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/* Global data structures. */ 64struct sctp_globals sctp_globals __read_mostly; 65 66struct idr sctp_assocs_id; 67DEFINE_SPINLOCK(sctp_assocs_id_lock); 68 69static struct sctp_pf *sctp_pf_inet6_specific; 70static struct sctp_pf *sctp_pf_inet_specific; 71static struct sctp_af *sctp_af_v4_specific; 72static struct sctp_af *sctp_af_v6_specific; 73 74struct kmem_cache *sctp_chunk_cachep __read_mostly; 75struct kmem_cache *sctp_bucket_cachep __read_mostly; 76 77long sysctl_sctp_mem[3]; 78int sysctl_sctp_rmem[3]; 79int sysctl_sctp_wmem[3]; 80 81/* Set up the proc fs entry for the SCTP protocol. */ 82static int __net_init sctp_proc_init(struct net *net) 83{ 84#ifdef CONFIG_PROC_FS 85 net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net); 86 if (!net->sctp.proc_net_sctp) 87 goto out_proc_net_sctp; 88 if (sctp_snmp_proc_init(net)) 89 goto out_snmp_proc_init; 90 if (sctp_eps_proc_init(net)) 91 goto out_eps_proc_init; 92 if (sctp_assocs_proc_init(net)) 93 goto out_assocs_proc_init; 94 if (sctp_remaddr_proc_init(net)) 95 goto out_remaddr_proc_init; 96 97 return 0; 98 99out_remaddr_proc_init: 100 sctp_assocs_proc_exit(net); 101out_assocs_proc_init: 102 sctp_eps_proc_exit(net); 103out_eps_proc_init: 104 sctp_snmp_proc_exit(net); 105out_snmp_proc_init: 106 remove_proc_entry("sctp", net->proc_net); 107 net->sctp.proc_net_sctp = NULL; 108out_proc_net_sctp: 109 return -ENOMEM; 110#endif /* CONFIG_PROC_FS */ 111 return 0; 112} 113 114/* Clean up the proc fs entry for the SCTP protocol. 115 * Note: Do not make this __exit as it is used in the init error 116 * path. 117 */ 118static void sctp_proc_exit(struct net *net) 119{ 120#ifdef CONFIG_PROC_FS 121 sctp_snmp_proc_exit(net); 122 sctp_eps_proc_exit(net); 123 sctp_assocs_proc_exit(net); 124 sctp_remaddr_proc_exit(net); 125 126 remove_proc_entry("sctp", net->proc_net); 127 net->sctp.proc_net_sctp = NULL; 128#endif 129} 130 131/* Private helper to extract ipv4 address and stash them in 132 * the protocol structure. 133 */ 134static void sctp_v4_copy_addrlist(struct list_head *addrlist, 135 struct net_device *dev) 136{ 137 struct in_device *in_dev; 138 struct in_ifaddr *ifa; 139 struct sctp_sockaddr_entry *addr; 140 141 rcu_read_lock(); 142 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 143 rcu_read_unlock(); 144 return; 145 } 146 147 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { 148 /* Add the address to the local list. */ 149 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 150 if (addr) { 151 addr->a.v4.sin_family = AF_INET; 152 addr->a.v4.sin_port = 0; 153 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 154 addr->valid = 1; 155 INIT_LIST_HEAD(&addr->list); 156 list_add_tail(&addr->list, addrlist); 157 } 158 } 159 160 rcu_read_unlock(); 161} 162 163/* Extract our IP addresses from the system and stash them in the 164 * protocol structure. 165 */ 166static void sctp_get_local_addr_list(struct net *net) 167{ 168 struct net_device *dev; 169 struct list_head *pos; 170 struct sctp_af *af; 171 172 rcu_read_lock(); 173 for_each_netdev_rcu(net, dev) { 174 list_for_each(pos, &sctp_address_families) { 175 af = list_entry(pos, struct sctp_af, list); 176 af->copy_addrlist(&net->sctp.local_addr_list, dev); 177 } 178 } 179 rcu_read_unlock(); 180} 181 182/* Free the existing local addresses. */ 183static void sctp_free_local_addr_list(struct net *net) 184{ 185 struct sctp_sockaddr_entry *addr; 186 struct list_head *pos, *temp; 187 188 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 189 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 190 list_del(pos); 191 kfree(addr); 192 } 193} 194 195/* Copy the local addresses which are valid for 'scope' into 