial delivery API will be 3285 * invoked to help free up rwnd space for the peer. Setting this to a 3286 * lower value will cause partial deliveries to happen more often. The 3287 * calls argument is an integer that sets or gets the partial delivery 3288 * point. Note also that the call will fail if the user attempts to set 3289 * this value larger than the socket receive buffer size. 3290 * 3291 * Note that any single message having a length smaller than or equal to 3292 * the SCTP partial delivery point will be delivered in one single read 3293 * call as long as the user provided buffer is large enough to hold the 3294 * message. 3295 */ 3296 static int sctp_setsockopt_partial_delivery_point(struct sock *sk, 3297 char __user *optval, 3298 unsigned int optlen) 3299 { 3300 u32 val; 3301 3302 if (optlen != sizeof(u32)) 3303 return -EINVAL; 3304 if (get_user(val, (int __user *)optval)) 3305 return -EFAULT; 3306 3307 /* Note: We double the receive buffer from what the user sets 3308 * it to be, also initial rwnd is based on rcvbuf/2. 3309 */ 3310 if (val > (sk->sk_rcvbuf >> 1)) 3311 return -EINVAL; 3312 3313 sctp_sk(sk)->pd_point = val; 3314 3315 return 0; /* is this the right error code? */ 3316 } 3317 3318 /* 3319 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST) 3320 * 3321 * This option will allow a user to change the maximum burst of packets 3322 * that can be emitted by this association. Note that the default value 3323 * is 4, and some implementations may restrict this setting so that it 3324 * can only be lowered. 3325 * 3326 * NOTE: This text doesn't seem right. Do this on a socket basis with 3327 * future associations inheriting the socket value. 3328 */ 3329 static int sctp_setsockopt_maxburst(struct sock *sk, 3330 char __user *optval, 3331 unsigned int optlen) 3332 { 3333 struct sctp_assoc_value params; 3334 struct sctp_sock *sp; 3335 struct sctp_association *asoc; 3336 int val; 3337 int assoc_id = 0; 3338 3339 if (optlen == sizeof(int)) { 3340 pr_warn_ratelimited(DEPRECATED 3341 "%s (pid %d) " 3342 "Use of int in max_burst socket option deprecated.\n" 3343 "Use struct sctp_assoc_value instead\n", 3344 current->comm, task_pid_nr(current)); 3345 if (copy_from_user(&val, optval, optlen)) 3346 return -EFAULT; 3347 } else if (optlen == sizeof(struct sctp_assoc_value)) { 3348 if (copy_from_user(¶ms, optval, optlen)) 3349 return -EFAULT; 3350 val = params.assoc_value; 3351 assoc_id = params.assoc_id; 3352 } else 3353 return -EINVAL; 3354 3355 sp = sctp_sk(sk); 3356 3357 if (assoc_id != 0) { 3358 asoc = sctp_id2assoc(sk, assoc_id); 3359 if (!asoc) 3360 return -EINVAL; 3361 asoc->max_burst = val; 3362 } else 3363 sp->max_burst = val; 3364 3365 return 0; 3366 } 3367 3368 /* 3369 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK) 3370 * 3371 * This set option adds a chunk type that the user is requesting to be 3372 * received only in an authenticated way. Changes to the list of chunks 3373 * will only effect future associations on the socket. 3374 */ 3375 static int sctp_setsockopt_auth_chunk(struct sock *sk, 3376 char __user *optval, 3377 unsigned int optlen) 3378 { 3379 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3380 struct sctp_authchunk val; 3381 3382 if (!ep->auth_enable) 3383 return -EACCES; 3384 3385 if (optlen != sizeof(struct sctp_authchunk)) 3386 return -EINVAL; 3387 if (copy_from_user(&val, optval, optlen)) 3388 return -EFAULT; 3389 3390 switch (val.sauth_chunk) { 3391 case SCTP_CID_INIT: 3392 case SCTP_CID_INIT_ACK: 3393 case SCTP_CID_SHUTDOWN_COMPLETE: 3394 case SCTP_CID_AUTH: 3395 return -EINVAL; 3396 } 3397 3398 /* add this chunk id to the endpoint */ 3399 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk); 3400 } 3401 3402 /* 3403 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT) 3404 * 3405 * This option gets or sets the list of HMAC algorithms that the local 3406 * endpoint requires the peer to use. 3407 */ 3408 static int sctp_setsockopt_hmac_ident(struct sock *sk, 3409 char __user *optval, 3410 unsigned int optlen) 3411 { 3412 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3413 struct sctp_hmacalgo *hmacs; 3414 u32 idents; 3415 int err; 3416 3417 if (!ep->auth_enable) 3418 return -EACCES; 3419 3420 if (optlen < sizeof(struct sctp_hmacalgo)) 3421 return -EINVAL; 3422 3423 hmacs = memdup_user(optval, optlen); 3424 if (IS_ERR(hmacs)) 3425 return PTR_ERR(hmacs); 3426 3427 idents = hmacs->shmac_num_idents; 3428 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS || 3429 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) { 3430 err = -EINVAL; 3431 goto out; 3432 } 3433 3434 err = sctp_auth_ep_set_hmacs(ep, hmacs); 3435 out: 3436 kfree(hmacs); 3437 return err; 3438 } 3439 3440 /* 3441 * 7.1.20. Set a shared key (SCTP_AUTH_KEY) 3442 * 3443 * This option will set a shared secret key which is used to build an 3444 * association shared key. 3445 */ 3446 static int sctp_setsockopt_auth_key(struct sock *sk, 3447 char __user *optval, 3448 unsigned int optlen) 3449 { 3450 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3451 struct sctp_authkey *authkey; 3452 struct sctp_association *asoc; 3453 int ret; 3454 3455 if (!ep->auth_enable) 3456 return -EACCES; 3457 3458 if (optlen <= sizeof(struct sctp_authkey)) 3459 return -EINVAL; 3460 3461 authkey = memdup_user(optval, optlen); 3462 if (IS_ERR(authkey)) 3463 return PTR_ERR(authkey); 3464 3465 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) { 3466 ret = -EINVAL; 3467 goto out; 3468 } 3469 3470 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id); 3471 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) { 3472 ret = -EINVAL; 3473 goto out; 3474 } 3475 3476 ret = sctp_auth_set_key(ep, asoc, authkey); 3477 out: 3478 kzfree(authkey); 3479 return ret; 3480 } 3481 3482 /* 3483 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY) 3484 * 3485 * This option will get or set the active shared key to be used to build 3486 * the association shared key. 3487 */ 3488 static int sctp_setsockopt_active_key(struct sock *sk, 3489 char __user *optval, 3490 unsigned int optlen) 3491 { 3492 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3493 struct sctp_authkeyid val; 3494 struct sctp_association *asoc; 3495 3496 if (!ep->auth_enable) 3497 return -EACCES; 3498 3499 if (optlen != sizeof(struct sctp_authkeyid)) 3500 return -EINVAL; 3501 if (copy_from_user(&val, optval, optlen)) 3502 return -EFAULT; 3503 3504 asoc = sctp_id2assoc(sk, val.scact_assoc_id); 3505 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP)) 3506 return -EINVAL; 3507 3508 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber); 3509 } 3510 3511 /* 3512 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY) 3513 * 3514 * This set option will delete a shared secret key from use. 3515 */ 3516 static int sctp_setsockopt_del_key(struct sock *sk, 3517 char __user *optval, 3518 unsigned int optlen) 3519 { 3520 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3521 struct sctp_authkeyid val; 3522 struct sctp_association *asoc; 3523 3524 if (!ep->auth_enable) 3525 return -EACCES; 3526 3527 if (optlen != sizeof(struct sctp_authkeyid)) 3528 return -EINVAL; 3529 if (copy_from_user(&val, optval, optlen)) 3530 return -EFAULT; 3531 3532 asoc = sctp_id2assoc(sk, val.scact_assoc_id); 3533 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP)) 3534 return -EINVAL; 3535 3536 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber); 3537 3538 } 3539 3540 /* 3541 * 8.1.23 SCTP_AUTO_ASCONF 3542 * 3543 * This option will enable or disable the use of the automatic generation of 3544 * ASCONF chunks to add and delete addresses to an existing association. Note 3545 * that this option has two caveats namely: a) it only affects sockets that 3546 * are bound to all addresses available to the SCTP stack, and b) the system 3547 * administrator may have an overriding control that turns the ASCONF feature 3548 * off no matter what setting the socket option may have. 3549 * This option expects an integer boolean flag, where a non-zero value turns on 3550 * the option, and a zero value turns off the option. 3551 * Note. In this implementation, socket operation overrides default parameter 3552 * being set by sysctl as well as FreeBSD implementation 3553 */ 3554 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval, 3555 unsigned int optlen) 3556 { 3557 int val; 3558 struct sctp_sock *sp = sctp_sk(sk); 3559 3560 if (optlen < sizeof(int)) 3561 return -EINVAL; 3562 if (get_user(val, (int __user *)optval)) 3563 return -EFAULT; 3564 if (!sctp_is_ep_boundall(sk) && val) 3565 return -EINVAL; 3566 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf)) 3567 return 0; 3568 3569 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock); 3570 if (val == 0 && sp->do_auto_asconf) { 3571 list_del(&sp->auto_asconf_list); 3572 sp->do_auto_asconf = 0; 3573 } else if (val && !sp->do_auto_asconf) { 3574 list_add_tail(&sp->auto_asconf_list, 3575 &sock_net(sk)->sctp.auto_asconf_splist); 3576 sp->do_auto_asconf = 1; 3577 } 3578 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock); 3579 return 0; 3580 } 3581 3582 /* 3583 * SCTP_PEER_ADDR_THLDS 3584 * 3585 * This option allows us to alter the partially failed threshold for one or all 3586 * transports in an association. See Section 6.1 of: 3587 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt 3588 */ 3589 static int sctp_setsockopt_paddr_thresholds(struct sock *sk, 3590 char __user *optval, 3591 unsigned int optlen) 3592 { 3593 struct sctp_paddrthlds val; 3594 struct sctp_transport *trans; 3595 struct sctp_association *asoc; 3596 3597 if (optlen < sizeof(struct sctp_paddrthlds)) 3598 return -EINVAL; 3599 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, 3600 sizeof(struct sctp_paddrthlds))) 3601 return -EFAULT; 3602 3603 3604 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) { 3605 asoc = sctp_id2assoc(sk, val.spt_assoc_id); 3606 if (!asoc) 3607 return -ENOENT; 3608 list_for_each_entry(trans, &asoc->peer.transport_addr_list, 3609 transports) { 3610 if (val.spt_pathmaxrxt) 3611 trans->pathmaxrxt = val.spt_pathmaxrxt; 3612 trans->pf_retrans = val.spt_pathpfthld; 3613 } 3614 3615 if (val.spt_pathmaxrxt) 3616 asoc->pathmaxrxt = val.spt_pathmaxrxt; 3617 asoc->pf_retrans = val.spt_pathpfthld; 3618 } else { 3619 trans = sctp_addr_id2transport(sk, &val.spt_address, 3620 val.spt_assoc_id); 3621 if (!trans) 3622 return -ENOENT; 3623 3624 if (val.spt_pathmaxrxt) 3625 trans->pathmaxrxt = val.spt_pathmaxrxt; 3626 trans->pf_retrans = val.spt_pathpfthld; 3627 } 3628 3629 return 0; 3630 } 3631 3632 static int sctp_setsockopt_recvrcvinfo(struct sock *sk, 3633 char __user *optval, 3634 unsigned int optlen) 3635 { 3636 int val; 3637 3638 if (optlen < sizeof(int)) 3639 return -EINVAL; 3640 if (get_user(val, (int __user *) optval)) 3641 return -EFAULT; 3642 3643 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1; 3644 3645 return 0; 3646 } 3647 3648 static int sctp_setsockopt_recvnxtinfo(struct sock *sk, 3649 char __user *optval, 3650 unsigned int optlen) 3651 { 3652 int val; 3653 3654 if (optlen < sizeof(int)) 3655 return -EINVAL; 3656 if (get_user(val, (int __user *) optval)) 3657 return -EFAULT; 3658 3659 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1; 3660 3661 return 0; 3662 } 3663 3664 /* API 6.2 setsockopt(), getsockopt() 3665 * 3666 * Applications use setsockopt() and getsockopt() to set or retrieve 3667 * socket options. Socket options are used to change the default 3668 * behavior of sockets calls. They are described in Section 7. 3669 * 3670 * The syntax is: 3671 * 3672 * ret = getsockopt(int sd, int level, int optname, void __user *optval, 3673 * int __user *optlen); 3674 * ret = setsockopt(int sd, int level, int optname, const void __user *optval, 3675 * int optlen); 3676 * 3677 * sd - the socket descript. 3678 * level - set to IPPROTO_SCTP for all SCTP options. 3679 * optname - the option name. 3680 * optval - the buffer to store the value of the option. 3681 * optlen - the size of the buffer. 3682 */ 3683 static int sctp_setsockopt(struct sock *sk, int level, int optname, 3684 char __user *optval, unsigned int optlen) 3685 { 3686 int retval = 0; 3687 3688 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname); 3689 3690 /* I can hardly begin to describe how wrong this is. This is 3691 * so broken as to be worse than useless. The API draft 3692 * REALLY is NOT helpful here... I am not convinced that the 3693 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP 3694 * are at all well-founded. 3695 */ 3696 if (level != SOL_SCTP) { 3697 struct sctp_af *af = sctp_sk(sk)->pf->af; 3698 retval = af->setsockopt(sk, level, optname, optval, optlen); 3699 goto out_nounlock; 3700 } 3701 3702 lock_sock(sk); 3703 3704 switch (optname) { 3705 case SCTP_SOCKOPT_BINDX_ADD: 3706 /* 'optlen' is the size of the addresses buffer. */ 3707 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval, 3708 optlen, SCTP_BINDX_ADD_ADDR); 3709 break; 3710 3711 case SCTP_SOCKOPT_BINDX_REM: 3712 /* 'optlen' is the size of the addresses buffer. */ 3713 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval, 3714 optlen, SCTP_BINDX_REM_ADDR); 3715 break; 3716 3717 case SCTP_SOCKOPT_CONNECTX_OLD: 3718 /* 'optlen' is the size of the addresses buffer. */ 3719 retval = sctp_setsockopt_connectx_old(sk, 3720 (struct sockaddr __user *)optval, 3721 optlen); 3722 break; 3723 3724 case SCTP_SOCKOPT_CONNECTX: 3725 /* 'optlen' is the size of the addresses buffer. */ 3726 retval = sctp_setsockopt_connectx(sk, 3727 (struct sockaddr __user *)optval, 3728 optlen); 3729 break; 3730 3731 case SCTP_DISABLE_FRAGMENTS: 3732 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen); 3733 break; 3734 3735 case SCTP_EVENTS: 3736 retval = sctp_setsockopt_events(sk, optval, optlen); 3737 break; 3738 3739 case SCTP_AUTOCLOSE: 3740 retval = sctp_setsockopt_autoclose(sk, optval, optlen); 3741 break; 3742 3743 case SCTP_PEER_ADDR_PARAMS: 3744 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen); 3745 break; 3746 3747 case SCTP_DELAYED_SACK: 3748 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen); 3749 break; 3750 case SCTP_PARTIAL_DELIVERY_POINT: 3751 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen); 3752 break; 3753 3754 case SCTP_INITMSG: 3755 retval = sctp_setsockopt_initmsg(sk, optval, optlen); 3756 break; 3757 case SCTP_DEFAULT_SEND_PARAM: 3758 retval = sctp_setsockopt_default_send_param(sk, optval, 3759 optlen); 3760 break; 3761 case SCTP_DEFAULT_SNDINFO: 3762 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen); 3763 break; 3764 case SCTP_PRIMARY_ADDR: 3765 retval = sctp_setsockopt_primary_addr(sk, optval, optlen); 3766 break; 3767 case SCTP_SET_PEER_PRIMARY_ADDR: 3768 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen); 3769 break; 3770 case SCTP_NODELAY: 3771 retval = sctp_setsockopt_nodelay(sk, optval, optlen); 3772 break; 3773 case SCTP_RTOINFO: 3774 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen); 3775 break; 3776 case SCTP_ASSOCINFO: 3777 retval = sctp_setsockopt_associnfo(sk, optval, optlen); 3778 break; 3779 case SCTP_I_WANT_MAPPED_V4_ADDR: 3780 retval = sctp_setsockopt_mappedv4(sk, optval, optlen); 3781 break; 3782 case SCTP_MAXSEG: 3783 retval = sctp_setsockopt_maxseg(sk, optval, optlen); 3784 break; 3785 case SCTP_ADAPTATION_LAYER: 3786 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen); 3787 break; 3788 case SCTP_CONTEXT: 3789 retval = sctp_setsockopt_context(sk, optval, optlen); 3790 break; 3791 case SCTP_FRAGMENT_INTERLEAVE: 3792 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen); 3793 break; 3794 case SCTP_MAX_BURST: 3795 retval = sctp_setsockopt_maxburst(sk, optval, optlen); 3796 break; 3797 case SCTP_AUTH_CHUNK: 3798 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen); 3799 break; 3800 case SCTP_HMAC_IDENT: 3801 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen); 3802 break; 3803 case SCTP_AUTH_KEY: 3804 retval = sctp_setsockopt_auth_key(sk, optval, optlen); 3805 break; 3806 case SCTP_AUTH_ACTIVE_KEY: 3807 retval = sctp_setsockopt_active_key(sk, optval, optlen); 3808 break; 3809 case SCTP_AUTH_DELETE_KEY: 3810 retval = sctp_setsockopt_del_key(sk, optval, optlen); 3811 break; 3812 case SCTP_AUTO_ASCONF: 3813 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen); 3814 break; 3815 case SCTP_PEER_ADDR_THLDS: 3816 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen); 3817 break; 3818 case SCTP_RECVRCVINFO: 3819 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen); 3820 break; 3821 case SCTP_RECVNXTINFO: 3822 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen); 3823 break; 3824 default: 3825 retval = -ENOPROTOOPT; 3826 break; 3827 } 3828 3829 release_sock(sk); 3830 3831 out_nounlock: 3832 return retval; 3833 } 3834 3835 /* API 3.1.6 connect() - UDP Style Syntax 3836 * 3837 * An application may use the connect() call in the UDP model to initiate an 3838 * association without sending data. 3839 * 3840 * The syntax is: 3841 * 3842 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len); 3843 * 3844 * sd: the socket descriptor to have a new association added to. 3845 * 3846 * nam: the address structure (either struct sockaddr_in or struct 3847 * sockaddr_in6 defined in RFC2553 [7]). 