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, see 27 * <http://www.gnu.org/licenses/>. 28 * 29 * Please send any bug reports or fixes you make to the 30 * email address(es): 31 * lksctp developers <linux-sctp@vger.kernel.org> 32 * 33 * Written or modified by: 34 * La Monte H.P. Yarroll <piggy@acm.org> 35 * Karl Knutson <karl@athena.chicago.il.us> 36 * Jon Grimm <jgrimm@us.ibm.com> 37 * Sridhar Samudrala <sri@us.ibm.com> 38 * Daisy Chang <daisyc@us.ibm.com> 39 * Ardelle Fan <ardelle.fan@intel.com> 40 */ 41 42#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 43 44#include <linux/module.h> 45#include <linux/init.h> 46#include <linux/netdevice.h> 47#include <linux/inetdevice.h> 48#include <linux/seq_file.h> 49#include <linux/bootmem.h> 50#include <linux/highmem.h> 51#include <linux/swap.h> 52#include <linux/slab.h> 53#include <net/net_namespace.h> 54#include <net/protocol.h> 55#include <net/ip.h> 56#include <net/ipv6.h> 57#include <net/route.h> 58#include <net/sctp/sctp.h> 59#include <net/addrconf.h> 60#include <net/inet_common.h> 61#include <net/inet_ecn.h> 62 63#define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024) 64 65/* Global data structures. */ 66struct sctp_globals sctp_globals __read_mostly; 67 68struct idr sctp_assocs_id; 69DEFINE_SPINLOCK(sctp_assocs_id_lock); 70 71static struct sctp_pf *sctp_pf_inet6_specific; 72static struct sctp_pf *sctp_pf_inet_specific; 73static struct sctp_af *sctp_af_v4_specific; 74static struct sctp_af *sctp_af_v6_specific; 75 76struct kmem_cache *sctp_chunk_cachep __read_mostly; 77struct kmem_cache *sctp_bucket_cachep __read_mostly; 78 79long sysctl_sctp_mem[3]; 80int sysctl_sctp_rmem[3]; 81int sysctl_sctp_wmem[3]; 82 83/* Set up the proc fs entry for the SCTP protocol. */ 84static int __net_init sctp_proc_init(struct net *net) 85{ 86#ifdef CONFIG_PROC_FS 87 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net); 88 if (!net->sctp.proc_net_sctp) 89 goto out_proc_net_sctp; 90 if (sctp_snmp_proc_init(net)) 91 goto out_snmp_proc_init; 92 if (sctp_eps_proc_init(net)) 93 goto out_eps_proc_init; 94 if (sctp_assocs_proc_init(net)) 95 goto out_assocs_proc_init; 96 if (sctp_remaddr_proc_init(net)) 97 goto out_remaddr_proc_init; 98 99 return 0; 100 101out_remaddr_proc_init: 102 sctp_assocs_proc_exit(net); 103out_assocs_proc_init: 104 sctp_eps_proc_exit(net); 105out_eps_proc_init: 106 sctp_snmp_proc_exit(net); 107out_snmp_proc_init: 108 remove_proc_entry("sctp", net->proc_net); 109 net->sctp.proc_net_sctp = NULL; 110out_proc_net_sctp: 111 return -ENOMEM; 112#endif /* CONFIG_PROC_FS */ 113 return 0; 114} 115 116/* Clean up the proc fs entry for the SCTP protocol. 117 * Note: Do not make this __exit as it is used in the init error 118 * path. 119 */ 120static void sctp_proc_exit(struct net *net) 121{ 122#ifdef CONFIG_PROC_FS 123 sctp_snmp_proc_exit(net); 124 sctp_eps_proc_exit(net); 125 sctp_assocs_proc_exit(net); 126 sctp_remaddr_proc_exit(net); 127 128 remove_proc_entry("sctp", net->proc_net); 129 net->sctp.proc_net_sctp = NULL; 130#endif 131} 132 133/* Private helper to extract ipv4 address and stash them in 134 * the protocol structure. 135 */ 136static void sctp_v4_copy_addrlist(struct list_head *addrlist, 137 struct net_device *dev) 138{ 139 struct in_device *in_dev; 140 struct in_ifaddr *ifa; 141 struct sctp_sockaddr_entry *addr; 142 143 rcu_read_lock(); 144 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 145 rcu_read_unlock(); 146 return; 147 } 148 149 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { 150 /* Add the address to the local list. */ 151 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 152 if (addr) { 153 addr->a.v4.sin_family = AF_INET; 154 addr->a.v4.sin_port = 0; 155 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 156 addr->valid = 1; 157 INIT_LIST_HEAD(&addr->list); 158 list_add_tail(&addr->list, addrlist); 159 } 160 } 161 162 rcu_read_unlock(); 163} 164 165/* Extract our IP addresses from the system and stash them in the 166 * protocol structure. 167 */ 168static void sctp_get_local_addr_list(struct net *net) 169{ 170 struct net_device *dev; 171 struct list_head *pos; 172 struct sctp_af *af; 173 174 rcu_read_lock(); 175 for_each_netdev_rcu(net, dev) { 176 list_for_each(pos, &sctp_address_families) { 177 af = list_entry(pos, struct sctp_af, list); 178 af->copy_addrlist(&net->sctp.local_addr_list, dev); 179 } 180 } 181 rcu_read_unlock(); 182} 183 184/* Free the existing local addresses. */ 185static void sctp_free_local_addr_list(struct net *net) 186{ 187 struct sctp_sockaddr_entry *addr; 188 struct list_head *pos, *temp; 189 190 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 191 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 192 list_del(pos); 193 kfree(addr); 194 } 195} 196 197/* Copy the local addresses which are valid for 'scope' into 'bp'. */ 198int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 199 sctp_scope_t scope, gfp_t gfp, int copy_flags) 200{ 201 struct sctp_sockaddr_entry *addr; 202 int error = 0; 203 204 rcu_read_lock(); 205 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 206 if (!addr->valid) 207 continue; 208 if (sctp_in_scope(net, &addr->a, scope)) { 209 /* Now that the address is in scope, check to see if 210 * the address type is really supported by the local 211 * sock as well as the remote peer. 212 */ 213 if ((((AF_INET == addr->a.sa.sa_family) && 214 (copy_flags & SCTP_ADDR4_PEERSUPP))) || 215 (((AF_INET6 == addr->a.sa.sa_family) && 216 (copy_flags & SCTP_ADDR6_ALLOWED) && 217 (copy_flags & SCTP_ADDR6_PEERSUPP)))) { 218 error = sctp_add_bind_addr(bp, &addr->a, 219 SCTP_ADDR_SRC, GFP_ATOMIC); 220 if (error) 221 goto end_copy; 222 } 223 } 224 } 225 226end_copy: 227 rcu_read_unlock(); 228 return