1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. 5 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. 6 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions are met: 10 * 11 * a) Redistributions of source code must retain the above copyright notice, 12 * this list of conditions and the following disclaimer. 13 * 14 * b) Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in 16 * the documentation and/or other materials provided with the distribution. 17 * 18 * c) Neither the name of Cisco Systems, Inc. nor the names of its 19 * contributors may be used to endorse or promote products derived 20 * from this software without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, 24 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 26 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF 32 * THE POSSIBILITY OF SUCH DAMAGE. 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD$"); 37 38 #include <netinet/sctp_os.h> 39 #include <sys/proc.h> 40 #include <netinet/sctp_var.h> 41 #include <netinet/sctp_sysctl.h> 42 #include <netinet/sctp_pcb.h> 43 #include <netinet/sctputil.h> 44 #include <netinet/sctp.h> 45 #include <netinet/sctp_header.h> 46 #include <netinet/sctp_asconf.h> 47 #include <netinet/sctp_output.h> 48 #include <netinet/sctp_timer.h> 49 #include <netinet/sctp_bsd_addr.h> 50 #if defined(INET) || defined(INET6) 51 #include <netinet/udp.h> 52 #endif 53 #ifdef INET6 54 #include <netinet6/ip6_var.h> 55 #endif 56 #include <sys/sched.h> 57 #include <sys/smp.h> 58 #include <sys/unistd.h> 59 60 VNET_DEFINE(struct sctp_base_info, system_base_info); 61 62 /* FIX: we don't handle multiple link local scopes */ 63 /* "scopeless" replacement IN6_ARE_ADDR_EQUAL */ 64 #ifdef INET6 65 int 66 SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b) 67 { 68 struct sockaddr_in6 tmp_a, tmp_b; 69 70 memcpy(&tmp_a, a, sizeof(struct sockaddr_in6)); 71 if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) { 72 return (0); 73 } 74 memcpy(&tmp_b, b, sizeof(struct sockaddr_in6)); 75 if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) { 76 return (0); 77 } 78 return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr)); 79 } 80 #endif 81 82 void 83 sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb) 84 { 85 /* 86 * We really don't need to lock this, but I will just because it 87 * does not hurt. 88 */ 89 SCTP_INP_INFO_RLOCK(); 90 spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep); 91 spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc); 92 spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr); 93 spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr); 94 spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk); 95 spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq); 96 spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq); 97 spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks); 98 SCTP_INP_INFO_RUNLOCK(); 99 } 100 101 /*- 102 * Addresses are added to VRF's (Virtual Router's). For BSD we 103 * have only the default VRF 0. We maintain a hash list of 104 * VRF's. Each VRF has its own list of sctp_ifn's. Each of 105 * these has a list of addresses. When we add a new address 106 * to a VRF we lookup the ifn/ifn_index, if the ifn does 107 * not exist we create it and add it to the list of IFN's 108 * within the VRF. Once we have the sctp_ifn, we add the 109 * address to the list. So we look something like: 110 * 111 * hash-vrf-table 112 * vrf-> ifn-> ifn -> ifn 113 * vrf | 114 * ... +--ifa-> ifa -> ifa 115 * vrf 116 * 117 * We keep these separate lists since the SCTP subsystem will 118 * point to these from its source address selection nets structure. 119 * When an address is deleted it does not happen right away on 120 * the SCTP side, it gets scheduled. What we do when a 121 * delete happens is immediately remove the address from 122 * the master list and decrement the refcount. As our 123 * addip iterator works through and frees the src address 124 * selection pointing to the sctp_ifa, eventually the refcount 125 * will reach 0 and we will delete it. Note that it is assumed 126 * that any locking on system level ifn/ifa is done at the 127 * caller of these functions and these routines will only 128 * lock the SCTP structures as they add or delete things. 129 * 130 * Other notes on VRF concepts. 131 * - An endpoint can be in multiple VRF's 132 * - An association lives within a VRF and only one VRF. 133 * - Any incoming packet we can deduce the VRF for by 134 * looking at the mbuf/pak inbound (for BSD its VRF=0 :D) 135 * - Any downward send call or connect call must supply the 136 * VRF via ancillary data or via some sort of set default 137 * VRF socket option call (again for BSD no brainer since 138 * the VRF is always 0). 139 * - An endpoint may add multiple VRF's to it. 140 * - Listening sockets can accept associations in any 141 * of the VRF's they are in but the assoc will end up 142 * in only one VRF (gotten from the packet or connect/send). 143 * 144 */ 145 146 struct sctp_vrf * 147 sctp_allocate_vrf(int vrf_id) 148 { 149 struct sctp_vrf *vrf = NULL; 150 struct sctp_vrflist *bucket; 151 152 /* First allocate the VRF structure */ 153 vrf = sctp_find_vrf(vrf_id); 154 if (vrf) { 155 /* Already allocated */ 156 return (vrf); 157 } 158 SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf), 159 SCTP_M_VRF); 160 if (vrf == NULL) { 161 /* No memory */ 162 #ifdef INVARIANTS 163 panic("No memory for VRF:%d", vrf_id); 164 #endif 165 return (NULL); 166 } 167 /* setup the VRF */ 168 memset(vrf, 0, sizeof(struct sctp_vrf)); 169 vrf->vrf_id = vrf_id; 170 LIST_INIT(&vrf->ifnlist); 171 vrf->total_ifa_count = 0; 172 vrf->refcount = 0; 173 /* now also setup table ids */ 174 SCTP_INIT_VRF_TABLEID(vrf); 175 /* Init the HASH of addresses */ 176 vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE, 177 &vrf->vrf_addr_hashmark); 178 if (vrf->vrf_addr_hash == NULL) { 179 /* No memory */ 180 #ifdef INVARIANTS 181 panic("No memory for VRF:%d", vrf_id); 182 #endif 183 SCTP_FREE(vrf, SCTP_M_VRF); 184 return (NULL); 185 } 186 187 /* Add it to the hash table */ 188 bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; 189 LIST_INSERT_HEAD(bucket, vrf, next_vrf); 190 atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); 191 return (vrf); 192 } 193 194 195 struct sctp_ifn * 196 sctp_find_ifn(void *ifn, uint32_t ifn_index) 197 { 198 struct sctp_ifn *sctp_ifnp; 199 struct sctp_ifnlist *hash_ifn_head; 200 201 /* 202 * We assume the lock is held for the addresses if that's wrong 203 * problems could occur :-) 204 */ 205 hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; 206 LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) { 207 if (sctp_ifnp->ifn_index == ifn_index) { 208 return (sctp_ifnp); 209 } 210 if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) { 211 return (sctp_ifnp); 212 } 213 } 214 return (NULL); 215 } 216 217 218 struct sctp_vrf * 219 sctp_find_vrf(uint32_t vrf_id) 220 { 221 struct sctp_vrflist *bucket; 222 struct sctp_vrf *liste; 223 224 bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; 225 LIST_FOREACH(liste, bucket, next_vrf) { 226 if (vrf_id == liste->vrf_id) { 227 return (liste); 228 } 229 } 230 return (NULL); 231 } 232 233 234 void 235 sctp_free_vrf(struct sctp_vrf *vrf) 236 { 237 if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) { 238 if (vrf->vrf_addr_hash) { 239 SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); 240 vrf->vrf_addr_hash = NULL; 241 } 242 /* We zero'd the count */ 243 LIST_REMOVE(vrf, next_vrf); 244 SCTP_FREE(vrf, SCTP_M_VRF); 245 atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); 246 } 247 } 248 249 250 void 251 sctp_free_ifn(struct sctp_ifn *sctp_ifnp) 252 { 253 if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) { 254 /* We zero'd the count */ 255 if (sctp_ifnp->vrf) { 256 sctp_free_vrf(sctp_ifnp->vrf); 257 } 258 SCTP_FREE(sctp_ifnp, SCTP_M_IFN); 259 atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); 260 } 261 } 262 263 264 void 265 sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu) 266 { 267 struct sctp_ifn *sctp_ifnp; 268 269 sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index); 270 if (sctp_ifnp != NULL) { 271 sctp_ifnp->ifn_mtu = mtu; 272 } 273 } 274 275 276 void 277 sctp_free_ifa(struct sctp_ifa *sctp_ifap) 278 { 279 if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) { 280 /* We zero'd the count */ 281 if (sctp_ifap->ifn_p) { 282 sctp_free_ifn(sctp_ifap->ifn_p); 283 } 284 SCTP_FREE(sctp_ifap, SCTP_M_IFA); 285 atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); 286 } 287 } 288 289 290 static void 291 sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock) 292 { 293 struct sctp_ifn *found; 294 295 found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index); 296 if (found == NULL) { 297 /* Not in the list.. sorry */ 298 return; 299 } 300 if (hold_addr_lock == 0) 301 SCTP_IPI_ADDR_WLOCK(); 302 LIST_REMOVE(sctp_ifnp, next_bucket); 303 LIST_REMOVE(sctp_ifnp, next_ifn); 304 if (hold_addr_lock == 0) 305 SCTP_IPI_ADDR_WUNLOCK(); 306 /* Take away the reference, and possibly free it */ 307 sctp_free_ifn(sctp_ifnp); 308 } 309 310 311 void 312 sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr, 313 const char *if_name, uint32_t ifn_index) 314 { 315 struct sctp_vrf *vrf; 316 struct sctp_ifa *sctp_ifap; 317 318 SCTP_IPI_ADDR_RLOCK(); 319 vrf = sctp_find_vrf(vrf_id); 320 if (vrf == NULL) { 321 SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); 322 goto out; 323 324 } 325 sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); 326 if (sctp_ifap == NULL) { 327 SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); 328 goto out; 329 } 330 if (sctp_ifap->ifn_p == NULL) { 331 SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n"); 332 goto out; 333 } 334 if (if_name) { 335 if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) { 336 SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", 337 sctp_ifap->ifn_p->ifn_name, if_name); 338 goto out; 339 } 340 } else { 341 if (sctp_ifap->ifn_p->ifn_index != ifn_index) { 342 SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", 343 sctp_ifap->ifn_p->ifn_index, ifn_index); 344 goto out; 345 } 346 } 347 348 sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID); 349 sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; 350 out: 351 SCTP_IPI_ADDR_RUNLOCK(); 352 } 353 354 355 void 356 sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr, 357 const char *if_name, uint32_t ifn_index) 358 { 359 struct sctp_vrf *vrf; 360 struct sctp_ifa *sctp_ifap; 361 362 SCTP_IPI_ADDR_RLOCK(); 363 vrf = sctp_find_vrf(vrf_id); 364 if (vrf == NULL) { 365 SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); 366 goto out; 367 368 } 369 sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); 370 if (sctp_ifap == NULL) { 371 SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); 372 goto out; 373 } 374 if (sctp_ifap->ifn_p == NULL) { 375 SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n"); 376 goto out; 377 } 378 if (if_name) { 379 if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) { 380 SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", 381 sctp_ifap->ifn_p->ifn_name, if_name); 382 goto out; 383 } 384 } else { 385 if (sctp_ifap->ifn_p->ifn_index != ifn_index) { 386 SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", 387 sctp_ifap->ifn_p->ifn_index, ifn_index); 388 goto out; 389 } 390 } 391 392 sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE); 393 sctp_ifap->localifa_flags |= SCTP_ADDR_VALID; 394 out: 395 SCTP_IPI_ADDR_RUNLOCK(); 396 } 397 398 399 /*- 400 * Add an ifa to an ifn. 401 * Register the interface as necessary. 402 * NOTE: ADDR write lock MUST be held. 403 */ 404 static void 405 sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap) 406 { 407 int ifa_af; 408 409 LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); 410 sctp_ifap->ifn_p = sctp_ifnp; 411 atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); 412 /* update address counts */ 413 sctp_ifnp->ifa_count++; 414 ifa_af = sctp_ifap->address.sa.sa_family; 415 switch (ifa_af) { 416 #ifdef INET 417 case AF_INET: 418 sctp_ifnp->num_v4++; 419 break; 420 #endif 421 #ifdef INET6 422 case AF_INET6: 423 sctp_ifnp->num_v6++; 424 break; 425 #endif 426 default: 427 break; 428 } 429 if (sctp_ifnp->ifa_count == 1) { 430 /* register the new interface */ 431 sctp_ifnp->registered_af = ifa_af; 432 } 433 } 434 435 436 /*- 437 * Remove an ifa from its ifn. 438 * If no more addresses exist, remove the ifn too. Otherwise, re-register 439 * the interface based on the remaining address families left. 440 * NOTE: ADDR write lock MUST be held. 441 */ 442 static void 443 sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap) 444 { 445 LIST_REMOVE(sctp_ifap, next_ifa); 446 if (sctp_ifap->ifn_p) { 447 /* update address counts */ 448 sctp_ifap->ifn_p->ifa_count--; 449 switch (sctp_ifap->address.sa.sa_family) { 450 #ifdef INET 451 case AF_INET: 452 sctp_ifap->ifn_p->num_v4--; 453 break; 454 #endif 455 #ifdef INET6 456 case AF_INET6: 457 sctp_ifap->ifn_p->num_v6--; 458 break; 459 #endif 460 default: 461 break; 462 } 463 464 if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) { 465 /* remove the ifn, possibly freeing it */ 466 sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED); 467 } else { 468 /* re-register address family type, if needed */ 469 if ((sctp_ifap->ifn_p->num_v6 == 0) && 470 (sctp_ifap->ifn_p->registered_af == AF_INET6)) { 471 sctp_ifap->ifn_p->registered_af = AF_INET; 472 } else if ((sctp_ifap->ifn_p->num_v4 == 0) && 473 (sctp_ifap->ifn_p->registered_af == AF_INET)) { 474 sctp_ifap->ifn_p->registered_af = AF_INET6; 475 } 476 /* free the ifn refcount */ 477 sctp_free_ifn(sctp_ifap->ifn_p); 478 } 479 sctp_ifap->ifn_p = NULL; 480 } 481 } 482 483 484 struct sctp_ifa * 485 sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index, 486 uint32_t ifn_type, const char *if_name, void *ifa, 487 struct sockaddr *addr, uint32_t ifa_flags, 488 int dynamic_add) 489 { 490 struct sctp_vrf *vrf; 491 struct sctp_ifn *sctp_ifnp = NULL; 492 struct sctp_ifa *sctp_ifap = NULL; 493 struct sctp_ifalist *hash_addr_head; 494 struct sctp_ifnlist *hash_ifn_head; 495 uint32_t hash_of_addr; 496 int new_ifn_af = 0; 497 498 #ifdef SCTP_DEBUG 499 SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id); 500 SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); 501 #endif 502 SCTP_IPI_ADDR_WLOCK(); 503 sctp_ifnp = sctp_find_ifn(ifn, ifn_index); 504 if (sctp_ifnp) { 505 vrf = sctp_ifnp->vrf; 506 } else { 507 vrf = sctp_find_vrf(vrf_id); 508 if (vrf == NULL) { 509 vrf = sctp_allocate_vrf(vrf_id); 510 if (vrf == NULL) { 511 SCTP_IPI_ADDR_WUNLOCK(); 512 return (NULL); 513 } 514 } 515 } 516 if (sctp_ifnp == NULL) { 517 /* 518 * build one and add it, can't hold lock until after malloc 519 * done though. 520 */ 521 SCTP_IPI_ADDR_WUNLOCK(); 522 SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *, 523 sizeof(struct sctp_ifn), SCTP_M_IFN); 524 if (sctp_ifnp == NULL) { 525 #ifdef INVARIANTS 526 panic("No memory for IFN"); 527 #endif 528 return (NULL); 529 } 530 memset(sctp_ifnp, 0, sizeof(struct sctp_ifn)); 531 sctp_ifnp->ifn_index = ifn_index; 532 sctp_ifnp->ifn_p = ifn; 533 sctp_ifnp->ifn_type = ifn_type; 534 sctp_ifnp->refcount = 0; 535 sctp_ifnp->vrf = vrf; 536 atomic_add_int(&vrf->refcount, 1); 537 sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family); 538 if (if_name != NULL) { 539 SCTP_SNPRINTF(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name); 540 } else { 541 SCTP_SNPRINTF(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown"); 542 } 543 hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; 544 LIST_INIT(&sctp_ifnp->ifalist); 545 SCTP_IPI_ADDR_WLOCK(); 546 LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket); 547 LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn); 548 atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); 549 new_ifn_af = 1; 550 } 551 sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); 552 if (sctp_ifap) { 553 /* Hmm, it already exists? */ 554 if ((sctp_ifap->ifn_p) && 555 (sctp_ifap->ifn_p->ifn_index == ifn_index)) { 556 SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n", 557 sctp_ifap->ifn_p->ifn_name, ifn_index, 558 (void *)sctp_ifap); 559 if (new_ifn_af) { 560 /* Remove the created one that we don't want */ 561 sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED); 562 } 563 if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) { 564 /* easy to solve, just switch back to active */ 565 SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n"); 566 sctp_ifap->localifa_flags = SCTP_ADDR_VALID; 567 sctp_ifap->ifn_p = sctp_ifnp; 568 atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); 569 } 570 exit_stage_left: 571 SCTP_IPI_ADDR_WUNLOCK(); 572 return (sctp_ifap); 573 } else { 574 if (sctp_ifap->ifn_p) { 575 /* 576 * The last IFN gets the address, remove the 577 * old one 578 */ 579 SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n", 580 (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name, 581 sctp_ifap->ifn_p->ifn_index, if_name, 582 ifn_index); 583 /* remove the address from the old ifn */ 584 sctp_remove_ifa_from_ifn(sctp_ifap); 585 /* move the address over to the new ifn */ 586 sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); 587 goto exit_stage_left; 588 } else { 589 /* repair ifnp which was NULL ? */ 590 sctp_ifap->localifa_flags = SCTP_ADDR_VALID; 591 SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n", 592 (void *)sctp_ifnp, (void *)sctp_ifap); 593 sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); 594 } 595 goto exit_stage_left; 596 } 597 } 598 SCTP_IPI_ADDR_WUNLOCK(); 599 SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA); 600 if (sctp_ifap == NULL) { 601 #ifdef INVARIANTS 602 panic("No memory for IFA"); 603 #endif 604 return (NULL); 605 } 606 memset(sctp_ifap, 0, sizeof(struct sctp_ifa)); 607 sctp_ifap->ifn_p = sctp_ifnp; 608 atomic_add_int(&sctp_ifnp->refcount, 1); 609 sctp_ifap->vrf_id = vrf_id; 610 sctp_ifap->ifa = ifa; 611 memcpy(&sctp_ifap->address, addr, addr->sa_len); 612 sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE; 613 sctp_ifap->flags = ifa_flags; 614 /* Set scope */ 615 switch (sctp_ifap->address.sa.sa_family) { 616 #ifdef INET 617 case AF_INET: 618 { 619 struct sockaddr_in *sin; 620 621 sin = &sctp_ifap->address.sin; 622 if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || 623 (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) { 624 sctp_ifap->src_is_loop = 1; 625 } 626 if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { 627 sctp_ifap->src_is_priv = 1; 628 } 629 sctp_ifnp->num_v4++; 630 if (new_ifn_af) 631 new_ifn_af = AF_INET; 632 break; 633 } 634 #endif 635 #ifdef INET6 636 case AF_INET6: 637 { 638 /* ok to use deprecated addresses? */ 639 struct sockaddr_in6 *sin6; 640 641 sin6 = &sctp_ifap->address.sin6; 642 if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || 643 (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) { 644 sctp_ifap->src_is_loop = 1; 645 } 646 if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { 647 sctp_ifap->src_is_priv = 1; 648 } 649 sctp_ifnp->num_v6++; 650 if (new_ifn_af) 651 new_ifn_af = AF_INET6; 652 break; 653 } 654 #endif 655 default: 656 new_ifn_af = 0; 657 break; 658 } 659 hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa); 660 661 if ((sctp_ifap->src_is_priv == 0) && 662 (sctp_ifap->src_is_loop == 0)) { 663 sctp_ifap->src_is_glob = 1; 664 } 665 SCTP_IPI_ADDR_WLOCK(); 666 hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)]; 667 LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket); 668 sctp_ifap->refcount = 1; 669 LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); 670 sctp_ifnp->ifa_count++; 671 vrf->total_ifa_count++; 672 atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); 673 if (new_ifn_af) { 674 sctp_ifnp->registered_af = new_ifn_af; 675 } 676 SCTP_IPI_ADDR_WUNLOCK(); 677 if (dynamic_add) { 678 /* 679 * Bump up the refcount so that when the timer completes it 680 * will drop back down. 681 */ 682 struct sctp_laddr *wi; 683 684 atomic_add_int(&sctp_ifap->refcount, 1); 685 wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); 686 if (wi == NULL) { 687 /* 688 * Gak, what can we do? We have lost an address 689 * change can you say HOSED? 690 */ 691 SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); 692 /* Opps, must decrement the count */ 693 sctp_del_addr_from_vrf(vrf_id, addr, ifn_index, 694 if_name); 695 return (NULL); 696 } 697 SCTP_INCR_LADDR_COUNT(); 698 memset(wi, 0, sizeof(*wi)); 699 (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); 700 wi->ifa = sctp_ifap; 701 wi->action = SCTP_ADD_IP_ADDRESS; 702 703 SCTP_WQ_ADDR_LOCK(); 704 LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); 705 sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, 706 (struct sctp_inpcb *)NULL, 707 (struct sctp_tcb *)NULL, 708 (struct sctp_nets *)NULL); 709 SCTP_WQ_ADDR_UNLOCK(); 710 } else { 711 /* it's ready for use */ 712 sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE; 713 } 714 return (sctp_ifap); 715 } 716 717 void 718 sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr, 719 uint32_t ifn_index, const char *if_name) 720 { 721 struct sctp_vrf *vrf; 722 struct sctp_ifa *sctp_ifap = NULL; 723 724 SCTP_IPI_ADDR_WLOCK(); 725 vrf = sctp_find_vrf(vrf_id); 726 if (vrf == NULL) { 727 SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); 728 goto out_now; 729 } 730 731 #ifdef SCTP_DEBUG 732 SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id); 733 SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); 734 #endif 735 sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); 736 if (sctp_ifap) { 737 /* Validate the delete */ 738 if (sctp_ifap->ifn_p) { 739 int valid = 0; 740 741 /*- 742 * The name has priority over the ifn_index 743 * if its given. 744 */ 745 if (if_name) { 746 if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) { 747 /* They match its a correct delete */ 748 valid = 1; 749 } 750 } 751 if (!valid) { 752 /* last ditch check ifn_index */ 753 if (ifn_index == sctp_ifap->ifn_p->ifn_index) { 754 valid = 1; 755 } 756 } 757 if (!valid) { 758 SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n", 759 ifn_index, ((if_name == NULL) ? "NULL" : if_name)); 760 SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n", 761 sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name); 762 SCTP_IPI_ADDR_WUNLOCK(); 763 return; 764 } 765 } 766 SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap); 767 sctp_ifap->localifa_flags &= SCTP_ADDR_VALID; 768 /* 769 * We don't set the flag. This means that the structure will 770 * hang around in EP's that have bound specific to it until 771 * they close. This gives us TCP like behavior if someone 772 * removes an address (or for that matter adds it right 773 * back). 774 */ 775 /* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */ 776 vrf->total_ifa_count--; 777 LIST_REMOVE(sctp_ifap, next_bucket); 778 sctp_remove_ifa_from_ifn(sctp_ifap); 779 } 780 #ifdef SCTP_DEBUG 781 else { 782 SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:", 783 ifn_index); 784 SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); 785 } 786 #endif 787 788 out_now: 789 SCTP_IPI_ADDR_WUNLOCK(); 790 if (sctp_ifap) { 791 struct sctp_laddr *wi; 792 793 wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); 794 if (wi == NULL) { 795 /* 796 * Gak, what can we do? We have lost an address 797 * change can you say HOSED? 798 */ 799 SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); 800 801 /* Oops, must decrement the count */ 802 sctp_free_ifa(sctp_ifap); 803 return; 804 } 805 SCTP_INCR_LADDR_COUNT(); 806 memset(wi, 0, sizeof(*wi)); 807 (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); 808 wi->ifa = sctp_ifap; 809 wi->action = SCTP_DEL_IP_ADDRESS; 810 SCTP_WQ_ADDR_LOCK(); 811 /* 812 * Should this really be a tailq? As it is we will process 813 * the newest first :-0 814 */ 815 LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); 816 sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, 817 (struct sctp_inpcb *)NULL, 818 (struct sctp_tcb *)NULL, 819 (struct sctp_nets *)NULL); 820 SCTP_WQ_ADDR_UNLOCK(); 821 } 822 return; 823 } 824 825 826 static int 827 sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to) 828 { 829 int loopback_scope; 830 #if defined(INET) 831 int ipv4_local_scope, ipv4_addr_legal; 832 #endif 833 #if defined(INET6) 834 int local_scope, site_scope, ipv6_addr_legal; 835 #endif 836 struct sctp_vrf *vrf; 837 struct sctp_ifn *sctp_ifn; 838 struct sctp_ifa *sctp_ifa; 839 840 loopback_scope = stcb->asoc.scope.loopback_scope; 841 #if defined(INET) 842 ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope; 843 ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal; 844 #endif 845 #if defined(INET6) 846 local_scope = stcb->asoc.scope.local_scope; 847 site_scope = stcb->asoc.scope.site_scope; 848 ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal; 849 #endif 850 851 SCTP_IPI_ADDR_RLOCK(); 852 vrf = sctp_find_vrf(stcb->asoc.vrf_id); 853 if (vrf == NULL) { 854 /* no vrf, no addresses */ 855 SCTP_IPI_ADDR_RUNLOCK(); 856 return (0); 857 } 858 859 if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { 860 LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { 861 if ((loopback_scope == 0) && 862 SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { 863 continue; 864 } 865 LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { 866 if (sctp_is_addr_restricted(stcb, sctp_ifa) && 867 (!sctp_is_addr_pending(stcb, sctp_ifa))) { 868 /* 869 * We allow pending addresses, where 870 * we have sent an asconf-add to be 871 * considered valid. 872 */ 873 continue; 874 } 875 if (sctp_ifa->address.sa.sa_family != to->sa_family) { 876 continue; 877 } 878 switch (sctp_ifa->address.sa.sa_family) { 879 #ifdef INET 880 case AF_INET: 881 if (ipv4_addr_legal) { 882 struct sockaddr_in *sin, 883 *rsin; 884 885 sin = &sctp_ifa->address.sin; 886 rsin = (struct sockaddr_in *)to; 887 if ((ipv4_local_scope == 0) && 888 IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { 889 continue; 890 } 891 if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred, 892 &sin->sin_addr) != 0) { 893 continue; 894 } 895 if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { 896 SCTP_IPI_ADDR_RUNLOCK(); 897 return (1); 898 } 899 } 900 break; 901 #endif 902 #ifdef INET6 903 case AF_INET6: 904 if (ipv6_addr_legal) { 905 struct sockaddr_in6 *sin6, 906 *rsin6; 907 908 sin6 = &sctp_ifa->address.sin6; 909 rsin6 = (struct sockaddr_in6 *)to; 910 if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred, 911 &sin6->sin6_addr) != 0) { 912 continue; 913 } 914 if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { 915 if (local_scope == 0) 916 continue; 917 if (sin6->sin6_scope_id == 0) { 918 if (sa6_recoverscope(sin6) != 0) 919 continue; 920 } 921 } 922 if ((site_scope == 0) && 923 (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { 924 continue; 925 } 926 if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { 927 SCTP_IPI_ADDR_RUNLOCK(); 928 return (1); 929 } 930 } 931 break; 932 #endif 933 default: 934 /* TSNH */ 935 break; 936 } 937 } 938 } 939 } else { 940 struct sctp_laddr *laddr; 941 942 LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { 943 if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { 944 SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n"); 945 continue; 946 } 947 if (sctp_is_addr_restricted(stcb, laddr->ifa) && 948 (!sctp_is_addr_pending(stcb, laddr->ifa))) { 949 /* 950 * We allow pending addresses, where we have 951 * sent an asconf-add to be considered 952 * valid. 