3848 * 3849 * len: the size of the address. 3850 */ 3851 static int sctp_connect(struct sock *sk, struct sockaddr *addr, 3852 int addr_len) 3853 { 3854 int err = 0; 3855 struct sctp_af *af; 3856 3857 lock_sock(sk); 3858 3859 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk, 3860 addr, addr_len); 3861 3862 /* Validate addr_len before calling common connect/connectx routine. */ 3863 af = sctp_get_af_specific(addr->sa_family); 3864 if (!af || addr_len < af->sockaddr_len) { 3865 err = -EINVAL; 3866 } else { 3867 /* Pass correct addr len to common routine (so it knows there 3868 * is only one address being passed. 3869 */ 3870 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL); 3871 } 3872 3873 release_sock(sk); 3874 return err; 3875 } 3876 3877 /* FIXME: Write comments. */ 3878 static int sctp_disconnect(struct sock *sk, int flags) 3879 { 3880 return -EOPNOTSUPP; /* STUB */ 3881 } 3882 3883 /* 4.1.4 accept() - TCP Style Syntax 3884 * 3885 * Applications use accept() call to remove an established SCTP 3886 * association from the accept queue of the endpoint. A new socket 3887 * descriptor will be returned from accept() to represent the newly 3888 * formed association. 3889 */ 3890 static struct sock *sctp_accept(struct sock *sk, int flags, int *err) 3891 { 3892 struct sctp_sock *sp; 3893 struct sctp_endpoint *ep; 3894 struct sock *newsk = NULL; 3895 struct sctp_association *asoc; 3896 long timeo; 3897 int error = 0; 3898 3899 lock_sock(sk); 3900 3901 sp = sctp_sk(sk); 3902 ep = sp->ep; 3903 3904 if (!sctp_style(sk, TCP)) { 3905 error = -EOPNOTSUPP; 3906 goto out; 3907 } 3908 3909 if (!sctp_sstate(sk, LISTENING)) { 3910 error = -EINVAL; 3911 goto out; 3912 } 3913 3914 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); 3915 3916 error = sctp_wait_for_accept(sk, timeo); 3917 if (error) 3918 goto out; 3919 3920 /* We treat the list of associations on the endpoint as the accept 3921 * queue and pick the first association on the list. 3922 */ 3923 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs); 3924 3925 newsk = sp->pf->create_accept_sk(sk, asoc); 3926 if (!newsk) { 3927 error = -ENOMEM; 3928 goto out; 3929 } 3930 3931 /* Populate the fields of the newsk from the oldsk and migrate the 3932 * asoc to the newsk. 3933 */ 3934 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP); 3935 3936 out: 3937 release_sock(sk); 3938 *err = error; 3939 return newsk; 3940 } 3941 3942 /* The SCTP ioctl handler. */ 3943 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg) 3944 { 3945 int rc = -ENOTCONN; 3946 3947 lock_sock(sk); 3948 3949 /* 3950 * SEQPACKET-style sockets in LISTENING state are valid, for 3951 * SCTP, so only discard TCP-style sockets in LISTENING state. 3952 */ 3953 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) 3954 goto out; 3955 3956 switch (cmd) { 3957 case SIOCINQ: { 3958 struct sk_buff *skb; 3959 unsigned int amount = 0; 3960 3961 skb = skb_peek(&sk->sk_receive_queue); 3962 if (skb != NULL) { 3963 /* 3964 * We will only return the amount of this packet since 3965 * that is all that will be read. 3966 */ 3967 amount = skb->len; 3968 } 3969 rc = put_user(amount, (int __user *)arg); 3970 break; 3971 } 3972 default: 3973 rc = -ENOIOCTLCMD; 3974 break; 3975 } 3976 out: 3977 release_sock(sk); 3978 return rc; 3979 } 3980 3981 /* This is the function which gets called during socket creation to 3982 * initialized the SCTP-specific portion of the sock. 3983 * The sock structure should already be zero-filled memory. 3984 */ 3985 static int sctp_init_sock(struct sock *sk) 3986 { 3987 struct net *net = sock_net(sk); 3988 struct sctp_sock *sp; 3989 3990 pr_debug("%s: sk:%p\n", __func__, sk); 3991 3992 sp = sctp_sk(sk); 3993 3994 /* Initialize the SCTP per socket area. */ 3995 switch (sk->sk_type) { 3996 case SOCK_SEQPACKET: 3997 sp->type = SCTP_SOCKET_UDP; 3998 break; 3999 case SOCK_STREAM: 4000 sp->type = SCTP_SOCKET_TCP; 4001 break; 4002 default: 4003 return -ESOCKTNOSUPPORT; 4004 } 4005 4006 /* Initialize default send parameters. These parameters can be 4007 * modified with the SCTP_DEFAULT_SEND_PARAM socket option. 4008 */ 4009 sp->default_stream = 0; 4010 sp->default_ppid = 0; 4011 sp->default_flags = 0; 4012 sp->default_context = 0; 4013 sp->default_timetolive = 0; 4014 4015 sp->default_rcv_context = 0; 4016 sp->max_burst = net->sctp.max_burst; 4017 4018 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg; 4019 4020 /* Initialize default setup parameters. These parameters 4021 * can be modified with the SCTP_INITMSG socket option or 4022 * overridden by the SCTP_INIT CMSG. 4023 */ 4024 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams; 4025 sp->initmsg.sinit_max_instreams = sctp_max_instreams; 4026 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init; 4027 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max; 4028 4029 /* Initialize default RTO related parameters. These parameters can 4030 * be modified for with the SCTP_RTOINFO socket option. 4031 */ 4032 sp->rtoinfo.srto_initial = net->sctp.rto_initial; 4033 sp->rtoinfo.srto_max = net->sctp.rto_max; 4034 sp->rtoinfo.srto_min = net->sctp.rto_min; 4035 4036 /* Initialize default association related parameters. These parameters 4037 * can be modified with the SCTP_ASSOCINFO socket option. 4038 */ 4039 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association; 4040 sp->assocparams.sasoc_number_peer_destinations = 0; 4041 sp->assocparams.sasoc_peer_rwnd = 0; 4042 sp->assocparams.sasoc_local_rwnd = 0; 4043 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life; 4044 4045 /* Initialize default event subscriptions. By default, all the 4046 * options are off. 4047 */ 4048 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe)); 4049 4050 /* Default Peer Address Parameters. These defaults can 4051 * be modified via SCTP_PEER_ADDR_PARAMS 4052 */ 4053 sp->hbinterval = net->sctp.hb_interval; 4054 sp->pathmaxrxt = net->sctp.max_retrans_path; 4055 sp->pathmtu = 0; /* allow default discovery */ 4056 sp->sackdelay = net->sctp.sack_timeout; 4057 sp->sackfreq = 2; 4058 sp->param_flags = SPP_HB_ENABLE | 4059 SPP_PMTUD_ENABLE | 4060 SPP_SACKDELAY_ENABLE; 4061 4062 /* If enabled no SCTP message fragmentation will be performed. 4063 * Configure through SCTP_DISABLE_FRAGMENTS socket option. 4064 */ 4065 sp->disable_fragments = 0; 4066 4067 /* Enable Nagle algorithm by default. */ 4068 sp->nodelay = 0; 4069 4070 sp->recvrcvinfo = 0; 4071 sp->recvnxtinfo = 0; 4072 4073 /* Enable by default. */ 4074 sp->v4mapped = 1; 4075 4076 /* Auto-close idle associations after the configured 4077 * number of seconds. A value of 0 disables this 4078 * feature. Configure through the SCTP_AUTOCLOSE socket option, 4079 * for UDP-style sockets only. 4080 */ 4081 sp->autoclose = 0; 4082 4083 /* User specified fragmentation limit. */ 4084 sp->user_frag = 0; 4085 4086 sp->adaptation_ind = 0; 4087 4088 sp->pf = sctp_get_pf_specific(sk->sk_family); 4089 4090 /* Control variables for partial data delivery. */ 4091 atomic_set(&sp->pd_mode, 0); 4092 skb_queue_head_init(&sp->pd_lobby); 4093 sp->frag_interleave = 0; 4094 4095 /* Create a per socket endpoint structure. Even if we 4096 * change the data structure relationships, this may still 4097 * be useful for storing pre-connect address information. 4098 */ 4099 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL); 4100 if (!sp->ep) 4101 return -ENOMEM; 4102 4103 sp->hmac = NULL; 4104 4105 sk->sk_destruct = sctp_destruct_sock; 4106 4107 SCTP_DBG_OBJCNT_INC(sock); 4108 4109 local_bh_disable(); 4110 percpu_counter_inc(&sctp_sockets_allocated); 4111 sock_prot_inuse_add(net, sk->sk_prot, 1); 4112 4113 /* Nothing can fail after this block, otherwise 4114 * sctp_destroy_sock() will be called without addr_wq_lock held 4115 */ 4116 if (net->sctp.default_auto_asconf) { 4117 spin_lock(&sock_net(sk)->sctp.addr_wq_lock); 4118 list_add_tail(&sp->auto_asconf_list, 4119 &net->sctp.auto_asconf_splist); 4120 sp->do_auto_asconf = 1; 4121 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock); 4122 } else { 4123 sp->do_auto_asconf = 0; 4124 } 4125 4126 local_bh_enable(); 4127 4128 return 0; 4129 } 4130 4131 /* Cleanup any SCTP per socket resources. Must be called with 4132 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true 4133 */ 4134 static void sctp_destroy_sock(struct sock *sk) 4135 { 4136 struct sctp_sock *sp; 4137 4138 pr_debug("%s: sk:%p\n", __func__, sk); 4139 4140 /* Release our hold on the endpoint. */ 4141 sp = sctp_sk(sk); 4142 /* This could happen during socket init, thus we bail out 4143 * early, since the rest of the below is not setup either. 4144 */ 4145 if (sp->ep == NULL) 4146 return; 4147 4148 if (sp->do_auto_asconf) { 4149 sp->do_auto_asconf = 0; 4150 list_del(&sp->auto_asconf_list); 4151 } 4152 sctp_endpoint_free(sp->ep); 4153 local_bh_disable(); 4154 percpu_counter_dec(&sctp_sockets_allocated); 4155 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 4156 local_bh_enable(); 4157 } 4158 4159 /* Triggered when there are no references on the socket anymore */ 4160 static void sctp_destruct_sock(struct sock *sk) 4161 { 4162 struct sctp_sock *sp = sctp_sk(sk); 4163 4164 /* Free up the HMAC transform. */ 4165 crypto_free_hash(sp->hmac); 4166 4167 inet_sock_destruct(sk); 4168 } 4169 4170 /* API 4.1.7 shutdown() - TCP Style Syntax 4171 * int shutdown(int socket, int how); 4172 * 4173 * sd - the socket descriptor of the association to be closed. 4174 * how - Specifies the type of shutdown. The values are 4175 * as follows: 4176 * SHUT_RD 4177 * Disables further receive operations. No SCTP 4178 * protocol action is taken. 4179 * SHUT_WR 4180 * Disables further send operations, and initiates 4181 * the SCTP shutdown sequence. 4182 * SHUT_RDWR 4183 * Disables further send and receive operations 4184 * and initiates the SCTP shutdown sequence. 4185 */ 4186 static void sctp_shutdown(struct sock *sk, int how) 4187 { 4188 struct net *net = sock_net(sk); 4189 struct sctp_endpoint *ep; 4190 struct sctp_association *asoc; 4191 4192 if (!sctp_style(sk, TCP)) 4193 return; 4194 4195 if (how & SEND_SHUTDOWN) { 4196 ep = sctp_sk(sk)->ep; 4197 if (!list_empty(&ep->asocs)) { 4198 asoc = list_entry(ep->asocs.next, 4199 struct sctp_association, asocs); 4200 sctp_primitive_SHUTDOWN(net, asoc, NULL); 4201 } 4202 } 4203 } 4204 4205 /* 7.2.1 Association Status (SCTP_STATUS) 4206 4207 * Applications can retrieve current status information about an 4208 * association, including association state, peer receiver window size, 4209 * number of unacked data chunks, and number of data chunks pending 4210 * receipt. This information is read-only. 4211 */ 4212 static int sctp_getsockopt_sctp_status(struct sock *sk, int len, 4213 char __user *optval, 4214 int __user *optlen) 4215 { 4216 struct sctp_status status; 4217 struct sctp_association *asoc = NULL; 4218 struct sctp_transport *transport; 4219 sctp_assoc_t associd; 4220 int retval = 0; 4221 4222 if (len < sizeof(status)) { 4223 retval = -EINVAL; 4224 goto out; 4225 } 4226 4227 len = sizeof(status); 4228 if (copy_from_user(&status, optval, len)) { 4229 retval = -EFAULT; 4230 goto out; 4231 } 4232 4233 associd = status.sstat_assoc_id; 4234 asoc = sctp_id2assoc(sk, associd); 4235 if (!asoc) { 4236 retval = -EINVAL; 4237 goto out; 4238 } 4239 4240 transport = asoc->peer.primary_path; 4241 4242 status.sstat_assoc_id = sctp_assoc2id(asoc); 4243 status.sstat_state = sctp_assoc_to_state(asoc); 4244 status.sstat_rwnd = asoc->peer.rwnd; 4245 status.sstat_unackdata = asoc->unack_data; 4246 4247 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map); 4248 status.sstat_instrms = asoc->c.sinit_max_instreams; 4249 status.sstat_outstrms = asoc->c.sinit_num_ostreams; 4250 status.sstat_fragmentation_point = asoc->frag_point; 4251 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc); 4252 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr, 4253 transport->af_specific->sockaddr_len); 4254 /* Map ipv4 address into v4-mapped-on-v6 address. */ 4255 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk), 4256 (union sctp_addr *)&status.sstat_primary.spinfo_address); 4257 status.sstat_primary.spinfo_state = transport->state; 4258 status.sstat_primary.spinfo_cwnd = transport->cwnd; 4259 status.sstat_primary.spinfo_srtt = transport->srtt; 4260 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto); 4261 status.sstat_primary.spinfo_mtu = transport->pathmtu; 4262 4263 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN) 4264 status.sstat_primary.spinfo_state = SCTP_ACTIVE; 4265 4266 if (put_user(len, optlen)) { 4267 retval = -EFAULT; 4268 goto out; 4269 } 4270 4271 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n", 4272 __func__, len, status.sstat_state, status.sstat_rwnd, 4273 status.sstat_assoc_id); 4274 4275 if (copy_to_user(optval, &status, len)) { 4276 retval = -EFAULT; 4277 goto out; 4278 } 4279 4280 out: 4281 return retval; 4282 } 4283 4284 4285 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO) 4286 * 4287 * Applications can retrieve information about a specific peer address 4288 * of an association, including its reachability state, congestion 4289 * window, and retransmission timer values. This information is 4290 * read-only. 4291 */ 4292 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len, 4293 char __user *optval, 4294 int __user *optlen) 4295 { 4296 struct sctp_paddrinfo pinfo; 4297 struct sctp_transport *transport; 4298 int retval = 0; 4299 4300 if (len < sizeof(pinfo)) { 4301 retval = -EINVAL; 4302 goto out; 4303 } 4304 4305 len = sizeof(pinfo); 4306 if (copy_from_user(&pinfo, optval, len)) { 4307 retval = -EFAULT; 4308 goto out; 4309 } 4310 4311 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address, 4312 pinfo.spinfo_assoc_id); 4313 if (!transport) 4314 return -EINVAL; 4315 4316 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc); 4317 pinfo.spinfo_state = transport->state; 4318 pinfo.spinfo_cwnd = transport->cwnd; 4319 pinfo.spinfo_srtt = transport->srtt; 4320 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto); 4321 pinfo.spinfo_mtu = transport->pathmtu; 4322 4323 if (pinfo.spinfo_state == SCTP_UNKNOWN) 4324 pinfo.spinfo_state = SCTP_ACTIVE; 4325 4326 if (put_user(len, optlen)) { 4327 retval = -EFAULT; 4328 goto out; 4329 } 4330 4331 if (copy_to_user(optval, &pinfo, len)) { 4332 retval = -EFAULT; 4333 goto out; 4334 } 4335 4336 out: 4337 return retval; 4338 } 4339 4340 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS) 4341 * 4342 * This option is a on/off flag. If enabled no SCTP message 4343 * fragmentation will be performed. Instead if a message being sent 4344 * exceeds the current PMTU size, the message will NOT be sent and 4345 * instead a error will be indicated to the user. 4346 */ 4347 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len, 4348 char __user *optval, int __user *optlen) 4349 { 4350 int val; 4351 4352 if (len < sizeof(int)) 4353 return -EINVAL; 4354 4355 len = sizeof(int); 4356 val = (sctp_sk(sk)->disable_fragments == 1); 4357 if (put_user(len, optlen)) 4358 return -EFAULT; 4359 if (copy_to_user(optval, &val, len)) 4360 return -EFAULT; 4361 return 0; 4362 } 4363 4364 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS) 4365 * 4366 * This socket option is used to specify various notifications and 4367 * ancillary data the user wishes to receive. 