error; 229} 230 231/* Initialize a sctp_addr from in incoming skb. */ 232static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 233 int is_saddr) 234{ 235 void *from; 236 __be16 *port; 237 struct sctphdr *sh; 238 239 port = &addr->v4.sin_port; 240 addr->v4.sin_family = AF_INET; 241 242 sh = sctp_hdr(skb); 243 if (is_saddr) { 244 *port = sh->source; 245 from = &ip_hdr(skb)->saddr; 246 } else { 247 *port = sh->dest; 248 from = &ip_hdr(skb)->daddr; 249 } 250 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr)); 251} 252 253/* Initialize an sctp_addr from a socket. */ 254static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 255{ 256 addr->v4.sin_family = AF_INET; 257 addr->v4.sin_port = 0; 258 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 259} 260 261/* Initialize sk->sk_rcv_saddr from sctp_addr. */ 262static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 263{ 264 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 265} 266 267/* Initialize sk->sk_daddr from sctp_addr. */ 268static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 269{ 270 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 271} 272 273/* Initialize a sctp_addr from an address parameter. */ 274static void sctp_v4_from_addr_param(union sctp_addr *addr, 275 union sctp_addr_param *param, 276 __be16 port, int iif) 277{ 278 addr->v4.sin_family = AF_INET; 279 addr->v4.sin_port = port; 280 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 281} 282 283/* Initialize an address parameter from a sctp_addr and return the length 284 * of the address parameter. 285 */ 286static int sctp_v4_to_addr_param(const union sctp_addr *addr, 287 union sctp_addr_param *param) 288{ 289 int length = sizeof(sctp_ipv4addr_param_t); 290 291 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 292 param->v4.param_hdr.length = htons(length); 293 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 294 295 return length; 296} 297 298/* Initialize a sctp_addr from a dst_entry. */ 299static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 300 __be16 port) 301{ 302 saddr->v4.sin_family = AF_INET; 303 saddr->v4.sin_port = port; 304 saddr->v4.sin_addr.s_addr = fl4->saddr; 305} 306 307/* Compare two addresses exactly. */ 308static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 309 const union sctp_addr *addr2) 310{ 311 if (addr1->sa.sa_family != addr2->sa.sa_family) 312 return 0; 313 if (addr1->v4.sin_port != addr2->v4.sin_port) 314 return 0; 315 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 316 return 0; 317 318 return 1; 319} 320 321/* Initialize addr struct to INADDR_ANY. */ 322static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 323{ 324 addr->v4.sin_family = AF_INET; 325 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 326 addr->v4.sin_port = port; 327} 328 329/* Is this a wildcard address? */ 330static int sctp_v4_is_any(const union sctp_addr *addr) 331{ 332 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 333} 334 335/* This function checks if the address is a valid address to be used for 336 * SCTP binding. 337 * 338 * Output: 339 * Return 0 - If the address is a non-unicast or an illegal address. 340 * Return 1 - If the address is a unicast. 341 */ 342static int sctp_v4_addr_valid(union sctp_addr *addr, 343 struct sctp_sock *sp, 344 const struct sk_buff *skb) 345{ 346 /* IPv4 addresses not allowed */ 347 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 348 return 0; 349 350 /* Is this a non-unicast address or a unusable SCTP address? */ 351 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 352 return 0; 353 354 /* Is this a broadcast address? */ 355 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 356 return 0; 357 358 return 1; 359} 360 361/* Should this be available for binding? */ 362static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 363{ 364 struct net *net = sock_net(&sp->inet.sk); 365 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr); 366 367 368 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 369 ret != RTN_LOCAL && 370 !sp->inet.freebind && 371 !net->ipv4.sysctl_ip_nonlocal_bind) 372 return 0; 373 374 if (ipv6_only_sock(sctp_opt2sk(sp))) 375 return 0; 376 377 return 1; 378} 379 380/* Checking the loopback, private and other address scopes as defined in 381 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 382 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 383 * 384 * Level 0 - unusable SCTP addresses 385 * Level 1 - loopback address 386 * Level 2 - link-local addresses 387 * Level 3 - private addresses. 388 * Level 4 - global addresses 389 * For INIT and INIT-ACK address list, let L be the level of 390 * of requested destination address, sender and receiver 391 * SHOULD include all of its addresses with level greater 392 * than or equal to L. 393 * 394 * IPv4 scoping can be controlled through sysctl option 395 * net.sctp.addr_scope_policy 396 */ 397static sctp_scope_t sctp_v4_scope(union sctp_addr *addr) 398{ 399 sctp_scope_t retval; 400 401 /* Check for unusable SCTP addresses. */ 402 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 403 retval = SCTP_SCOPE_UNUSABLE; 404 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 405 retval = SCTP_SCOPE_LOOPBACK; 406 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 407 retval = SCTP_SCOPE_LINK; 408 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 409 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 410 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) { 411 retval = SCTP_SCOPE_PRIVATE; 412 } else { 413 retval = SCTP_SCOPE_GLOBAL; 414 } 415 416 return retval; 417} 418 419/* Returns a valid dst cache entry for the given source and destination ip 420 * addresses. If an association is passed, trys