1 /* SCTP kernel implementation 2 * (C) Copyright IBM Corp. 2001, 2004 3 * Copyright (c) 1999-2000 Cisco, Inc. 4 * Copyright (c) 1999-2001 Motorola, Inc. 5 * Copyright (c) 2001 Intel Corp. 6 * Copyright (c) 2001 Nokia, Inc. 7 * Copyright (c) 2001 La Monte H.P. Yarroll 8 * 9 * This file is part of the SCTP kernel implementation 10 * 11 * Initialization/cleanup for SCTP protocol support. 12 * 13 * This SCTP implementation is free software; 14 * you can redistribute it and/or modify it under the terms of 15 * the GNU General Public License as published by 16 * the Free Software Foundation; either version 2, or (at your option) 17 * any later version. 18 * 19 * This SCTP implementation is distributed in the hope that it 20 * will be useful, but WITHOUT ANY WARRANTY; without even the implied 21 * ************************ 22 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. 23 * See the GNU General Public License for more details. 24 * 25 * You should have received a copy of the GNU General Public License 26 * along with GNU CC; see the file COPYING. If not, write to 27 * the Free Software Foundation, 59 Temple Place - Suite 330, 28 * Boston, MA 02111-1307, USA. 29 * 30 * Please send any bug reports or fixes you make to the 31 * email address(es): 32 * lksctp developers <linux-sctp@vger.kernel.org> 33 * 34 * Written or modified by: 35 * La Monte H.P. Yarroll <piggy@acm.org> 36 * Karl Knutson <karl@athena.chicago.il.us> 37 * Jon Grimm <jgrimm@us.ibm.com> 38 * Sridhar Samudrala <sri@us.ibm.com> 39 * Daisy Chang <daisyc@us.ibm.com> 40 * Ardelle Fan <ardelle.fan@intel.com> 41 */ 42 43 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 44 45 #include <linux/module.h> 46 #include <linux/init.h> 47 #include <linux/netdevice.h> 48 #include <linux/inetdevice.h> 49 #include <linux/seq_file.h> 50 #include <linux/bootmem.h> 51 #include <linux/highmem.h> 52 #include <linux/swap.h> 53 #include <linux/slab.h> 54 #include <net/net_namespace.h> 55 #include <net/protocol.h> 56 #include <net/ip.h> 57 #include <net/ipv6.h> 58 #include <net/route.h> 59 #include <net/sctp/sctp.h> 60 #include <net/addrconf.h> 61 #include <net/inet_common.h> 62 #include <net/inet_ecn.h> 63 64 /* Global data structures. */ 65 struct sctp_globals sctp_globals __read_mostly; 66 67 struct idr sctp_assocs_id; 68 DEFINE_SPINLOCK(sctp_assocs_id_lock); 69 70 static struct sctp_pf *sctp_pf_inet6_specific; 71 static struct sctp_pf *sctp_pf_inet_specific; 72 static struct sctp_af *sctp_af_v4_specific; 73 static struct sctp_af *sctp_af_v6_specific; 74 75 struct kmem_cache *sctp_chunk_cachep __read_mostly; 76 struct kmem_cache *sctp_bucket_cachep __read_mostly; 77 78 long sysctl_sctp_mem[3]; 79 int sysctl_sctp_rmem[3]; 80 int sysctl_sctp_wmem[3]; 81 82 /* Set up the proc fs entry for the SCTP protocol. */ 83 static int __net_init sctp_proc_init(struct net *net) 84 { 85 #ifdef CONFIG_PROC_FS 86 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net); 87 if (!net->sctp.proc_net_sctp) 88 goto out_proc_net_sctp; 89 if (sctp_snmp_proc_init(net)) 90 goto out_snmp_proc_init; 91 if (sctp_eps_proc_init(net)) 92 goto out_eps_proc_init; 93 if (sctp_assocs_proc_init(net)) 94 goto out_assocs_proc_init; 95 if (sctp_remaddr_proc_init(net)) 96 goto out_remaddr_proc_init; 97 98 return 0; 99 100 out_remaddr_proc_init: 101 sctp_assocs_proc_exit(net); 102 out_assocs_proc_init: 103 sctp_eps_proc_exit(net); 104 out_eps_proc_init: 105 sctp_snmp_proc_exit(net); 106 out_snmp_proc_init: 107 remove_proc_entry("sctp", net->proc_net); 108 net->sctp.proc_net_sctp = NULL; 109 out_proc_net_sctp: 110 return -ENOMEM; 111 #endif /* CONFIG_PROC_FS */ 112 return 0; 113 } 114 115 /* Clean up the proc fs entry for the SCTP protocol. 116 * Note: Do not make this __exit as it is used in the init error 117 * path. 118 */ 119 static void sctp_proc_exit(struct net *net) 120 { 121 #ifdef CONFIG_PROC_FS 122 sctp_snmp_proc_exit(net); 123 sctp_eps_proc_exit(net); 124 sctp_assocs_proc_exit(net); 125 sctp_remaddr_proc_exit(net); 126 127 remove_proc_entry("sctp", net->proc_net); 128 net->sctp.proc_net_sctp = NULL; 129 #endif 130 } 131 132 /* Private helper to extract ipv4 address and stash them in 133 * the protocol structure. 134 */ 135 static void sctp_v4_copy_addrlist(struct list_head *addrlist, 136 struct net_device *dev) 137 { 138 struct in_device *in_dev; 139 struct in_ifaddr *ifa; 140 struct sctp_sockaddr_entry *addr; 141 142 rcu_read_lock(); 143 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 144 rcu_read_unlock(); 145 return; 146 } 147 148 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { 149 /* Add the address to the local list. */ 150 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 151 if (addr) { 152 addr->a.v4.sin_family = AF_INET; 153 addr->a.v4.sin_port = 0; 154 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 155 addr->valid = 1; 156 INIT_LIST_HEAD(&addr->list); 157 list_add_tail(&addr->list, addrlist); 158 } 159 } 160 161 rcu_read_unlock(); 162 } 163 164 /* Extract our IP addresses from the system and stash them in the 165 * protocol structure. 