1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #pragma ident "%Z%%M% %I% %E% SMI" 28 29 #include <sys/types.h> 30 #include <sys/systm.h> 31 #include <sys/stream.h> 32 #include <sys/cmn_err.h> 33 #include <sys/kmem.h> 34 #define _SUN_TPI_VERSION 2 35 #include <sys/tihdr.h> 36 #include <sys/stropts.h> 37 #include <sys/socket.h> 38 #include <sys/random.h> 39 #include <sys/policy.h> 40 41 #include <netinet/in.h> 42 #include <netinet/ip6.h> 43 44 #include <inet/common.h> 45 #include <inet/ip.h> 46 #include <inet/ip6.h> 47 #include <inet/ipclassifier.h> 48 #include "sctp_impl.h" 49 #include "sctp_asconf.h" 50 #include "sctp_addr.h" 51 52 uint_t sctp_next_port_to_try; 53 54 /* 55 * Returns 0 on success, EACCES on permission failure. 56 */ 57 static int 58 sctp_select_port(sctp_t *sctp, in_port_t *requested_port, int *user_specified) 59 { 60 /* 61 * Get a valid port (within the anonymous range and should not 62 * be a privileged one) to use if the user has not given a port. 63 * If multiple threads are here, they may all start with 64 * with the same initial port. But, it should be fine as long as 65 * sctp_bindi will ensure that no two threads will be assigned 66 * the same port. 67 */ 68 if (*requested_port == 0) { 69 *requested_port = sctp_update_next_port(sctp_next_port_to_try); 70 *user_specified = 0; 71 } else { 72 int i; 73 boolean_t priv = B_FALSE; 74 75 /* 76 * If the requested_port is in the well-known privileged range, 77 * verify that the stream was opened by a privileged user. 78 * Note: No locks are held when inspecting sctp_g_*epriv_ports 79 * but instead the code relies on: 80 * - the fact that the address of the array and its size never 81 * changes 82 * - the atomic assignment of the elements of the array 83 */ 84 if (*requested_port < sctp_smallest_nonpriv_port) { 85 priv = B_TRUE; 86 } else { 87 for (i = 0; i < sctp_g_num_epriv_ports; i++) { 88 if (*requested_port == sctp_g_epriv_ports[i]) { 89 priv = B_TRUE; 90 break; 91 } 92 } 93 } 94 if (priv) { 95 /* 96 * sctp_bind() should take a cred_t argument so that 97 * we can use it here. 98 */ 99 if (secpolicy_net_privaddr(sctp->sctp_credp, 100 *requested_port) != 0) { 101 dprint(1, 102 ("sctp_bind(x): no prive for port %d", 103 *requested_port)); 104 return (TACCES); 105 } 106 } 107 *user_specified = 1; 108 } 109 110 return (0); 111 } 112 113 int 114 sctp_listen(sctp_t *sctp) 115 { 116 sctp_tf_t *tf; 117 118 RUN_SCTP(sctp); 119 /* 120 * TCP handles listen() increasing the backlog, need to check 121 * if it should be handled here too - VENU. 122 */ 123 if (sctp->sctp_state > SCTPS_BOUND) { 124 WAKE_SCTP(sctp); 125 return (EINVAL); 126 } 127 128 /* Do an anonymous bind for unbound socket doing listen(). */ 129 if (sctp->sctp_nsaddrs == 0) { 130 struct sockaddr_storage ss; 131 int ret; 132 133 bzero(&ss, sizeof (ss)); 134 ss.ss_family = sctp->sctp_family; 135 136 WAKE_SCTP(sctp); 137 if ((ret = sctp_bind(sctp, (struct sockaddr *)&ss, 138 sizeof (ss))) != 0) 139 return (ret); 140 RUN_SCTP(sctp) 141 } 142 143 sctp->sctp_state = SCTPS_LISTEN; 144 (void) random_get_pseudo_bytes(sctp->sctp_secret, SCTP_SECRET_LEN); 145 sctp->sctp_last_secret_update = lbolt64; 146 bzero(sctp->sctp_old_secret, SCTP_SECRET_LEN); 147 tf = &sctp_listen_fanout[SCTP_LISTEN_HASH(ntohs(sctp->sctp_lport))]; 148 sctp_listen_hash_insert(tf, sctp); 149 150 WAKE_SCTP(sctp); 151 return (0); 152 } 153 154 /* 155 * Bind the sctp_t to a sockaddr, which includes an address and other 156 * information, such as port or flowinfo. 