953 */ 954 continue; 955 } 956 if (laddr->ifa->address.sa.sa_family != to->sa_family) { 957 continue; 958 } 959 switch (to->sa_family) { 960 #ifdef INET 961 case AF_INET: 962 { 963 struct sockaddr_in *sin, *rsin; 964 965 sin = &laddr->ifa->address.sin; 966 rsin = (struct sockaddr_in *)to; 967 if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { 968 SCTP_IPI_ADDR_RUNLOCK(); 969 return (1); 970 } 971 break; 972 } 973 #endif 974 #ifdef INET6 975 case AF_INET6: 976 { 977 struct sockaddr_in6 *sin6, *rsin6; 978 979 sin6 = &laddr->ifa->address.sin6; 980 rsin6 = (struct sockaddr_in6 *)to; 981 if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { 982 SCTP_IPI_ADDR_RUNLOCK(); 983 return (1); 984 } 985 break; 986 } 987 988 #endif 989 default: 990 /* TSNH */ 991 break; 992 } 993 994 } 995 } 996 SCTP_IPI_ADDR_RUNLOCK(); 997 return (0); 998 } 999 1000 1001 static struct sctp_tcb * 1002 sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from, 1003 struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id) 1004 { 1005 /**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */ 1006 /* 1007 * If we support the TCP model, then we must now dig through to see 1008 * if we can find our endpoint in the list of tcp ep's. 1009 */ 1010 uint16_t lport, rport; 1011 struct sctppcbhead *ephead; 1012 struct sctp_inpcb *inp; 1013 struct sctp_laddr *laddr; 1014 struct sctp_tcb *stcb; 1015 struct sctp_nets *net; 1016 1017 if ((to == NULL) || (from == NULL)) { 1018 return (NULL); 1019 } 1020 1021 switch (to->sa_family) { 1022 #ifdef INET 1023 case AF_INET: 1024 if (from->sa_family == AF_INET) { 1025 lport = ((struct sockaddr_in *)to)->sin_port; 1026 rport = ((struct sockaddr_in *)from)->sin_port; 1027 } else { 1028 return (NULL); 1029 } 1030 break; 1031 #endif 1032 #ifdef INET6 1033 case AF_INET6: 1034 if (from->sa_family == AF_INET6) { 1035 lport = ((struct sockaddr_in6 *)to)->sin6_port; 1036 rport = ((struct sockaddr_in6 *)from)->sin6_port; 1037 } else { 1038 return (NULL); 1039 } 1040 break; 1041 #endif 1042 default: 1043 return (NULL); 1044 } 1045 ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))]; 1046 /* 1047 * Ok now for each of the guys in this bucket we must look and see: 1048 * - Does the remote port match. - Does there single association's 1049 * addresses match this address (to). If so we update p_ep to point 1050 * to this ep and return the tcb from it. 1051 */ 1052 LIST_FOREACH(inp, ephead, sctp_hash) { 1053 SCTP_INP_RLOCK(inp); 1054 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 1055 SCTP_INP_RUNLOCK(inp); 1056 continue; 1057 } 1058 if (lport != inp->sctp_lport) { 1059 SCTP_INP_RUNLOCK(inp); 1060 continue; 1061 } 1062 switch (to->sa_family) { 1063 #ifdef INET 1064 case AF_INET: 1065 { 1066 struct sockaddr_in *sin; 1067 1068 sin = (struct sockaddr_in *)to; 1069 if (prison_check_ip4(inp->ip_inp.inp.inp_cred, 1070 &sin->sin_addr) != 0) { 1071 SCTP_INP_RUNLOCK(inp); 1072 continue; 1073 } 1074 break; 1075 } 1076 #endif 1077 #ifdef INET6 1078 case AF_INET6: 1079 { 1080 struct sockaddr_in6 *sin6; 1081 1082 sin6 = (struct sockaddr_in6 *)to; 1083 if (prison_check_ip6(inp->ip_inp.inp.inp_cred, 1084 &sin6->sin6_addr) != 0) { 1085 SCTP_INP_RUNLOCK(inp); 1086 continue; 1087 } 1088 break; 1089 } 1090 #endif 1091 default: 1092 SCTP_INP_RUNLOCK(inp); 1093 continue; 1094 } 1095 if (inp->def_vrf_id != vrf_id) { 1096 SCTP_INP_RUNLOCK(inp); 1097 continue; 1098 } 1099 /* check to see if the ep has one of the addresses */ 1100 if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { 1101 /* We are NOT bound all, so look further */ 1102 int match = 0; 1103 1104 LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { 1105 1106 if (laddr->ifa == NULL) { 1107 SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__); 1108 continue; 1109 } 1110 if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { 1111 SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n"); 1112 continue; 1113 } 1114 if (laddr->ifa->address.sa.sa_family == 1115 to->sa_family) { 1116 /* see if it matches */ 1117 #ifdef INET 1118 if (from->sa_family == AF_INET) { 1119 struct sockaddr_in *intf_addr, 1120 *sin; 1121 1122 intf_addr = &laddr->ifa->address.sin; 1123 sin = (struct sockaddr_in *)to; 1124 if (sin->sin_addr.s_addr == 1125 intf_addr->sin_addr.s_addr) { 1126 match = 1; 1127 break; 1128 } 1129 } 1130 #endif 1131 #ifdef INET6 1132 if (from->sa_family == AF_INET6) { 1133 struct sockaddr_in6 *intf_addr6; 1134 struct sockaddr_in6 *sin6; 1135 1136 sin6 = (struct sockaddr_in6 *) 1137 to; 1138 intf_addr6 = &laddr->ifa->address.sin6; 1139 1140 if (SCTP6_ARE_ADDR_EQUAL(sin6, 1141 intf_addr6)) { 1142 match = 1; 1143 break; 1144 } 1145 } 1146 #endif 1147 } 1148 } 1149 if (match == 0) { 1150 /* This endpoint does not have this address */ 1151 SCTP_INP_RUNLOCK(inp); 1152 continue; 1153 } 1154 } 1155 /* 1156 * Ok if we hit here the ep has the address, does it hold 1157 * the tcb? 1158 */ 1159 /* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */ 1160 stcb = LIST_FIRST(&inp->sctp_asoc_list); 1161 if (stcb == NULL) { 1162 SCTP_INP_RUNLOCK(inp); 1163 continue; 1164 } 1165 SCTP_TCB_LOCK(stcb); 1166 if (!sctp_does_stcb_own_this_addr(stcb, to)) { 1167 SCTP_TCB_UNLOCK(stcb); 1168 SCTP_INP_RUNLOCK(inp); 1169 continue; 1170 } 1171 if (stcb->rport != rport) { 1172 /* remote port does not match. */ 1173 SCTP_TCB_UNLOCK(stcb); 1174 SCTP_INP_RUNLOCK(inp); 1175 continue; 1176 } 1177 if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { 1178 SCTP_TCB_UNLOCK(stcb); 1179 SCTP_INP_RUNLOCK(inp); 1180 continue; 1181 } 1182 if (!sctp_does_stcb_own_this_addr(stcb, to)) { 1183 SCTP_TCB_UNLOCK(stcb); 1184 SCTP_INP_RUNLOCK(inp); 1185 continue; 1186 } 1187 /* Does this TCB have a matching address? */ 1188 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 1189 1190 if (net->ro._l_addr.sa.sa_family != from->sa_family) { 1191 /* not the same family, can't be a match */ 1192 continue; 1193 } 1194 switch (from->sa_family) { 1195 #ifdef INET 1196 case AF_INET: 1197 { 1198 struct sockaddr_in *sin, *rsin; 1199 1200 sin = (struct sockaddr_in *)&net->ro._l_addr; 1201 rsin = (struct sockaddr_in *)from; 1202 if (sin->sin_addr.s_addr == 1203 rsin->sin_addr.s_addr) { 1204 /* found it */ 1205 if (netp != NULL) { 1206 *netp = net; 1207 } 1208 /* 1209 * Update the endpoint 1210 * pointer 1211 */ 1212 *inp_p = inp; 1213 SCTP_INP_RUNLOCK(inp); 1214 return (stcb); 1215 } 1216 break; 1217 } 1218 #endif 1219 #ifdef INET6 1220 case AF_INET6: 1221 { 1222 struct sockaddr_in6 *sin6, *rsin6; 1223 1224 sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; 1225 rsin6 = (struct sockaddr_in6 *)from; 1226 if (SCTP6_ARE_ADDR_EQUAL(sin6, 1227 rsin6)) { 1228 /* found it */ 1229 if (netp != NULL) { 1230 *netp = net; 1231 } 1232 /* 1233 * Update the endpoint 1234 * pointer 1235 */ 1236 *inp_p = inp; 1237 SCTP_INP_RUNLOCK(inp); 1238 return (stcb); 1239 } 1240 break; 1241 } 1242 #endif 1243 default: 1244 /* TSNH */ 1245 break; 1246 } 1247 } 1248 SCTP_TCB_UNLOCK(stcb); 1249 SCTP_INP_RUNLOCK(inp); 1250 } 1251 return (NULL); 1252 } 1253 1254 1255 /* 1256 * rules for use 1257 * 1258 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an 1259 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done 1260 * (if locked == NULL). 3) Decrement happens on return ONLY if locked == 1261 * NULL. 1262 */ 1263 1264 struct sctp_tcb * 1265 sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote, 1266 struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb) 1267 { 1268 struct sctpasochead *head; 1269 struct sctp_inpcb *inp; 1270 struct sctp_tcb *stcb = NULL; 1271 struct sctp_nets *net; 1272 uint16_t rport; 1273 1274 inp = *inp_p; 1275 switch (remote->sa_family) { 1276 #ifdef INET 1277 case AF_INET: 1278 rport = (((struct sockaddr_in *)remote)->sin_port); 1279 break; 1280 #endif 1281 #ifdef INET6 1282 case AF_INET6: 1283 rport = (((struct sockaddr_in6 *)remote)->sin6_port); 1284 break; 1285 #endif 1286 default: 1287 return (NULL); 1288 } 1289 if (locked_tcb) { 1290 /* 1291 * UN-lock so we can do proper locking here this occurs when 1292 * called from load_addresses_from_init. 1293 */ 1294 atomic_add_int(&locked_tcb->asoc.refcnt, 1); 1295 SCTP_TCB_UNLOCK(locked_tcb); 1296 } 1297 SCTP_INP_INFO_RLOCK(); 1298 if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || 1299 (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { 1300 /*- 1301 * Now either this guy is our listener or it's the 1302 * connector. If it is the one that issued the connect, then 1303 * it's only chance is to be the first TCB in the list. If 1304 * it is the acceptor, then do the special_lookup to hash 1305 * and find the real inp. 1306 */ 1307 if ((inp->sctp_socket) && SCTP_IS_LISTENING(inp)) { 1308 /* to is peer addr, from is my addr */ 1309 stcb = sctp_tcb_special_locate(inp_p, remote, local, 1310 netp, inp->def_vrf_id); 1311 if ((stcb != NULL) && (locked_tcb == NULL)) { 1312 /* we have a locked tcb, lower refcount */ 1313 SCTP_INP_DECR_REF(inp); 1314 } 1315 if ((locked_tcb != NULL) && (locked_tcb != stcb)) { 1316 SCTP_INP_RLOCK(locked_tcb->sctp_ep); 1317 SCTP_TCB_LOCK(locked_tcb); 1318 atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); 1319 SCTP_INP_RUNLOCK(locked_tcb->sctp_ep); 1320 } 1321 SCTP_INP_INFO_RUNLOCK(); 1322 return (stcb); 1323 } else { 1324 SCTP_INP_WLOCK(inp); 1325 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 1326 goto null_return; 1327 } 1328 stcb = LIST_FIRST(&inp->sctp_asoc_list); 1329 if (stcb == NULL) { 1330 goto null_return; 1331 } 1332 SCTP_TCB_LOCK(stcb); 1333 1334 if (stcb->rport != rport) { 1335 /* remote port does not match. */ 1336 SCTP_TCB_UNLOCK(stcb); 1337 goto null_return; 1338 } 1339 if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { 1340 SCTP_TCB_UNLOCK(stcb); 1341 goto null_return; 1342 } 1343 if (local && !sctp_does_stcb_own_this_addr(stcb, local)) { 1344 SCTP_TCB_UNLOCK(stcb); 1345 goto null_return; 1346 } 1347 /* now look at the list of remote addresses */ 1348 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 1349 #ifdef INVARIANTS 1350 if (net == (TAILQ_NEXT(net, sctp_next))) { 1351 panic("Corrupt net list"); 1352 } 1353 #endif 1354 if (net->ro._l_addr.sa.sa_family != 1355 remote->sa_family) { 1356 /* not the same family */ 1357 continue; 1358 } 1359 switch (remote->sa_family) { 1360 #ifdef INET 1361 case AF_INET: 1362 { 1363 struct sockaddr_in *sin, 1364 *rsin; 1365 1366 sin = (struct sockaddr_in *) 1367 &net->ro._l_addr; 1368 rsin = (struct sockaddr_in *)remote; 1369 if (sin->sin_addr.s_addr == 1370 rsin->sin_addr.s_addr) { 1371 /* found it */ 1372 if (netp != NULL) { 1373 *netp = net; 1374 } 1375 if (locked_tcb == NULL) { 1376 SCTP_INP_DECR_REF(inp); 1377 } else if (locked_tcb != stcb) { 1378 SCTP_TCB_LOCK(locked_tcb); 1379 } 1380 if (locked_tcb) { 1381 atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); 1382 } 1383 1384 SCTP_INP_WUNLOCK(inp); 1385 SCTP_INP_INFO_RUNLOCK(); 1386 return (stcb); 1387 } 1388 break; 1389 } 1390 #endif 1391 #ifdef INET6 1392 case AF_INET6: 1393 { 1394 struct sockaddr_in6 *sin6, 1395 *rsin6; 1396 1397 sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; 1398 rsin6 = (struct sockaddr_in6 *)remote; 1399 if (SCTP6_ARE_ADDR_EQUAL(sin6, 1400 rsin6)) { 1401 /* found it */ 1402 if (netp != NULL) { 1403 *netp = net; 1404 } 1405 if (locked_tcb == NULL) { 1406 SCTP_INP_DECR_REF(inp); 1407 } else if (locked_tcb != stcb) { 1408 SCTP_TCB_LOCK(locked_tcb); 1409 } 1410 if (locked_tcb) { 1411 atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); 1412 } 1413 SCTP_INP_WUNLOCK(inp); 1414 SCTP_INP_INFO_RUNLOCK(); 1415 return (stcb); 1416 } 1417 break; 1418 } 1419 #endif 1420 default: 1421 /* TSNH */ 1422 break; 1423 } 1424 } 1425 SCTP_TCB_UNLOCK(stcb); 1426 } 1427 } else { 1428 SCTP_INP_WLOCK(inp); 1429 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 1430 goto null_return; 1431 } 1432 head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport, 1433 inp->sctp_hashmark)]; 1434 LIST_FOREACH(stcb, head, sctp_tcbhash) { 1435 if (stcb->rport != rport) { 1436 /* remote port does not match */ 1437 continue; 1438 } 1439 SCTP_TCB_LOCK(stcb); 1440 if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { 1441 SCTP_TCB_UNLOCK(stcb); 1442 continue; 1443 } 1444 if (local && !sctp_does_stcb_own_this_addr(stcb, local)) { 1445 SCTP_TCB_UNLOCK(stcb); 1446 continue; 1447 } 1448 /* now look at the list of remote addresses */ 1449 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 1450 #ifdef INVARIANTS 1451 if (net == (TAILQ_NEXT(net, sctp_next))) { 1452 panic("Corrupt net list"); 1453 } 1454 #endif 1455 if (net->ro._l_addr.sa.sa_family != 1456 remote->sa_family) { 1457 /* not the same family */ 1458 continue; 1459 } 1460 switch (remote->sa_family) { 1461 #ifdef INET 1462 case AF_INET: 1463 { 1464 struct sockaddr_in *sin, 1465 *rsin; 1466 1467 sin = (struct sockaddr_in *) 1468 &net->ro._l_addr; 1469 rsin = (struct sockaddr_in *)remote; 1470 if (sin->sin_addr.s_addr == 1471 rsin->sin_addr.s_addr) { 1472 /* found it */ 1473 if (netp != NULL) { 1474 *netp = net; 1475 } 1476 if (locked_tcb == NULL) { 1477 SCTP_INP_DECR_REF(inp); 1478 } else if (locked_tcb != stcb) { 1479 SCTP_TCB_LOCK(locked_tcb); 1480 } 1481 if (locked_tcb) { 1482 atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); 1483 } 1484 SCTP_INP_WUNLOCK(inp); 1485 SCTP_INP_INFO_RUNLOCK(); 1486 return (stcb); 1487 } 1488 break; 1489 } 1490 #endif 1491 #ifdef INET6 1492 case AF_INET6: 1493 { 1494 struct sockaddr_in6 *sin6, 1495 *rsin6; 1496 1497 sin6 = (struct sockaddr_in6 *) 1498 &net->ro._l_addr; 1499 rsin6 = (struct sockaddr_in6 *)remote; 1500 if (SCTP6_ARE_ADDR_EQUAL(sin6, 1501 rsin6)) { 1502 /* found it */ 1503 if (netp != NULL) { 1504 *netp = net; 1505 } 1506 if (locked_tcb == NULL) { 1507 SCTP_INP_DECR_REF(inp); 1508 } else if (locked_tcb != stcb) { 1509 SCTP_TCB_LOCK(locked_tcb); 1510 } 1511 if (locked_tcb) { 1512 atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); 1513 } 1514 SCTP_INP_WUNLOCK(inp); 1515 SCTP_INP_INFO_RUNLOCK(); 1516 return (stcb); 1517 } 1518 break; 1519 } 1520 #endif 1521 default: 1522 /* TSNH */ 1523 break; 1524 } 1525 } 1526 SCTP_TCB_UNLOCK(stcb); 1527 } 1528 } 1529 null_return: 1530 /* clean up for returning null */ 1531 if (locked_tcb) { 1532 SCTP_TCB_LOCK(locked_tcb); 1533 atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); 1534 } 1535 SCTP_INP_WUNLOCK(inp); 1536 SCTP_INP_INFO_RUNLOCK(); 1537 /* not found */ 1538 return (NULL); 1539 } 1540 1541 1542 /* 1543 * Find an association for a specific endpoint using the association id given 1544 * out in the COMM_UP notification 1545 */ 1546 struct sctp_tcb * 1547 sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) 1548 { 1549 /* 1550 * Use my the assoc_id to find a endpoint 1551 */ 1552 struct sctpasochead *head; 1553 struct sctp_tcb *stcb; 1554 uint32_t id; 1555 1556 if (inp == NULL) { 1557 SCTP_PRINTF("TSNH ep_associd\n"); 1558 return (NULL); 1559 } 1560 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 1561 SCTP_PRINTF("TSNH ep_associd0\n"); 1562 return (NULL); 1563 } 1564 id = (uint32_t)asoc_id; 1565 head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)]; 1566 if (head == NULL) { 1567 /* invalid id TSNH */ 1568 SCTP_PRINTF("TSNH ep_associd1\n"); 1569 return (NULL); 1570 } 1571 LIST_FOREACH(stcb, head, sctp_tcbasocidhash) { 1572 if (stcb->asoc.assoc_id == id) { 1573 if (inp != stcb->sctp_ep) { 1574 /* 1575 * some other guy has the same id active (id 1576 * collision ??). 1577 */ 1578 SCTP_PRINTF("TSNH ep_associd2\n"); 1579 continue; 1580 } 1581 if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { 1582 continue; 1583 } 1584 if (want_lock) { 1585 SCTP_TCB_LOCK(stcb); 1586 } 1587 return (stcb); 1588 } 1589 } 1590 return (NULL); 1591 } 1592 1593 1594 struct sctp_tcb * 1595 sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) 1596 { 1597 struct sctp_tcb *stcb; 1598 1599 SCTP_INP_RLOCK(inp); 1600 stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock); 1601 SCTP_INP_RUNLOCK(inp); 1602 return (stcb); 1603 } 1604 1605 1606 /* 1607 * Endpoint probe expects that the INP_INFO is locked. 1608 */ 1609 static struct sctp_inpcb * 1610 sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head, 1611 uint16_t lport, uint32_t vrf_id) 1612 { 1613 struct sctp_inpcb *inp; 1614 struct sctp_laddr *laddr; 1615 #ifdef INET 1616 struct sockaddr_in *sin; 1617 #endif 1618 #ifdef INET6 1619 struct sockaddr_in6 *sin6; 1620 struct sockaddr_in6 *intf_addr6; 1621 #endif 1622 int fnd; 1623 1624 #ifdef INET 1625 sin = NULL; 1626 #endif 1627 #ifdef INET6 1628 sin6 = NULL; 1629 #endif 1630 switch (nam->sa_family) { 1631 #ifdef INET 1632 case AF_INET: 1633 sin = (struct sockaddr_in *)nam; 1634 break; 1635 #endif 1636 #ifdef INET6 1637 case AF_INET6: 1638 sin6 = (struct sockaddr_in6 *)nam; 1639 break; 1640 #endif 1641 default: 1642 /* unsupported family */ 1643 return (NULL); 1644 } 1645 1646 if (head == NULL) 1647 return (NULL); 1648 1649 LIST_FOREACH(inp, head, sctp_hash) { 1650 SCTP_INP_RLOCK(inp); 1651 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 1652 SCTP_INP_RUNLOCK(inp); 1653 continue; 1654 } 1655 if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) && 1656 (inp->sctp_lport == lport)) { 1657 /* got it */ 1658 switch (nam->sa_family) { 1659 #ifdef INET 1660 case AF_INET: 1661 if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && 1662 SCTP_IPV6_V6ONLY(inp)) { 1663 /* 1664 * IPv4 on a IPv6 socket with ONLY 1665 * IPv6 set 1666 */ 1667 SCTP_INP_RUNLOCK(inp); 1668 continue; 1669 } 1670 if (prison_check_ip4(inp->ip_inp.inp.inp_cred, 1671 &sin->sin_addr) != 0) { 1672 SCTP_INP_RUNLOCK(inp); 1673 continue; 1674 } 1675 break; 1676 #endif 1677 #ifdef INET6 1678 case AF_INET6: 1679 /* 1680 * A V6 address and the endpoint is NOT 1681 * bound V6 1682 */ 1683 if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { 1684 SCTP_INP_RUNLOCK(inp); 1685 continue; 1686 } 1687 if (prison_check_ip6(inp->ip_inp.inp.inp_cred, 1688 &sin6->sin6_addr) != 0) { 1689 SCTP_INP_RUNLOCK(inp); 1690 continue; 1691 } 1692 break; 1693 #endif 1694 default: 1695 break; 1696 } 1697 /* does a VRF id match? */ 1698 fnd = 0; 1699 if (inp->def_vrf_id == vrf_id) 1700 fnd = 1; 1701 1702 SCTP_INP_RUNLOCK(inp); 1703 if (!fnd) 1704 continue; 1705 return (inp); 1706 } 1707 SCTP_INP_RUNLOCK(inp); 1708 } 1709 switch (nam->sa_family) { 1710 #ifdef INET 1711 case AF_INET: 1712 if (sin->sin_addr.s_addr == INADDR_ANY) { 1713 /* Can't hunt for one that has no address specified */ 1714 return (NULL); 1715 } 1716 break; 1717 #endif 1718 #ifdef INET6 1719 case AF_INET6: 1720 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1721 /* Can't hunt for one that has no address specified */ 1722 return (NULL); 1723 } 1724 break; 1725 #endif 1726 default: 1727 break; 1728 } 1729 /* 1730 * ok, not bound to all so see if we can find a EP bound to this 1731 * address. 1732 */ 1733 LIST_FOREACH(inp, head, sctp_hash) { 1734 SCTP_INP_RLOCK(inp); 1735 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 1736 SCTP_INP_RUNLOCK(inp); 1737 continue; 1738 } 1739 if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) { 1740 SCTP_INP_RUNLOCK(inp); 1741 continue; 1742 } 1743 /* 1744 * Ok this could be a likely candidate, look at all of its 1745 * addresses 1746 */ 1747 if (inp->sctp_lport != lport) { 1748 SCTP_INP_RUNLOCK(inp); 1749 continue; 1750 } 1751 /* does a VRF id match? */ 1752 fnd = 0; 1753 if (inp->def_vrf_id == vrf_id) 1754 fnd = 1; 1755 1756 if (!fnd) { 1757 SCTP_INP_RUNLOCK(inp); 1758 continue; 1759 } 1760 LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { 1761 if (laddr->ifa == NULL) { 1762 SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", 1763 __func__); 1764 continue; 1765 } 1766 SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ", 1767 (void *)laddr->ifa); 1768 if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { 1769 SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n"); 1770 continue; 1771 } 1772 if (laddr->ifa->address.sa.sa_family == nam->sa_family) { 1773 /* possible, see if it matches */ 1774 switch (nam->sa_family) { 1775 #ifdef INET 1776 case AF_INET: 1777 if (sin->sin_addr.s_addr == 1778 laddr->ifa->address.sin.sin_addr.s_addr) { 1779 SCTP_INP_RUNLOCK(inp); 1780 return (inp); 1781 } 1782 break; 1783 #endif 1784 #ifdef INET6 1785 case AF_INET6: 1786 intf_addr6 = &laddr->ifa->address.sin6; 1787 if (SCTP6_ARE_ADDR_EQUAL(sin6, 1788 intf_addr6)) { 1789 SCTP_INP_RUNLOCK(inp); 1790 return (inp); 1791 } 1792 break; 1793 #endif 1794 } 1795 } 1796 } 1797 SCTP_INP_RUNLOCK(inp); 1798 } 1799 return (NULL); 1800 } 1801 1802 1803 static struct sctp_inpcb * 1804 sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id) 1805 { 1806 struct sctppcbhead *head; 1807 struct sctp_inpcb *t_inp; 1808 int fnd; 1809 1810 head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, 1811 SCTP_BASE_INFO(hashmark))]; 1812 LIST_FOREACH(t_inp, head, sctp_hash) { 1813 if (t_inp->sctp_lport != lport) { 1814 continue; 1815 } 1816 /* is it in the VRF in question */ 1817 fnd = 0; 1818 if (t_inp->def_vrf_id == vrf_id) 1819 fnd = 1; 1820 if (!fnd) 1821 continue; 1822 1823 /* This one is in use. */ 1824 /* check the v6/v4 binding issue */ 1825 if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && 1826 SCTP_IPV6_V6ONLY(t_inp)) { 1827 if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { 1828 /* collision in V6 space */ 1829 return (t_inp); 1830 } else { 1831 /* inp is BOUND_V4 no conflict */ 1832 continue; 1833 } 1834 } else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { 1835 /* t_inp is bound v4 and v6, conflict always */ 1836 return (t_inp); 1837 } else { 1838 /* t_inp is bound only V4 */ 1839 if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && 1840 SCTP_IPV6_V6ONLY(inp)) { 1841 /* no conflict */ 1842 continue; 1843 } 1844 /* else fall through to conflict */ 1845 } 1846 return (t_inp); 1847 } 1848 return (NULL); 1849 } 1850 1851 1852 int 1853 sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp) 1854 { 1855 /* For 1-2-1 with port reuse */ 1856 struct sctppcbhead *head; 1857 struct sctp_inpcb *tinp, *ninp; 1858 1859 if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) { 1860 /* only works with port reuse on */ 1861 return (-1); 1862 } 1863 if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) { 1864 return (0); 1865 } 1866 SCTP_INP_RUNLOCK(inp); 1867 SCTP_INP_INFO_WLOCK(); 1868 head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, 1869 SCTP_BASE_INFO(hashmark))]; 1870 /* Kick out all non-listeners to the TCP hash */ 1871 LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) { 1872 if (tinp->sctp_lport != inp->sctp_lport) { 1873 continue; 1874 } 1875 if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 1876 continue; 1877 } 1878 if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { 1879 continue; 1880 } 1881 if (SCTP_IS_LISTENING(tinp)) { 1882 continue; 1883 } 1884 SCTP_INP_WLOCK(tinp); 1885 LIST_REMOVE(tinp, sctp_hash); 1886 head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))]; 1887 tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; 1888 LIST_INSERT_HEAD(head, tinp, sctp_hash); 1889 SCTP_INP_WUNLOCK(tinp); 1890 } 1891 SCTP_INP_WLOCK(inp); 1892 /* Pull from where he was */ 1893 LIST_REMOVE(inp, sctp_hash); 1894 inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL; 1895 head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))]; 1896 LIST_INSERT_HEAD(head, inp, sctp_hash); 1897 SCTP_INP_WUNLOCK(inp); 1898 SCTP_INP_RLOCK(inp); 1899 SCTP_INP_INFO_WUNLOCK(); 1900 return (0); 1901 } 1902 1903 1904 struct sctp_inpcb * 1905 sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock, 1906 uint32_t vrf_id) 1907 { 1908 /* 1909 * First we check the hash table to see if someone has this port 1910 * bound with just the port. 1911 */ 1912 struct sctp_inpcb *inp; 1913 struct sctppcbhead *head; 1914 int lport; 1915 unsigned int i; 1916 #ifdef INET 1917 struct sockaddr_in *sin; 1918 #endif 1919 #ifdef INET6 1920 struct sockaddr_in6 *sin6; 1921 #endif 1922 1923 switch (nam->sa_family) { 1924 #ifdef INET 1925 case AF_INET: 1926 sin = (struct sockaddr_in *)nam; 1927 lport = sin->sin_port; 1928 break; 1929 #endif 1930 #ifdef INET6 1931 case AF_INET6: 1932 sin6 = (struct sockaddr_in6 *)nam; 1933 lport = sin6->sin6_port; 1934 break; 1935 #endif 1936 default: 1937 return (NULL); 1938 } 1939 /* 1940 * I could cheat here and just cast to one of the types but we will 1941 * do it right. It also provides the check against an Unsupported 1942 * type too. 1943 */ 1944 /* Find the head of the ALLADDR chain */ 1945 if (have_lock == 0) { 1946 SCTP_INP_INFO_RLOCK(); 1947 } 1948 head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, 1949 SCTP_BASE_INFO(hashmark))]; 1950 inp = sctp_endpoint_probe(nam, head, lport, vrf_id); 1951 1952 /* 1953 * If the TCP model exists it could be that the main listening 1954 * endpoint is gone but there still exists a connected socket for 1955 * this guy. If so we can return the first one that we find. This 1956 * may NOT be the correct one so the caller should be wary on the 1957 * returned INP. Currently the only caller that sets find_tcp_pool 1958 * is in bindx where we are verifying that a user CAN bind the 1959 * address. He either has bound it already, or someone else has, or 1960 * its open to bind, so this is good enough. 1961 */ 1962 if (inp == NULL && find_tcp_pool) { 1963 for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) { 1964 head = &SCTP_BASE_INFO(sctp_tcpephash)[i]; 1965 inp = sctp_endpoint_probe(nam, head, lport, vrf_id); 1966 if (inp) { 1967 break; 1968 } 1969 } 1970 } 1971 if (inp) { 1972 SCTP_INP_INCR_REF(inp); 1973 } 1974 if (have_lock == 0) { 1975 SCTP_INP_INFO_RUNLOCK(); 1976 } 1977 return (inp); 1978 } 1979 1980 1981 /* 1982 * Find an association for an endpoint with the pointer to whom you want to 1983 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may 1984 * need to change the *to to some other struct like a mbuf... 1985 */ 1986 struct sctp_tcb * 1987 sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to, 1988 struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool, 1989 uint32_t vrf_id) 1990 { 1991 struct sctp_inpcb *inp = NULL; 1992 struct sctp_tcb *stcb; 1993 1994 SCTP_INP_INFO_RLOCK(); 1995 if (find_tcp_pool) { 1996 if (inp_p != NULL) { 1997 stcb = sctp_tcb_special_locate(inp_p, from, to, netp, 1998 vrf_id); 1999 } else { 2000 stcb = sctp_tcb_special_locate(&inp, from, to, netp, 2001 vrf_id); 2002 } 2003 if (stcb != NULL) { 2004 SCTP_INP_INFO_RUNLOCK(); 2005 return (stcb); 2006 } 2007 } 2008 inp = sctp_pcb_findep(to, 0, 1, vrf_id); 2009 if (inp_p != NULL) { 2010 *inp_p = inp; 2011 } 2012 SCTP_INP_INFO_RUNLOCK(); 2013 if (inp == NULL) { 2014 return (NULL); 2015 } 2016 /* 2017 * ok, we have an endpoint, now lets find the assoc for it (if any) 2018 * we now place the source address or from in the to of the find 2019 * endpoint call. Since in reality this chain is used from the 2020 * inbound packet side. 2021 */ 2022 if (inp_p != NULL) { 2023 stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to, 2024 NULL); 2025 } else { 2026 stcb = sctp_findassociation_ep_addr(&inp, from, netp, to, 2027 NULL); 2028 } 2029 return (stcb); 2030 } 2031 2032 2033 /* 2034 * This routine will grub through the mbuf that is a INIT or INIT-ACK and 2035 * find all addresses that the sender has specified in any address list. Each 2036 * address will be used to lookup the TCB and see if one exits. 