166 */ 167 static void sctp_get_local_addr_list(struct net *net) 168 { 169 struct net_device *dev; 170 struct list_head *pos; 171 struct sctp_af *af; 172 173 rcu_read_lock(); 174 for_each_netdev_rcu(net, dev) { 175 list_for_each(pos, &sctp_address_families) { 176 af = list_entry(pos, struct sctp_af, list); 177 af->copy_addrlist(&net->sctp.local_addr_list, dev); 178 } 179 } 180 rcu_read_unlock(); 181 } 182 183 /* Free the existing local addresses. */ 184 static void sctp_free_local_addr_list(struct net *net) 185 { 186 struct sctp_sockaddr_entry *addr; 187 struct list_head *pos, *temp; 188 189 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 190 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 191 list_del(pos); 192 kfree(addr); 193 } 194 } 195 196 /* Copy the local addresses which are valid for 'scope' into 'bp'. */ 197 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 198 sctp_scope_t scope, gfp_t gfp, int copy_flags) 199 { 200 struct sctp_sockaddr_entry *addr; 201 int error = 0; 202 203 rcu_read_lock(); 204 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 205 if (!addr->valid) 206 continue; 207 if (sctp_in_scope(net, &addr->a, scope)) { 208 /* Now that the address is in scope, check to see if 209 * the address type is really supported by the local 210 * sock as well as the remote peer. 211 */ 212 if ((((AF_INET == addr->a.sa.sa_family) && 213 (copy_flags & SCTP_ADDR4_PEERSUPP))) || 214 (((AF_INET6 == addr->a.sa.sa_family) && 215 (copy_flags & SCTP_ADDR6_ALLOWED) && 216 (copy_flags & SCTP_ADDR6_PEERSUPP)))) { 217 error = sctp_add_bind_addr(bp, &addr->a, 218 SCTP_ADDR_SRC, GFP_ATOMIC); 219 if (error) 220 goto end_copy; 221 } 222 } 223 } 224 225 end_copy: 226 rcu_read_unlock(); 227 return error; 228 } 229 230 /* Initialize a sctp_addr from in incoming skb. */ 231 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 232 int is_saddr) 233 { 234 void *from; 235 __be16 *port; 236 struct sctphdr *sh; 237 238 port = &addr->v4.sin_port; 239 addr->v4.sin_family = AF_INET; 240 241 sh = sctp_hdr(skb); 242 if (is_saddr) { 243 *port = sh->source; 244 from = &ip_hdr(skb)->saddr; 245 } else { 246 *port = sh->dest; 247 from = &ip_hdr(skb)->daddr; 248 } 249 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr)); 250 } 251 252 /* Initialize an sctp_addr from a socket. */ 253 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 254 { 255 addr->v4.sin_family = AF_INET; 256 addr->v4.sin_port = 0; 257 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 258 } 259 260 /* Initialize sk->sk_rcv_saddr from sctp_addr. */ 261 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 262 { 263 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 264 } 265 266 /* Initialize sk->sk_daddr from sctp_addr. */ 267 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 268 { 269 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 270 } 271 272 /* Initialize a sctp_addr from an address parameter. */ 273 static void sctp_v4_from_addr_param(union sctp_addr *addr, 274 union sctp_addr_param *param, 275 __be16 port, int iif) 276 { 277 addr->v4.sin_family = AF_INET; 278 addr->v4.sin_port = port; 279 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 280 } 281 282 /* Initialize an address parameter from a sctp_addr and return the length 283 * of the address parameter. 284 */ 285 static int sctp_v4_to_addr_param(const union sctp_addr *addr, 286 union sctp_addr_param *param) 287 { 288 int length = sizeof(sctp_ipv4addr_param_t); 289 290 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 291 param->v4.param_hdr.length = htons(length); 292 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 293 294 return length; 295 } 296 297 /* Initialize a sctp_addr from a dst_entry. */ 298 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 299 __be16 port) 300 { 301 saddr->v4.sin_family = AF_INET; 302 saddr->v4.sin_port = port; 303 saddr->v4.sin_addr.s_addr = fl4->saddr; 304 } 305 306 /* Compare two addresses exactly. */ 307 static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 308 const union sctp_addr *addr2) 309 { 310 if (addr1->sa.sa_family != addr2->sa.sa_family) 311 return 0; 312 if (addr1->v4.sin_port != addr2->v4.sin_port) 313 return 0; 314 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 315 return 0; 316 317 return 1; 318 } 319 320 /* Initialize addr struct to INADDR_ANY. */ 321 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 322 { 323 addr->v4.sin_family = AF_INET; 324 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 325 addr->v4.sin_port = port; 326 } 327 328 /* Is this a wildcard address? */ 329 static int sctp_v4_is_any(const union sctp_addr *addr) 330 { 331 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 332 } 333 334 /* This function checks if the address is a valid address to be used for 335 * SCTP binding. 336 * 337 * Output: 338 * Return 0 - If the address is a non-unicast or an illegal address. 339 * Return 1 - If the address is a unicast. 340 */ 341 static int sctp_v4_addr_valid(union sctp_addr *addr, 342 struct sctp_sock *sp, 343 const struct sk_buff *skb) 344 { 345 /* IPv4 addresses not allowed */ 346 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 347 return 0; 348 349 /* Is this a non-unicast address or a unusable SCTP address? */ 350 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 351 return 0; 352 353 /* Is this a broadcast address? */ 354 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 355 return 0; 356 357 return 1; 358 } 359 360 /* Should this be available for binding? */ 361 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 362 { 363 struct net *net = sock_net(&sp->inet.sk); 364 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr); 365 366 367 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 368 ret != RTN_LOCAL && 369 !sp->inet.freebind && 370 !sysctl_ip_nonlocal_bind) 371 return 0; 372 373 if (ipv6_only_sock(sctp_opt2sk(sp))) 374 return 0; 375 376 return 1; 377 } 378 379 /* Checking the loopback, private and other address scopes as defined in 380 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 381 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 382 * 383 * Level 0 - unusable SCTP addresses 384 * Level 1 - loopback address 385 * Level 2 - link-local addresses 386 * Level 3 - private addresses. 