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_association_free(asoc); 1232 } 1233 return err; 1234 } 1235 1236 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt() 1237 * 1238 * API 8.9 1239 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt, 1240 * sctp_assoc_t *asoc); 1241 * 1242 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses. 1243 * If the sd is an IPv6 socket, the addresses passed can either be IPv4 1244 * or IPv6 addresses. 1245 * 1246 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see 1247 * Section 3.1.2 for this usage. 1248 * 1249 * addrs is a pointer to an array of one or more socket addresses. Each 1250 * address is contained in its appropriate structure (i.e. struct 1251 * sockaddr_in or struct sockaddr_in6) the family of the address type 1252 * must be used to distengish the address length (note that this 1253 * representation is termed a "packed array" of addresses). The caller 1254 * specifies the number of addresses in the array with addrcnt. 1255 * 1256 * On success, sctp_connectx() returns 0. It also sets the assoc_id to 1257 * the association id of the new association. On failure, sctp_connectx() 1258 * returns -1, and sets errno to the appropriate error code. The assoc_id 1259 * is not touched by the kernel. 1260 * 1261 * For SCTP, the port given in each socket address must be the same, or 1262 * sctp_connectx() will fail, setting errno to EINVAL. 1263 * 1264 * An application can use sctp_connectx to initiate an association with 1265 * an endpoint that is multi-homed. Much like sctp_bindx() this call 1266 * allows a caller to specify multiple addresses at which a peer can be 1267 * reached. The way the SCTP stack uses the list of addresses to set up 1268 * the association is implementation dependent. This function only 1269 * specifies that the stack will try to make use of all the addresses in 1270 * the list when needed. 1271 * 1272 * Note that the list of addresses passed in is only used for setting up 1273 * the association. It does not necessarily equal the set of addresses 1274 * the peer uses for the resulting association. If the caller wants to 1275 * find out the set of peer addresses, it must use sctp_getpaddrs() to 1276 * retrieve them after the association has been set up. 1277 * 1278 * Basically do nothing but copying the addresses from user to kernel 1279 * land and invoking either sctp_connectx(). This is used for tunneling 1280 * the sctp_connectx() request through sctp_setsockopt() from userspace. 1281 * 1282 * We don't use copy_from_user() for optimization: we first do the 1283 * sanity checks (buffer size -fast- and access check-healthy 1284 * pointer); if all of those succeed, then we can alloc the memory 1285 * (expensive operation) needed to copy the data to kernel. Then we do 1286 * the copying without checking the user space area 1287 * (__copy_from_user()). 1288 * 1289 * On exit there is no need to do sockfd_put(), sys_setsockopt() does 1290 * it. 1291 * 1292 * sk The sk of the socket 1293 * addrs The pointer to the addresses in user land 1294 * addrssize Size of the addrs buffer 1295 * 1296 * Returns >=0 if ok, <0 errno code on error. 1297 */ 1298 static int __sctp_setsockopt_connectx(struct sock *sk, 1299 struct sockaddr __user *addrs, 1300 int addrs_size, 1301 sctp_assoc_t *assoc_id) 1302 { 1303 struct sockaddr *kaddrs; 1304 gfp_t gfp = GFP_KERNEL; 1305 int err = 0; 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 if (sk->sk_socket->file) 1319 gfp = GFP_USER | __GFP_NOWARN; 1320 kaddrs = kmalloc(addrs_size, gfp); 1321 if (unlikely(!kaddrs)) 1322 return -ENOMEM; 1323 1324 if (__copy_from_user(kaddrs, addrs, addrs_size)) { 1325 err = -EFAULT; 1326 } else { 1327 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id); 1328 } 1329 1330 kfree(kaddrs); 1331 1332 return err; 1333 } 1334 1335 /* 1336 * This is an older interface. It's kept for backward compatibility 1337 * to the option that doesn't provide association id. 1338 */ 1339 static int sctp_setsockopt_connectx_old(struct sock *sk, 1340 struct sockaddr __user *addrs, 1341 int addrs_size) 1342 { 1343 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL); 1344 } 1345 1346 /* 1347 * New interface for the API. The since the API is done with a socket 1348 * option, to make it simple we feed back the association id is as a return 1349 * indication to the call. Error is always negative and association id is 1350 * always positive. 1351 */ 1352 static int sctp_setsockopt_connectx(struct sock *sk, 1353 struct sockaddr __user *addrs, 1354 int addrs_size) 1355 { 1356 sctp_assoc_t assoc_id = 0; 1357 int err = 0; 1358 1359 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id); 1360 1361 if (err) 1362 return err; 1363 else 1364 return assoc_id; 1365 } 1366 1367 /* 1368 * New (hopefully final) interface for the API. 1369 * We use the sctp_getaddrs_old structure so that use-space library 1370 * can avoid any unnecessary allocations. The only different part 1371 * is that we store the actual length of the address buffer into the 1372 * addrs_num structure member. That way we can re-use the existing 1373 * code. 1374 */ 1375 #ifdef CONFIG_COMPAT 1376 struct compat_sctp_getaddrs_old { 1377 sctp_assoc_t assoc_id; 1378 s32 addr_num; 1379 compat_uptr_t addrs; /* struct sockaddr * */ 1380 }; 1381 #endif 1382 1383 static int sctp_getsockopt_connectx3(struct sock *sk, int len, 1384 char __user *optval, 1385 int __user *optlen) 1386 { 1387 struct sctp_getaddrs_old param; 1388 sctp_assoc_t assoc_id = 0; 1389 int err = 0; 1390 1391 #ifdef CONFIG_COMPAT 1392 if (is_compat_task()) { 1393 struct compat_sctp_getaddrs_old param32; 1394 1395 if (len < sizeof(param32)) 1396 return -EINVAL; 1397 if (copy_from_user(¶m32, optval, sizeof(param32))) 1398 return -EFAULT; 1399 1400 param.assoc_id = param32.assoc_id; 1401 param.addr_num = param32.addr_num; 1402 param.addrs = compat_ptr(param32.addrs); 1403 } else 1404 #endif 1405 { 1406 if (len < sizeof(param)) 1407 return -EINVAL; 1408 if (copy_from_user(¶m, optval, sizeof(param))) 1409 return -EFAULT; 1410 } 1411 1412 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *) 1413 param.addrs, param.addr_num, 1414 &assoc_id); 1415 if (err == 0 || err == -EINPROGRESS) { 1416 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id))) 1417 return -EFAULT; 1418 if (put_user(sizeof(assoc_id), optlen)) 1419 return -EFAULT; 1420 } 1421 1422 return err; 1423 } 1424 1425 /* API 3.1.4 close() - UDP Style Syntax 1426 * Applications use close() to perform graceful shutdown (as described in 1427 * Section 10.1 of [SCTP]) on ALL the associations currently represented 1428 * by a UDP-style socket. 1429 * 1430 * The syntax is 1431 * 1432 * ret = close(int sd); 1433 * 1434 * sd - the socket descriptor of the associations to be closed. 1435 * 1436 * To gracefully shutdown a specific association represented by the 1437 * UDP-style socket, an application should use the sendmsg() call, 1438 * passing no user data, but including the appropriate flag in the 1439 * ancillary data (see Section xxxx). 1440 * 1441 * If sd in the close() call is a branched-off socket representing only 1442 * one association, the shutdown is performed on that association only. 1443 * 1444 * 4.1.6 close() - TCP Style Syntax 1445 * 1446 * Applications use close() to gracefully close down an association. 1447 * 1448 * The syntax is: 1449 * 1450 * int close(int sd); 1451 * 1452 * sd - the socket descriptor of the association to be closed. 1453 * 1454 * After an application calls close() on a socket descriptor, no further 1455 * socket operations will succeed on that descriptor. 1456 * 1457 * API 7.1.4 SO_LINGER 1458 * 1459 * An application using the TCP-style socket can use this option to 1460 * perform the SCTP ABORT primitive. The linger option structure is: 1461 * 1462 * struct linger { 1463 * int l_onoff; // option on/off 1464 * int l_linger; // linger time 1465 * }; 1466 * 1467 * To enable the option, set l_onoff to 1. If the l_linger value is set 1468 * to 0, calling close() is the same as the ABORT primitive. If the 1469 * value is set to a negative value, the setsockopt() call will return 1470 * an error. If the value is set to a positive value linger_time, the 1471 * close() can be blocked for at most linger_time ms. If the graceful 1472 * shutdown phase does not finish during this period, close() will 1473 * return but the graceful shutdown phase continues in the system. 1474 */ 1475 static void sctp_close(struct sock *sk, long timeout) 1476 { 1477 struct net *net = sock_net(sk); 1478 struct sctp_endpoint *ep; 1479 struct sctp_association *asoc; 1480 struct list_head *pos, *temp; 1481 unsigned int data_was_unread; 1482 1483 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout); 1484 1485 lock_sock(sk); 1486 sk->sk_shutdown = SHUTDOWN_MASK; 1487 sk->sk_state = SCTP_SS_CLOSING; 1488 1489 ep = sctp_sk(sk)->ep; 1490 1491 /* Clean up any skbs sitting on the receive queue. */ 1492 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue); 1493 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby); 1494 1495 /* Walk all associations on an endpoint. */ 1496 list_for_each_safe(pos, temp, &ep->asocs) { 1497 asoc = list_entry(pos, struct sctp_association, asocs); 1498 1499 if (sctp_style(sk, TCP)) { 1500 /* A closed association can still be in the list if 1501 * it belongs to a TCP-style listening socket that is 1502 * not yet accepted. If so, free it. If not, send an 1503 * ABORT or SHUTDOWN based on the linger options. 1504 */ 1505 if (sctp_state(asoc, CLOSED)) { 1506 sctp_association_free(asoc); 1507 continue; 1508 } 1509 } 1510 1511 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) || 1512 !skb_queue_empty(&asoc->ulpq.reasm) || 1513 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) { 1514 struct sctp_chunk *chunk; 1515 1516 chunk = sctp_make_abort_user(asoc, NULL, 0); 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_association_free(asoc); 1989 out_unlock: 1990 release_sock(sk); 1991 1992 out_nounlock: 1993 return sctp_error(sk, msg_flags, err); 1994 1995 #if 0 1996 do_sock_err: 1997 if (msg_len) 1998 err = msg_len; 1999 else 2000 err = sock_error(sk); 2001 goto out; 2002 2003 do_interrupted: 2004 if (msg_len) 2005 err = msg_len; 2006 goto out; 2007 #endif /* 0 */ 2008 } 2009 2010 /* This is an extended version of skb_pull() that removes the data from the 2011 * start of a skb even when data is spread across the list of skb's in the 2012 * frag_list. len specifies the total amount of data that needs to be removed. 2013 * when 'len' bytes could be removed from the skb, it returns 0. 2014 * If 'len' exceeds the total skb length, it returns the no. of bytes that 2015 * could not be removed. 2016 */ 2017 static int sctp_skb_pull(struct sk_buff *skb, int len) 2018 { 2019 struct sk_buff *list; 2020 int skb_len = skb_headlen(skb); 2021 int rlen; 2022 2023 if (len <= skb_len) { 2024 __skb_pull(skb, len); 2025 return 0; 2026 } 2027 len -= skb_len; 2028 __skb_pull(skb, skb_len); 2029 2030 skb_walk_frags(skb, list) { 2031 rlen = sctp_skb_pull(list, len); 2032 skb->len -= (len-rlen); 2033 skb->data_len -= (len-rlen); 2034 2035 if (!rlen) 2036 return 0; 2037 2038 len = rlen; 2039 } 2040 2041 return len; 2042 } 2043 2044 /* API 3.1.3 recvmsg() - UDP Style Syntax 2045 * 2046 * ssize_t recvmsg(int socket, struct msghdr *message, 2047 * int flags); 2048 * 2049 * socket - the socket descriptor of the endpoint. 2050 * message - pointer to the msghdr structure which contains a single 2051 * user message and possibly some ancillary data. 2052 * 2053 * See Section 5 for complete description of the data 2054 * structures. 2055 * 2056 * flags - flags sent or received with the user message, see Section 2057 * 5 for complete description of the flags. 2058 */ 2059 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 2060 int noblock, int flags, int *addr_len) 2061 { 2062 struct sctp_ulpevent *event = NULL; 2063 struct sctp_sock *sp = sctp_sk(sk); 2064 struct sk_buff *skb; 2065 int copied; 2066 int err = 0; 2067 int skb_len; 2068 2069 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, " 2070 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags, 2071 addr_len); 2072 2073 lock_sock(sk); 2074 2075 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) { 2076 err = -ENOTCONN; 2077 goto out; 2078 } 2079 2080 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err); 2081 if (!skb) 2082 goto out; 2083 2084 /* Get the total length of the skb including any skb's in the 2085 * frag_list. 2086 */ 2087 skb_len = skb->len; 2088 2089 copied = skb_len; 2090 if (copied > len) 2091 copied = len; 2092 2093 err = skb_copy_datagram_msg(skb, 0, msg, copied); 2094 2095 event = sctp_skb2event(skb); 2096 2097 if (err) 2098 goto out_free; 2099 2100 sock_recv_ts_and_drops(msg, sk, skb); 2101 if (sctp_ulpevent_is_notification(event)) { 2102 msg->msg_flags |= MSG_NOTIFICATION; 2103 sp->pf->event_msgname(event, msg->msg_name, addr_len); 2104 } else { 2105 sp->pf->skb_msgname(skb, msg->msg_name, addr_len); 2106 } 2107 2108 /* Check if we allow SCTP_NXTINFO. */ 2109 if (sp->recvnxtinfo) 2110 sctp_ulpevent_read_nxtinfo(event, msg, sk); 2111 /* Check if we allow SCTP_RCVINFO. */ 2112 if (sp->recvrcvinfo) 2113 sctp_ulpevent_read_rcvinfo(event, msg); 2114 /* Check if we allow SCTP_SNDRCVINFO. */ 2115 if (sp->subscribe.sctp_data_io_event) 2116 sctp_ulpevent_read_sndrcvinfo(event, msg); 2117 2118 err = copied; 2119 2120 /* If skb's length exceeds the user's buffer, update the skb and 2121 * push it back to the receive_queue so that the next call to 2122 * recvmsg() will return the remaining data. Don't set MSG_EOR. 2123 */ 2124 if (skb_len > copied) { 2125 msg->msg_flags &= ~MSG_EOR; 2126 if (flags & MSG_PEEK) 2127 goto out_free; 2128 sctp_skb_pull(skb, copied); 2129 skb_queue_head(&sk->sk_receive_queue, skb); 2130 2131 /* When only partial message is copied to the user, increase 2132 * rwnd by that amount. If all the data in the skb is read, 2133 * rwnd is updated when the event is freed. 2134 */ 2135 if (!sctp_ulpevent_is_notification(event)) 2136 sctp_assoc_rwnd_increase(event->asoc, copied); 2137 goto out; 2138 } else if ((event->msg_flags & MSG_NOTIFICATION) || 2139 (event->msg_flags & MSG_EOR)) 2140 msg->msg_flags |= MSG_EOR; 2141 else 2142 msg->msg_flags &= ~MSG_EOR; 2143 2144 out_free: 2145 if (flags & MSG_PEEK) { 2146 /* Release the skb reference acquired after peeking the skb in 2147 * sctp_skb_recv_datagram(). 2148 */ 2149 kfree_skb(skb); 2150 } else { 2151 /* Free the event which includes releasing the reference to 2152 * the owner of the skb, freeing the skb and updating the 2153 * rwnd. 2154 */ 2155 sctp_ulpevent_free(event); 2156 } 2157 out: 2158 release_sock(sk); 2159 return err; 2160 } 2161 2162 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS) 2163 * 2164 * This option is a on/off flag. If enabled no SCTP message 2165 * fragmentation will be performed. Instead if a message being sent 2166 * exceeds the current PMTU size, the message will NOT be sent and 2167 * instead a error will be indicated to the user. 2168 */ 2169 static int sctp_setsockopt_disable_fragments(struct sock *sk, 2170 char __user *optval, 2171 unsigned int optlen) 2172 { 2173 int val; 2174 2175 if (optlen < sizeof(int)) 2176 return -EINVAL; 2177 2178 if (get_user(val, (int __user *)optval)) 2179 return -EFAULT; 2180 2181 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1; 2182 2183 return 0; 2184 } 2185 2186 static int sctp_setsockopt_events(struct sock *sk, char __user *optval, 2187 unsigned int optlen) 2188 { 2189 struct sctp_association *asoc; 2190 struct sctp_ulpevent *event; 2191 2192 if (optlen > sizeof(struct sctp_event_subscribe)) 2193 return -EINVAL; 2194 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen)) 2195 return -EFAULT; 2196 2197 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT, 2198 * if there is no data to be sent or retransmit, the stack will 2199 * immediately send up this notification. 2200 */ 2201 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT, 2202 &sctp_sk(sk)->subscribe)) { 2203 asoc = sctp_id2assoc(sk, 0); 2204 2205 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) { 2206 event = sctp_ulpevent_make_sender_dry_event(asoc, 2207 GFP_ATOMIC); 2208 if (!event) 2209 return -ENOMEM; 2210 2211 sctp_ulpq_tail_event(&asoc->ulpq, event); 2212 } 2213 } 2214 2215 return 0; 2216 } 2217 2218 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE) 2219 * 2220 * This socket option is applicable to the UDP-style socket only. When 2221 * set it will cause associations that are idle for more than the 2222 * specified number of seconds to automatically close. An association 2223 * being idle is defined an association that has NOT sent or received 2224 * user data. The special value of '0' indicates that no automatic 2225 * close of any associations should be performed. The option expects an 2226 * integer defining the number of seconds of idle time before an 2227 * association is closed. 2228 */ 2229 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval, 2230 unsigned int optlen) 2231 { 2232 struct sctp_sock *sp = sctp_sk(sk); 2233 struct net *net = sock_net(sk); 2234 2235 /* Applicable to UDP-style socket only */ 2236 if (sctp_style(sk, TCP)) 2237 return -EOPNOTSUPP; 2238 if (optlen != sizeof(int)) 2239 return -EINVAL; 2240 if (copy_from_user(&sp->autoclose, optval, optlen)) 2241 return -EFAULT; 2242 2243 if (sp->autoclose > net->sctp.max_autoclose) 2244 sp->autoclose = net->sctp.max_autoclose; 2245 2246 return 0; 2247 } 2248 2249 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS) 2250 * 2251 * Applications can enable or disable heartbeats for any peer address of 2252 * an association, modify an address's heartbeat interval, force a 2253 * heartbeat to be sent immediately, and adjust the address's maximum 2254 * number of retransmissions sent before an address is considered 2255 * unreachable. The following structure is used to access and modify an 2256 * address's parameters: 2257 * 2258 * struct sctp_paddrparams { 2259 * sctp_assoc_t spp_assoc_id; 2260 * struct sockaddr_storage spp_address; 2261 * uint32_t spp_hbinterval; 2262 * uint16_t spp_pathmaxrxt; 2263 * uint32_t spp_pathmtu; 2264 * uint32_t spp_sackdelay; 2265 * uint32_t spp_flags; 2266 * }; 2267 * 2268 * spp_assoc_id - (one-to-many style socket) This is filled in the 2269 * application, and identifies the association for 2270 * this query. 2271 * spp_address - This specifies which address is of interest. 2272 * spp_hbinterval - This contains the value of the heartbeat interval, 2273 * in milliseconds. If a value of zero 2274 * is present in this field then no changes are to 2275 * be made to this parameter. 2276 * spp_pathmaxrxt - This contains the maximum number of 2277 * retransmissions before this address shall be 2278 * considered unreachable. If a value of zero 2279 * is present in this field then no changes are to 2280 * be made to this parameter. 2281 * spp_pathmtu - When Path MTU discovery is disabled the value 2282 * specified here will be the "fixed" path mtu. 2283 * Note that if the spp_address field is empty 2284 * then all associations on this address will 2285 * have this fixed path mtu set upon them. 2286 * 2287 * spp_sackdelay - When delayed sack is enabled, this value specifies 2288 * the number of milliseconds that sacks will be delayed 2289 * for. This value will apply to all addresses of an 2290 * association if the spp_address field is empty. Note 2291 * also, that if delayed sack is enabled and this 2292 * value is set to 0, no change is made to the last 2293 * recorded delayed sack timer value. 2294 * 2295 * spp_flags - These flags are used to control various features 2296 * on an association. The flag field may contain 2297 * zero or more of the following options. 2298 * 2299 * SPP_HB_ENABLE - Enable heartbeats on the 2300 * specified address. Note that if the address 2301 * field is empty all addresses for the association 2302 * have heartbeats enabled upon them. 2303 * 2304 * SPP_HB_DISABLE - Disable heartbeats on the 2305 * speicifed address. Note that if the address 2306 * field is empty all addresses for the association 2307 * will have their heartbeats disabled. Note also 2308 * that SPP_HB_ENABLE and SPP_HB_DISABLE are 2309 * mutually exclusive, only one of these two should 2310 * be specified. Enabling both fields will have 2311 * undetermined results. 