to get a dst entry with a 421 * source address that matches an address in the bind address list. 422 */ 423static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 424 struct flowi *fl, struct sock *sk) 425{ 426 struct sctp_association *asoc = t->asoc; 427 struct rtable *rt; 428 struct flowi4 *fl4 = &fl->u.ip4; 429 struct sctp_bind_addr *bp; 430 struct sctp_sockaddr_entry *laddr; 431 struct dst_entry *dst = NULL; 432 union sctp_addr *daddr = &t->ipaddr; 433 union sctp_addr dst_saddr; 434 435 memset(fl4, 0x0, sizeof(struct flowi4)); 436 fl4->daddr = daddr->v4.sin_addr.s_addr; 437 fl4->fl4_dport = daddr->v4.sin_port; 438 fl4->flowi4_proto = IPPROTO_SCTP; 439 if (asoc) { 440 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk); 441 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 442 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 443 } 444 if (saddr) { 445 fl4->saddr = saddr->v4.sin_addr.s_addr; 446 fl4->fl4_sport = saddr->v4.sin_port; 447 } 448 449 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 450 &fl4->saddr); 451 452 rt = ip_route_output_key(sock_net(sk), fl4); 453 if (!IS_ERR(rt)) 454 dst = &rt->dst; 455 456 /* If there is no association or if a source address is passed, no 457 * more validation is required. 458 */ 459 if (!asoc || saddr) 460 goto out; 461 462 bp = &asoc->base.bind_addr; 463 464 if (dst) { 465 /* Walk through the bind address list and look for a bind 466 * address that matches the source address of the returned dst. 467 */ 468 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 469 rcu_read_lock(); 470 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 471 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 472 (laddr->state != SCTP_ADDR_SRC && 473 !asoc->src_out_of_asoc_ok)) 474 continue; 475 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 476 goto out_unlock; 477 } 478 rcu_read_unlock(); 479 480 /* None of the bound addresses match the source address of the 481 * dst. So release it. 482 */ 483 dst_release(dst); 484 dst = NULL; 485 } 486 487 /* Walk through the bind address list and try to get a dst that 488 * matches a bind address as the source address. 489 */ 490 rcu_read_lock(); 491 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 492 if (!laddr->valid) 493 continue; 494 if ((laddr->state == SCTP_ADDR_SRC) && 495 (AF_INET == laddr->a.sa.sa_family)) { 496 fl4->fl4_sport = laddr->a.v4.sin_port; 497 flowi4_update_output(fl4, 498 asoc->base.sk->sk_bound_dev_if, 499 RT_CONN_FLAGS(asoc->base.sk), 500 daddr->v4.sin_addr.s_addr, 501 laddr->a.v4.sin_addr.s_addr); 502 503 rt = ip_route_output_key(sock_net(sk), fl4); 504 if (!IS_ERR(rt)) { 505 dst = &rt->dst; 506 goto out_unlock; 507 } 508 } 509 } 510 511out_unlock: 512 rcu_read_unlock(); 513out: 514 t->dst = dst; 515 if (dst) 516 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 517 &fl4->daddr, &fl4->saddr); 518 else 519 pr_debug("no route\n"); 520} 521 522/* For v4, the source address is cached in the route entry(dst). So no need 523 * to cache it separately and hence this is an empty routine. 524 */ 525static void sctp_v4_get_saddr(struct sctp_sock *sk, 526 struct sctp_transport *t, 527 struct flowi *fl) 528{ 529 union sctp_addr *saddr = &t->saddr; 530 struct rtable *rt = (struct rtable *)t->dst; 531 532 if (rt) { 533 saddr->v4.sin_family = AF_INET; 534 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 535 } 536} 537 538/* What interface did this skb arrive on? */ 539static int sctp_v4_skb_iif(const struct sk_buff *skb) 540{ 541 return inet_iif(skb); 542} 543 544/* Was this packet marked by Explicit Congestion Notification? */ 545static int sctp_v4_is_ce(const struct sk_buff *skb) 546{ 547 return INET_ECN_is_ce(ip_hdr(skb)->tos); 548} 549 550/* Create and initialize a new sk for the socket returned by accept(). */ 551static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 552 struct sctp_association *asoc) 553{ 554 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 555 sk->sk_prot); 556 struct inet_sock *newinet; 557 558 if (!newsk) 559 goto out; 560 561 sock_init_data(NULL, newsk); 562 563 sctp_copy_sock(newsk, sk, asoc); 564 sock_reset_flag(newsk, SOCK_ZAPPED); 565 566 newinet = inet_sk(newsk); 567 568 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 569 570 sk_refcnt_debug_inc(newsk); 571 572 if (newsk->sk_prot->init(newsk)) { 573 sk_common_release(newsk); 574 newsk = NULL; 575 } 576 577out: 578 return newsk; 579} 580 581static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 582{ 583 /* No address mapping for V4 sockets */ 584 return sizeof(struct sockaddr_in); 585} 586 587/* Dump the v4 addr to the seq file. */ 588static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 589{ 590 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 591} 592 593static void sctp_v4_ecn_capable(struct sock *sk) 594{ 595 INET_ECN_xmit(sk); 596} 597 598static void sctp_addr_wq_timeout_handler(unsigned long arg) 599{ 600 struct net *net = (struct net *)arg; 601 struct sctp_sockaddr_entry *addrw, *temp; 602 struct sctp_sock *sp; 603 604 spin_lock_bh(&net->sctp.addr_wq_lock); 605 606 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 607 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 608 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 609 addrw->state, addrw); 610 611#if IS_ENABLED(CONFIG_IPV6) 612 /* Now we send an ASCONF for each association */ 613 /* Note. we currently don't handle link local IPv6 addressees */ 614 if (addrw->a.sa.sa_family == AF_INET6) { 615 struct in6_addr *in6; 616 617 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 618 IPV6_ADDR_LINKLOCAL) 619 goto free_next; 620 621 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 622 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 