387 * Level 4 - global addresses 388 * For INIT and INIT-ACK address list, let L be the level of 389 * of requested destination address, sender and receiver 390 * SHOULD include all of its addresses with level greater 391 * than or equal to L. 392 * 393 * IPv4 scoping can be controlled through sysctl option 394 * net.sctp.addr_scope_policy 395 */ 396 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr) 397 { 398 sctp_scope_t retval; 399 400 /* Check for unusable SCTP addresses. */ 401 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 402 retval = SCTP_SCOPE_UNUSABLE; 403 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 404 retval = SCTP_SCOPE_LOOPBACK; 405 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 406 retval = SCTP_SCOPE_LINK; 407 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 408 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 409 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) { 410 retval = SCTP_SCOPE_PRIVATE; 411 } else { 412 retval = SCTP_SCOPE_GLOBAL; 413 } 414 415 return retval; 416 } 417 418 /* Returns a valid dst cache entry for the given source and destination ip 419 * addresses. If an association is passed, trys to get a dst entry with a 420 * source address that matches an address in the bind address list. 421 */ 422 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 423 struct flowi *fl, struct sock *sk) 424 { 425 struct sctp_association *asoc = t->asoc; 426 struct rtable *rt; 427 struct flowi4 *fl4 = &fl->u.ip4; 428 struct sctp_bind_addr *bp; 429 struct sctp_sockaddr_entry *laddr; 430 struct dst_entry *dst = NULL; 431 union sctp_addr *daddr = &t->ipaddr; 432 union sctp_addr dst_saddr; 433 434 memset(fl4, 0x0, sizeof(struct flowi4)); 435 fl4->daddr = daddr->v4.sin_addr.s_addr; 436 fl4->fl4_dport = daddr->v4.sin_port; 437 fl4->flowi4_proto = IPPROTO_SCTP; 438 if (asoc) { 439 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk); 440 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 441 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 442 } 443 if (saddr) { 444 fl4->saddr = saddr->v4.sin_addr.s_addr; 445 fl4->fl4_sport = saddr->v4.sin_port; 446 } 447 448 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 449 &fl4->saddr); 450 451 rt = ip_route_output_key(sock_net(sk), fl4); 452 if (!IS_ERR(rt)) 453 dst = &rt->dst; 454 455 /* If there is no association or if a source address is passed, no 456 * more validation is required. 457 */ 458 if (!asoc || saddr) 459 goto out; 460 461 bp = &asoc->base.bind_addr; 462 463 if (dst) { 464 /* Walk through the bind address list and look for a bind 465 * address that matches the source address of the returned dst. 466 */ 467 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 468 rcu_read_lock(); 469 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 470 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 471 (laddr->state != SCTP_ADDR_SRC && 472 !asoc->src_out_of_asoc_ok)) 473 continue; 474 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 475 goto out_unlock; 476 } 477 rcu_read_unlock(); 478 479 /* None of the bound addresses match the source address of the 480 * dst. So release it. 481 */ 482 dst_release(dst); 483 dst = NULL; 484 } 485 486 /* Walk through the bind address list and try to get a dst that 487 * matches a bind address as the source address. 488 */ 489 rcu_read_lock(); 490 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 491 if (!laddr->valid) 492 continue; 493 if ((laddr->state == SCTP_ADDR_SRC) && 494 (AF_INET == laddr->a.sa.sa_family)) { 495 fl4->saddr = laddr->a.v4.sin_addr.s_addr; 496 fl4->fl4_sport = laddr->a.v4.sin_port; 497 rt = ip_route_output_key(sock_net(sk), fl4); 498 if (!IS_ERR(rt)) { 499 dst = &rt->dst; 500 goto out_unlock; 501 } 502 } 503 } 504 505 out_unlock: 506 rcu_read_unlock(); 507 out: 508 t->dst = dst; 509 if (dst) 510 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 511 &fl4->daddr, &fl4->saddr); 512 else 513 pr_debug("no route\n"); 514 } 515 516 /* For v4, the source address is cached in the route entry(dst). So no need 517 * to cache it separately and hence this is an empty routine. 518 */ 519 static void sctp_v4_get_saddr(struct sctp_sock *sk, 520 struct sctp_transport *t, 521 struct flowi *fl) 522 { 523 union sctp_addr *saddr = &t->saddr; 524 struct rtable *rt = (struct rtable *)t->dst; 525 526 if (rt) { 527 saddr->v4.sin_family = AF_INET; 528 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 529 } 530 } 531 532 /* What interface did this skb arrive on? */ 533 static int sctp_v4_skb_iif(const struct sk_buff *skb) 534 { 535 return inet_iif(skb); 536 } 537 538 /* Was this packet marked by Explicit Congestion Notification? */ 539 static int sctp_v4_is_ce(const struct sk_buff *skb) 540 { 541 return INET_ECN_is_ce(ip_hdr(skb)->tos); 542 } 543 544 /* Create and initialize a new sk for the socket returned by accept(). */ 545 static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 546 struct sctp_association *asoc) 547 { 548 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 549 sk->sk_prot); 550 struct inet_sock *newinet; 551 552 if (!newsk) 553 goto out; 554 555 sock_init_data(NULL, newsk); 556 557 sctp_copy_sock(newsk, sk, asoc); 558 sock_reset_flag(newsk, SOCK_ZAPPED); 559 560 newinet = inet_sk(newsk); 561 562 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 563 564 sk_refcnt_debug_inc(newsk); 565 566 if (newsk->sk_prot->init(newsk)) { 567 sk_common_release(newsk); 568 newsk = NULL; 569 } 570 571 out: 572 return newsk; 573 } 574 575 /* Map address, empty for v4 family */ 576 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr) 577 { 578 /* Empty */ 579 } 580 581 /* Dump the v4 addr to the seq file. */ 582 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 583 { 584 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 585 } 586 587 static void sctp_v4_ecn_capable(struct sock *sk) 588 { 589 