2312 * 2313 * SPP_HB_DEMAND - Request a user initiated heartbeat 2314 * to be made immediately. 2315 * 2316 * SPP_HB_TIME_IS_ZERO - Specify's that the time for 2317 * heartbeat delayis to be set to the value of 0 2318 * milliseconds. 2319 * 2320 * SPP_PMTUD_ENABLE - This field will enable PMTU 2321 * discovery upon the specified address. Note that 2322 * if the address feild is empty then all addresses 2323 * on the association are effected. 2324 * 2325 * SPP_PMTUD_DISABLE - This field will disable PMTU 2326 * discovery upon the specified address. Note that 2327 * if the address feild is empty then all addresses 2328 * on the association are effected. Not also that 2329 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually 2330 * exclusive. Enabling both will have undetermined 2331 * results. 2332 * 2333 * SPP_SACKDELAY_ENABLE - Setting this flag turns 2334 * on delayed sack. The time specified in spp_sackdelay 2335 * is used to specify the sack delay for this address. Note 2336 * that if spp_address is empty then all addresses will 2337 * enable delayed sack and take on the sack delay 2338 * value specified in spp_sackdelay. 2339 * SPP_SACKDELAY_DISABLE - Setting this flag turns 2340 * off delayed sack. If the spp_address field is blank then 2341 * delayed sack is disabled for the entire association. Note 2342 * also that this field is mutually exclusive to 2343 * SPP_SACKDELAY_ENABLE, setting both will have undefined 2344 * results. 2345 */ 2346 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params, 2347 struct sctp_transport *trans, 2348 struct sctp_association *asoc, 2349 struct sctp_sock *sp, 2350 int hb_change, 2351 int pmtud_change, 2352 int sackdelay_change) 2353 { 2354 int error; 2355 2356 if (params->spp_flags & SPP_HB_DEMAND && trans) { 2357 struct net *net = sock_net(trans->asoc->base.sk); 2358 2359 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans); 2360 if (error) 2361 return error; 2362 } 2363 2364 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of 2365 * this field is ignored. Note also that a value of zero indicates 2366 * the current setting should be left unchanged. 2367 */ 2368 if (params->spp_flags & SPP_HB_ENABLE) { 2369 2370 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is 2371 * set. This lets us use 0 value when this flag 2372 * is set. 2373 */ 2374 if (params->spp_flags & SPP_HB_TIME_IS_ZERO) 2375 params->spp_hbinterval = 0; 2376 2377 if (params->spp_hbinterval || 2378 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) { 2379 if (trans) { 2380 trans->hbinterval = 2381 msecs_to_jiffies(params->spp_hbinterval); 2382 } else if (asoc) { 2383 asoc->hbinterval = 2384 msecs_to_jiffies(params->spp_hbinterval); 2385 } else { 2386 sp->hbinterval = params->spp_hbinterval; 2387 } 2388 } 2389 } 2390 2391 if (hb_change) { 2392 if (trans) { 2393 trans->param_flags = 2394 (trans->param_flags & ~SPP_HB) | hb_change; 2395 } else if (asoc) { 2396 asoc->param_flags = 2397 (asoc->param_flags & ~SPP_HB) | hb_change; 2398 } else { 2399 sp->param_flags = 2400 (sp->param_flags & ~SPP_HB) | hb_change; 2401 } 2402 } 2403 2404 /* When Path MTU discovery is disabled the value specified here will 2405 * be the "fixed" path mtu (i.e. the value of the spp_flags field must 2406 * include the flag SPP_PMTUD_DISABLE for this field to have any 2407 * effect). 2408 */ 2409 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) { 2410 if (trans) { 2411 trans->pathmtu = params->spp_pathmtu; 2412 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc); 2413 } else if (asoc) { 2414 asoc->pathmtu = params->spp_pathmtu; 2415 sctp_frag_point(asoc, params->spp_pathmtu); 2416 } else { 2417 sp->pathmtu = params->spp_pathmtu; 2418 } 2419 } 2420 2421 if (pmtud_change) { 2422 if (trans) { 2423 int update = (trans->param_flags & SPP_PMTUD_DISABLE) && 2424 (params->spp_flags & SPP_PMTUD_ENABLE); 2425 trans->param_flags = 2426 (trans->param_flags & ~SPP_PMTUD) | pmtud_change; 2427 if (update) { 2428 sctp_transport_pmtu(trans, sctp_opt2sk(sp)); 2429 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc); 2430 } 2431 } else if (asoc) { 2432 asoc->param_flags = 2433 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change; 2434 } else { 2435 sp->param_flags = 2436 (sp->param_flags & ~SPP_PMTUD) | pmtud_change; 2437 } 2438 } 2439 2440 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the 2441 * value of this field is ignored. Note also that a value of zero 2442 * indicates the current setting should be left unchanged. 2443 */ 2444 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) { 2445 if (trans) { 2446 trans->sackdelay = 2447 msecs_to_jiffies(params->spp_sackdelay); 2448 } else if (asoc) { 2449 asoc->sackdelay = 2450 msecs_to_jiffies(params->spp_sackdelay); 2451 } else { 2452 sp->sackdelay = params->spp_sackdelay; 2453 } 2454 } 2455 2456 if (sackdelay_change) { 2457 if (trans) { 2458 trans->param_flags = 2459 (trans->param_flags & ~SPP_SACKDELAY) | 2460 sackdelay_change; 2461 } else if (asoc) { 2462 asoc->param_flags = 2463 (asoc->param_flags & ~SPP_SACKDELAY) | 2464 sackdelay_change; 2465 } else { 2466 sp->param_flags = 2467 (sp->param_flags & ~SPP_SACKDELAY) | 2468 sackdelay_change; 2469 } 2470 } 2471 2472 /* Note that a value of zero indicates the current setting should be 2473 left unchanged. 2474 */ 2475 if (params->spp_pathmaxrxt) { 2476 if (trans) { 2477 trans->pathmaxrxt = params->spp_pathmaxrxt; 2478 } else if (asoc) { 2479 asoc->pathmaxrxt = params->spp_pathmaxrxt; 2480 } else { 2481 sp->pathmaxrxt = params->spp_pathmaxrxt; 2482 } 2483 } 2484 2485 return 0; 2486 } 2487 2488 static int sctp_setsockopt_peer_addr_params(struct sock *sk, 2489 char __user *optval, 2490 unsigned int optlen) 2491 { 2492 struct sctp_paddrparams params; 2493 struct sctp_transport *trans = NULL; 2494 struct sctp_association *asoc = NULL; 2495 struct sctp_sock *sp = sctp_sk(sk); 2496 int error; 2497 int hb_change, pmtud_change, sackdelay_change; 2498 2499 if (optlen != sizeof(struct sctp_paddrparams)) 2500 return -EINVAL; 2501 2502 if (copy_from_user(¶ms, optval, optlen)) 2503 return -EFAULT; 2504 2505 /* Validate flags and value parameters. */ 2506 hb_change = params.spp_flags & SPP_HB; 2507 pmtud_change = params.spp_flags & SPP_PMTUD; 2508 sackdelay_change = params.spp_flags & SPP_SACKDELAY; 2509 2510 if (hb_change == SPP_HB || 2511 pmtud_change == SPP_PMTUD || 2512 sackdelay_change == SPP_SACKDELAY || 2513 params.spp_sackdelay > 500 || 2514 (params.spp_pathmtu && 2515 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT)) 2516 return -EINVAL; 2517 2518 /* If an address other than INADDR_ANY is specified, and 2519 * no transport is found, then the request is invalid. 2520 */ 2521 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) { 2522 trans = sctp_addr_id2transport(sk, ¶ms.spp_address, 2523 params.spp_assoc_id); 2524 if (!trans) 2525 return -EINVAL; 2526 } 2527 2528 /* Get association, if assoc_id != 0 and the socket is a one 2529 * to many style socket, and an association was not found, then 2530 * the id was invalid. 2531 */ 2532 asoc = sctp_id2assoc(sk, params.spp_assoc_id); 2533 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) 2534 return -EINVAL; 2535 2536 /* Heartbeat demand can only be sent on a transport or 2537 * association, but not a socket. 2538 */ 2539 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc) 2540 return -EINVAL; 2541 2542 /* Process parameters. */ 2543 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp, 2544 hb_change, pmtud_change, 2545 sackdelay_change); 2546 2547 if (error) 2548 return error; 2549 2550 /* If changes are for association, also apply parameters to each 2551 * transport. 2552 */ 2553 if (!trans && asoc) { 2554 list_for_each_entry(trans, &asoc->peer.transport_addr_list, 2555 transports) { 2556 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp, 2557 hb_change, pmtud_change, 2558 sackdelay_change); 2559 } 2560 } 2561 2562 return 0; 2563 } 2564 2565 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags) 2566 { 2567 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE; 2568 } 2569 2570 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags) 2571 { 2572 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE; 2573 } 2574 2575 /* 2576 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK) 2577 * 2578 * This option will effect the way delayed acks are performed. This 2579 * option allows you to get or set the delayed ack time, in 2580 * milliseconds. It also allows changing the delayed ack frequency. 2581 * Changing the frequency to 1 disables the delayed sack algorithm. If 2582 * the assoc_id is 0, then this sets or gets the endpoints default 2583 * values. If the assoc_id field is non-zero, then the set or get 2584 * effects the specified association for the one to many model (the 2585 * assoc_id field is ignored by the one to one model). Note that if 2586 * sack_delay or sack_freq are 0 when setting this option, then the 2587 * current values will remain unchanged. 2588 * 2589 * struct sctp_sack_info { 2590 * sctp_assoc_t sack_assoc_id; 2591 * uint32_t sack_delay; 2592 * uint32_t sack_freq; 2593 * }; 2594 * 2595 * sack_assoc_id - This parameter, indicates which association the user 2596 * is performing an action upon. Note that if this field's value is 2597 * zero then the endpoints default value is changed (effecting future 2598 * associations only). 2599 * 2600 * sack_delay - This parameter contains the number of milliseconds that 2601 * the user is requesting the delayed ACK timer be set to. Note that 2602 * this value is defined in the standard to be between 200 and 500 2603 * milliseconds. 2604 * 2605 * sack_freq - This parameter contains the number of packets that must 2606 * be received before a sack is sent without waiting for the delay 2607 * timer to expire. The default value for this is 2, setting this 2608 * value to 1 will disable the delayed sack algorithm. 2609 */ 2610 2611 static int sctp_setsockopt_delayed_ack(struct sock *sk, 2612 char __user *optval, unsigned int optlen) 2613 { 2614 struct sctp_sack_info params; 2615 struct sctp_transport *trans = NULL; 2616 struct sctp_association *asoc = NULL; 2617 struct sctp_sock *sp = sctp_sk(sk); 2618 2619 if (optlen == sizeof(struct sctp_sack_info)) { 2620 if (copy_from_user(¶ms, optval, optlen)) 2621 return -EFAULT; 2622 2623 if (params.sack_delay == 0 && params.sack_freq == 0) 2624 return 0; 2625 } else if (optlen == sizeof(struct sctp_assoc_value)) { 2626 pr_warn_ratelimited(DEPRECATED 2627 "%s (pid %d) " 2628 "Use of struct sctp_assoc_value in delayed_ack socket option.\n" 2629 "Use struct sctp_sack_info instead\n", 2630 current->comm, task_pid_nr(current)); 2631 if (copy_from_user(¶ms, optval, optlen)) 2632 return -EFAULT; 2633 2634 if (params.sack_delay == 0) 2635 params.sack_freq = 1; 2636 else 2637 params.sack_freq = 0; 2638 } else 2639 return -EINVAL; 2640 2641 /* Validate value parameter. */ 2642 if (params.sack_delay > 500) 2643 return -EINVAL; 2644 2645 /* Get association, if sack_assoc_id != 0 and the socket is a one 2646 * to many style socket, and an association was not found, then 2647 * the id was invalid. 2648 */ 2649 asoc = sctp_id2assoc(sk, params.sack_assoc_id); 2650 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP)) 2651 return -EINVAL; 2652 2653 if (params.sack_delay) { 2654 if (asoc) { 2655 asoc->sackdelay = 2656 msecs_to_jiffies(params.sack_delay); 2657 asoc->param_flags = 2658 sctp_spp_sackdelay_enable(asoc->param_flags); 2659 } else { 2660 sp->sackdelay = params.sack_delay; 2661 sp->param_flags = 2662 sctp_spp_sackdelay_enable(sp->param_flags); 2663 } 2664 } 2665 2666 if (params.sack_freq == 1) { 2667 if (asoc) { 2668 asoc->param_flags = 2669 sctp_spp_sackdelay_disable(asoc->param_flags); 2670 } else { 2671 sp->param_flags = 2672 sctp_spp_sackdelay_disable(sp->param_flags); 2673 } 2674 } else if (params.sack_freq > 1) { 2675 if (asoc) { 2676 asoc->sackfreq = params.sack_freq; 2677 asoc->param_flags = 2678 sctp_spp_sackdelay_enable(asoc->param_flags); 2679 } else { 2680 sp->sackfreq = params.sack_freq; 2681 sp->param_flags = 2682 sctp_spp_sackdelay_enable(sp->param_flags); 2683 } 2684 } 2685 2686 /* If change is for association, also apply to each transport. */ 2687 if (asoc) { 2688 list_for_each_entry(trans, &asoc->peer.transport_addr_list, 2689 transports) { 2690 if (params.sack_delay) { 2691 trans->sackdelay = 2692 msecs_to_jiffies(params.sack_delay); 2693 trans->param_flags = 2694 sctp_spp_sackdelay_enable(trans->param_flags); 2695 } 2696 if (params.sack_freq == 1) { 2697 trans->param_flags = 2698 sctp_spp_sackdelay_disable(trans->param_flags); 2699 } else if (params.sack_freq > 1) { 2700 trans->sackfreq = params.sack_freq; 2701 trans->param_flags = 2702 sctp_spp_sackdelay_enable(trans->param_flags); 2703 } 2704 } 2705 } 2706 2707 return 0; 2708 } 2709 2710 /* 7.1.3 Initialization Parameters (SCTP_INITMSG) 2711 * 2712 * Applications can specify protocol parameters for the default association 2713 * initialization. The option name argument to setsockopt() and getsockopt() 2714 * is SCTP_INITMSG. 2715 * 2716 * Setting initialization parameters is effective only on an unconnected 2717 * socket (for UDP-style sockets only future associations are effected 2718 * by the change). With TCP-style sockets, this option is inherited by 2719 * sockets derived from a listener socket. 2720 */ 2721 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen) 2722 { 2723 struct sctp_initmsg sinit; 2724 struct sctp_sock *sp = sctp_sk(sk); 2725 2726 if (optlen != sizeof(struct sctp_initmsg)) 2727 return -EINVAL; 2728 if (copy_from_user(&sinit, optval, optlen)) 2729 return -EFAULT; 2730 2731 if (sinit.sinit_num_ostreams) 2732 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams; 2733 if (sinit.sinit_max_instreams) 2734 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams; 2735 if (sinit.sinit_max_attempts) 2736 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts; 2737 if (sinit.sinit_max_init_timeo) 2738 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo; 2739 2740 return 0; 2741 } 2742 2743 /* 2744 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM) 2745 * 2746 * Applications that wish to use the sendto() system call may wish to 2747 * specify a default set of parameters that would normally be supplied 2748 * through the inclusion of ancillary data. This socket option allows 2749 * such an application to set the default sctp_sndrcvinfo structure. 2750 * The application that wishes to use this socket option simply passes 2751 * in to this call the sctp_sndrcvinfo structure defined in Section 2752 * 5.2.2) The input parameters accepted by this call include 2753 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context, 2754 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in 2755 * to this call if the caller is using the UDP model. 2756 */ 2757 static int sctp_setsockopt_default_send_param(struct sock *sk, 2758 char __user *optval, 2759 unsigned int optlen) 2760 { 2761 struct sctp_sock *sp = sctp_sk(sk); 2762 struct sctp_association *asoc; 2763 struct sctp_sndrcvinfo info; 2764 2765 if (optlen != sizeof(info)) 2766 return -EINVAL; 2767 if (copy_from_user(&info, optval, optlen)) 2768 return -EFAULT; 2769 if (info.sinfo_flags & 2770 ~(SCTP_UNORDERED | SCTP_ADDR_OVER | 2771 SCTP_ABORT | SCTP_EOF)) 2772 return -EINVAL; 2773 2774 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id); 2775 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP)) 2776 return -EINVAL; 2777 if (asoc) { 2778 asoc->default_stream = info.sinfo_stream; 2779 asoc->default_flags = info.sinfo_flags; 2780 asoc->default_ppid = info.sinfo_ppid; 2781 asoc->default_context = info.sinfo_context; 2782 asoc->default_timetolive = info.sinfo_timetolive; 2783 } else { 2784 sp->default_stream = info.sinfo_stream; 2785 sp->default_flags = info.sinfo_flags; 2786 sp->default_ppid = info.sinfo_ppid; 2787 sp->default_context = info.sinfo_context; 2788 sp->default_timetolive = info.sinfo_timetolive; 2789 } 2790 2791 return 0; 2792 } 2793 2794 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters 2795 * (SCTP_DEFAULT_SNDINFO) 2796 */ 2797 static int sctp_setsockopt_default_sndinfo(struct sock *sk, 2798 char __user *optval, 2799 unsigned int optlen) 2800 { 2801 struct sctp_sock *sp = sctp_sk(sk); 2802 struct sctp_association *asoc; 2803 struct sctp_sndinfo info; 2804 2805 if (optlen != sizeof(info)) 2806 return -EINVAL; 2807 if (copy_from_user(&info, optval, optlen)) 2808 return -EFAULT; 2809 if (info.snd_flags & 2810 ~(SCTP_UNORDERED | SCTP_ADDR_OVER | 2811 SCTP_ABORT | SCTP_EOF)) 2812 return -EINVAL; 2813 2814 asoc = sctp_id2assoc(sk, info.snd_assoc_id); 2815 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP)) 2816 return -EINVAL; 2817 if (asoc) { 2818 asoc->default_stream = info.snd_sid; 2819 asoc->default_flags = info.snd_flags; 2820 asoc->default_ppid = info.snd_ppid; 2821 asoc->default_context = info.snd_context; 2822 } else { 2823 sp->default_stream = info.snd_sid; 2824 sp->default_flags = info.snd_flags; 2825 sp->default_ppid = info.snd_ppid; 2826 sp->default_context = info.snd_context; 2827 } 2828 2829 return 0; 2830 } 2831 2832 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR) 2833 * 2834 * Requests that the local SCTP stack use the enclosed peer address as 2835 * the association primary. The enclosed address must be one of the 2836 * association peer's addresses. 2837 */ 2838 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval, 2839 unsigned int optlen) 2840 { 2841 struct sctp_prim prim; 2842 struct sctp_transport *trans; 2843 2844 if (optlen != sizeof(struct sctp_prim)) 2845 return -EINVAL; 2846 2847 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim))) 2848 return -EFAULT; 2849 2850 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id); 2851 if (!trans) 2852 return -EINVAL; 2853 2854 sctp_assoc_set_primary(trans->asoc, trans); 2855 2856 return 0; 2857 } 2858 2859 /* 2860 * 7.1.5 SCTP_NODELAY 2861 * 2862 * Turn on/off any Nagle-like algorithm. This means that packets are 2863 * generally sent as soon as possible and no unnecessary delays are 2864 * introduced, at the cost of more packets in the network. Expects an 2865 * integer boolean flag. 2866 */ 2867 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval, 2868 unsigned int optlen) 2869 { 2870 int val; 2871 2872 if (optlen < sizeof(int)) 2873 return -EINVAL; 2874 if (get_user(val, (int __user *)optval)) 2875 return -EFAULT; 2876 2877 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1; 2878 return 0; 2879 } 2880 2881 /* 2882 * 2883 * 7.1.1 SCTP_RTOINFO 2884 * 2885 * The protocol parameters used to initialize and bound retransmission 2886 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access 2887 * and modify these parameters. 2888 * All parameters are time values, in milliseconds. A value of 0, when 2889 * modifying the parameters, indicates that the current value should not 2890 * be changed. 2891 * 2892 */ 2893 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen) 2894 { 2895 struct sctp_rtoinfo rtoinfo; 2896 struct sctp_association *asoc; 2897 unsigned long rto_min, rto_max; 2898 struct sctp_sock *sp = sctp_sk(sk); 2899 2900 if (optlen != sizeof (struct sctp_rtoinfo)) 2901 return -EINVAL; 2902 2903 if (copy_from_user(&rtoinfo, optval, optlen)) 2904 return -EFAULT; 2905 2906 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id); 2907 2908 /* Set the values to the specific association */ 2909 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP)) 2910 return -EINVAL; 2911 2912 rto_max = rtoinfo.srto_max; 2913 rto_min = rtoinfo.srto_min; 2914 2915 if (rto_max) 2916 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max; 2917 else 2918 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max; 2919 2920 if (rto_min) 2921 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min; 2922 else 2923 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min; 2924 2925 if (rto_min > rto_max) 2926 return -EINVAL; 2927 2928 if (asoc) { 2929 if (rtoinfo.srto_initial != 0) 2930 asoc->rto_initial = 2931 msecs_to_jiffies(rtoinfo.srto_initial); 2932 asoc->rto_max = rto_max; 2933 asoc->rto_min = rto_min; 2934 } else { 2935 /* If there is no association or the association-id = 0 2936 * set the values to the endpoint. 