623 addrw->state == SCTP_ADDR_NEW) { 624 unsigned long timeo_val; 625 626 pr_debug("%s: this is on DAD, trying %d sec " 627 "later\n", __func__, 628 SCTP_ADDRESS_TICK_DELAY); 629 630 timeo_val = jiffies; 631 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 632 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 633 break; 634 } 635 } 636#endif 637 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 638 struct sock *sk; 639 640 sk = sctp_opt2sk(sp); 641 /* ignore bound-specific endpoints */ 642 if (!sctp_is_ep_boundall(sk)) 643 continue; 644 bh_lock_sock(sk); 645 if (sctp_asconf_mgmt(sp, addrw) < 0) 646 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 647 bh_unlock_sock(sk); 648 } 649#if IS_ENABLED(CONFIG_IPV6) 650free_next: 651#endif 652 list_del(&addrw->list); 653 kfree(addrw); 654 } 655 spin_unlock_bh(&net->sctp.addr_wq_lock); 656} 657 658static void sctp_free_addr_wq(struct net *net) 659{ 660 struct sctp_sockaddr_entry *addrw; 661 struct sctp_sockaddr_entry *temp; 662 663 spin_lock_bh(&net->sctp.addr_wq_lock); 664 del_timer(&net->sctp.addr_wq_timer); 665 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 666 list_del(&addrw->list); 667 kfree(addrw); 668 } 669 spin_unlock_bh(&net->sctp.addr_wq_lock); 670} 671 672/* lookup the entry for the same address in the addr_waitq 673 * sctp_addr_wq MUST be locked 674 */ 675static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 676 struct sctp_sockaddr_entry *addr) 677{ 678 struct sctp_sockaddr_entry *addrw; 679 680 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 681 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 682 continue; 683 if (addrw->a.sa.sa_family == AF_INET) { 684 if (addrw->a.v4.sin_addr.s_addr == 685 addr->a.v4.sin_addr.s_addr) 686 return addrw; 687 } else if (addrw->a.sa.sa_family == AF_INET6) { 688 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 689 &addr->a.v6.sin6_addr)) 690 return addrw; 691 } 692 } 693 return NULL; 694} 695 696void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 697{ 698 struct sctp_sockaddr_entry *addrw; 699 unsigned long timeo_val; 700 701 /* first, we check if an opposite message already exist in the queue. 702 * If we found such message, it is removed. 703 * This operation is a bit stupid, but the DHCP client attaches the 704 * new address after a couple of addition and deletion of that address 705 */ 706 707 spin_lock_bh(&net->sctp.addr_wq_lock); 708 /* Offsets existing events in addr_wq */ 709 addrw = sctp_addr_wq_lookup(net, addr); 710 if (addrw) { 711 if (addrw->state != cmd) { 712 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 713 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 714 &net->sctp.addr_waitq); 715 716 list_del(&addrw->list); 717 kfree(addrw); 718 } 719 spin_unlock_bh(&net->sctp.addr_wq_lock); 720 return; 721 } 722 723 /* OK, we have to add the new address to the wait queue */ 724 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 725 if (addrw == NULL) { 726 spin_unlock_bh(&net->sctp.addr_wq_lock); 727 return; 728 } 729 addrw->state = cmd; 730 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 731 732 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 733 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 734 735 if (!timer_pending(&net->sctp.addr_wq_timer)) { 736 timeo_val = jiffies; 737 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 738 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 739 } 740 spin_unlock_bh(&net->sctp.addr_wq_lock); 741} 742 743/* Event handler for inet address addition/deletion events. 744 * The sctp_local_addr_list needs to be protocted by a spin lock since 745 * multiple notifiers (say IPv4 and IPv6) may be running at the same 746 * time and thus corrupt the list. 747 * The reader side is protected with RCU. 748 */ 749static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 750 void *ptr) 751{ 752 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 753 struct sctp_sockaddr_entry *addr = NULL; 754 struct sctp_sockaddr_entry *temp; 755 struct net *net = dev_net(ifa->ifa_dev->dev); 756 int found = 0; 757 758 switch (ev) { 759 case NETDEV_UP: 760 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 761 if (addr) { 762 addr->a.v4.sin_family = AF_INET; 763 addr->a.v4.sin_port = 0; 764 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 765 addr->valid = 1; 766 spin_lock_bh(&net->sctp.local_addr_lock); 767 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 768 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 769 spin_unlock_bh(&net->sctp.local_addr_lock); 770 } 771 break; 772 case NETDEV_DOWN: 773 spin_lock_bh(&net->sctp.local_addr_lock); 774 list_for_each_entry_safe(addr, temp, 775 &net->sctp.local_addr_list, list) { 776 if (addr->a.sa.sa_family == AF_INET && 777 addr->a.v4.sin_addr.s_addr == 778 ifa->ifa_local) { 779 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 780 found = 1; 781 addr->valid = 0; 782 list_del_rcu(&addr->list); 783 break; 784 } 785 } 786 spin_unlock_bh(&net->sctp.local_addr_lock); 787 if (found) 788 kfree_rcu(addr, rcu); 789 break; 790 } 791 792 return NOTIFY_DONE; 793} 794 795/* 796 * Initialize the control inode/socket with a control endpoint data 797 * structure. This endpoint is reserved exclusively for the OOTB processing. 