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