2937 */ 2938 if (rtoinfo.srto_initial != 0) 2939 sp->rtoinfo.srto_initial = rtoinfo.srto_initial; 2940 sp->rtoinfo.srto_max = rto_max; 2941 sp->rtoinfo.srto_min = rto_min; 2942 } 2943 2944 return 0; 2945 } 2946 2947 /* 2948 * 2949 * 7.1.2 SCTP_ASSOCINFO 2950 * 2951 * This option is used to tune the maximum retransmission attempts 2952 * of the association. 2953 * Returns an error if the new association retransmission value is 2954 * greater than the sum of the retransmission value of the peer. 2955 * See [SCTP] for more information. 2956 * 2957 */ 2958 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen) 2959 { 2960 2961 struct sctp_assocparams assocparams; 2962 struct sctp_association *asoc; 2963 2964 if (optlen != sizeof(struct sctp_assocparams)) 2965 return -EINVAL; 2966 if (copy_from_user(&assocparams, optval, optlen)) 2967 return -EFAULT; 2968 2969 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id); 2970 2971 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP)) 2972 return -EINVAL; 2973 2974 /* Set the values to the specific association */ 2975 if (asoc) { 2976 if (assocparams.sasoc_asocmaxrxt != 0) { 2977 __u32 path_sum = 0; 2978 int paths = 0; 2979 struct sctp_transport *peer_addr; 2980 2981 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list, 2982 transports) { 2983 path_sum += peer_addr->pathmaxrxt; 2984 paths++; 2985 } 2986 2987 /* Only validate asocmaxrxt if we have more than 2988 * one path/transport. We do this because path 2989 * retransmissions are only counted when we have more 2990 * then one path. 2991 */ 2992 if (paths > 1 && 2993 assocparams.sasoc_asocmaxrxt > path_sum) 2994 return -EINVAL; 2995 2996 asoc->max_retrans = assocparams.sasoc_asocmaxrxt; 2997 } 2998 2999 if (assocparams.sasoc_cookie_life != 0) 3000 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life); 3001 } else { 3002 /* Set the values to the endpoint */ 3003 struct sctp_sock *sp = sctp_sk(sk); 3004 3005 if (assocparams.sasoc_asocmaxrxt != 0) 3006 sp->assocparams.sasoc_asocmaxrxt = 3007 assocparams.sasoc_asocmaxrxt; 3008 if (assocparams.sasoc_cookie_life != 0) 3009 sp->assocparams.sasoc_cookie_life = 3010 assocparams.sasoc_cookie_life; 3011 } 3012 return 0; 3013 } 3014 3015 /* 3016 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR) 3017 * 3018 * This socket option is a boolean flag which turns on or off mapped V4 3019 * addresses. If this option is turned on and the socket is type 3020 * PF_INET6, then IPv4 addresses will be mapped to V6 representation. 3021 * If this option is turned off, then no mapping will be done of V4 3022 * addresses and a user will receive both PF_INET6 and PF_INET type 3023 * addresses on the socket. 3024 */ 3025 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen) 3026 { 3027 int val; 3028 struct sctp_sock *sp = sctp_sk(sk); 3029 3030 if (optlen < sizeof(int)) 3031 return -EINVAL; 3032 if (get_user(val, (int __user *)optval)) 3033 return -EFAULT; 3034 if (val) 3035 sp->v4mapped = 1; 3036 else 3037 sp->v4mapped = 0; 3038 3039 return 0; 3040 } 3041 3042 /* 3043 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG) 3044 * This option will get or set the maximum size to put in any outgoing 3045 * SCTP DATA chunk. If a message is larger than this size it will be 3046 * fragmented by SCTP into the specified size. Note that the underlying 3047 * SCTP implementation may fragment into smaller sized chunks when the 3048 * PMTU of the underlying association is smaller than the value set by 3049 * the user. The default value for this option is '0' which indicates 3050 * the user is NOT limiting fragmentation and only the PMTU will effect 3051 * SCTP's choice of DATA chunk size. Note also that values set larger 3052 * than the maximum size of an IP datagram will effectively let SCTP 3053 * control fragmentation (i.e. the same as setting this option to 0). 3054 * 3055 * The following structure is used to access and modify this parameter: 3056 * 3057 * struct sctp_assoc_value { 3058 * sctp_assoc_t assoc_id; 3059 * uint32_t assoc_value; 3060 * }; 3061 * 3062 * assoc_id: This parameter is ignored for one-to-one style sockets. 3063 * For one-to-many style sockets this parameter indicates which 3064 * association the user is performing an action upon. Note that if 3065 * this field's value is zero then the endpoints default value is 3066 * changed (effecting future associations only). 3067 * assoc_value: This parameter specifies the maximum size in bytes. 3068 */ 3069 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen) 3070 { 3071 struct sctp_assoc_value params; 3072 struct sctp_association *asoc; 3073 struct sctp_sock *sp = sctp_sk(sk); 3074 int val; 3075 3076 if (optlen == sizeof(int)) { 3077 pr_warn_ratelimited(DEPRECATED 3078 "%s (pid %d) " 3079 "Use of int in maxseg socket option.\n" 3080 "Use struct sctp_assoc_value instead\n", 3081 current->comm, task_pid_nr(current)); 3082 if (copy_from_user(&val, optval, optlen)) 3083 return -EFAULT; 3084 params.assoc_id = 0; 3085 } else if (optlen == sizeof(struct sctp_assoc_value)) { 3086 if (copy_from_user(¶ms, optval, optlen)) 3087 return -EFAULT; 3088 val = params.assoc_value; 3089 } else 3090 return -EINVAL; 3091 3092 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN))) 3093 return -EINVAL; 3094 3095 asoc = sctp_id2assoc(sk, params.assoc_id); 3096 if (!asoc && params.assoc_id && sctp_style(sk, UDP)) 3097 return -EINVAL; 3098 3099 if (asoc) { 3100 if (val == 0) { 3101 val = asoc->pathmtu; 3102 val -= sp->pf->af->net_header_len; 3103 val -= sizeof(struct sctphdr) + 3104 sizeof(struct sctp_data_chunk); 3105 } 3106 asoc->user_frag = val; 3107 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu); 3108 } else { 3109 sp->user_frag = val; 3110 } 3111 3112 return 0; 3113 } 3114 3115 3116 /* 3117 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR) 3118 * 3119 * Requests that the peer mark the enclosed address as the association 3120 * primary. The enclosed address must be one of the association's 3121 * locally bound addresses. The following structure is used to make a 3122 * set primary request: 3123 */ 3124 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval, 3125 unsigned int optlen) 3126 { 3127 struct net *net = sock_net(sk); 3128 struct sctp_sock *sp; 3129 struct sctp_association *asoc = NULL; 3130 struct sctp_setpeerprim prim; 3131 struct sctp_chunk *chunk; 3132 struct sctp_af *af; 3133 int err; 3134 3135 sp = sctp_sk(sk); 3136 3137 if (!net->sctp.addip_enable) 3138 return -EPERM; 3139 3140 if (optlen != sizeof(struct sctp_setpeerprim)) 3141 return -EINVAL; 3142 3143 if (copy_from_user(&prim, optval, optlen)) 3144 return -EFAULT; 3145 3146 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id); 3147 if (!asoc) 3148 return -EINVAL; 3149 3150 if (!asoc->peer.asconf_capable) 3151 return -EPERM; 3152 3153 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY) 3154 return -EPERM; 3155 3156 if (!sctp_state(asoc, ESTABLISHED)) 3157 return -ENOTCONN; 3158 3159 af = sctp_get_af_specific(prim.sspp_addr.ss_family); 3160 if (!af) 3161 return -EINVAL; 3162 3163 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL)) 3164 return -EADDRNOTAVAIL; 3165 3166 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr)) 3167 return -EADDRNOTAVAIL; 3168 3169 /* Create an ASCONF chunk with SET_PRIMARY parameter */ 3170 chunk = sctp_make_asconf_set_prim(asoc, 3171 (union sctp_addr *)&prim.sspp_addr); 3172 if (!chunk) 3173 return -ENOMEM; 3174 3175 err = sctp_send_asconf(asoc, chunk); 3176 3177 pr_debug("%s: we set peer primary addr primitively\n", __func__); 3178 3179 return err; 3180 } 3181 3182 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval, 3183 unsigned int optlen) 3184 { 3185 struct sctp_setadaptation adaptation; 3186 3187 if (optlen != sizeof(struct sctp_setadaptation)) 3188 return -EINVAL; 3189 if (copy_from_user(&adaptation, optval, optlen)) 3190 return -EFAULT; 3191 3192 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind; 3193 3194 return 0; 3195 } 3196 3197 /* 3198 * 7.1.29. Set or Get the default context (SCTP_CONTEXT) 3199 * 3200 * The context field in the sctp_sndrcvinfo structure is normally only 3201 * used when a failed message is retrieved holding the value that was 3202 * sent down on the actual send call. This option allows the setting of 3203 * a default context on an association basis that will be received on 3204 * reading messages from the peer. This is especially helpful in the 3205 * one-2-many model for an application to keep some reference to an 3206 * internal state machine that is processing messages on the 3207 * association. Note that the setting of this value only effects 3208 * received messages from the peer and does not effect the value that is 3209 * saved with outbound messages. 3210 */ 3211 static int sctp_setsockopt_context(struct sock *sk, char __user *optval, 3212 unsigned int optlen) 3213 { 3214 struct sctp_assoc_value params; 3215 struct sctp_sock *sp; 3216 struct sctp_association *asoc; 3217 3218 if (optlen != sizeof(struct sctp_assoc_value)) 3219 return -EINVAL; 3220 if (copy_from_user(¶ms, optval, optlen)) 3221 return -EFAULT; 3222 3223 sp = sctp_sk(sk); 3224 3225 if (params.assoc_id != 0) { 3226 asoc = sctp_id2assoc(sk, params.assoc_id); 3227 if (!asoc) 3228 return -EINVAL; 3229 asoc->default_rcv_context = params.assoc_value; 3230 } else { 3231 sp->default_rcv_context = params.assoc_value; 3232 } 3233 3234 return 0; 3235 } 3236 3237 /* 3238 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE) 3239 * 3240 * This options will at a minimum specify if the implementation is doing 3241 * fragmented interleave. Fragmented interleave, for a one to many 3242 * socket, is when subsequent calls to receive a message may return 3243 * parts of messages from different associations. Some implementations 3244 * may allow you to turn this value on or off. If so, when turned off, 3245 * no fragment interleave will occur (which will cause a head of line 3246 * blocking amongst multiple associations sharing the same one to many 3247 * socket). When this option is turned on, then each receive call may 3248 * come from a different association (thus the user must receive data 3249 * with the extended calls (e.g. sctp_recvmsg) to keep track of which 3250 * association each receive belongs to. 3251 * 3252 * This option takes a boolean value. A non-zero value indicates that 3253 * fragmented interleave is on. A value of zero indicates that 3254 * fragmented interleave is off. 3255 * 3256 * Note that it is important that an implementation that allows this 3257 * option to be turned on, have it off by default. Otherwise an unaware 3258 * application using the one to many model may become confused and act 3259 * incorrectly. 3260 */ 3261 static int sctp_setsockopt_fragment_interleave(struct sock *sk, 3262 char __user *optval, 3263 unsigned int optlen) 3264 { 3265 int val; 3266 3267 if (optlen != sizeof(int)) 3268 return -EINVAL; 3269 if (get_user(val, (int __user *)optval)) 3270 return -EFAULT; 3271 3272 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1; 3273 3274 return 0; 3275 } 3276 3277 /* 3278 * 8.1.21. Set or Get the SCTP Partial Delivery Point 3279 * (SCTP_PARTIAL_DELIVERY_POINT) 3280 * 3281 * This option will set or get the SCTP partial delivery point. This 3282 * point is the size of a message where the partial delivery API will be 3283 * invoked to help free up rwnd space for the peer. Setting this to a 3284 * lower value will cause partial deliveries to happen more often. The 3285 * calls argument is an integer that sets or gets the partial delivery 3286 * point. Note also that the call will fail if the user attempts to set 3287 * this value larger than the socket receive buffer size. 3288 * 3289 * Note that any single message having a length smaller than or equal to 3290 * the SCTP partial delivery point will be delivered in one single read 3291 * call as long as the user provided buffer is large enough to hold the 3292 * message. 3293 */ 3294 static int sctp_setsockopt_partial_delivery_point(struct sock *sk, 3295 char __user *optval, 3296 unsigned int optlen) 3297 { 3298 u32 val; 3299 3300 if (optlen != sizeof(u32)) 3301 return -EINVAL; 3302 if (get_user(val, (int __user *)optval)) 3303 return -EFAULT; 3304 3305 /* Note: We double the receive buffer from what the user sets 3306 * it to be, also initial rwnd is based on rcvbuf/2. 3307 */ 3308 if (val > (sk->sk_rcvbuf >> 1)) 3309 return -EINVAL; 3310 3311 sctp_sk(sk)->pd_point = val; 3312 3313 return 0; /* is this the right error code? */ 3314 } 3315 3316 /* 3317 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST) 3318 * 3319 * This option will allow a user to change the maximum burst of packets 3320 * that can be emitted by this association. Note that the default value 3321 * is 4, and some implementations may restrict this setting so that it 3322 * can only be lowered. 3323 * 3324 * NOTE: This text doesn't seem right. Do this on a socket basis with 3325 * future associations inheriting the socket value. 3326 */ 3327 static int sctp_setsockopt_maxburst(struct sock *sk, 3328 char __user *optval, 3329 unsigned int optlen) 3330 { 3331 struct sctp_assoc_value params; 3332 struct sctp_sock *sp; 3333 struct sctp_association *asoc; 3334 int val; 3335 int assoc_id = 0; 3336 3337 if (optlen == sizeof(int)) { 3338 pr_warn_ratelimited(DEPRECATED 3339 "%s (pid %d) " 3340 "Use of int in max_burst socket option deprecated.\n" 3341 "Use struct sctp_assoc_value instead\n", 3342 current->comm, task_pid_nr(current)); 3343 if (copy_from_user(&val, optval, optlen)) 3344 return -EFAULT; 3345 } else if (optlen == sizeof(struct sctp_assoc_value)) { 3346 if (copy_from_user(¶ms, optval, optlen)) 3347 return -EFAULT; 3348 val = params.assoc_value; 3349 assoc_id = params.assoc_id; 3350 } else 3351 return -EINVAL; 3352 3353 sp = sctp_sk(sk); 3354 3355 if (assoc_id != 0) { 3356 asoc = sctp_id2assoc(sk, assoc_id); 3357 if (!asoc) 3358 return -EINVAL; 3359 asoc->max_burst = val; 3360 } else 3361 sp->max_burst = val; 3362 3363 return 0; 3364 } 3365 3366 /* 3367 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK) 3368 * 3369 * This set option adds a chunk type that the user is requesting to be 3370 * received only in an authenticated way. Changes to the list of chunks 3371 * will only effect future associations on the socket. 3372 */ 3373 static int sctp_setsockopt_auth_chunk(struct sock *sk, 3374 char __user *optval, 3375 unsigned int optlen) 3376 { 3377 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3378 struct sctp_authchunk val; 3379 3380 if (!ep->auth_enable) 3381 return -EACCES; 3382 3383 if (optlen != sizeof(struct sctp_authchunk)) 3384 return -EINVAL; 3385 if (copy_from_user(&val, optval, optlen)) 3386 return -EFAULT; 3387 3388 switch (val.sauth_chunk) { 3389 case SCTP_CID_INIT: 3390 case SCTP_CID_INIT_ACK: 3391 case SCTP_CID_SHUTDOWN_COMPLETE: 3392 case SCTP_CID_AUTH: 3393 return -EINVAL; 3394 } 3395 3396 /* add this chunk id to the endpoint */ 3397 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk); 3398 } 3399 3400 /* 3401 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT) 3402 * 3403 * This option gets or sets the list of HMAC algorithms that the local 3404 * endpoint requires the peer to use. 3405 */ 3406 static int sctp_setsockopt_hmac_ident(struct sock *sk, 3407 char __user *optval, 3408 unsigned int optlen) 3409 { 3410 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3411 struct sctp_hmacalgo *hmacs; 3412 u32 idents; 3413 int err; 3414 3415 if (!ep->auth_enable) 3416 return -EACCES; 3417 3418 if (optlen < sizeof(struct sctp_hmacalgo)) 3419 return -EINVAL; 3420 3421 hmacs = memdup_user(optval, optlen); 3422 if (IS_ERR(hmacs)) 3423 return PTR_ERR(hmacs); 3424 3425 idents = hmacs->shmac_num_idents; 3426 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS || 3427 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) { 3428 err = -EINVAL; 3429 goto out; 3430 } 3431 3432 err = sctp_auth_ep_set_hmacs(ep, hmacs); 3433 out: 3434 kfree(hmacs); 3435 return err; 3436 } 3437 3438 /* 3439 * 7.1.20. Set a shared key (SCTP_AUTH_KEY) 3440 * 3441 * This option will set a shared secret key which is used to build an 3442 * association shared key. 3443 */ 3444 static int sctp_setsockopt_auth_key(struct sock *sk, 3445 char __user *optval, 3446 unsigned int optlen) 3447 { 3448 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3449 struct sctp_authkey *authkey; 3450 struct sctp_association *asoc; 3451 int ret; 3452 3453 if (!ep->auth_enable) 3454 return -EACCES; 3455 3456 if (optlen <= sizeof(struct sctp_authkey)) 3457 return -EINVAL; 3458 3459 authkey = memdup_user(optval, optlen); 3460 if (IS_ERR(authkey)) 3461 return PTR_ERR(authkey); 3462 3463 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) { 3464 ret = -EINVAL; 3465 goto out; 3466 } 3467 3468 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id); 3469 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) { 3470 ret = -EINVAL; 3471 goto out; 3472 } 3473 3474 ret = sctp_auth_set_key(ep, asoc, authkey); 3475 out: 3476 kzfree(authkey); 3477 return ret; 3478 } 3479 3480 /* 3481 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY) 3482 * 3483 * This option will get or set the active shared key to be used to build 3484 * the association shared key. 3485 */ 3486 static int sctp_setsockopt_active_key(struct sock *sk, 3487 char __user *optval, 3488 unsigned int optlen) 3489 { 3490 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3491 struct sctp_authkeyid val; 3492 struct sctp_association *asoc; 3493 3494 if (!ep->auth_enable) 3495 return -EACCES; 3496 3497 if (optlen != sizeof(struct sctp_authkeyid)) 3498 return -EINVAL; 3499 if (copy_from_user(&val, optval, optlen)) 3500 return -EFAULT; 3501 3502 asoc = sctp_id2assoc(sk, val.scact_assoc_id); 3503 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP)) 3504 return -EINVAL; 3505 3506 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber); 3507 } 3508 3509 /* 3510 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY) 3511 * 3512 * This set option will delete a shared secret key from use. 3513 */ 3514 static int sctp_setsockopt_del_key(struct sock *sk, 3515 char __user *optval, 3516 unsigned int optlen) 3517 { 3518 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 3519 struct sctp_authkeyid val; 3520 struct sctp_association *asoc; 3521 3522 if (!ep->auth_enable) 3523 return -EACCES; 3524 3525 if (optlen != sizeof(struct sctp_authkeyid)) 3526 return -EINVAL; 3527 if (copy_from_user(&val, optval, optlen)) 3528 return -EFAULT; 3529 3530 asoc = sctp_id2assoc(sk, val.scact_assoc_id); 3531 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP)) 3532 return -EINVAL; 3533 3534 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber); 3535 3536 } 3537 3538 /* 3539 * 8.1.23 SCTP_AUTO_ASCONF 3540 * 3541 * This option will enable or disable the use of the automatic generation of 3542 * ASCONF chunks to add and delete addresses to an existing association. Note 3543 * that this option has two caveats namely: a) it only affects sockets that 3544 * are bound to all addresses available to the SCTP stack, and b) the system 3545 * administrator may have an overriding control that turns the ASCONF feature 3546 * off no matter what setting the socket option may have. 3547 * This option expects an integer boolean flag, where a non-zero value turns on 3548 * the option, and a zero value turns off the option. 