798 */ 799static int sctp_ctl_sock_init(struct net *net) 800{ 801 int err; 802 sa_family_t family = PF_INET; 803 804 if (sctp_get_pf_specific(PF_INET6)) 805 family = PF_INET6; 806 807 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 808 SOCK_SEQPACKET, IPPROTO_SCTP, net); 809 810 /* If IPv6 socket could not be created, try the IPv4 socket */ 811 if (err < 0 && family == PF_INET6) 812 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 813 SOCK_SEQPACKET, IPPROTO_SCTP, 814 net); 815 816 if (err < 0) { 817 pr_err("Failed to create the SCTP control socket\n"); 818 return err; 819 } 820 return 0; 821} 822 823/* Register address family specific functions. */ 824int sctp_register_af(struct sctp_af *af) 825{ 826 switch (af->sa_family) { 827 case AF_INET: 828 if (sctp_af_v4_specific) 829 return 0; 830 sctp_af_v4_specific = af; 831 break; 832 case AF_INET6: 833 if (sctp_af_v6_specific) 834 return 0; 835 sctp_af_v6_specific = af; 836 break; 837 default: 838 return 0; 839 } 840 841 INIT_LIST_HEAD(&af->list); 842 list_add_tail(&af->list, &sctp_address_families); 843 return 1; 844} 845 846/* Get the table of functions for manipulating a particular address 847 * family. 848 */ 849struct sctp_af *sctp_get_af_specific(sa_family_t family) 850{ 851 switch (family) { 852 case AF_INET: 853 return sctp_af_v4_specific; 854 case AF_INET6: 855 return sctp_af_v6_specific; 856 default: 857 return NULL; 858 } 859} 860 861/* Common code to initialize a AF_INET msg_name. */ 862static void sctp_inet_msgname(char *msgname, int *addr_len) 863{ 864 struct sockaddr_in *sin; 865 866 sin = (struct sockaddr_in *)msgname; 867 *addr_len = sizeof(struct sockaddr_in); 868 sin->sin_family = AF_INET; 869 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 870} 871 872/* Copy the primary address of the peer primary address as the msg_name. */ 873static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 874 int *addr_len) 875{ 876 struct sockaddr_in *sin, *sinfrom; 877 878 if (msgname) { 879 struct sctp_association *asoc; 880 881 asoc = event->asoc; 882 sctp_inet_msgname(msgname, addr_len); 883 sin = (struct sockaddr_in *)msgname; 884 sinfrom = &asoc->peer.primary_addr.v4; 885 sin->sin_port = htons(asoc->peer.port); 886 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 887 } 888} 889 890/* Initialize and copy out a msgname from an inbound skb. */ 891static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 892{ 893 if (msgname) { 894 struct sctphdr *sh = sctp_hdr(skb); 895 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 896 897 sctp_inet_msgname(msgname, len); 898 sin->sin_port = sh->source; 899 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 900 } 901} 902 903/* Do we support this AF? */ 904static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 905{ 906 /* PF_INET only supports AF_INET addresses. */ 907 return AF_INET == family; 908} 909 910/* Address matching with wildcards allowed. */ 911static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 912 const union sctp_addr *addr2, 913 struct sctp_sock *opt) 914{ 915 /* PF_INET only supports AF_INET addresses. */ 916 if (addr1->sa.sa_family != addr2->sa.sa_family) 917 return 0; 918 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 919 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 920 return 1; 921 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 922 return 1; 923 924 return 0; 925} 926 927/* Verify that provided sockaddr looks bindable. Common verification has 928 * already been taken care of. 929 */ 930static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 931{ 932 return sctp_v4_available(addr, opt); 933} 934 935/* Verify that sockaddr looks sendable. Common verification has already 936 * been taken care of. 937 */ 938static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 939{ 940 return 1; 941} 942 943/* Fill in Supported Address Type information for INIT and INIT-ACK 944 * chunks. Returns number of addresses supported. 945 */ 946static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 947 __be16 *types) 948{ 949 types[0] = SCTP_PARAM_IPV4_ADDRESS; 950 return 1; 951} 952 953/* Wrapper routine that calls the ip transmit routine. */ 954static inline int sctp_v4_xmit(struct sk_buff *skb, 955 struct sctp_transport *transport) 956{ 957 struct inet_sock *inet = inet_sk(skb->sk); 958 959 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 960 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr); 961 962 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ? 963 IP_PMTUDISC_DO : IP_PMTUDISC_DONT; 964 965 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS); 966 967 return ip_queue_xmit(&inet->sk, skb, &transport->fl); 968} 969 970static struct sctp_af sctp_af_inet; 971 972static struct sctp_pf sctp_pf_inet = { 973 .event_msgname = sctp_inet_event_msgname, 974 .skb_msgname = sctp_inet_skb_msgname, 975 .af_supported = sctp_inet_af_supported, 976 .cmp_addr = sctp_inet_cmp_addr, 977 .bind_verify = sctp_inet_bind_verify, 978 .send_verify = sctp_inet_send_verify, 979 .supported_addrs = sctp_inet_supported_addrs, 980 .create_accept_sk = sctp_v4_create_accept_sk, 981 .addr_to_user = sctp_v4_addr_to_user, 982 .to_sk_saddr = sctp_v4_to_sk_saddr, 983 .to_sk_daddr = sctp_v4_to_sk_daddr, 984 .af = &sctp_af_inet 985}; 986 987/* Notifier for inetaddr addition/deletion events. */ 988static struct notifier_block sctp_inetaddr_notifier = { 989 .notifier_call = sctp_inetaddr_event, 990}; 991 992/* Socket operations. */ 993static const struct proto_ops inet_seqpacket_ops = { 994 .family = PF_INET, 995 .owner = THIS_MODULE, 996 .release = inet_release, /* Needs to be wrapped... */ 997 .bind = inet_bind, 998 .connect = inet_dgram_connect, 999 .socketpair = sock_no_socketpair, 1000 .accept = inet_accept, 1001 .getname = inet_getname, /* Semantics are different. */ 1002 .poll = sctp_poll, 1003 .ioctl = inet_ioctl, 1004 .listen = sctp_inet_listen, 1005 .shutdown = inet_shutdown, /* Looks harmless. */ 1006 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1007 .getsockopt = sock_common_getsockopt, 1008 .sendmsg = inet_sendmsg, 1009 .recvmsg = sock_common_recvmsg, 1010 .mmap = sock_no_mmap, 1011 .sendpage = sock_no_sendpage, 1012#ifdef