2037 */ 2038 static struct sctp_tcb * 2039 sctp_findassociation_special_addr(struct mbuf *m, int offset, 2040 struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, 2041 struct sockaddr *dst) 2042 { 2043 struct sctp_paramhdr *phdr, param_buf; 2044 #if defined(INET) || defined(INET6) 2045 struct sctp_tcb *stcb; 2046 uint16_t ptype; 2047 #endif 2048 uint16_t plen; 2049 #ifdef INET 2050 struct sockaddr_in sin4; 2051 #endif 2052 #ifdef INET6 2053 struct sockaddr_in6 sin6; 2054 #endif 2055 2056 #ifdef INET 2057 memset(&sin4, 0, sizeof(sin4)); 2058 sin4.sin_len = sizeof(sin4); 2059 sin4.sin_family = AF_INET; 2060 sin4.sin_port = sh->src_port; 2061 #endif 2062 #ifdef INET6 2063 memset(&sin6, 0, sizeof(sin6)); 2064 sin6.sin6_len = sizeof(sin6); 2065 sin6.sin6_family = AF_INET6; 2066 sin6.sin6_port = sh->src_port; 2067 #endif 2068 2069 offset += sizeof(struct sctp_init_chunk); 2070 2071 phdr = sctp_get_next_param(m, offset, ¶m_buf, sizeof(param_buf)); 2072 while (phdr != NULL) { 2073 /* now we must see if we want the parameter */ 2074 #if defined(INET) || defined(INET6) 2075 ptype = ntohs(phdr->param_type); 2076 #endif 2077 plen = ntohs(phdr->param_length); 2078 if (plen == 0) { 2079 break; 2080 } 2081 #ifdef INET 2082 if (ptype == SCTP_IPV4_ADDRESS && 2083 plen == sizeof(struct sctp_ipv4addr_param)) { 2084 /* Get the rest of the address */ 2085 struct sctp_ipv4addr_param ip4_param, *p4; 2086 2087 phdr = sctp_get_next_param(m, offset, 2088 (struct sctp_paramhdr *)&ip4_param, sizeof(ip4_param)); 2089 if (phdr == NULL) { 2090 return (NULL); 2091 } 2092 p4 = (struct sctp_ipv4addr_param *)phdr; 2093 memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr)); 2094 /* look it up */ 2095 stcb = sctp_findassociation_ep_addr(inp_p, 2096 (struct sockaddr *)&sin4, netp, dst, NULL); 2097 if (stcb != NULL) { 2098 return (stcb); 2099 } 2100 } 2101 #endif 2102 #ifdef INET6 2103 if (ptype == SCTP_IPV6_ADDRESS && 2104 plen == sizeof(struct sctp_ipv6addr_param)) { 2105 /* Get the rest of the address */ 2106 struct sctp_ipv6addr_param ip6_param, *p6; 2107 2108 phdr = sctp_get_next_param(m, offset, 2109 (struct sctp_paramhdr *)&ip6_param, sizeof(ip6_param)); 2110 if (phdr == NULL) { 2111 return (NULL); 2112 } 2113 p6 = (struct sctp_ipv6addr_param *)phdr; 2114 memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr)); 2115 /* look it up */ 2116 stcb = sctp_findassociation_ep_addr(inp_p, 2117 (struct sockaddr *)&sin6, netp, dst, NULL); 2118 if (stcb != NULL) { 2119 return (stcb); 2120 } 2121 } 2122 #endif 2123 offset += SCTP_SIZE32(plen); 2124 phdr = sctp_get_next_param(m, offset, ¶m_buf, 2125 sizeof(param_buf)); 2126 } 2127 return (NULL); 2128 } 2129 2130 static struct sctp_tcb * 2131 sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag, 2132 struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport, 2133 uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag) 2134 { 2135 /* 2136 * Use my vtag to hash. If we find it we then verify the source addr 2137 * is in the assoc. If all goes well we save a bit on rec of a 2138 * packet. 2139 */ 2140 struct sctpasochead *head; 2141 struct sctp_nets *net; 2142 struct sctp_tcb *stcb; 2143 2144 SCTP_INP_INFO_RLOCK(); 2145 head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag, 2146 SCTP_BASE_INFO(hashasocmark))]; 2147 LIST_FOREACH(stcb, head, sctp_asocs) { 2148 SCTP_INP_RLOCK(stcb->sctp_ep); 2149 if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 2150 SCTP_INP_RUNLOCK(stcb->sctp_ep); 2151 continue; 2152 } 2153 if (stcb->sctp_ep->def_vrf_id != vrf_id) { 2154 SCTP_INP_RUNLOCK(stcb->sctp_ep); 2155 continue; 2156 } 2157 SCTP_TCB_LOCK(stcb); 2158 SCTP_INP_RUNLOCK(stcb->sctp_ep); 2159 if (stcb->asoc.my_vtag == vtag) { 2160 /* candidate */ 2161 if (stcb->rport != rport) { 2162 SCTP_TCB_UNLOCK(stcb); 2163 continue; 2164 } 2165 if (stcb->sctp_ep->sctp_lport != lport) { 2166 SCTP_TCB_UNLOCK(stcb); 2167 continue; 2168 } 2169 if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { 2170 SCTP_TCB_UNLOCK(stcb); 2171 continue; 2172 } 2173 /* RRS:Need toaddr check here */ 2174 if (sctp_does_stcb_own_this_addr(stcb, to) == 0) { 2175 /* Endpoint does not own this address */ 2176 SCTP_TCB_UNLOCK(stcb); 2177 continue; 2178 } 2179 if (remote_tag) { 2180 /* 2181 * If we have both vtags that's all we match 2182 * on 2183 */ 2184 if (stcb->asoc.peer_vtag == remote_tag) { 2185 /* 2186 * If both tags match we consider it 2187 * conclusive and check NO 2188 * source/destination addresses 2189 */ 2190 goto conclusive; 2191 } 2192 } 2193 if (skip_src_check) { 2194 conclusive: 2195 if (from) { 2196 *netp = sctp_findnet(stcb, from); 2197 } else { 2198 *netp = NULL; /* unknown */ 2199 } 2200 if (inp_p) 2201 *inp_p = stcb->sctp_ep; 2202 SCTP_INP_INFO_RUNLOCK(); 2203 return (stcb); 2204 } 2205 net = sctp_findnet(stcb, from); 2206 if (net) { 2207 /* yep its him. */ 2208 *netp = net; 2209 SCTP_STAT_INCR(sctps_vtagexpress); 2210 *inp_p = stcb->sctp_ep; 2211 SCTP_INP_INFO_RUNLOCK(); 2212 return (stcb); 2213 } else { 2214 /* 2215 * not him, this should only happen in rare 2216 * cases so I peg it. 2217 */ 2218 SCTP_STAT_INCR(sctps_vtagbogus); 2219 } 2220 } 2221 SCTP_TCB_UNLOCK(stcb); 2222 } 2223 SCTP_INP_INFO_RUNLOCK(); 2224 return (NULL); 2225 } 2226 2227 2228 /* 2229 * Find an association with the pointer to the inbound IP packet. This can be 2230 * a IPv4 or IPv6 packet. 2231 */ 2232 struct sctp_tcb * 2233 sctp_findassociation_addr(struct mbuf *m, int offset, 2234 struct sockaddr *src, struct sockaddr *dst, 2235 struct sctphdr *sh, struct sctp_chunkhdr *ch, 2236 struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) 2237 { 2238 struct sctp_tcb *stcb; 2239 struct sctp_inpcb *inp; 2240 2241 if (sh->v_tag) { 2242 /* we only go down this path if vtag is non-zero */ 2243 stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag), 2244 inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0); 2245 if (stcb) { 2246 return (stcb); 2247 } 2248 } 2249 2250 if (inp_p) { 2251 stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp, 2252 1, vrf_id); 2253 inp = *inp_p; 2254 } else { 2255 stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp, 2256 1, vrf_id); 2257 } 2258 SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp); 2259 if (stcb == NULL && inp) { 2260 /* Found a EP but not this address */ 2261 if ((ch->chunk_type == SCTP_INITIATION) || 2262 (ch->chunk_type == SCTP_INITIATION_ACK)) { 2263 /*- 2264 * special hook, we do NOT return linp or an 2265 * association that is linked to an existing 2266 * association that is under the TCP pool (i.e. no 2267 * listener exists). The endpoint finding routine 2268 * will always find a listener before examining the 2269 * TCP pool. 2270 */ 2271 if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) { 2272 if (inp_p) { 2273 *inp_p = NULL; 2274 } 2275 return (NULL); 2276 } 2277 stcb = sctp_findassociation_special_addr(m, 2278 offset, sh, &inp, netp, dst); 2279 if (inp_p != NULL) { 2280 *inp_p = inp; 2281 } 2282 } 2283 } 2284 SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb); 2285 return (stcb); 2286 } 2287 2288 /* 2289 * lookup an association by an ASCONF lookup address. 2290 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup 2291 */ 2292 struct sctp_tcb * 2293 sctp_findassociation_ep_asconf(struct mbuf *m, int offset, 2294 struct sockaddr *dst, struct sctphdr *sh, 2295 struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) 2296 { 2297 struct sctp_tcb *stcb; 2298 union sctp_sockstore remote_store; 2299 struct sctp_paramhdr param_buf, *phdr; 2300 int ptype; 2301 int zero_address = 0; 2302 #ifdef INET 2303 struct sockaddr_in *sin; 2304 #endif 2305 #ifdef INET6 2306 struct sockaddr_in6 *sin6; 2307 #endif 2308 2309 memset(&remote_store, 0, sizeof(remote_store)); 2310 phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), 2311 ¶m_buf, sizeof(struct sctp_paramhdr)); 2312 if (phdr == NULL) { 2313 SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n", 2314 __func__); 2315 return NULL; 2316 } 2317 ptype = (int)((uint32_t)ntohs(phdr->param_type)); 2318 /* get the correlation address */ 2319 switch (ptype) { 2320 #ifdef INET6 2321 case SCTP_IPV6_ADDRESS: 2322 { 2323 /* ipv6 address param */ 2324 struct sctp_ipv6addr_param *p6, p6_buf; 2325 2326 if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) { 2327 return NULL; 2328 } 2329 p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m, 2330 offset + sizeof(struct sctp_asconf_chunk), 2331 &p6_buf.ph, sizeof(p6_buf)); 2332 if (p6 == NULL) { 2333 SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n", 2334 __func__); 2335 return (NULL); 2336 } 2337 sin6 = &remote_store.sin6; 2338 sin6->sin6_family = AF_INET6; 2339 sin6->sin6_len = sizeof(*sin6); 2340 sin6->sin6_port = sh->src_port; 2341 memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr)); 2342 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 2343 zero_address = 1; 2344 break; 2345 } 2346 #endif 2347 #ifdef INET 2348 case SCTP_IPV4_ADDRESS: 2349 { 2350 /* ipv4 address param */ 2351 struct sctp_ipv4addr_param *p4, p4_buf; 2352 2353 if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) { 2354 return NULL; 2355 } 2356 p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m, 2357 offset + sizeof(struct sctp_asconf_chunk), 2358 &p4_buf.ph, sizeof(p4_buf)); 2359 if (p4 == NULL) { 2360 SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n", 2361 __func__); 2362 return (NULL); 2363 } 2364 sin = &remote_store.sin; 2365 sin->sin_family = AF_INET; 2366 sin->sin_len = sizeof(*sin); 2367 sin->sin_port = sh->src_port; 2368 memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr)); 2369 if (sin->sin_addr.s_addr == INADDR_ANY) 2370 zero_address = 1; 2371 break; 2372 } 2373 #endif 2374 default: 2375 /* invalid address param type */ 2376 return NULL; 2377 } 2378 2379 if (zero_address) { 2380 stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p, 2381 netp, sh->src_port, sh->dest_port, 1, vrf_id, 0); 2382 if (stcb != NULL) { 2383 SCTP_INP_DECR_REF(*inp_p); 2384 } 2385 } else { 2386 stcb = sctp_findassociation_ep_addr(inp_p, 2387 &remote_store.sa, netp, 2388 dst, NULL); 2389 } 2390 return (stcb); 2391 } 2392 2393 2394 /* 2395 * allocate a sctp_inpcb and setup a temporary binding to a port/all 2396 * addresses. This way if we don't get a bind we by default pick a ephemeral 2397 * port with all addresses bound. 2398 */ 2399 int 2400 sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id) 2401 { 2402 /* 2403 * we get called when a new endpoint starts up. We need to allocate 2404 * the sctp_inpcb structure from the zone and init it. Mark it as 2405 * unbound and find a port that we can use as an ephemeral with 2406 * INADDR_ANY. If the user binds later no problem we can then add in 2407 * the specific addresses. And setup the default parameters for the 2408 * EP. 2409 */ 2410 int i, error; 2411 struct sctp_inpcb *inp; 2412 struct sctp_pcb *m; 2413 struct timeval time; 2414 sctp_sharedkey_t *null_key; 2415 2416 error = 0; 2417 2418 SCTP_INP_INFO_WLOCK(); 2419 inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb); 2420 if (inp == NULL) { 2421 SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n"); 2422 SCTP_INP_INFO_WUNLOCK(); 2423 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); 2424 return (ENOBUFS); 2425 } 2426 /* zap it */ 2427 memset(inp, 0, sizeof(*inp)); 2428 2429 /* bump generations */ 2430 /* setup socket pointers */ 2431 inp->sctp_socket = so; 2432 inp->ip_inp.inp.inp_socket = so; 2433 inp->ip_inp.inp.inp_cred = crhold(so->so_cred); 2434 #ifdef INET6 2435 if (INP_SOCKAF(so) == AF_INET6) { 2436 if (MODULE_GLOBAL(ip6_auto_flowlabel)) { 2437 inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL; 2438 } 2439 if (MODULE_GLOBAL(ip6_v6only)) { 2440 inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY; 2441 } 2442 } 2443 #endif 2444 inp->sctp_associd_counter = 1; 2445 inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT; 2446 inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT; 2447 inp->max_cwnd = 0; 2448 inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off); 2449 inp->ecn_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_ecn_enable); 2450 inp->prsctp_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pr_enable); 2451 inp->auth_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_auth_enable); 2452 inp->asconf_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_asconf_enable); 2453 inp->reconfig_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_reconfig_enable); 2454 inp->nrsack_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_nrsack_enable); 2455 inp->pktdrop_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pktdrop_enable); 2456 inp->idata_supported = 0; 2457 2458 inp->fibnum = so->so_fibnum; 2459 /* init the small hash table we use to track asocid <-> tcb */ 2460 inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark); 2461 if (inp->sctp_asocidhash == NULL) { 2462 crfree(inp->ip_inp.inp.inp_cred); 2463 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); 2464 SCTP_INP_INFO_WUNLOCK(); 2465 return (ENOBUFS); 2466 } 2467 SCTP_INCR_EP_COUNT(); 2468 inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl); 2469 SCTP_INP_INFO_WUNLOCK(); 2470 2471 so->so_pcb = (caddr_t)inp; 2472 2473 if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) { 2474 /* UDP style socket */ 2475 inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE | 2476 SCTP_PCB_FLAGS_UNBOUND); 2477 /* Be sure it is NON-BLOCKING IO for UDP */ 2478 /* SCTP_SET_SO_NBIO(so); */ 2479 } else if (SCTP_SO_TYPE(so) == SOCK_STREAM) { 2480 /* TCP style socket */ 2481 inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE | 2482 SCTP_PCB_FLAGS_UNBOUND); 2483 /* Be sure we have blocking IO by default */ 2484 SOCK_LOCK(so); 2485 SCTP_CLEAR_SO_NBIO(so); 2486 SOCK_UNLOCK(so); 2487 } else { 2488 /* 2489 * unsupported socket type (RAW, etc)- in case we missed it 2490 * in protosw 2491 */ 2492 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP); 2493 so->so_pcb = NULL; 2494 crfree(inp->ip_inp.inp.inp_cred); 2495 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); 2496 return (EOPNOTSUPP); 2497 } 2498 if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) { 2499 sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); 2500 sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); 2501 } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) { 2502 sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); 2503 sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); 2504 } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) { 2505 sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); 2506 sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); 2507 } 2508 inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize), 2509 &inp->sctp_hashmark); 2510 if (inp->sctp_tcbhash == NULL) { 2511 SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n"); 2512 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); 2513 so->so_pcb = NULL; 2514 crfree(inp->ip_inp.inp.inp_cred); 2515 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); 2516 return (ENOBUFS); 2517 } 2518 inp->def_vrf_id = vrf_id; 2519 2520 SCTP_INP_INFO_WLOCK(); 2521 SCTP_INP_LOCK_INIT(inp); 2522 INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp"); 2523 SCTP_INP_READ_INIT(inp); 2524 SCTP_ASOC_CREATE_LOCK_INIT(inp); 2525 /* lock the new ep */ 2526 SCTP_INP_WLOCK(inp); 2527 2528 /* add it to the info area */ 2529 LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list); 2530 SCTP_INP_INFO_WUNLOCK(); 2531 2532 TAILQ_INIT(&inp->read_queue); 2533 LIST_INIT(&inp->sctp_addr_list); 2534 2535 LIST_INIT(&inp->sctp_asoc_list); 2536 2537 #ifdef SCTP_TRACK_FREED_ASOCS 2538 /* TEMP CODE */ 2539 LIST_INIT(&inp->sctp_asoc_free_list); 2540 #endif 2541 /* Init the timer structure for signature change */ 2542 SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer); 2543 inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE; 2544 2545 /* now init the actual endpoint default data */ 2546 m = &inp->sctp_ep; 2547 2548 /* setup the base timeout information */ 2549 m->sctp_timeoutticks[SCTP_TIMER_SEND] = sctp_secs_to_ticks(SCTP_SEND_SEC); /* needed ? */ 2550 m->sctp_timeoutticks[SCTP_TIMER_INIT] = sctp_secs_to_ticks(SCTP_INIT_SEC); /* needed ? */ 2551 m->sctp_timeoutticks[SCTP_TIMER_RECV] = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default)); 2552 m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default)); 2553 m->sctp_timeoutticks[SCTP_TIMER_PMTU] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default)); 2554 m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default)); 2555 m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default)); 2556 /* all max/min max are in ms */ 2557 m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default); 2558 m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default); 2559 m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default); 2560 m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default); 2561 m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default); 2562 m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default); 2563 m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default); 2564 m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default); 2565 m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold); 2566 m->sctp_sws_sender = SCTP_SWS_SENDER_DEF; 2567 m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF; 2568 m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default); 2569 m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default); 2570 2571 m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module); 2572 m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module); 2573 m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default); 2574 /* number of streams to pre-open on a association */ 2575 m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default); 2576 2577 m->default_mtu = 0; 2578 /* Add adaptation cookie */ 2579 m->adaptation_layer_indicator = 0; 2580 m->adaptation_layer_indicator_provided = 0; 2581 2582 /* seed random number generator */ 2583 m->random_counter = 1; 2584 m->store_at = SCTP_SIGNATURE_SIZE; 2585 SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers)); 2586 sctp_fill_random_store(m); 2587 2588 /* Minimum cookie size */ 2589 m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) + 2590 sizeof(struct sctp_state_cookie); 2591 m->size_of_a_cookie += SCTP_SIGNATURE_SIZE; 2592 2593 /* Setup the initial secret */ 2594 (void)SCTP_GETTIME_TIMEVAL(&time); 2595 m->time_of_secret_change = time.tv_sec; 2596 2597 for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) { 2598 m->secret_key[0][i] = sctp_select_initial_TSN(m); 2599 } 2600 sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL); 2601 2602 /* How long is a cookie good for ? */ 2603 m->def_cookie_life = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default)); 2604 /* 2605 * Initialize authentication parameters 2606 */ 2607 m->local_hmacs = sctp_default_supported_hmaclist(); 2608 m->local_auth_chunks = sctp_alloc_chunklist(); 2609 if (inp->asconf_supported) { 2610 sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks); 2611 sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks); 2612 } 2613 m->default_dscp = 0; 2614 #ifdef INET6 2615 m->default_flowlabel = 0; 2616 #endif 2617 m->port = 0; /* encapsulation disabled by default */ 2618 LIST_INIT(&m->shared_keys); 2619 /* add default NULL key as key id 0 */ 2620 null_key = sctp_alloc_sharedkey(); 2621 sctp_insert_sharedkey(&m->shared_keys, null_key); 2622 SCTP_INP_WUNLOCK(inp); 2623 #ifdef SCTP_LOG_CLOSING 2624 sctp_log_closing(inp, NULL, 12); 2625 #endif 2626 return (error); 2627 } 2628 2629 2630 void 2631 sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp, 2632 struct sctp_tcb *stcb) 2633 { 2634 struct sctp_nets *net; 2635 uint16_t lport, rport; 2636 struct sctppcbhead *head; 2637 struct sctp_laddr *laddr, *oladdr; 2638 2639 atomic_add_int(&stcb->asoc.refcnt, 1); 2640 SCTP_TCB_UNLOCK(stcb); 2641 SCTP_INP_INFO_WLOCK(); 2642 SCTP_INP_WLOCK(old_inp); 2643 SCTP_INP_WLOCK(new_inp); 2644 SCTP_TCB_LOCK(stcb); 2645 atomic_subtract_int(&stcb->asoc.refcnt, 1); 2646 2647 new_inp->sctp_ep.time_of_secret_change = 2648 old_inp->sctp_ep.time_of_secret_change; 2649 memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key, 2650 sizeof(old_inp->sctp_ep.secret_key)); 2651 new_inp->sctp_ep.current_secret_number = 2652 old_inp->sctp_ep.current_secret_number; 2653 new_inp->sctp_ep.last_secret_number = 2654 old_inp->sctp_ep.last_secret_number; 2655 new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie; 2656 2657 /* make it so new data pours into the new socket */ 2658 stcb->sctp_socket = new_inp->sctp_socket; 2659 stcb->sctp_ep = new_inp; 2660 2661 /* Copy the port across */ 2662 lport = new_inp->sctp_lport = old_inp->sctp_lport; 2663 rport = stcb->rport; 2664 /* Pull the tcb from the old association */ 2665 LIST_REMOVE(stcb, sctp_tcbhash); 2666 LIST_REMOVE(stcb, sctp_tcblist); 2667 if (stcb->asoc.in_asocid_hash) { 2668 LIST_REMOVE(stcb, sctp_tcbasocidhash); 2669 } 2670 /* Now insert the new_inp into the TCP connected hash */ 2671 head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))]; 2672 2673 LIST_INSERT_HEAD(head, new_inp, sctp_hash); 2674 /* Its safe to access */ 2675 new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; 2676 2677 /* Now move the tcb into the endpoint list */ 2678 LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist); 2679 /* 2680 * Question, do we even need to worry about the ep-hash since we 2681 * only have one connection? Probably not :> so lets get rid of it 2682 * and not suck up any kernel memory in that. 2683 */ 2684 if (stcb->asoc.in_asocid_hash) { 2685 struct sctpasochead *lhd; 2686 2687 lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id, 2688 new_inp->hashasocidmark)]; 2689 LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash); 2690 } 2691 /* Ok. Let's restart timer. */ 2692 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 2693 sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp, 2694 stcb, net); 2695 } 2696 2697 SCTP_INP_INFO_WUNLOCK(); 2698 if (new_inp->sctp_tcbhash != NULL) { 2699 SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark); 2700 new_inp->sctp_tcbhash = NULL; 2701 } 2702 if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { 2703 /* Subset bound, so copy in the laddr list from the old_inp */ 2704 LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) { 2705 laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); 2706 if (laddr == NULL) { 2707 /* 2708 * Gak, what can we do? This assoc is really 2709 * HOSED. We probably should send an abort 2710 * here. 2711 */ 2712 SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n"); 2713 continue; 2714 } 2715 SCTP_INCR_LADDR_COUNT(); 2716 memset(laddr, 0, sizeof(*laddr)); 2717 (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); 2718 laddr->ifa = oladdr->ifa; 2719 atomic_add_int(&laddr->ifa->refcount, 1); 2720 LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr, 2721 sctp_nxt_addr); 2722 new_inp->laddr_count++; 2723 if (oladdr == stcb->asoc.last_used_address) { 2724 stcb->asoc.last_used_address = laddr; 2725 } 2726 } 2727 } 2728 /* 2729 * Now any running timers need to be adjusted since we really don't 2730 * care if they are running or not just blast in the new_inp into 2731 * all of them. 2732 */ 2733 2734 stcb->asoc.dack_timer.ep = (void *)new_inp; 2735 stcb->asoc.asconf_timer.ep = (void *)new_inp; 2736 stcb->asoc.strreset_timer.ep = (void *)new_inp; 2737 stcb->asoc.shut_guard_timer.ep = (void *)new_inp; 2738 stcb->asoc.autoclose_timer.ep = (void *)new_inp; 2739 stcb->asoc.delete_prim_timer.ep = (void *)new_inp; 2740 /* now what about the nets? */ 2741 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 2742 net->pmtu_timer.ep = (void *)new_inp; 2743 net->hb_timer.ep = (void *)new_inp; 2744 net->rxt_timer.ep = (void *)new_inp; 2745 } 2746 SCTP_INP_WUNLOCK(new_inp); 2747 SCTP_INP_WUNLOCK(old_inp); 2748 } 2749 2750 /* 2751 * insert an laddr entry with the given ifa for the desired list 2752 */ 2753 static int 2754 sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act) 2755 { 2756 struct sctp_laddr *laddr; 2757 2758 laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); 2759 if (laddr == NULL) { 2760 /* out of memory? */ 2761 SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 2762 return (EINVAL); 2763 } 2764 SCTP_INCR_LADDR_COUNT(); 2765 memset(laddr, 0, sizeof(*laddr)); 2766 (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); 2767 laddr->ifa = ifa; 2768 laddr->action = act; 2769 atomic_add_int(&ifa->refcount, 1); 2770 /* insert it */ 2771 LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr); 2772 2773 return (0); 2774 } 2775 2776 /* 2777 * Remove an laddr entry from the local address list (on an assoc) 2778 */ 2779 static void 2780 sctp_remove_laddr(struct sctp_laddr *laddr) 2781 { 2782 2783 /* remove from the list */ 2784 LIST_REMOVE(laddr, sctp_nxt_addr); 2785 sctp_free_ifa(laddr->ifa); 2786 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr); 2787 SCTP_DECR_LADDR_COUNT(); 2788 } 2789 2790 2791 2792 /* sctp_ifap is used to bypass normal local address validation checks */ 2793 int 2794 sctp_inpcb_bind(struct socket *so, struct sockaddr *addr, 2795 struct sctp_ifa *sctp_ifap, struct thread *p) 2796 { 2797 /* bind a ep to a socket address */ 2798 struct sctppcbhead *head; 2799 struct sctp_inpcb *inp, *inp_tmp; 2800 struct inpcb *ip_inp; 2801 int port_reuse_active = 0; 2802 int bindall; 2803 uint16_t lport; 2804 int error; 2805 uint32_t vrf_id; 2806 2807 lport = 0; 2808 bindall = 1; 2809 inp = (struct sctp_inpcb *)so->so_pcb; 2810 ip_inp = (struct inpcb *)so->so_pcb; 2811 #ifdef SCTP_DEBUG 2812 if (addr) { 2813 SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n", 2814 ntohs(((struct sockaddr_in *)addr)->sin_port)); 2815 SCTPDBG(SCTP_DEBUG_PCB1, "Addr: "); 2816 SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); 2817 } 2818 #endif 2819 if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { 2820 /* already did a bind, subsequent binds NOT allowed ! */ 2821 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 2822 return (EINVAL); 2823 } 2824 #ifdef INVARIANTS 2825 if (p == NULL) 2826 panic("null proc/thread"); 2827 #endif 2828 if (addr != NULL) { 2829 switch (addr->sa_family) { 2830 #ifdef INET 2831 case AF_INET: 2832 { 2833 struct sockaddr_in *sin; 2834 2835 /* IPV6_V6ONLY socket? */ 2836 if (SCTP_IPV6_V6ONLY(inp)) { 2837 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 2838 return (EINVAL); 2839 } 2840 if (addr->sa_len != sizeof(*sin)) { 2841 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 2842 return (EINVAL); 2843 } 2844 2845 sin = (struct sockaddr_in *)addr; 2846 lport = sin->sin_port; 2847 /* 2848 * For LOOPBACK the prison_local_ip4() call 2849 * will transmute the ip address to the 2850 * proper value. 