3549 * Note. In this implementation, socket operation overrides default parameter 3550 * being set by sysctl as well as FreeBSD implementation 3551 */ 3552 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval, 3553 unsigned int optlen) 3554 { 3555 int val; 3556 struct sctp_sock *sp = sctp_sk(sk); 3557 3558 if (optlen < sizeof(int)) 3559 return -EINVAL; 3560 if (get_user(val, (int __user *)optval)) 3561 return -EFAULT; 3562 if (!sctp_is_ep_boundall(sk) && val) 3563 return -EINVAL; 3564 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf)) 3565 return 0; 3566 3567 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock); 3568 if (val == 0 && sp->do_auto_asconf) { 3569 list_del(&sp->auto_asconf_list); 3570 sp->do_auto_asconf = 0; 3571 } else if (val && !sp->do_auto_asconf) { 3572 list_add_tail(&sp->auto_asconf_list, 3573 &sock_net(sk)->sctp.auto_asconf_splist); 3574 sp->do_auto_asconf = 1; 3575 } 3576 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock); 3577 return 0; 3578 } 3579 3580 /* 3581 * SCTP_PEER_ADDR_THLDS 3582 * 3583 * This option allows us to alter the partially failed threshold for one or all 3584 * transports in an association. See Section 6.1 of: 3585 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt 3586 */ 3587 static int sctp_setsockopt_paddr_thresholds(struct sock *sk, 3588 char __user *optval, 3589 unsigned int optlen) 3590 { 3591 struct sctp_paddrthlds val; 3592 struct sctp_transport *trans; 3593 struct sctp_association *asoc; 3594 3595 if (optlen < sizeof(struct sctp_paddrthlds)) 3596 return -EINVAL; 3597 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, 3598 sizeof(struct sctp_paddrthlds))) 3599 return -EFAULT; 3600 3601 3602 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) { 3603 asoc = sctp_id2assoc(sk, val.spt_assoc_id); 3604 if (!asoc) 3605 return -ENOENT; 3606 list_for_each_entry(trans, &asoc->peer.transport_addr_list, 3607 transports) { 3608 if (val.spt_pathmaxrxt) 3609 trans->pathmaxrxt = val.spt_pathmaxrxt; 3610 trans->pf_retrans = val.spt_pathpfthld; 3611 } 3612 3613 if (val.spt_pathmaxrxt) 3614 asoc->pathmaxrxt = val.spt_pathmaxrxt; 3615 asoc->pf_retrans = val.spt_pathpfthld; 3616 } else { 3617 trans = sctp_addr_id2transport(sk, &val.spt_address, 3618 val.spt_assoc_id); 3619 if (!trans) 3620 return -ENOENT; 3621 3622 if (val.spt_pathmaxrxt) 3623 trans->pathmaxrxt = val.spt_pathmaxrxt; 3624 trans->pf_retrans = val.spt_pathpfthld; 3625 } 3626 3627 return 0; 3628 } 3629 3630 static int sctp_setsockopt_recvrcvinfo(struct sock *sk, 3631 char __user *optval, 3632 unsigned int optlen) 3633 { 3634 int val; 3635 3636 if (optlen < sizeof(int)) 3637 return -EINVAL; 3638 if (get_user(val, (int __user *) optval)) 3639 return -EFAULT; 3640 3641 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1; 3642 3643 return 0; 3644 } 3645 3646 static int sctp_setsockopt_recvnxtinfo(struct sock *sk, 3647 char __user *optval, 3648 unsigned int optlen) 3649 { 3650 int val; 3651 3652 if (optlen < sizeof(int)) 3653 return -EINVAL; 3654 if (get_user(val, (int __user *) optval)) 3655 return -EFAULT; 3656 3657 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1; 3658 3659 return 0; 3660 } 3661 3662 /* API 6.2 setsockopt(), getsockopt() 3663 * 3664 * Applications use setsockopt() and getsockopt() to set or retrieve 3665 * socket options. Socket options are used to change the default 3666 * behavior of sockets calls. They are described in Section 7. 3667 * 3668 * The syntax is: 3669 * 3670 * ret = getsockopt(int sd, int level, int optname, void __user *optval, 3671 * int __user *optlen); 3672 * ret = setsockopt(int sd, int level, int optname, const void __user *optval, 3673 * int optlen); 3674 * 3675 * sd - the socket descript. 3676 * level - set to IPPROTO_SCTP for all SCTP options. 3677 * optname - the option name. 3678 * optval - the buffer to store the value of the option. 3679 * optlen - the size of the buffer. 3680 */ 3681 static int sctp_setsockopt(struct sock *sk, int level, int optname, 3682 char __user *optval, unsigned int optlen) 3683 { 3684 int retval = 0; 3685 3686 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname); 3687 3688 /* I can hardly begin to describe how wrong this is. This is 3689 * so broken as to be worse than useless. The API draft 3690 * REALLY is NOT helpful here... I am not convinced that the 3691 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP 3692 * are at all well-founded. 3693 */ 3694 if (level != SOL_SCTP) { 3695 struct sctp_af *af = sctp_sk(sk)->pf->af; 3696 retval = af->setsockopt(sk, level, optname, optval, optlen); 3697 goto out_nounlock; 3698 } 3699 3700 lock_sock(sk); 3701 3702 switch (optname) { 3703 case SCTP_SOCKOPT_BINDX_ADD: 3704 /* 'optlen' is the size of the addresses buffer. */ 3705 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval, 3706 optlen, SCTP_BINDX_ADD_ADDR); 3707 break; 3708 3709 case SCTP_SOCKOPT_BINDX_REM: 3710 /* 'optlen' is the size of the addresses buffer. */ 3711 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval, 3712 optlen, SCTP_BINDX_REM_ADDR); 3713 break; 3714 3715 case SCTP_SOCKOPT_CONNECTX_OLD: 3716 /* 'optlen' is the size of the addresses buffer. */ 3717 retval = sctp_setsockopt_connectx_old(sk, 3718 (struct sockaddr __user *)optval, 3719 optlen); 3720 break; 3721 3722 case SCTP_SOCKOPT_CONNECTX: 3723 /* 'optlen' is the size of the addresses buffer. */ 3724 retval = sctp_setsockopt_connectx(sk, 3725 (struct sockaddr __user *)optval, 3726 optlen); 3727 break; 3728 3729 case SCTP_DISABLE_FRAGMENTS: 3730 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen); 3731 break; 3732 3733 case SCTP_EVENTS: 3734 retval = sctp_setsockopt_events(sk, optval, optlen); 3735 break; 3736 3737 case SCTP_AUTOCLOSE: 3738 retval = sctp_setsockopt_autoclose(sk, optval, optlen); 3739 break; 3740 3741 case SCTP_PEER_ADDR_PARAMS: 3742 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen); 3743 break; 3744 3745 case SCTP_DELAYED_SACK: 3746 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen); 3747 break; 3748 case SCTP_PARTIAL_DELIVERY_POINT: 3749 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen); 3750 break; 3751 3752 case SCTP_INITMSG: 3753 retval = sctp_setsockopt_initmsg(sk, optval, optlen); 3754 break; 3755 case SCTP_DEFAULT_SEND_PARAM: 3756 retval = sctp_setsockopt_default_send_param(sk, optval, 3757 optlen); 3758 break; 3759 case SCTP_DEFAULT_SNDINFO: 3760 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen); 3761 break; 3762 case SCTP_PRIMARY_ADDR: 3763 retval = sctp_setsockopt_primary_addr(sk, optval, optlen); 3764 break; 3765 case SCTP_SET_PEER_PRIMARY_ADDR: 3766 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen); 3767 break; 3768 case SCTP_NODELAY: 3769 retval = sctp_setsockopt_nodelay(sk, optval, optlen); 3770 break; 3771 case SCTP_RTOINFO: 3772 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen); 3773 break; 3774 case SCTP_ASSOCINFO: 3775 retval = sctp_setsockopt_associnfo(sk, optval, optlen); 3776 break; 3777 case SCTP_I_WANT_MAPPED_V4_ADDR: 3778 retval = sctp_setsockopt_mappedv4(sk, optval, optlen); 3779 break; 3780 case SCTP_MAXSEG: 3781 retval = sctp_setsockopt_maxseg(sk, optval, optlen); 3782 break; 3783 case SCTP_ADAPTATION_LAYER: 3784 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen); 3785 break; 3786 case SCTP_CONTEXT: 3787 retval = sctp_setsockopt_context(sk, optval, optlen); 3788 break; 3789 case SCTP_FRAGMENT_INTERLEAVE: 3790 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen); 3791 break; 3792 case SCTP_MAX_BURST: 3793 retval = sctp_setsockopt_maxburst(sk, optval, optlen); 3794 break; 3795 case SCTP_AUTH_CHUNK: 3796 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen); 3797 break; 3798 case SCTP_HMAC_IDENT: 3799 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen); 3800 break; 3801 case SCTP_AUTH_KEY: 3802 retval = sctp_setsockopt_auth_key(sk, optval, optlen); 3803 break; 3804 case SCTP_AUTH_ACTIVE_KEY: 3805 retval = sctp_setsockopt_active_key(sk, optval, optlen); 3806 break; 3807 case SCTP_AUTH_DELETE_KEY: 3808 retval = sctp_setsockopt_del_key(sk, optval, optlen); 3809 break; 3810 case SCTP_AUTO_ASCONF: 3811 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen); 3812 break; 3813 case SCTP_PEER_ADDR_THLDS: 3814 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen); 3815 break; 3816 case SCTP_RECVRCVINFO: 3817 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen); 3818 break; 3819 case SCTP_RECVNXTINFO: 3820 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen); 3821 break; 3822 default: 3823 retval = -ENOPROTOOPT; 3824 break; 3825 } 3826 3827 release_sock(sk); 3828 3829 out_nounlock: 3830 return retval; 3831 } 3832 3833 /* API 3.1.6 connect() - UDP Style Syntax 3834 * 3835 * An application may use the connect() call in the UDP model to initiate an 3836 * association without sending data. 3837 * 3838 * The syntax is: 3839 * 3840 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len); 3841 * 3842 * sd: the socket descriptor to have a new association added to. 3843 * 3844 * nam: the address structure (either struct sockaddr_in or struct 3845 * sockaddr_in6 defined in RFC2553 [7]). 3846 * 3847 * len: the size of the address. 3848 */ 3849 static int sctp_connect(struct sock *sk, struct sockaddr *addr, 3850 int addr_len) 3851 { 3852 int err = 0; 3853 struct sctp_af *af; 3854 3855 lock_sock(sk); 3856 3857 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk, 3858 addr, addr_len); 3859 3860 /* Validate addr_len before calling common connect/connectx routine. */ 3861 af = sctp_get_af_specific(addr->sa_family); 3862 if (!af || addr_len < af->sockaddr_len) { 3863 err = -EINVAL; 3864 } else { 3865 /* Pass correct addr len to common routine (so it knows there 3866 * is only one address being passed. 3867 */ 3868 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL); 3869 } 3870 3871 release_sock(sk); 3872 return err; 3873 } 3874 3875 /* FIXME: Write comments. */ 3876 static int sctp_disconnect(struct sock *sk, int flags) 3877 { 3878 return -EOPNOTSUPP; /* STUB */ 3879 } 3880 3881 /* 4.1.4 accept() - TCP Style Syntax 3882 * 3883 * Applications use accept() call to remove an established SCTP 3884 * association from the accept queue of the endpoint. A new socket 3885 * descriptor will be returned from accept() to represent the newly 3886 * formed association. 3887 */ 3888 static struct sock *sctp_accept(struct sock *sk, int flags, int *err) 3889 { 3890 struct sctp_sock *sp; 3891 struct sctp_endpoint *ep; 3892 struct sock *newsk = NULL; 3893 struct sctp_association *asoc; 3894 long timeo; 3895 int error = 0; 3896 3897 lock_sock(sk); 3898 3899 sp = sctp_sk(sk); 3900 ep = sp->ep; 3901 3902 if (!sctp_style(sk, TCP)) { 3903 error = -EOPNOTSUPP; 3904 goto out; 3905 } 3906 3907 if (!sctp_sstate(sk, LISTENING)) { 3908 error = -EINVAL; 3909 goto out; 3910 } 3911 3912 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); 3913 3914 error = sctp_wait_for_accept(sk, timeo); 3915 if (error) 3916 goto out; 3917 3918 /* We treat the list of associations on the endpoint as the accept 3919 * queue and pick the first association on the list. 3920 */ 3921 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs); 3922 3923 newsk = sp->pf->create_accept_sk(sk, asoc); 3924 if (!newsk) { 3925 error = -ENOMEM; 3926 goto out; 3927 } 3928 3929 /* Populate the fields of the newsk from the oldsk and migrate the 3930 * asoc to the newsk. 3931 */ 3932 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP); 3933 3934 out: 3935 release_sock(sk); 3936 *err = error; 3937 return newsk; 3938 } 3939 3940 /* The SCTP ioctl handler. */ 3941 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg) 3942 { 3943 int rc = -ENOTCONN; 3944 3945 lock_sock(sk); 3946 3947 /* 3948 * SEQPACKET-style sockets in LISTENING state are valid, for 3949 * SCTP, so only discard TCP-style sockets in LISTENING state. 3950 */ 3951 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) 3952 goto out; 3953 3954 switch (cmd) { 3955 case SIOCINQ: { 3956 struct sk_buff *skb; 3957 unsigned int amount = 0; 3958 3959 skb = skb_peek(&sk->sk_receive_queue); 3960 if (skb != NULL) { 3961 /* 3962 * We will only return the amount of this packet since 3963 * that is all that will be read. 3964 */ 3965 amount = skb->len; 3966 } 3967 rc = put_user(amount, (int __user *)arg); 3968 break; 3969 } 3970 default: 3971 rc = -ENOIOCTLCMD; 3972 break; 3973 } 3974 out: 3975 release_sock(sk); 3976 return rc; 3977 } 3978 3979 /* This is the function which gets called during socket creation to 3980 * initialized the SCTP-specific portion of the sock. 3981 * The sock structure should already be zero-filled memory. 3982 */ 3983 static int sctp_init_sock(struct sock *sk) 3984 { 3985 struct net *net = sock_net(sk); 3986 struct sctp_sock *sp; 3987 3988 pr_debug("%s: sk:%p\n", __func__, sk); 3989 3990 sp = sctp_sk(sk); 3991 3992 /* Initialize the SCTP per socket area. */ 3993 switch (sk->sk_type) { 3994 case SOCK_SEQPACKET: 3995 sp->type = SCTP_SOCKET_UDP; 3996 break; 3997 case SOCK_STREAM: 3998 sp->type = SCTP_SOCKET_TCP; 3999 break; 4000 default: 4001 return -ESOCKTNOSUPPORT; 4002 } 4003 4004 /* Initialize default send parameters. These parameters can be 4005 * modified with the SCTP_DEFAULT_SEND_PARAM socket option. 4006 */ 4007 sp->default_stream = 0; 4008 sp->default_ppid = 0; 4009 sp->default_flags = 0; 4010 sp->default_context = 0; 4011 sp->default_timetolive = 0; 4012 4013 sp->default_rcv_context = 0; 4014 sp->max_burst = net->sctp.max_burst; 4015 4016 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg; 4017 4018 /* Initialize default setup parameters. These parameters 4019 * can be modified with the SCTP_INITMSG socket option or 4020 * overridden by the SCTP_INIT CMSG. 4021 */ 4022 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams; 4023 sp->initmsg.sinit_max_instreams = sctp_max_instreams; 4024 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init; 4025 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max; 4026 4027 /* Initialize default RTO related parameters. These parameters can 4028 * be modified for with the SCTP_RTOINFO socket option. 4029 */ 4030 sp->rtoinfo.srto_initial = net->sctp.rto_initial; 4031 sp->rtoinfo.srto_max = net->sctp.rto_max; 4032 sp->rtoinfo.srto_min = net->sctp.rto_min; 4033 4034 /* Initialize default association related parameters. These parameters 4035 * can be modified with the SCTP_ASSOCINFO socket option. 4036 */ 4037 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association; 4038 sp->assocparams.sasoc_number_peer_destinations = 0; 4039 sp->assocparams.sasoc_peer_rwnd = 0; 4040 sp->assocparams.sasoc_local_rwnd = 0; 4041 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life; 4042 4043 /* Initialize default event subscriptions. By default, all the 4044 * options are off. 4045 */ 4046 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe)); 4047 4048 /* Default Peer Address Parameters. These defaults can 4049 * be modified via SCTP_PEER_ADDR_PARAMS 4050 */ 4051 sp->hbinterval = net->sctp.hb_interval; 4052 sp->pathmaxrxt = net->sctp.max_retrans_path; 4053 sp->pathmtu = 0; /* allow default discovery */ 4054 sp->sackdelay = net->sctp.sack_timeout; 4055 sp->sackfreq = 2; 4056 sp->param_flags = SPP_HB_ENABLE | 4057 SPP_PMTUD_ENABLE | 4058 SPP_SACKDELAY_ENABLE; 4059 4060 /* If enabled no SCTP message fragmentation will be performed. 4061 * Configure through SCTP_DISABLE_FRAGMENTS socket option. 4062 */ 4063 sp->disable_fragments = 0; 4064 4065 /* Enable Nagle algorithm by default. */ 4066 sp->nodelay = 0; 4067 4068 sp->recvrcvinfo = 0; 4069 sp->recvnxtinfo = 0; 4070 4071 /* Enable by default. */ 4072 sp->v4mapped = 1; 4073 4074 /* Auto-close idle associations after the configured 4075 * number of seconds. A value of 0 disables this 4076 * feature. Configure through the SCTP_AUTOCLOSE socket option, 4077 * for UDP-style sockets only. 4078 */ 4079 sp->autoclose = 0; 4080 4081 /* User specified fragmentation limit. */ 4082 sp->user_frag = 0; 4083 4084 sp->adaptation_ind = 0; 4085 4086 sp->pf = sctp_get_pf_specific(sk->sk_family); 4087 4088 /* Control variables for partial data delivery. */ 4089 atomic_set(&sp->pd_mode, 0); 4090 skb_queue_head_init(&sp->pd_lobby); 4091 sp->frag_interleave = 0; 4092 4093 /* Create a per socket endpoint structure. Even if we 4094 * change the data structure relationships, this may still 4095 * be useful for storing pre-connect address information. 4096 */ 4097 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL); 4098 if (!sp->ep) 4099 return -ENOMEM; 4100 4101 sp->hmac = NULL; 4102 4103 sk->sk_destruct = sctp_destruct_sock; 4104 4105 SCTP_DBG_OBJCNT_INC(sock); 4106 4107 local_bh_disable(); 4108 percpu_counter_inc(&sctp_sockets_allocated); 4109 sock_prot_inuse_add(net, sk->sk_prot, 1); 4110 4111 /* Nothing can fail after this block, otherwise 4112 * sctp_destroy_sock() will be called without addr_wq_lock held 4113 */ 4114 if (net->sctp.default_auto_asconf) { 4115 spin_lock(&sock_net(sk)->sctp.addr_wq_lock); 4116 list_add_tail(&sp->auto_asconf_list, 4117 &net->sctp.auto_asconf_splist); 4118 sp->do_auto_asconf = 1; 4119 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock); 4120 } else { 4121 sp->do_auto_asconf = 0; 4122 } 4123 4124 local_bh_enable(); 4125 4126 return 0; 4127 } 4128 4129 /* Cleanup any SCTP per socket resources. Must be called with 4130 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true 4131 */ 4132 static void sctp_destroy_sock(struct sock *sk) 4133 { 4134 struct sctp_sock *sp; 4135 4136 pr_debug("%s: sk:%p\n", __func__, sk); 4137 4138 /* Release our hold on the endpoint. */ 4139 sp = sctp_sk(sk); 4140 /* This could happen during socket init, thus we bail out 4141 * early, since the rest of the below is not setup either. 4142 */ 4143 if (sp->ep == NULL) 4144 return; 4145 4146 if (sp->do_auto_asconf) { 4147 sp->do_auto_asconf = 0; 4148 list_del(&sp->auto_asconf_list); 4149 } 4150 sctp_endpoint_free(sp->ep); 4151 local_bh_disable(); 4152 percpu_counter_dec(&sctp_sockets_allocated); 4153 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 4154 local_bh_enable(); 4155 } 4156 4157 /* Triggered when there are no references on the socket anymore */ 4158 static void sctp_destruct_sock(struct sock *sk) 4159 { 4160 struct sctp_sock *sp = sctp_sk(sk); 4161 4162 /* Free up the HMAC transform. */ 4163 crypto_free_hash(sp->hmac); 4164 4165 inet_sock_destruct(sk); 4166 } 4167 4168 /* API 4.1.7 shutdown() - TCP Style Syntax 4169 * int shutdown(int socket, int how); 4170 * 4171 * sd - the socket descriptor of the association to be closed. 4172 * how - Specifies the type of shutdown. The values are 4173 * as follows: 4174 * SHUT_RD 4175 * Disables further receive operations. No SCTP 4176 * protocol action is taken. 4177 * SHUT_WR 4178 * Disables further send operations, and initiates 4179 * the SCTP shutdown sequence. 4180 * SHUT_RDWR 4181 * Disables further send and receive operations 4182 * and initiates the SCTP shutdown sequence. 