4368 */ 4369 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval, 4370 int __user *optlen) 4371 { 4372 if (len <= 0) 4373 return -EINVAL; 4374 if (len > sizeof(struct sctp_event_subscribe)) 4375 len = sizeof(struct sctp_event_subscribe); 4376 if (put_user(len, optlen)) 4377 return -EFAULT; 4378 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len)) 4379 return -EFAULT; 4380 return 0; 4381 } 4382 4383 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE) 4384 * 4385 * This socket option is applicable to the UDP-style socket only. When 4386 * set it will cause associations that are idle for more than the 4387 * specified number of seconds to automatically close. An association 4388 * being idle is defined an association that has NOT sent or received 4389 * user data. The special value of '0' indicates that no automatic 4390 * close of any associations should be performed. The option expects an 4391 * integer defining the number of seconds of idle time before an 4392 * association is closed. 4393 */ 4394 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen) 4395 { 4396 /* Applicable to UDP-style socket only */ 4397 if (sctp_style(sk, TCP)) 4398 return -EOPNOTSUPP; 4399 if (len < sizeof(int)) 4400 return -EINVAL; 4401 len = sizeof(int); 4402 if (put_user(len, optlen)) 4403 return -EFAULT; 4404 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int))) 4405 return -EFAULT; 4406 return 0; 4407 } 4408 4409 /* Helper routine to branch off an association to a new socket. */ 4410 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp) 4411 { 4412 struct sctp_association *asoc = sctp_id2assoc(sk, id); 4413 struct sctp_sock *sp = sctp_sk(sk); 4414 struct socket *sock; 4415 int err = 0; 4416 4417 if (!asoc) 4418 return -EINVAL; 4419 4420 /* An association cannot be branched off from an already peeled-off 4421 * socket, nor is this supported for tcp style sockets. 4422 */ 4423 if (!sctp_style(sk, UDP)) 4424 return -EINVAL; 4425 4426 /* Create a new socket. */ 4427 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock); 4428 if (err < 0) 4429 return err; 4430 4431 sctp_copy_sock(sock->sk, sk, asoc); 4432 4433 /* Make peeled-off sockets more like 1-1 accepted sockets. 4434 * Set the daddr and initialize id to something more random 4435 */ 4436 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk); 4437 4438 /* Populate the fields of the newsk from the oldsk and migrate the 4439 * asoc to the newsk. 4440 */ 4441 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH); 4442 4443 *sockp = sock; 4444 4445 return err; 4446 } 4447 EXPORT_SYMBOL(sctp_do_peeloff); 4448 4449 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen) 4450 { 4451 sctp_peeloff_arg_t peeloff; 4452 struct socket *newsock; 4453 struct file *newfile; 4454 int retval = 0; 4455 4456 if (len < sizeof(sctp_peeloff_arg_t)) 4457 return -EINVAL; 4458 len = sizeof(sctp_peeloff_arg_t); 4459 if (copy_from_user(&peeloff, optval, len)) 4460 return -EFAULT; 4461 4462 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock); 4463 if (retval < 0) 4464 goto out; 4465 4466 /* Map the socket to an unused fd that can be returned to the user. */ 4467 retval = get_unused_fd_flags(0); 4468 if (retval < 0) { 4469 sock_release(newsock); 4470 goto out; 4471 } 4472 4473 newfile = sock_alloc_file(newsock, 0, NULL); 4474 if (IS_ERR(newfile)) { 4475 put_unused_fd(retval); 4476 sock_release(newsock); 4477 return PTR_ERR(newfile); 4478 } 4479 4480 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk, 4481 retval); 4482 4483 /* Return the fd mapped to the new socket. */ 4484 if (put_user(len, optlen)) { 4485 fput(newfile); 4486 put_unused_fd(retval); 4487 return -EFAULT; 4488 } 4489 peeloff.sd = retval; 4490 if (copy_to_user(optval, &peeloff, len)) { 4491 fput(newfile); 4492 put_unused_fd(retval); 4493 return -EFAULT; 4494 } 4495 fd_install(retval, newfile); 4496 out: 4497 return retval; 4498 } 4499 4500 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS) 4501 * 4502 * Applications can enable or disable heartbeats for any peer address of 4503 * an association, modify an address's heartbeat interval, force a 4504 * heartbeat to be sent immediately, and adjust the address's maximum 4505 * number of retransmissions sent before an address is considered 4506 * unreachable. The following structure is used to access and modify an 4507 * address's parameters: 4508 * 4509 * struct sctp_paddrparams { 4510 * sctp_assoc_t spp_assoc_id; 4511 * struct sockaddr_storage spp_address; 4512 * uint32_t spp_hbinterval; 4513 * uint16_t spp_pathmaxrxt; 4514 * uint32_t spp_pathmtu; 4515 * uint32_t spp_sackdelay; 4516 * uint32_t spp_flags; 4517 * }; 4518 * 4519 * spp_assoc_id - (one-to-many style socket) This is filled in the 4520 * application, and identifies the association for 4521 * this query. 4522 * spp_address - This specifies which address is of interest. 4523 * spp_hbinterval - This contains the value of the heartbeat interval, 4524 * in milliseconds. If a value of zero 4525 * is present in this field then no changes are to 4526 * be made to this parameter. 4527 * spp_pathmaxrxt - This contains the maximum number of 4528 * retransmissions before this address shall be 4529 * considered unreachable. If a value of zero 4530 * is present in this field then no changes are to 4531 * be made to this parameter. 4532 * spp_pathmtu - When Path MTU discovery is disabled the value 4533 * specified here will be the "fixed" path mtu. 4534 * Note that if the spp_address field is empty 4535 * then all associations on this address will 4536 * have this fixed path mtu set upon them. 4537 * 4538 * spp_sackdelay - When delayed sack is enabled, this value specifies 4539 * the number of milliseconds that sacks will be delayed 4540 * for. This value will apply to all addresses of an 4541 * association if the spp_address field is empty. Note 4542 * also, that if delayed sack is enabled and this 4543 * value is set to 0, no change is made to the last 4544 * recorded delayed sack timer value. 4545 * 4546 * spp_flags - These flags are used to control various features 4547 * on an association. The flag field may contain 4548 * zero or more of the following options. 4549 * 4550 * SPP_HB_ENABLE - Enable heartbeats on the 4551 * specified address. Note that if the address 4552 * field is empty all addresses for the association 4553 * have heartbeats enabled upon them. 4554 * 4555 * SPP_HB_DISABLE - Disable heartbeats on the 4556 * speicifed address. Note that if the address 4557 * field is empty all addresses for the association 4558 * will have their heartbeats disabled. Note also 4559 * that SPP_HB_ENABLE and SPP_HB_DISABLE are 4560 * mutually exclusive, only one of these two should 4561 * be specified. Enabling both fields will have 4562 * undetermined results. 4563 * 4564 * SPP_HB_DEMAND - Request a user initiated heartbeat 4565 * to be made immediately. 4566 * 4567 * SPP_PMTUD_ENABLE - This field will enable PMTU 4568 * discovery upon the specified address. Note that 4569 * if the address feild is empty then all addresses 4570 * on the association are effected. 4571 * 4572 * SPP_PMTUD_DISABLE - This field will disable PMTU 4573 * discovery upon the specified address. Note that 4574 * if the address feild is empty then all addresses 4575 * on the association are effected. Not also that 4576 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually 4577 * exclusive. Enabling both will have undetermined 4578 * results. 4579 * 4580 * SPP_SACKDELAY_ENABLE - Setting this flag turns 4581 * on delayed sack. The time specified in spp_sackdelay 4582 * is used to specify the sack delay for this address. Note 4583 * that if spp_address is empty then all addresses will 4584 * enable delayed sack and take on the sack delay 4585 * value specified in spp_sackdelay. 4586 * SPP_SACKDELAY_DISABLE - Setting this flag turns 4587 * off delayed sack. If the spp_address field is blank then 4588 * delayed sack is disabled for the entire association. Note 4589 * also that this field is mutually exclusive to 4590 * SPP_SACKDELAY_ENABLE, setting both will have undefined 4591 * results. 4592 */ 4593 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len, 4594 char __user *optval, int __user *optlen) 4595 { 4596 struct sctp_paddrparams params; 4597 struct sctp_transport *trans = NULL; 4598 struct sctp_association *asoc = NULL; 4599 struct sctp_sock *sp = sctp_sk(sk); 4600 4601 if (len < sizeof(struct sctp_paddrparams)) 4602 return -EINVAL; 4603 len = sizeof(struct sctp_paddrparams); 4604 if (copy_from_user(¶ms, optval, len)) 4605 return -EFAULT; 4606 4607 /* If an address other than INADDR_ANY is specified, and 4608 * no transport is found, then the request is invalid. 4609 */ 4610 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) { 4611 trans = sctp_addr_id2transport(sk, ¶ms.spp_address, 4612 params.spp_assoc_id); 4613 if (!trans) { 4614 pr_debug("%s: failed no transport\n", __func__); 4615 return -EINVAL; 4616 } 4617 } 4618 4619 /* Get association, if assoc_id != 0 and the socket is a one 4620 * to many style socket, and an association was not found, then 4621 * the id was invalid. 4622 */ 4623 asoc = sctp_id2assoc(sk, params.spp_assoc_id); 4624 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) { 4625 pr_debug("%s: failed no association\n", __func__); 4626 return -EINVAL; 4627 } 4628 4629 if (trans) { 4630 /* Fetch transport values. */ 4631 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval); 4632 params.spp_pathmtu = trans->pathmtu; 4633 params.spp_pathmaxrxt = trans->pathmaxrxt; 4634 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay); 4635 4636 /*draft-11 doesn't say what to return in spp_flags*/ 4637 params.spp_flags = trans->param_flags; 4638 } else if (asoc) { 4639 /* Fetch association values. */ 4640 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval); 4641 params.spp_pathmtu = asoc->pathmtu; 4642 params.spp_pathmaxrxt = asoc->pathmaxrxt; 4643 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay); 4644 4645 /*draft-11 doesn't say what to return in spp_flags*/ 4646 params.spp_flags = asoc->param_flags; 4647 } else { 4648 /* Fetch socket values. */ 4649 params.spp_hbinterval = sp->hbinterval; 4650 params.spp_pathmtu = sp->pathmtu; 4651 params.spp_sackdelay = sp->sackdelay; 4652 params.spp_pathmaxrxt = sp->pathmaxrxt; 4653 4654 /*draft-11 doesn't say what to return in spp_flags*/ 4655 params.spp_flags = sp->param_flags; 4656 } 4657 4658 if (copy_to_user(optval, ¶ms, len)) 4659 return -EFAULT; 4660 4661 if (put_user(len, optlen)) 4662 return -EFAULT; 4663 4664 return 0; 4665 } 4666 4667 /* 4668 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK) 4669 * 4670 * This option will effect the way delayed acks are performed. This 4671 * option allows you to get or set the delayed ack time, in 4672 * milliseconds. It also allows changing the delayed ack frequency. 4673 * Changing the frequency to 1 disables the delayed sack algorithm. If 4674 * the assoc_id is 0, then this sets or gets the endpoints default 4675 * values. If the assoc_id field is non-zero, then the set or get 4676 * effects the specified association for the one to many model (the 4677 * assoc_id field is ignored by the one to one model). Note that if 4678 * sack_delay or sack_freq are 0 when setting this option, then the 4679 * current values will remain unchanged. 4680 * 4681 * struct sctp_sack_info { 4682 * sctp_assoc_t sack_assoc_id; 4683 * uint32_t sack_delay; 4684 * uint32_t sack_freq; 4685 * }; 4686 * 4687 * sack_assoc_id - This parameter, indicates which association the user 4688 * is performing an action upon. Note that if this field's value is 4689 * zero then the endpoints default value is changed (effecting future 4690 * associations only). 4691 * 4692 * sack_delay - This parameter contains the number of milliseconds that 4693 * the user is requesting the delayed ACK timer be set to. Note that 4694 * this value is defined in the standard to be between 200 and 500 4695 * milliseconds. 4696 * 4697 * sack_freq - This parameter contains the number of packets that must 4698 * be received before a sack is sent without waiting for the delay 4699 * timer to expire. The default value for this is 2, setting this 4700 * value to 1 will disable the delayed sack algorithm. 4701 */ 4702 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len, 4703 char __user *optval, 4704 int __user *optlen) 4705 { 4706 struct sctp_sack_info params; 4707 struct sctp_association *asoc = NULL; 4708 struct sctp_sock *sp = sctp_sk(sk); 4709 4710 if (len >= sizeof(struct sctp_sack_info)) { 4711 len = sizeof(struct sctp_sack_info); 4712 4713 if (copy_from_user(¶ms, optval, len)) 4714 return -EFAULT; 4715 } else if (len == sizeof(struct sctp_assoc_value)) { 4716 pr_warn_ratelimited(DEPRECATED 4717 "%s (pid %d) " 4718 "Use of struct sctp_assoc_value in delayed_ack socket option.\n" 4719 "Use struct sctp_sack_info instead\n", 4720 current->comm, task_pid_nr(current)); 4721 if (copy_from_user(¶ms, optval, len)) 4722 return -EFAULT; 4723 } else 4724 return -EINVAL; 4725 4726 /* Get association, if sack_assoc_id != 0 and the socket is a one 4727 * to many style socket, and an association was not found, then 4728 * the id was invalid. 4729 */ 4730 asoc = sctp_id2assoc(sk, params.sack_assoc_id); 4731 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP)) 4732 return -EINVAL; 4733 4734 if (asoc) { 4735 /* Fetch association values. */ 4736 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) { 4737 params.sack_delay = jiffies_to_msecs( 4738 asoc->sackdelay); 4739 params.sack_freq = asoc->sackfreq; 4740 4741 } else { 4742 params.sack_delay = 0; 4743 params.sack_freq = 1; 4744 } 4745 } else { 4746 /* Fetch socket values. */ 4747 if (sp->param_flags & SPP_SACKDELAY_ENABLE) { 4748 params.sack_delay = sp->sackdelay; 4749 params.sack_freq = sp->sackfreq; 4750 } else { 4751 params.sack_delay = 0; 4752 params.sack_freq = 1; 4753 } 4754 } 4755 4756 if (copy_to_user(optval, ¶ms, len)) 4757 return -EFAULT; 4758 4759 if (put_user(len, optlen)) 4760 return -EFAULT; 4761 4762 return 0; 4763 } 4764 4765 /* 7.1.3 Initialization Parameters (SCTP_INITMSG) 4766 * 4767 * Applications can specify protocol parameters for the default association 4768 * initialization. The option name argument to setsockopt() and getsockopt() 4769 * is SCTP_INITMSG. 