2851 */ 2852 if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) { 2853 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); 2854 return (error); 2855 } 2856 if (sin->sin_addr.s_addr != INADDR_ANY) { 2857 bindall = 0; 2858 } 2859 break; 2860 } 2861 #endif 2862 #ifdef INET6 2863 case AF_INET6: 2864 { 2865 /* 2866 * Only for pure IPv6 Address. (No IPv4 2867 * Mapped!) 2868 */ 2869 struct sockaddr_in6 *sin6; 2870 2871 sin6 = (struct sockaddr_in6 *)addr; 2872 2873 if (addr->sa_len != sizeof(*sin6)) { 2874 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 2875 return (EINVAL); 2876 } 2877 lport = sin6->sin6_port; 2878 /* 2879 * For LOOPBACK the prison_local_ip6() call 2880 * will transmute the ipv6 address to the 2881 * proper value. 2882 */ 2883 if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr, 2884 (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) { 2885 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); 2886 return (error); 2887 } 2888 if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 2889 bindall = 0; 2890 /* KAME hack: embed scopeid */ 2891 if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) { 2892 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 2893 return (EINVAL); 2894 } 2895 } 2896 /* this must be cleared for ifa_ifwithaddr() */ 2897 sin6->sin6_scope_id = 0; 2898 break; 2899 } 2900 #endif 2901 default: 2902 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT); 2903 return (EAFNOSUPPORT); 2904 } 2905 } 2906 SCTP_INP_INFO_WLOCK(); 2907 SCTP_INP_WLOCK(inp); 2908 /* Setup a vrf_id to be the default for the non-bind-all case. */ 2909 vrf_id = inp->def_vrf_id; 2910 2911 /* increase our count due to the unlock we do */ 2912 SCTP_INP_INCR_REF(inp); 2913 if (lport) { 2914 /* 2915 * Did the caller specify a port? if so we must see if an ep 2916 * already has this one bound. 2917 */ 2918 /* got to be root to get at low ports */ 2919 if (ntohs(lport) < IPPORT_RESERVED) { 2920 if ((p != NULL) && ((error = 2921 priv_check(p, PRIV_NETINET_RESERVEDPORT) 2922 ) != 0)) { 2923 SCTP_INP_DECR_REF(inp); 2924 SCTP_INP_WUNLOCK(inp); 2925 SCTP_INP_INFO_WUNLOCK(); 2926 return (error); 2927 } 2928 } 2929 SCTP_INP_WUNLOCK(inp); 2930 if (bindall) { 2931 vrf_id = inp->def_vrf_id; 2932 inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); 2933 if (inp_tmp != NULL) { 2934 /* 2935 * lock guy returned and lower count note 2936 * that we are not bound so inp_tmp should 2937 * NEVER be inp. And it is this inp 2938 * (inp_tmp) that gets the reference bump, 2939 * so we must lower it. 2940 */ 2941 SCTP_INP_DECR_REF(inp_tmp); 2942 /* unlock info */ 2943 if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && 2944 (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { 2945 /* 2946 * Ok, must be one-2-one and 2947 * allowing port re-use 2948 */ 2949 port_reuse_active = 1; 2950 goto continue_anyway; 2951 } 2952 SCTP_INP_DECR_REF(inp); 2953 SCTP_INP_INFO_WUNLOCK(); 2954 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); 2955 return (EADDRINUSE); 2956 } 2957 } else { 2958 inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); 2959 if (inp_tmp != NULL) { 2960 /* 2961 * lock guy returned and lower count note 2962 * that we are not bound so inp_tmp should 2963 * NEVER be inp. And it is this inp 2964 * (inp_tmp) that gets the reference bump, 2965 * so we must lower it. 2966 */ 2967 SCTP_INP_DECR_REF(inp_tmp); 2968 /* unlock info */ 2969 if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && 2970 (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { 2971 /* 2972 * Ok, must be one-2-one and 2973 * allowing port re-use 2974 */ 2975 port_reuse_active = 1; 2976 goto continue_anyway; 2977 } 2978 SCTP_INP_DECR_REF(inp); 2979 SCTP_INP_INFO_WUNLOCK(); 2980 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); 2981 return (EADDRINUSE); 2982 } 2983 } 2984 continue_anyway: 2985 SCTP_INP_WLOCK(inp); 2986 if (bindall) { 2987 /* verify that no lport is not used by a singleton */ 2988 if ((port_reuse_active == 0) && 2989 (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) { 2990 /* Sorry someone already has this one bound */ 2991 if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && 2992 (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { 2993 port_reuse_active = 1; 2994 } else { 2995 SCTP_INP_DECR_REF(inp); 2996 SCTP_INP_WUNLOCK(inp); 2997 SCTP_INP_INFO_WUNLOCK(); 2998 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); 2999 return (EADDRINUSE); 3000 } 3001 } 3002 } 3003 } else { 3004 uint16_t first, last, candidate; 3005 uint16_t count; 3006 int done; 3007 3008 if (ip_inp->inp_flags & INP_HIGHPORT) { 3009 first = MODULE_GLOBAL(ipport_hifirstauto); 3010 last = MODULE_GLOBAL(ipport_hilastauto); 3011 } else if (ip_inp->inp_flags & INP_LOWPORT) { 3012 if (p && (error = 3013 priv_check(p, PRIV_NETINET_RESERVEDPORT) 3014 )) { 3015 SCTP_INP_DECR_REF(inp); 3016 SCTP_INP_WUNLOCK(inp); 3017 SCTP_INP_INFO_WUNLOCK(); 3018 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); 3019 return (error); 3020 } 3021 first = MODULE_GLOBAL(ipport_lowfirstauto); 3022 last = MODULE_GLOBAL(ipport_lowlastauto); 3023 } else { 3024 first = MODULE_GLOBAL(ipport_firstauto); 3025 last = MODULE_GLOBAL(ipport_lastauto); 3026 } 3027 if (first > last) { 3028 uint16_t temp; 3029 3030 temp = first; 3031 first = last; 3032 last = temp; 3033 } 3034 count = last - first + 1; /* number of candidates */ 3035 candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count); 3036 3037 done = 0; 3038 while (!done) { 3039 if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) { 3040 done = 1; 3041 } 3042 if (!done) { 3043 if (--count == 0) { 3044 SCTP_INP_DECR_REF(inp); 3045 SCTP_INP_WUNLOCK(inp); 3046 SCTP_INP_INFO_WUNLOCK(); 3047 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); 3048 return (EADDRINUSE); 3049 } 3050 if (candidate == last) 3051 candidate = first; 3052 else 3053 candidate = candidate + 1; 3054 } 3055 } 3056 lport = htons(candidate); 3057 } 3058 SCTP_INP_DECR_REF(inp); 3059 if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | 3060 SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { 3061 /* 3062 * this really should not happen. The guy did a non-blocking 3063 * bind and then did a close at the same time. 3064 */ 3065 SCTP_INP_WUNLOCK(inp); 3066 SCTP_INP_INFO_WUNLOCK(); 3067 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 3068 return (EINVAL); 3069 } 3070 /* ok we look clear to give out this port, so lets setup the binding */ 3071 if (bindall) { 3072 /* binding to all addresses, so just set in the proper flags */ 3073 inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL; 3074 /* set the automatic addr changes from kernel flag */ 3075 if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) { 3076 sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF); 3077 sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); 3078 } else { 3079 sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); 3080 sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); 3081 } 3082 if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) { 3083 sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); 3084 } else { 3085 sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); 3086 } 3087 /* 3088 * set the automatic mobility_base from kernel flag (by 3089 * micchie) 3090 */ 3091 if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) { 3092 sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE); 3093 sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); 3094 } else { 3095 sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE); 3096 sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); 3097 } 3098 /* 3099 * set the automatic mobility_fasthandoff from kernel flag 3100 * (by micchie) 3101 */ 3102 if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) { 3103 sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF); 3104 sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); 3105 } else { 3106 sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF); 3107 sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); 3108 } 3109 } else { 3110 /* 3111 * bind specific, make sure flags is off and add a new 3112 * address structure to the sctp_addr_list inside the ep 3113 * structure. 3114 * 3115 * We will need to allocate one and insert it at the head. 3116 * The socketopt call can just insert new addresses in there 3117 * as well. It will also have to do the embed scope kame 3118 * hack too (before adding). 3119 */ 3120 struct sctp_ifa *ifa; 3121 union sctp_sockstore store; 3122 3123 memset(&store, 0, sizeof(store)); 3124 switch (addr->sa_family) { 3125 #ifdef INET 3126 case AF_INET: 3127 memcpy(&store.sin, addr, sizeof(struct sockaddr_in)); 3128 store.sin.sin_port = 0; 3129 break; 3130 #endif 3131 #ifdef INET6 3132 case AF_INET6: 3133 memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6)); 3134 store.sin6.sin6_port = 0; 3135 break; 3136 #endif 3137 default: 3138 break; 3139 } 3140 /* 3141 * first find the interface with the bound address need to 3142 * zero out the port to find the address! yuck! can't do 3143 * this earlier since need port for sctp_pcb_findep() 3144 */ 3145 if (sctp_ifap != NULL) { 3146 ifa = sctp_ifap; 3147 } else { 3148 /* 3149 * Note for BSD we hit here always other O/S's will 3150 * pass things in via the sctp_ifap argument. 3151 */ 3152 ifa = sctp_find_ifa_by_addr(&store.sa, 3153 vrf_id, SCTP_ADDR_NOT_LOCKED); 3154 } 3155 if (ifa == NULL) { 3156 /* Can't find an interface with that address */ 3157 SCTP_INP_WUNLOCK(inp); 3158 SCTP_INP_INFO_WUNLOCK(); 3159 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL); 3160 return (EADDRNOTAVAIL); 3161 } 3162 #ifdef INET6 3163 if (addr->sa_family == AF_INET6) { 3164 /* GAK, more FIXME IFA lock? */ 3165 if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { 3166 /* Can't bind a non-existent addr. */ 3167 SCTP_INP_WUNLOCK(inp); 3168 SCTP_INP_INFO_WUNLOCK(); 3169 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 3170 return (EINVAL); 3171 } 3172 } 3173 #endif 3174 /* we're not bound all */ 3175 inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL; 3176 /* allow bindx() to send ASCONF's for binding changes */ 3177 sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); 3178 /* clear automatic addr changes from kernel flag */ 3179 sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); 3180 3181 /* add this address to the endpoint list */ 3182 error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0); 3183 if (error != 0) { 3184 SCTP_INP_WUNLOCK(inp); 3185 SCTP_INP_INFO_WUNLOCK(); 3186 return (error); 3187 } 3188 inp->laddr_count++; 3189 } 3190 /* find the bucket */ 3191 if (port_reuse_active) { 3192 /* Put it into tcp 1-2-1 hash */ 3193 head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))]; 3194 inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; 3195 } else { 3196 head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; 3197 } 3198 /* put it in the bucket */ 3199 LIST_INSERT_HEAD(head, inp, sctp_hash); 3200 SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n", 3201 (void *)head, ntohs(lport), port_reuse_active); 3202 /* set in the port */ 3203 inp->sctp_lport = lport; 3204 3205 /* turn off just the unbound flag */ 3206 inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; 3207 SCTP_INP_WUNLOCK(inp); 3208 SCTP_INP_INFO_WUNLOCK(); 3209 return (0); 3210 } 3211 3212 3213 static void 3214 sctp_iterator_inp_being_freed(struct sctp_inpcb *inp) 3215 { 3216 struct sctp_iterator *it, *nit; 3217 3218 /* 3219 * We enter with the only the ITERATOR_LOCK in place and a write 3220 * lock on the inp_info stuff. 3221 */ 3222 it = sctp_it_ctl.cur_it; 3223 if (it && (it->vn != curvnet)) { 3224 /* Its not looking at our VNET */ 3225 return; 3226 } 3227 if (it && (it->inp == inp)) { 3228 /* 3229 * This is tricky and we hold the iterator lock, but when it 3230 * returns and gets the lock (when we release it) the 3231 * iterator will try to operate on inp. We need to stop that 3232 * from happening. But of course the iterator has a 3233 * reference on the stcb and inp. We can mark it and it will 3234 * stop. 3235 * 3236 * If its a single iterator situation, we set the end 3237 * iterator flag. Otherwise we set the iterator to go to the 3238 * next inp. 3239 * 3240 */ 3241 if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { 3242 sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT; 3243 } else { 3244 sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP; 3245 } 3246 } 3247 /* 3248 * Now go through and remove any single reference to our inp that 3249 * may be still pending on the list 3250 */ 3251 SCTP_IPI_ITERATOR_WQ_LOCK(); 3252 TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { 3253 if (it->vn != curvnet) { 3254 continue; 3255 } 3256 if (it->inp == inp) { 3257 /* This one points to me is it inp specific? */ 3258 if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { 3259 /* Remove and free this one */ 3260 TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, 3261 it, sctp_nxt_itr); 3262 if (it->function_atend != NULL) { 3263 (*it->function_atend) (it->pointer, it->val); 3264 } 3265 SCTP_FREE(it, SCTP_M_ITER); 3266 } else { 3267 it->inp = LIST_NEXT(it->inp, sctp_list); 3268 if (it->inp) { 3269 SCTP_INP_INCR_REF(it->inp); 3270 } 3271 } 3272 /* 3273 * When its put in the refcnt is incremented so decr 3274 * it 3275 */ 3276 SCTP_INP_DECR_REF(inp); 3277 } 3278 } 3279 SCTP_IPI_ITERATOR_WQ_UNLOCK(); 3280 } 3281 3282 /* release sctp_inpcb unbind the port */ 3283 void 3284 sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from) 3285 { 3286 /* 3287 * Here we free a endpoint. We must find it (if it is in the Hash 3288 * table) and remove it from there. Then we must also find it in the 3289 * overall list and remove it from there. After all removals are 3290 * complete then any timer has to be stopped. Then start the actual 3291 * freeing. a) Any local lists. b) Any associations. c) The hash of 3292 * all associations. d) finally the ep itself. 3293 */ 3294 struct sctp_tcb *asoc, *nasoc; 3295 struct sctp_laddr *laddr, *nladdr; 3296 struct inpcb *ip_pcb; 3297 struct socket *so; 3298 int being_refed = 0; 3299 struct sctp_queued_to_read *sq, *nsq; 3300 int cnt; 3301 sctp_sharedkey_t *shared_key, *nshared_key; 3302 3303 3304 #ifdef SCTP_LOG_CLOSING 3305 sctp_log_closing(inp, NULL, 0); 3306 #endif 3307 SCTP_ITERATOR_LOCK(); 3308 /* mark any iterators on the list or being processed */ 3309 sctp_iterator_inp_being_freed(inp); 3310 SCTP_ITERATOR_UNLOCK(); 3311 so = inp->sctp_socket; 3312 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 3313 /* been here before.. eeks.. get out of here */ 3314 SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate); 3315 #ifdef SCTP_LOG_CLOSING 3316 sctp_log_closing(inp, NULL, 1); 3317 #endif 3318 return; 3319 } 3320 SCTP_ASOC_CREATE_LOCK(inp); 3321 SCTP_INP_INFO_WLOCK(); 3322 3323 SCTP_INP_WLOCK(inp); 3324 if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) { 3325 inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP; 3326 /* socket is gone, so no more wakeups allowed */ 3327 inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE; 3328 inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT; 3329 inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT; 3330 3331 } 3332 /* First time through we have the socket lock, after that no more. */ 3333 sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL, 3334 SCTP_FROM_SCTP_PCB + SCTP_LOC_1); 3335 3336 if (inp->control) { 3337 sctp_m_freem(inp->control); 3338 inp->control = NULL; 3339 } 3340 if (inp->pkt) { 3341 sctp_m_freem(inp->pkt); 3342 inp->pkt = NULL; 3343 } 3344 ip_pcb = &inp->ip_inp.inp; /* we could just cast the main pointer 3345 * here but I will be nice :> (i.e. 3346 * ip_pcb = ep;) */ 3347 if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) { 3348 int cnt_in_sd; 3349 3350 cnt_in_sd = 0; 3351 LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) { 3352 SCTP_TCB_LOCK(asoc); 3353 if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { 3354 /* Skip guys being freed */ 3355 cnt_in_sd++; 3356 if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) { 3357 /* 3358 * Special case - we did not start a 3359 * kill timer on the asoc due to it 3360 * was not closed. So go ahead and 3361 * start it now. 3362 */ 3363 SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_IN_ACCEPT_QUEUE); 3364 sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL); 3365 } 3366 SCTP_TCB_UNLOCK(asoc); 3367 continue; 3368 } 3369 if (((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || 3370 (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED)) && 3371 (asoc->asoc.total_output_queue_size == 0)) { 3372 /* 3373 * If we have data in queue, we don't want 3374 * to just free since the app may have done, 3375 * send()/close or connect/send/close. And 3376 * it wants the data to get across first. 3377 */ 3378 /* Just abandon things in the front states */ 3379 if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, 3380 SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) { 3381 cnt_in_sd++; 3382 } 3383 continue; 3384 } 3385 /* Disconnect the socket please */ 3386 asoc->sctp_socket = NULL; 3387 SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_CLOSED_SOCKET); 3388 if ((asoc->asoc.size_on_reasm_queue > 0) || 3389 (asoc->asoc.control_pdapi) || 3390 (asoc->asoc.size_on_all_streams > 0) || 3391 (so && (so->so_rcv.sb_cc > 0))) { 3392 /* Left with Data unread */ 3393 struct mbuf *op_err; 3394 3395 op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); 3396 asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3; 3397 sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); 3398 SCTP_STAT_INCR_COUNTER32(sctps_aborted); 3399 if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || 3400 (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { 3401 SCTP_STAT_DECR_GAUGE32(sctps_currestab); 3402 } 3403 if (sctp_free_assoc(inp, asoc, 3404 SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) { 3405 cnt_in_sd++; 3406 } 3407 continue; 3408 } else if (TAILQ_EMPTY(&asoc->asoc.send_queue) && 3409 TAILQ_EMPTY(&asoc->asoc.sent_queue) && 3410 (asoc->asoc.stream_queue_cnt == 0)) { 3411 if ((*asoc->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (asoc, &asoc->asoc)) { 3412 goto abort_anyway; 3413 } 3414 if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && 3415 (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { 3416 struct sctp_nets *netp; 3417 3418 /* 3419 * there is nothing queued to send, 3420 * so I send shutdown 3421 */ 3422 if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || 3423 (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { 3424 SCTP_STAT_DECR_GAUGE32(sctps_currestab); 3425 } 3426 SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); 3427 sctp_stop_timers_for_shutdown(asoc); 3428 if (asoc->asoc.alternate) { 3429 netp = asoc->asoc.alternate; 3430 } else { 3431 netp = asoc->asoc.primary_destination; 3432 } 3433 sctp_send_shutdown(asoc, netp); 3434 sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc, 3435 netp); 3436 sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, NULL); 3437 sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED); 3438 } 3439 } else { 3440 /* mark into shutdown pending */ 3441 SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); 3442 sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, NULL); 3443 if ((*asoc->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (asoc, &asoc->asoc)) { 3444 SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_PARTIAL_MSG_LEFT); 3445 } 3446 if (TAILQ_EMPTY(&asoc->asoc.send_queue) && 3447 TAILQ_EMPTY(&asoc->asoc.sent_queue) && 3448 (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) { 3449 struct mbuf *op_err; 3450 3451 abort_anyway: 3452 op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); 3453 asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5; 3454 sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); 3455 SCTP_STAT_INCR_COUNTER32(sctps_aborted); 3456 if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || 3457 (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { 3458 SCTP_STAT_DECR_GAUGE32(sctps_currestab); 3459 } 3460 if (sctp_free_assoc(inp, asoc, 3461 SCTP_PCBFREE_NOFORCE, 3462 SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) { 3463 cnt_in_sd++; 3464 } 3465 continue; 3466 } else { 3467 sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); 3468 } 3469 } 3470 cnt_in_sd++; 3471 SCTP_TCB_UNLOCK(asoc); 3472 } 3473 /* now is there some left in our SHUTDOWN state? */ 3474 if (cnt_in_sd) { 3475 #ifdef SCTP_LOG_CLOSING 3476 sctp_log_closing(inp, NULL, 2); 3477 #endif 3478 inp->sctp_socket = NULL; 3479 SCTP_INP_WUNLOCK(inp); 3480 SCTP_ASOC_CREATE_UNLOCK(inp); 3481 SCTP_INP_INFO_WUNLOCK(); 3482 return; 3483 } 3484 } 3485 inp->sctp_socket = NULL; 3486 if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) != 3487 SCTP_PCB_FLAGS_UNBOUND) { 3488 /* 3489 * ok, this guy has been bound. It's port is somewhere in 3490 * the SCTP_BASE_INFO(hash table). Remove it! 3491 */ 3492 LIST_REMOVE(inp, sctp_hash); 3493 inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND; 3494 } 3495 3496 /* 3497 * If there is a timer running to kill us, forget it, since it may 3498 * have a contest on the INP lock.. which would cause us to die ... 3499 */ 3500 cnt = 0; 3501 LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) { 3502 SCTP_TCB_LOCK(asoc); 3503 if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { 3504 if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) { 3505 SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_IN_ACCEPT_QUEUE); 3506 sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL); 3507 } 3508 cnt++; 3509 SCTP_TCB_UNLOCK(asoc); 3510 continue; 3511 } 3512 /* Free associations that are NOT killing us */ 3513 if ((SCTP_GET_STATE(asoc) != SCTP_STATE_COOKIE_WAIT) && 3514 ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) { 3515 struct mbuf *op_err; 3516 3517 op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); 3518 asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7; 3519 sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); 3520 SCTP_STAT_INCR_COUNTER32(sctps_aborted); 3521 } else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { 3522 cnt++; 3523 SCTP_TCB_UNLOCK(asoc); 3524 continue; 3525 } 3526 if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || 3527 (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { 3528 SCTP_STAT_DECR_GAUGE32(sctps_currestab); 3529 } 3530 if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, 3531 SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) { 3532 cnt++; 3533 } 3534 } 3535 if (cnt) { 3536 /* Ok we have someone out there that will kill us */ 3537 #ifdef SCTP_LOG_CLOSING 3538 sctp_log_closing(inp, NULL, 3); 3539 #endif 3540 SCTP_INP_WUNLOCK(inp); 3541 SCTP_ASOC_CREATE_UNLOCK(inp); 3542 SCTP_INP_INFO_WUNLOCK(); 3543 return; 3544 } 3545 if (SCTP_INP_LOCK_CONTENDED(inp)) 3546 being_refed++; 3547 if (SCTP_INP_READ_CONTENDED(inp)) 3548 being_refed++; 3549 if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp)) 3550 being_refed++; 3551 3552 if ((inp->refcount) || 3553 (being_refed) || 3554 (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) { 3555 #ifdef SCTP_LOG_CLOSING 3556 sctp_log_closing(inp, NULL, 4); 3557 #endif 3558 sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL); 3559 SCTP_INP_WUNLOCK(inp); 3560 SCTP_ASOC_CREATE_UNLOCK(inp); 3561 SCTP_INP_INFO_WUNLOCK(); 3562 return; 3563 } 3564 inp->sctp_ep.signature_change.type = 0; 3565 inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE; 3566 /* 3567 * Remove it from the list .. last thing we need a lock for. 3568 */ 3569 LIST_REMOVE(inp, sctp_list); 3570 SCTP_INP_WUNLOCK(inp); 3571 SCTP_ASOC_CREATE_UNLOCK(inp); 3572 SCTP_INP_INFO_WUNLOCK(); 3573 /* 3574 * Now we release all locks. Since this INP cannot be found anymore 3575 * except possibly by the kill timer that might be running. We call 3576 * the drain function here. It should hit the case were it sees the 3577 * ACTIVE flag cleared and exit out freeing us to proceed and 3578 * destroy everything. 3579 */ 3580 if (from != SCTP_CALLED_FROM_INPKILL_TIMER) { 3581 (void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer); 3582 } else { 3583 /* Probably un-needed */ 3584 (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); 3585 } 3586 3587 #ifdef SCTP_LOG_CLOSING 3588 sctp_log_closing(inp, NULL, 5); 3589 #endif 3590 3591 3592 if ((inp->sctp_asocidhash) != NULL) { 3593 SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark); 3594 inp->sctp_asocidhash = NULL; 3595 } 3596 /* sa_ignore FREED_MEMORY */ 3597 TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) { 3598 /* Its only abandoned if it had data left */ 3599 if (sq->length) 3600 SCTP_STAT_INCR(sctps_left_abandon); 3601 3602 TAILQ_REMOVE(&inp->read_queue, sq, next); 3603 sctp_free_remote_addr(sq->whoFrom); 3604 if (so) 3605 so->so_rcv.sb_cc -= sq->length; 3606 if (sq->data) { 3607 sctp_m_freem(sq->data); 3608 sq->data = NULL; 3609 } 3610 /* 3611 * no need to free the net count, since at this point all 3612 * assoc's are gone. 3613 */ 3614 sctp_free_a_readq(NULL, sq); 3615 } 3616 /* Now the sctp_pcb things */ 3617 /* 3618 * free each asoc if it is not already closed/free. we can't use the 3619 * macro here since le_next will get freed as part of the 3620 * sctp_free_assoc() call. 3621 */ 3622 if (ip_pcb->inp_options) { 3623 (void)sctp_m_free(ip_pcb->inp_options); 3624 ip_pcb->inp_options = 0; 3625 } 3626 3627 3628 #ifdef INET6 3629 if (ip_pcb->inp_vflag & INP_IPV6) { 3630 ip6_freepcbopts(ip_pcb->in6p_outputopts); 3631 } 3632 #endif /* INET6 */ 3633 ip_pcb->inp_vflag = 0; 3634 /* free up authentication fields */ 3635 if (inp->sctp_ep.local_auth_chunks != NULL) 3636 sctp_free_chunklist(inp->sctp_ep.local_auth_chunks); 3637 if (inp->sctp_ep.local_hmacs != NULL) 3638 sctp_free_hmaclist(inp->sctp_ep.local_hmacs); 3639 3640 LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) { 3641 LIST_REMOVE(shared_key, next); 3642 sctp_free_sharedkey(shared_key); 3643 /* sa_ignore FREED_MEMORY */ 3644 } 3645 3646 /* 3647 * if we have an address list the following will free the list of 3648 * ifaddr's that are set into this ep. Again macro limitations here, 3649 * since the LIST_FOREACH could be a bad idea. 3650 */ 3651 LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) { 3652 sctp_remove_laddr(laddr); 3653 } 3654 3655 #ifdef SCTP_TRACK_FREED_ASOCS 3656 /* TEMP CODE */ 3657 LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) { 3658 LIST_REMOVE(asoc, sctp_tcblist); 3659 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc); 3660 SCTP_DECR_ASOC_COUNT(); 3661 } 3662 /* *** END TEMP CODE *** */ 3663 #endif 3664 /* Now lets see about freeing the EP hash table. */ 3665 if (inp->sctp_tcbhash != NULL) { 3666 SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark); 3667 inp->sctp_tcbhash = NULL; 3668 } 3669 /* Now we must put the ep memory back into the zone pool */ 3670 crfree(inp->ip_inp.inp.inp_cred); 3671 INP_LOCK_DESTROY(&inp->ip_inp.inp); 3672 SCTP_INP_LOCK_DESTROY(inp); 3673 SCTP_INP_READ_DESTROY(inp); 3674 SCTP_ASOC_CREATE_LOCK_DESTROY(inp); 3675 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); 3676 SCTP_DECR_EP_COUNT(); 3677 } 3678 3679 3680 struct sctp_nets * 3681 sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr) 3682 { 3683 struct sctp_nets *net; 3684 3685 /* locate the address */ 3686 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 3687 if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr)) 3688 return (net); 3689 } 3690 return (NULL); 3691 } 3692 3693 3694 int 3695 sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id) 3696 { 3697 struct sctp_ifa *sctp_ifa; 3698 3699 sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED); 3700 if (sctp_ifa) { 3701 return (1); 3702 } else { 3703 return (0); 3704 } 3705 } 3706 3707 /* 3708 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as 3709 * when a ASCONF arrives that adds it. It will also initialize all the cwnd 3710 * stats of stuff. 