4183 */ 4184 static void sctp_shutdown(struct sock *sk, int how) 4185 { 4186 struct net *net = sock_net(sk); 4187 struct sctp_endpoint *ep; 4188 struct sctp_association *asoc; 4189 4190 if (!sctp_style(sk, TCP)) 4191 return; 4192 4193 if (how & SEND_SHUTDOWN) { 4194 ep = sctp_sk(sk)->ep; 4195 if (!list_empty(&ep->asocs)) { 4196 asoc = list_entry(ep->asocs.next, 4197 struct sctp_association, asocs); 4198 sctp_primitive_SHUTDOWN(net, asoc, NULL); 4199 } 4200 } 4201 } 4202 4203 /* 7.2.1 Association Status (SCTP_STATUS) 4204 4205 * Applications can retrieve current status information about an 4206 * association, including association state, peer receiver window size, 4207 * number of unacked data chunks, and number of data chunks pending 4208 * receipt. This information is read-only. 4209 */ 4210 static int sctp_getsockopt_sctp_status(struct sock *sk, int len, 4211 char __user *optval, 4212 int __user *optlen) 4213 { 4214 struct sctp_status status; 4215 struct sctp_association *asoc = NULL; 4216 struct sctp_transport *transport; 4217 sctp_assoc_t associd; 4218 int retval = 0; 4219 4220 if (len < sizeof(status)) { 4221 retval = -EINVAL; 4222 goto out; 4223 } 4224 4225 len = sizeof(status); 4226 if (copy_from_user(&status, optval, len)) { 4227 retval = -EFAULT; 4228 goto out; 4229 } 4230 4231 associd = status.sstat_assoc_id; 4232 asoc = sctp_id2assoc(sk, associd); 4233 if (!asoc) { 4234 retval = -EINVAL; 4235 goto out; 4236 } 4237 4238 transport = asoc->peer.primary_path; 4239 4240 status.sstat_assoc_id = sctp_assoc2id(asoc); 4241 status.sstat_state = sctp_assoc_to_state(asoc); 4242 status.sstat_rwnd = asoc->peer.rwnd; 4243 status.sstat_unackdata = asoc->unack_data; 4244 4245 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map); 4246 status.sstat_instrms = asoc->c.sinit_max_instreams; 4247 status.sstat_outstrms = asoc->c.sinit_num_ostreams; 4248 status.sstat_fragmentation_point = asoc->frag_point; 4249 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc); 4250 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr, 4251 transport->af_specific->sockaddr_len); 4252 /* Map ipv4 address into v4-mapped-on-v6 address. */ 4253 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk), 4254 (union sctp_addr *)&status.sstat_primary.spinfo_address); 4255 status.sstat_primary.spinfo_state = transport->state; 4256 status.sstat_primary.spinfo_cwnd = transport->cwnd; 4257 status.sstat_primary.spinfo_srtt = transport->srtt; 4258 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto); 4259 status.sstat_primary.spinfo_mtu = transport->pathmtu; 4260 4261 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN) 4262 status.sstat_primary.spinfo_state = SCTP_ACTIVE; 4263 4264 if (put_user(len, optlen)) { 4265 retval = -EFAULT; 4266 goto out; 4267 } 4268 4269 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n", 4270 __func__, len, status.sstat_state, status.sstat_rwnd, 4271 status.sstat_assoc_id); 4272 4273 if (copy_to_user(optval, &status, len)) { 4274 retval = -EFAULT; 4275 goto out; 4276 } 4277 4278 out: 4279 return retval; 4280 } 4281 4282 4283 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO) 4284 * 4285 * Applications can retrieve information about a specific peer address 4286 * of an association, including its reachability state, congestion 4287 * window, and retransmission timer values. This information is 4288 * read-only. 4289 */ 4290 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len, 4291 char __user *optval, 4292 int __user *optlen) 4293 { 4294 struct sctp_paddrinfo pinfo; 4295 struct sctp_transport *transport; 4296 int retval = 0; 4297 4298 if (len < sizeof(pinfo)) { 4299 retval = -EINVAL; 4300 goto out; 4301 } 4302 4303 len = sizeof(pinfo); 4304 if (copy_from_user(&pinfo, optval, len)) { 4305 retval = -EFAULT; 4306 goto out; 4307 } 4308 4309 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address, 4310 pinfo.spinfo_assoc_id); 4311 if (!transport) 4312 return -EINVAL; 4313 4314 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc); 4315 pinfo.spinfo_state = transport->state; 4316 pinfo.spinfo_cwnd = transport->cwnd; 4317 pinfo.spinfo_srtt = transport->srtt; 4318 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto); 4319 pinfo.spinfo_mtu = transport->pathmtu; 4320 4321 if (pinfo.spinfo_state == SCTP_UNKNOWN) 4322 pinfo.spinfo_state = SCTP_ACTIVE; 4323 4324 if (put_user(len, optlen)) { 4325 retval = -EFAULT; 4326 goto out; 4327 } 4328 4329 if (copy_to_user(optval, &pinfo, len)) { 4330 retval = -EFAULT; 4331 goto out; 4332 } 4333 4334 out: 4335 return retval; 4336 } 4337 4338 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS) 4339 * 4340 * This option is a on/off flag. If enabled no SCTP message 4341 * fragmentation will be performed. Instead if a message being sent 4342 * exceeds the current PMTU size, the message will NOT be sent and 4343 * instead a error will be indicated to the user. 4344 */ 4345 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len, 4346 char __user *optval, int __user *optlen) 4347 { 4348 int val; 4349 4350 if (len < sizeof(int)) 4351 return -EINVAL; 4352 4353 len = sizeof(int); 4354 val = (sctp_sk(sk)->disable_fragments == 1); 4355 if (put_user(len, optlen)) 4356 return -EFAULT; 4357 if (copy_to_user(optval, &val, len)) 4358 return -EFAULT; 4359 return 0; 4360 } 4361 4362 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS) 4363 * 4364 * This socket option is used to specify various notifications and 4365 * ancillary data the user wishes to receive. 4366 */ 4367 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval, 4368 int __user *optlen) 4369 { 4370 if (len <= 0) 4371 return -EINVAL; 4372 if (len > sizeof(struct sctp_event_subscribe)) 4373 len = sizeof(struct sctp_event_subscribe); 4374 if (put_user(len, optlen)) 4375 return -EFAULT; 4376 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len)) 4377 return -EFAULT; 4378 return 0; 4379 } 4380 4381 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE) 4382 * 4383 * This socket option is applicable to the UDP-style socket only. When 4384 * set it will cause associations that are idle for more than the 4385 * specified number of seconds to automatically close. An association 4386 * being idle is defined an association that has NOT sent or received 4387 * user data. The special value of '0' indicates that no automatic 4388 * close of any associations should be performed. The option expects an 4389 * integer defining the number of seconds of idle time before an 4390 * association is closed. 4391 */ 4392 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen) 4393 { 4394 /* Applicable to UDP-style socket only */ 4395 if (sctp_style(sk, TCP)) 4396 return -EOPNOTSUPP; 4397 if (len < sizeof(int)) 4398 return -EINVAL; 4399 len = sizeof(int); 4400 if (put_user(len, optlen)) 4401 return -EFAULT; 4402 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int))) 4403 return -EFAULT; 4404 return 0; 4405 } 4406 4407 /* Helper routine to branch off an association to a new socket. */ 4408 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp) 4409 { 4410 struct sctp_association *asoc = sctp_id2assoc(sk, id); 4411 struct sctp_sock *sp = sctp_sk(sk); 4412 struct socket *sock; 4413 int err = 0; 4414 4415 if (!asoc) 4416 return -EINVAL; 4417 4418 /* An association cannot be branched off from an already peeled-off 4419 * socket, nor is this supported for tcp style sockets. 4420 */ 4421 if (!sctp_style(sk, UDP)) 4422 return -EINVAL; 4423 4424 /* Create a new socket. */ 4425 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock); 4426 if (err < 0) 4427 return err; 4428 4429 sctp_copy_sock(sock->sk, sk, asoc); 4430 4431 /* Make peeled-off sockets more like 1-1 accepted sockets. 4432 * Set the daddr and initialize id to something more random 4433 */ 4434 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk); 4435 4436 /* Populate the fields of the newsk from the oldsk and migrate the 4437 * asoc to the newsk. 4438 */ 4439 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH); 4440 4441 *sockp = sock; 4442 4443 return err; 4444 } 4445 EXPORT_SYMBOL(sctp_do_peeloff); 4446 4447 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen) 4448 { 4449 sctp_peeloff_arg_t peeloff; 4450 struct socket *newsock; 4451 struct file *newfile; 4452 int retval = 0; 4453 4454 if (len < sizeof(sctp_peeloff_arg_t)) 4455 return -EINVAL; 4456 len = sizeof(sctp_peeloff_arg_t); 4457 if (copy_from_user(&peeloff, optval, len)) 4458 return -EFAULT; 4459 4460 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock); 4461 if (retval < 0) 4462 goto out; 4463 4464 /* Map the socket to an unused fd that can be returned to the user. */ 4465 retval = get_unused_fd_flags(0); 4466 if (retval < 0) { 4467 sock_release(newsock); 4468 goto out; 4469 } 4470 4471 newfile = sock_alloc_file(newsock, 0, NULL); 4472 if (IS_ERR(newfile)) { 4473 put_unused_fd(retval); 4474 sock_release(newsock); 4475 return PTR_ERR(newfile); 4476 } 4477 4478 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk, 4479 retval); 4480 4481 /* Return the fd mapped to the new socket. */ 4482 if (put_user(len, optlen)) { 4483 fput(newfile); 4484 put_unused_fd(retval); 4485 return -EFAULT; 4486 } 4487 peeloff.sd = retval; 4488 if (copy_to_user(optval, &peeloff, len)) { 4489 fput(newfile); 4490 put_unused_fd(retval); 4491 return -EFAULT; 4492 } 4493 fd_install(retval, newfile); 4494 out: 4495 return retval; 4496 } 4497 4498 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS) 4499 * 4500 * Applications can enable or disable heartbeats for any peer address of 4501 * an association, modify an address's heartbeat interval, force a 4502 * heartbeat to be sent immediately, and adjust the address's maximum 4503 * number of retransmissions sent before an address is considered 4504 * unreachable. The following structure is used to access and modify an 4505 * address's parameters: 4506 * 4507 * struct sctp_paddrparams { 4508 * sctp_assoc_t spp_assoc_id; 4509 * struct sockaddr_storage spp_address; 4510 * uint32_t spp_hbinterval; 4511 * uint16_t spp_pathmaxrxt; 4512 * uint32_t spp_pathmtu; 4513 * uint32_t spp_sackdelay; 4514 * uint32_t spp_flags; 4515 * }; 4516 * 4517 * spp_assoc_id - (one-to-many style socket) This is filled in the 4518 * application, and identifies the association for 4519 * this query. 4520 * spp_address - This specifies which address is of interest. 4521 * spp_hbinterval - This contains the value of the heartbeat interval, 4522 * in milliseconds. If a value of zero 4523 * is present in this field then no changes are to 4524 * be made to this parameter. 4525 * spp_pathmaxrxt - This contains the maximum number of 4526 * retransmissions before this address shall be 4527 * considered unreachable. If a value of zero 4528 * is present in this field then no changes are to 4529 * be made to this parameter. 4530 * spp_pathmtu - When Path MTU discovery is disabled the value 4531 * specified here will be the "fixed" path mtu. 4532 * Note that if the spp_address field is empty 4533 * then all associations on this address will 4534 * have this fixed path mtu set upon them. 4535 * 4536 * spp_sackdelay - When delayed sack is enabled, this value specifies 4537 * the number of milliseconds that sacks will be delayed 4538 * for. This value will apply to all addresses of an 4539 * association if the spp_address field is empty. Note 4540 * also, that if delayed sack is enabled and this 4541 * value is set to 0, no change is made to the last 4542 * recorded delayed sack timer value. 4543 * 4544 * spp_flags - These flags are used to control various features 4545 * on an association. The flag field may contain 4546 * zero or more of the following options. 4547 * 4548 * SPP_HB_ENABLE - Enable heartbeats on the 4549 * specified address. Note that if the address 4550 * field is empty all addresses for the association 4551 * have heartbeats enabled upon them. 4552 * 4553 * SPP_HB_DISABLE - Disable heartbeats on the 4554 * speicifed address. Note that if the address 4555 * field is empty all addresses for the association 4556 * will have their heartbeats disabled. Note also 4557 * that SPP_HB_ENABLE and SPP_HB_DISABLE are 4558 * mutually exclusive, only one of these two should 4559 * be specified. Enabling both fields will have 4560 * undetermined results. 4561 * 4562 * SPP_HB_DEMAND - Request a user initiated heartbeat 4563 * to be made immediately. 4564 * 4565 * SPP_PMTUD_ENABLE - This field will enable PMTU 4566 * discovery upon the specified address. Note that 4567 * if the address feild is empty then all addresses 4568 * on the association are effected. 4569 * 4570 * SPP_PMTUD_DISABLE - This field will disable PMTU 4571 * discovery upon the specified address. Note that 4572 * if the address feild is empty then all addresses 4573 * on the association are effected. Not also that 4574 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually 4575 * exclusive. Enabling both will have undetermined 4576 * results. 4577 * 4578 * SPP_SACKDELAY_ENABLE - Setting this flag turns 4579 * on delayed sack. The time specified in spp_sackdelay 4580 * is used to specify the sack delay for this address. Note 4581 * that if spp_address is empty then all addresses will 4582 * enable delayed sack and take on the sack delay 4583 * value specified in spp_sackdelay. 4584 * SPP_SACKDELAY_DISABLE - Setting this flag turns 4585 * off delayed sack. If the spp_address field is blank then 4586 * delayed sack is disabled for the entire association. Note 4587 * also that this field is mutually exclusive to 4588 * SPP_SACKDELAY_ENABLE, setting both will have undefined 4589 * results. 4590 */ 4591 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len, 4592 char __user *optval, int __user *optlen) 4593 { 4594 struct sctp_paddrparams params; 4595 struct sctp_transport *trans = NULL; 4596 struct sctp_association *asoc = NULL; 4597 struct sctp_sock *sp = sctp_sk(sk); 4598 4599 if (len < sizeof(struct sctp_paddrparams)) 4600 return -EINVAL; 4601 len = sizeof(struct sctp_paddrparams); 4602 if (copy_from_user(¶ms, optval, len)) 4603 return -EFAULT; 4604 4605 /* If an address other than INADDR_ANY is specified, and 4606 * no transport is found, then the request is invalid. 4607 */ 4608 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) { 4609 trans = sctp_addr_id2transport(sk, ¶ms.spp_address, 4610 params.spp_assoc_id); 4611 if (!trans) { 4612 pr_debug("%s: failed no transport\n", __func__); 4613 return -EINVAL; 4614 } 4615 } 4616 4617 /* Get association, if assoc_id != 0 and the socket is a one 4618 * to many style socket, and an association was not found, then 4619 * the id was invalid. 4620 */ 4621 asoc = sctp_id2assoc(sk, params.spp_assoc_id); 4622 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) { 4623 pr_debug("%s: failed no association\n", __func__); 4624 return -EINVAL; 4625 } 4626 4627 if (trans) { 4628 /* Fetch transport values. */ 4629 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval); 4630 params.spp_pathmtu = trans->pathmtu; 4631 params.spp_pathmaxrxt = trans->pathmaxrxt; 4632 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay); 4633 4634 /*draft-11 doesn't say what to return in spp_flags*/ 4635 params.spp_flags = trans->param_flags; 4636 } else if (asoc) { 4637 /* Fetch association values. */ 4638 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval); 4639 params.spp_pathmtu = asoc->pathmtu; 4640 params.spp_pathmaxrxt = asoc->pathmaxrxt; 4641 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay); 4642 4643 /*draft-11 doesn't say what to return in spp_flags*/ 4644 params.spp_flags = asoc->param_flags; 4645 } else { 4646 /* Fetch socket values. */ 4647 params.spp_hbinterval = sp->hbinterval; 4648 params.spp_pathmtu = sp->pathmtu; 4649 params.spp_sackdelay = sp->sackdelay; 4650 params.spp_pathmaxrxt = sp->pathmaxrxt; 4651 4652 /*draft-11 doesn't say what to return in spp_flags*/ 4653 params.spp_flags = sp->param_flags; 4654 } 4655 4656 if (copy_to_user(optval, ¶ms, len)) 4657 return -EFAULT; 4658 4659 if (put_user(len, optlen)) 4660 return -EFAULT; 4661 4662 return 0; 4663 } 4664 4665 /* 4666 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK) 4667 * 4668 * This option will effect the way delayed acks are performed. This 4669 * option allows you to get or set the delayed ack time, in 4670 * milliseconds. It also allows changing the delayed ack frequency. 4671 * Changing the frequency to 1 disables the delayed sack algorithm. If 4672 * the assoc_id is 0, then this sets or gets the endpoints default 4673 * values. If the assoc_id field is non-zero, then the set or get 4674 * effects the specified association for the one to many model (the 4675 * assoc_id field is ignored by the one to one model). Note that if 4676 * sack_delay or sack_freq are 0 when setting this option, then the 4677 * current values will remain unchanged. 4678 * 4679 * struct sctp_sack_info { 4680 * sctp_assoc_t sack_assoc_id; 4681 * uint32_t sack_delay; 4682 * uint32_t sack_freq; 4683 * }; 4684 * 4685 * sack_assoc_id - This parameter, indicates which association the user 4686 * is performing an action upon. Note that if this field's value is 4687 * zero then the endpoints default value is changed (effecting future 4688 * associations only). 4689 * 4690 * sack_delay - This parameter contains the number of milliseconds that 4691 * the user is requesting the delayed ACK timer be set to. Note that 4692 * this value is defined in the standard to be between 200 and 500 4693 * milliseconds. 4694 * 4695 * sack_freq - This parameter contains the number of packets that must 4696 * be received before a sack is sent without waiting for the delay 4697 * timer to expire. The default value for this is 2, setting this 4698 * value to 1 will disable the delayed sack algorithm. 4699 */ 4700 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len, 4701 char __user *optval, 4702 int __user *optlen) 4703 { 4704 struct sctp_sack_info params; 4705 struct sctp_association *asoc = NULL; 4706 struct sctp_sock *sp = sctp_sk(sk); 4707 4708 if (len >= sizeof(struct sctp_sack_info)) { 4709 len = sizeof(struct sctp_sack_info); 4710 4711 if (copy_from_user(¶ms, optval, len)) 4712 return -EFAULT; 4713 } else if (len == sizeof(struct sctp_assoc_value)) { 4714 pr_warn_ratelimited(DEPRECATED 4715 "%s (pid %d) " 4716 "Use of struct sctp_assoc_value in delayed_ack socket option.\n" 4717 "Use struct sctp_sack_info instead\n", 4718 current->comm, task_pid_nr(current)); 4719 if (copy_from_user(¶ms, optval, len)) 4720 return -EFAULT; 4721 } else 4722 return -EINVAL; 4723 4724 /* Get association, if sack_assoc_id != 0 and the socket is a one 4725 * to many style socket, and an association was not found, then 4726 * the id was invalid. 4727 */ 4728 asoc = sctp_id2assoc(sk, params.sack_assoc_id); 4729 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP)) 4730 return -EINVAL; 4731 4732 if (asoc) { 4733 /* Fetch association values. */ 4734 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) { 4735 params.sack_delay = jiffies_to_msecs( 4736 asoc->sackdelay); 4737 params.sack_freq = asoc->sackfreq; 4738 4739 } else { 4740 params.sack_delay = 0; 4741 params.sack_freq = 1; 4742 } 4743 } else { 4744 /* Fetch socket values. */ 4745 if (sp->param_flags & SPP_SACKDELAY_ENABLE) { 4746 params.sack_delay = sp->sackdelay; 4747 params.sack_freq = sp->sackfreq; 4748 } else { 4749 params.sack_delay = 0; 4750 params.sack_freq = 1; 4751 } 4752 } 4753 4754 if (copy_to_user(optval, ¶ms, len)) 4755 return -EFAULT; 4756 4757 if (put_user(len, optlen)) 4758 return -EFAULT; 4759 4760 return 0; 4761 } 4762 4763 /* 7.1.3 Initialization Parameters (SCTP_INITMSG) 4764 * 4765 * Applications can specify protocol parameters for the default association 4766 * initialization. The option name argument to setsockopt() and getsockopt() 4767 * is SCTP_INITMSG. 4768 * 4769 * Setting initialization parameters is effective only on an unconnected 4770 * socket (for UDP-style sockets only future associations are effected 4771 * by the change). With TCP-style sockets, this option is inherited by 4772 * sockets derived from a listener socket. 4773 */ 4774 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen) 4775 { 4776 if (len < sizeof(struct sctp_initmsg)) 4777 return -EINVAL; 4778 len = sizeof(struct sctp_initmsg); 4779 if (put_user(len, optlen)) 4780 return -EFAULT; 4781 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len)) 4782 return -EFAULT; 4783 return 0; 4784 } 4785 4786 4787 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len, 4788 char __user *optval, int __user *optlen) 4789 { 4790 struct sctp_association *asoc; 4791 int cnt = 0; 4792 struct sctp_getaddrs getaddrs; 4793 struct sctp_transport *from; 4794 void __user *to; 4795 union sctp_addr temp; 4796 struct sctp_sock *sp = sctp_sk(sk); 4797 int addrlen; 4798 size_t space_left; 4799 int bytes_copied; 4800 4801 if (len < sizeof(struct sctp_getaddrs)) 4802 return -EINVAL; 4803 4804 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs))) 4805 return -EFAULT; 4806 4807 /* For UDP-style sockets, id specifies the association to query. */ 4808 asoc = sctp_id2assoc(sk, getaddrs.assoc_id); 4809 if (!asoc) 4810 return -EINVAL; 4811 4812 to = optval + offsetof(struct sctp_getaddrs, addrs); 4813 space_left = len - offsetof(struct sctp_getaddrs, addrs); 4814 4815 list_for_each_entry(from, &asoc->peer.transport_addr_list, 4816 transports) { 4817 memcpy(&temp, &from->ipaddr, sizeof(temp)); 4818 addrlen = sctp_get_pf_specific(sk->sk_family) 4819 ->addr_to_user(sp, &temp); 4820 if (space_left < addrlen) 4821 return -ENOMEM; 4822 if (copy_to_user(to, &temp, addrlen)) 4823 return -EFAULT; 4824 to += addrlen; 4825 cnt++; 4826 space_left -= addrlen; 4827 } 4828 4829 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) 4830 return -EFAULT; 4831 bytes_copied = ((char __user *)to) - optval; 4832 if (put_user(bytes_copied, optlen)) 4833 return -EFAULT; 4834 4835 return 0; 4836 } 4837 4838 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to, 4839 size_t space_left, int *bytes_copied) 4840 { 4841 struct sctp_sockaddr_entry *addr; 4842 union sctp_addr temp; 4843 int cnt = 0; 4844 int addrlen; 4845 struct net *net = sock_net(sk); 4846 4847 rcu_read_lock(); 4848 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 4849 if (!addr->valid) 4850 continue; 4851 4852 if ((PF_INET == sk->sk_family) && 4853 (AF_INET6 == addr->a.sa.sa_family)) 4854 continue; 4855 if ((PF_INET6 == sk->sk_family) && 4856 inet_v6_ipv6only(sk) && 4857 (AF_INET == addr->a.sa.sa_family)) 4858 continue; 4859 memcpy(&temp, &addr->a, sizeof(temp)); 4860 if (!temp.v4.sin_port) 4861 temp.v4.sin_port = htons(port); 4862 4863 addrlen = sctp_get_pf_specific(sk->sk_family) 4864 ->addr_to_user(sctp_sk(sk), &temp); 4865 4866 if (space_left < addrlen) { 4867 cnt = -ENOMEM; 4868 break; 4869 } 4870 memcpy(to, &temp, addrlen); 4871 4872 to += addrlen; 4873 cnt++; 4874 space_left -= addrlen; 4875 *bytes_copied += addrlen; 4876 } 4877 rcu_read_unlock(); 4878 4879 return cnt; 4880 } 4881 4882 4883 static int sctp_getsockopt_local_addrs(struct sock *sk, int len, 4884 char __user *optval, int __user *optlen) 4885 { 4886 struct sctp_bind_addr *bp; 4887 struct sctp_association *asoc; 4888 int cnt = 0; 4889 struct sctp_getaddrs getaddrs; 4890 struct sctp_sockaddr_entry *addr; 4891 void __user *to; 4892 union sctp_addr temp; 4893 struct sctp_sock *sp = sctp_sk(sk); 4894 int addrlen; 4895 int err = 0; 4896 size_t space_left; 4897 int bytes_copied = 0; 4898 void *addrs; 4899 void *buf; 4900 4901 if (len < sizeof(struct sctp_getaddrs)) 4902 return -EINVAL; 4903 4904 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs))) 4905 return -EFAULT; 4906 4907 /* 4908 * For UDP-style sockets, id specifies the association to query. 