INET_ECN_xmit(sk); 590 } 591 592 static void sctp_addr_wq_timeout_handler(unsigned long arg) 593 { 594 struct net *net = (struct net *)arg; 595 struct sctp_sockaddr_entry *addrw, *temp; 596 struct sctp_sock *sp; 597 598 spin_lock_bh(&net->sctp.addr_wq_lock); 599 600 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 601 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 602 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 603 addrw->state, addrw); 604 605 #if IS_ENABLED(CONFIG_IPV6) 606 /* Now we send an ASCONF for each association */ 607 /* Note. we currently don't handle link local IPv6 addressees */ 608 if (addrw->a.sa.sa_family == AF_INET6) { 609 struct in6_addr *in6; 610 611 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 612 IPV6_ADDR_LINKLOCAL) 613 goto free_next; 614 615 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 616 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 617 addrw->state == SCTP_ADDR_NEW) { 618 unsigned long timeo_val; 619 620 pr_debug("%s: this is on DAD, trying %d sec " 621 "later\n", __func__, 622 SCTP_ADDRESS_TICK_DELAY); 623 624 timeo_val = jiffies; 625 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 626 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 627 break; 628 } 629 } 630 #endif 631 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 632 struct sock *sk; 633 634 sk = sctp_opt2sk(sp); 635 /* ignore bound-specific endpoints */ 636 if (!sctp_is_ep_boundall(sk)) 637 continue; 638 sctp_bh_lock_sock(sk); 639 if (sctp_asconf_mgmt(sp, addrw) < 0) 640 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 641 sctp_bh_unlock_sock(sk); 642 } 643 #if IS_ENABLED(CONFIG_IPV6) 644 free_next: 645 #endif 646 list_del(&addrw->list); 647 kfree(addrw); 648 } 649 spin_unlock_bh(&net->sctp.addr_wq_lock); 650 } 651 652 static void sctp_free_addr_wq(struct net *net) 653 { 654 struct sctp_sockaddr_entry *addrw; 655 struct sctp_sockaddr_entry *temp; 656 657 spin_lock_bh(&net->sctp.addr_wq_lock); 658 del_timer(&net->sctp.addr_wq_timer); 659 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 660 list_del(&addrw->list); 661 kfree(addrw); 662 } 663 spin_unlock_bh(&net->sctp.addr_wq_lock); 664 } 665 666 /* lookup the entry for the same address in the addr_waitq 667 * sctp_addr_wq MUST be locked 668 */ 669 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 670 struct sctp_sockaddr_entry *addr) 671 { 672 struct sctp_sockaddr_entry *addrw; 673 674 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 675 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 676 continue; 677 if (addrw->a.sa.sa_family == AF_INET) { 678 if (addrw->a.v4.sin_addr.s_addr == 679 addr->a.v4.sin_addr.s_addr) 680 return addrw; 681 } else if (addrw->a.sa.sa_family == AF_INET6) { 682 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 683 &addr->a.v6.sin6_addr)) 684 return addrw; 685 } 686 } 687 return NULL; 688 } 689 690 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 691 { 692 struct sctp_sockaddr_entry *addrw; 693 unsigned long timeo_val; 694 695 /* first, we check if an opposite message already exist in the queue. 696 * If we found such message, it is removed. 697 * This operation is a bit stupid, but the DHCP client attaches the 698 * new address after a couple of addition and deletion of that address 699 */ 700 701 spin_lock_bh(&net->sctp.addr_wq_lock); 702 /* Offsets existing events in addr_wq */ 703 addrw = sctp_addr_wq_lookup(net, addr); 704 if (addrw) { 705 if (addrw->state != cmd) { 706 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 707 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 708 &net->sctp.addr_waitq); 709 710 list_del(&addrw->list); 711 kfree(addrw); 712 } 713 spin_unlock_bh(&net->sctp.addr_wq_lock); 714 return; 715 } 716 717 /* OK, we have to add the new address to the wait queue */ 718 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 719 if (addrw == NULL) { 720 spin_unlock_bh(&net->sctp.addr_wq_lock); 721 return; 722 } 723 addrw->state = cmd; 724 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 725 726 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 727 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 728 729 if (!timer_pending(&net->sctp.addr_wq_timer)) { 730 timeo_val = jiffies; 731 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 732 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 733 } 734 spin_unlock_bh(&net->sctp.addr_wq_lock); 735 } 736 737 /* Event handler for inet address addition/deletion events. 738 * The sctp_local_addr_list needs to be protocted by a spin lock since 739 * multiple notifiers (say IPv4 and IPv6) may be running at the same 740 * time and thus corrupt the list. 741 * The reader side is protected with RCU. 742 */ 743 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 744 void *ptr) 745 { 746 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 747 struct sctp_sockaddr_entry *addr = NULL; 748 struct sctp_sockaddr_entry *temp; 749 struct net *net = dev_net(ifa->ifa_dev->dev); 750 int found = 0; 751 752 switch (ev) { 753 case NETDEV_UP: 754 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 755 if (addr) { 756 addr->a.v4.sin_family = AF_INET; 757 addr->a.v4.sin_port = 0; 758 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 759 addr->valid = 1; 760 spin_lock_bh(&net->sctp.local_addr_lock); 761 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 762 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 763 spin_unlock_bh(&net->sctp.local_addr_lock); 764 } 765 break; 766 case NETDEV_DOWN: 767 spin_lock_bh(&net->sctp.local_addr_lock); 768 list_for_each_entry_safe(addr, temp, 769 &net->sctp.local_addr_list, list) { 770 if (addr->a.sa.sa_family == AF_INET && 771 addr->a.v4.sin_addr.s_addr == 772 ifa->ifa_local) { 773 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 774 found = 1; 775 addr->valid = 0; 776 list_del_rcu(&addr->list); 777 break; 778 } 779 } 780 spin_unlock_bh(&net->sctp.local_addr_lock); 781 if (found) 782 kfree_rcu(addr, rcu); 783 break; 784 } 785 786 return NOTIFY_DONE; 787 } 788 789 /* 790 * Initialize the control inode/socket with a control endpoint data 791 * structure. This endpoint is reserved exclusively for the OOTB processing. 