'bp'. */ 196int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 197 sctp_scope_t scope, gfp_t gfp, int copy_flags) 198{ 199 struct sctp_sockaddr_entry *addr; 200 int error = 0; 201 202 rcu_read_lock(); 203 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 204 if (!addr->valid) 205 continue; 206 if (sctp_in_scope(net, &addr->a, scope)) { 207 /* Now that the address is in scope, check to see if 208 * the address type is really supported by the local 209 * sock as well as the remote peer. 210 */ 211 if ((((AF_INET == addr->a.sa.sa_family) && 212 (copy_flags & SCTP_ADDR4_PEERSUPP))) || 213 (((AF_INET6 == addr->a.sa.sa_family) && 214 (copy_flags & SCTP_ADDR6_ALLOWED) && 215 (copy_flags & SCTP_ADDR6_PEERSUPP)))) { 216 error = sctp_add_bind_addr(bp, &addr->a, 217 SCTP_ADDR_SRC, GFP_ATOMIC); 218 if (error) 219 goto end_copy; 220 } 221 } 222 } 223 224end_copy: 225 rcu_read_unlock(); 226 return error; 227} 228 229/* Initialize a sctp_addr from in incoming skb. */ 230static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 231 int is_saddr) 232{ 233 void *from; 234 __be16 *port; 235 struct sctphdr *sh; 236 237 port = &addr->v4.sin_port; 238 addr->v4.sin_family = AF_INET; 239 240 sh = sctp_hdr(skb); 241 if (is_saddr) { 242 *port = sh->source; 243 from = &ip_hdr(skb)->saddr; 244 } else { 245 *port = sh->dest; 246 from = &ip_hdr(skb)->daddr; 247 } 248 memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr)); 249} 250 251/* Initialize an sctp_addr from a socket. */ 252static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 253{ 254 addr->v4.sin_family = AF_INET; 255 addr->v4.sin_port = 0; 256 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 257} 258 259/* Initialize sk->sk_rcv_saddr from sctp_addr. */ 260static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 261{ 262 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 263} 264 265/* Initialize sk->sk_daddr from sctp_addr. */ 266static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 267{ 268 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 269} 270 271/* Initialize a sctp_addr from an address parameter. */ 272static void sctp_v4_from_addr_param(union sctp_addr *addr, 273 union sctp_addr_param *param, 274 __be16 port, int iif) 275{ 276 addr->v4.sin_family = AF_INET; 277 addr->v4.sin_port = port; 278 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 279} 280 281/* Initialize an address parameter from a sctp_addr and return the length 282 * of the address parameter. 283 */ 284static int sctp_v4_to_addr_param(const union sctp_addr *addr, 285 union sctp_addr_param *param) 286{ 287 int length = sizeof(sctp_ipv4addr_param_t); 288 289 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 290 param->v4.param_hdr.length = htons(length); 291 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 292 293 return length; 294} 295 296/* Initialize a sctp_addr from a dst_entry. */ 297static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 298 __be16 port) 299{ 300 saddr->v4.sin_family = AF_INET; 301 saddr->v4.sin_port = port; 302 saddr->v4.sin_addr.s_addr = fl4->saddr; 303} 304 305/* Compare two addresses exactly. */ 306static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 307 const union sctp_addr *addr2) 308{ 309 if (addr1->sa.sa_family != addr2->sa.sa_family) 310 return 0; 311 if (addr1->v4.sin_port != addr2->v4.sin_port) 312 return 0; 313 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 314 return 0; 315 316 return 1; 317} 318 319/* Initialize addr struct to INADDR_ANY. */ 320static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 321{ 322 addr->v4.sin_family = AF_INET; 323 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 324 addr->v4.sin_port = port; 325} 326 327/* Is this a wildcard address? */ 328static int sctp_v4_is_any(const union sctp_addr *addr) 329{ 330 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 331} 332 333/* This function checks if the address is a valid address to be used for 334 * SCTP binding. 