CONFIG_COMPAT 1013 .compat_setsockopt = compat_sock_common_setsockopt, 1014 .compat_getsockopt = compat_sock_common_getsockopt, 1015#endif 1016}; 1017 1018/* Registration with AF_INET family. */ 1019static struct inet_protosw sctp_seqpacket_protosw = { 1020 .type = SOCK_SEQPACKET, 1021 .protocol = IPPROTO_SCTP, 1022 .prot = &sctp_prot, 1023 .ops = &inet_seqpacket_ops, 1024 .flags = SCTP_PROTOSW_FLAG 1025}; 1026static struct inet_protosw sctp_stream_protosw = { 1027 .type = SOCK_STREAM, 1028 .protocol = IPPROTO_SCTP, 1029 .prot = &sctp_prot, 1030 .ops = &inet_seqpacket_ops, 1031 .flags = SCTP_PROTOSW_FLAG 1032}; 1033 1034/* Register with IP layer. */ 1035static const struct net_protocol sctp_protocol = { 1036 .handler = sctp_rcv, 1037 .err_handler = sctp_v4_err, 1038 .no_policy = 1, 1039 .netns_ok = 1, 1040 .icmp_strict_tag_validation = 1, 1041}; 1042 1043/* IPv4 address related functions. */ 1044static struct sctp_af sctp_af_inet = { 1045 .sa_family = AF_INET, 1046 .sctp_xmit = sctp_v4_xmit, 1047 .setsockopt = ip_setsockopt, 1048 .getsockopt = ip_getsockopt, 1049 .get_dst = sctp_v4_get_dst, 1050 .get_saddr = sctp_v4_get_saddr, 1051 .copy_addrlist = sctp_v4_copy_addrlist, 1052 .from_skb = sctp_v4_from_skb, 1053 .from_sk = sctp_v4_from_sk, 1054 .from_addr_param = sctp_v4_from_addr_param, 1055 .to_addr_param = sctp_v4_to_addr_param, 1056 .cmp_addr = sctp_v4_cmp_addr, 1057 .addr_valid = sctp_v4_addr_valid, 1058 .inaddr_any = sctp_v4_inaddr_any, 1059 .is_any = sctp_v4_is_any, 1060 .available = sctp_v4_available, 1061 .scope = sctp_v4_scope, 1062 .skb_iif = sctp_v4_skb_iif, 1063 .is_ce = sctp_v4_is_ce, 1064 .seq_dump_addr = sctp_v4_seq_dump_addr, 1065 .ecn_capable = sctp_v4_ecn_capable, 1066 .net_header_len = sizeof(struct iphdr), 1067 .sockaddr_len = sizeof(struct sockaddr_in), 1068#ifdef CONFIG_COMPAT 1069 .compat_setsockopt = compat_ip_setsockopt, 1070 .compat_getsockopt = compat_ip_getsockopt, 1071#endif 1072}; 1073 1074struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1075{ 1076 switch (family) { 1077 case PF_INET: 1078 return sctp_pf_inet_specific; 1079 case PF_INET6: 1080 return sctp_pf_inet6_specific; 1081 default: 1082 return NULL; 1083 } 1084} 1085 1086/* Register the PF specific function table. */ 1087int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1088{ 1089 switch (family) { 1090 case PF_INET: 1091 if (sctp_pf_inet_specific) 1092 return 0; 1093 sctp_pf_inet_specific = pf; 1094 break; 1095 case PF_INET6: 1096 if (sctp_pf_inet6_specific) 1097 return 0; 1098 sctp_pf_inet6_specific = pf; 1099 break; 1100 default: 1101 return 0; 1102 } 1103 return 1; 1104} 1105 1106static inline int init_sctp_mibs(struct net *net) 1107{ 1108 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1109 if (!net->sctp.sctp_statistics) 1110 return -ENOMEM; 1111 return 0; 1112} 1113 1114static inline void cleanup_sctp_mibs(struct net *net) 1115{ 1116 free_percpu(net->sctp.sctp_statistics); 1117} 1118 1119static void sctp_v4_pf_init(void) 1120{ 1121 /* Initialize the SCTP specific PF functions. */ 1122 sctp_register_pf(&sctp_pf_inet, PF_INET); 1123 sctp_register_af(&sctp_af_inet); 1124} 1125 1126static void sctp_v4_pf_exit(void) 1127{ 1128 list_del(&sctp_af_inet.list); 1129} 1130 1131static int sctp_v4_protosw_init(void) 1132{ 1133 int rc; 1134 1135 rc = proto_register(&sctp_prot, 1); 1136 if (rc) 1137 return rc; 1138 1139 /* Register SCTP(UDP and TCP style) with socket layer. */ 1140 inet_register_protosw(&sctp_seqpacket_protosw); 1141 inet_register_protosw(&sctp_stream_protosw); 1142 1143 return 0; 1144} 1145 1146static void sctp_v4_protosw_exit(void) 1147{ 1148 inet_unregister_protosw(&sctp_stream_protosw); 1149 inet_unregister_protosw(&sctp_seqpacket_protosw); 1150 proto_unregister(&sctp_prot); 1151} 1152 1153static int sctp_v4_add_protocol(void) 1154{ 1155 /* Register notifier for inet address additions/deletions. */ 1156 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1157 1158 /* Register SCTP with inet layer. */ 1159 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1160 return -EAGAIN; 1161 1162 return 0; 1163} 1164 1165static void sctp_v4_del_protocol(void) 1166{ 1167 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1168 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1169} 1170 1171static int __net_init sctp_defaults_init(struct net *net) 1172{ 1173 int status; 1174 1175 /* 1176 * 14. Suggested SCTP Protocol Parameter Values 1177 */ 1178 /* The following protocol parameters are RECOMMENDED: */ 1179 /* RTO.Initial - 3 seconds */ 1180 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1181 /* RTO.Min - 1 second */ 1182 net->sctp.rto_min = SCTP_RTO_MIN; 1183 /* RTO.Max - 60 seconds */ 1184 net->sctp.rto_max = SCTP_RTO_MAX; 1185 /* RTO.Alpha - 1/8 */ 1186 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1187 /* RTO.Beta - 1/4 */ 1188 net->sctp.rto_beta = SCTP_RTO_BETA; 1189 1190 /* Valid.Cookie.Life - 60 seconds */ 1191 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1192 1193 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1194 net->sctp.cookie_preserve_enable = 1; 1195 1196 /* Default sctp sockets to use md5 as their hmac alg */ 1197#if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1198 net->sctp.sctp_hmac_alg = "md5"; 1199#elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1200 net->sctp.sctp_hmac_alg = "sha1"; 1201#else 1202 net->sctp.sctp_hmac_alg = NULL; 1203#endif 1204 1205 /* Max.Burst - 4 */ 1206 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1207 1208 /* Association.Max.Retrans - 10 attempts 1209 * Path.Max.Retrans - 5 attempts (per destination address) 1210 * Max.Init.Retransmits - 8 attempts 1211 */ 1212 net->sctp.max_retrans_association = 10; 1213 net->sctp.max_retrans_path = 5; 1214 net->sctp.max_retrans_init = 8; 1215 1216 /* Sendbuffer growth - do per-socket accounting */ 1217 net->sctp.sndbuf_policy = 0; 1218 1219 /* Rcvbuffer growth - do per-socket accounting */ 1220 net->sctp.rcvbuf_policy = 0; 1221 1222 /* HB.interval - 30 seconds */ 1223 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1224 1225 /* delayed SACK timeout */ 1226 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1227 1228 /* Disable ADDIP by default. */ 1229 net->sctp.addip_enable = 0; 1230 net->sctp.addip_noauth = 0; 1231 net->sctp.default_auto_asconf = 0; 1232 1233 /* Enable PR-SCTP by default. */ 1234 net->sctp.prsctp_enable = 1; 1235 1236 /* Disable AUTH by default. */ 1237 net->sctp.auth_enable = 0; 1238 1239 /* Set SCOPE policy to enabled */ 1240 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1241 1242 /* Set the default rwnd update threshold */ 1243 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1244 1245 /* Initialize maximum autoclose timeout. */ 1246 net->sctp.max_autoclose = INT_MAX / HZ; 1247 1248 status = sctp_sysctl_net_register(net); 1249 if (status) 1250 goto err_sysctl_register; 1251 1252 /* Allocate and initialise sctp mibs. */ 1253 status = init_sctp_mibs(net); 1254 if (status) 1255 goto err_init_mibs; 1256 1257 /* Initialize proc fs directory. */ 1258 status = sctp_proc_init(net); 1259 if (status) 1260 goto err_init_proc; 1261 1262 sctp_dbg_objcnt_init(net); 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 1279err_init_proc: 1280 cleanup_sctp_mibs(net); 1281err_init_mibs: 1282 sctp_sysctl_net_unregister(net); 1283err_sysctl_register: 1284 return status; 1285} 1286 1287static void __net_exit sctp_defaults_exit(struct net *net) 1288{ 1289 /* Free the local address list */ 1290 sctp_free_addr_wq(net); 1291 sctp_free_local_addr_list(net); 1292 1293 sctp_dbg_objcnt_exit(net); 1294 1295 sctp_proc_exit(net); 1296 cleanup_sctp_mibs(net); 1297 sctp_sysctl_net_unregister(net); 1298} 1299 1300static struct pernet_operations sctp_defaults_ops = { 1301 .init = sctp_defaults_init, 1302 .exit = sctp_defaults_exit, 1303}; 1304 1305static int __net_init sctp_ctrlsock_init(struct net *net) 1306{ 1307 int status; 1308 1309 /* Initialize the control inode/socket for handling OOTB packets. */ 1310 status = sctp_ctl_sock_init(net); 1311 if (status) 1312 pr_err("Failed to initialize the SCTP control sock\n"); 1313 1314 return status; 1315} 1316 1317static void __net_init sctp_ctrlsock_exit(struct net *net) 1318{ 1319 /* Free the control endpoint. */ 1320 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1321} 1322 1323static struct pernet_operations sctp_ctrlsock_ops = { 1324 .init = sctp_ctrlsock_init, 1325 .exit = sctp_ctrlsock_exit, 1326}; 1327 1328/* Initialize the universe into something sensible. */ 1329static __init int sctp_init(void) 1330{ 1331 int i; 1332 int status = -EINVAL; 1333 unsigned long goal; 1334 unsigned long limit; 1335 int max_share; 1336 int order; 1337 int num_entries; 1338 int max_entry_order; 1339 1340 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1341 1342 /* Allocate bind_bucket and chunk caches. */ 1343 status = -ENOBUFS; 1344 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket", 1345 sizeof(struct sctp_bind_bucket), 1346 0, SLAB_HWCACHE_ALIGN, 1347 NULL); 1348 if (!sctp_bucket_cachep) 1349 goto out; 1350 1351 sctp_chunk_cachep = kmem_cache_create("sctp_chunk", 1352 sizeof(struct sctp_chunk), 1353 0, SLAB_HWCACHE_ALIGN, 1354 NULL); 1355 if (!sctp_chunk_cachep) 1356 goto err_chunk_cachep; 1357 1358 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1359 if (status) 1360 goto err_percpu_counter_init; 1361 1362 /* Implementation specific variables. */ 1363 1364 /* Initialize default stream count setup information. */ 1365 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1366 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1367 1368 /* Initialize handle used for association ids. */ 1369 idr_init(&sctp_assocs_id); 1370 1371 limit = nr_free_buffer_pages() / 8; 1372 limit = max(limit, 128UL); 1373 sysctl_sctp_mem[0] = limit / 4 * 3; 1374 sysctl_sctp_mem[1] = limit; 1375 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1376 1377 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1378 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1379 max_share = min(4UL*1024*1024, limit); 1380 1381 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */ 1382 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1383 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1384 1385 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM; 1386 sysctl_sctp_wmem[1] = 16*1024; 1387 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1388 1389 /* Size and allocate the association hash table. 1390 * The methodology is similar to that of the tcp hash tables. 1391 * Though not identical. Start by getting a goal size 1392 */ 1393 if (totalram_pages >= (128 * 1024)) 1394 goal = totalram_pages >> (22 - PAGE_SHIFT); 1395 else 1396 goal = totalram_pages >> (24 - PAGE_SHIFT); 1397 1398 /* Then compute the page order for said goal */ 1399 order = get_order(goal); 1400 1401 /* Now compute the required page order for the maximum sized table we 1402 * want to create 1403 */ 1404 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1405 sizeof(struct sctp_bind_hashbucket)); 1406 1407 /* Limit the page order by that maximum hash table size */ 1408 order = min(order, max_entry_order); 1409 1410 do { 1411 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE / 1412 sizeof(struct sctp_hashbucket); 1413 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0) 1414 continue; 1415 sctp_assoc_hashtable = (struct sctp_hashbucket *) 1416 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order); 