792 */ 793 static int sctp_ctl_sock_init(struct net *net) 794 { 795 int err; 796 sa_family_t family = PF_INET; 797 798 if (sctp_get_pf_specific(PF_INET6)) 799 family = PF_INET6; 800 801 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 802 SOCK_SEQPACKET, IPPROTO_SCTP, net); 803 804 /* If IPv6 socket could not be created, try the IPv4 socket */ 805 if (err < 0 && family == PF_INET6) 806 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 807 SOCK_SEQPACKET, IPPROTO_SCTP, 808 net); 809 810 if (err < 0) { 811 pr_err("Failed to create the SCTP control socket\n"); 812 return err; 813 } 814 return 0; 815 } 816 817 /* Register address family specific functions. */ 818 int sctp_register_af(struct sctp_af *af) 819 { 820 switch (af->sa_family) { 821 case AF_INET: 822 if (sctp_af_v4_specific) 823 return 0; 824 sctp_af_v4_specific = af; 825 break; 826 case AF_INET6: 827 if (sctp_af_v6_specific) 828 return 0; 829 sctp_af_v6_specific = af; 830 break; 831 default: 832 return 0; 833 } 834 835 INIT_LIST_HEAD(&af->list); 836 list_add_tail(&af->list, &sctp_address_families); 837 return 1; 838 } 839 840 /* Get the table of functions for manipulating a particular address 841 * family. 842 */ 843 struct sctp_af *sctp_get_af_specific(sa_family_t family) 844 { 845 switch (family) { 846 case AF_INET: 847 return sctp_af_v4_specific; 848 case AF_INET6: 849 return sctp_af_v6_specific; 850 default: 851 return NULL; 852 } 853 } 854 855 /* Common code to initialize a AF_INET msg_name. */ 856 static void sctp_inet_msgname(char *msgname, int *addr_len) 857 { 858 struct sockaddr_in *sin; 859 860 sin = (struct sockaddr_in *)msgname; 861 *addr_len = sizeof(struct sockaddr_in); 862 sin->sin_family = AF_INET; 863 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 864 } 865 866 /* Copy the primary address of the peer primary address as the msg_name. */ 867 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 868 int *addr_len) 869 { 870 struct sockaddr_in *sin, *sinfrom; 871 872 if (msgname) { 873 struct sctp_association *asoc; 874 875 asoc = event->asoc; 876 sctp_inet_msgname(msgname, addr_len); 877 sin = (struct sockaddr_in *)msgname; 878 sinfrom = &asoc->peer.primary_addr.v4; 879 sin->sin_port = htons(asoc->peer.port); 880 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 881 } 882 } 883 884 /* Initialize and copy out a msgname from an inbound skb. */ 885 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 886 { 887 if (msgname) { 888 struct sctphdr *sh = sctp_hdr(skb); 889 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 890 891 sctp_inet_msgname(msgname, len); 892 sin->sin_port = sh->source; 893 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 894 } 895 } 896 897 /* Do we support this AF? */ 898 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 899 { 900 /* PF_INET only supports AF_INET addresses. */ 901 return AF_INET == family; 902 } 903 904 /* Address matching with wildcards allowed. */ 905 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 906 const union sctp_addr *addr2, 907 struct sctp_sock *opt) 908 { 909 /* PF_INET only supports AF_INET addresses. */ 910 if (addr1->sa.sa_family != addr2->sa.sa_family) 911 return 0; 912 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 913 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 914 return 1; 915 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 916 return 1; 917 918 return 0; 919 } 920 921 /* Verify that provided sockaddr looks bindable. Common verification has 922 * already been taken care of. 923 */ 924 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 925 { 926 return sctp_v4_available(addr, opt); 927 } 928 929 /* Verify that sockaddr looks sendable. Common verification has already 930 * been taken care of. 931 */ 932 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 933 { 934 return 1; 935 } 936 937 /* Fill in Supported Address Type information for INIT and INIT-ACK 938 * chunks. Returns number of addresses supported. 939 */ 940 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 941 __be16 *types) 942 { 943 types[0] = SCTP_PARAM_IPV4_ADDRESS; 944 return 1; 945 } 946 947 /* Wrapper routine that calls the ip transmit routine. */ 948 static inline int sctp_v4_xmit(struct sk_buff *skb, 949 struct sctp_transport *transport) 950 { 951 struct inet_sock *inet = inet_sk(skb->sk); 952 953 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 954 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr); 955 956 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ? 