4770 * 4771 * Setting initialization parameters is effective only on an unconnected 4772 * socket (for UDP-style sockets only future associations are effected 4773 * by the change). With TCP-style sockets, this option is inherited by 4774 * sockets derived from a listener socket. 4775 */ 4776 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen) 4777 { 4778 if (len < sizeof(struct sctp_initmsg)) 4779 return -EINVAL; 4780 len = sizeof(struct sctp_initmsg); 4781 if (put_user(len, optlen)) 4782 return -EFAULT; 4783 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len)) 4784 return -EFAULT; 4785 return 0; 4786 } 4787 4788 4789 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len, 4790 char __user *optval, int __user *optlen) 4791 { 4792 struct sctp_association *asoc; 4793 int cnt = 0; 4794 struct sctp_getaddrs getaddrs; 4795 struct sctp_transport *from; 4796 void __user *to; 4797 union sctp_addr temp; 4798 struct sctp_sock *sp = sctp_sk(sk); 4799 int addrlen; 4800 size_t space_left; 4801 int bytes_copied; 4802 4803 if (len < sizeof(struct sctp_getaddrs)) 4804 return -EINVAL; 4805 4806 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs))) 4807 return -EFAULT; 4808 4809 /* For UDP-style sockets, id specifies the association to query. */ 4810 asoc = sctp_id2assoc(sk, getaddrs.assoc_id); 4811 if (!asoc) 4812 return -EINVAL; 4813 4814 to = optval + offsetof(struct sctp_getaddrs, addrs); 4815 space_left = len - offsetof(struct sctp_getaddrs, addrs); 4816 4817 list_for_each_entry(from, &asoc->peer.transport_addr_list, 4818 transports) { 4819 memcpy(&temp, &from->ipaddr, sizeof(temp)); 4820 addrlen = sctp_get_pf_specific(sk->sk_family) 4821 ->addr_to_user(sp, &temp); 4822 if (space_left < addrlen) 4823 return -ENOMEM; 4824 if (copy_to_user(to, &temp, addrlen)) 4825 return -EFAULT; 4826 to += addrlen; 4827 cnt++; 4828 space_left -= addrlen; 4829 } 4830 4831 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) 4832 return -EFAULT; 4833 bytes_copied = ((char __user *)to) - optval; 4834 if (put_user(bytes_copied, optlen)) 4835 return -EFAULT; 4836 4837 return 0; 4838 } 4839 4840 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to, 4841 size_t space_left, int *bytes_copied) 4842 { 4843 struct sctp_sockaddr_entry *addr; 4844 union sctp_addr temp; 4845 int cnt = 0; 4846 int addrlen; 4847 struct net *net = sock_net(sk); 4848 4849 rcu_read_lock(); 4850 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 4851 if (!addr->valid) 4852 continue; 4853 4854 if ((PF_INET == sk->sk_family) && 4855 (AF_INET6 == addr->a.sa.sa_family)) 4856 continue; 4857 if ((PF_INET6 == sk->sk_family) && 4858 inet_v6_ipv6only(sk) && 4859 (AF_INET == addr->a.sa.sa_family)) 4860 continue; 4861 memcpy(&temp, &addr->a, sizeof(temp)); 4862 if (!temp.v4.sin_port) 4863 temp.v4.sin_port = htons(port); 4864 4865 addrlen = sctp_get_pf_specific(sk->sk_family) 4866 ->addr_to_user(sctp_sk(sk), &temp); 4867 4868 if (space_left < addrlen) { 4869 cnt = -ENOMEM; 4870 break; 4871 } 4872 memcpy(to, &temp, addrlen); 4873 4874 to += addrlen; 4875 cnt++; 4876 space_left -= addrlen; 4877 *bytes_copied += addrlen; 4878 } 4879 rcu_read_unlock(); 4880 4881 return cnt; 4882 } 4883 4884 4885 static int sctp_getsockopt_local_addrs(struct sock *sk, int len, 4886 char __user *optval, int __user *optlen) 4887 { 4888 struct sctp_bind_addr *bp; 4889 struct sctp_association *asoc; 4890 int cnt = 0; 4891 struct sctp_getaddrs getaddrs; 4892 struct sctp_sockaddr_entry *addr; 4893 void __user *to; 4894 union sctp_addr temp; 4895 struct sctp_sock *sp = sctp_sk(sk); 4896 int addrlen; 4897 int err = 0; 4898 size_t space_left; 4899 int bytes_copied = 0; 4900 void *addrs; 4901 void *buf; 4902 4903 if (len < sizeof(struct sctp_getaddrs)) 4904 return -EINVAL; 4905 4906 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs))) 4907 return -EFAULT; 4908 4909 /* 4910 * For UDP-style sockets, id specifies the association to query. 4911 * If the id field is set to the value '0' then the locally bound 4912 * addresses are returned without regard to any particular 4913 * association. 4914 */ 4915 if (0 == getaddrs.assoc_id) { 4916 bp = &sctp_sk(sk)->ep->base.bind_addr; 4917 } else { 4918 asoc = sctp_id2assoc(sk, getaddrs.assoc_id); 4919 if (!asoc) 4920 return -EINVAL; 4921 bp = &asoc->base.bind_addr; 4922 } 4923 4924 to = optval + offsetof(struct sctp_getaddrs, addrs); 4925 space_left = len - offsetof(struct sctp_getaddrs, addrs); 4926 4927 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN); 4928 if (!addrs) 4929 return -ENOMEM; 4930 4931 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid 4932 * addresses from the global local address list. 4933 */ 4934 if (sctp_list_single_entry(&bp->address_list)) { 4935 addr = list_entry(bp->address_list.next, 4936 struct sctp_sockaddr_entry, list); 4937 if (sctp_is_any(sk, &addr->a)) { 4938 cnt = sctp_copy_laddrs(sk, bp->port, addrs, 4939 space_left, &bytes_copied); 4940 if (cnt < 0) { 4941 err = cnt; 4942 goto out; 4943 } 4944 goto copy_getaddrs; 4945 } 4946 } 4947 4948 buf = addrs; 4949 /* Protection on the bound address list is not needed since 4950 * in the socket option context we hold a socket lock and 4951 * thus the bound address list can't change. 4952 */ 4953 list_for_each_entry(addr, &bp->address_list, list) { 4954 memcpy(&temp, &addr->a, sizeof(temp)); 4955 addrlen = sctp_get_pf_specific(sk->sk_family) 4956 ->addr_to_user(sp, &temp); 4957 if (space_left < addrlen) { 4958 err = -ENOMEM; /*fixme: right error?*/ 4959 goto out; 4960 } 4961 memcpy(buf, &temp, addrlen); 4962 buf += addrlen; 4963 bytes_copied += addrlen; 4964 cnt++; 4965 space_left -= addrlen; 4966 } 4967 4968 copy_getaddrs: 4969 if (copy_to_user(to, addrs, bytes_copied)) { 4970 err = -EFAULT; 4971 goto out; 4972 } 4973 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) { 4974 err = -EFAULT; 4975 goto out; 4976 } 4977 if (put_user(bytes_copied, optlen)) 4978 err = -EFAULT; 4979 out: 4980 kfree(addrs); 4981 return err; 4982 } 4983 4984 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR) 4985 * 4986 * Requests that the local SCTP stack use the enclosed peer address as 4987 * the association primary. The enclosed address must be one of the 4988 * association peer's addresses. 4989 */ 4990 static int sctp_getsockopt_primary_addr(struct sock *sk, int len, 4991 char __user *optval, int __user *optlen) 4992 { 4993 struct sctp_prim prim; 4994 struct sctp_association *asoc; 4995 struct sctp_sock *sp = sctp_sk(sk); 4996 4997 if (len < sizeof(struct sctp_prim)) 4998 return -EINVAL; 4999 5000 len = sizeof(struct sctp_prim); 5001 5002 if (copy_from_user(&prim, optval, len)) 5003 return -EFAULT; 5004 5005 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id); 5006 if (!asoc) 5007 return -EINVAL; 5008 5009 if (!asoc->peer.primary_path) 5010 return -ENOTCONN; 5011 5012 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr, 5013 asoc->peer.primary_path->af_specific->sockaddr_len); 5014 5015 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp, 5016 (union sctp_addr *)&prim.ssp_addr); 5017 5018 if (put_user(len, optlen)) 5019 return -EFAULT; 5020 if (copy_to_user(optval, &prim, len)) 5021 return -EFAULT; 5022 5023 return 0; 5024 } 5025 5026 /* 5027 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER) 5028 * 5029 * Requests that the local endpoint set the specified Adaptation Layer 5030 * Indication parameter for all future INIT and INIT-ACK exchanges. 5031 */ 5032 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len, 5033 char __user *optval, int __user *optlen) 5034 { 5035 struct sctp_setadaptation adaptation; 5036 5037 if (len < sizeof(struct sctp_setadaptation)) 5038 return -EINVAL; 5039 5040 len = sizeof(struct sctp_setadaptation); 5041 5042 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind; 5043 5044 if (put_user(len, optlen)) 5045 return -EFAULT; 5046 if (copy_to_user(optval, &adaptation, len)) 5047 return -EFAULT; 5048 5049 return 0; 5050 } 5051 5052 /* 5053 * 5054 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM) 5055 * 5056 * Applications that wish to use the sendto() system call may wish to 5057 * specify a default set of parameters that would normally be supplied 5058 * through the inclusion of ancillary data. This socket option allows 5059 * such an application to set the default sctp_sndrcvinfo structure. 5060 5061 5062 * The application that wishes to use this socket option simply passes 5063 * in to this call the sctp_sndrcvinfo structure defined in Section 5064 * 5.2.2) The input parameters accepted by this call include 5065 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context, 5066 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in 5067 * to this call if the caller is using the UDP model. 5068 * 5069 * For getsockopt, it get the default sctp_sndrcvinfo structure. 5070 */ 5071 static int sctp_getsockopt_default_send_param(struct sock *sk, 5072 int len, char __user *optval, 5073 int __user *optlen) 5074 { 5075 struct sctp_sock *sp = sctp_sk(sk); 5076 struct sctp_association *asoc; 5077 struct sctp_sndrcvinfo info; 5078 5079 if (len < sizeof(info)) 5080 return -EINVAL; 5081 5082 len = sizeof(info); 5083 5084 if (copy_from_user(&info, optval, len)) 5085 return -EFAULT; 5086 5087 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id); 5088 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP)) 5089 return -EINVAL; 5090 if (asoc) { 5091 info.sinfo_stream = asoc->default_stream; 5092 info.sinfo_flags = asoc->default_flags; 5093 info.sinfo_ppid = asoc->default_ppid; 5094 info.sinfo_context = asoc->default_context; 5095 info.sinfo_timetolive = asoc->default_timetolive; 5096 } else { 5097 info.sinfo_stream = sp->default_stream; 5098 info.sinfo_flags = sp->default_flags; 5099 info.sinfo_ppid = sp->default_ppid; 5100 info.sinfo_context = sp->default_context; 5101 info.sinfo_timetolive = sp->default_timetolive; 5102 } 5103 5104 if (put_user(len, optlen)) 5105 return -EFAULT; 5106 if (copy_to_user(optval, &info, len)) 5107 return -EFAULT; 5108 5109 return 0; 5110 } 5111 5112 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters 5113 * (SCTP_DEFAULT_SNDINFO) 5114 */ 5115 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len, 5116 char __user *optval, 5117 int __user *optlen) 5118 { 5119 struct sctp_sock *sp = sctp_sk(sk); 5120 struct sctp_association *asoc; 5121 struct sctp_sndinfo info; 5122 5123 if (len < sizeof(info)) 5124 return -EINVAL; 5125 5126 len = sizeof(info); 5127 5128 if (copy_from_user(&info, optval, len)) 5129 return -EFAULT; 5130 5131 asoc = sctp_id2assoc(sk, info.snd_assoc_id); 5132 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP)) 5133 return -EINVAL; 5134 if (asoc) { 5135 info.snd_sid = asoc->default_stream; 5136 info.snd_flags = asoc->default_flags; 5137 info.snd_ppid = asoc->default_ppid; 5138 info.snd_context = asoc->default_context; 5139 } else { 5140 info.snd_sid = sp->default_stream; 5141 info.snd_flags = sp->default_flags; 5142 info.snd_ppid = sp->default_ppid; 5143 info.snd_context = sp->default_context; 5144 } 5145 5146 if (put_user(len, optlen)) 5147 return -EFAULT; 5148 if (copy_to_user(optval, &info, len)) 5149 return -EFAULT; 5150 5151 return 0; 5152 } 5153 5154 /* 5155 * 5156 * 7.1.5 SCTP_NODELAY 5157 * 5158 * Turn on/off any Nagle-like algorithm. This means that packets are 5159 * generally sent as soon as possible and no unnecessary delays are 5160 * introduced, at the cost of more packets in the network. Expects an 5161 * integer boolean flag. 5162 */ 5163 5164 static int sctp_getsockopt_nodelay(struct sock *sk, int len, 5165 char __user *optval, int __user *optlen) 5166 { 5167 int val; 5168 5169 if (len < sizeof(int)) 5170 return -EINVAL; 5171 5172 len = sizeof(int); 5173 val = (sctp_sk(sk)->nodelay == 1); 5174 if (put_user(len, optlen)) 5175 return -EFAULT; 5176 if (copy_to_user(optval, &val, len)) 5177 return -EFAULT; 5178 return 0; 5179 } 5180 5181 /* 5182 * 5183 * 7.1.1 SCTP_RTOINFO 5184 * 5185 * The protocol parameters used to initialize and bound retransmission 5186 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access 5187 * and modify these parameters. 5188 * All parameters are time values, in milliseconds. A value of 0, when 5189 * modifying the parameters, indicates that the current value should not 5190 * be changed. 5191 * 5192 */ 5193 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len, 5194 char __user *optval, 5195 int __user *optlen) { 5196 struct sctp_rtoinfo rtoinfo; 5197 struct sctp_association *asoc; 5198 5199 if (len < sizeof (struct sctp_rtoinfo)) 5200 return -EINVAL; 5201 5202 len = sizeof(struct sctp_rtoinfo); 5203 5204 if (copy_from_user(&rtoinfo, optval, len)) 5205 return -EFAULT; 5206 5207 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id); 5208 5209 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP)) 5210 return -EINVAL; 5211 5212 /* Values corresponding to the specific association. */ 5213 if (asoc) { 5214 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial); 5215 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max); 5216 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min); 5217 } else { 5218 /* Values corresponding to the endpoint. */ 5219 struct sctp_sock *sp = sctp_sk(sk); 5220 5221 rtoinfo.srto_initial = sp->rtoinfo.srto_initial; 5222 rtoinfo.srto_max = sp->rtoinfo.srto_max; 5223 rtoinfo.srto_min = sp->rtoinfo.srto_min; 5224 } 5225 5226 if (put_user(len, optlen)) 5227 return -EFAULT; 5228 5229 if (copy_to_user(optval, &rtoinfo, len)) 5230 return -EFAULT; 5231 5232 return 0; 5233 } 5234 5235 /* 5236 * 5237 * 7.1.2 SCTP_ASSOCINFO 5238 * 5239 * This option is used to tune the maximum retransmission attempts 5240 * of the association. 5241 * Returns an error if the new association retransmission value is 5242 * greater than the sum of the retransmission value of the peer. 5243 * See [SCTP] for more information. 5244 * 5245 */ 5246 static int sctp_getsockopt_associnfo(struct sock *sk, int len, 5247 char __user *optval, 5248 int __user *optlen) 5249 { 5250 5251 struct sctp_assocparams assocparams; 5252 struct sctp_association *asoc; 5253 struct list_head *pos; 5254 int cnt = 0; 5255 5256 if (len < sizeof (struct sctp_assocparams)) 5257 return -EINVAL; 5258 5259 len = sizeof(struct sctp_assocparams); 5260 5261 if (copy_from_user(&assocparams, optval, len)) 5262 return -EFAULT; 5263 5264 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id); 5265 5266 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP)) 5267 return -EINVAL; 5268 5269 /* Values correspoinding to the specific association */ 5270 if (asoc) { 5271 assocparams.sasoc_asocmaxrxt = asoc->max_retrans; 5272 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd; 5273 assocparams.sasoc_local_rwnd = asoc->a_rwnd; 5274 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life); 5275 5276 list_for_each(pos, &asoc->peer.transport_addr_list) { 5277 cnt++; 5278 } 5279 5280 assocparams.sasoc_number_peer_destinations = cnt; 5281 } else { 5282 /* Values corresponding to the endpoint */ 5283 struct sctp_sock *sp = sctp_sk(sk); 5284 5285 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt; 5286 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd; 5287 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd; 5288 assocparams.sasoc_cookie_life = 5289 sp->assocparams.sasoc_cookie_life; 5290 assocparams.sasoc_number_peer_destinations = 5291 sp->assocparams. 5292 sasoc_number_peer_destinations; 5293 } 5294 5295 if (put_user(len, optlen)) 5296 return -EFAULT; 5297 5298 if (copy_to_user(optval, &assocparams, len)) 5299 return -EFAULT; 5300 5301 return 0; 5302 } 5303 5304 /* 5305 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR) 5306 * 5307 * This socket option is a boolean flag which turns on or off mapped V4 5308 * addresses. If this option is turned on and the socket is type 5309 * PF_INET6, then IPv4 addresses will be mapped to V6 representation. 