3711 */ 3712 int 3713 sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr, 3714 struct sctp_nets **netp, uint16_t port, int set_scope, int from) 3715 { 3716 /* 3717 * The following is redundant to the same lines in the 3718 * sctp_aloc_assoc() but is needed since others call the add address 3719 * function 3720 */ 3721 struct sctp_nets *net, *netfirst; 3722 int addr_inscope; 3723 3724 SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ", 3725 from); 3726 SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr); 3727 3728 netfirst = sctp_findnet(stcb, newaddr); 3729 if (netfirst) { 3730 /* 3731 * Lie and return ok, we don't want to make the association 3732 * go away for this behavior. It will happen in the TCP 3733 * model in a connected socket. It does not reach the hash 3734 * table until after the association is built so it can't be 3735 * found. Mark as reachable, since the initial creation will 3736 * have been cleared and the NOT_IN_ASSOC flag will have 3737 * been added... and we don't want to end up removing it 3738 * back out. 3739 */ 3740 if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) { 3741 netfirst->dest_state = (SCTP_ADDR_REACHABLE | 3742 SCTP_ADDR_UNCONFIRMED); 3743 } else { 3744 netfirst->dest_state = SCTP_ADDR_REACHABLE; 3745 } 3746 3747 return (0); 3748 } 3749 addr_inscope = 1; 3750 switch (newaddr->sa_family) { 3751 #ifdef INET 3752 case AF_INET: 3753 { 3754 struct sockaddr_in *sin; 3755 3756 sin = (struct sockaddr_in *)newaddr; 3757 if (sin->sin_addr.s_addr == 0) { 3758 /* Invalid address */ 3759 return (-1); 3760 } 3761 /* zero out the zero area */ 3762 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 3763 3764 /* assure len is set */ 3765 sin->sin_len = sizeof(struct sockaddr_in); 3766 if (set_scope) { 3767 if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { 3768 stcb->asoc.scope.ipv4_local_scope = 1; 3769 } 3770 } else { 3771 /* Validate the address is in scope */ 3772 if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) && 3773 (stcb->asoc.scope.ipv4_local_scope == 0)) { 3774 addr_inscope = 0; 3775 } 3776 } 3777 break; 3778 } 3779 #endif 3780 #ifdef INET6 3781 case AF_INET6: 3782 { 3783 struct sockaddr_in6 *sin6; 3784 3785 sin6 = (struct sockaddr_in6 *)newaddr; 3786 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 3787 /* Invalid address */ 3788 return (-1); 3789 } 3790 /* assure len is set */ 3791 sin6->sin6_len = sizeof(struct sockaddr_in6); 3792 if (set_scope) { 3793 if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) { 3794 stcb->asoc.scope.loopback_scope = 1; 3795 stcb->asoc.scope.local_scope = 0; 3796 stcb->asoc.scope.ipv4_local_scope = 1; 3797 stcb->asoc.scope.site_scope = 1; 3798 } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { 3799 /* 3800 * If the new destination is a 3801 * LINK_LOCAL we must have common 3802 * site scope. Don't set the local 3803 * scope since we may not share all 3804 * links, only loopback can do this. 3805 * Links on the local network would 3806 * also be on our private network 3807 * for v4 too. 3808 */ 3809 stcb->asoc.scope.ipv4_local_scope = 1; 3810 stcb->asoc.scope.site_scope = 1; 3811 } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) { 3812 /* 3813 * If the new destination is 3814 * SITE_LOCAL then we must have site 3815 * scope in common. 3816 */ 3817 stcb->asoc.scope.site_scope = 1; 3818 } 3819 } else { 3820 /* Validate the address is in scope */ 3821 if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) && 3822 (stcb->asoc.scope.loopback_scope == 0)) { 3823 addr_inscope = 0; 3824 } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && 3825 (stcb->asoc.scope.local_scope == 0)) { 3826 addr_inscope = 0; 3827 } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && 3828 (stcb->asoc.scope.site_scope == 0)) { 3829 addr_inscope = 0; 3830 } 3831 } 3832 break; 3833 } 3834 #endif 3835 default: 3836 /* not supported family type */ 3837 return (-1); 3838 } 3839 net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets); 3840 if (net == NULL) { 3841 return (-1); 3842 } 3843 SCTP_INCR_RADDR_COUNT(); 3844 memset(net, 0, sizeof(struct sctp_nets)); 3845 (void)SCTP_GETTIME_TIMEVAL(&net->start_time); 3846 memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len); 3847 switch (newaddr->sa_family) { 3848 #ifdef INET 3849 case AF_INET: 3850 ((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport; 3851 break; 3852 #endif 3853 #ifdef INET6 3854 case AF_INET6: 3855 ((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport; 3856 break; 3857 #endif 3858 default: 3859 break; 3860 } 3861 net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id); 3862 if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) { 3863 stcb->asoc.scope.loopback_scope = 1; 3864 stcb->asoc.scope.ipv4_local_scope = 1; 3865 stcb->asoc.scope.local_scope = 0; 3866 stcb->asoc.scope.site_scope = 1; 3867 addr_inscope = 1; 3868 } 3869 net->failure_threshold = stcb->asoc.def_net_failure; 3870 net->pf_threshold = stcb->asoc.def_net_pf_threshold; 3871 if (addr_inscope == 0) { 3872 net->dest_state = (SCTP_ADDR_REACHABLE | 3873 SCTP_ADDR_OUT_OF_SCOPE); 3874 } else { 3875 if (from == SCTP_ADDR_IS_CONFIRMED) 3876 /* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */ 3877 net->dest_state = SCTP_ADDR_REACHABLE; 3878 else 3879 net->dest_state = SCTP_ADDR_REACHABLE | 3880 SCTP_ADDR_UNCONFIRMED; 3881 } 3882 /* 3883 * We set this to 0, the timer code knows that this means its an 3884 * initial value 3885 */ 3886 net->rto_needed = 1; 3887 net->RTO = 0; 3888 net->RTO_measured = 0; 3889 stcb->asoc.numnets++; 3890 net->ref_count = 1; 3891 net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1; 3892 net->port = port; 3893 net->dscp = stcb->asoc.default_dscp; 3894 #ifdef INET6 3895 net->flowlabel = stcb->asoc.default_flowlabel; 3896 #endif 3897 if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) { 3898 net->dest_state |= SCTP_ADDR_NOHB; 3899 } else { 3900 net->dest_state &= ~SCTP_ADDR_NOHB; 3901 } 3902 if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) { 3903 net->dest_state |= SCTP_ADDR_NO_PMTUD; 3904 } else { 3905 net->dest_state &= ~SCTP_ADDR_NO_PMTUD; 3906 } 3907 net->heart_beat_delay = stcb->asoc.heart_beat_delay; 3908 /* Init the timer structure */ 3909 SCTP_OS_TIMER_INIT(&net->rxt_timer.timer); 3910 SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer); 3911 SCTP_OS_TIMER_INIT(&net->hb_timer.timer); 3912 3913 /* Now generate a route for this guy */ 3914 #ifdef INET6 3915 /* KAME hack: embed scopeid */ 3916 if (newaddr->sa_family == AF_INET6) { 3917 struct sockaddr_in6 *sin6; 3918 3919 sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; 3920 (void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)); 3921 sin6->sin6_scope_id = 0; 3922 } 3923 #endif 3924 SCTP_RTALLOC((sctp_route_t *)&net->ro, 3925 stcb->asoc.vrf_id, 3926 stcb->sctp_ep->fibnum); 3927 3928 net->src_addr_selected = 0; 3929 if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) { 3930 /* Get source address */ 3931 net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep, 3932 stcb, 3933 (sctp_route_t *)&net->ro, 3934 net, 3935 0, 3936 stcb->asoc.vrf_id); 3937 if (stcb->asoc.default_mtu > 0) { 3938 net->mtu = stcb->asoc.default_mtu; 3939 switch (net->ro._l_addr.sa.sa_family) { 3940 #ifdef INET 3941 case AF_INET: 3942 net->mtu += SCTP_MIN_V4_OVERHEAD; 3943 break; 3944 #endif 3945 #ifdef INET6 3946 case AF_INET6: 3947 net->mtu += SCTP_MIN_OVERHEAD; 3948 break; 3949 #endif 3950 default: 3951 break; 3952 } 3953 #if defined(INET) || defined(INET6) 3954 if (net->port) { 3955 net->mtu += (uint32_t)sizeof(struct udphdr); 3956 } 3957 #endif 3958 } else if (net->ro._s_addr != NULL) { 3959 uint32_t imtu, rmtu, hcmtu; 3960 3961 net->src_addr_selected = 1; 3962 /* Now get the interface MTU */ 3963 if (net->ro._s_addr->ifn_p != NULL) { 3964 imtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p); 3965 } else { 3966 imtu = 0; 3967 } 3968 rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_nh); 3969 hcmtu = sctp_hc_get_mtu(&net->ro._l_addr, stcb->sctp_ep->fibnum); 3970 net->mtu = sctp_min_mtu(hcmtu, rmtu, imtu); 3971 } 3972 } 3973 if (net->mtu == 0) { 3974 if (stcb->asoc.default_mtu > 0) { 3975 net->mtu = stcb->asoc.default_mtu; 3976 switch (net->ro._l_addr.sa.sa_family) { 3977 #ifdef INET 3978 case AF_INET: 3979 net->mtu += SCTP_MIN_V4_OVERHEAD; 3980 break; 3981 #endif 3982 #ifdef INET6 3983 case AF_INET6: 3984 net->mtu += SCTP_MIN_OVERHEAD; 3985 break; 3986 #endif 3987 default: 3988 break; 3989 } 3990 #if defined(INET) || defined(INET6) 3991 if (net->port) { 3992 net->mtu += (uint32_t)sizeof(struct udphdr); 3993 } 3994 #endif 3995 } else { 3996 switch (newaddr->sa_family) { 3997 #ifdef INET 3998 case AF_INET: 3999 net->mtu = SCTP_DEFAULT_MTU; 4000 break; 4001 #endif 4002 #ifdef INET6 4003 case AF_INET6: 4004 net->mtu = 1280; 4005 break; 4006 #endif 4007 default: 4008 break; 4009 } 4010 } 4011 } 4012 #if defined(INET) || defined(INET6) 4013 if (net->port) { 4014 net->mtu -= (uint32_t)sizeof(struct udphdr); 4015 } 4016 #endif 4017 if (from == SCTP_ALLOC_ASOC) { 4018 stcb->asoc.smallest_mtu = net->mtu; 4019 } 4020 if (stcb->asoc.smallest_mtu > net->mtu) { 4021 sctp_pathmtu_adjustment(stcb, net->mtu); 4022 } 4023 #ifdef INET6 4024 if (newaddr->sa_family == AF_INET6) { 4025 struct sockaddr_in6 *sin6; 4026 4027 sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; 4028 (void)sa6_recoverscope(sin6); 4029 } 4030 #endif 4031 4032 /* JRS - Use the congestion control given in the CC module */ 4033 if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL) 4034 (*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net); 4035 4036 /* 4037 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning 4038 * of assoc (2005/06/27, iyengar@cis.udel.edu) 4039 */ 4040 net->find_pseudo_cumack = 1; 4041 net->find_rtx_pseudo_cumack = 1; 4042 /* Choose an initial flowid. */ 4043 net->flowid = stcb->asoc.my_vtag ^ 4044 ntohs(stcb->rport) ^ 4045 ntohs(stcb->sctp_ep->sctp_lport); 4046 net->flowtype = M_HASHTYPE_OPAQUE_HASH; 4047 if (netp) { 4048 *netp = net; 4049 } 4050 netfirst = TAILQ_FIRST(&stcb->asoc.nets); 4051 if (net->ro.ro_nh == NULL) { 4052 /* Since we have no route put it at the back */ 4053 TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); 4054 } else if (netfirst == NULL) { 4055 /* We are the first one in the pool. */ 4056 TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); 4057 } else if (netfirst->ro.ro_nh == NULL) { 4058 /* 4059 * First one has NO route. Place this one ahead of the first 4060 * one. 4061 */ 4062 TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); 4063 } else if (net->ro.ro_nh->nh_ifp != netfirst->ro.ro_nh->nh_ifp) { 4064 /* 4065 * This one has a different interface than the one at the 4066 * top of the list. Place it ahead. 4067 */ 4068 TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); 4069 } else { 4070 /* 4071 * Ok we have the same interface as the first one. Move 4072 * forward until we find either a) one with a NULL route... 4073 * insert ahead of that b) one with a different ifp.. insert 4074 * after that. c) end of the list.. insert at the tail. 4075 */ 4076 struct sctp_nets *netlook; 4077 4078 do { 4079 netlook = TAILQ_NEXT(netfirst, sctp_next); 4080 if (netlook == NULL) { 4081 /* End of the list */ 4082 TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); 4083 break; 4084 } else if (netlook->ro.ro_nh == NULL) { 4085 /* next one has NO route */ 4086 TAILQ_INSERT_BEFORE(netfirst, net, sctp_next); 4087 break; 4088 } else if (netlook->ro.ro_nh->nh_ifp != net->ro.ro_nh->nh_ifp) { 4089 TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook, 4090 net, sctp_next); 4091 break; 4092 } 4093 /* Shift forward */ 4094 netfirst = netlook; 4095 } while (netlook != NULL); 4096 } 4097 4098 /* got to have a primary set */ 4099 if (stcb->asoc.primary_destination == 0) { 4100 stcb->asoc.primary_destination = net; 4101 } else if ((stcb->asoc.primary_destination->ro.ro_nh == NULL) && 4102 (net->ro.ro_nh) && 4103 ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) { 4104 /* No route to current primary adopt new primary */ 4105 stcb->asoc.primary_destination = net; 4106 } 4107 /* Validate primary is first */ 4108 net = TAILQ_FIRST(&stcb->asoc.nets); 4109 if ((net != stcb->asoc.primary_destination) && 4110 (stcb->asoc.primary_destination)) { 4111 /* 4112 * first one on the list is NOT the primary sctp_cmpaddr() 4113 * is much more efficient if the primary is the first on the 4114 * list, make it so. 4115 */ 4116 TAILQ_REMOVE(&stcb->asoc.nets, 4117 stcb->asoc.primary_destination, sctp_next); 4118 TAILQ_INSERT_HEAD(&stcb->asoc.nets, 4119 stcb->asoc.primary_destination, sctp_next); 4120 } 4121 return (0); 4122 } 4123 4124 4125 static uint32_t 4126 sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb) 4127 { 4128 uint32_t id; 4129 struct sctpasochead *head; 4130 struct sctp_tcb *lstcb; 4131 4132 try_again: 4133 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 4134 /* TSNH */ 4135 return (0); 4136 } 4137 /* 4138 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC, 4139 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC. 4140 */ 4141 if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) { 4142 inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1; 4143 } 4144 id = inp->sctp_associd_counter; 4145 inp->sctp_associd_counter++; 4146 lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t)id, 0); 4147 if (lstcb) { 4148 goto try_again; 4149 } 4150 head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)]; 4151 LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash); 4152 stcb->asoc.in_asocid_hash = 1; 4153 return (id); 4154 } 4155 4156 /* 4157 * allocate an association and add it to the endpoint. The caller must be 4158 * careful to add all additional addresses once they are know right away or 4159 * else the assoc will be may experience a blackout scenario. 4160 */ 4161 struct sctp_tcb * 4162 sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr, 4163 int *error, uint32_t override_tag, uint32_t vrf_id, 4164 uint16_t o_streams, uint16_t port, 4165 struct thread *p, 4166 int initialize_auth_params) 4167 { 4168 /* note the p argument is only valid in unbound sockets */ 4169 4170 struct sctp_tcb *stcb; 4171 struct sctp_association *asoc; 4172 struct sctpasochead *head; 4173 uint16_t rport; 4174 int err; 4175 4176 /* 4177 * Assumption made here: Caller has done a 4178 * sctp_findassociation_ep_addr(ep, addr's); to make sure the 4179 * address does not exist already. 4180 */ 4181 if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) { 4182 /* Hit max assoc, sorry no more */ 4183 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); 4184 *error = ENOBUFS; 4185 return (NULL); 4186 } 4187 if (firstaddr == NULL) { 4188 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 4189 *error = EINVAL; 4190 return (NULL); 4191 } 4192 SCTP_INP_RLOCK(inp); 4193 if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && 4194 ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) || 4195 (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) { 4196 /* 4197 * If its in the TCP pool, its NOT allowed to create an 4198 * association. The parent listener needs to call 4199 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled 4200 * off, or connected one does this.. its an error. 4201 */ 4202 SCTP_INP_RUNLOCK(inp); 4203 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 4204 *error = EINVAL; 4205 return (NULL); 4206 } 4207 if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || 4208 (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) { 4209 if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) || 4210 (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) { 4211 SCTP_INP_RUNLOCK(inp); 4212 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 4213 *error = EINVAL; 4214 return (NULL); 4215 } 4216 } 4217 SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:"); 4218 #ifdef SCTP_DEBUG 4219 if (firstaddr) { 4220 SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr); 4221 switch (firstaddr->sa_family) { 4222 #ifdef INET 4223 case AF_INET: 4224 SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", 4225 ntohs(((struct sockaddr_in *)firstaddr)->sin_port)); 4226 break; 4227 #endif 4228 #ifdef INET6 4229 case AF_INET6: 4230 SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", 4231 ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port)); 4232 break; 4233 #endif 4234 default: 4235 break; 4236 } 4237 } else { 4238 SCTPDBG(SCTP_DEBUG_PCB3, "None\n"); 4239 } 4240 #endif /* SCTP_DEBUG */ 4241 switch (firstaddr->sa_family) { 4242 #ifdef INET 4243 case AF_INET: 4244 { 4245 struct sockaddr_in *sin; 4246 4247 sin = (struct sockaddr_in *)firstaddr; 4248 if ((ntohs(sin->sin_port) == 0) || 4249 (sin->sin_addr.s_addr == INADDR_ANY) || 4250 (sin->sin_addr.s_addr == INADDR_BROADCAST) || 4251 IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 4252 /* Invalid address */ 4253 SCTP_INP_RUNLOCK(inp); 4254 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 4255 *error = EINVAL; 4256 return (NULL); 4257 } 4258 rport = sin->sin_port; 4259 break; 4260 } 4261 #endif 4262 #ifdef INET6 4263 case AF_INET6: 4264 { 4265 struct sockaddr_in6 *sin6; 4266 4267 sin6 = (struct sockaddr_in6 *)firstaddr; 4268 if ((ntohs(sin6->sin6_port) == 0) || 4269 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || 4270 IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 4271 /* Invalid address */ 4272 SCTP_INP_RUNLOCK(inp); 4273 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 4274 *error = EINVAL; 4275 return (NULL); 4276 } 4277 rport = sin6->sin6_port; 4278 break; 4279 } 4280 #endif 4281 default: 4282 /* not supported family type */ 4283 SCTP_INP_RUNLOCK(inp); 4284 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 4285 *error = EINVAL; 4286 return (NULL); 4287 } 4288 SCTP_INP_RUNLOCK(inp); 4289 if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { 4290 /* 4291 * If you have not performed a bind, then we need to do the 4292 * ephemeral bind for you. 4293 */ 4294 if ((err = sctp_inpcb_bind(inp->sctp_socket, NULL, NULL, p))) { 4295 /* bind error, probably perm */ 4296 *error = err; 4297 return (NULL); 4298 } 4299 } 4300 stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb); 4301 if (stcb == NULL) { 4302 /* out of memory? */ 4303 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); 4304 *error = ENOMEM; 4305 return (NULL); 4306 } 4307 SCTP_INCR_ASOC_COUNT(); 4308 4309 memset(stcb, 0, sizeof(*stcb)); 4310 asoc = &stcb->asoc; 4311 4312 SCTP_TCB_LOCK_INIT(stcb); 4313 SCTP_TCB_SEND_LOCK_INIT(stcb); 4314 stcb->rport = rport; 4315 /* setup back pointer's */ 4316 stcb->sctp_ep = inp; 4317 stcb->sctp_socket = inp->sctp_socket; 4318 if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id, o_streams))) { 4319 /* failed */ 4320 SCTP_TCB_LOCK_DESTROY(stcb); 4321 SCTP_TCB_SEND_LOCK_DESTROY(stcb); 4322 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); 4323 SCTP_DECR_ASOC_COUNT(); 4324 *error = err; 4325 return (NULL); 4326 } 4327 /* and the port */ 4328 SCTP_INP_INFO_WLOCK(); 4329 SCTP_INP_WLOCK(inp); 4330 if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { 4331 /* inpcb freed while alloc going on */ 4332 SCTP_TCB_LOCK_DESTROY(stcb); 4333 SCTP_TCB_SEND_LOCK_DESTROY(stcb); 4334 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); 4335 SCTP_INP_WUNLOCK(inp); 4336 SCTP_INP_INFO_WUNLOCK(); 4337 SCTP_DECR_ASOC_COUNT(); 4338 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); 4339 *error = EINVAL; 4340 return (NULL); 4341 } 4342 SCTP_TCB_LOCK(stcb); 4343 4344 asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb); 4345 /* now that my_vtag is set, add it to the hash */ 4346 head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; 4347 /* put it in the bucket in the vtag hash of assoc's for the system */ 4348 LIST_INSERT_HEAD(head, stcb, sctp_asocs); 4349 SCTP_INP_INFO_WUNLOCK(); 4350 4351 if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, port, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) { 4352 /* failure.. memory error? */ 4353 if (asoc->strmout) { 4354 SCTP_FREE(asoc->strmout, SCTP_M_STRMO); 4355 asoc->strmout = NULL; 4356 } 4357 if (asoc->mapping_array) { 4358 SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); 4359 asoc->mapping_array = NULL; 4360 } 4361 if (asoc->nr_mapping_array) { 4362 SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); 4363 asoc->nr_mapping_array = NULL; 4364 } 4365 SCTP_DECR_ASOC_COUNT(); 4366 SCTP_TCB_UNLOCK(stcb); 4367 SCTP_TCB_LOCK_DESTROY(stcb); 4368 SCTP_TCB_SEND_LOCK_DESTROY(stcb); 4369 LIST_REMOVE(stcb, sctp_tcbasocidhash); 4370 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); 4371 SCTP_INP_WUNLOCK(inp); 4372 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); 4373 *error = ENOBUFS; 4374 return (NULL); 4375 } 4376 /* Init all the timers */ 4377 SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer); 4378 SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer); 4379 SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer); 4380 SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer); 4381 SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer); 4382 SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer); 4383 4384 LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist); 4385 /* now file the port under the hash as well */ 4386 if (inp->sctp_tcbhash != NULL) { 4387 head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport, 4388 inp->sctp_hashmark)]; 4389 LIST_INSERT_HEAD(head, stcb, sctp_tcbhash); 4390 } 4391 if (initialize_auth_params == SCTP_INITIALIZE_AUTH_PARAMS) { 4392 sctp_initialize_auth_params(inp, stcb); 4393 } 4394 SCTP_INP_WUNLOCK(inp); 4395 SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb); 4396 return (stcb); 4397 } 4398 4399 4400 void 4401 sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net) 4402 { 4403 struct sctp_inpcb *inp; 4404 struct sctp_association *asoc; 4405 4406 inp = stcb->sctp_ep; 4407 asoc = &stcb->asoc; 4408 asoc->numnets--; 4409 TAILQ_REMOVE(&asoc->nets, net, sctp_next); 4410 if (net == asoc->primary_destination) { 4411 /* Reset primary */ 4412 struct sctp_nets *lnet; 4413 4414 lnet = TAILQ_FIRST(&asoc->nets); 4415 /* 4416 * Mobility adaptation Ideally, if deleted destination is 4417 * the primary, it becomes a fast retransmission trigger by 4418 * the subsequent SET PRIMARY. (by micchie) 4419 */ 4420 if (sctp_is_mobility_feature_on(stcb->sctp_ep, 4421 SCTP_MOBILITY_BASE) || 4422 sctp_is_mobility_feature_on(stcb->sctp_ep, 4423 SCTP_MOBILITY_FASTHANDOFF)) { 4424 SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n"); 4425 if (asoc->deleted_primary != NULL) { 4426 SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n"); 4427 goto out; 4428 } 4429 asoc->deleted_primary = net; 4430 atomic_add_int(&net->ref_count, 1); 4431 memset(&net->lastsa, 0, sizeof(net->lastsa)); 4432 memset(&net->lastsv, 0, sizeof(net->lastsv)); 4433 sctp_mobility_feature_on(stcb->sctp_ep, 4434 SCTP_MOBILITY_PRIM_DELETED); 4435 sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED, 4436 stcb->sctp_ep, stcb, NULL); 4437 } 4438 out: 4439 /* Try to find a confirmed primary */ 4440 asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0); 4441 } 4442 if (net == asoc->last_data_chunk_from) { 4443 /* Reset primary */ 4444 asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets); 4445 } 4446 if (net == asoc->last_control_chunk_from) { 4447 /* Clear net */ 4448 asoc->last_control_chunk_from = NULL; 4449 } 4450 if (net == stcb->asoc.alternate) { 4451 sctp_free_remote_addr(stcb->asoc.alternate); 4452 stcb->asoc.alternate = NULL; 4453 } 4454 sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, 4455 SCTP_FROM_SCTP_PCB + SCTP_LOC_9); 4456 sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, 4457 SCTP_FROM_SCTP_PCB + SCTP_LOC_10); 4458 net->dest_state |= SCTP_ADDR_BEING_DELETED; 4459 sctp_free_remote_addr(net); 4460 } 4461 4462 /* 4463 * remove a remote endpoint address from an association, it will fail if the 4464 * address does not exist. 4465 */ 4466 int 4467 sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr) 4468 { 4469 /* 4470 * Here we need to remove a remote address. This is quite simple, we 4471 * first find it in the list of address for the association 4472 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE 4473 * on that item. Note we do not allow it to be removed if there are 4474 * no other addresses. 