957 IP_PMTUDISC_DO : IP_PMTUDISC_DONT; 958 959 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS); 960 961 return ip_queue_xmit(skb, &transport->fl); 962 } 963 964 static struct sctp_af sctp_af_inet; 965 966 static struct sctp_pf sctp_pf_inet = { 967 .event_msgname = sctp_inet_event_msgname, 968 .skb_msgname = sctp_inet_skb_msgname, 969 .af_supported = sctp_inet_af_supported, 970 .cmp_addr = sctp_inet_cmp_addr, 971 .bind_verify = sctp_inet_bind_verify, 972 .send_verify = sctp_inet_send_verify, 973 .supported_addrs = sctp_inet_supported_addrs, 974 .create_accept_sk = sctp_v4_create_accept_sk, 975 .addr_v4map = sctp_v4_addr_v4map, 976 .af = &sctp_af_inet 977 }; 978 979 /* Notifier for inetaddr addition/deletion events. */ 980 static struct notifier_block sctp_inetaddr_notifier = { 981 .notifier_call = sctp_inetaddr_event, 982 }; 983 984 /* Socket operations. */ 985 static const struct proto_ops inet_seqpacket_ops = { 986 .family = PF_INET, 987 .owner = THIS_MODULE, 988 .release = inet_release, /* Needs to be wrapped... */ 989 .bind = inet_bind, 990 .connect = inet_dgram_connect, 991 .socketpair = sock_no_socketpair, 992 .accept = inet_accept, 993 .getname = inet_getname, /* Semantics are different. */ 994 .poll = sctp_poll, 995 .ioctl = inet_ioctl, 996 .listen = sctp_inet_listen, 997 .shutdown = inet_shutdown, /* Looks harmless. */ 998 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 999 .getsockopt = sock_common_getsockopt, 1000 .sendmsg = inet_sendmsg, 1001 .recvmsg = sock_common_recvmsg, 1002 .mmap = sock_no_mmap, 1003 .sendpage = sock_no_sendpage, 1004 #ifdef CONFIG_COMPAT 1005 .compat_setsockopt = compat_sock_common_setsockopt, 1006 .compat_getsockopt = compat_sock_common_getsockopt, 1007 #endif 1008 }; 1009 1010 /* Registration with AF_INET family. */ 1011 static struct inet_protosw sctp_seqpacket_protosw = { 1012 .type = SOCK_SEQPACKET, 1013 .protocol = IPPROTO_SCTP, 1014 .prot = &sctp_prot, 1015 .ops = &inet_seqpacket_ops, 1016 .no_check = 0, 1017 .flags = SCTP_PROTOSW_FLAG 1018 }; 1019 static struct inet_protosw sctp_stream_protosw = { 1020 .type = SOCK_STREAM, 1021 .protocol = IPPROTO_SCTP, 1022 .prot = &sctp_prot, 1023 .ops = &inet_seqpacket_ops, 1024 .no_check = 0, 1025 .flags = SCTP_PROTOSW_FLAG 1026 }; 1027 1028 /* Register with IP layer. */ 1029 static const struct net_protocol sctp_protocol = { 1030 .handler = sctp_rcv, 1031 .err_handler = sctp_v4_err, 1032 .no_policy = 1, 1033 .netns_ok = 1, 1034 }; 1035 1036 /* IPv4 address related functions. */ 1037 static struct sctp_af sctp_af_inet = { 1038 .sa_family = AF_INET, 1039 .sctp_xmit = sctp_v4_xmit, 1040 .setsockopt = ip_setsockopt, 1041 .getsockopt = ip_getsockopt, 1042 .get_dst = sctp_v4_get_dst, 1043 .get_saddr = sctp_v4_get_saddr, 1044 .copy_addrlist = sctp_v4_copy_addrlist, 1045 .from_skb = sctp_v4_from_skb, 1046 .from_sk = sctp_v4_from_sk, 1047 .to_sk_saddr = sctp_v4_to_sk_saddr, 1048 .to_sk_daddr = sctp_v4_to_sk_daddr, 1049 .from_addr_param = sctp_v4_from_addr_param, 1050 .to_addr_param = sctp_v4_to_addr_param, 1051 .cmp_addr = sctp_v4_cmp_addr, 1052 .addr_valid = sctp_v4_addr_valid, 1053 .inaddr_any = sctp_v4_inaddr_any, 1054 .is_any = sctp_v4_is_any, 1055 .available = sctp_v4_available, 1056 .scope = sctp_v4_scope, 1057 .skb_iif = sctp_v4_skb_iif, 1058 .is_ce = sctp_v4_is_ce, 1059 .seq_dump_addr = sctp_v4_seq_dump_addr, 1060 .ecn_capable = sctp_v4_ecn_capable, 1061 .net_header_len = sizeof(struct iphdr), 1062 .sockaddr_len = sizeof(struct sockaddr_in), 1063 #ifdef CONFIG_COMPAT 1064 .compat_setsockopt = compat_ip_setsockopt, 1065 .compat_getsockopt = compat_ip_getsockopt, 1066 #endif 1067 }; 1068 1069 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) { 1070 1071 switch (family) { 1072 case PF_INET: 1073 return sctp_pf_inet_specific; 1074 case PF_INET6: 1075 return sctp_pf_inet6_specific; 1076 default: 1077 return NULL; 1078 } 1079 } 1080 1081 /* Register the PF specific function table. */ 1082 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1083 { 1084 switch (family) { 1085 case PF_INET: 1086 if (sctp_pf_inet_specific) 1087 return 0; 1088 sctp_pf_inet_specific = pf; 1089 break; 1090 case PF_INET6: 1091 if (sctp_pf_inet6_specific) 1092 return 0; 1093 sctp_pf_inet6_specific = pf; 1094 break; 1095 default: 1096 return 0; 1097 } 1098 return 1; 1099 } 1100 1101 static inline int init_sctp_mibs(struct net *net) 1102 { 1103 return snmp_mib_init((void __percpu **)net->sctp.sctp_statistics, 1104 sizeof(struct sctp_mib), 1105 __alignof__(struct sctp_mib)); 1106 } 1107 1108 static inline void cleanup_sctp_mibs(struct net *net) 1109 { 1110 snmp_mib_free((void __percpu **)net->sctp.sctp_statistics); 1111 } 1112 1113 static void sctp_v4_pf_init(void) 1114 { 1115 /* Initialize the SCTP specific PF functions. */ 1116 sctp_register_pf(&sctp_pf_inet, PF_INET); 1117 sctp_register_af(&sctp_af_inet); 1118 } 1119 1120 static void sctp_v4_pf_exit(void) 1121 { 1122 list_del(&sctp_af_inet.list); 1123 } 1124 1125 static int sctp_v4_protosw_init(void) 1126 { 1127 int rc; 1128 1129 rc = proto_register(&sctp_prot, 1); 1130 if (rc) 1131 return rc; 1132 1133 /* Register SCTP(UDP and TCP style) with socket layer. */ 1134 inet_register_protosw(&sctp_seqpacket_protosw); 1135 inet_register_protosw(&sctp_stream_protosw); 1136 1137 return 0; 1138 } 1139 1140 static void sctp_v4_protosw_exit(void) 1141 { 1142 inet_unregister_protosw(&sctp_stream_protosw); 1143 inet_unregister_protosw(&sctp_seqpacket_protosw); 1144 proto_unregister(&sctp_prot); 1145 } 1146 1147 static int sctp_v4_add_protocol(void) 1148 { 1149 /* Register notifier for inet address additions/deletions. */ 1150 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1151 1152 /* Register SCTP with inet layer. */ 1153 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1154 return -EAGAIN; 1155 1156 return 0; 1157 } 1158 1159 static void sctp_v4_del_protocol(void) 1160 { 1161 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1162 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1163 } 1164 1165 static int __net_init sctp_net_init(struct net *net) 1166 { 1167 int status; 1168 1169 /* 1170 * 14. Suggested SCTP Protocol Parameter Values 1171 */ 1172 /* The following protocol parameters are RECOMMENDED: */ 1173 /* RTO.Initial - 3 seconds */ 1174 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1175 /* RTO.Min - 1 second */ 1176 net->sctp.rto_min = SCTP_RTO_MIN; 1177 /* RTO.Max - 60 seconds */ 1178 net->sctp.rto_max = SCTP_RTO_MAX; 1179 /* RTO.Alpha - 1/8 */ 1180 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1181 /* RTO.Beta - 1/4 */ 1182 net->sctp.rto_beta = SCTP_RTO_BETA; 1183 1184 /* Valid.Cookie.Life - 60 seconds */ 1185 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1186 1187 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1188 net->sctp.cookie_preserve_enable = 1; 1189 1190 /* Default sctp sockets to use md5 as their hmac alg */ 1191 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1192 net->sctp.sctp_hmac_alg = "md5"; 1193 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1194 net->sctp.sctp_hmac_alg = "sha1"; 1195 #else 1196 net->sctp.sctp_hmac_alg = NULL; 1197 #endif 1198 1199 /* Max.Burst - 4 */ 1200 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1201 1202 /* Association.Max.Retrans - 10 attempts 1203 * Path.Max.Retrans - 5 attempts (per destination address) 1204 * Max.Init.Retransmits - 8 attempts 1205 */ 1206 net->sctp.max_retrans_association = 10; 1207 net->sctp.max_retrans_path = 5; 1208 net->sctp.max_retrans_init = 8; 1209 1210 /* Sendbuffer growth - do per-socket accounting */ 1211 net->sctp.sndbuf_policy = 0; 1212 1213 /* Rcvbuffer growth - do per-socket accounting */ 1214 net->sctp.rcvbuf_policy = 0; 1215 1216 /* HB.interval - 30 seconds */ 1217 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1218 1219 /* delayed SACK timeout */ 1220 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1221 1222 /* Disable ADDIP by default. */ 1223 net->sctp.addip_enable = 0; 1224 net->sctp.addip_noauth = 0; 1225 net->sctp.default_auto_asconf = 0; 1226 1227 /* Enable PR-SCTP by default. */ 1228 net->sctp.prsctp_enable = 1; 1229 1230 /* Disable AUTH by default. */ 1231 net->sctp.auth_enable = 0; 1232 1233 /* Set SCOPE policy to enabled */ 1234 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1235 1236 /* Set the default rwnd update threshold */ 1237 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1238 1239 /* Initialize maximum autoclose timeout. */ 1240 net->sctp.max_autoclose = INT_MAX / HZ; 1241 1242 status = sctp_sysctl_net_register(net); 1243 if (status) 1244 goto err_sysctl_register; 1245 1246 /* Allocate and initialise sctp mibs. */ 1247 status = init_sctp_mibs(net); 1248 if (status) 1249 goto err_init_mibs; 1250 1251 /* Initialize proc fs directory. */ 1252 status = sctp_proc_init(net); 1253 if (status) 1254 goto err_init_proc; 1255 1256 sctp_dbg_objcnt_init(net); 1257 1258 /* Initialize the control inode/socket for handling OOTB packets. */ 1259 if ((status = sctp_ctl_sock_init(net))) { 1260 pr_err("Failed to initialize the SCTP control sock\n"); 1261 goto err_ctl_sock_init; 1262 } 1263 1264 /* Initialize the local address list. */ 1265 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1266 spin_lock_init(&net->sctp.local_addr_lock); 1267 sctp_get_local_addr_list(net); 1268 1269 /* Initialize the address event list */ 1270 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1271 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1272 spin_lock_init(&net->sctp.addr_wq_lock); 1273 net->sctp.addr_wq_timer.expires = 0; 1274 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 1275 (unsigned long)net); 1276 1277 return 0; 1278 1279 err_ctl_sock_init: 1280 sctp_dbg_objcnt_exit(net); 1281 sctp_proc_exit(net); 1282 err_init_proc: 1283 cleanup_sctp_mibs(net); 1284 err_init_mibs: 1285 sctp_sysctl_net_unregister(net); 1286 err_sysctl_register: 1287 return status; 1288 } 1289 1290 static void __net_exit sctp_net_exit(struct net *net) 1291 { 1292 /* Free the local address list */ 1293 sctp_free_addr_wq(net); 1294 sctp_free_local_addr_list(net); 1295 1296 /* Free the control endpoint. */ 1297 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1298 1299 sctp_dbg_objcnt_exit(net); 1300 1301 sctp_proc_exit(net); 1302 cleanup_sctp_mibs(net); 1303 sctp_sysctl_net_unregister(net); 1304 } 1305 1306 static struct pernet_operations sctp_net_ops = { 1307 .init = sctp_net_init, 1308 .exit = sctp_net_exit, 1309 }; 1310 1311 /* Initialize the universe into something sensible. */ 1312 static __init int sctp_init(void) 1313 { 1314 int i; 1315 int status = -EINVAL; 1316 unsigned long goal; 1317 unsigned long limit; 1318 int max_share; 1319 int order; 1320 1321 BUILD_BUG_ON(sizeof(struct sctp_ulpevent) > 1322 sizeof(((struct sk_buff *) 0)->cb)); 1323 1324 /* Allocate bind_bucket and chunk caches. */ 1325 status = -ENOBUFS; 1326 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket", 1327 sizeof(struct sctp_bind_bucket), 1328 0, SLAB_HWCACHE_ALIGN, 1329 NULL); 1330 if (!sctp_bucket_cachep) 1331 goto out; 1332 1333 sctp_chunk_cachep = kmem_cache_create("sctp_chunk", 1334 sizeof(struct sctp_chunk), 1335 0, SLAB_HWCACHE_ALIGN, 1336 NULL); 1337 if (!sctp_chunk_cachep) 1338 goto err_chunk_cachep; 1339 1340 status = percpu_counter_init(&sctp_sockets_allocated, 0); 1341 if (status) 1342 goto err_percpu_counter_init; 1343 1344 /* Implementation specific variables. */ 1345 1346 /* Initialize default stream count setup information. */ 1347 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1348 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1349 1350 /* Initialize handle used for association ids. */ 1351 idr_init(&sctp_assocs_id); 1352 1353 limit = nr_free_buffer_pages() / 8; 1354 limit = max(limit, 128UL); 1355 sysctl_sctp_mem[0] = limit / 4 * 3; 1356 sysctl_sctp_mem[1] = limit; 1357 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1358 1359 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1360 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1361 max_share = min(4UL*1024*1024, limit); 1362 1363 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */ 1364 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1365 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1366 1367 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM; 1368 sysctl_sctp_wmem[1] = 16*1024; 1369 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1370 1371 /* Size and allocate the association hash table. 