1417 } while (!sctp_assoc_hashtable && --order > 0); 1418 if (!sctp_assoc_hashtable) { 1419 pr_err("Failed association hash alloc\n"); 1420 status = -ENOMEM; 1421 goto err_ahash_alloc; 1422 } 1423 for (i = 0; i < sctp_assoc_hashsize; i++) { 1424 rwlock_init(&sctp_assoc_hashtable[i].lock); 1425 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain); 1426 } 1427 1428 /* Allocate and initialize the endpoint hash table. */ 1429 sctp_ep_hashsize = 64; 1430 sctp_ep_hashtable = 1431 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL); 1432 if (!sctp_ep_hashtable) { 1433 pr_err("Failed endpoint_hash alloc\n"); 1434 status = -ENOMEM; 1435 goto err_ehash_alloc; 1436 } 1437 for (i = 0; i < sctp_ep_hashsize; i++) { 1438 rwlock_init(&sctp_ep_hashtable[i].lock); 1439 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1440 } 1441 1442 /* Allocate and initialize the SCTP port hash table. 1443 * Note that order is initalized to start at the max sized 1444 * table we want to support. If we can't get that many pages 1445 * reduce the order and try again 1446 */ 1447 do { 1448 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1449 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order); 1450 } while (!sctp_port_hashtable && --order > 0); 1451 1452 if (!sctp_port_hashtable) { 1453 pr_err("Failed bind hash alloc\n"); 1454 status = -ENOMEM; 1455 goto err_bhash_alloc; 1456 } 1457 1458 /* Now compute the number of entries that will fit in the 1459 * port hash space we allocated 1460 */ 1461 num_entries = (1UL << order) * PAGE_SIZE / 1462 sizeof(struct sctp_bind_hashbucket); 1463 1464 /* And finish by rounding it down to the nearest power of two 1465 * this wastes some memory of course, but its needed because 1466 * the hash function operates based on the assumption that 1467 * that the number of entries is a power of two 1468 */ 1469 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1470 1471 for (i = 0; i < sctp_port_hashsize; i++) { 1472 spin_lock_init(&sctp_port_hashtable[i].lock); 1473 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1474 } 1475 1476 pr_info("Hash tables configured (established %d bind %d/%d)\n", 1477 sctp_assoc_hashsize, sctp_port_hashsize, num_entries); 1478 1479 sctp_sysctl_register(); 1480 1481 INIT_LIST_HEAD(&sctp_address_families); 1482 sctp_v4_pf_init(); 1483 sctp_v6_pf_init(); 1484 1485 status = register_pernet_subsys(&sctp_defaults_ops); 1486 if (status) 1487 goto err_register_defaults; 1488 1489 status = sctp_v4_protosw_init(); 1490 if (status) 1491 goto err_protosw_init; 1492 1493 status = sctp_v6_protosw_init(); 1494 if (status) 1495 goto err_v6_protosw_init; 1496 1497 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1498 if (status) 1499 goto err_register_ctrlsock; 1500 1501 status = sctp_v4_add_protocol(); 1502 if (status) 1503 goto err_add_protocol; 1504 1505 /* Register SCTP with inet6 layer. */ 1506 status = sctp_v6_add_protocol(); 1507 if (status) 1508 goto err_v6_add_protocol; 1509 1510out: 1511 return status; 1512err_v6_add_protocol: 1513 sctp_v4_del_protocol(); 1514err_add_protocol: 1515 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1516err_register_ctrlsock: 1517 sctp_v6_protosw_exit(); 1518err_v6_protosw_init: 1519 sctp_v4_protosw_exit(); 1520err_protosw_init: 1521 unregister_pernet_subsys(&sctp_defaults_ops); 1522err_register_defaults: 1523 sctp_v4_pf_exit(); 1524 sctp_v6_pf_exit(); 1525 sctp_sysctl_unregister(); 1526 free_pages((unsigned long)sctp_port_hashtable, 1527 get_order(sctp_port_hashsize * 1528 sizeof(struct sctp_bind_hashbucket))); 1529err_bhash_alloc: 1530 kfree(sctp_ep_hashtable); 1531err_ehash_alloc: 1532 free_pages((unsigned long)sctp_assoc_hashtable, 1533 get_order(sctp_assoc_hashsize * 1534 sizeof(struct sctp_hashbucket))); 1535err_ahash_alloc: 1536 percpu_counter_destroy(&sctp_sockets_allocated); 1537err_percpu_counter_init: 1538 kmem_cache_destroy(sctp_chunk_cachep); 1539err_chunk_cachep: 1540 kmem_cache_destroy(sctp_bucket_cachep); 1541 goto out; 1542} 1543 1544/* Exit handler for the SCTP protocol. */ 1545static __exit void sctp_exit(void) 1546{ 1547 /* BUG. This should probably do something useful like clean 1548 * up all the remaining associations and all that memory. 1549 */ 1550 1551 /* Unregister with inet6/inet layers. */ 1552 sctp_v6_del_protocol(); 1553 sctp_v4_del_protocol(); 1554 1555 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1556 1557 /* Free protosw registrations */ 1558 sctp_v6_protosw_exit(); 1559 sctp_v4_protosw_exit(); 1560 1561 unregister_pernet_subsys(&sctp_defaults_ops); 1562 1563 /* Unregister with socket layer. */ 1564 sctp_v6_pf_exit(); 1565 sctp_v4_pf_exit(); 1566 1567 sctp_sysctl_unregister(); 1568 1569 free_pages((unsigned long)sctp_assoc_hashtable, 1570 get_order(sctp_assoc_hashsize * 1571 sizeof(struct sctp_hashbucket))); 1572 kfree(sctp_ep_hashtable); 1573 free_pages((unsigned long)sctp_port_hashtable, 1574 get_order(sctp_port_hashsize * 1575 sizeof(struct sctp_bind_hashbucket))); 1576 1577 percpu_counter_destroy(&sctp_sockets_allocated); 1578 1579 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1580 1581 kmem_cache_destroy(sctp_chunk_cachep); 1582 kmem_cache_destroy(sctp_bucket_cachep); 1583} 1584 1585module_init(sctp_init); 1586module_exit(sctp_exit); 1587 1588/* 1589 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1590 */ 1591MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1592MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1593MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1594MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1595module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1596MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1597MODULE_LICENSE("GPL"); 1598