335 * 336 * Output: 337 * Return 0 - If the address is a non-unicast or an illegal address. 338 * Return 1 - If the address is a unicast. 339 */ 340static int sctp_v4_addr_valid(union sctp_addr *addr, 341 struct sctp_sock *sp, 342 const struct sk_buff *skb) 343{ 344 /* IPv4 addresses not allowed */ 345 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 346 return 0; 347 348 /* Is this a non-unicast address or a unusable SCTP address? */ 349 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 350 return 0; 351 352 /* Is this a broadcast address? */ 353 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 354 return 0; 355 356 return 1; 357} 358 359/* Should this be available for binding? */ 360static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 361{ 362 struct net *net = sock_net(&sp->inet.sk); 363 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr); 364 365 366 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 367 ret != RTN_LOCAL && 368 !sp->inet.freebind && 369 !net->ipv4.sysctl_ip_nonlocal_bind) 370 return 0; 371 372 if (ipv6_only_sock(sctp_opt2sk(sp))) 373 return 0; 374 375 return 1; 376} 377 378/* Checking the loopback, private and other address scopes as defined in 379 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 380 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 381 * 382 * Level 0 - unusable SCTP addresses 383 * Level 1 - loopback address 384 * Level 2 - link-local addresses 385 * Level 3 - private addresses. 386 * Level 4 - global addresses 387 * For INIT and INIT-ACK address list, let L be the level of 388 * of requested destination address, sender and receiver 389 * SHOULD include all of its addresses with level greater 390 * than or equal to L. 391 * 392 * IPv4 scoping can be controlled through sysctl option 393 * net.sctp.addr_scope_policy 394 */ 395static sctp_scope_t sctp_v4_scope(union sctp_addr *addr) 396{ 397 sctp_scope_t retval; 398 399 /* Check for unusable SCTP addresses. */ 400 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 401 retval = SCTP_SCOPE_UNUSABLE; 402 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 403 retval = SCTP_SCOPE_LOOPBACK; 404 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 405 retval = SCTP_SCOPE_LINK; 406 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 407 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 408 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) { 409 retval = SCTP_SCOPE_PRIVATE; 410 } else { 411 retval = SCTP_SCOPE_GLOBAL; 412 } 413 414 return retval; 415} 416 417/* Returns a valid dst cache entry for the given source and destination ip 418 * addresses. If an association is passed, trys to get a dst entry with a 419 * source address that matches an address in the bind address list. 420 */ 421static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 422 struct flowi *fl, struct sock *sk) 423{ 424 struct sctp_association *asoc = t->asoc; 425 struct rtable *rt; 426 struct flowi4 *fl4 = &fl->u.ip4; 427 struct sctp_bind_addr *bp; 428 struct sctp_sockaddr_entry *laddr; 429 struct dst_entry *dst = NULL; 430 union sctp_addr *daddr = &t->ipaddr; 431 union sctp_addr dst_saddr; 432 433 memset(fl4, 0x0, sizeof(struct flowi4)); 434 fl4->daddr = daddr->v4.sin_addr.s_addr; 435 fl4->fl4_dport = daddr->v4.sin_port; 436 fl4->flowi4_proto = IPPROTO_SCTP; 437 if (asoc) { 438 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk); 439 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 440 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 441 } 442 if (saddr) { 443 fl4->saddr = saddr->v4.sin_addr.s_addr; 444 fl4->fl4_sport = saddr->v4.sin_port; 445 } 446 447 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 448 &fl4->saddr); 449 450 rt = ip_route_output_key(sock_net(sk), fl4); 451 if (!IS_ERR(rt)) 452 dst = &rt->dst; 453 454 /* If there is no association or if a source address is passed, no 455 * more validation is required. 