4475 */ 4476 struct sctp_association *asoc; 4477 struct sctp_nets *net, *nnet; 4478 4479 asoc = &stcb->asoc; 4480 4481 /* locate the address */ 4482 TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) { 4483 if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) { 4484 continue; 4485 } 4486 if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr, 4487 remaddr)) { 4488 /* we found the guy */ 4489 if (asoc->numnets < 2) { 4490 /* Must have at LEAST two remote addresses */ 4491 return (-1); 4492 } else { 4493 sctp_remove_net(stcb, net); 4494 return (0); 4495 } 4496 } 4497 } 4498 /* not found. */ 4499 return (-2); 4500 } 4501 4502 void 4503 sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport) 4504 { 4505 struct sctpvtaghead *chain; 4506 struct sctp_tagblock *twait_block; 4507 int found = 0; 4508 int i; 4509 4510 chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; 4511 LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { 4512 for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { 4513 if ((twait_block->vtag_block[i].v_tag == tag) && 4514 (twait_block->vtag_block[i].lport == lport) && 4515 (twait_block->vtag_block[i].rport == rport)) { 4516 twait_block->vtag_block[i].tv_sec_at_expire = 0; 4517 twait_block->vtag_block[i].v_tag = 0; 4518 twait_block->vtag_block[i].lport = 0; 4519 twait_block->vtag_block[i].rport = 0; 4520 found = 1; 4521 break; 4522 } 4523 } 4524 if (found) 4525 break; 4526 } 4527 } 4528 4529 int 4530 sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport) 4531 { 4532 struct sctpvtaghead *chain; 4533 struct sctp_tagblock *twait_block; 4534 int found = 0; 4535 int i; 4536 4537 SCTP_INP_INFO_WLOCK(); 4538 chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; 4539 LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { 4540 for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { 4541 if ((twait_block->vtag_block[i].v_tag == tag) && 4542 (twait_block->vtag_block[i].lport == lport) && 4543 (twait_block->vtag_block[i].rport == rport)) { 4544 found = 1; 4545 break; 4546 } 4547 } 4548 if (found) 4549 break; 4550 } 4551 SCTP_INP_INFO_WUNLOCK(); 4552 return (found); 4553 } 4554 4555 4556 void 4557 sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport) 4558 { 4559 struct sctpvtaghead *chain; 4560 struct sctp_tagblock *twait_block; 4561 struct timeval now; 4562 int set, i; 4563 4564 if (time == 0) { 4565 /* Its disabled */ 4566 return; 4567 } 4568 (void)SCTP_GETTIME_TIMEVAL(&now); 4569 chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; 4570 set = 0; 4571 LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { 4572 /* Block(s) present, lets find space, and expire on the fly */ 4573 for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { 4574 if ((twait_block->vtag_block[i].v_tag == 0) && 4575 !set) { 4576 twait_block->vtag_block[i].tv_sec_at_expire = 4577 now.tv_sec + time; 4578 twait_block->vtag_block[i].v_tag = tag; 4579 twait_block->vtag_block[i].lport = lport; 4580 twait_block->vtag_block[i].rport = rport; 4581 set = 1; 4582 } else if ((twait_block->vtag_block[i].v_tag) && 4583 ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) { 4584 /* Audit expires this guy */ 4585 twait_block->vtag_block[i].tv_sec_at_expire = 0; 4586 twait_block->vtag_block[i].v_tag = 0; 4587 twait_block->vtag_block[i].lport = 0; 4588 twait_block->vtag_block[i].rport = 0; 4589 if (set == 0) { 4590 /* Reuse it for my new tag */ 4591 twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time; 4592 twait_block->vtag_block[i].v_tag = tag; 4593 twait_block->vtag_block[i].lport = lport; 4594 twait_block->vtag_block[i].rport = rport; 4595 set = 1; 4596 } 4597 } 4598 } 4599 if (set) { 4600 /* 4601 * We only do up to the block where we can place our 4602 * tag for audits 4603 */ 4604 break; 4605 } 4606 } 4607 /* Need to add a new block to chain */ 4608 if (!set) { 4609 SCTP_MALLOC(twait_block, struct sctp_tagblock *, 4610 sizeof(struct sctp_tagblock), SCTP_M_TIMW); 4611 if (twait_block == NULL) { 4612 return; 4613 } 4614 memset(twait_block, 0, sizeof(struct sctp_tagblock)); 4615 LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock); 4616 twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time; 4617 twait_block->vtag_block[0].v_tag = tag; 4618 twait_block->vtag_block[0].lport = lport; 4619 twait_block->vtag_block[0].rport = rport; 4620 } 4621 } 4622 4623 void 4624 sctp_clean_up_stream(struct sctp_tcb *stcb, struct sctp_readhead *rh) 4625 { 4626 struct sctp_tmit_chunk *chk, *nchk; 4627 struct sctp_queued_to_read *control, *ncontrol; 4628 4629 TAILQ_FOREACH_SAFE(control, rh, next_instrm, ncontrol) { 4630 TAILQ_REMOVE(rh, control, next_instrm); 4631 control->on_strm_q = 0; 4632 if (control->on_read_q == 0) { 4633 sctp_free_remote_addr(control->whoFrom); 4634 if (control->data) { 4635 sctp_m_freem(control->data); 4636 control->data = NULL; 4637 } 4638 } 4639 /* Reassembly free? */ 4640 TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) { 4641 TAILQ_REMOVE(&control->reasm, chk, sctp_next); 4642 if (chk->data) { 4643 sctp_m_freem(chk->data); 4644 chk->data = NULL; 4645 } 4646 if (chk->holds_key_ref) 4647 sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); 4648 sctp_free_remote_addr(chk->whoTo); 4649 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); 4650 SCTP_DECR_CHK_COUNT(); 4651 /* sa_ignore FREED_MEMORY */ 4652 } 4653 /* 4654 * We don't free the address here since all the net's were 4655 * freed above. 4656 */ 4657 if (control->on_read_q == 0) { 4658 sctp_free_a_readq(stcb, control); 4659 } 4660 } 4661 } 4662 4663 /*- 4664 * Free the association after un-hashing the remote port. This 4665 * function ALWAYS returns holding NO LOCK on the stcb. It DOES 4666 * expect that the input to this function IS a locked TCB. 4667 * It will return 0, if it did NOT destroy the association (instead 4668 * it unlocks it. It will return NON-zero if it either destroyed the 4669 * association OR the association is already destroyed. 4670 */ 4671 int 4672 sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location) 4673 { 4674 int i; 4675 struct sctp_association *asoc; 4676 struct sctp_nets *net, *nnet; 4677 struct sctp_laddr *laddr, *naddr; 4678 struct sctp_tmit_chunk *chk, *nchk; 4679 struct sctp_asconf_addr *aparam, *naparam; 4680 struct sctp_asconf_ack *aack, *naack; 4681 struct sctp_stream_reset_list *strrst, *nstrrst; 4682 struct sctp_queued_to_read *sq, *nsq; 4683 struct sctp_stream_queue_pending *sp, *nsp; 4684 sctp_sharedkey_t *shared_key, *nshared_key; 4685 struct socket *so; 4686 4687 /* first, lets purge the entry from the hash table. */ 4688 4689 #ifdef SCTP_LOG_CLOSING 4690 sctp_log_closing(inp, stcb, 6); 4691 #endif 4692 if (stcb->asoc.state == 0) { 4693 #ifdef SCTP_LOG_CLOSING 4694 sctp_log_closing(inp, NULL, 7); 4695 #endif 4696 /* there is no asoc, really TSNH :-0 */ 4697 return (1); 4698 } 4699 if (stcb->asoc.alternate) { 4700 sctp_free_remote_addr(stcb->asoc.alternate); 4701 stcb->asoc.alternate = NULL; 4702 } 4703 /* TEMP CODE */ 4704 if (stcb->freed_from_where == 0) { 4705 /* Only record the first place free happened from */ 4706 stcb->freed_from_where = from_location; 4707 } 4708 /* TEMP CODE */ 4709 4710 asoc = &stcb->asoc; 4711 if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || 4712 (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) 4713 /* nothing around */ 4714 so = NULL; 4715 else 4716 so = inp->sctp_socket; 4717 4718 /* 4719 * We used timer based freeing if a reader or writer is in the way. 4720 * So we first check if we are actually being called from a timer, 4721 * if so we abort early if a reader or writer is still in the way. 4722 */ 4723 if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) && 4724 (from_inpcbfree == SCTP_NORMAL_PROC)) { 4725 /* 4726 * is it the timer driving us? if so are the reader/writers 4727 * gone? 4728 */ 4729 if (stcb->asoc.refcnt) { 4730 /* nope, reader or writer in the way */ 4731 sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); 4732 /* no asoc destroyed */ 4733 SCTP_TCB_UNLOCK(stcb); 4734 #ifdef SCTP_LOG_CLOSING 4735 sctp_log_closing(inp, stcb, 8); 4736 #endif 4737 return (0); 4738 } 4739 } 4740 /* Now clean up any other timers */ 4741 sctp_stop_association_timers(stcb, false); 4742 /* Now the read queue needs to be cleaned up (only once) */ 4743 if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) { 4744 SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_ABOUT_TO_BE_FREED); 4745 SCTP_INP_READ_LOCK(inp); 4746 TAILQ_FOREACH(sq, &inp->read_queue, next) { 4747 if (sq->stcb == stcb) { 4748 sq->do_not_ref_stcb = 1; 4749 sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn; 4750 /* 4751 * If there is no end, there never will be 4752 * now. 4753 */ 4754 if (sq->end_added == 0) { 4755 /* Held for PD-API clear that. */ 4756 sq->pdapi_aborted = 1; 4757 sq->held_length = 0; 4758 if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) { 4759 /* 4760 * Need to add a PD-API 4761 * aborted indication. 4762 * Setting the control_pdapi 4763 * assures that it will be 4764 * added right after this 4765 * msg. 4766 */ 4767 uint32_t strseq; 4768 4769 stcb->asoc.control_pdapi = sq; 4770 strseq = (sq->sinfo_stream << 16) | (sq->mid & 0x0000ffff); 4771 sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION, 4772 stcb, 4773 SCTP_PARTIAL_DELIVERY_ABORTED, 4774 (void *)&strseq, 4775 SCTP_SO_LOCKED); 4776 stcb->asoc.control_pdapi = NULL; 4777 } 4778 } 4779 /* Add an end to wake them */ 4780 sq->end_added = 1; 4781 } 4782 } 4783 SCTP_INP_READ_UNLOCK(inp); 4784 if (stcb->block_entry) { 4785 SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET); 4786 stcb->block_entry->error = ECONNRESET; 4787 stcb->block_entry = NULL; 4788 } 4789 } 4790 if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) { 4791 /* 4792 * Someone holds a reference OR the socket is unaccepted 4793 * yet. 4794 */ 4795 if ((stcb->asoc.refcnt) || 4796 (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || 4797 (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { 4798 SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE); 4799 sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); 4800 } 4801 SCTP_TCB_UNLOCK(stcb); 4802 if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || 4803 (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) 4804 /* nothing around */ 4805 so = NULL; 4806 if (so) { 4807 /* Wake any reader/writers */ 4808 sctp_sorwakeup(inp, so); 4809 sctp_sowwakeup(inp, so); 4810 } 4811 4812 #ifdef SCTP_LOG_CLOSING 4813 sctp_log_closing(inp, stcb, 9); 4814 #endif 4815 /* no asoc destroyed */ 4816 return (0); 4817 } 4818 #ifdef SCTP_LOG_CLOSING 4819 sctp_log_closing(inp, stcb, 10); 4820 #endif 4821 /* 4822 * When I reach here, no others want to kill the assoc yet.. and I 4823 * own the lock. Now its possible an abort comes in when I do the 4824 * lock exchange below to grab all the locks to do the final take 4825 * out. to prevent this we increment the count, which will start a 4826 * timer and blow out above thus assuring us that we hold exclusive 4827 * killing of the asoc. Note that after getting back the TCB lock we 4828 * will go ahead and increment the counter back up and stop any 4829 * timer a passing stranger may have started :-S 4830 */ 4831 if (from_inpcbfree == SCTP_NORMAL_PROC) { 4832 atomic_add_int(&stcb->asoc.refcnt, 1); 4833 4834 SCTP_TCB_UNLOCK(stcb); 4835 SCTP_INP_INFO_WLOCK(); 4836 SCTP_INP_WLOCK(inp); 4837 SCTP_TCB_LOCK(stcb); 4838 } 4839 /* Double check the GONE flag */ 4840 if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || 4841 (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) 4842 /* nothing around */ 4843 so = NULL; 4844 4845 if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || 4846 (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { 4847 /* 4848 * For TCP type we need special handling when we are 4849 * connected. We also include the peel'ed off ones to. 4850 */ 4851 if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { 4852 inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED; 4853 inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED; 4854 if (so) { 4855 SOCKBUF_LOCK(&so->so_rcv); 4856 so->so_state &= ~(SS_ISCONNECTING | 4857 SS_ISDISCONNECTING | 4858 SS_ISCONFIRMING | 4859 SS_ISCONNECTED); 4860 so->so_state |= SS_ISDISCONNECTED; 4861 socantrcvmore_locked(so); 4862 socantsendmore(so); 4863 sctp_sowwakeup(inp, so); 4864 sctp_sorwakeup(inp, so); 4865 SCTP_SOWAKEUP(so); 4866 } 4867 } 4868 } 4869 4870 /* 4871 * Make it invalid too, that way if its about to run it will abort 4872 * and return. 4873 */ 4874 /* re-increment the lock */ 4875 if (from_inpcbfree == SCTP_NORMAL_PROC) { 4876 atomic_add_int(&stcb->asoc.refcnt, -1); 4877 } 4878 if (stcb->asoc.refcnt) { 4879 SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE); 4880 sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); 4881 if (from_inpcbfree == SCTP_NORMAL_PROC) { 4882 SCTP_INP_INFO_WUNLOCK(); 4883 SCTP_INP_WUNLOCK(inp); 4884 } 4885 SCTP_TCB_UNLOCK(stcb); 4886 return (0); 4887 } 4888 asoc->state = 0; 4889 if (inp->sctp_tcbhash) { 4890 LIST_REMOVE(stcb, sctp_tcbhash); 4891 } 4892 if (stcb->asoc.in_asocid_hash) { 4893 LIST_REMOVE(stcb, sctp_tcbasocidhash); 4894 } 4895 /* Now lets remove it from the list of ALL associations in the EP */ 4896 LIST_REMOVE(stcb, sctp_tcblist); 4897 if (from_inpcbfree == SCTP_NORMAL_PROC) { 4898 SCTP_INP_INCR_REF(inp); 4899 SCTP_INP_WUNLOCK(inp); 4900 } 4901 /* pull from vtag hash */ 4902 LIST_REMOVE(stcb, sctp_asocs); 4903 sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait), 4904 inp->sctp_lport, stcb->rport); 4905 4906 /* 4907 * Now restop the timers to be sure this is paranoia at is finest! 4908 */ 4909 sctp_stop_association_timers(stcb, true); 4910 4911 /* 4912 * The chunk lists and such SHOULD be empty but we check them just 4913 * in case. 4914 */ 4915 /* anything on the wheel needs to be removed */ 4916 SCTP_TCB_SEND_LOCK(stcb); 4917 for (i = 0; i < asoc->streamoutcnt; i++) { 4918 struct sctp_stream_out *outs; 4919 4920 outs = &asoc->strmout[i]; 4921 /* now clean up any chunks here */ 4922 TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) { 4923 atomic_subtract_int(&asoc->stream_queue_cnt, 1); 4924 TAILQ_REMOVE(&outs->outqueue, sp, next); 4925 stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, outs, sp, 1); 4926 sctp_free_spbufspace(stcb, asoc, sp); 4927 if (sp->data) { 4928 if (so) { 4929 /* Still an open socket - report */ 4930 sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb, 4931 0, (void *)sp, SCTP_SO_LOCKED); 4932 } 4933 if (sp->data) { 4934 sctp_m_freem(sp->data); 4935 sp->data = NULL; 4936 sp->tail_mbuf = NULL; 4937 sp->length = 0; 4938 } 4939 } 4940 if (sp->net) { 4941 sctp_free_remote_addr(sp->net); 4942 sp->net = NULL; 4943 } 4944 sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED); 4945 } 4946 } 4947 SCTP_TCB_SEND_UNLOCK(stcb); 4948 /* sa_ignore FREED_MEMORY */ 4949 TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) { 4950 TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp); 4951 SCTP_FREE(strrst, SCTP_M_STRESET); 4952 } 4953 TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) { 4954 TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next); 4955 if (sq->data) { 4956 sctp_m_freem(sq->data); 4957 sq->data = NULL; 4958 } 4959 sctp_free_remote_addr(sq->whoFrom); 4960 sq->whoFrom = NULL; 4961 sq->stcb = NULL; 4962 /* Free the ctl entry */ 4963 sctp_free_a_readq(stcb, sq); 4964 /* sa_ignore FREED_MEMORY */ 4965 } 4966 TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) { 4967 TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next); 4968 if (chk->data) { 4969 sctp_m_freem(chk->data); 4970 chk->data = NULL; 4971 } 4972 if (chk->holds_key_ref) 4973 sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); 4974 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); 4975 SCTP_DECR_CHK_COUNT(); 4976 atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); 4977 asoc->free_chunk_cnt--; 4978 /* sa_ignore FREED_MEMORY */ 4979 } 4980 /* pending send queue SHOULD be empty */ 4981 TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { 4982 if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) { 4983 asoc->strmout[chk->rec.data.sid].chunks_on_queues--; 4984 #ifdef INVARIANTS 4985 } else { 4986 panic("No chunks on the queues for sid %u.", chk->rec.data.sid); 4987 #endif 4988 } 4989 TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); 4990 if (chk->data) { 4991 if (so) { 4992 /* Still a socket? */ 4993 sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb, 4994 0, chk, SCTP_SO_LOCKED); 4995 } 4996 if (chk->data) { 4997 sctp_m_freem(chk->data); 4998 chk->data = NULL; 4999 } 5000 } 5001 if (chk->holds_key_ref) 5002 sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); 5003 if (chk->whoTo) { 5004 sctp_free_remote_addr(chk->whoTo); 5005 chk->whoTo = NULL; 5006 } 5007 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); 5008 SCTP_DECR_CHK_COUNT(); 5009 /* sa_ignore FREED_MEMORY */ 5010 } 5011 /* sent queue SHOULD be empty */ 5012 TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) { 5013 if (chk->sent != SCTP_DATAGRAM_NR_ACKED) { 5014 if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) { 5015 asoc->strmout[chk->rec.data.sid].chunks_on_queues--; 5016 #ifdef INVARIANTS 5017 } else { 5018 panic("No chunks on the queues for sid %u.", chk->rec.data.sid); 5019 #endif 5020 } 5021 } 5022 TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); 5023 if (chk->data) { 5024 if (so) { 5025 /* Still a socket? */ 5026 sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb, 5027 0, chk, SCTP_SO_LOCKED); 5028 } 5029 if (chk->data) { 5030 sctp_m_freem(chk->data); 5031 chk->data = NULL; 5032 } 5033 } 5034 if (chk->holds_key_ref) 5035 sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); 5036 sctp_free_remote_addr(chk->whoTo); 5037 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); 5038 SCTP_DECR_CHK_COUNT(); 5039 /* sa_ignore FREED_MEMORY */ 5040 } 5041 #ifdef INVARIANTS 5042 for (i = 0; i < stcb->asoc.streamoutcnt; i++) { 5043 if (stcb->asoc.strmout[i].chunks_on_queues > 0) { 5044 panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i); 5045 } 5046 } 5047 #endif 5048 /* control queue MAY not be empty */ 5049 TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { 5050 TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); 5051 if (chk->data) { 5052 sctp_m_freem(chk->data); 5053 chk->data = NULL; 5054 } 5055 if (chk->holds_key_ref) 5056 sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); 5057 sctp_free_remote_addr(chk->whoTo); 5058 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); 5059 SCTP_DECR_CHK_COUNT(); 5060 /* sa_ignore FREED_MEMORY */ 5061 } 5062 /* ASCONF queue MAY not be empty */ 5063 TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { 5064 TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); 5065 if (chk->data) { 5066 sctp_m_freem(chk->data); 5067 chk->data = NULL; 5068 } 5069 if (chk->holds_key_ref) 5070 sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); 5071 sctp_free_remote_addr(chk->whoTo); 5072 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); 5073 SCTP_DECR_CHK_COUNT(); 5074 /* sa_ignore FREED_MEMORY */ 5075 } 5076 if (asoc->mapping_array) { 5077 SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); 5078 asoc->mapping_array = NULL; 5079 } 5080 if (asoc->nr_mapping_array) { 5081 SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); 5082 asoc->nr_mapping_array = NULL; 5083 } 5084 /* the stream outs */ 5085 if (asoc->strmout) { 5086 SCTP_FREE(asoc->strmout, SCTP_M_STRMO); 5087 asoc->strmout = NULL; 5088 } 5089 asoc->strm_realoutsize = asoc->streamoutcnt = 0; 5090 if (asoc->strmin) { 5091 for (i = 0; i < asoc->streamincnt; i++) { 5092 sctp_clean_up_stream(stcb, &asoc->strmin[i].inqueue); 5093 sctp_clean_up_stream(stcb, &asoc->strmin[i].uno_inqueue); 5094 } 5095 SCTP_FREE(asoc->strmin, SCTP_M_STRMI); 5096 asoc->strmin = NULL; 5097 } 5098 asoc->streamincnt = 0; 5099 TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) { 5100 #ifdef INVARIANTS 5101 if (SCTP_BASE_INFO(ipi_count_raddr) == 0) { 5102 panic("no net's left alloc'ed, or list points to itself"); 5103 } 5104 #endif 5105 TAILQ_REMOVE(&asoc->nets, net, sctp_next); 5106 sctp_free_remote_addr(net); 5107 } 5108 LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) { 5109 /* sa_ignore FREED_MEMORY */ 5110 sctp_remove_laddr(laddr); 5111 } 5112 5113 /* pending asconf (address) parameters */ 5114 TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) { 5115 /* sa_ignore FREED_MEMORY */ 5116 TAILQ_REMOVE(&asoc->asconf_queue, aparam, next); 5117 SCTP_FREE(aparam, SCTP_M_ASC_ADDR); 5118 } 5119 TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) { 5120 /* sa_ignore FREED_MEMORY */ 5121 TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next); 5122 if (aack->data != NULL) { 5123 sctp_m_freem(aack->data); 5124 } 5125 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack); 5126 } 5127 /* clean up auth stuff */ 5128 if (asoc->local_hmacs) 5129 sctp_free_hmaclist(asoc->local_hmacs); 5130 if (asoc->peer_hmacs) 5131 sctp_free_hmaclist(asoc->peer_hmacs); 5132 5133 if (asoc->local_auth_chunks) 5134 sctp_free_chunklist(asoc->local_auth_chunks); 5135 if (asoc->peer_auth_chunks) 5136 sctp_free_chunklist(asoc->peer_auth_chunks); 5137 5138 sctp_free_authinfo(&asoc->authinfo); 5139 5140 LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) { 5141 LIST_REMOVE(shared_key, next); 5142 sctp_free_sharedkey(shared_key); 5143 /* sa_ignore FREED_MEMORY */ 5144 } 5145 5146 /* Insert new items here :> */ 5147 5148 /* Get rid of LOCK */ 5149 SCTP_TCB_UNLOCK(stcb); 5150 SCTP_TCB_LOCK_DESTROY(stcb); 5151 SCTP_TCB_SEND_LOCK_DESTROY(stcb); 5152 if (from_inpcbfree == SCTP_NORMAL_PROC) { 5153 SCTP_INP_INFO_WUNLOCK(); 5154 SCTP_INP_RLOCK(inp); 5155 } 5156 #ifdef SCTP_TRACK_FREED_ASOCS 5157 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { 5158 /* now clean up the tasoc itself */ 5159 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); 5160 SCTP_DECR_ASOC_COUNT(); 5161 } else { 5162 LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist); 5163 } 5164 #else 5165 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); 5166 SCTP_DECR_ASOC_COUNT(); 5167 #endif 5168 if (from_inpcbfree == SCTP_NORMAL_PROC) { 5169 if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { 5170 /* 5171 * If its NOT the inp_free calling us AND sctp_close 5172 * as been called, we call back... 5173 */ 5174 SCTP_INP_RUNLOCK(inp); 5175 /* 5176 * This will start the kill timer (if we are the 5177 * last one) since we hold an increment yet. But 5178 * this is the only safe way to do this since 5179 * otherwise if the socket closes at the same time 5180 * we are here we might collide in the cleanup. 5181 */ 5182 sctp_inpcb_free(inp, 5183 SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, 5184 SCTP_CALLED_DIRECTLY_NOCMPSET); 5185 SCTP_INP_DECR_REF(inp); 5186 } else { 5187 /* The socket is still open. */ 5188 SCTP_INP_DECR_REF(inp); 5189 SCTP_INP_RUNLOCK(inp); 5190 } 5191 } 5192 /* destroyed the asoc */ 5193 #ifdef SCTP_LOG_CLOSING 5194 sctp_log_closing(inp, NULL, 11); 5195 #endif 5196 return (1); 5197 } 5198 5199 5200 5201 /* 5202 * determine if a destination is "reachable" based upon the addresses bound 5203 * to the current endpoint (e.g. only v4 or v6 currently bound) 5204 */ 5205 /* 5206 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use 5207 * assoc level v4/v6 flags, as the assoc *may* not have the same address 5208 * types bound as its endpoint 5209 */ 5210 int 5211 sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr) 5212 { 5213 struct sctp_inpcb *inp; 5214 int answer; 5215 5216 /* 5217 * No locks here, the TCB, in all cases is already locked and an 5218 * assoc is up. There is either a INP lock by the caller applied (in 5219 * asconf case when deleting an address) or NOT in the HB case, 5220 * however if HB then the INP increment is up and the INP will not 5221 * be removed (on top of the fact that we have a TCB lock). So we 5222 * only want to read the sctp_flags, which is either bound-all or 5223 * not.. no protection needed since once an assoc is up you can't be 5224 * changing your binding. 5225 */ 5226 inp = stcb->sctp_ep; 5227 if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { 5228 /* if bound all, destination is not restricted */ 5229 /* 5230 * RRS: Question during lock work: Is this correct? If you 5231 * are bound-all you still might need to obey the V4--V6 5232 * flags??? IMO this bound-all stuff needs to be removed! 5233 */ 5234 return (1); 5235 } 5236 /* NOTE: all "scope" checks are done when local addresses are added */ 5237 switch (destaddr->sa_family) { 5238 #ifdef INET6 5239 case AF_INET6: 5240 answer = inp->ip_inp.inp.inp_vflag & INP_IPV6; 5241 break; 5242 #endif 5243 #ifdef INET 5244 case AF_INET: 5245 answer = inp->ip_inp.inp.inp_vflag & INP_IPV4; 5246 break; 5247 #endif 5248 default: 5249 /* invalid family, so it's unreachable */ 5250 answer = 0; 5251 break; 5252 } 5253 return (answer); 5254 } 5255 5256 /* 5257 * update the inp_vflags on an endpoint 5258 */ 5259 static void 5260 sctp_update_ep_vflag(struct sctp_inpcb *inp) 5261 { 5262 struct sctp_laddr *laddr; 5263 5264 /* first clear the flag */ 5265 inp->ip_inp.inp.inp_vflag = 0; 5266 /* set the flag based on addresses on the ep list */ 5267 LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { 5268 if (laddr->ifa == NULL) { 5269 SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", 5270 __func__); 5271 continue; 5272 } 5273 5274 if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { 5275 continue; 5276 } 5277 switch (laddr->ifa->address.sa.sa_family) { 5278 #ifdef INET6 5279 case AF_INET6: 5280 inp->ip_inp.inp.inp_vflag |= INP_IPV6; 5281 break; 5282 #endif 5283 #ifdef INET 5284 case AF_INET: 5285 inp->ip_inp.inp.inp_vflag |= INP_IPV4; 5286 break; 5287 #endif 5288 default: 5289 break; 5290 } 5291 } 5292 } 5293 5294 /* 5295 * Add the address to the endpoint local address list There is nothing to be 5296 * done if we are bound to all addresses 5297 */ 5298 void 5299 sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action) 5300 { 5301 struct sctp_laddr *laddr; 5302 struct sctp_tcb *stcb; 5303 int fnd, error = 0; 5304 5305 fnd = 0; 5306 5307 if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { 5308 /* You are already bound to all. You have it already */ 5309 return; 5310 } 5311 #ifdef INET6 5312 if (ifa->address.sa.sa_family == AF_INET6) { 5313 if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { 5314 /* Can't bind a non-useable addr. */ 5315 return; 5316 } 5317 } 5318 #endif 5319 /* first, is it already present? */ 5320 LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { 5321 if (laddr->ifa == ifa) { 5322 fnd = 1; 5323 break; 5324 } 5325 } 5326 5327 if (fnd == 0) { 5328 /* Not in the ep list */ 5329 error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action); 5330 if (error != 0) 5331 return; 5332 inp->laddr_count++; 5333 /* update inp_vflag flags */ 5334 switch (ifa->address.sa.sa_family) { 5335 #ifdef INET6 5336 case AF_INET6: 5337 inp->ip_inp.inp.inp_vflag |= INP_IPV6; 5338 break; 5339 #endif 5340 #ifdef INET 5341 case AF_INET: 5342 inp->ip_inp.inp.inp_vflag |= INP_IPV4; 5343 break; 5344 #endif 5345 default: 5346 break; 5347 } 5348 LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { 5349 sctp_add_local_addr_restricted(stcb, ifa); 5350 } 5351 } 5352 return; 5353 } 5354 5355 5356 /* 5357 * select a new (hopefully reachable) destination net (should only be used 5358 * when we deleted an ep addr that is the only usable source address to reach 5359 * the destination net) 5360 */ 5361 static void 5362 sctp_select_primary_destination(struct sctp_tcb *stcb) 5363 { 5364 struct sctp_nets *net; 5365 5366 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 5367 /* for now, we'll just pick the first reachable one we find */ 5368 if (net->dest_state & SCTP_ADDR_UNCONFIRMED) 5369 continue; 5370 if (sctp_destination_is_reachable(stcb, 5371 (struct sockaddr *)&net->ro._l_addr)) { 5372 /* found a reachable destination */ 5373 stcb->asoc.primary_destination = net; 5374 } 5375 } 5376 /* I can't there from here! ...we're gonna die shortly... */ 5377 } 5378 5379 5380 /* 5381 * Delete the address from the endpoint local address list. There is nothing 5382 * to be done if we are bound to all addresses 5383 */ 5384 void 5385 sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa) 5386 { 5387 struct sctp_laddr *laddr; 5388 int fnd; 5389 5390 fnd = 0; 5391 if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { 5392 /* You are already bound to all. You have it already */ 5393 return; 5394 } 5395 LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { 5396 if (laddr->ifa == ifa) { 5397 fnd = 1; 5398 break; 5399 } 5400 } 5401 if (fnd && (inp->laddr_count < 2)) { 5402 /* can't delete unless there are at LEAST 2 addresses */ 5403 return; 5404 } 5405 if (fnd) { 5406 /* 5407 * clean up any use of this address go through our 5408 * associations and clear any last_used_address that match 5409 * this one for each assoc, see if a new primary_destination 5410 * is needed 5411 */ 5412 struct sctp_tcb *stcb; 5413 5414 /* clean up "next_addr_touse" */ 5415 if (inp->next_addr_touse == laddr) 5416 /* delete this address */ 5417 inp->next_addr_touse = NULL; 5418 5419 /* clean up "last_used_address" */ 5420 LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { 5421 struct sctp_nets *net; 5422 5423 SCTP_TCB_LOCK(stcb); 5424 if (stcb->asoc.last_used_address == laddr) 5425 /* delete this address */ 5426 stcb->asoc.last_used_address = NULL; 5427 /* 5428 * Now spin through all the nets and purge any ref 5429 * to laddr 5430 */ 5431 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 5432 if (net->ro._s_addr == laddr->ifa) { 5433 /* Yep, purge src address selected */ 5434 RO_NHFREE(&net->ro); 5435 sctp_free_ifa(net->ro._s_addr); 5436 net->ro._s_addr = NULL; 5437 net->src_addr_selected = 0; 5438 } 5439 } 5440 SCTP_TCB_UNLOCK(stcb); 5441 } /* for each tcb */ 5442 /* remove it from the ep list */ 5443 sctp_remove_laddr(laddr); 5444 inp->laddr_count--; 5445 /* update inp_vflag flags */ 5446 sctp_update_ep_vflag(inp); 5447 } 5448 return; 5449 } 5450 5451 /* 5452 * Add the address to the TCB local address restricted list. 5453 * This is a "pending" address list (eg. addresses waiting for an 5454 * ASCONF-ACK response) and cannot be used as a valid source address. 