157 */ 158 int 159 sctp_bind(sctp_t *sctp, struct sockaddr *sa, socklen_t len) 160 { 161 int user_specified; 162 boolean_t bind_to_req_port_only; 163 in_port_t requested_port; 164 in_port_t allocated_port; 165 int err = 0; 166 167 ASSERT(sctp != NULL); 168 ASSERT(sa); 169 170 RUN_SCTP(sctp); 171 172 if (sctp->sctp_state > SCTPS_BOUND) { 173 err = EINVAL; 174 goto done; 175 } 176 177 switch (sa->sa_family) { 178 case AF_INET: 179 if (len < sizeof (struct sockaddr_in) || 180 sctp->sctp_family == AF_INET6) { 181 err = EINVAL; 182 goto done; 183 } 184 requested_port = ntohs(((struct sockaddr_in *)sa)->sin_port); 185 break; 186 case AF_INET6: 187 if (len < sizeof (struct sockaddr_in6) || 188 sctp->sctp_family == AF_INET) { 189 err = EINVAL; 190 goto done; 191 } 192 requested_port = ntohs(((struct sockaddr_in6 *)sa)->sin6_port); 193 /* Set the flowinfo. */ 194 sctp->sctp_ip6h->ip6_vcf = 195 (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) | 196 (((struct sockaddr_in6 *)sa)->sin6_flowinfo & 197 ~IPV6_VERS_AND_FLOW_MASK); 198 break; 199 default: 200 err = EAFNOSUPPORT; 201 goto done; 202 } 203 bind_to_req_port_only = requested_port == 0 ? B_FALSE : B_TRUE; 204 205 if (sctp_select_port(sctp, &requested_port, &user_specified) != 0) { 206 err = EPERM; 207 goto done; 208 } 209 210 if ((err = sctp_bind_add(sctp, sa, 1, B_TRUE)) != 0) 211 goto done; 212 213 allocated_port = sctp_bindi(sctp, requested_port, 214 bind_to_req_port_only, user_specified); 215 if (allocated_port == 0) { 216 sctp_free_saddrs(sctp); 217 if (bind_to_req_port_only) { 218 err = EADDRINUSE; 219 goto done; 220 } else { 221 err = EADDRNOTAVAIL; 222 goto done; 223 } 224 } 225 ASSERT(sctp->sctp_state == SCTPS_BOUND); 226 done: 227 WAKE_SCTP(sctp); 228 return (err); 229 } 230 231 /* 232 * Perform bind/unbind operation of a list of addresses on a sctp_t 233 */ 234 int 235 sctp_bindx(sctp_t *sctp, const void *addrs, int addrcnt, int bindop) 236 { 237 ASSERT(sctp != NULL); 238 ASSERT(addrs != NULL); 239 ASSERT(addrcnt > 0); 240 241 switch (bindop) { 242 case SCTP_BINDX_ADD_ADDR: 243 return (sctp_bind_add(sctp, addrs, addrcnt, B_FALSE)); 244 case SCTP_BINDX_REM_ADDR: 245 return (sctp_bind_del(sctp, addrs, addrcnt, B_FALSE)); 246 default: 247 return (EINVAL); 248 } 249 } 250 251 /* 252 * Add a list of addresses to a sctp_t. 253 */ 254 int 255 sctp_bind_add(sctp_t *sctp, const void *addrs, uint32_t addrcnt, 256 boolean_t caller_hold_lock) 257 { 258 int err = 0; 259 boolean_t do_asconf = B_FALSE; 260 261 if (!caller_hold_lock) 262 RUN_SCTP(sctp); 263 264 if (sctp->sctp_state > SCTPS_ESTABLISHED) { 265 if (!caller_hold_lock) 266 WAKE_SCTP(sctp); 267 return (EINVAL); 268 } 269 270 if (sctp->sctp_state > SCTPS_LISTEN) { 271 /* 272 * Let's do some checking here rather than undoing the 273 * add later (for these reasons). 274 */ 275 if (!sctp_addip_enabled || !sctp->sctp_understands_asconf || 276 !sctp->sctp_understands_addip) { 277 if (!caller_hold_lock) 278 WAKE_SCTP(sctp); 279 return (EINVAL); 280 } 281 do_asconf = B_TRUE; 282 } 283 err = sctp_valid_addr_list(sctp, addrs, addrcnt); 284 if (err != 0) { 285 if (!caller_hold_lock) 286 WAKE_SCTP(sctp); 287 return (err); 288 } 289 290 /* Need to send ASCONF messages */ 291 if (do_asconf) { 292 err = sctp_add_ip(sctp, addrs, addrcnt); 293 if (err != 0) { 294 sctp_del_saddr_list(sctp, addrs, addrcnt, B_FALSE); 