1372 * The methodology is similar to that of the tcp hash tables. 1373 */ 1374 if (totalram_pages >= (128 * 1024)) 1375 goal = totalram_pages >> (22 - PAGE_SHIFT); 1376 else 1377 goal = totalram_pages >> (24 - PAGE_SHIFT); 1378 1379 for (order = 0; (1UL << order) < goal; order++) 1380 ; 1381 1382 do { 1383 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE / 1384 sizeof(struct sctp_hashbucket); 1385 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0) 1386 continue; 1387 sctp_assoc_hashtable = (struct sctp_hashbucket *) 1388 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order); 1389 } while (!sctp_assoc_hashtable && --order > 0); 1390 if (!sctp_assoc_hashtable) { 1391 pr_err("Failed association hash alloc\n"); 1392 status = -ENOMEM; 1393 goto err_ahash_alloc; 1394 } 1395 for (i = 0; i < sctp_assoc_hashsize; i++) { 1396 rwlock_init(&sctp_assoc_hashtable[i].lock); 1397 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain); 1398 } 1399 1400 /* Allocate and initialize the endpoint hash table. */ 1401 sctp_ep_hashsize = 64; 1402 sctp_ep_hashtable = 1403 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL); 1404 if (!sctp_ep_hashtable) { 1405 pr_err("Failed endpoint_hash alloc\n"); 1406 status = -ENOMEM; 1407 goto err_ehash_alloc; 1408 } 1409 for (i = 0; i < sctp_ep_hashsize; i++) { 1410 rwlock_init(&sctp_ep_hashtable[i].lock); 1411 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1412 } 1413 1414 /* Allocate and initialize the SCTP port hash table. */ 1415 do { 1416 sctp_port_hashsize = (1UL << order) * PAGE_SIZE / 1417 sizeof(struct sctp_bind_hashbucket); 1418 if ((sctp_port_hashsize > (64 * 1024)) && order > 0) 1419 continue; 1420 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1421 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order); 1422 } while (!sctp_port_hashtable && --order > 0); 1423 if (!sctp_port_hashtable) { 1424 pr_err("Failed bind hash alloc\n"); 1425 status = -ENOMEM; 1426 goto err_bhash_alloc; 1427 } 1428 for (i = 0; i < sctp_port_hashsize; i++) { 1429 spin_lock_init(&sctp_port_hashtable[i].lock); 1430 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1431 } 1432 1433 pr_info("Hash tables configured (established %d bind %d)\n", 1434 sctp_assoc_hashsize, sctp_port_hashsize); 1435 1436 sctp_sysctl_register(); 1437 1438 INIT_LIST_HEAD(&sctp_address_families); 1439 sctp_v4_pf_init(); 1440 sctp_v6_pf_init(); 1441 1442 status = sctp_v4_protosw_init(); 1443 1444 if (status) 1445 goto err_protosw_init; 1446 1447 status = sctp_v6_protosw_init(); 1448 if (status) 1449 goto err_v6_protosw_init; 1450 1451 status = register_pernet_subsys(&sctp_net_ops); 1452 if (status) 1453 goto err_register_pernet_subsys; 1454 1455 status = sctp_v4_add_protocol(); 1456 if (status) 1457 goto err_add_protocol; 1458 1459 /* Register SCTP with inet6 layer. */ 1460 status = sctp_v6_add_protocol(); 1461 if (status) 1462 goto err_v6_add_protocol; 1463 1464 status = 0; 1465 out: 1466 return status; 1467 err_v6_add_protocol: 1468 sctp_v4_del_protocol(); 1469 err_add_protocol: 1470 unregister_pernet_subsys(&sctp_net_ops); 1471 err_register_pernet_subsys: 1472 sctp_v6_protosw_exit(); 1473 err_v6_protosw_init: 1474 sctp_v4_protosw_exit(); 1475 err_protosw_init: 1476 sctp_v4_pf_exit(); 1477 sctp_v6_pf_exit(); 1478 sctp_sysctl_unregister(); 1479 free_pages((unsigned long)sctp_port_hashtable, 1480 get_order(sctp_port_hashsize * 1481 sizeof(struct sctp_bind_hashbucket))); 1482 err_bhash_alloc: 1483 kfree(sctp_ep_hashtable); 1484 err_ehash_alloc: 1485 free_pages((unsigned long)sctp_assoc_hashtable, 1486 get_order(sctp_assoc_hashsize * 1487 sizeof(struct sctp_hashbucket))); 1488 err_ahash_alloc: 1489 percpu_counter_destroy(&sctp_sockets_allocated); 1490 err_percpu_counter_init: 1491 kmem_cache_destroy(sctp_chunk_cachep); 1492 err_chunk_cachep: 1493 kmem_cache_destroy(sctp_bucket_cachep); 1494 goto out; 1495 } 1496 1497 /* Exit handler for the SCTP protocol. */ 1498 static __exit void sctp_exit(void) 1499 { 1500 /* BUG. This should probably do something useful like clean 1501 * up all the remaining associations and all that memory. 1502 */ 1503 1504 /* Unregister with inet6/inet layers. */ 1505 sctp_v6_del_protocol(); 1506 sctp_v4_del_protocol(); 1507 1508 unregister_pernet_subsys(&sctp_net_ops); 1509 1510 /* Free protosw registrations */ 1511 sctp_v6_protosw_exit(); 1512 sctp_v4_protosw_exit(); 1513 1514 /* Unregister with socket layer. */ 1515 sctp_v6_pf_exit(); 1516 sctp_v4_pf_exit(); 1517 1518 sctp_sysctl_unregister(); 1519 1520 free_pages((unsigned long)sctp_assoc_hashtable, 1521 get_order(sctp_assoc_hashsize * 1522 sizeof(struct sctp_hashbucket))); 1523 kfree(sctp_ep_hashtable); 1524 free_pages((unsigned long)sctp_port_hashtable, 1525 get_order(sctp_port_hashsize * 1526 sizeof(struct sctp_bind_hashbucket))); 1527 1528 percpu_counter_destroy(&sctp_sockets_allocated); 1529 1530 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1531 1532 kmem_cache_destroy(sctp_chunk_cachep); 1533 kmem_cache_destroy(sctp_bucket_cachep); 1534 } 1535 1536 module_init(sctp_init); 1537 module_exit(sctp_exit); 1538 1539 /* 1540 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1541 */ 1542 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1543 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1544 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1545 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1546 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1547 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1548 MODULE_LICENSE("GPL"); 1549