4909 * If the id field is set to the value '0' then the locally bound 4910 * addresses are returned without regard to any particular 4911 * association. 4912 */ 4913 if (0 == getaddrs.assoc_id) { 4914 bp = &sctp_sk(sk)->ep->base.bind_addr; 4915 } else { 4916 asoc = sctp_id2assoc(sk, getaddrs.assoc_id); 4917 if (!asoc) 4918 return -EINVAL; 4919 bp = &asoc->base.bind_addr; 4920 } 4921 4922 to = optval + offsetof(struct sctp_getaddrs, addrs); 4923 space_left = len - offsetof(struct sctp_getaddrs, addrs); 4924 4925 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN); 4926 if (!addrs) 4927 return -ENOMEM; 4928 4929 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid 4930 * addresses from the global local address list. 4931 */ 4932 if (sctp_list_single_entry(&bp->address_list)) { 4933 addr = list_entry(bp->address_list.next, 4934 struct sctp_sockaddr_entry, list); 4935 if (sctp_is_any(sk, &addr->a)) { 4936 cnt = sctp_copy_laddrs(sk, bp->port, addrs, 4937 space_left, &bytes_copied); 4938 if (cnt < 0) { 4939 err = cnt; 4940 goto out; 4941 } 4942 goto copy_getaddrs; 4943 } 4944 } 4945 4946 buf = addrs; 4947 /* Protection on the bound address list is not needed since 4948 * in the socket option context we hold a socket lock and 4949 * thus the bound address list can't change. 4950 */ 4951 list_for_each_entry(addr, &bp->address_list, list) { 4952 memcpy(&temp, &addr->a, sizeof(temp)); 4953 addrlen = sctp_get_pf_specific(sk->sk_family) 4954 ->addr_to_user(sp, &temp); 4955 if (space_left < addrlen) { 4956 err = -ENOMEM; /*fixme: right error?*/ 4957 goto out; 4958 } 4959 memcpy(buf, &temp, addrlen); 4960 buf += addrlen; 4961 bytes_copied += addrlen; 4962 cnt++; 4963 space_left -= addrlen; 4964 } 4965 4966 copy_getaddrs: 4967 if (copy_to_user(to, addrs, bytes_copied)) { 4968 err = -EFAULT; 4969 goto out; 4970 } 4971 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) { 4972 err = -EFAULT; 4973 goto out; 4974 } 4975 if (put_user(bytes_copied, optlen)) 4976 err = -EFAULT; 4977 out: 4978 kfree(addrs); 4979 return err; 4980 } 4981 4982 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR) 4983 * 4984 * Requests that the local SCTP stack use the enclosed peer address as 4985 * the association primary. The enclosed address must be one of the 4986 * association peer's addresses. 4987 */ 4988 static int sctp_getsockopt_primary_addr(struct sock *sk, int len, 4989 char __user *optval, int __user *optlen) 4990 { 4991 struct sctp_prim prim; 4992 struct sctp_association *asoc; 4993 struct sctp_sock *sp = sctp_sk(sk); 4994 4995 if (len < sizeof(struct sctp_prim)) 4996 return -EINVAL; 4997 4998 len = sizeof(struct sctp_prim); 4999 5000 if (copy_from_user(&prim, optval, len)) 5001 return -EFAULT; 5002 5003 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id); 5004 if (!asoc) 5005 return -EINVAL; 5006 5007 if (!asoc->peer.primary_path) 5008 return -ENOTCONN; 5009 5010 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr, 5011 asoc->peer.primary_path->af_specific->sockaddr_len); 5012 5013 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp, 5014 (union sctp_addr *)&prim.ssp_addr); 5015 5016 if (put_user(len, optlen)) 5017 return -EFAULT; 5018 if (copy_to_user(optval, &prim, len)) 5019 return -EFAULT; 5020 5021 return 0; 5022 } 5023 5024 /* 5025 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER) 5026 * 5027 * Requests that the local endpoint set the specified Adaptation Layer 5028 * Indication parameter for all future INIT and INIT-ACK exchanges. 5029 */ 5030 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len, 5031 char __user *optval, int __user *optlen) 5032 { 5033 struct sctp_setadaptation adaptation; 5034 5035 if (len < sizeof(struct sctp_setadaptation)) 5036 return -EINVAL; 5037 5038 len = sizeof(struct sctp_setadaptation); 5039 5040 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind; 5041 5042 if (put_user(len, optlen)) 5043 return -EFAULT; 5044 if (copy_to_user(optval, &adaptation, len)) 5045 return -EFAULT; 5046 5047 return 0; 5048 } 5049 5050 /* 5051 * 5052 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM) 5053 * 5054 * Applications that wish to use the sendto() system call may wish to 5055 * specify a default set of parameters that would normally be supplied 5056 * through the inclusion of ancillary data. This socket option allows 5057 * such an application to set the default sctp_sndrcvinfo structure. 5058 5059 5060 * The application that wishes to use this socket option simply passes 5061 * in to this call the sctp_sndrcvinfo structure defined in Section 5062 * 5.2.2) The input parameters accepted by this call include 5063 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context, 5064 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in 5065 * to this call if the caller is using the UDP model. 5066 * 5067 * For getsockopt, it get the default sctp_sndrcvinfo structure. 5068 */ 5069 static int sctp_getsockopt_default_send_param(struct sock *sk, 5070 int len, char __user *optval, 5071 int __user *optlen) 5072 { 5073 struct sctp_sock *sp = sctp_sk(sk); 5074 struct sctp_association *asoc; 5075 struct sctp_sndrcvinfo info; 5076 5077 if (len < sizeof(info)) 5078 return -EINVAL; 5079 5080 len = sizeof(info); 5081 5082 if (copy_from_user(&info, optval, len)) 5083 return -EFAULT; 5084 5085 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id); 5086 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP)) 5087 return -EINVAL; 5088 if (asoc) { 5089 info.sinfo_stream = asoc->default_stream; 5090 info.sinfo_flags = asoc->default_flags; 5091 info.sinfo_ppid = asoc->default_ppid; 5092 info.sinfo_context = asoc->default_context; 5093 info.sinfo_timetolive = asoc->default_timetolive; 5094 } else { 5095 info.sinfo_stream = sp->default_stream; 5096 info.sinfo_flags = sp->default_flags; 5097 info.sinfo_ppid = sp->default_ppid; 5098 info.sinfo_context = sp->default_context; 5099 info.sinfo_timetolive = sp->default_timetolive; 5100 } 5101 5102 if (put_user(len, optlen)) 5103 return -EFAULT; 5104 if (copy_to_user(optval, &info, len)) 5105 return -EFAULT; 5106 5107 return 0; 5108 } 5109 5110 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters 5111 * (SCTP_DEFAULT_SNDINFO) 5112 */ 5113 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len, 5114 char __user *optval, 5115 int __user *optlen) 5116 { 5117 struct sctp_sock *sp = sctp_sk(sk); 5118 struct sctp_association *asoc; 5119 struct sctp_sndinfo info; 5120 5121 if (len < sizeof(info)) 5122 return -EINVAL; 5123 5124 len = sizeof(info); 5125 5126 if (copy_from_user(&info, optval, len)) 5127 return -EFAULT; 5128 5129 asoc = sctp_id2assoc(sk, info.snd_assoc_id); 5130 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP)) 5131 return -EINVAL; 5132 if (asoc) { 5133 info.snd_sid = asoc->default_stream; 5134 info.snd_flags = asoc->default_flags; 5135 info.snd_ppid = asoc->default_ppid; 5136 info.snd_context = asoc->default_context; 5137 } else { 5138 info.snd_sid = sp->default_stream; 5139 info.snd_flags = sp->default_flags; 5140 info.snd_ppid = sp->default_ppid; 5141 info.snd_context = sp->default_context; 5142 } 5143 5144 if (put_user(len, optlen)) 5145 return -EFAULT; 5146 if (copy_to_user(optval, &info, len)) 5147 return -EFAULT; 5148 5149 return 0; 5150 } 5151 5152 /* 5153 * 5154 * 7.1.5 SCTP_NODELAY 5155 * 5156 * Turn on/off any Nagle-like algorithm. This means that packets are 5157 * generally sent as soon as possible and no unnecessary delays are 5158 * introduced, at the cost of more packets in the network. Expects an 5159 * integer boolean flag. 5160 */ 5161 5162 static int sctp_getsockopt_nodelay(struct sock *sk, int len, 5163 char __user *optval, int __user *optlen) 5164 { 5165 int val; 5166 5167 if (len < sizeof(int)) 5168 return -EINVAL; 5169 5170 len = sizeof(int); 5171 val = (sctp_sk(sk)->nodelay == 1); 5172 if (put_user(len, optlen)) 5173 return -EFAULT; 5174 if (copy_to_user(optval, &val, len)) 5175 return -EFAULT; 5176 return 0; 5177 } 5178 5179 /* 5180 * 5181 * 7.1.1 SCTP_RTOINFO 5182 * 5183 * The protocol parameters used to initialize and bound retransmission 5184 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access 5185 * and modify these parameters. 5186 * All parameters are time values, in milliseconds. A value of 0, when 5187 * modifying the parameters, indicates that the current value should not 5188 * be changed. 5189 * 5190 */ 5191 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len, 5192 char __user *optval, 5193 int __user *optlen) { 5194 struct sctp_rtoinfo rtoinfo; 5195 struct sctp_association *asoc; 5196 5197 if (len < sizeof (struct sctp_rtoinfo)) 5198 return -EINVAL; 5199 5200 len = sizeof(struct sctp_rtoinfo); 5201 5202 if (copy_from_user(&rtoinfo, optval, len)) 5203 return -EFAULT; 5204 5205 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id); 5206 5207 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP)) 5208 return -EINVAL; 5209 5210 /* Values corresponding to the specific association. */ 5211 if (asoc) { 5212 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial); 5213 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max); 5214 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min); 5215 } else { 5216 /* Values corresponding to the endpoint. */ 5217 struct sctp_sock *sp = sctp_sk(sk); 5218 5219 rtoinfo.srto_initial = sp->rtoinfo.srto_initial; 5220 rtoinfo.srto_max = sp->rtoinfo.srto_max; 5221 rtoinfo.srto_min = sp->rtoinfo.srto_min; 5222 } 5223 5224 if (put_user(len, optlen)) 5225 return -EFAULT; 5226 5227 if (copy_to_user(optval, &rtoinfo, len)) 5228 return -EFAULT; 5229 5230 return 0; 5231 } 5232 5233 /* 5234 * 5235 * 7.1.2 SCTP_ASSOCINFO 5236 * 5237 * This option is used to tune the maximum retransmission attempts 5238 * of the association. 5239 * Returns an error if the new association retransmission value is 5240 * greater than the sum of the retransmission value of the peer. 5241 * See [SCTP] for more information. 5242 * 5243 */ 5244 static int sctp_getsockopt_associnfo(struct sock *sk, int len, 5245 char __user *optval, 5246 int __user *optlen) 5247 { 5248 5249 struct sctp_assocparams assocparams; 5250 struct sctp_association *asoc; 5251 struct list_head *pos; 5252 int cnt = 0; 5253 5254 if (len < sizeof (struct sctp_assocparams)) 5255 return -EINVAL; 5256 5257 len = sizeof(struct sctp_assocparams); 5258 5259 if (copy_from_user(&assocparams, optval, len)) 5260 return -EFAULT; 5261 5262 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id); 5263 5264 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP)) 5265 return -EINVAL; 5266 5267 /* Values correspoinding to the specific association */ 5268 if (asoc) { 5269 assocparams.sasoc_asocmaxrxt = asoc->max_retrans; 5270 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd; 5271 assocparams.sasoc_local_rwnd = asoc->a_rwnd; 5272 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life); 5273 5274 list_for_each(pos, &asoc->peer.transport_addr_list) { 5275 cnt++; 5276 } 5277 5278 assocparams.sasoc_number_peer_destinations = cnt; 5279 } else { 5280 /* Values corresponding to the endpoint */ 5281 struct sctp_sock *sp = sctp_sk(sk); 5282 5283 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt; 5284 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd; 5285 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd; 5286 assocparams.sasoc_cookie_life = 5287 sp->assocparams.sasoc_cookie_life; 5288 assocparams.sasoc_number_peer_destinations = 5289 sp->assocparams. 5290 sasoc_number_peer_destinations; 5291 } 5292 5293 if (put_user(len, optlen)) 5294 return -EFAULT; 5295 5296 if (copy_to_user(optval, &assocparams, len)) 5297 return -EFAULT; 5298 5299 return 0; 5300 } 5301 5302 /* 5303 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR) 5304 * 5305 * This socket option is a boolean flag which turns on or off mapped V4 5306 * addresses. If this option is turned on and the socket is type 5307 * PF_INET6, then IPv4 addresses will be mapped to V6 representation. 5308 * If this option is turned off, then no mapping will be done of V4 5309 * addresses and a user will receive both PF_INET6 and PF_INET type 5310 * addresses on the socket. 5311 */ 5312 static int sctp_getsockopt_mappedv4(struct sock *sk, int len, 5313 char __user *optval, int __user *optlen) 5314 { 5315 int val; 5316 struct sctp_sock *sp = sctp_sk(sk); 5317 5318 if (len < sizeof(int)) 5319 return -EINVAL; 5320 5321 len = sizeof(int); 5322 val = sp->v4mapped; 5323 if (put_user(len, optlen)) 5324 return -EFAULT; 5325 if (copy_to_user(optval, &val, len)) 5326 return -EFAULT; 5327 5328 return 0; 5329 } 5330 5331 /* 5332 * 7.1.29. Set or Get the default context (SCTP_CONTEXT) 5333 * (chapter and verse is quoted at sctp_setsockopt_context()) 5334 */ 5335 static int sctp_getsockopt_context(struct sock *sk, int len, 5336 char __user *optval, int __user *optlen) 5337 { 5338 struct sctp_assoc_value params; 5339 struct sctp_sock *sp; 5340 struct sctp_association *asoc; 5341 5342 if (len < sizeof(struct sctp_assoc_value)) 5343 return -EINVAL; 5344 5345 len = sizeof(struct sctp_assoc_value); 5346 5347 if (copy_from_user(¶ms, optval, len)) 5348 return -EFAULT; 5349 5350 sp = sctp_sk(sk); 5351 5352 if (params.assoc_id != 0) { 5353 asoc = sctp_id2assoc(sk, params.assoc_id); 5354 if (!asoc) 5355 return -EINVAL; 5356 params.assoc_value = asoc->default_rcv_context; 5357 } else { 5358 params.assoc_value = sp->default_rcv_context; 5359 } 5360 5361 if (put_user(len, optlen)) 5362 return -EFAULT; 5363 if (copy_to_user(optval, ¶ms, len)) 5364 return -EFAULT; 5365 5366 return 0; 5367 } 5368 5369 /* 5370 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG) 5371 * This option will get or set the maximum size to put in any outgoing 5372 * SCTP DATA chunk. If a message is larger than this size it will be 5373 * fragmented by SCTP into the specified size. Note that the underlying 5374 * SCTP implementation may fragment into smaller sized chunks when the 5375 * PMTU of the underlying association is smaller than the value set by 5376 * the user. The default value for this option is '0' which indicates 5377 * the user is NOT limiting fragmentation and only the PMTU will effect 5378 * SCTP's choice of DATA chunk size. Note also that values set larger 5379 * than the maximum size of an IP datagram will effectively let SCTP 5380 * control fragmentation (i.e. the same as setting this option to 0). 5381 * 5382 * The following structure is used to access and modify this parameter: 5383 * 5384 * struct sctp_assoc_value { 5385 * sctp_assoc_t assoc_id; 5386 * uint32_t assoc_value; 5387 * }; 5388 * 5389 * assoc_id: This parameter is ignored for one-to-one style sockets. 5390 * For one-to-many style sockets this parameter indicates which 5391 * association the user is performing an action upon. Note that if 5392 * this field's value is zero then the endpoints default value is 5393 * changed (effecting future associations only). 5394 * assoc_value: This parameter specifies the maximum size in bytes. 5395 */ 5396 static int sctp_getsockopt_maxseg(struct sock *sk, int len, 5397 char __user *optval, int __user *optlen) 5398 { 5399 struct sctp_assoc_value params; 5400 struct sctp_association *asoc; 5401 5402 if (len == sizeof(int)) { 5403 pr_warn_ratelimited(DEPRECATED 5404 "%s (pid %d) " 5405 "Use of int in maxseg socket option.\n" 5406 "Use struct sctp_assoc_value instead\n", 5407 current->comm, task_pid_nr(current)); 5408 params.assoc_id = 0; 5409 } else if (len >= sizeof(struct sctp_assoc_value)) { 5410 len = sizeof(struct sctp_assoc_value); 5411 if (copy_from_user(¶ms, optval, sizeof(params))) 5412 return -EFAULT; 5413 } else 5414 return -EINVAL; 5415 5416 asoc = sctp_id2assoc(sk, params.assoc_id); 5417 if (!asoc && params.assoc_id && sctp_style(sk, UDP)) 5418 return -EINVAL; 5419 5420 if (asoc) 5421 params.assoc_value = asoc->frag_point; 5422 else 5423 params.assoc_value = sctp_sk(sk)->user_frag; 5424 5425 if (put_user(len, optlen)) 5426 return -EFAULT; 5427 if (len == sizeof(int)) { 5428 if (copy_to_user(optval, ¶ms.assoc_value, len)) 5429 return -EFAULT; 5430 } else { 5431 if (copy_to_user(optval, ¶ms, len)) 5432 return -EFAULT; 5433 } 5434 5435 return 0; 5436 } 5437 5438 /* 5439 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE) 5440 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave()) 5441 */ 5442 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len, 5443 char __user *optval, int __user *optlen) 5444 { 5445 int val; 5446 5447 if (len < sizeof(int)) 5448 return -EINVAL; 5449 5450 len = sizeof(int); 5451 5452 val = sctp_sk(sk)->frag_interleave; 5453 if (put_user(len, optlen)) 5454 return -EFAULT; 5455 if (copy_to_user(optval, &val, len)) 5456 return -EFAULT; 5457 5458 return 0; 5459 } 5460 5461 /* 5462 * 7.1.25. Set or Get the sctp partial delivery point 5463 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point()) 5464 */ 5465 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len, 5466 char __user *optval, 5467 int __user *optlen) 5468 { 5469 u32 val; 5470 5471 if (len < sizeof(u32)) 5472 return -EINVAL; 5473 5474 len = sizeof(u32); 5475 5476 val = sctp_sk(sk)->pd_point; 5477 if (put_user(len, optlen)) 5478 return -EFAULT; 5479 if (copy_to_user(optval, &val, len)) 5480 return -EFAULT; 5481 5482 return 0; 5483 } 5484 5485 /* 5486 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST) 5487 * (chapter and verse is quoted at sctp_setsockopt_maxburst()) 5488 */ 5489 static int sctp_getsockopt_maxburst(struct sock *sk, int len, 5490 char __user *optval, 5491 int __user *optlen) 5492 { 5493 struct sctp_assoc_value params; 5494 struct sctp_sock *sp; 5495 struct sctp_association *asoc; 5496 5497 if (len == sizeof(int)) { 5498 pr_warn_ratelimited(DEPRECATED 5499 "%s (pid %d) " 5500 "Use of int in max_burst socket option.