295 if (!caller_hold_lock) 296 WAKE_SCTP(sctp); 297 return (err); 298 } 299 } 300 if (!caller_hold_lock) 301 WAKE_SCTP(sctp); 302 if (do_asconf) 303 sctp_process_sendq(sctp); 304 return (0); 305 } 306 307 /* 308 * Remove one or more addresses bound to the sctp_t. 309 */ 310 int 311 sctp_bind_del(sctp_t *sctp, const void *addrs, uint32_t addrcnt, 312 boolean_t caller_hold_lock) 313 { 314 int error = 0; 315 boolean_t do_asconf = B_FALSE; 316 317 if (!caller_hold_lock) 318 RUN_SCTP(sctp); 319 320 if (sctp->sctp_state > SCTPS_ESTABLISHED) { 321 if (!caller_hold_lock) 322 WAKE_SCTP(sctp); 323 return (EINVAL); 324 } 325 /* 326 * Fail the remove if we are beyond listen, but can't send this 327 * to the peer. 328 */ 329 if (sctp->sctp_state > SCTPS_LISTEN) { 330 if (!sctp_addip_enabled || !sctp->sctp_understands_asconf || 331 !sctp->sctp_understands_addip) { 332 if (!caller_hold_lock) 333 WAKE_SCTP(sctp); 334 return (EINVAL); 335 } 336 do_asconf = B_TRUE; 337 } 338 339 /* Can't delete the last address nor all of the addresses */ 340 if (sctp->sctp_nsaddrs == 1 || addrcnt >= sctp->sctp_nsaddrs) { 341 if (!caller_hold_lock) 342 WAKE_SCTP(sctp); 343 return (EINVAL); 344 } 345 346 error = sctp_del_ip(sctp, addrs, addrcnt); 347 if (!caller_hold_lock) 348 WAKE_SCTP(sctp); 349 if (error == 0 && do_asconf) 350 sctp_process_sendq(sctp); 351 return (error); 352 } 353 354 /* 355 * If the "bind_to_req_port_only" parameter is set, if the requested port 356 * number is available, return it, If not return 0 357 * 358 * If "bind_to_req_port_only" parameter is not set and 359 * If the requested port number is available, return it. If not, return 360 * the first anonymous port we happen across. If no anonymous ports are 361 * available, return 0. addr is the requested local address, if any. 362 * 363 * In either case, when succeeding update the sctp_t to record the port number 364 * and insert it in the bind hash table. 365 */ 366 in_port_t 367 sctp_bindi(sctp_t *sctp, in_port_t port, int bind_to_req_port_only, 368 int user_specified) 369 { 370 /* number of times we have run around the loop */ 371 int count = 0; 372 /* maximum number of times to run around the loop */ 373 int loopmax; 374 zoneid_t zoneid = sctp->sctp_zoneid; 375 376 /* 377 * Lookup for free addresses is done in a loop and "loopmax" 378 * influences how long we spin in the loop 379 */ 380 if (bind_to_req_port_only) { 381 /* 382 * If the requested port is busy, don't bother to look 383 * for a new one. Setting loop maximum count to 1 has 384 * that effect. 385 */ 386 loopmax = 1; 387 } else { 388 /* 389 * If the requested port is busy, look for a free one 390 * in the anonymous port range. 391 * Set loopmax appropriately so that one does not look 392 * forever in the case all of the anonymous ports are in use. 393 */ 394 loopmax = (sctp_largest_anon_port - 395 sctp_smallest_anon_port + 1); 396 } 397 do { 398 uint16_t lport; 399 sctp_tf_t *tbf; 400 sctp_t *lsctp; 401 int addrcmp; 402 403 lport = htons(port); 404 405 /* 406 * Ensure that the sctp_t is not currently in the bind hash. 407 * Hold the lock on the hash bucket to ensure that 408 * the duplicate check plus the insertion is an atomic 409 * operation. 