5455 */ 5456 void 5457 sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) 5458 { 5459 struct sctp_laddr *laddr; 5460 struct sctpladdr *list; 5461 5462 /* 5463 * Assumes TCB is locked.. and possibly the INP. May need to 5464 * confirm/fix that if we need it and is not the case. 5465 */ 5466 list = &stcb->asoc.sctp_restricted_addrs; 5467 5468 #ifdef INET6 5469 if (ifa->address.sa.sa_family == AF_INET6) { 5470 if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { 5471 /* Can't bind a non-existent addr. */ 5472 return; 5473 } 5474 } 5475 #endif 5476 /* does the address already exist? */ 5477 LIST_FOREACH(laddr, list, sctp_nxt_addr) { 5478 if (laddr->ifa == ifa) { 5479 return; 5480 } 5481 } 5482 5483 /* add to the list */ 5484 (void)sctp_insert_laddr(list, ifa, 0); 5485 return; 5486 } 5487 5488 /* 5489 * Remove a local address from the TCB local address restricted list 5490 */ 5491 void 5492 sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) 5493 { 5494 struct sctp_inpcb *inp; 5495 struct sctp_laddr *laddr; 5496 5497 /* 5498 * This is called by asconf work. It is assumed that a) The TCB is 5499 * locked and b) The INP is locked. This is true in as much as I can 5500 * trace through the entry asconf code where I did these locks. 5501 * Again, the ASCONF code is a bit different in that it does lock 5502 * the INP during its work often times. This must be since we don't 5503 * want other proc's looking up things while what they are looking 5504 * up is changing :-D 5505 */ 5506 5507 inp = stcb->sctp_ep; 5508 /* if subset bound and don't allow ASCONF's, can't delete last */ 5509 if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) && 5510 sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) { 5511 if (stcb->sctp_ep->laddr_count < 2) { 5512 /* can't delete last address */ 5513 return; 5514 } 5515 } 5516 LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) { 5517 /* remove the address if it exists */ 5518 if (laddr->ifa == NULL) 5519 continue; 5520 if (laddr->ifa == ifa) { 5521 sctp_remove_laddr(laddr); 5522 return; 5523 } 5524 } 5525 5526 /* address not found! */ 5527 return; 5528 } 5529 5530 /* sysctl */ 5531 static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC; 5532 static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR; 5533 5534 #if defined(SCTP_MCORE_INPUT) && defined(SMP) 5535 struct sctp_mcore_ctrl *sctp_mcore_workers = NULL; 5536 int *sctp_cpuarry = NULL; 5537 5538 void 5539 sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use) 5540 { 5541 /* Queue a packet to a processor for the specified core */ 5542 struct sctp_mcore_queue *qent; 5543 struct sctp_mcore_ctrl *wkq; 5544 int need_wake = 0; 5545 5546 if (sctp_mcore_workers == NULL) { 5547 /* Something went way bad during setup */ 5548 sctp_input_with_port(m, off, 0); 5549 return; 5550 } 5551 SCTP_MALLOC(qent, struct sctp_mcore_queue *, 5552 (sizeof(struct sctp_mcore_queue)), 5553 SCTP_M_MCORE); 5554 if (qent == NULL) { 5555 /* This is trouble */ 5556 sctp_input_with_port(m, off, 0); 5557 return; 5558 } 5559 qent->vn = curvnet; 5560 qent->m = m; 5561 qent->off = off; 5562 qent->v6 = 0; 5563 wkq = &sctp_mcore_workers[cpu_to_use]; 5564 SCTP_MCORE_QLOCK(wkq); 5565 5566 TAILQ_INSERT_TAIL(&wkq->que, qent, next); 5567 if (wkq->running == 0) { 5568 need_wake = 1; 5569 } 5570 SCTP_MCORE_QUNLOCK(wkq); 5571 if (need_wake) { 5572 wakeup(&wkq->running); 5573 } 5574 } 5575 5576 static void 5577 sctp_mcore_thread(void *arg) 5578 { 5579 5580 struct sctp_mcore_ctrl *wkq; 5581 struct sctp_mcore_queue *qent; 5582 5583 wkq = (struct sctp_mcore_ctrl *)arg; 5584 struct mbuf *m; 5585 int off, v6; 5586 5587 /* Wait for first tickle */ 5588 SCTP_MCORE_LOCK(wkq); 5589 wkq->running = 0; 5590 msleep(&wkq->running, 5591 &wkq->core_mtx, 5592 0, "wait for pkt", 0); 5593 SCTP_MCORE_UNLOCK(wkq); 5594 5595 /* Bind to our cpu */ 5596 thread_lock(curthread); 5597 sched_bind(curthread, wkq->cpuid); 5598 thread_unlock(curthread); 5599 5600 /* Now lets start working */ 5601 SCTP_MCORE_LOCK(wkq); 5602 /* Now grab lock and go */ 5603 for (;;) { 5604 SCTP_MCORE_QLOCK(wkq); 5605 skip_sleep: 5606 wkq->running = 1; 5607 qent = TAILQ_FIRST(&wkq->que); 5608 if (qent) { 5609 TAILQ_REMOVE(&wkq->que, qent, next); 5610 SCTP_MCORE_QUNLOCK(wkq); 5611 CURVNET_SET(qent->vn); 5612 m = qent->m; 5613 off = qent->off; 5614 v6 = qent->v6; 5615 SCTP_FREE(qent, SCTP_M_MCORE); 5616 if (v6 == 0) { 5617 sctp_input_with_port(m, off, 0); 5618 } else { 5619 SCTP_PRINTF("V6 not yet supported\n"); 5620 sctp_m_freem(m); 5621 } 5622 CURVNET_RESTORE(); 5623 SCTP_MCORE_QLOCK(wkq); 5624 } 5625 wkq->running = 0; 5626 if (!TAILQ_EMPTY(&wkq->que)) { 5627 goto skip_sleep; 5628 } 5629 SCTP_MCORE_QUNLOCK(wkq); 5630 msleep(&wkq->running, 5631 &wkq->core_mtx, 5632 0, "wait for pkt", 0); 5633 } 5634 } 5635 5636 static void 5637 sctp_startup_mcore_threads(void) 5638 { 5639 int i, cpu; 5640 5641 if (mp_ncpus == 1) 5642 return; 5643 5644 if (sctp_mcore_workers != NULL) { 5645 /* 5646 * Already been here in some previous vnet? 5647 */ 5648 return; 5649 } 5650 SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *, 5651 ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)), 5652 SCTP_M_MCORE); 5653 if (sctp_mcore_workers == NULL) { 5654 /* TSNH I hope */ 5655 return; 5656 } 5657 memset(sctp_mcore_workers, 0, ((mp_maxid + 1) * 5658 sizeof(struct sctp_mcore_ctrl))); 5659 /* Init the structures */ 5660 for (i = 0; i <= mp_maxid; i++) { 5661 TAILQ_INIT(&sctp_mcore_workers[i].que); 5662 SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]); 5663 SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]); 5664 sctp_mcore_workers[i].cpuid = i; 5665 } 5666 if (sctp_cpuarry == NULL) { 5667 SCTP_MALLOC(sctp_cpuarry, int *, 5668 (mp_ncpus * sizeof(int)), 5669 SCTP_M_MCORE); 5670 i = 0; 5671 CPU_FOREACH(cpu) { 5672 sctp_cpuarry[i] = cpu; 5673 i++; 5674 } 5675 } 5676 /* Now start them all */ 5677 CPU_FOREACH(cpu) { 5678 (void)kproc_create(sctp_mcore_thread, 5679 (void *)&sctp_mcore_workers[cpu], 5680 &sctp_mcore_workers[cpu].thread_proc, 5681 RFPROC, 5682 SCTP_KTHREAD_PAGES, 5683 SCTP_MCORE_NAME); 5684 } 5685 } 5686 #endif 5687 5688 void 5689 sctp_pcb_init(void) 5690 { 5691 /* 5692 * SCTP initialization for the PCB structures should be called by 5693 * the sctp_init() function. 5694 */ 5695 int i; 5696 struct timeval tv; 5697 5698 if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) { 5699 /* error I was called twice */ 5700 return; 5701 } 5702 SCTP_BASE_VAR(sctp_pcb_initialized) = 1; 5703 5704 #if defined(SCTP_LOCAL_TRACE_BUF) 5705 memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log)); 5706 #endif 5707 #if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) 5708 SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *, 5709 ((mp_maxid + 1) * sizeof(struct sctpstat)), 5710 SCTP_M_MCORE); 5711 #endif 5712 (void)SCTP_GETTIME_TIMEVAL(&tv); 5713 #if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) 5714 memset(SCTP_BASE_STATS, 0, sizeof(struct sctpstat) * (mp_maxid + 1)); 5715 SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t)tv.tv_sec; 5716 SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t)tv.tv_usec; 5717 #else 5718 memset(&SCTP_BASE_STATS, 0, sizeof(struct sctpstat)); 5719 SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t)tv.tv_sec; 5720 SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t)tv.tv_usec; 5721 #endif 5722 /* init the empty list of (All) Endpoints */ 5723 LIST_INIT(&SCTP_BASE_INFO(listhead)); 5724 5725 5726 /* init the hash table of endpoints */ 5727 TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize)); 5728 TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize)); 5729 TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale)); 5730 SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31), 5731 &SCTP_BASE_INFO(hashasocmark)); 5732 SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), 5733 &SCTP_BASE_INFO(hashmark)); 5734 SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), 5735 &SCTP_BASE_INFO(hashtcpmark)); 5736 SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize); 5737 5738 5739 SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH, 5740 &SCTP_BASE_INFO(hashvrfmark)); 5741 5742 SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE, 5743 &SCTP_BASE_INFO(vrf_ifn_hashmark)); 5744 /* init the zones */ 5745 /* 5746 * FIX ME: Should check for NULL returns, but if it does fail we are 5747 * doomed to panic anyways... add later maybe. 5748 */ 5749 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep", 5750 sizeof(struct sctp_inpcb), maxsockets); 5751 5752 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc", 5753 sizeof(struct sctp_tcb), sctp_max_number_of_assoc); 5754 5755 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr", 5756 sizeof(struct sctp_laddr), 5757 (sctp_max_number_of_assoc * sctp_scale_up_for_address)); 5758 5759 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr", 5760 sizeof(struct sctp_nets), 5761 (sctp_max_number_of_assoc * sctp_scale_up_for_address)); 5762 5763 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk", 5764 sizeof(struct sctp_tmit_chunk), 5765 (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); 5766 5767 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq", 5768 sizeof(struct sctp_queued_to_read), 5769 (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); 5770 5771 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out", 5772 sizeof(struct sctp_stream_queue_pending), 5773 (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); 5774 5775 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf", 5776 sizeof(struct sctp_asconf), 5777 (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); 5778 5779 SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack", 5780 sizeof(struct sctp_asconf_ack), 5781 (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); 5782 5783 5784 /* Master Lock INIT for info structure */ 5785 SCTP_INP_INFO_LOCK_INIT(); 5786 SCTP_STATLOG_INIT_LOCK(); 5787 5788 SCTP_IPI_COUNT_INIT(); 5789 SCTP_IPI_ADDR_INIT(); 5790 #ifdef SCTP_PACKET_LOGGING 5791 SCTP_IP_PKTLOG_INIT(); 5792 #endif 5793 LIST_INIT(&SCTP_BASE_INFO(addr_wq)); 5794 5795 SCTP_WQ_ADDR_INIT(); 5796 /* not sure if we need all the counts */ 5797 SCTP_BASE_INFO(ipi_count_ep) = 0; 5798 /* assoc/tcb zone info */ 5799 SCTP_BASE_INFO(ipi_count_asoc) = 0; 5800 /* local addrlist zone info */ 5801 SCTP_BASE_INFO(ipi_count_laddr) = 0; 5802 /* remote addrlist zone info */ 5803 SCTP_BASE_INFO(ipi_count_raddr) = 0; 5804 /* chunk info */ 5805 SCTP_BASE_INFO(ipi_count_chunk) = 0; 5806 5807 /* socket queue zone info */ 5808 SCTP_BASE_INFO(ipi_count_readq) = 0; 5809 5810 /* stream out queue cont */ 5811 SCTP_BASE_INFO(ipi_count_strmoq) = 0; 5812 5813 SCTP_BASE_INFO(ipi_free_strmoq) = 0; 5814 SCTP_BASE_INFO(ipi_free_chunks) = 0; 5815 5816 SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer)); 5817 5818 /* Init the TIMEWAIT list */ 5819 for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) { 5820 LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]); 5821 } 5822 sctp_startup_iterator(); 5823 5824 #if defined(SCTP_MCORE_INPUT) && defined(SMP) 5825 sctp_startup_mcore_threads(); 5826 #endif 5827 5828 /* 5829 * INIT the default VRF which for BSD is the only one, other O/S's 5830 * may have more. But initially they must start with one and then 5831 * add the VRF's as addresses are added. 5832 */ 5833 sctp_init_vrf_list(SCTP_DEFAULT_VRF); 5834 } 5835 5836 /* 5837 * Assumes that the SCTP_BASE_INFO() lock is NOT held. 5838 */ 5839 void 5840 sctp_pcb_finish(void) 5841 { 5842 struct sctp_vrflist *vrf_bucket; 5843 struct sctp_vrf *vrf, *nvrf; 5844 struct sctp_ifn *ifn, *nifn; 5845 struct sctp_ifa *ifa, *nifa; 5846 struct sctpvtaghead *chain; 5847 struct sctp_tagblock *twait_block, *prev_twait_block; 5848 struct sctp_laddr *wi, *nwi; 5849 int i; 5850 struct sctp_iterator *it, *nit; 5851 5852 if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { 5853 SCTP_PRINTF("%s: race condition on teardown.\n", __func__); 5854 return; 5855 } 5856 SCTP_BASE_VAR(sctp_pcb_initialized) = 0; 5857 /* 5858 * In FreeBSD the iterator thread never exits but we do clean up. 5859 * The only way FreeBSD reaches here is if we have VRF's but we 5860 * still add the ifdef to make it compile on old versions. 5861 */ 5862 retry: 5863 SCTP_IPI_ITERATOR_WQ_LOCK(); 5864 /* 5865 * sctp_iterator_worker() might be working on an it entry without 5866 * holding the lock. We won't find it on the list either and 5867 * continue and free/destroy it. While holding the lock, spin, to 5868 * avoid the race condition as sctp_iterator_worker() will have to 5869 * wait to re-acquire the lock. 5870 */ 5871 if (sctp_it_ctl.iterator_running != 0 || sctp_it_ctl.cur_it != NULL) { 5872 SCTP_IPI_ITERATOR_WQ_UNLOCK(); 5873 SCTP_PRINTF("%s: Iterator running while we held the lock. Retry. " 5874 "cur_it=%p\n", __func__, sctp_it_ctl.cur_it); 5875 DELAY(10); 5876 goto retry; 5877 } 5878 TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { 5879 if (it->vn != curvnet) { 5880 continue; 5881 } 5882 TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); 5883 if (it->function_atend != NULL) { 5884 (*it->function_atend) (it->pointer, it->val); 5885 } 5886 SCTP_FREE(it, SCTP_M_ITER); 5887 } 5888 SCTP_IPI_ITERATOR_WQ_UNLOCK(); 5889 SCTP_ITERATOR_LOCK(); 5890 if ((sctp_it_ctl.cur_it) && 5891 (sctp_it_ctl.cur_it->vn == curvnet)) { 5892 sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT; 5893 } 5894 SCTP_ITERATOR_UNLOCK(); 5895 SCTP_OS_TIMER_STOP_DRAIN(&SCTP_BASE_INFO(addr_wq_timer.timer)); 5896 SCTP_WQ_ADDR_LOCK(); 5897 LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) { 5898 LIST_REMOVE(wi, sctp_nxt_addr); 5899 SCTP_DECR_LADDR_COUNT(); 5900 if (wi->action == SCTP_DEL_IP_ADDRESS) { 5901 SCTP_FREE(wi->ifa, SCTP_M_IFA); 5902 } 5903 SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi); 5904 } 5905 SCTP_WQ_ADDR_UNLOCK(); 5906 5907 /* 5908 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is 5909 * destroyed first). 5910 */ 5911 vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))]; 5912 LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) { 5913 LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) { 5914 LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) { 5915 /* free the ifa */ 5916 LIST_REMOVE(ifa, next_bucket); 5917 LIST_REMOVE(ifa, next_ifa); 5918 SCTP_FREE(ifa, SCTP_M_IFA); 5919 } 5920 /* free the ifn */ 5921 LIST_REMOVE(ifn, next_bucket); 5922 LIST_REMOVE(ifn, next_ifn); 5923 SCTP_FREE(ifn, SCTP_M_IFN); 5924 } 5925 SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); 5926 /* free the vrf */ 5927 LIST_REMOVE(vrf, next_vrf); 5928 SCTP_FREE(vrf, SCTP_M_VRF); 5929 } 5930 /* free the vrf hashes */ 5931 SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark)); 5932 SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark)); 5933 5934 /* 5935 * free the TIMEWAIT list elements malloc'd in the function 5936 * sctp_add_vtag_to_timewait()... 5937 */ 5938 for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) { 5939 chain = &SCTP_BASE_INFO(vtag_timewait)[i]; 5940 if (!LIST_EMPTY(chain)) { 5941 prev_twait_block = NULL; 5942 LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { 5943 if (prev_twait_block) { 5944 SCTP_FREE(prev_twait_block, SCTP_M_TIMW); 5945 } 5946 prev_twait_block = twait_block; 5947 } 5948 SCTP_FREE(prev_twait_block, SCTP_M_TIMW); 5949 } 5950 } 5951 5952 /* free the locks and mutexes */ 5953 #ifdef SCTP_PACKET_LOGGING 5954 SCTP_IP_PKTLOG_DESTROY(); 5955 #endif 5956 SCTP_IPI_ADDR_DESTROY(); 5957 SCTP_STATLOG_DESTROY(); 5958 SCTP_INP_INFO_LOCK_DESTROY(); 5959 5960 SCTP_WQ_ADDR_DESTROY(); 5961 5962 /* Get rid of other stuff too. */ 5963 if (SCTP_BASE_INFO(sctp_asochash) != NULL) 5964 SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark)); 5965 if (SCTP_BASE_INFO(sctp_ephash) != NULL) 5966 SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark)); 5967 if (SCTP_BASE_INFO(sctp_tcpephash) != NULL) 5968 SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark)); 5969 5970 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep)); 5971 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc)); 5972 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr)); 5973 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net)); 5974 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk)); 5975 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq)); 5976 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq)); 5977 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf)); 5978 SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack)); 5979 #if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) 5980 SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE); 5981 #endif 5982 } 5983 5984 5985 int 5986 sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m, 5987 int offset, int limit, 5988 struct sockaddr *src, struct sockaddr *dst, 5989 struct sockaddr *altsa, uint16_t port) 5990 { 5991 /* 5992 * grub through the INIT pulling addresses and loading them to the 5993 * nets structure in the asoc. The from address in the mbuf should 5994 * also be loaded (if it is not already). This routine can be called 5995 * with either INIT or INIT-ACK's as long as the m points to the IP 5996 * packet and the offset points to the beginning of the parameters. 5997 */ 5998 struct sctp_inpcb *inp; 5999 struct sctp_nets *net, *nnet, *net_tmp; 6000 struct sctp_paramhdr *phdr, param_buf; 6001 struct sctp_tcb *stcb_tmp; 6002 uint16_t ptype, plen; 6003 struct sockaddr *sa; 6004 uint8_t random_store[SCTP_PARAM_BUFFER_SIZE]; 6005 struct sctp_auth_random *p_random = NULL; 6006 uint16_t random_len = 0; 6007 uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE]; 6008 struct sctp_auth_hmac_algo *hmacs = NULL; 6009 uint16_t hmacs_len = 0; 6010 uint8_t saw_asconf = 0; 6011 uint8_t saw_asconf_ack = 0; 6012 uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE]; 6013 struct sctp_auth_chunk_list *chunks = NULL; 6014 uint16_t num_chunks = 0; 6015 sctp_key_t *new_key; 6016 uint32_t keylen; 6017 int got_random = 0, got_hmacs = 0, got_chklist = 0; 6018 uint8_t peer_supports_ecn; 6019 uint8_t peer_supports_prsctp; 6020 uint8_t peer_supports_auth; 6021 uint8_t peer_supports_asconf; 6022 uint8_t peer_supports_asconf_ack; 6023 uint8_t peer_supports_reconfig; 6024 uint8_t peer_supports_nrsack; 6025 uint8_t peer_supports_pktdrop; 6026 uint8_t peer_supports_idata; 6027 #ifdef INET 6028 struct sockaddr_in sin; 6029 #endif 6030 #ifdef INET6 6031 struct sockaddr_in6 sin6; 6032 #endif 6033 6034 /* First get the destination address setup too. */ 6035 #ifdef INET 6036 memset(&sin, 0, sizeof(sin)); 6037 sin.sin_family = AF_INET; 6038 sin.sin_len = sizeof(sin); 6039 sin.sin_port = stcb->rport; 6040 #endif 6041 #ifdef INET6 6042 memset(&sin6, 0, sizeof(sin6)); 6043 sin6.sin6_family = AF_INET6; 6044 sin6.sin6_len = sizeof(struct sockaddr_in6); 6045 sin6.sin6_port = stcb->rport; 6046 #endif 6047 if (altsa) { 6048 sa = altsa; 6049 } else { 6050 sa = src; 6051 } 6052 peer_supports_idata = 0; 6053 peer_supports_ecn = 0; 6054 peer_supports_prsctp = 0; 6055 peer_supports_auth = 0; 6056 peer_supports_asconf = 0; 6057 peer_supports_reconfig = 0; 6058 peer_supports_nrsack = 0; 6059 peer_supports_pktdrop = 0; 6060 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { 6061 /* mark all addresses that we have currently on the list */ 6062 net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC; 6063 } 6064 /* does the source address already exist? if so skip it */ 6065 inp = stcb->sctp_ep; 6066 atomic_add_int(&stcb->asoc.refcnt, 1); 6067 stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb); 6068 atomic_add_int(&stcb->asoc.refcnt, -1); 6069 6070 if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) { 6071 /* we must add the source address */ 6072 /* no scope set here since we have a tcb already. */ 6073 switch (sa->sa_family) { 6074 #ifdef INET 6075 case AF_INET: 6076 if (stcb->asoc.scope.ipv4_addr_legal) { 6077 if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) { 6078 return (-1); 6079 } 6080 } 6081 break; 6082 #endif 6083 #ifdef INET6 6084 case AF_INET6: 6085 if (stcb->asoc.scope.ipv6_addr_legal) { 6086 if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) { 6087 return (-2); 6088 } 6089 } 6090 break; 6091 #endif 6092 default: 6093 break; 6094 } 6095 } else { 6096 if (net_tmp != NULL && stcb_tmp == stcb) { 6097 net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; 6098 } else if (stcb_tmp != stcb) { 6099 /* It belongs to another association? */ 6100 if (stcb_tmp) 6101 SCTP_TCB_UNLOCK(stcb_tmp); 6102 return (-3); 6103 } 6104 } 6105 if (stcb->asoc.state == 0) { 6106 /* the assoc was freed? */ 6107 return (-4); 6108 } 6109 /* now we must go through each of the params. */ 6110 phdr = sctp_get_next_param(m, offset, ¶m_buf, sizeof(param_buf)); 6111 while (phdr) { 6112 ptype = ntohs(phdr->param_type); 6113 plen = ntohs(phdr->param_length); 6114 /* 6115 * SCTP_PRINTF("ptype => %0x, plen => %d\n", 6116 * (uint32_t)ptype, (int)plen); 6117 */ 6118 if (offset + plen > limit) { 6119 break; 6120 } 6121 if (plen < sizeof(struct sctp_paramhdr)) { 6122 break; 6123 } 6124 #ifdef INET 6125 if (ptype == SCTP_IPV4_ADDRESS) { 6126 if (stcb->asoc.scope.ipv4_addr_legal) { 6127 struct sctp_ipv4addr_param *p4, p4_buf; 6128 6129 /* ok get the v4 address and check/add */ 6130 phdr = sctp_get_next_param(m, offset, 6131 (struct sctp_paramhdr *)&p4_buf, 6132 sizeof(p4_buf)); 6133 if (plen != sizeof(struct sctp_ipv4addr_param) || 6134 phdr == NULL) { 6135 return (-5); 6136 } 6137 p4 = (struct sctp_ipv4addr_param *)phdr; 6138 sin.sin_addr.s_addr = p4->addr; 6139 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 6140 /* Skip multi-cast addresses */ 6141 goto next_param; 6142 } 6143 if ((sin.sin_addr.s_addr == INADDR_BROADCAST) || 6144 (sin.sin_addr.s_addr == INADDR_ANY)) { 6145 goto next_param; 6146 } 6147 sa = (struct sockaddr *)&sin; 6148 inp = stcb->sctp_ep; 6149 atomic_add_int(&stcb->asoc.refcnt, 1); 6150 stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, 6151 dst, stcb); 6152 atomic_add_int(&stcb->asoc.refcnt, -1); 6153 6154 if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || 6155 inp == NULL) { 6156 /* we must add the source address */ 6157 /* 6158 * no scope set since we have a tcb 6159 * already 6160 */ 6161 6162 /* 6163 * we must validate the state again 6164 * here 6165 */ 6166 add_it_now: 6167 if (stcb->asoc.state == 0) { 6168 /* the assoc was freed? */ 6169 return (-7); 6170 } 6171 if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) { 6172 return (-8); 6173 } 6174 } else if (stcb_tmp == stcb) { 6175 if (stcb->asoc.state == 0) { 6176 /* the assoc was freed? */ 6177 return (-10); 6178 } 6179 if (net != NULL) { 6180 /* clear flag */ 6181 net->dest_state &= 6182 ~SCTP_ADDR_NOT_IN_ASSOC; 6183 } 6184 } else { 6185 /* 6186 * strange, address is in another 6187 * assoc? straighten out locks. 6188 */ 6189 if (stcb_tmp) { 6190 if (SCTP_GET_STATE(stcb_tmp) == SCTP_STATE_COOKIE_WAIT) { 6191 struct mbuf *op_err; 6192 char msg[SCTP_DIAG_INFO_LEN]; 6193 6194 /* 6195 * in setup state we 6196 * abort this guy 6197 */ 6198 SCTP_SNPRINTF(msg, sizeof(msg), 6199 "%s:%d at %s", __FILE__, __LINE__, __func__); 6200 op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), 6201 msg); 6202 sctp_abort_an_association(stcb_tmp->sctp_ep, 6203 stcb_tmp, op_err, 6204 SCTP_SO_NOT_LOCKED); 6205 goto add_it_now; 6206 } 6207 SCTP_TCB_UNLOCK(stcb_tmp); 6208 } 6209 6210 if (stcb->asoc.state == 0) { 6211 /* the assoc was freed? */ 6212 return (-12); 6213 } 6214 return (-13); 6215 } 6216 } 6217 } else 6218 #endif 6219 #ifdef INET6 6220 if (ptype == SCTP_IPV6_ADDRESS) { 6221 if (stcb->asoc.scope.ipv6_addr_legal) { 6222 /* ok get the v6 address and check/add */ 6223 struct sctp_ipv6addr_param *p6, p6_buf; 6224 6225 phdr = sctp_get_next_param(m, offset, 6226 (struct sctp_paramhdr *)&p6_buf, 6227 sizeof(p6_buf)); 6228 if (plen != sizeof(struct sctp_ipv6addr_param) || 6229 phdr == NULL) { 6230 return (-14); 6231 } 6232 p6 = (struct sctp_ipv6addr_param *)phdr; 6233 memcpy((caddr_t)&sin6.sin6_addr, p6->addr, 6234 sizeof(p6->addr)); 6235 if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) { 6236 /* Skip multi-cast addresses */ 6237 goto next_param; 6238 } 6239 if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) { 6240 /* 6241 * Link local make no sense without 6242 * scope 6243 */ 6244 goto next_param; 6245 } 6246 sa = (struct sockaddr *)&sin6; 6247 inp = stcb->sctp_ep; 6248 atomic_add_int(&stcb->asoc.refcnt, 1); 6249 stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, 6250 dst, stcb); 6251 atomic_add_int(&stcb->asoc.refcnt, -1); 6252 if (stcb_tmp == NULL && 6253 (inp == stcb->sctp_ep || inp == NULL)) { 6254 /* 6255 * we must validate the state again 6256 * here 6257 */ 6258 add_it_now6: 6259 if (stcb->asoc.state == 0) { 6260 /* the assoc was freed? */ 6261 return (-16); 6262 } 6263 /* 6264 * we must add the address, no scope 6265 * set 6266 */ 6267 if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) { 6268 return (-17); 6269 } 6270 } else if (stcb_tmp == stcb) { 6271 /* 6272 * we must validate the state again 6273 * here 6274 */ 6275 if (stcb->asoc.state == 0) { 6276 /* the assoc was freed? */ 6277 return (-19); 6278 } 6279 if (net != NULL) { 6280 /* clear flag */ 6281 net->dest_state &= 6282 ~SCTP_ADDR_NOT_IN_ASSOC; 6283 } 6284 } else { 6285 /* 6286 * strange, address is in another 6287 * assoc? straighten out locks. 