\n" 5501 "Use struct sctp_assoc_value instead\n", 5502 current->comm, task_pid_nr(current)); 5503 params.assoc_id = 0; 5504 } else if (len >= sizeof(struct sctp_assoc_value)) { 5505 len = sizeof(struct sctp_assoc_value); 5506 if (copy_from_user(¶ms, optval, len)) 5507 return -EFAULT; 5508 } else 5509 return -EINVAL; 5510 5511 sp = sctp_sk(sk); 5512 5513 if (params.assoc_id != 0) { 5514 asoc = sctp_id2assoc(sk, params.assoc_id); 5515 if (!asoc) 5516 return -EINVAL; 5517 params.assoc_value = asoc->max_burst; 5518 } else 5519 params.assoc_value = sp->max_burst; 5520 5521 if (len == sizeof(int)) { 5522 if (copy_to_user(optval, ¶ms.assoc_value, len)) 5523 return -EFAULT; 5524 } else { 5525 if (copy_to_user(optval, ¶ms, len)) 5526 return -EFAULT; 5527 } 5528 5529 return 0; 5530 5531 } 5532 5533 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len, 5534 char __user *optval, int __user *optlen) 5535 { 5536 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 5537 struct sctp_hmacalgo __user *p = (void __user *)optval; 5538 struct sctp_hmac_algo_param *hmacs; 5539 __u16 data_len = 0; 5540 u32 num_idents; 5541 5542 if (!ep->auth_enable) 5543 return -EACCES; 5544 5545 hmacs = ep->auth_hmacs_list; 5546 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t); 5547 5548 if (len < sizeof(struct sctp_hmacalgo) + data_len) 5549 return -EINVAL; 5550 5551 len = sizeof(struct sctp_hmacalgo) + data_len; 5552 num_idents = data_len / sizeof(u16); 5553 5554 if (put_user(len, optlen)) 5555 return -EFAULT; 5556 if (put_user(num_idents, &p->shmac_num_idents)) 5557 return -EFAULT; 5558 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len)) 5559 return -EFAULT; 5560 return 0; 5561 } 5562 5563 static int sctp_getsockopt_active_key(struct sock *sk, int len, 5564 char __user *optval, int __user *optlen) 5565 { 5566 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 5567 struct sctp_authkeyid val; 5568 struct sctp_association *asoc; 5569 5570 if (!ep->auth_enable) 5571 return -EACCES; 5572 5573 if (len < sizeof(struct sctp_authkeyid)) 5574 return -EINVAL; 5575 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid))) 5576 return -EFAULT; 5577 5578 asoc = sctp_id2assoc(sk, val.scact_assoc_id); 5579 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP)) 5580 return -EINVAL; 5581 5582 if (asoc) 5583 val.scact_keynumber = asoc->active_key_id; 5584 else 5585 val.scact_keynumber = ep->active_key_id; 5586 5587 len = sizeof(struct sctp_authkeyid); 5588 if (put_user(len, optlen)) 5589 return -EFAULT; 5590 if (copy_to_user(optval, &val, len)) 5591 return -EFAULT; 5592 5593 return 0; 5594 } 5595 5596 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len, 5597 char __user *optval, int __user *optlen) 5598 { 5599 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 5600 struct sctp_authchunks __user *p = (void __user *)optval; 5601 struct sctp_authchunks val; 5602 struct sctp_association *asoc; 5603 struct sctp_chunks_param *ch; 5604 u32 num_chunks = 0; 5605 char __user *to; 5606 5607 if (!ep->auth_enable) 5608 return -EACCES; 5609 5610 if (len < sizeof(struct sctp_authchunks)) 5611 return -EINVAL; 5612 5613 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks))) 5614 return -EFAULT; 5615 5616 to = p->gauth_chunks; 5617 asoc = sctp_id2assoc(sk, val.gauth_assoc_id); 5618 if (!asoc) 5619 return -EINVAL; 5620 5621 ch = asoc->peer.peer_chunks; 5622 if (!ch) 5623 goto num; 5624 5625 /* See if the user provided enough room for all the data */ 5626 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t); 5627 if (len < num_chunks) 5628 return -EINVAL; 5629 5630 if (copy_to_user(to, ch->chunks, num_chunks)) 5631 return -EFAULT; 5632 num: 5633 len = sizeof(struct sctp_authchunks) + num_chunks; 5634 if (put_user(len, optlen)) 5635 return -EFAULT; 5636 if (put_user(num_chunks, &p->gauth_number_of_chunks)) 5637 return -EFAULT; 5638 return 0; 5639 } 5640 5641 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len, 5642 char __user *optval, int __user *optlen) 5643 { 5644 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 5645 struct sctp_authchunks __user *p = (void __user *)optval; 5646 struct sctp_authchunks val; 5647 struct sctp_association *asoc; 5648 struct sctp_chunks_param *ch; 5649 u32 num_chunks = 0; 5650 char __user *to; 5651 5652 if (!ep->auth_enable) 5653 return -EACCES; 5654 5655 if (len < sizeof(struct sctp_authchunks)) 5656 return -EINVAL; 5657 5658 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks))) 5659 return -EFAULT; 5660 5661 to = p->gauth_chunks; 5662 asoc = sctp_id2assoc(sk, val.gauth_assoc_id); 5663 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP)) 5664 return -EINVAL; 5665 5666 if (asoc) 5667 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks; 5668 else 5669 ch = ep->auth_chunk_list; 5670 5671 if (!ch) 5672 goto num; 5673 5674 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t); 5675 if (len < sizeof(struct sctp_authchunks) + num_chunks) 5676 return -EINVAL; 5677 5678 if (copy_to_user(to, ch->chunks, num_chunks)) 5679 return -EFAULT; 5680 num: 5681 len = sizeof(struct sctp_authchunks) + num_chunks; 5682 if (put_user(len, optlen)) 5683 return -EFAULT; 5684 if (put_user(num_chunks, &p->gauth_number_of_chunks)) 5685 return -EFAULT; 5686 5687 return 0; 5688 } 5689 5690 /* 5691 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER) 5692 * This option gets the current number of associations that are attached 5693 * to a one-to-many style socket. The option value is an uint32_t. 5694 */ 5695 static int sctp_getsockopt_assoc_number(struct sock *sk, int len, 5696 char __user *optval, int __user *optlen) 5697 { 5698 struct sctp_sock *sp = sctp_sk(sk); 5699 struct sctp_association *asoc; 5700 u32 val = 0; 5701 5702 if (sctp_style(sk, TCP)) 5703 return -EOPNOTSUPP; 5704 5705 if (len < sizeof(u32)) 5706 return -EINVAL; 5707 5708 len = sizeof(u32); 5709 5710 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) { 5711 val++; 5712 } 5713 5714 if (put_user(len, optlen)) 5715 return -EFAULT; 5716 if (copy_to_user(optval, &val, len)) 5717 return -EFAULT; 5718 5719 return 0; 5720 } 5721 5722 /* 5723 * 8.1.23 SCTP_AUTO_ASCONF 5724 * See the corresponding setsockopt entry as description 5725 */ 5726 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len, 5727 char __user *optval, int __user *optlen) 5728 { 5729 int val = 0; 5730 5731 if (len < sizeof(int)) 5732 return -EINVAL; 5733 5734 len = sizeof(int); 5735 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk)) 5736 val = 1; 5737 if (put_user(len, optlen)) 5738 return -EFAULT; 5739 if (copy_to_user(optval, &val, len)) 5740 return -EFAULT; 5741 return 0; 5742 } 5743 5744 /* 5745 * 8.2.6. Get the Current Identifiers of Associations 5746 * (SCTP_GET_ASSOC_ID_LIST) 5747 * 5748 * This option gets the current list of SCTP association identifiers of 5749 * the SCTP associations handled by a one-to-many style socket. 5750 */ 5751 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len, 5752 char __user *optval, int __user *optlen) 5753 { 5754 struct sctp_sock *sp = sctp_sk(sk); 5755 struct sctp_association *asoc; 5756 struct sctp_assoc_ids *ids; 5757 u32 num = 0; 5758 5759 if (sctp_style(sk, TCP)) 5760 return -EOPNOTSUPP; 5761 5762 if (len < sizeof(struct sctp_assoc_ids)) 5763 return -EINVAL; 5764 5765 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) { 5766 num++; 5767 } 5768 5769 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num) 5770 return -EINVAL; 5771 5772 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num; 5773 5774 ids = kmalloc(len, GFP_USER | __GFP_NOWARN); 5775 if (unlikely(!ids)) 5776 return -ENOMEM; 5777 5778 ids->gaids_number_of_ids = num; 5779 num = 0; 5780 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) { 5781 ids->gaids_assoc_id[num++] = asoc->assoc_id; 5782 } 5783 5784 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) { 5785 kfree(ids); 5786 return -EFAULT; 5787 } 5788 5789 kfree(ids); 5790 return 0; 5791 } 5792 5793 /* 5794 * SCTP_PEER_ADDR_THLDS 5795 * 5796 * This option allows us to fetch the partially failed threshold for one or all 5797 * transports in an association. See Section 6.1 of: 5798 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt 5799 */ 5800 static int sctp_getsockopt_paddr_thresholds(struct sock *sk, 5801 char __user *optval, 5802 int len, 5803 int __user *optlen) 5804 { 5805 struct sctp_paddrthlds val; 5806 struct sctp_transport *trans; 5807 struct sctp_association *asoc; 5808 5809 if (len < sizeof(struct sctp_paddrthlds)) 5810 return -EINVAL; 5811 len = sizeof(struct sctp_paddrthlds); 5812 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len)) 5813 return -EFAULT; 5814 5815 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) { 5816 asoc = sctp_id2assoc(sk, val.spt_assoc_id); 5817 if (!asoc) 5818 return -ENOENT; 5819 5820 val.spt_pathpfthld = asoc->pf_retrans; 5821 val.spt_pathmaxrxt = asoc->pathmaxrxt; 5822 } else { 5823 trans = sctp_addr_id2transport(sk, &val.spt_address, 5824 val.spt_assoc_id); 5825 if (!trans) 5826 return -ENOENT; 5827 5828 val.spt_pathmaxrxt = trans->pathmaxrxt; 5829 val.spt_pathpfthld = trans->pf_retrans; 5830 } 5831 5832 if (put_user(len, optlen) || copy_to_user(optval, &val, len)) 5833 return -EFAULT; 5834 5835 return 0; 5836 } 5837 5838 /* 5839 * SCTP_GET_ASSOC_STATS 5840 * 5841 * This option retrieves local per endpoint statistics. It is modeled 5842 * after OpenSolaris' implementation 5843 */ 5844 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len, 5845 char __user *optval, 5846 int __user *optlen) 5847 { 5848 struct sctp_assoc_stats sas; 5849 struct sctp_association *asoc = NULL; 5850 5851 /* User must provide at least the assoc id */ 5852 if (len < sizeof(sctp_assoc_t)) 5853 return -EINVAL; 5854 5855 /* Allow the struct to grow and fill in as much as possible */ 5856 len = min_t(size_t, len, sizeof(sas)); 5857 5858 if (copy_from_user(&sas, optval, len)) 5859 return -EFAULT; 5860 5861 asoc = sctp_id2assoc(sk, sas.sas_assoc_id); 5862 if (!asoc) 5863 return -EINVAL; 5864 5865 sas.sas_rtxchunks = asoc->stats.rtxchunks; 5866 sas.sas_gapcnt = asoc->stats.gapcnt; 5867 sas.sas_outofseqtsns = asoc->stats.outofseqtsns; 5868 sas.sas_osacks = asoc->stats.osacks; 5869 sas.sas_isacks = asoc->stats.isacks; 5870 sas.sas_octrlchunks = asoc->stats.octrlchunks; 5871 sas.sas_ictrlchunks = asoc->stats.ictrlchunks; 5872 sas.sas_oodchunks = asoc->stats.oodchunks; 5873 sas.sas_iodchunks = asoc->stats.iodchunks; 5874 sas.sas_ouodchunks = asoc->stats.ouodchunks; 5875 sas.sas_iuodchunks = asoc->stats.iuodchunks; 5876 sas.sas_idupchunks = asoc->stats.idupchunks; 5877 sas.sas_opackets = asoc->stats.opackets; 5878 sas.sas_ipackets = asoc->stats.ipackets; 5879 5880 /* New high max rto observed, will return 0 if not a single 5881 * RTO update took place. obs_rto_ipaddr will be bogus 5882 * in such a case 5883 */ 5884 sas.sas_maxrto = asoc->stats.max_obs_rto; 5885 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr, 5886 sizeof(struct sockaddr_storage)); 5887 5888 /* Mark beginning of a new observation period */ 5889 asoc->stats.max_obs_rto = asoc->rto_min; 5890 5891 if (put_user(len, optlen)) 5892 return -EFAULT; 5893 5894 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id); 5895 5896 if (copy_to_user(optval, &sas, len)) 5897 return -EFAULT; 5898 5899 return 0; 5900 } 5901 5902 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len, 5903 char __user *optval, 5904 int __user *optlen) 5905 { 5906 int val = 0; 5907 5908 if (len < sizeof(int)) 5909 return -EINVAL; 5910 5911 len = sizeof(int); 5912 if (sctp_sk(sk)->recvrcvinfo) 5913 val = 1; 5914 if (put_user(len, optlen)) 5915 return -EFAULT; 5916 if (copy_to_user(optval, &val, len)) 5917 return -EFAULT; 5918 5919 return 0; 5920 } 5921 5922 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len, 5923 char __user *optval, 5924 int __user *optlen) 5925 { 5926 int val = 0; 5927 5928 if (len < sizeof(int)) 5929 return -EINVAL; 5930 5931 len = sizeof(int); 5932 if (sctp_sk(sk)->recvnxtinfo) 5933 val = 1; 5934 if (put_user(len, optlen)) 5935 return -EFAULT; 5936 if (copy_to_user(optval, &val, len)) 5937 return -EFAULT; 5938 5939 return 0; 5940 } 5941 5942 static int sctp_getsockopt(struct sock *sk, int level, int optname, 5943 char __user *optval, int __user *optlen) 5944 { 5945 int retval = 0; 5946 int len; 5947 5948 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname); 5949 5950 /* I can hardly begin to describe how wrong this is. This is 5951 * so broken as to be worse than useless. The API draft 5952 * REALLY is NOT helpful here... I am not convinced that the 5953 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP 5954 * are at all well-founded. 5955 */ 5956 if (level != SOL_SCTP) { 5957 struct sctp_af *af = sctp_sk(sk)->pf->af; 5958 5959 retval = af->getsockopt(sk, level, optname, optval, optlen); 5960 return retval; 5961 } 5962 5963 if (get_user(len, optlen)) 5964 return -EFAULT; 5965 5966 lock_sock(sk); 5967 5968 switch (optname) { 5969 case SCTP_STATUS: 5970 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen); 5971 break; 5972 case SCTP_DISABLE_FRAGMENTS: 5973 retval = sctp_getsockopt_disable_fragments(sk, len, optval, 5974 optlen); 5975 break; 5976 case SCTP_EVENTS: 5977 retval = sctp_getsockopt_events(sk, len, optval, optlen); 5978 break; 5979 case SCTP_AUTOCLOSE: 5980 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen); 5981 break; 5982 case SCTP_SOCKOPT_PEELOFF: 5983 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen); 5984 break; 5985 case SCTP_PEER_ADDR_PARAMS: 5986 retval = sctp_getsockopt_peer_addr_params(sk, len, optval, 5987 optlen); 5988 break; 5989 case SCTP_DELAYED_SACK: 5990 retval = sctp_getsockopt_delayed_ack(sk, len, optval, 5991 optlen); 5992 break; 5993 case SCTP_INITMSG: 5994 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen); 5995 break; 5996 case SCTP_GET_PEER_ADDRS: 5997 retval = sctp_getsockopt_peer_addrs(sk, len, optval, 5998 optlen); 5999 break; 6000 case SCTP_GET_LOCAL_ADDRS: 6001 retval = sctp_getsockopt_local_addrs(sk, len, optval, 6002 optlen); 6003 break; 6004 case SCTP_SOCKOPT_CONNECTX3: 6005 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen); 6006 break; 6007 case SCTP_DEFAULT_SEND_PARAM: 6008 retval = sctp_getsockopt_default_send_param(sk, len, 6009 optval, optlen); 6010 break; 6011 case SCTP_DEFAULT_SNDINFO: 6012 retval = sctp_getsockopt_default_sndinfo(sk, len, 6013 optval, optlen); 6014 break; 6015 case SCTP_PRIMARY_ADDR: 6016 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen); 6017 break; 6018 case SCTP_NODELAY: 6019 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen); 6020 break; 6021 case SCTP_RTOINFO: 6022 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen); 6023 break; 6024 case SCTP_ASSOCINFO: 6025 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen); 6026 break; 6027 case SCTP_I_WANT_MAPPED_V4_ADDR: 6028 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen); 6029 break; 6030 case SCTP_MAXSEG: 6031 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen); 6032 break; 6033 case SCTP_GET_PEER_ADDR_INFO: 6034 retval = sctp_getsockopt_peer_addr_info(sk, len, optval, 6035 optlen); 6036 break; 6037 case SCTP_ADAPTATION_LAYER: 6038 retval = sctp_getsockopt_adaptation_layer(sk, len, optval, 6039 optlen); 6040 break; 6041 case SCTP_CONTEXT: 6042 retval = sctp_getsockopt_context(sk, len, optval, optlen); 6043 break; 6044 case SCTP_FRAGMENT_INTERLEAVE: 6045 retval = sctp_getsockopt_fragment_interleave(sk, len, optval, 6046 optlen); 6047 break; 6048 case SCTP_PARTIAL_DELIVERY_POINT: 6049 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval, 6050 optlen); 6051 break; 6052 case SCTP_MAX_BURST: 6053 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen); 6054 break; 6055 case SCTP_AUTH_KEY: 6056 case SCTP_AUTH_CHUNK: 6057 case SCTP_AUTH_DELETE_KEY: 6058 retval = -EOPNOTSUPP; 6059 break; 6060 case SCTP_HMAC_IDENT: 6061 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen); 6062 break; 6063 case SCTP_AUTH_ACTIVE_KEY: 6064 retval = sctp_getsockopt_active_key(sk, len, optval, optlen); 6065 break; 6066 case SCTP_PEER_AUTH_CHUNKS: 6067 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval, 6068 optlen); 6069 break; 6070 case SCTP_LOCAL_AUTH_CHUNKS: 6071 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval, 6072 optlen); 6073 break; 6074 case SCTP_GET_ASSOC_NUMBER: 6075 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen); 6076 break; 6077 case SCTP_GET_ASSOC_ID_LIST: 6078 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen); 6079 break; 6080 case SCTP_AUTO_ASCONF: 6081 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen); 6082 break; 6083 case SCTP_PEER_ADDR_THLDS: 6084 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen); 6085 break; 6086 case SCTP_GET_ASSOC_STATS: 6087 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen); 6088 break; 6089 case SCTP_RECVRCVINFO: 6090 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen); 6091 break; 6092 case SCTP_RECVNXTINFO: 6093 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen); 6094 break; 6095 default: 6096 retval = -ENOPROTOOPT; 6097 break; 6098 } 6099 6100 release_sock(sk); 6101 return retval; 6102 } 6103 6104 static void sctp_hash(struct sock *sk) 6105 { 6106 /* STUB */ 6107 } 6108 6109 static void sctp_unhash(struct sock *sk) 6110 { 6111 /* STUB */ 6112 } 6113 6114 /* Check if port is acceptable. Possibly find first available port. 6115 * 6116 * The port hash table (contained in the 'global' SCTP protocol storage 6117 * returned by struct sctp_protocol *sctp_get_protocol()). The hash 6118 * table is an array of 4096 lists (sctp_bind_hashbucket). Each 6119 * list (the list number is the port number hashed out, so as you 6120 * would expect from a hash function, all the ports in a given list have 6121 * such a number that hashes out to the same list number; you were 6122 * expecting that, right?); so each list has a set of ports, with a 6123 * link to the socket (struct sock) that uses it, the port number and 6124 * a fastreuse flag (FIXME: NPI ipg). 6125 */ 6126 static struct sctp_bind_bucket *sctp_bucket_create( 6127 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum); 6128 6129 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr) 6130 { 6131 struct sctp_bind_hashbucket *head; /* hash list */ 6132 struct sctp_bind_bucket *pp; 6133 unsigned short snum; 6134 int ret; 6135 6136 snum = ntohs(addr->v4.sin_port); 6137 6138 pr_debug("%s: begins, snum:%d\n", __func__, snum); 6139 6140 local_bh_disable(); 6141 6142 if (snum == 0) { 6143 /* Search for an available port. */ 6144 int low, high, remaining, index; 6145 unsigned int rover; 6146 struct net *net = sock_net(sk); 6147 6148 inet_get_local_port_range(net, &low, &high); 6149 remaining = (high - low) + 1; 6150 rover = prandom_u32() % remaining + low; 6151 6152 do { 6153 rover++; 6154 if ((rover < low) || (rover > high)) 6155 rover = low; 6156 if (inet_is_local_reserved_port(net, rover)) 6157 continue; 6158 index = sctp_phashfn(sock_net(sk), rover); 6159 head = &sctp_port_hashtable[index]; 6160 spin_lock(&head->lock); 6161 sctp_for_each_hentry(pp, &head->chain) 6162 if ((pp->port == rover) && 6163 net_eq(sock_net(sk), pp->net)) 6164 goto next; 6165 break; 6166 next: 6167 spin_unlock(&head->lock); 6168 } while (--remaining > 0); 6169 6170 /* Exhausted local port range during search? */ 6171 ret = 1; 6172 if (remaining <= 0) 6173 goto fail; 6174 6175 /* OK, here is the one we will use. HEAD (the port 6176 * hash table list entry) is non-NULL and we hold it's 6177 * mutex. 6178 */ 6179 snum = rover; 6180 } else { 6181 /* We are given an specific port number; we verify 6182 * that it is not being used. If it is used, we will 6183 * exahust the search in the hash list corresponding 6184 * to the port number (snum) - we detect that with the 6185 * port iterator, pp being NULL. 6186 */ 6187 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)]; 6188 spin_lock(&head->lock); 6189 sctp_for_each_hentry(pp, &head->chain) { 6190 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk))) 6191 goto pp_found; 6192 } 6193 } 6194 pp = NULL; 6195 goto pp_not_found; 6196 pp_found: 6197 if (!hlist_empty(&pp->owner)) { 6198 /* We had a port hash table hit - there is an 6199 * available port (pp != NULL) and it is being 6200 * used by other socket (pp->owner not empty); that other 6201 * socket is going to be sk2. 6202 */ 6203 int reuse = sk->sk_reuse; 6204 struct sock *sk2; 6205 6206 pr_debug("%s: found a possible match\n", __func__); 6207 6208 if (pp->fastreuse && sk->sk_reuse && 6209 sk->sk_state != SCTP_SS_LISTENING) 6210 goto success; 6211 6212 /* Run through the list of sockets bound to the port 6213 * (pp->port) [via the pointers bind_next and 6214 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one, 6215 * we get the endpoint they describe and run through 6216 * the endpoint's list of IP (v4 or v6) addresses, 6217 * comparing each of the addresses with the address of 6218 * the socket sk. If we find a match, then that means 6219 * that this port/socket (sk) combination are already 6220 * in an endpoint. 6221 */ 6222 sk_for_each_bound(sk2, &pp->owner) { 6223 struct sctp_endpoint *ep2; 6224 ep2 = sctp_sk(sk2)->ep; 6225 6226 if (sk == sk2 || 6227 (reuse && sk2->sk_reuse && 6228 sk2->sk_state != SCTP_SS_LISTENING)) 6229 continue; 6230 6231 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr, 6232 sctp_sk(sk2), sctp_sk(sk))) { 6233 ret = (long)sk2; 6234 goto fail_unlock; 6235 } 6236 } 6237 6238 pr_debug("%s: found a match\n", __func__); 6239 } 6240 pp_not_found: 6241 /* If there was a hash table miss, create a new port. */ 6242 ret = 1; 6243 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum))) 6244 goto fail_unlock; 6245 6246 /* In either case (hit or miss), make sure fastreuse is 1 only 6247 * if sk->sk_reuse is too (that is, if the caller requested 6248 * SO_REUSEADDR on this socket -sk-). 