410 * 411 * This function does an inline lookup on the bind hash list 412 * Make sure that we access only members of sctp_t 413 * and that we don't look at sctp_sctp, since we are not 414 * doing a SCTPB_REFHOLD. For more details please see the notes 415 * in sctp_compress() 416 */ 417 sctp_bind_hash_remove(sctp); 418 tbf = &sctp_bind_fanout[SCTP_BIND_HASH(port)]; 419 mutex_enter(&tbf->tf_lock); 420 for (lsctp = tbf->tf_sctp; lsctp != NULL; 421 lsctp = lsctp->sctp_bind_hash) { 422 423 if (lport != lsctp->sctp_lport || 424 lsctp->sctp_zoneid != zoneid || 425 lsctp->sctp_state < SCTPS_BOUND) 426 continue; 427 428 addrcmp = sctp_compare_saddrs(sctp, lsctp); 429 if (addrcmp != SCTP_ADDR_DISJOINT) { 430 if (!sctp->sctp_reuseaddr) { 431 /* in use */ 432 break; 433 } else if (lsctp->sctp_state == SCTPS_BOUND || 434 lsctp->sctp_state == SCTPS_LISTEN) { 435 /* 436 * socket option SO_REUSEADDR is set 437 * on the binding sctp_t. 438 * 439 * We have found a match of IP source 440 * address and source port, which is 441 * refused regardless of the 442 * SO_REUSEADDR setting, so we break. 443 */ 444 break; 445 } 446 } 447 } 448 if (lsctp != NULL) { 449 /* The port number is busy */ 450 mutex_exit(&tbf->tf_lock); 451 } else { 452 /* 453 * This port is ours. Insert in fanout and mark as 454 * bound to prevent others from getting the port 455 * number. 456 */ 457 sctp->sctp_state = SCTPS_BOUND; 458 sctp->sctp_lport = lport; 459 sctp->sctp_sctph->sh_sport = sctp->sctp_lport; 460 461 ASSERT(&sctp_bind_fanout[SCTP_BIND_HASH(port)] == tbf); 462 sctp_bind_hash_insert(tbf, sctp, 1); 463 464 mutex_exit(&tbf->tf_lock); 465 466 /* 467 * We don't want sctp_next_port_to_try to "inherit" 468 * a port number supplied by the user in a bind. 469 */ 470 if (user_specified != 0) 471 return (port); 472 473 /* 474 * This is the only place where sctp_next_port_to_try 475 * is updated. After the update, it may or may not 476 * be in the valid range. 477 */ 478 sctp_next_port_to_try = port + 1; 479 return (port); 480 } 481 482 if ((count == 0) && (user_specified)) { 483 /* 484 * We may have to return an anonymous port. So 485 * get one to start with. 486 */ 487 port = sctp_update_next_port(sctp_next_port_to_try); 488 user_specified = 0; 489 } else { 490 port = sctp_update_next_port(port + 1); 491 } 492 493 /* 494 * Don't let this loop run forever in the case where 495 * all of the anonymous ports are in use. 496 */ 497 } while (++count < loopmax); 498 return (0); 499 } 500 501 /* 502 * Don't let port fall into the privileged range. 503 * Since the extra privileged ports can be arbitrary we also 504 * ensure that we exclude those from consideration. 505 * sctp_g_epriv_ports is not sorted thus we loop over it until 506 * there are no changes. 507 * 508 * Note: No locks are held when inspecting sctp_g_*epriv_ports 509 * but instead the code relies on: 510 * - the fact that the address of the array and its size never changes 511 * - the atomic assignment of the elements of the array 512 */ 513 in_port_t 514 sctp_update_next_port(in_port_t port) 515 { 516 int i; 517 518 retry: 519 if (port < sctp_smallest_anon_port || port > sctp_largest_anon_port) 520 port = sctp_smallest_anon_port; 521 522 if (port < sctp_smallest_nonpriv_port) 523 port = sctp_smallest_nonpriv_port; 524 525 for (i = 0; i < sctp_g_num_epriv_ports; i++) { 526 if (port == sctp_g_epriv_ports[i]) { 527 port++; 528 /* 529 * Make sure whether the port is in the 530 * valid range. 531 * 532 * XXX Note that if sctp_g_epriv_ports contains 533 * all the anonymous ports this will be an 534 * infinite loop. 535 */ 536 goto retry; 537 } 538 } 539 return (port); 540 } 541