6288 */ 6289 if (stcb_tmp) { 6290 if (SCTP_GET_STATE(stcb_tmp) == SCTP_STATE_COOKIE_WAIT) { 6291 struct mbuf *op_err; 6292 char msg[SCTP_DIAG_INFO_LEN]; 6293 6294 /* 6295 * in setup state we 6296 * abort this guy 6297 */ 6298 SCTP_SNPRINTF(msg, sizeof(msg), 6299 "%s:%d at %s", __FILE__, __LINE__, __func__); 6300 op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), 6301 msg); 6302 sctp_abort_an_association(stcb_tmp->sctp_ep, 6303 stcb_tmp, op_err, 6304 SCTP_SO_NOT_LOCKED); 6305 goto add_it_now6; 6306 } 6307 SCTP_TCB_UNLOCK(stcb_tmp); 6308 } 6309 if (stcb->asoc.state == 0) { 6310 /* the assoc was freed? */ 6311 return (-21); 6312 } 6313 return (-22); 6314 } 6315 } 6316 } else 6317 #endif 6318 if (ptype == SCTP_ECN_CAPABLE) { 6319 peer_supports_ecn = 1; 6320 } else if (ptype == SCTP_ULP_ADAPTATION) { 6321 if (stcb->asoc.state != SCTP_STATE_OPEN) { 6322 struct sctp_adaptation_layer_indication ai, 6323 *aip; 6324 6325 phdr = sctp_get_next_param(m, offset, 6326 (struct sctp_paramhdr *)&ai, sizeof(ai)); 6327 aip = (struct sctp_adaptation_layer_indication *)phdr; 6328 if (aip) { 6329 stcb->asoc.peers_adaptation = ntohl(aip->indication); 6330 stcb->asoc.adaptation_needed = 1; 6331 } 6332 } 6333 } else if (ptype == SCTP_SET_PRIM_ADDR) { 6334 struct sctp_asconf_addr_param lstore, *fee; 6335 int lptype; 6336 struct sockaddr *lsa = NULL; 6337 #ifdef INET 6338 struct sctp_asconf_addrv4_param *fii; 6339 #endif 6340 6341 if (stcb->asoc.asconf_supported == 0) { 6342 return (-100); 6343 } 6344 if (plen > sizeof(lstore)) { 6345 return (-23); 6346 } 6347 if (plen < sizeof(struct sctp_asconf_addrv4_param)) { 6348 return (-101); 6349 } 6350 phdr = sctp_get_next_param(m, offset, 6351 (struct sctp_paramhdr *)&lstore, 6352 plen); 6353 if (phdr == NULL) { 6354 return (-24); 6355 } 6356 fee = (struct sctp_asconf_addr_param *)phdr; 6357 lptype = ntohs(fee->addrp.ph.param_type); 6358 switch (lptype) { 6359 #ifdef INET 6360 case SCTP_IPV4_ADDRESS: 6361 if (plen != 6362 sizeof(struct sctp_asconf_addrv4_param)) { 6363 SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n", 6364 (int)sizeof(struct sctp_asconf_addrv4_param), 6365 plen); 6366 } else { 6367 fii = (struct sctp_asconf_addrv4_param *)fee; 6368 sin.sin_addr.s_addr = fii->addrp.addr; 6369 lsa = (struct sockaddr *)&sin; 6370 } 6371 break; 6372 #endif 6373 #ifdef INET6 6374 case SCTP_IPV6_ADDRESS: 6375 if (plen != 6376 sizeof(struct sctp_asconf_addr_param)) { 6377 SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n", 6378 (int)sizeof(struct sctp_asconf_addr_param), 6379 plen); 6380 } else { 6381 memcpy(sin6.sin6_addr.s6_addr, 6382 fee->addrp.addr, 6383 sizeof(fee->addrp.addr)); 6384 lsa = (struct sockaddr *)&sin6; 6385 } 6386 break; 6387 #endif 6388 default: 6389 break; 6390 } 6391 if (lsa) { 6392 (void)sctp_set_primary_addr(stcb, sa, NULL); 6393 } 6394 } else if (ptype == SCTP_HAS_NAT_SUPPORT) { 6395 stcb->asoc.peer_supports_nat = 1; 6396 } else if (ptype == SCTP_PRSCTP_SUPPORTED) { 6397 /* Peer supports pr-sctp */ 6398 peer_supports_prsctp = 1; 6399 } else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) { 6400 /* A supported extension chunk */ 6401 struct sctp_supported_chunk_types_param *pr_supported; 6402 uint8_t local_store[SCTP_PARAM_BUFFER_SIZE]; 6403 int num_ent, i; 6404 6405 if (plen > sizeof(local_store)) { 6406 return (-35); 6407 } 6408 phdr = sctp_get_next_param(m, offset, 6409 (struct sctp_paramhdr *)&local_store, plen); 6410 if (phdr == NULL) { 6411 return (-25); 6412 } 6413 pr_supported = (struct sctp_supported_chunk_types_param *)phdr; 6414 num_ent = plen - sizeof(struct sctp_paramhdr); 6415 for (i = 0; i < num_ent; i++) { 6416 switch (pr_supported->chunk_types[i]) { 6417 case SCTP_ASCONF: 6418 peer_supports_asconf = 1; 6419 break; 6420 case SCTP_ASCONF_ACK: 6421 peer_supports_asconf_ack = 1; 6422 break; 6423 case SCTP_FORWARD_CUM_TSN: 6424 peer_supports_prsctp = 1; 6425 break; 6426 case SCTP_PACKET_DROPPED: 6427 peer_supports_pktdrop = 1; 6428 break; 6429 case SCTP_NR_SELECTIVE_ACK: 6430 peer_supports_nrsack = 1; 6431 break; 6432 case SCTP_STREAM_RESET: 6433 peer_supports_reconfig = 1; 6434 break; 6435 case SCTP_AUTHENTICATION: 6436 peer_supports_auth = 1; 6437 break; 6438 case SCTP_IDATA: 6439 peer_supports_idata = 1; 6440 break; 6441 default: 6442 /* one I have not learned yet */ 6443 break; 6444 6445 } 6446 } 6447 } else if (ptype == SCTP_RANDOM) { 6448 if (plen > sizeof(random_store)) 6449 break; 6450 if (got_random) { 6451 /* already processed a RANDOM */ 6452 goto next_param; 6453 } 6454 phdr = sctp_get_next_param(m, offset, 6455 (struct sctp_paramhdr *)random_store, 6456 plen); 6457 if (phdr == NULL) 6458 return (-26); 6459 p_random = (struct sctp_auth_random *)phdr; 6460 random_len = plen - sizeof(*p_random); 6461 /* enforce the random length */ 6462 if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) { 6463 SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n"); 6464 return (-27); 6465 } 6466 got_random = 1; 6467 } else if (ptype == SCTP_HMAC_LIST) { 6468 uint16_t num_hmacs; 6469 uint16_t i; 6470 6471 if (plen > sizeof(hmacs_store)) 6472 break; 6473 if (got_hmacs) { 6474 /* already processed a HMAC list */ 6475 goto next_param; 6476 } 6477 phdr = sctp_get_next_param(m, offset, 6478 (struct sctp_paramhdr *)hmacs_store, 6479 plen); 6480 if (phdr == NULL) 6481 return (-28); 6482 hmacs = (struct sctp_auth_hmac_algo *)phdr; 6483 hmacs_len = plen - sizeof(*hmacs); 6484 num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]); 6485 /* validate the hmac list */ 6486 if (sctp_verify_hmac_param(hmacs, num_hmacs)) { 6487 return (-29); 6488 } 6489 if (stcb->asoc.peer_hmacs != NULL) 6490 sctp_free_hmaclist(stcb->asoc.peer_hmacs); 6491 stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs); 6492 if (stcb->asoc.peer_hmacs != NULL) { 6493 for (i = 0; i < num_hmacs; i++) { 6494 (void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs, 6495 ntohs(hmacs->hmac_ids[i])); 6496 } 6497 } 6498 got_hmacs = 1; 6499 } else if (ptype == SCTP_CHUNK_LIST) { 6500 int i; 6501 6502 if (plen > sizeof(chunks_store)) 6503 break; 6504 if (got_chklist) { 6505 /* already processed a Chunks list */ 6506 goto next_param; 6507 } 6508 phdr = sctp_get_next_param(m, offset, 6509 (struct sctp_paramhdr *)chunks_store, 6510 plen); 6511 if (phdr == NULL) 6512 return (-30); 6513 chunks = (struct sctp_auth_chunk_list *)phdr; 6514 num_chunks = plen - sizeof(*chunks); 6515 if (stcb->asoc.peer_auth_chunks != NULL) 6516 sctp_clear_chunklist(stcb->asoc.peer_auth_chunks); 6517 else 6518 stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist(); 6519 for (i = 0; i < num_chunks; i++) { 6520 (void)sctp_auth_add_chunk(chunks->chunk_types[i], 6521 stcb->asoc.peer_auth_chunks); 6522 /* record asconf/asconf-ack if listed */ 6523 if (chunks->chunk_types[i] == SCTP_ASCONF) 6524 saw_asconf = 1; 6525 if (chunks->chunk_types[i] == SCTP_ASCONF_ACK) 6526 saw_asconf_ack = 1; 6527 6528 } 6529 got_chklist = 1; 6530 } else if ((ptype == SCTP_HEARTBEAT_INFO) || 6531 (ptype == SCTP_STATE_COOKIE) || 6532 (ptype == SCTP_UNRECOG_PARAM) || 6533 (ptype == SCTP_COOKIE_PRESERVE) || 6534 (ptype == SCTP_SUPPORTED_ADDRTYPE) || 6535 (ptype == SCTP_ADD_IP_ADDRESS) || 6536 (ptype == SCTP_DEL_IP_ADDRESS) || 6537 (ptype == SCTP_ERROR_CAUSE_IND) || 6538 (ptype == SCTP_SUCCESS_REPORT)) { 6539 /* don't care */ ; 6540 } else { 6541 if ((ptype & 0x8000) == 0x0000) { 6542 /* 6543 * must stop processing the rest of the 6544 * param's. Any report bits were handled 6545 * with the call to 6546 * sctp_arethere_unrecognized_parameters() 6547 * when the INIT or INIT-ACK was first seen. 6548 */ 6549 break; 6550 } 6551 } 6552 6553 next_param: 6554 offset += SCTP_SIZE32(plen); 6555 if (offset >= limit) { 6556 break; 6557 } 6558 phdr = sctp_get_next_param(m, offset, ¶m_buf, 6559 sizeof(param_buf)); 6560 } 6561 /* Now check to see if we need to purge any addresses */ 6562 TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) { 6563 if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) == 6564 SCTP_ADDR_NOT_IN_ASSOC) { 6565 /* This address has been removed from the asoc */ 6566 /* remove and free it */ 6567 stcb->asoc.numnets--; 6568 TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next); 6569 sctp_free_remote_addr(net); 6570 if (net == stcb->asoc.primary_destination) { 6571 stcb->asoc.primary_destination = NULL; 6572 sctp_select_primary_destination(stcb); 6573 } 6574 } 6575 } 6576 if ((stcb->asoc.ecn_supported == 1) && 6577 (peer_supports_ecn == 0)) { 6578 stcb->asoc.ecn_supported = 0; 6579 } 6580 if ((stcb->asoc.prsctp_supported == 1) && 6581 (peer_supports_prsctp == 0)) { 6582 stcb->asoc.prsctp_supported = 0; 6583 } 6584 if ((stcb->asoc.auth_supported == 1) && 6585 ((peer_supports_auth == 0) || 6586 (got_random == 0) || (got_hmacs == 0))) { 6587 stcb->asoc.auth_supported = 0; 6588 } 6589 if ((stcb->asoc.asconf_supported == 1) && 6590 ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) || 6591 (stcb->asoc.auth_supported == 0) || 6592 (saw_asconf == 0) || (saw_asconf_ack == 0))) { 6593 stcb->asoc.asconf_supported = 0; 6594 } 6595 if ((stcb->asoc.reconfig_supported == 1) && 6596 (peer_supports_reconfig == 0)) { 6597 stcb->asoc.reconfig_supported = 0; 6598 } 6599 if ((stcb->asoc.idata_supported == 1) && 6600 (peer_supports_idata == 0)) { 6601 stcb->asoc.idata_supported = 0; 6602 } 6603 if ((stcb->asoc.nrsack_supported == 1) && 6604 (peer_supports_nrsack == 0)) { 6605 stcb->asoc.nrsack_supported = 0; 6606 } 6607 if ((stcb->asoc.pktdrop_supported == 1) && 6608 (peer_supports_pktdrop == 0)) { 6609 stcb->asoc.pktdrop_supported = 0; 6610 } 6611 /* validate authentication required parameters */ 6612 if ((peer_supports_auth == 0) && (got_chklist == 1)) { 6613 /* peer does not support auth but sent a chunks list? */ 6614 return (-31); 6615 } 6616 if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) { 6617 /* peer supports asconf but not auth? */ 6618 return (-32); 6619 } else if ((peer_supports_asconf == 1) && 6620 (peer_supports_auth == 1) && 6621 ((saw_asconf == 0) || (saw_asconf_ack == 0))) { 6622 return (-33); 6623 } 6624 /* concatenate the full random key */ 6625 keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len; 6626 if (chunks != NULL) { 6627 keylen += sizeof(*chunks) + num_chunks; 6628 } 6629 new_key = sctp_alloc_key(keylen); 6630 if (new_key != NULL) { 6631 /* copy in the RANDOM */ 6632 if (p_random != NULL) { 6633 keylen = sizeof(*p_random) + random_len; 6634 memcpy(new_key->key, p_random, keylen); 6635 } else { 6636 keylen = 0; 6637 } 6638 /* append in the AUTH chunks */ 6639 if (chunks != NULL) { 6640 memcpy(new_key->key + keylen, chunks, 6641 sizeof(*chunks) + num_chunks); 6642 keylen += sizeof(*chunks) + num_chunks; 6643 } 6644 /* append in the HMACs */ 6645 if (hmacs != NULL) { 6646 memcpy(new_key->key + keylen, hmacs, 6647 sizeof(*hmacs) + hmacs_len); 6648 } 6649 } else { 6650 /* failed to get memory for the key */ 6651 return (-34); 6652 } 6653 if (stcb->asoc.authinfo.peer_random != NULL) 6654 sctp_free_key(stcb->asoc.authinfo.peer_random); 6655 stcb->asoc.authinfo.peer_random = new_key; 6656 sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid); 6657 sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid); 6658 6659 return (0); 6660 } 6661 6662 int 6663 sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa, 6664 struct sctp_nets *net) 6665 { 6666 /* make sure the requested primary address exists in the assoc */ 6667 if (net == NULL && sa) 6668 net = sctp_findnet(stcb, sa); 6669 6670 if (net == NULL) { 6671 /* didn't find the requested primary address! */ 6672 return (-1); 6673 } else { 6674 /* set the primary address */ 6675 if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { 6676 /* Must be confirmed, so queue to set */ 6677 net->dest_state |= SCTP_ADDR_REQ_PRIMARY; 6678 return (0); 6679 } 6680 stcb->asoc.primary_destination = net; 6681 if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) { 6682 sctp_free_remote_addr(stcb->asoc.alternate); 6683 stcb->asoc.alternate = NULL; 6684 } 6685 net = TAILQ_FIRST(&stcb->asoc.nets); 6686 if (net != stcb->asoc.primary_destination) { 6687 /* 6688 * first one on the list is NOT the primary 6689 * sctp_cmpaddr() is much more efficient if the 6690 * primary is the first on the list, make it so. 6691 */ 6692 TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); 6693 TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); 6694 } 6695 return (0); 6696 } 6697 } 6698 6699 int 6700 sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now) 6701 { 6702 /* 6703 * This function serves two purposes. It will see if a TAG can be 6704 * re-used and return 1 for yes it is ok and 0 for don't use that 6705 * tag. A secondary function it will do is purge out old tags that 6706 * can be removed. 6707 */ 6708 struct sctpvtaghead *chain; 6709 struct sctp_tagblock *twait_block; 6710 struct sctpasochead *head; 6711 struct sctp_tcb *stcb; 6712 int i; 6713 6714 SCTP_INP_INFO_RLOCK(); 6715 head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag, 6716 SCTP_BASE_INFO(hashasocmark))]; 6717 LIST_FOREACH(stcb, head, sctp_asocs) { 6718 /* 6719 * We choose not to lock anything here. TCB's can't be 6720 * removed since we have the read lock, so they can't be 6721 * freed on us, same thing for the INP. I may be wrong with 6722 * this assumption, but we will go with it for now :-) 6723 */ 6724 if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { 6725 continue; 6726 } 6727 if (stcb->asoc.my_vtag == tag) { 6728 /* candidate */ 6729 if (stcb->rport != rport) { 6730 continue; 6731 } 6732 if (stcb->sctp_ep->sctp_lport != lport) { 6733 continue; 6734 } 6735 /* Its a used tag set */ 6736 SCTP_INP_INFO_RUNLOCK(); 6737 return (0); 6738 } 6739 } 6740 chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; 6741 /* Now what about timed wait ? */ 6742 LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { 6743 /* 6744 * Block(s) are present, lets see if we have this tag in the 6745 * list 6746 */ 6747 for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { 6748 if (twait_block->vtag_block[i].v_tag == 0) { 6749 /* not used */ 6750 continue; 6751 } else if ((long)twait_block->vtag_block[i].tv_sec_at_expire < 6752 now->tv_sec) { 6753 /* Audit expires this guy */ 6754 twait_block->vtag_block[i].tv_sec_at_expire = 0; 6755 twait_block->vtag_block[i].v_tag = 0; 6756 twait_block->vtag_block[i].lport = 0; 6757 twait_block->vtag_block[i].rport = 0; 6758 } else if ((twait_block->vtag_block[i].v_tag == tag) && 6759 (twait_block->vtag_block[i].lport == lport) && 6760 (twait_block->vtag_block[i].rport == rport)) { 6761 /* Bad tag, sorry :< */ 6762 SCTP_INP_INFO_RUNLOCK(); 6763 return (0); 6764 } 6765 } 6766 } 6767 SCTP_INP_INFO_RUNLOCK(); 6768 return (1); 6769 } 6770 6771 static void 6772 sctp_drain_mbufs(struct sctp_tcb *stcb) 6773 { 6774 /* 6775 * We must hunt this association for MBUF's past the cumack (i.e. 6776 * out of order data that we can renege on). 6777 */ 6778 struct sctp_association *asoc; 6779 struct sctp_tmit_chunk *chk, *nchk; 6780 uint32_t cumulative_tsn_p1; 6781 struct sctp_queued_to_read *control, *ncontrol; 6782 int cnt, strmat; 6783 uint32_t gap, i; 6784 int fnd = 0; 6785 6786 /* We look for anything larger than the cum-ack + 1 */ 6787 6788 asoc = &stcb->asoc; 6789 if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) { 6790 /* none we can reneg on. */ 6791 return; 6792 } 6793 SCTP_STAT_INCR(sctps_protocol_drains_done); 6794 cumulative_tsn_p1 = asoc->cumulative_tsn + 1; 6795 cnt = 0; 6796 /* Ok that was fun, now we will drain all the inbound streams? */ 6797 for (strmat = 0; strmat < asoc->streamincnt; strmat++) { 6798 TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].inqueue, next_instrm, ncontrol) { 6799 #ifdef INVARIANTS 6800 if (control->on_strm_q != SCTP_ON_ORDERED) { 6801 panic("Huh control: %p on_q: %d -- not ordered?", 6802 control, control->on_strm_q); 6803 } 6804 #endif 6805 if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) { 6806 /* Yep it is above cum-ack */ 6807 cnt++; 6808 SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn); 6809 KASSERT(control->length > 0, ("control has zero length")); 6810 if (asoc->size_on_all_streams >= control->length) { 6811 asoc->size_on_all_streams -= control->length; 6812 } else { 6813 #ifdef INVARIANTS 6814 panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); 6815 #else 6816 asoc->size_on_all_streams = 0; 6817 #endif 6818 } 6819 sctp_ucount_decr(asoc->cnt_on_all_streams); 6820 SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); 6821 if (control->on_read_q) { 6822 TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next); 6823 control->on_read_q = 0; 6824 } 6825 TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, control, next_instrm); 6826 control->on_strm_q = 0; 6827 if (control->data) { 6828 sctp_m_freem(control->data); 6829 control->data = NULL; 6830 } 6831 sctp_free_remote_addr(control->whoFrom); 6832 /* Now its reasm? */ 6833 TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) { 6834 cnt++; 6835 SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn); 6836 KASSERT(chk->send_size > 0, ("chunk has zero length")); 6837 if (asoc->size_on_reasm_queue >= chk->send_size) { 6838 asoc->size_on_reasm_queue -= chk->send_size; 6839 } else { 6840 #ifdef INVARIANTS 6841 panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size); 6842 #else 6843 asoc->size_on_reasm_queue = 0; 6844 #endif 6845 } 6846 sctp_ucount_decr(asoc->cnt_on_reasm_queue); 6847 SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); 6848 TAILQ_REMOVE(&control->reasm, chk, sctp_next); 6849 if (chk->data) { 6850 sctp_m_freem(chk->data); 6851 chk->data = NULL; 6852 } 6853 sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); 6854 } 6855 sctp_free_a_readq(stcb, control); 6856 } 6857 } 6858 TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].uno_inqueue, next_instrm, ncontrol) { 6859 #ifdef INVARIANTS 6860 if (control->on_strm_q != SCTP_ON_UNORDERED) { 6861 panic("Huh control: %p on_q: %d -- not unordered?", 6862 control, control->on_strm_q); 6863 } 6864 #endif 6865 if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) { 6866 /* Yep it is above cum-ack */ 6867 cnt++; 6868 SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn); 6869 KASSERT(control->length > 0, ("control has zero length")); 6870 if (asoc->size_on_all_streams >= control->length) { 6871 asoc->size_on_all_streams -= control->length; 6872 } else { 6873 #ifdef INVARIANTS 6874 panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); 6875 #else 6876 asoc->size_on_all_streams = 0; 6877 #endif 6878 } 6879 sctp_ucount_decr(asoc->cnt_on_all_streams); 6880 SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); 6881 if (control->on_read_q) { 6882 TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next); 6883 control->on_read_q = 0; 6884 } 6885 TAILQ_REMOVE(&asoc->strmin[strmat].uno_inqueue, control, next_instrm); 6886 control->on_strm_q = 0; 6887 if (control->data) { 6888 sctp_m_freem(control->data); 6889 control->data = NULL; 6890 } 6891 sctp_free_remote_addr(control->whoFrom); 6892 /* Now its reasm? */ 6893 TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) { 6894 cnt++; 6895 SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn); 6896 KASSERT(chk->send_size > 0, ("chunk has zero length")); 6897 if (asoc->size_on_reasm_queue >= chk->send_size) { 6898 asoc->size_on_reasm_queue -= chk->send_size; 6899 } else { 6900 #ifdef INVARIANTS 6901 panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size); 6902 #else 6903 asoc->size_on_reasm_queue = 0; 6904 #endif 6905 } 6906 sctp_ucount_decr(asoc->cnt_on_reasm_queue); 6907 SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); 6908 TAILQ_REMOVE(&control->reasm, chk, sctp_next); 6909 if (chk->data) { 6910 sctp_m_freem(chk->data); 6911 chk->data = NULL; 6912 } 6913 sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); 6914 } 6915 sctp_free_a_readq(stcb, control); 6916 } 6917 } 6918 } 6919 if (cnt) { 6920 /* We must back down to see what the new highest is */ 6921 for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) { 6922 SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn); 6923 if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { 6924 asoc->highest_tsn_inside_map = i; 6925 fnd = 1; 6926 break; 6927 } 6928 } 6929 if (!fnd) { 6930 asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1; 6931 } 6932 6933 /* 6934 * Question, should we go through the delivery queue? The 6935 * only reason things are on here is the app not reading OR 6936 * a p-d-api up. An attacker COULD send enough in to 6937 * initiate the PD-API and then send a bunch of stuff to 6938 * other streams... these would wind up on the delivery 6939 * queue.. and then we would not get to them. But in order 6940 * to do this I then have to back-track and un-deliver 6941 * sequence numbers in streams.. el-yucko. I think for now 6942 * we will NOT look at the delivery queue and leave it to be 6943 * something to consider later. An alternative would be to 6944 * abort the P-D-API with a notification and then deliver 6945 * the data.... Or another method might be to keep track of 6946 * how many times the situation occurs and if we see a 6947 * possible attack underway just abort the association. 6948 */ 6949 #ifdef SCTP_DEBUG 6950 SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt); 6951 #endif 6952 /* 6953 * Now do we need to find a new 6954 * asoc->highest_tsn_inside_map? 6955 */ 6956 asoc->last_revoke_count = cnt; 6957 sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, 6958 SCTP_FROM_SCTP_PCB + SCTP_LOC_11); 6959 /* sa_ignore NO_NULL_CHK */ 6960 sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED); 6961 sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED); 6962 } 6963 /* 6964 * Another issue, in un-setting the TSN's in the mapping array we 6965 * DID NOT adjust the highest_tsn marker. This will cause one of 6966 * two things to occur. It may cause us to do extra work in checking 6967 * for our mapping array movement. More importantly it may cause us 6968 * to SACK every datagram. This may not be a bad thing though since 6969 * we will recover once we get our cum-ack above and all this stuff 6970 * we dumped recovered. 6971 */ 6972 } 6973 6974 void 6975 sctp_drain() 6976 { 6977 /* 6978 * We must walk the PCB lists for ALL associations here. The system 6979 * is LOW on MBUF's and needs help. This is where reneging will 6980 * occur. We really hope this does NOT happen! 6981 */ 6982 VNET_ITERATOR_DECL(vnet_iter); 6983 VNET_LIST_RLOCK_NOSLEEP(); 6984 VNET_FOREACH(vnet_iter) { 6985 CURVNET_SET(vnet_iter); 6986 struct sctp_inpcb *inp; 6987 struct sctp_tcb *stcb; 6988 6989 SCTP_STAT_INCR(sctps_protocol_drain_calls); 6990 if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { 6991 #ifdef VIMAGE 6992 continue; 6993 #else 6994 return; 6995 #endif 6996 } 6997 SCTP_INP_INFO_RLOCK(); 6998 LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { 6999 /* For each endpoint */ 7000 SCTP_INP_RLOCK(inp); 7001 LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { 7002 /* For each association */ 7003 SCTP_TCB_LOCK(stcb); 7004 sctp_drain_mbufs(stcb); 7005 SCTP_TCB_UNLOCK(stcb); 7006 } 7007 SCTP_INP_RUNLOCK(inp); 7008 } 7009 SCTP_INP_INFO_RUNLOCK(); 7010 CURVNET_RESTORE(); 7011 } 7012 VNET_LIST_RUNLOCK_NOSLEEP(); 7013 } 7014 7015 /* 7016 * start a new iterator 7017 * iterates through all endpoints and associations based on the pcb_state 7018 * flags and asoc_state. "af" (mandatory) is executed for all matching 7019 * assocs and "ef" (optional) is executed when the iterator completes. 7020 * "inpf" (optional) is executed for each new endpoint as it is being 7021 * iterated through. inpe (optional) is called when the inp completes 7022 * its way through all the stcbs. 7023 */ 7024 int 7025 sctp_initiate_iterator(inp_func inpf, 7026 asoc_func af, 7027 inp_func inpe, 7028 uint32_t pcb_state, 7029 uint32_t pcb_features, 7030 uint32_t asoc_state, 7031 void *argp, 7032 uint32_t argi, 7033 end_func ef, 7034 struct sctp_inpcb *s_inp, 7035 uint8_t chunk_output_off) 7036 { 7037 struct sctp_iterator *it = NULL; 7038 7039 if (af == NULL) { 7040 return (-1); 7041 } 7042 if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { 7043 SCTP_PRINTF("%s: abort on initialize being %d\n", __func__, 7044 SCTP_BASE_VAR(sctp_pcb_initialized)); 7045 return (-1); 7046 } 7047 SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator), 7048 SCTP_M_ITER); 7049 if (it == NULL) { 7050 SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); 7051 return (-1); 7052 } 7053 memset(it, 0, sizeof(*it)); 7054 it->function_assoc = af; 7055 it->function_inp = inpf; 7056 if (inpf) 7057 it->done_current_ep = 0; 7058 else 7059 it->done_current_ep = 1; 7060 it->function_atend = ef; 7061 it->pointer = argp; 7062 it->val = argi; 7063 it->pcb_flags = pcb_state; 7064 it->pcb_features = pcb_features; 7065 it->asoc_state = asoc_state; 7066 it->function_inp_end = inpe; 7067 it->no_chunk_output = chunk_output_off; 7068 it->vn = curvnet; 7069 if (s_inp) { 7070 /* Assume lock is held here */ 7071 it->inp = s_inp; 7072 SCTP_INP_INCR_REF(it->inp); 7073 it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP; 7074 } else { 7075 SCTP_INP_INFO_RLOCK(); 7076 it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead)); 7077 if (it->inp) { 7078 SCTP_INP_INCR_REF(it->inp); 7079 } 7080 SCTP_INP_INFO_RUNLOCK(); 7081 it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP; 7082 7083 } 7084 SCTP_IPI_ITERATOR_WQ_LOCK(); 7085 if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { 7086 SCTP_IPI_ITERATOR_WQ_UNLOCK(); 7087 SCTP_PRINTF("%s: rollback on initialize being %d it=%p\n", __func__, 7088 SCTP_BASE_VAR(sctp_pcb_initialized), it); 7089 SCTP_FREE(it, SCTP_M_ITER); 7090 return (-1); 7091 } 7092 TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); 7093 if (sctp_it_ctl.iterator_running == 0) { 7094 sctp_wakeup_iterator(); 7095 } 7096 SCTP_IPI_ITERATOR_WQ_UNLOCK(); 7097 /* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */ 7098 return (0); 7099 } 7100