5310 * If this option is turned off, then no mapping will be done of V4 5311 * addresses and a user will receive both PF_INET6 and PF_INET type 5312 * addresses on the socket. 5313 */ 5314 static int sctp_getsockopt_mappedv4(struct sock *sk, int len, 5315 char __user *optval, int __user *optlen) 5316 { 5317 int val; 5318 struct sctp_sock *sp = sctp_sk(sk); 5319 5320 if (len < sizeof(int)) 5321 return -EINVAL; 5322 5323 len = sizeof(int); 5324 val = sp->v4mapped; 5325 if (put_user(len, optlen)) 5326 return -EFAULT; 5327 if (copy_to_user(optval, &val, len)) 5328 return -EFAULT; 5329 5330 return 0; 5331 } 5332 5333 /* 5334 * 7.1.29. Set or Get the default context (SCTP_CONTEXT) 5335 * (chapter and verse is quoted at sctp_setsockopt_context()) 5336 */ 5337 static int sctp_getsockopt_context(struct sock *sk, int len, 5338 char __user *optval, int __user *optlen) 5339 { 5340 struct sctp_assoc_value params; 5341 struct sctp_sock *sp; 5342 struct sctp_association *asoc; 5343 5344 if (len < sizeof(struct sctp_assoc_value)) 5345 return -EINVAL; 5346 5347 len = sizeof(struct sctp_assoc_value); 5348 5349 if (copy_from_user(¶ms, optval, len)) 5350 return -EFAULT; 5351 5352 sp = sctp_sk(sk); 5353 5354 if (params.assoc_id != 0) { 5355 asoc = sctp_id2assoc(sk, params.assoc_id); 5356 if (!asoc) 5357 return -EINVAL; 5358 params.assoc_value = asoc->default_rcv_context; 5359 } else { 5360 params.assoc_value = sp->default_rcv_context; 5361 } 5362 5363 if (put_user(len, optlen)) 5364 return -EFAULT; 5365 if (copy_to_user(optval, ¶ms, len)) 5366 return -EFAULT; 5367 5368 return 0; 5369 } 5370 5371 /* 5372 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG) 5373 * This option will get or set the maximum size to put in any outgoing 5374 * SCTP DATA chunk. If a message is larger than this size it will be 5375 * fragmented by SCTP into the specified size. Note that the underlying 5376 * SCTP implementation may fragment into smaller sized chunks when the 5377 * PMTU of the underlying association is smaller than the value set by 5378 * the user. The default value for this option is '0' which indicates 5379 * the user is NOT limiting fragmentation and only the PMTU will effect 5380 * SCTP's choice of DATA chunk size. Note also that values set larger 5381 * than the maximum size of an IP datagram will effectively let SCTP 5382 * control fragmentation (i.e. the same as setting this option to 0). 5383 * 5384 * The following structure is used to access and modify this parameter: 5385 * 5386 * struct sctp_assoc_value { 5387 * sctp_assoc_t assoc_id; 5388 * uint32_t assoc_value; 5389 * }; 5390 * 5391 * assoc_id: This parameter is ignored for one-to-one style sockets. 5392 * For one-to-many style sockets this parameter indicates which 5393 * association the user is performing an action upon. Note that if 5394 * this field's value is zero then the endpoints default value is 5395 * changed (effecting future associations only). 5396 * assoc_value: This parameter specifies the maximum size in bytes. 5397 */ 5398 static int sctp_getsockopt_maxseg(struct sock *sk, int len, 5399 char __user *optval, int __user *optlen) 5400 { 5401 struct sctp_assoc_value params; 5402 struct sctp_association *asoc; 5403 5404 if (len == sizeof(int)) { 5405 pr_warn_ratelimited(DEPRECATED 5406 "%s (pid %d) " 5407 "Use of int in maxseg socket option.\n" 5408 "Use struct sctp_assoc_value instead\n", 5409 current->comm, task_pid_nr(current)); 5410 params.assoc_id = 0; 5411 } else if (len >= sizeof(struct sctp_assoc_value)) { 5412 len = sizeof(struct sctp_assoc_value); 5413 if (copy_from_user(¶ms, optval, sizeof(params))) 5414 return -EFAULT; 5415 } else 5416 return -EINVAL; 5417 5418 asoc = sctp_id2assoc(sk, params.assoc_id); 5419 if (!asoc && params.assoc_id && sctp_style(sk, UDP)) 5420 return -EINVAL; 5421 5422 if (asoc) 5423 params.assoc_value = asoc->frag_point; 5424 else 5425 params.assoc_value = sctp_sk(sk)->user_frag; 5426 5427 if (put_user(len, optlen)) 5428 return -EFAULT; 5429 if (len == sizeof(int)) { 5430 if (copy_to_user(optval, ¶ms.assoc_value, len)) 5431 return -EFAULT; 5432 } else { 5433 if (copy_to_user(optval, ¶ms, len)) 5434 return -EFAULT; 5435 } 5436 5437 return 0; 5438 } 5439 5440 /* 5441 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE) 5442 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave()) 5443 */ 5444 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len, 5445 char __user *optval, int __user *optlen) 5446 { 5447 int val; 5448 5449 if (len < sizeof(int)) 5450 return -EINVAL; 5451 5452 len = sizeof(int); 5453 5454 val = sctp_sk(sk)->frag_interleave; 5455 if (put_user(len, optlen)) 5456 return -EFAULT; 5457 if (copy_to_user(optval, &val, len)) 5458 return -EFAULT; 5459 5460 return 0; 5461 } 5462 5463 /* 5464 * 7.1.25. Set or Get the sctp partial delivery point 5465 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point()) 5466 */ 5467 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len, 5468 char __user *optval, 5469 int __user *optlen) 5470 { 5471 u32 val; 5472 5473 if (len < sizeof(u32)) 5474 return -EINVAL; 5475 5476 len = sizeof(u32); 5477 5478 val = sctp_sk(sk)->pd_point; 5479 if (put_user(len, optlen)) 5480 return -EFAULT; 5481 if (copy_to_user(optval, &val, len)) 5482 return -EFAULT; 5483 5484 return 0; 5485 } 5486 5487 /* 5488 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST) 5489 * (chapter and verse is quoted at sctp_setsockopt_maxburst()) 5490 */ 5491 static int sctp_getsockopt_maxburst(struct sock *sk, int len, 5492 char __user *optval, 5493 int __user *optlen) 5494 { 5495 struct sctp_assoc_value params; 5496 struct sctp_sock *sp; 5497 struct sctp_association *asoc; 5498 5499 if (len == sizeof(int)) { 5500 pr_warn_ratelimited(DEPRECATED 5501 "%s (pid %d) " 5502 "Use of int in max_burst socket option.\n" 5503 "Use struct sctp_assoc_value instead\n", 5504 current->comm, task_pid_nr(current)); 5505 params.assoc_id = 0; 5506 } else if (len >= sizeof(struct sctp_assoc_value)) { 5507 len = sizeof(struct sctp_assoc_value); 5508 if (copy_from_user(¶ms, optval, len)) 5509 return -EFAULT; 5510 } else 5511 return -EINVAL; 5512 5513 sp = sctp_sk(sk); 5514 5515 if (params.assoc_id != 0) { 5516 asoc = sctp_id2assoc(sk, params.assoc_id); 5517 if (!asoc) 5518 return -EINVAL; 5519 params.assoc_value = asoc->max_burst; 5520 } else 5521 params.assoc_value = sp->max_burst; 5522 5523 if (len == sizeof(int)) { 5524 if (copy_to_user(optval, ¶ms.assoc_value, len)) 5525 return -EFAULT; 5526 } else { 5527 if (copy_to_user(optval, ¶ms, len)) 5528 return -EFAULT; 5529 } 5530 5531 return 0; 5532 5533 } 5534 5535 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len, 5536 char __user *optval, int __user *optlen) 5537 { 5538 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 5539 struct sctp_hmacalgo __user *p = (void __user *)optval; 5540 struct sctp_hmac_algo_param *hmacs; 5541 __u16 data_len = 0; 5542 u32 num_idents; 5543 5544 if (!ep->auth_enable) 5545 return -EACCES; 5546 5547 hmacs = ep->auth_hmacs_list; 5548 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t); 5549 5550 if (len < sizeof(struct sctp_hmacalgo) + data_len) 5551 return -EINVAL; 5552 5553 len = sizeof(struct sctp_hmacalgo) + data_len; 5554 num_idents = data_len / sizeof(u16); 5555 5556 if (put_user(len, optlen)) 5557 return -EFAULT; 5558 if (put_user(num_idents, &p->shmac_num_idents)) 5559 return -EFAULT; 5560 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len)) 5561 return -EFAULT; 5562 return 0; 5563 } 5564 5565 static int sctp_getsockopt_active_key(struct sock *sk, int len, 5566 char __user *optval, int __user *optlen) 5567 { 5568 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 5569 struct sctp_authkeyid val; 5570 struct sctp_association *asoc; 5571 5572 if (!ep->auth_enable) 5573 return -EACCES; 5574 5575 if (len < sizeof(struct sctp_authkeyid)) 5576 return -EINVAL; 5577 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid))) 5578 return -EFAULT; 5579 5580 asoc = sctp_id2assoc(sk, val.scact_assoc_id); 5581 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP)) 5582 return -EINVAL; 5583 5584 if (asoc) 5585 val.scact_keynumber = asoc->active_key_id; 5586 else 5587 val.scact_keynumber = ep->active_key_id; 5588 5589 len = sizeof(struct sctp_authkeyid); 5590 if (put_user(len, optlen)) 5591 return -EFAULT; 5592 if (copy_to_user(optval, &val, len)) 5593 return -EFAULT; 5594 5595 return 0; 5596 } 5597 5598 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len, 5599 char __user *optval, int __user *optlen) 5600 { 5601 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 5602 struct sctp_authchunks __user *p = (void __user *)optval; 5603 struct sctp_authchunks val; 5604 struct sctp_association *asoc; 5605 struct sctp_chunks_param *ch; 5606 u32 num_chunks = 0; 5607 char __user *to; 5608 5609 if (!ep->auth_enable) 5610 return -EACCES; 5611 5612 if (len < sizeof(struct sctp_authchunks)) 5613 return -EINVAL; 5614 5615 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks))) 5616 return -EFAULT; 5617 5618 to = p->gauth_chunks; 5619 asoc = sctp_id2assoc(sk, val.gauth_assoc_id); 5620 if (!asoc) 5621 return -EINVAL; 5622 5623 ch = asoc->peer.peer_chunks; 5624 if (!ch) 5625 goto num; 5626 5627 /* See if the user provided enough room for all the data */ 5628 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t); 5629 if (len < num_chunks) 5630 return -EINVAL; 5631 5632 if (copy_to_user(to, ch->chunks, num_chunks)) 5633 return -EFAULT; 5634 num: 5635 len = sizeof(struct sctp_authchunks) + num_chunks; 5636 if (put_user(len, optlen)) 5637 return -EFAULT; 5638 if (put_user(num_chunks, &p->gauth_number_of_chunks)) 5639 return -EFAULT; 5640 return 0; 5641 } 5642 5643 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len, 5644 char __user *optval, int __user *optlen) 5645 { 5646 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 5647 struct sctp_authchunks __user *p = (void __user *)optval; 5648 struct sctp_authchunks val; 5649 struct sctp_association *asoc; 5650 struct sctp_chunks_param *ch; 5651 u32 num_chunks = 0; 5652 char __user *to; 5653 5654 if (!ep->auth_enable) 5655 return -EACCES; 5656 5657 if (len < sizeof(struct sctp_authchunks)) 5658 return -EINVAL; 5659 5660 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks))) 5661 return -EFAULT; 5662 5663 to = p->gauth_chunks; 5664 asoc = sctp_id2assoc(sk, val.gauth_assoc_id); 5665 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP)) 5666 return -EINVAL; 5667 5668 if (asoc) 5669 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks; 5670 else 5671 ch = ep->auth_chunk_list; 5672 5673 if (!ch) 5674 goto num; 5675 5676 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t); 5677 if (len < sizeof(struct sctp_authchunks) + num_chunks) 5678 return -EINVAL; 5679 5680 if (copy_to_user(to, ch->chunks, num_chunks)) 5681 return -EFAULT; 5682 num: 5683 len = sizeof(struct sctp_authchunks) + num_chunks; 5684 if (put_user(len, optlen)) 5685 return -EFAULT; 5686 if (put_user(num_chunks, &p->gauth_number_of_chunks)) 5687 return -EFAULT; 5688 5689 return 0; 5690 } 5691 5692 /* 5693 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER) 5694 * This option gets the current number of associations that are attached 5695 * to a one-to-many style socket. The option value is an uint32_t. 5696 */ 5697 static int sctp_getsockopt_assoc_number(struct sock *sk, int len, 5698 char __user *optval, int __user *optlen) 5699 { 5700 struct sctp_sock *sp = sctp_sk(sk); 5701 struct sctp_association *asoc; 5702 u32 val = 0; 5703 5704 if (sctp_style(sk, TCP)) 5705 return -EOPNOTSUPP; 5706 5707 if (len < sizeof(u32)) 5708 return -EINVAL; 5709 5710 len = sizeof(u32); 5711 5712 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) { 5713 val++; 5714 } 5715 5716 if (put_user(len, optlen)) 5717 return -EFAULT; 5718 if (copy_to_user(optval, &val, len)) 5719 return -EFAULT; 5720 5721 return 0; 5722 } 5723 5724 /* 5725 * 8.1.23 SCTP_AUTO_ASCONF 5726 * See the corresponding setsockopt entry as description 5727 */ 5728 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len, 5729 char __user *optval, int __user *optlen) 5730 { 5731 int val = 0; 5732 5733 if (len < sizeof(int)) 5734 return -EINVAL; 5735 5736 len = sizeof(int); 5737 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk)) 5738 val = 1; 5739 if (put_user(len, optlen)) 5740 return -EFAULT; 5741 if (copy_to_user(optval, &val, len)) 5742 return -EFAULT; 5743 return 0; 5744 } 5745 5746 /* 5747 * 8.2.6. Get the Current Identifiers of Associations 5748 * (SCTP_GET_ASSOC_ID_LIST) 5749 * 5750 * This option gets the current list of SCTP association identifiers of 5751 * the SCTP associations handled by a one-to-many style socket. 5752 */ 5753 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len, 5754 char __user *optval, int __user *optlen) 5755 { 5756 struct sctp_sock *sp = sctp_sk(sk); 5757 struct sctp_association *asoc; 5758 struct sctp_assoc_ids *ids; 5759 u32 num = 0; 5760 5761 if (sctp_style(sk, TCP)) 5762 return -EOPNOTSUPP; 5763 5764 if (len < sizeof(struct sctp_assoc_ids)) 5765 return -EINVAL; 5766 5767 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) { 5768 num++; 5769 } 5770 5771 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num) 5772 return -EINVAL; 5773 5774 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num; 5775 5776 ids = kmalloc(len, GFP_KERNEL); 5777 if (unlikely(!ids)) 5778 return -ENOMEM; 5779 5780 ids->gaids_number_of_ids = num; 5781 num = 0; 5782 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) { 5783 ids->gaids_assoc_id[num++] = asoc->assoc_id; 5784 } 5785 5786 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) { 5787 kfree(ids); 5788 return -EFAULT; 5789 } 5790 5791 kfree(ids); 5792 return 0; 5793 } 5794 5795 /* 5796 * SCTP_PEER_ADDR_THLDS 5797 * 5798 * This option allows us to fetch the partially failed threshold for one or all 5799 * transports in an association. See Section 6.1 of: 5800 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt 5801 */ 5802 static int sctp_getsockopt_paddr_thresholds(struct sock *sk, 5803 char __user *optval, 5804 int len, 5805 int __user *optlen) 5806 { 5807 struct sctp_paddrthlds val; 5808 struct sctp_transport *trans; 5809 struct sctp_association *asoc; 5810 5811 if (len < sizeof(struct sctp_paddrthlds)) 5812 return -EINVAL; 5813 len = sizeof(struct sctp_paddrthlds); 5814 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len)) 5815 return -EFAULT; 5816 5817 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) { 5818 asoc = sctp_id2assoc(sk, val.spt_assoc_id); 5819 if (!asoc) 5820 return -ENOENT; 5821 5822 val.spt_pathpfthld = asoc->pf_retrans; 5823 val.spt_pathmaxrxt = asoc->pathmaxrxt; 5824 } else { 5825 trans = sctp_addr_id2transport(sk, &val.spt_address, 5826 val.spt_assoc_id); 5827 if (!trans) 5828 return -ENOENT; 5829 5830 val.spt_pathmaxrxt = trans->pathmaxrxt; 5831 val.spt_pathpfthld = trans->pf_retrans; 5832 } 5833 5834 if (put_user(len, optlen) || copy_to_user(optval, &val, len)) 5835 return -EFAULT; 5836 5837 return 0; 5838 } 5839 5840 /* 5841 * SCTP_GET_ASSOC_STATS 5842 * 5843 * This option retrieves local per endpoint statistics. It is modeled 5844 * after OpenSolaris' implementation 5845 */ 5846 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len, 5847 char __user *optval, 5848 int __user *optlen) 5849 { 5850 struct sctp_assoc_stats sas; 5851 struct sctp_association *asoc = NULL; 5852 5853 /* User must provide at least the assoc id */ 5854 if (len < sizeof(sctp_assoc_t)) 5855 return -EINVAL; 5856 5857 /* Allow the struct to grow and fill in as much as possible */ 5858 len = min_t(size_t, len, sizeof(sas)); 5859 5860 if (copy_from_user(&sas, optval, len)) 5861 return -EFAULT; 5862 5863 asoc = sctp_id2assoc(sk, sas.sas_assoc_id); 5864 if (!asoc) 5865 return -EINVAL; 5866 5867 sas.sas_rtxchunks = asoc->stats.rtxchunks; 5868 sas.sas_gapcnt = asoc->stats.gapcnt; 5869 sas.sas_outofseqtsns = asoc->stats.outofseqtsns; 5870 sas.sas_osacks = asoc->stats.osacks; 5871 sas.sas_isacks = asoc->stats.isacks; 5872 sas.sas_octrlchunks = asoc->stats.octrlchunks; 5873 sas.sas_ictrlchunks = asoc->stats.ictrlchunks; 5874 sas.sas_oodchunks = asoc->stats.oodchunks; 5875 sas.sas_iodchunks = asoc->stats.iodchunks; 5876 sas.sas_ouodchunks = asoc->stats.ouodchunks; 5877 sas.sas_iuodchunks = asoc->stats.iuodchunks; 5878 sas.sas_idupchunks = asoc->stats.idupchunks; 5879 sas.sas_opackets = asoc->stats.opackets; 5880 sas.sas_ipackets = asoc->stats.ipackets; 5881 5882 /* New high max rto observed, will return 0 if not a single 5883 * RTO update took place. obs_rto_ipaddr will be bogus 5884 * in such a case 5885 */ 5886 sas.sas_maxrto = asoc->stats.max_obs_rto; 5887 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr, 5888 sizeof(struct sockaddr_storage)); 5889 5890 /* Mark beginning of a new observation period */ 5891 asoc->stats.max_obs_rto = asoc->rto_min; 5892 5893 if (put_user(len, optlen)) 5894 return -EFAULT; 5895 5896 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id); 5897 5898 if (copy_to_user(optval, &sas, len)) 5899 return -EFAULT; 5900 5901 return 0; 5902 } 5903 5904 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len, 5905 char __user *optval, 5906 int __user *optlen) 5907 { 5908 int val = 0; 5909 5910 if (len < sizeof(int)) 5911 return -EINVAL; 5912 5913 len = sizeof(int); 5914 if (sctp_sk(sk)->recvrcvinfo) 5915 val = 1; 5916 if (put_user(len, optlen)) 5917 return -EFAULT; 5918 if (copy_to_user(optval, &val, len)) 5919 return -EFAULT; 5920 5921 return 0; 5922 } 5923 5924 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len, 5925 char __user *optval, 5926 int __user *optlen) 5927 { 5928 int val = 0; 5929 5930 if (len < sizeof(int)) 5931 return -EINVAL; 5932 5933 len = sizeof(int); 5934 if (sctp_sk(sk)->recvnxtinfo) 5935 val = 1; 5936 if (put_user(len, optlen)) 5937 return -EFAULT; 5938 if (copy_to_user(optval, &val, len)) 5939 return -EFAULT; 5940 5941 return 0; 5942 } 5943 5944 static int sctp_getsockopt(struct sock *sk, int level, int optname, 5945 char __user *optval, int __user *optlen) 5946 { 5947 int retval = 0; 5948 int len; 5949 5950 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname); 5951 5952 /* I can hardly begin to describe how wrong this is. This is 5953 * so broken as to be worse than useless. The API draft 5954 * REALLY is NOT helpful here... I am not convinced that the 5955 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP 5956 * are at all well-founded. 5957 */ 5958 if (level != SOL_SCTP) { 5959 struct sctp_af *af = sctp_sk(sk)->pf->af; 5960 5961 retval = af->getsockopt(sk, level, optname, optval, optlen); 5962 return retval; 5963 } 5964 5965 if (get_user(len, optlen)) 5966 return -EFAULT; 5967 5968 lock_sock(sk); 5969 5970 switch (optname) { 5971 case SCTP_STATUS: 5972 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen); 5973 break; 5974 case SCTP_DISABLE_FRAGMENTS: 5975 retval = sctp_getsockopt_disable_fragments(sk, len, optval, 5976 optlen); 5977 break; 5978 case SCTP_EVENTS: 5979 retval = sctp_getsockopt_events(sk, len, optval, optlen); 5980 break; 5981 case SCTP_AUTOCLOSE: 5982 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen); 5983 break; 5984 case SCTP_SOCKOPT_PEELOFF: 5985 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen); 5986 break; 5987 case SCTP_PEER_ADDR_PARAMS: 5988 retval = sctp_getsockopt_peer_addr_params(sk, len, optval, 5989 optlen); 5990 break; 5991 case SCTP_DELAYED_SACK: 5992 retval = sctp_getsockopt_delayed_ack(sk, len, optval, 5993 optlen); 5994 break; 5995 case SCTP_INITMSG: 5996 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen); 5997 break; 5998 case SCTP_GET_PEER_ADDRS: 5999 retval = sctp_getsockopt_peer_addrs(sk, len, optval, 6000 optlen); 6001 break; 6002 case SCTP_GET_LOCAL_ADDRS: 6003 retval = sctp_getsockopt_local_addrs(sk, len, optval, 6004 optlen); 6005 break; 6006 case SCTP_SOCKOPT_CONNECTX3: 6007 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen); 6008 break; 6009 case SCTP_DEFAULT_SEND_PARAM: 6010 retval = sctp_getsockopt_default_send_param(sk, len, 6011 optval, optlen); 6012 break; 6013 case SCTP_DEFAULT_SNDINFO: 6014 retval = sctp_getsockopt_default_sndinfo(sk, len, 6015 optval, optlen); 6016 break; 6017 case SCTP_PRIMARY_ADDR: 6018 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen); 6019 break; 6020 case SCTP_NODELAY: 6021 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen); 6022 break; 6023 case SCTP_RTOINFO: 6024 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen); 6025 break; 6026 case SCTP_ASSOCINFO: 6027 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen); 6028 break; 6029 case SCTP_I_WANT_MAPPED_V4_ADDR: 6030 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen); 6031 break; 6032 case SCTP_MAXSEG: 6033 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen); 6034 break; 6035 case SCTP_GET_PEER_ADDR_INFO: 6036 retval = sctp_getsockopt_peer_addr_info(sk, len, optval, 6037 optlen); 6038 break; 6039 case SCTP_ADAPTATION_LAYER: 6040 retval = sctp_getsockopt_adaptation_layer(sk, len, optval, 6041 optlen); 6042 break; 6043 case SCTP_CONTEXT: 6044 retval = sctp_getsockopt_context(sk, len, optval, optlen); 6045 break; 6046 case SCTP_FRAGMENT_INTERLEAVE: 6047 retval = sctp_getsockopt_fragment_interleave(sk, len, optval, 6048 optlen); 6049 break; 6050 case SCTP_PARTIAL_DELIVERY_POINT: 6051 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval, 6052 optlen); 6053 break; 6054 case SCTP_MAX_BURST: 6055 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen); 6056 break; 6057 case SCTP_AUTH_KEY: 6058 case SCTP_AUTH_CHUNK: 6059 case SCTP_AUTH_DELETE_KEY: 6060 retval = -EOPNOTSUPP; 6061 break; 6062 case SCTP_HMAC_IDENT: 6063 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen); 6064 break; 6065 case SCTP_AUTH_ACTIVE_KEY: 6066 retval = sctp_getsockopt_active_key(sk, len, optval, optlen); 6067 break; 6068 case SCTP_PEER_AUTH_CHUNKS: 6069 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval, 6070 optlen); 6071 break; 6072 case SCTP_LOCAL_AUTH_CHUNKS: 6073 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval, 6074 optlen); 6075 break; 6076 case SCTP_GET_ASSOC_NUMBER: 6077 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen); 6078 break; 6079 case SCTP_GET_ASSOC_ID_LIST: 6080 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen); 6081 break; 6082 case SCTP_AUTO_ASCONF: 6083 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen); 6084 break; 6085 case SCTP_PEER_ADDR_THLDS: 6086 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen); 6087 break; 6088 case SCTP_GET_ASSOC_STATS: 6089 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen); 6090 break; 6091 case SCTP_RECVRCVINFO: 6092 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen); 6093 break; 6094 case SCTP_RECVNXTINFO: 6095 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen); 6096 break; 6097 default: 6098 retval = -ENOPROTOOPT; 6099 break; 6100 } 6101 6102 release_sock(sk); 6103 return retval; 6104 } 6105 6106 static void sctp_hash(struct sock *sk) 6107 { 6108 /* STUB */ 6109 } 6110 6111 static void sctp_unhash(struct sock *sk) 6112 { 6113 /* STUB */ 6114 } 6115 6116 /* Check if port is acceptable. Possibly find first available port. 6117 * 6118 * The port hash table (contained in the 'global' SCTP protocol storage 6119 * returned by struct sctp_protocol *sctp_get_protocol()). The hash 6120 * table is an array of 4096 lists (sctp_bind_hashbucket). Each 6121 * list (the list number is the port number hashed out, so as you 6122 * would expect from a hash function, all the ports in a given list have 6123 * such a number that hashes out to the same list number; you were 6124 * expecting that, right?); so each list has a set of ports, with a 6125 * link to the socket (struct sock) that uses it, the port number and 6126 * a fastreuse flag (FIXME: NPI ipg). 6127 */ 6128 static struct sctp_bind_bucket *sctp_bucket_create( 6129 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum); 6130 6131 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr) 6132 { 6133 struct sctp_bind_hashbucket *head; /* hash list */ 6134 struct sctp_bind_bucket *pp; 6135 unsigned short snum; 6136 int ret; 6137 6138 snum = ntohs(addr->v4.sin_port); 6139 6140 pr_debug("%s: begins, snum:%d\n", __func__, snum); 6141 6142 local_bh_disable(); 6143 6144 if (snum == 0) { 6145 /* Search for an available port. */ 6146 int low, high, remaining, index; 6147 unsigned int rover; 6148 struct net *net = sock_net(sk); 6149 6150 inet_get_local_port_range(net, &low, &high); 6151 remaining = (high - low) + 1; 6152 rover = prandom_u32() % remaining + low; 6153 6154 do { 6155 rover++; 6156 if ((rover < low) || (rover > high)) 6157 rover = low; 6158 if (inet_is_local_reserved_port(net, rover)) 6159 continue; 6160 index = sctp_phashfn(sock_net(sk), rover); 6161 head = &sctp_port_hashtable[index]; 6162 spin_lock(&head->lock); 6163 sctp_for_each_hentry(pp, &head->chain) 6164 if ((pp->port == rover) && 6165 net_eq(sock_net(sk), pp->net)) 6166 goto next; 6167 break; 6168 next: 6169 spin_unlock(&head->lock); 6170 } while (--remaining > 0); 6171 6172 /* Exhausted local port range during search? */ 6173 ret = 1; 6174 if (remaining <= 0) 6175 goto fail; 6176 6177 /* OK, here is the one we will use. HEAD (the port 6178 * hash table list entry) is non-NULL and we hold it's 6179 * mutex. 6180 */ 6181 snum = rover; 6182 } else { 6183 /* We are given an specific port number; we verify 6184 * that it is not being used. If it is used, we will 6185 * exahust the search in the hash list corresponding 6186 * to the port number (snum) - we detect that with the 6187 * port iterator, pp being NULL. 6188 */ 6189 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)]; 6190 spin_lock(&head->lock); 6191 sctp_for_each_hentry(pp, &head->chain) { 6192 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk))) 6193 goto pp_found; 6194 } 6195 } 6196 pp = NULL; 6197 goto pp_not_found; 6198 pp_found: 6199 if (!hlist_empty(&pp->owner)) { 6200 /* We had a port hash table hit - there is an 6201 * available port (pp != NULL) and it is being 6202 * used by other socket (pp->owner not empty); that other 6203 * socket is going to be sk2. 6204 */ 6205 int reuse = sk->sk_reuse; 6206 struct sock *sk2; 6207 6208 pr_debug("%s: found a possible match\n", __func__); 6209 6210 if (pp->fastreuse && sk->sk_reuse && 6211 sk->sk_state != SCTP_SS_LISTENING) 6212 goto success; 6213 6214 /* Run through the list of sockets bound to the port 6215 * (pp->port) [via the pointers bind_next and 6216 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one, 6217 * we get the endpoint they describe and run through 6218 * the endpoint's list of IP (v4 or v6) addresses, 6219 * comparing each of the addresses with the address of 6220 * the socket sk. If we find a match, then that means 6221 * that this port/socket (sk) combination are already 6222 * in an endpoint. 6223 */ 6224 sk_for_each_bound(sk2, &pp->owner) { 6225 struct sctp_endpoint *ep2; 6226 ep2 = sctp_sk(sk2)->ep; 6227 6228 if (sk == sk2 || 6229 (reuse && sk2->sk_reuse && 6230 sk2->sk_state != SCTP_SS_LISTENING)) 6231 continue; 6232 6233 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr, 6234 sctp_sk(sk2), sctp_sk(sk))) { 6235 ret = (long)sk2; 6236 goto fail_unlock; 6237 } 6238 } 6239 6240 pr_debug("%s: found a match\n", __func__); 6241 } 6242 pp_not_found: 6243 /* If there was a hash table miss, create a new port. */ 6244 ret = 1; 6245 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum))) 6246 goto fail_unlock; 6247 6248 /* In either case (hit or miss), make sure fastreuse is 1 only 6249 * if sk->sk_reuse is too (that is, if the caller requested 6250 * SO_REUSEADDR on this socket -sk-). 6251 */ 6252 if (hlist_empty(&pp->owner)) { 6253 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING) 6254 pp->fastreuse = 1; 6255 else 6256 pp->fastreuse = 0; 6257 } else if (pp->fastreuse && 6258 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING)) 6259 pp->fastreuse = 0; 6260 6261 /* We are set, so fill up all the data in the hash table 6262 * entry, tie the socket list information with the rest of the 6263 * sockets FIXME: Blurry, NPI (ipg). 6264 */ 6265 success: 6266 if (!sctp_sk(sk)->bind_hash) { 6267 inet_sk(sk)->inet_num = snum; 6268 sk_add_bind_node(sk, &pp->owner); 6269 sctp_sk(sk)->bind_hash = pp; 6270 } 6271 ret = 0; 6272 6273 fail_unlock: 6274 spin_unlock(&head->lock); 6275 6276 fail: 6277 local_bh_enable(); 6278 return ret; 6279 } 6280 6281 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral 6282 * port is requested. 6283 */ 6284 static int sctp_get_port(struct sock *sk, unsigned short snum) 6285 { 6286 union sctp_addr addr; 6287 struct sctp_af *af = sctp_sk(sk)->pf->af; 6288 6289 /* Set up a dummy address struct from the sk. */ 6290 af->from_sk(&addr, sk); 6291 addr.v4.sin_port = htons(snum); 6292 6293 /* Note: sk->sk_num gets filled in if ephemeral port request. */ 6294 return !!sctp_get_port_local(sk, &addr); 6295 } 6296 6297 /* 6298 * Move a socket to LISTENING state. 6299 */ 6300 static int sctp_listen_start(struct sock *sk, int backlog) 6301 { 6302 struct sctp_sock *sp = sctp_sk(sk); 6303 struct sctp_endpoint *ep = sp->ep; 6304 struct crypto_hash *tfm = NULL; 6305 char alg[32]; 6306 6307 /* Allocate HMAC for generating cookie. */ 6308 if (!sp->hmac && sp->sctp_hmac_alg) { 6309 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg); 6310 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC); 6311 if (IS_ERR(tfm)) { 6312 net_info_ratelimited("failed to load transform for %s: %ld\n", 6313 sp->sctp_hmac_alg, PTR_ERR(tfm)); 6314 return -ENOSYS; 6315 } 6316 sctp_sk(sk)->hmac = tfm; 6317 } 6318 6319 /* 6320 * If a bind() or sctp_bindx() is not called prior to a listen() 6321 * call that allows new associations to be accepted, the system 6322 * picks an ephemeral port and will choose an address set equivalent 6323 * to binding with a wildcard address. 6324 * 6325 * This is not currently spelled out in the SCTP sockets 6326 * extensions draft, but follows the practice as seen in TCP 6327 * sockets. 6328 * 6329 */ 6330 sk->sk_state = SCTP_SS_LISTENING; 6331 if (!ep->base.bind_addr.port) { 6332 if (sctp_autobind(sk)) 6333 return -EAGAIN; 6334 } else { 6335 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) { 6336 sk->sk_state = SCTP_SS_CLOSED; 6337 return -EADDRINUSE; 6338 } 6339 } 6340 6341 sk->sk_max_ack_backlog = backlog; 6342 sctp_hash_endpoint(ep); 6343 return 0; 6344 } 6345 6346 /* 6347 * 4.1.3 / 5.1.3 listen() 6348 * 6349 * By default, new associations are not accepted for UDP style sockets. 6350 * An application uses listen() to mark a socket as being able to 6351 * accept new associations. 6352 * 6353 * On TCP style sockets, applications use listen() to ready the SCTP 6354 * endpoint for accepting inbound associations. 6355 * 6356 * On both types of endpoints a backlog of '0' disables listening. 6357 * 6358 * Move a socket to LISTENING state. 