6249 */ 6250 if (hlist_empty(&pp->owner)) { 6251 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING) 6252 pp->fastreuse = 1; 6253 else 6254 pp->fastreuse = 0; 6255 } else if (pp->fastreuse && 6256 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING)) 6257 pp->fastreuse = 0; 6258 6259 /* We are set, so fill up all the data in the hash table 6260 * entry, tie the socket list information with the rest of the 6261 * sockets FIXME: Blurry, NPI (ipg). 6262 */ 6263 success: 6264 if (!sctp_sk(sk)->bind_hash) { 6265 inet_sk(sk)->inet_num = snum; 6266 sk_add_bind_node(sk, &pp->owner); 6267 sctp_sk(sk)->bind_hash = pp; 6268 } 6269 ret = 0; 6270 6271 fail_unlock: 6272 spin_unlock(&head->lock); 6273 6274 fail: 6275 local_bh_enable(); 6276 return ret; 6277 } 6278 6279 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral 6280 * port is requested. 6281 */ 6282 static int sctp_get_port(struct sock *sk, unsigned short snum) 6283 { 6284 union sctp_addr addr; 6285 struct sctp_af *af = sctp_sk(sk)->pf->af; 6286 6287 /* Set up a dummy address struct from the sk. */ 6288 af->from_sk(&addr, sk); 6289 addr.v4.sin_port = htons(snum); 6290 6291 /* Note: sk->sk_num gets filled in if ephemeral port request. */ 6292 return !!sctp_get_port_local(sk, &addr); 6293 } 6294 6295 /* 6296 * Move a socket to LISTENING state. 6297 */ 6298 static int sctp_listen_start(struct sock *sk, int backlog) 6299 { 6300 struct sctp_sock *sp = sctp_sk(sk); 6301 struct sctp_endpoint *ep = sp->ep; 6302 struct crypto_hash *tfm = NULL; 6303 char alg[32]; 6304 6305 /* Allocate HMAC for generating cookie. */ 6306 if (!sp->hmac && sp->sctp_hmac_alg) { 6307 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg); 6308 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC); 6309 if (IS_ERR(tfm)) { 6310 net_info_ratelimited("failed to load transform for %s: %ld\n", 6311 sp->sctp_hmac_alg, PTR_ERR(tfm)); 6312 return -ENOSYS; 6313 } 6314 sctp_sk(sk)->hmac = tfm; 6315 } 6316 6317 /* 6318 * If a bind() or sctp_bindx() is not called prior to a listen() 6319 * call that allows new associations to be accepted, the system 6320 * picks an ephemeral port and will choose an address set equivalent 6321 * to binding with a wildcard address. 6322 * 6323 * This is not currently spelled out in the SCTP sockets 6324 * extensions draft, but follows the practice as seen in TCP 6325 * sockets. 6326 * 6327 */ 6328 sk->sk_state = SCTP_SS_LISTENING; 6329 if (!ep->base.bind_addr.port) { 6330 if (sctp_autobind(sk)) 6331 return -EAGAIN; 6332 } else { 6333 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) { 6334 sk->sk_state = SCTP_SS_CLOSED; 6335 return -EADDRINUSE; 6336 } 6337 } 6338 6339 sk->sk_max_ack_backlog = backlog; 6340 sctp_hash_endpoint(ep); 6341 return 0; 6342 } 6343 6344 /* 6345 * 4.1.3 / 5.1.3 listen() 6346 * 6347 * By default, new associations are not accepted for UDP style sockets. 6348 * An application uses listen() to mark a socket as being able to 6349 * accept new associations. 6350 * 6351 * On TCP style sockets, applications use listen() to ready the SCTP 6352 * endpoint for accepting inbound associations. 6353 * 6354 * On both types of endpoints a backlog of '0' disables listening. 6355 * 6356 * Move a socket to LISTENING state. 6357 */ 6358 int sctp_inet_listen(struct socket *sock, int backlog) 6359 { 6360 struct sock *sk = sock->sk; 6361 struct sctp_endpoint *ep = sctp_sk(sk)->ep; 6362 int err = -EINVAL; 6363 6364 if (unlikely(backlog < 0)) 6365 return err; 6366 6367 lock_sock(sk); 6368 6369 /* Peeled-off sockets are not allowed to listen(). */ 6370 if (sctp_style(sk, UDP_HIGH_BANDWIDTH)) 6371 goto out; 6372 6373 if (sock->state != SS_UNCONNECTED) 6374 goto out; 6375 6376 /* If backlog is zero, disable listening. */ 6377 if (!backlog) { 6378 if (sctp_sstate(sk, CLOSED)) 6379 goto out; 6380 6381 err = 0; 6382 sctp_unhash_endpoint(ep); 6383 sk->sk_state = SCTP_SS_CLOSED; 6384 if (sk->sk_reuse) 6385 sctp_sk(sk)->bind_hash->fastreuse = 1; 6386 goto out; 6387 } 6388 6389 /* If we are already listening, just update the backlog */ 6390 if (sctp_sstate(sk, LISTENING)) 6391 sk->sk_max_ack_backlog = backlog; 6392 else { 6393 err = sctp_listen_start(sk, backlog); 6394 if (err) 6395 goto out; 6396 } 6397 6398 err = 0; 6399 out: 6400 release_sock(sk); 6401 return err; 6402 } 6403 6404 /* 6405 * This function is done by modeling the current datagram_poll() and the 6406 * tcp_poll(). Note that, based on these implementations, we don't 6407 * lock the socket in this function, even though it seems that, 6408 * ideally, locking or some other mechanisms can be used to ensure 6409 * the integrity of the counters (sndbuf and wmem_alloc) used 6410 * in this place. We assume that we don't need locks either until proven 6411 * otherwise. 6412 * 6413 * Another thing to note is that we include the Async I/O support 6414 * here, again, by modeling the current TCP/UDP code. We don't have 6415 * a good way to test with it yet. 6416 */ 6417 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait) 6418 { 6419 struct sock *sk = sock->sk; 6420 struct sctp_sock *sp = sctp_sk(sk); 6421 unsigned int mask; 6422 6423 poll_wait(file, sk_sleep(sk), wait); 6424 6425 /* A TCP-style listening socket becomes readable when the accept queue 6426 * is not empty. 6427 */ 6428 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) 6429 return (!list_empty(&sp->ep->asocs)) ? 6430 (POLLIN | POLLRDNORM) : 0; 6431 6432 mask = 0; 6433 6434 /* Is there any exceptional events? */ 6435 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) 6436 mask |= POLLERR | 6437 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0); 6438 if (sk->sk_shutdown & RCV_SHUTDOWN) 6439 mask |= POLLRDHUP | POLLIN | POLLRDNORM; 6440 if (sk->sk_shutdown == SHUTDOWN_MASK) 6441 mask |= POLLHUP; 6442 6443 /* Is it readable? Reconsider this code with TCP-style support. */ 6444 if (!skb_queue_empty(&sk->sk_receive_queue)) 6445 mask |= POLLIN | POLLRDNORM; 6446 6447 /* The association is either gone or not ready. */ 6448 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED)) 6449 return mask; 6450 6451 /* Is it writable? */ 6452 if (sctp_writeable(sk)) { 6453 mask |= POLLOUT | POLLWRNORM; 6454 } else { 6455 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk); 6456 /* 6457 * Since the socket is not locked, the buffer 6458 * might be made available after the writeable check and 6459 * before the bit is set. This could cause a lost I/O 6460 * signal. tcp_poll() has a race breaker for this race 6461 * condition. Based on their implementation, we put 6462 * in the following code to cover it as well. 6463 */ 6464 if (sctp_writeable(sk)) 6465 mask |= POLLOUT | POLLWRNORM; 6466 } 6467 return mask; 6468 } 6469 6470 /******************************************************************** 6471 * 2nd Level Abstractions 6472 ********************************************************************/ 6473 6474 static struct sctp_bind_bucket *sctp_bucket_create( 6475 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum) 6476 { 6477 struct sctp_bind_bucket *pp; 6478 6479 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC); 6480 if (pp) { 6481 SCTP_DBG_OBJCNT_INC(bind_bucket); 6482 pp->port = snum; 6483 pp->fastreuse = 0; 6484 INIT_HLIST_HEAD(&pp->owner); 6485 pp->net = net; 6486 hlist_add_head(&pp->node, &head->chain); 6487 } 6488 return pp; 6489 } 6490 6491 /* Caller must hold hashbucket lock for this tb with local BH disabled */ 6492 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp) 6493 { 6494 if (pp && hlist_empty(&pp->owner)) { 6495 __hlist_del(&pp->node); 6496 kmem_cache_free(sctp_bucket_cachep, pp); 6497 SCTP_DBG_OBJCNT_DEC(bind_bucket); 6498 } 6499 } 6500 6501 /* Release this socket's reference to a local port. */ 6502 static inline void __sctp_put_port(struct sock *sk) 6503 { 6504 struct sctp_bind_hashbucket *head = 6505 &sctp_port_hashtable[sctp_phashfn(sock_net(sk), 6506 inet_sk(sk)->inet_num)]; 6507 struct sctp_bind_bucket *pp; 6508 6509 spin_lock(&head->lock); 6510 pp = sctp_sk(sk)->bind_hash; 6511 __sk_del_bind_node(sk); 6512 sctp_sk(sk)->bind_hash = NULL; 6513 inet_sk(sk)->inet_num = 0; 6514 sctp_bucket_destroy(pp); 6515 spin_unlock(&head->lock); 6516 } 6517 6518 void sctp_put_port(struct sock *sk) 6519 { 6520 local_bh_disable(); 6521 __sctp_put_port(sk); 6522 local_bh_enable(); 6523 } 6524 6525 /* 6526 * The system picks an ephemeral port and choose an address set equivalent 6527 * to binding with a wildcard address. 6528 * One of those addresses will be the primary address for the association. 6529 * This automatically enables the multihoming capability of SCTP. 6530 */ 6531 static int sctp_autobind(struct sock *sk) 6532 { 6533 union sctp_addr autoaddr; 6534 struct sctp_af *af; 6535 __be16 port; 6536 6537 /* Initialize a local sockaddr structure to INADDR_ANY. */ 6538 af = sctp_sk(sk)->pf->af; 6539 6540 port = htons(inet_sk(sk)->inet_num); 6541 af->inaddr_any(&autoaddr, port); 6542 6543 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len); 6544 } 6545 6546 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation. 6547 * 6548 * From RFC 2292 6549 * 4.2 The cmsghdr Structure * 6550 * 6551 * When ancillary data is sent or received, any number of ancillary data 6552 * objects can be specified by the msg_control and msg_controllen members of 6553 * the msghdr structure, because each object is preceded by 6554 * a cmsghdr structure defining the object's length (the cmsg_len member). 6555 * Historically Berkeley-derived implementations have passed only one object 6556 * at a time, but this API allows multiple objects to be 6557 * passed in a single call to sendmsg() or recvmsg(). The following example 6558 * shows two ancillary data objects in a control buffer. 6559 * 6560 * |<--------------------------- msg_controllen -------------------------->| 6561 * | | 6562 * 6563 * |<----- ancillary data object ----->|<----- ancillary data object ----->| 6564 * 6565 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->| 6566 * | | | 6567 * 6568 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| | 6569 * 6570 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| | 6571 * | | | | | 6572 * 6573 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+ 6574 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX| 6575 * 6576 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX| 6577 * 6578 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+ 6579 * ^ 6580 * | 6581 * 6582 * msg_control 6583 * points here 6584 */ 6585 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs) 6586 { 6587 struct cmsghdr *cmsg; 6588 struct msghdr *my_msg = (struct msghdr *)msg; 6589 6590 for_each_cmsghdr(cmsg, my_msg) { 6591 if (!CMSG_OK(my_msg, cmsg)) 6592 return -EINVAL; 6593 6594 /* Should we parse this header or ignore? */ 6595 if (cmsg->cmsg_level != IPPROTO_SCTP) 6596 continue; 6597 6598 /* Strictly check lengths following example in SCM code. */ 6599 switch (cmsg->cmsg_type) { 6600 case SCTP_INIT: 6601 /* SCTP Socket API Extension 6602 * 5.3.1 SCTP Initiation Structure (SCTP_INIT) 6603 * 6604 * This cmsghdr structure provides information for 6605 * initializing new SCTP associations with sendmsg(). 6606 * The SCTP_INITMSG socket option uses this same data 6607 * structure. This structure is not used for 6608 * recvmsg(). 6609 * 6610 * cmsg_level cmsg_type cmsg_data[] 6611 * ------------ ------------ ---------------------- 6612 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg 6613 */ 6614 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg))) 6615 return -EINVAL; 6616 6617 cmsgs->init = CMSG_DATA(cmsg); 6618 break; 6619 6620 case SCTP_SNDRCV: 6621 /* SCTP Socket API Extension 6622 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV) 6623 * 6624 * This cmsghdr structure specifies SCTP options for 6625 * sendmsg() and describes SCTP header information 6626 * about a received message through recvmsg(). 6627 * 6628 * cmsg_level cmsg_type cmsg_data[] 6629 * ------------ ------------ ---------------------- 6630 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo 6631 */ 6632 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo))) 6633 return -EINVAL; 6634 6635 cmsgs->srinfo = CMSG_DATA(cmsg); 6636 6637 if (cmsgs->srinfo->sinfo_flags & 6638 ~(SCTP_UNORDERED | SCTP_ADDR_OVER | 6639 SCTP_SACK_IMMEDIATELY | 6640 SCTP_ABORT | SCTP_EOF)) 6641 return -EINVAL; 6642 break; 6643 6644 case SCTP_SNDINFO: 6645 /* SCTP Socket API Extension 6646 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO) 6647 * 6648 * This cmsghdr structure specifies SCTP options for 6649 * sendmsg(). This structure and SCTP_RCVINFO replaces 6650 * SCTP_SNDRCV which has been deprecated. 6651 * 6652 * cmsg_level cmsg_type cmsg_data[] 6653 * ------------ ------------ --------------------- 6654 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo 6655 */ 6656 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo))) 6657 return -EINVAL; 6658 6659 cmsgs->sinfo = CMSG_DATA(cmsg); 6660 6661 if (cmsgs->sinfo->snd_flags & 6662 ~(SCTP_UNORDERED | SCTP_ADDR_OVER | 6663 SCTP_SACK_IMMEDIATELY | 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 (skwq_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 security_sk_clone(sk, newsk); 7204 } 7205 7206 static inline void sctp_copy_descendant(struct sock *sk_to, 7207 const struct sock *sk_from) 7208 { 7209 int ancestor_size = sizeof(struct inet_sock) + 7210 sizeof(struct sctp_sock) - 7211 offsetof(struct sctp_sock, auto_asconf_list); 7212 7213 if (sk_from->sk_family == PF_INET6) 7214 ancestor_size += sizeof(struct ipv6_pinfo); 7215 7216 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size); 7217 } 7218 7219 /* Populate the fields of the newsk from the oldsk and migrate the assoc 7220 * and its messages to the newsk. 7221 */ 7222 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk, 7223 struct sctp_association *assoc, 7224 sctp_socket_type_t type) 7225 { 7226 struct sctp_sock *oldsp = sctp_sk(oldsk); 7227 struct sctp_sock *newsp = sctp_sk(newsk); 7228 struct sctp_bind_bucket *pp; /* hash list port iterator */ 7229 struct sctp_endpoint *newep = newsp->ep; 7230 struct sk_buff *skb, *tmp; 7231 struct sctp_ulpevent *event; 7232 struct sctp_bind_hashbucket *head; 7233 7234 /* Migrate socket buffer sizes and all the socket level options to the 7235 * new socket. 7236 */ 7237 newsk->sk_sndbuf = oldsk->sk_sndbuf; 7238 newsk->sk_rcvbuf = oldsk->sk_rcvbuf; 7239 /* Brute force copy old sctp opt. */ 7240 sctp_copy_descendant(newsk, oldsk); 7241 7242 /* Restore the ep value that was overwritten with the above structure 7243 * copy. 7244 */ 7245 newsp->ep = newep; 7246 newsp->hmac = NULL; 7247 7248 /* Hook this new socket in to the bind_hash list. */ 7249 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk), 7250 inet_sk(oldsk)->inet_num)]; 7251 local_bh_disable(); 7252 spin_lock(&head->lock); 7253 pp = sctp_sk(oldsk)->bind_hash; 7254 sk_add_bind_node(newsk, &pp->owner); 7255 sctp_sk(newsk)->bind_hash = pp; 7256 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num; 7257 spin_unlock(&head->lock); 7258 local_bh_enable(); 7259 7260 /* Copy the bind_addr list from the original endpoint to the new 7261 * endpoint so that we can handle restarts properly 7262 */ 7263 sctp_bind_addr_dup(&newsp->ep->base.bind_addr, 7264 &oldsp->ep->base.bind_addr, GFP_KERNEL); 7265 7266 /* Move any messages in the old socket's receive queue that are for the 7267 * peeled off association to the new socket's receive queue. 7268 */ 7269 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) { 7270 event = sctp_skb2event(skb); 7271 if (event->asoc == assoc) { 7272 __skb_unlink(skb, &oldsk->sk_receive_queue); 7273 __skb_queue_tail(&newsk->sk_receive_queue, skb); 7274 sctp_skb_set_owner_r_frag(skb, newsk); 7275 } 7276 } 7277 7278 /* Clean up any messages pending delivery due to partial 7279 * delivery. Three cases: 7280 * 1) No partial deliver; no work. 7281 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby. 7282 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue. 7283 */ 7284 skb_queue_head_init(&newsp->pd_lobby); 7285 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode); 7286 7287 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) { 7288 struct sk_buff_head *queue; 7289 7290 /* Decide which queue to move pd_lobby skbs to. */ 7291 if (assoc->ulpq.pd_mode) { 7292 queue = &newsp->pd_lobby; 7293 } else 7294 queue = &newsk->sk_receive_queue; 7295 7296 /* Walk through the pd_lobby, looking for skbs that 7297 * need moved to the new socket. 7298 */ 7299 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) { 7300 event = sctp_skb2event(skb); 7301 if (event->asoc == assoc) { 7302 __skb_unlink(skb, &oldsp->pd_lobby); 7303 __skb_queue_tail(queue, skb); 7304 sctp_skb_set_owner_r_frag(skb, newsk); 7305 } 7306 } 7307 7308 /* Clear up any skbs waiting for the partial 7309 * delivery to finish. 7310 */ 7311 if (assoc->ulpq.pd_mode) 7312 sctp_clear_pd(oldsk, NULL); 7313 7314 } 7315 7316 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) 7317 sctp_skb_set_owner_r_frag(skb, newsk); 7318 7319 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) 7320 sctp_skb_set_owner_r_frag(skb, newsk); 7321 7322 /* Set the type of socket to indicate that it is peeled off from the 7323 * original UDP-style socket or created with the accept() call on a 7324 * TCP-style socket.. 7325 */ 7326 newsp->type = type; 7327 7328 /* Mark the new socket "in-use" by the user so that any packets 7329 * that may arrive on the association after we've moved it are 7330 * queued to the backlog. This prevents a potential race between 7331 * backlog processing on the old socket and new-packet processing 7332 * on the new socket. 7333 * 7334 * The caller has just allocated newsk so we can guarantee that other 7335 * paths won't try to lock it and then oldsk. 7336 */ 7337 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING); 7338 sctp_assoc_migrate(assoc, newsk); 7339 7340 /* If the association on the newsk is already closed before accept() 7341 * is called, set RCV_SHUTDOWN flag. 7342 */ 7343 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) 7344 newsk->sk_shutdown |= RCV_SHUTDOWN; 7345 7346 newsk->sk_state = SCTP_SS_ESTABLISHED; 7347 release_sock(newsk); 7348 } 7349 7350 7351 /* This proto struct describes the ULP interface for SCTP. */ 7352 struct proto sctp_prot = { 7353 .name = "SCTP", 7354 .owner = THIS_MODULE, 7355 .close = sctp_close, 7356 .connect = sctp_connect, 7357 .disconnect = sctp_disconnect, 7358 .accept = sctp_accept, 7359 .ioctl = sctp_ioctl, 7360 .init = sctp_init_sock, 7361 .destroy = sctp_destroy_sock, 7362 .shutdown = sctp_shutdown, 7363 .setsockopt = sctp_setsockopt, 7364 .getsockopt = sctp_getsockopt, 7365 .sendmsg = sctp_sendmsg, 7366 .recvmsg = sctp_recvmsg, 7367 .bind = sctp_bind, 7368 .backlog_rcv = sctp_backlog_rcv, 7369 .hash = sctp_hash, 7370 .unhash = sctp_unhash, 7371 .get_port = sctp_get_port, 7372 .obj_size = sizeof(struct sctp_sock), 7373 .sysctl_mem = sysctl_sctp_mem, 7374 .sysctl_rmem = sysctl_sctp_rmem, 7375 .sysctl_wmem = sysctl_sctp_wmem, 7376 .memory_pressure = &sctp_memory_pressure, 7377 .enter_memory_pressure = sctp_enter_memory_pressure, 7378 .memory_allocated = &sctp_memory_allocated, 7379 .sockets_allocated = &sctp_sockets_allocated, 7380 }; 7381 7382 #if IS_ENABLED(CONFIG_IPV6) 7383 7384 #include <net/transp_v6.h> 7385 static void sctp_v6_destroy_sock(struct sock *sk) 7386 { 7387 sctp_destroy_sock(sk); 7388 inet6_destroy_sock(sk); 7389 } 7390 7391 struct proto sctpv6_prot = { 7392 .name = "SCTPv6", 7393 .owner = THIS_MODULE, 7394 .close = sctp_close, 7395 .connect = sctp_connect, 7396 .disconnect = sctp_disconnect, 7397 .accept = sctp_accept, 7398 .ioctl = sctp_ioctl, 7399 .init = sctp_init_sock, 7400 .destroy = sctp_v6_destroy_sock, 7401 .shutdown = sctp_shutdown, 7402 .setsockopt = sctp_setsockopt, 7403 .getsockopt = sctp_getsockopt, 7404 .sendmsg = sctp_sendmsg, 7405 .recvmsg = sctp_recvmsg, 7406 .bind = sctp_bind, 7407 .backlog_rcv = sctp_backlog_rcv, 7408 .hash = sctp_hash, 7409 .unhash = sctp_unhash, 7410 .get_port = sctp_get_port, 7411 .obj_size = sizeof(struct sctp6_sock), 7412 .sysctl_mem = sysctl_sctp_mem, 7413 .sysctl_rmem = sysctl_sctp_rmem, 7414 .sysctl_wmem = sysctl_sctp_wmem, 7415 .memory_pressure = &sctp_memory_pressure, 7416 .enter_memory_pressure = sctp_enter_memory_pressure, 7417 .memory_allocated = &sctp_memory_allocated, 7418 .sockets_allocated = &sctp_sockets_allocated, 7419 }; 7420 #endif /* IS_ENABLED(CONFIG_IPV6) */ 7421