6359 */ 6360 int sctp_inet_listen(struct socket *sock, int backlog) 6361 { 6362 struct sock *sk = sock->sk; 6363 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 6364 int err = -EINVAL; 6365 6366 if (unlikely(backlog < 0)) 6367 return err; 6368 6369 lock_sock(sk); 6370 6371 /* Peeled-off sockets are not allowed to listen(). */ 6372 if (sctp_style(sk, UDP_HIGH_BANDWIDTH)) 6373 goto out; 6374 6375 if (sock->state != SS_UNCONNECTED) 6376 goto out; 6377 6378 /* If backlog is zero, disable listening. */ 6379 if (!backlog) { 6380 if (sctp_sstate(sk, CLOSED)) 6381 goto out; 6382 6383 err = 0; 6384 sctp_unhash_endpoint(ep); 6385 sk->sk_state = SCTP_SS_CLOSED; 6386 if (sk->sk_reuse) 6387 sctp_sk(sk)->bind_hash->fastreuse = 1; 6388 goto out; 6389 } 6390 6391 /* If we are already listening, just update the backlog */ 6392 if (sctp_sstate(sk, LISTENING)) 6393 sk->sk_max_ack_backlog = backlog; 6394 else { 6395 err = sctp_listen_start(sk, backlog); 6396 if (err) 6397 goto out; 6398 } 6399 6400 err = 0; 6401 out: 6402 release_sock(sk); 6403 return err; 6404 } 6405 6406 /* 6407 * This function is done by modeling the current datagram_poll() and the 6408 * tcp_poll(). Note that, based on these implementations, we don't 6409 * lock the socket in this function, even though it seems that, 6410 * ideally, locking or some other mechanisms can be used to ensure 6411 * the integrity of the counters (sndbuf and wmem_alloc) used 6412 * in this place. We assume that we don't need locks either until proven 6413 * otherwise. 6414 * 6415 * Another thing to note is that we include the Async I/O support 6416 * here, again, by modeling the current TCP/UDP code. We don't have 6417 * a good way to test with it yet. 6418 */ 6419 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait) 6420 { 6421 struct sock *sk = sock->sk; 6422 struct sctp_sock *sp = sctp_sk(sk); 6423 unsigned int mask; 6424 6425 poll_wait(file, sk_sleep(sk), wait); 6426 6427 /* A TCP-style listening socket becomes readable when the accept queue 6428 * is not empty. 6429 */ 6430 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) 6431 return (!list_empty(&sp->ep->asocs)) ? 6432 (POLLIN | POLLRDNORM) : 0; 6433 6434 mask = 0; 6435 6436 /* Is there any exceptional events? */ 6437 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) 6438 mask |= POLLERR | 6439 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0); 6440 if (sk->sk_shutdown & RCV_SHUTDOWN) 6441 mask |= POLLRDHUP | POLLIN | POLLRDNORM; 6442 if (sk->sk_shutdown == SHUTDOWN_MASK) 6443 mask |= POLLHUP; 6444 6445 /* Is it readable? Reconsider this code with TCP-style support. */ 6446 if (!skb_queue_empty(&sk->sk_receive_queue)) 6447 mask |= POLLIN | POLLRDNORM; 6448 6449 /* The association is either gone or not ready. */ 6450 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED)) 6451 return mask; 6452 6453 /* Is it writable? */ 6454 if (sctp_writeable(sk)) { 6455 mask |= POLLOUT | POLLWRNORM; 6456 } else { 6457 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk); 6458 /* 6459 * Since the socket is not locked, the buffer 6460 * might be made available after the writeable check and 6461 * before the bit is set. This could cause a lost I/O 6462 * signal. tcp_poll() has a race breaker for this race 6463 * condition. Based on their implementation, we put 6464 * in the following code to cover it as well. 6465 */ 6466 if (sctp_writeable(sk)) 6467 mask |= POLLOUT | POLLWRNORM; 6468 } 6469 return mask; 6470 } 6471 6472 /******************************************************************** 6473 * 2nd Level Abstractions 6474 ********************************************************************/ 6475 6476 static struct sctp_bind_bucket *sctp_bucket_create( 6477 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum) 6478 { 6479 struct sctp_bind_bucket *pp; 6480 6481 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC); 6482 if (pp) { 6483 SCTP_DBG_OBJCNT_INC(bind_bucket); 6484 pp->port = snum; 6485 pp->fastreuse = 0; 6486 INIT_HLIST_HEAD(&pp->owner); 6487 pp->net = net; 6488 hlist_add_head(&pp->node, &head->chain); 6489 } 6490 return pp; 6491 } 6492 6493 /* Caller must hold hashbucket lock for this tb with local BH disabled */ 6494 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp) 6495 { 6496 if (pp && hlist_empty(&pp->owner)) { 6497 __hlist_del(&pp->node); 6498 kmem_cache_free(sctp_bucket_cachep, pp); 6499 SCTP_DBG_OBJCNT_DEC(bind_bucket); 6500 } 6501 } 6502 6503 /* Release this socket's reference to a local port. */ 6504 static inline void __sctp_put_port(struct sock *sk) 6505 { 6506 struct sctp_bind_hashbucket *head = 6507 &sctp_port_hashtable[sctp_phashfn(sock_net(sk), 6508 inet_sk(sk)->inet_num)]; 6509 struct sctp_bind_bucket *pp; 6510 6511 spin_lock(&head->lock); 6512 pp = sctp_sk(sk)->bind_hash; 6513 __sk_del_bind_node(sk); 6514 sctp_sk(sk)->bind_hash = NULL; 6515 inet_sk(sk)->inet_num = 0; 6516 sctp_bucket_destroy(pp); 6517 spin_unlock(&head->lock); 6518 } 6519 6520 void sctp_put_port(struct sock *sk) 6521 { 6522 local_bh_disable(); 6523 __sctp_put_port(sk); 6524 local_bh_enable(); 6525 } 6526 6527 /* 6528 * The system picks an ephemeral port and choose an address set equivalent 6529 * to binding with a wildcard address. 6530 * One of those addresses will be the primary address for the association. 6531 * This automatically enables the multihoming capability of SCTP. 6532 */ 6533 static int sctp_autobind(struct sock *sk) 6534 { 6535 union sctp_addr autoaddr; 6536 struct sctp_af *af; 6537 __be16 port; 6538 6539 /* Initialize a local sockaddr structure to INADDR_ANY. */ 6540 af = sctp_sk(sk)->pf->af; 6541 6542 port = htons(inet_sk(sk)->inet_num); 6543 af->inaddr_any(&autoaddr, port); 6544 6545 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len); 6546 } 6547 6548 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation. 6549 * 6550 * From RFC 2292 6551 * 4.2 The cmsghdr Structure * 6552 * 6553 * When ancillary data is sent or received, any number of ancillary data 6554 * objects can be specified by the msg_control and msg_controllen members of 6555 * the msghdr structure, because each object is preceded by 6556 * a cmsghdr structure defining the object's length (the cmsg_len member). 6557 * Historically Berkeley-derived implementations have passed only one object 6558 * at a time, but this API allows multiple objects to be 6559 * passed in a single call to sendmsg() or recvmsg(). The following example 6560 * shows two ancillary data objects in a control buffer. 6561 * 6562 * |<--------------------------- msg_controllen -------------------------->| 6563 * | | 6564 * 6565 * |<----- ancillary data object ----->|<----- ancillary data object ----->| 6566 * 6567 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->| 6568 * | | | 6569 * 6570 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| | 6571 * 6572 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| | 6573 * | | | | | 6574 * 6575 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+ 6576 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX| 6577 * 6578 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX| 6579 * 6580 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+ 6581 * ^ 6582 * | 6583 * 6584 * msg_control 6585 * points here 6586 */ 6587 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs) 6588 { 6589 struct cmsghdr *cmsg; 6590 struct msghdr *my_msg = (struct msghdr *)msg; 6591 6592 for_each_cmsghdr(cmsg, my_msg) { 6593 if (!CMSG_OK(my_msg, cmsg)) 6594 return -EINVAL; 6595 6596 /* Should we parse this header or ignore? */ 6597 if (cmsg->cmsg_level != IPPROTO_SCTP) 6598 continue; 6599 6600 /* Strictly check lengths following example in SCM code. */ 6601 switch (cmsg->cmsg_type) { 6602 case SCTP_INIT: 6603 /* SCTP Socket API Extension 6604 * 5.3.1 SCTP Initiation Structure (SCTP_INIT) 6605 * 6606 * This cmsghdr structure provides information for 6607 * initializing new SCTP associations with sendmsg(). 6608 * The SCTP_INITMSG socket option uses this same data 6609 * structure. This structure is not used for 6610 * recvmsg(). 6611 * 6612 * cmsg_level cmsg_type cmsg_data[] 6613 * ------------ ------------ ---------------------- 6614 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg 6615 */ 6616 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg))) 6617 return -EINVAL; 6618 6619 cmsgs->init = CMSG_DATA(cmsg); 6620 break; 6621 6622 case SCTP_SNDRCV: 6623 /* SCTP Socket API Extension 6624 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV) 6625 * 6626 * This cmsghdr structure specifies SCTP options for 6627 * sendmsg() and describes SCTP header information 6628 * about a received message through recvmsg(). 6629 * 6630 * cmsg_level cmsg_type cmsg_data[] 6631 * ------------ ------------ ---------------------- 6632 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo 6633 */ 6634 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo))) 6635 return -EINVAL; 6636 6637 cmsgs->srinfo = CMSG_DATA(cmsg); 6638 6639 if (cmsgs->srinfo->sinfo_flags & 6640 ~(SCTP_UNORDERED | SCTP_ADDR_OVER | 6641 SCTP_ABORT | SCTP_EOF)) 6642 return -EINVAL; 6643 break; 6644 6645 case SCTP_SNDINFO: 6646 /* SCTP Socket API Extension 6647 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO) 6648 * 6649 * This cmsghdr structure specifies SCTP options for 6650 * sendmsg(). This structure and SCTP_RCVINFO replaces 6651 * SCTP_SNDRCV which has been deprecated. 6652 * 6653 * cmsg_level cmsg_type cmsg_data[] 6654 * ------------ ------------ --------------------- 6655 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo 6656 */ 6657 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo))) 6658 return -EINVAL; 6659 6660 cmsgs->sinfo = CMSG_DATA(cmsg); 6661 6662 if (cmsgs->sinfo->snd_flags & 6663 ~(SCTP_UNORDERED | SCTP_ADDR_OVER | 6664 SCTP_ABORT | SCTP_EOF)) 6665 return -EINVAL; 6666 break; 6667 default: 6668 return -EINVAL; 6669 } 6670 } 6671 6672 return 0; 6673 } 6674 6675 /* 6676 * Wait for a packet.. 6677 * Note: This function is the same function as in core/datagram.c 6678 * with a few modifications to make lksctp work. 6679 */ 6680 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p) 6681 { 6682 int error; 6683 DEFINE_WAIT(wait); 6684 6685 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 6686 6687 /* Socket errors? */ 6688 error = sock_error(sk); 6689 if (error) 6690 goto out; 6691 6692 if (!skb_queue_empty(&sk->sk_receive_queue)) 6693 goto ready; 6694 6695 /* Socket shut down? */ 6696 if (sk->sk_shutdown & RCV_SHUTDOWN) 6697 goto out; 6698 6699 /* Sequenced packets can come disconnected. If so we report the 6700 * problem. 6701 */ 6702 error = -ENOTCONN; 6703 6704 /* Is there a good reason to think that we may receive some data? */ 6705 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING)) 6706 goto out; 6707 6708 /* Handle signals. */ 6709 if (signal_pending(current)) 6710 goto interrupted; 6711 6712 /* Let another process have a go. Since we are going to sleep 6713 * anyway. Note: This may cause odd behaviors if the message 6714 * does not fit in the user's buffer, but this seems to be the 6715 * only way to honor MSG_DONTWAIT realistically. 6716 */ 6717 release_sock(sk); 6718 *timeo_p = schedule_timeout(*timeo_p); 6719 lock_sock(sk); 6720 6721 ready: 6722 finish_wait(sk_sleep(sk), &wait); 6723 return 0; 6724 6725 interrupted: 6726 error = sock_intr_errno(*timeo_p); 6727 6728 out: 6729 finish_wait(sk_sleep(sk), &wait); 6730 *err = error; 6731 return error; 6732 } 6733 6734 /* Receive a datagram. 6735 * Note: This is pretty much the same routine as in core/datagram.c 6736 * with a few changes to make lksctp work. 6737 */ 6738 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags, 6739 int noblock, int *err) 6740 { 6741 int error; 6742 struct sk_buff *skb; 6743 long timeo; 6744 6745 timeo = sock_rcvtimeo(sk, noblock); 6746 6747 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo, 6748 MAX_SCHEDULE_TIMEOUT); 6749 6750 do { 6751 /* Again only user level code calls this function, 6752 * so nothing interrupt level 6753 * will suddenly eat the receive_queue. 6754 * 6755 * Look at current nfs client by the way... 6756 * However, this function was correct in any case. 8) 6757 */ 6758 if (flags & MSG_PEEK) { 6759 spin_lock_bh(&sk->sk_receive_queue.lock); 6760 skb = skb_peek(&sk->sk_receive_queue); 6761 if (skb) 6762 atomic_inc(&skb->users); 6763 spin_unlock_bh(&sk->sk_receive_queue.lock); 6764 } else { 6765 skb = skb_dequeue(&sk->sk_receive_queue); 6766 } 6767 6768 if (skb) 6769 return skb; 6770 6771 /* Caller is allowed not to check sk->sk_err before calling. */ 6772 error = sock_error(sk); 6773 if (error) 6774 goto no_packet; 6775 6776 if (sk->sk_shutdown & RCV_SHUTDOWN) 6777 break; 6778 6779 if (sk_can_busy_loop(sk) && 6780 sk_busy_loop(sk, noblock)) 6781 continue; 6782 6783 /* User doesn't want to wait. */ 6784 error = -EAGAIN; 6785 if (!timeo) 6786 goto no_packet; 6787 } while (sctp_wait_for_packet(sk, err, &timeo) == 0); 6788 6789 return NULL; 6790 6791 no_packet: 6792 *err = error; 6793 return NULL; 6794 } 6795 6796 /* If sndbuf has changed, wake up per association sndbuf waiters. */ 6797 static void __sctp_write_space(struct sctp_association *asoc) 6798 { 6799 struct sock *sk = asoc->base.sk; 6800 6801 if (sctp_wspace(asoc) <= 0) 6802 return; 6803 6804 if (waitqueue_active(&asoc->wait)) 6805 wake_up_interruptible(&asoc->wait); 6806 6807 if (sctp_writeable(sk)) { 6808 struct socket_wq *wq; 6809 6810 rcu_read_lock(); 6811 wq = rcu_dereference(sk->sk_wq); 6812 if (wq) { 6813 if (waitqueue_active(&wq->wait)) 6814 wake_up_interruptible(&wq->wait); 6815 6816 /* Note that we try to include the Async I/O support 6817 * here by modeling from the current TCP/UDP code. 6818 * We have not tested with it yet. 6819 */ 6820 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) 6821 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT); 6822 } 6823 rcu_read_unlock(); 6824 } 6825 } 6826 6827 static void sctp_wake_up_waiters(struct sock *sk, 6828 struct sctp_association *asoc) 6829 { 6830 struct sctp_association *tmp = asoc; 6831 6832 /* We do accounting for the sndbuf space per association, 6833 * so we only need to wake our own association. 6834 */ 6835 if (asoc->ep->sndbuf_policy) 6836 return __sctp_write_space(asoc); 6837 6838 /* If association goes down and is just flushing its 6839 * outq, then just normally notify others. 6840 */ 6841 if (asoc->base.dead) 6842 return sctp_write_space(sk); 6843 6844 /* Accounting for the sndbuf space is per socket, so we 6845 * need to wake up others, try to be fair and in case of 6846 * other associations, let them have a go first instead 6847 * of just doing a sctp_write_space() call. 6848 * 6849 * Note that we reach sctp_wake_up_waiters() only when 6850 * associations free up queued chunks, thus we are under 6851 * lock and the list of associations on a socket is 6852 * guaranteed not to change. 6853 */ 6854 for (tmp = list_next_entry(tmp, asocs); 1; 6855 tmp = list_next_entry(tmp, asocs)) { 6856 /* Manually skip the head element. */ 6857 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs)) 6858 continue; 6859 /* Wake up association. */ 6860 __sctp_write_space(tmp); 6861 /* We've reached the end. */ 6862 if (tmp == asoc) 6863 break; 6864 } 6865 } 6866 6867 /* Do accounting for the sndbuf space. 6868 * Decrement the used sndbuf space of the corresponding association by the 6869 * data size which was just transmitted(freed). 6870 */ 6871 static void sctp_wfree(struct sk_buff *skb) 6872 { 6873 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg; 6874 struct sctp_association *asoc = chunk->asoc; 6875 struct sock *sk = asoc->base.sk; 6876 6877 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) + 6878 sizeof(struct sk_buff) + 6879 sizeof(struct sctp_chunk); 6880 6881 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc); 6882 6883 /* 6884 * This undoes what is done via sctp_set_owner_w and sk_mem_charge 6885 */ 6886 sk->sk_wmem_queued -= skb->truesize; 6887 sk_mem_uncharge(sk, skb->truesize); 6888 6889 sock_wfree(skb); 6890 sctp_wake_up_waiters(sk, asoc); 6891 6892 sctp_association_put(asoc); 6893 } 6894 6895 /* Do accounting for the receive space on the socket. 6896 * Accounting for the association is done in ulpevent.c 6897 * We set this as a destructor for the cloned data skbs so that 6898 * accounting is done at the correct time. 6899 */ 6900 void sctp_sock_rfree(struct sk_buff *skb) 6901 { 6902 struct sock *sk = skb->sk; 6903 struct sctp_ulpevent *event = sctp_skb2event(skb); 6904 6905 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc); 6906 6907 /* 6908 * Mimic the behavior of sock_rfree 6909 */ 6910 sk_mem_uncharge(sk, event->rmem_len); 6911 } 6912 6913 6914 /* Helper function to wait for space in the sndbuf. */ 6915 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p, 6916 size_t msg_len) 6917 { 6918 struct sock *sk = asoc->base.sk; 6919 int err = 0; 6920 long current_timeo = *timeo_p; 6921 DEFINE_WAIT(wait); 6922 6923 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc, 6924 *timeo_p, msg_len); 6925 6926 /* Increment the association's refcnt. */ 6927 sctp_association_hold(asoc); 6928 6929 /* Wait on the association specific sndbuf space. */ 6930 for (;;) { 6931 prepare_to_wait_exclusive(&asoc->wait, &wait, 6932 TASK_INTERRUPTIBLE); 6933 if (!*timeo_p) 6934 goto do_nonblock; 6935 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING || 6936 asoc->base.dead) 6937 goto do_error; 6938 if (signal_pending(current)) 6939 goto do_interrupted; 6940 if (msg_len <= sctp_wspace(asoc)) 6941 break; 6942 6943 /* Let another process have a go. Since we are going 6944 * to sleep anyway. 6945 */ 6946 release_sock(sk); 6947 current_timeo = schedule_timeout(current_timeo); 6948 BUG_ON(sk != asoc->base.sk); 6949 lock_sock(sk); 6950 6951 *timeo_p = current_timeo; 6952 } 6953 6954 out: 6955 finish_wait(&asoc->wait, &wait); 6956 6957 /* Release the association's refcnt. */ 6958 sctp_association_put(asoc); 6959 6960 return err; 6961 6962 do_error: 6963 err = -EPIPE; 6964 goto out; 6965 6966 do_interrupted: 6967 err = sock_intr_errno(*timeo_p); 6968 goto out; 6969 6970 do_nonblock: 6971 err = -EAGAIN; 6972 goto out; 6973 } 6974 6975 void sctp_data_ready(struct sock *sk) 6976 { 6977 struct socket_wq *wq; 6978 6979 rcu_read_lock(); 6980 wq = rcu_dereference(sk->sk_wq); 6981 if (wq_has_sleeper(wq)) 6982 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | 6983 POLLRDNORM | POLLRDBAND); 6984 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN); 6985 rcu_read_unlock(); 6986 } 6987 6988 /* If socket sndbuf has changed, wake up all per association waiters. */ 6989 void sctp_write_space(struct sock *sk) 6990 { 6991 struct sctp_association *asoc; 6992 6993 /* Wake up the tasks in each wait queue. */ 6994 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) { 6995 __sctp_write_space(asoc); 6996 } 6997 } 6998 6999 /* Is there any sndbuf space available on the socket? 7000 * 7001 * Note that sk_wmem_alloc is the sum of the send buffers on all of the 7002 * associations on the same socket. For a UDP-style socket with 7003 * multiple associations, it is possible for it to be "unwriteable" 7004 * prematurely. I assume that this is acceptable because 7005 * a premature "unwriteable" is better than an accidental "writeable" which 7006 * would cause an unwanted block under certain circumstances. For the 1-1 7007 * UDP-style sockets or TCP-style sockets, this code should work. 7008 * - Daisy 7009 */ 7010 static int sctp_writeable(struct sock *sk) 7011 { 7012 int amt = 0; 7013 7014 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk); 7015 if (amt < 0) 7016 amt = 0; 7017 return amt; 7018 } 7019 7020 /* Wait for an association to go into ESTABLISHED state. If timeout is 0, 7021 * returns immediately with EINPROGRESS. 7022 */ 7023 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p) 7024 { 7025 struct sock *sk = asoc->base.sk; 7026 int err = 0; 7027 long current_timeo = *timeo_p; 7028 DEFINE_WAIT(wait); 7029 7030 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p); 7031 7032 /* Increment the association's refcnt. */ 7033 sctp_association_hold(asoc); 7034 7035 for (;;) { 7036 prepare_to_wait_exclusive(&asoc->wait, &wait, 7037 TASK_INTERRUPTIBLE); 7038 if (!*timeo_p) 7039 goto do_nonblock; 7040 if (sk->sk_shutdown & RCV_SHUTDOWN) 7041 break; 7042 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING || 7043 asoc->base.dead) 7044 goto do_error; 7045 if (signal_pending(current)) 7046 goto do_interrupted; 7047 7048 if (sctp_state(asoc, ESTABLISHED)) 7049 break; 7050 7051 /* Let another process have a go. Since we are going 7052 * to sleep anyway. 7053 */ 7054 release_sock(sk); 7055 current_timeo = schedule_timeout(current_timeo); 7056 lock_sock(sk); 7057 7058 *timeo_p = current_timeo; 7059 } 7060 7061 out: 7062 finish_wait(&asoc->wait, &wait); 7063 7064 /* Release the association's refcnt. */ 7065 sctp_association_put(asoc); 7066 7067 return err; 7068 7069 do_error: 7070 if (asoc->init_err_counter + 1 > asoc->max_init_attempts) 7071 err = -ETIMEDOUT; 7072 else 7073 err = -ECONNREFUSED; 7074 goto out; 7075 7076 do_interrupted: 7077 err = sock_intr_errno(*timeo_p); 7078 goto out; 7079 7080 do_nonblock: 7081 err = -EINPROGRESS; 7082 goto out; 7083 } 7084 7085 static int sctp_wait_for_accept(struct sock *sk, long timeo) 7086 { 7087 struct sctp_endpoint *ep; 7088 int err = 0; 7089 DEFINE_WAIT(wait); 7090 7091 ep = sctp_sk(sk)->ep; 7092 7093 7094 for (;;) { 7095 prepare_to_wait_exclusive(sk_sleep(sk), &wait, 7096 TASK_INTERRUPTIBLE); 7097 7098 if (list_empty(&ep->asocs)) { 7099 release_sock(sk); 7100 timeo = schedule_timeout(timeo); 7101 lock_sock(sk); 7102 } 7103 7104 err = -EINVAL; 7105 if (!sctp_sstate(sk, LISTENING)) 7106 break; 7107 7108 err = 0; 7109 if (!list_empty(&ep->asocs)) 7110 break; 7111 7112 err = sock_intr_errno(timeo); 7113 if (signal_pending(current)) 7114 break; 7115 7116 err = -EAGAIN; 7117 if (!timeo) 7118 break; 7119 } 7120 7121 finish_wait(sk_sleep(sk), &wait); 7122 7123 return err; 7124 } 7125 7126 static void sctp_wait_for_close(struct sock *sk, long timeout) 7127 { 7128 DEFINE_WAIT(wait); 7129 7130 do { 7131 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); 7132 if (list_empty(&sctp_sk(sk)->ep->asocs)) 7133 break; 7134 release_sock(sk); 7135 timeout = schedule_timeout(timeout); 7136 lock_sock(sk); 7137 } while (!signal_pending(current) && timeout); 7138 7139 finish_wait(sk_sleep(sk), &wait); 7140 } 7141 7142 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk) 7143 { 7144 struct sk_buff *frag; 7145 7146 if (!skb->data_len) 7147 goto done; 7148 7149 /* Don't forget the fragments. */ 7150 skb_walk_frags(skb, frag) 7151 sctp_skb_set_owner_r_frag(frag, sk); 7152 7153 done: 7154 sctp_skb_set_owner_r(skb, sk); 7155 } 7156 7157 void sctp_copy_sock(struct sock *newsk, struct sock *sk, 7158 struct sctp_association *asoc) 7159 { 7160 struct inet_sock *inet = inet_sk(sk); 7161 struct inet_sock *newinet; 7162 7163 newsk->sk_type = sk->sk_type; 7164 newsk->sk_bound_dev_if = sk->sk_bound_dev_if; 7165 newsk->sk_flags = sk->sk_flags; 7166 newsk->sk_tsflags = sk->sk_tsflags; 7167 newsk->sk_no_check_tx = sk->sk_no_check_tx; 7168 newsk->sk_no_check_rx = sk->sk_no_check_rx; 7169 newsk->sk_reuse = sk->sk_reuse; 7170 7171 newsk->sk_shutdown = sk->sk_shutdown; 7172 newsk->sk_destruct = sctp_destruct_sock; 7173 newsk->sk_family = sk->sk_family; 7174 newsk->sk_protocol = IPPROTO_SCTP; 7175 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; 7176 newsk->sk_sndbuf = sk->sk_sndbuf; 7177 newsk->sk_rcvbuf = sk->sk_rcvbuf; 7178 newsk->sk_lingertime = sk->sk_lingertime; 7179 newsk->sk_rcvtimeo = sk->sk_rcvtimeo; 7180 newsk->sk_sndtimeo = sk->sk_sndtimeo; 7181 7182 newinet = inet_sk(newsk); 7183 7184 /* Initialize sk's sport, dport, rcv_saddr and daddr for 7185 * getsockname() and getpeername() 7186 */ 7187 newinet->inet_sport = inet->inet_sport; 7188 newinet->inet_saddr = inet->inet_saddr; 7189 newinet->inet_rcv_saddr = inet->inet_rcv_saddr; 7190 newinet->inet_dport = htons(asoc->peer.port); 7191 newinet->pmtudisc = inet->pmtudisc; 7192 newinet->inet_id = asoc->next_tsn ^ jiffies; 7193 7194 newinet->uc_ttl = inet->uc_ttl; 7195 newinet->mc_loop = 1; 7196 newinet->mc_ttl = 1; 7197 newinet->mc_index = 0; 7198 newinet->mc_list = NULL; 7199 7200 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP) 7201 net_enable_timestamp(); 7202 } 7203 7204 static inline void sctp_copy_descendant(struct sock *sk_to, 7205 const struct sock *sk_from) 7206 { 7207 int ancestor_size = sizeof(struct inet_sock) + 7208 sizeof(struct sctp_sock) - 7209 offsetof(struct sctp_sock, auto_asconf_list); 7210 7211 if (sk_from->sk_family == PF_INET6) 7212 ancestor_size += sizeof(struct ipv6_pinfo); 7213 7214 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size); 7215 } 7216 7217 /* Populate the fields of the newsk from the oldsk and migrate the assoc 7218 * and its messages to the newsk. 7219 */ 7220 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk, 7221 struct sctp_association *assoc, 7222 sctp_socket_type_t type) 7223 { 7224 struct sctp_sock *oldsp = sctp_sk(oldsk); 7225 struct sctp_sock *newsp = sctp_sk(newsk); 7226 struct sctp_bind_bucket *pp; /* hash list port iterator */ 7227 struct sctp_endpoint *newep = newsp->ep; 7228 struct sk_buff *skb, *tmp; 7229 struct sctp_ulpevent *event; 7230 struct sctp_bind_hashbucket *head; 7231 7232 /* Migrate socket buffer sizes and all the socket level options to the 7233 * new socket. 7234 */ 7235 newsk->sk_sndbuf = oldsk->sk_sndbuf; 7236 newsk->sk_rcvbuf = oldsk->sk_rcvbuf; 7237 /* Brute force copy old sctp opt. */ 7238 sctp_copy_descendant(newsk, oldsk); 7239 7240 /* Restore the ep value that was overwritten with the above structure 7241 * copy. 7242 */ 7243 newsp->ep = newep; 7244 newsp->hmac = NULL; 7245 7246 /* Hook this new socket in to the bind_hash list. */ 7247 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk), 7248 inet_sk(oldsk)->inet_num)]; 7249 local_bh_disable(); 7250 spin_lock(&head->lock); 7251 pp = sctp_sk(oldsk)->bind_hash; 7252 sk_add_bind_node(newsk, &pp->owner); 7253 sctp_sk(newsk)->bind_hash = pp; 7254 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num; 7255 spin_unlock(&head->lock); 7256 local_bh_enable(); 7257 7258 /* Copy the bind_addr list from the original endpoint to the new 7259 * endpoint so that we can handle restarts properly 7260 */ 7261 sctp_bind_addr_dup(&newsp->ep->base.bind_addr, 7262 &oldsp->ep->base.bind_addr, GFP_KERNEL); 7263 7264 /* Move any messages in the old socket's receive queue that are for the 7265 * peeled off association to the new socket's receive queue. 7266 */ 7267 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) { 7268 event = sctp_skb2event(skb); 7269 if (event->asoc == assoc) { 7270 __skb_unlink(skb, &oldsk->sk_receive_queue); 7271 __skb_queue_tail(&newsk->sk_receive_queue, skb); 7272 sctp_skb_set_owner_r_frag(skb, newsk); 7273 } 7274 } 7275 7276 /* Clean up any messages pending delivery due to partial 7277 * delivery. Three cases: 7278 * 1) No partial deliver; no work. 7279 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby. 7280 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue. 7281 */ 7282 skb_queue_head_init(&newsp->pd_lobby); 7283 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode); 7284 7285 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) { 7286 struct sk_buff_head *queue; 7287 7288 /* Decide which queue to move pd_lobby skbs to. */ 7289 if (assoc->ulpq.pd_mode) { 7290 queue = &newsp->pd_lobby; 7291 } else 7292 queue = &newsk->sk_receive_queue; 7293 7294 /* Walk through the pd_lobby, looking for skbs that 7295 * need moved to the new socket. 7296 */ 7297 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) { 7298 event = sctp_skb2event(skb); 7299 if (event->asoc == assoc) { 7300 __skb_unlink(skb, &oldsp->pd_lobby); 7301 __skb_queue_tail(queue, skb); 7302 sctp_skb_set_owner_r_frag(skb, newsk); 7303 } 7304 } 7305 7306 /* Clear up any skbs waiting for the partial 7307 * delivery to finish. 7308 */ 7309 if (assoc->ulpq.pd_mode) 7310 sctp_clear_pd(oldsk, NULL); 7311 7312 } 7313 7314 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) 7315 sctp_skb_set_owner_r_frag(skb, newsk); 7316 7317 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) 7318 sctp_skb_set_owner_r_frag(skb, newsk); 7319 7320 /* Set the type of socket to indicate that it is peeled off from the 7321 * original UDP-style socket or created with the accept() call on a 7322 * TCP-style socket.. 7323 */ 7324 newsp->type = type; 7325 7326 /* Mark the new socket "in-use" by the user so that any packets 7327 * that may arrive on the association after we've moved it are 7328 * queued to the backlog. This prevents a potential race between 7329 * backlog processing on the old socket and new-packet processing 7330 * on the new socket. 7331 * 7332 * The caller has just allocated newsk so we can guarantee that other 7333 * paths won't try to lock it and then oldsk. 7334 */ 7335 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING); 7336 sctp_assoc_migrate(assoc, newsk); 7337 7338 /* If the association on the newsk is already closed before accept() 7339 * is called, set RCV_SHUTDOWN flag. 7340 */ 7341 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) 7342 newsk->sk_shutdown |= RCV_SHUTDOWN; 7343 7344 newsk->sk_state = SCTP_SS_ESTABLISHED; 7345 release_sock(newsk); 7346 } 7347 7348 7349 /* This proto struct describes the ULP interface for SCTP. */ 7350 struct proto sctp_prot = { 7351 .name = "SCTP", 7352 .owner = THIS_MODULE, 7353 .close = sctp_close, 7354 .connect = sctp_connect, 7355 .disconnect = sctp_disconnect, 7356 .accept = sctp_accept, 7357 .ioctl = sctp_ioctl, 7358 .init = sctp_init_sock, 7359 .destroy = sctp_destroy_sock, 7360 .shutdown = sctp_shutdown, 7361 .setsockopt = sctp_setsockopt, 7362 .getsockopt = sctp_getsockopt, 7363 .sendmsg = sctp_sendmsg, 7364 .recvmsg = sctp_recvmsg, 7365 .bind = sctp_bind, 7366 .backlog_rcv = sctp_backlog_rcv, 7367 .hash = sctp_hash, 7368 .unhash = sctp_unhash, 7369 .get_port = sctp_get_port, 7370 .obj_size = sizeof(struct sctp_sock), 7371 .sysctl_mem = sysctl_sctp_mem, 7372 .sysctl_rmem = sysctl_sctp_rmem, 7373 .sysctl_wmem = sysctl_sctp_wmem, 7374 .memory_pressure = &sctp_memory_pressure, 7375 .enter_memory_pressure = sctp_enter_memory_pressure, 7376 .memory_allocated = &sctp_memory_allocated, 7377 .sockets_allocated = &sctp_sockets_allocated, 7378 }; 7379 7380 #if IS_ENABLED(CONFIG_IPV6) 7381 7382 #include <net/transp_v6.h> 7383 static void sctp_v6_destroy_sock(struct sock *sk) 7384 { 7385 sctp_destroy_sock(sk); 7386 inet6_destroy_sock(sk); 7387 } 7388 7389 struct proto sctpv6_prot = { 7390 .name = "SCTPv6", 7391 .owner = THIS_MODULE, 7392 .close = sctp_close, 7393 .connect = sctp_connect, 7394 .disconnect = sctp_disconnect, 7395 .accept = sctp_accept, 7396 .ioctl = sctp_ioctl, 7397 .init = sctp_init_sock, 7398 .destroy = sctp_v6_destroy_sock, 7399 .shutdown = sctp_shutdown, 7400 .setsockopt = sctp_setsockopt, 7401 .getsockopt = sctp_getsockopt, 7402 .sendmsg = sctp_sendmsg, 7403 .recvmsg = sctp_recvmsg, 7404 .bind = sctp_bind, 7405 .backlog_rcv = sctp_backlog_rcv, 7406 .hash = sctp_hash, 7407 .unhash = sctp_unhash, 7408 .get_port = sctp_get_port, 7409 .obj_size = sizeof(struct sctp6_sock), 7410 .sysctl_mem = sysctl_sctp_mem, 7411 .sysctl_rmem = sysctl_sctp_rmem, 7412 .sysctl_wmem = sysctl_sctp_wmem, 7413 .memory_pressure = &sctp_memory_pressure, 7414 .enter_memory_pressure = sctp_enter_memory_pressure, 7415 .memory_allocated = &sctp_memory_allocated, 7416 .sockets_allocated = &sctp_sockets_allocated, 7417 }; 7418 #endif /* IS_ENABLED(CONFIG_IPV6) */ 7419