456 */ 457 if (!asoc || saddr) 458 goto out; 459 460 bp = &asoc->base.bind_addr; 461 462 if (dst) { 463 /* Walk through the bind address list and look for a bind 464 * address that matches the source address of the returned dst. 465 */ 466 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 467 rcu_read_lock(); 468 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 469 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 470 (laddr->state != SCTP_ADDR_SRC && 471 !asoc->src_out_of_asoc_ok)) 472 continue; 473 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 474 goto out_unlock; 475 } 476 rcu_read_unlock(); 477 478 /* None of the bound addresses match the source address of the 479 * dst. So release it. 480 */ 481 dst_release(dst); 482 dst = NULL; 483 } 484 485 /* Walk through the bind address list and try to get a dst that 486 * matches a bind address as the source address. 487 */ 488 rcu_read_lock(); 489 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 490 if (!laddr->valid) 491 continue; 492 if ((laddr->state == SCTP_ADDR_SRC) && 493 (AF_INET == laddr->a.sa.sa_family)) { 494 fl4->fl4_sport = laddr->a.v4.sin_port; 495 flowi4_update_output(fl4, 496 asoc->base.sk->sk_bound_dev_if, 497 RT_CONN_FLAGS(asoc->base.sk), 498 daddr->v4.sin_addr.s_addr, 499 laddr->a.v4.sin_addr.s_addr); 500 501 rt = ip_route_output_key(sock_net(sk), fl4); 502 if (!IS_ERR(rt)) { 503 dst = &rt->dst; 504 goto out_unlock; 505 } 506 } 507 } 508 509out_unlock: 510 rcu_read_unlock(); 511out: 512 t->dst = dst; 513 if (dst) 514 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 515 &fl4->daddr, &fl4->saddr); 516 else 517 pr_debug("no route\n"); 518} 519 520/* For v4, the source address is cached in the route entry(dst). So no need 521 * to cache it separately and hence this is an empty routine. 522 */ 523static void sctp_v4_get_saddr(struct sctp_sock *sk, 524 struct sctp_transport *t, 525 struct flowi *fl) 526{ 527 union sctp_addr *saddr = &t->saddr; 528 struct rtable *rt = (struct rtable *)t->dst; 529 530 if (rt) { 531 saddr->v4.sin_family = AF_INET; 532 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 533 } 534} 535 536/* What interface did this skb arrive on? */ 537static int sctp_v4_skb_iif(const struct sk_buff *skb) 538{ 539 return inet_iif(skb); 540} 541 542/* Was this packet marked by Explicit Congestion Notification? */ 543static int sctp_v4_is_ce(const struct sk_buff *skb) 544{ 545 return INET_ECN_is_ce(ip_hdr(skb)->tos); 546} 547 548/* Create and initialize a new sk for the socket returned by accept(). */ 549static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 550 struct sctp_association *asoc) 551{ 552 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 553 sk->sk_prot); 554 struct inet_sock *newinet; 555 556 if (!newsk) 557 goto out; 558 559 sock_init_data(NULL, newsk); 560 561 sctp_copy_sock(newsk, sk, asoc); 562 sock_reset_flag(newsk, SOCK_ZAPPED); 563 564 newinet = inet_sk(newsk); 565 566 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 567 568 sk_refcnt_debug_inc(newsk); 569 570 if (newsk->sk_prot->init(newsk)) { 571 sk_common_release(newsk); 572 newsk = NULL; 573 } 574 575out: 576 return newsk; 577} 578 579static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 580{ 581 /* No address mapping for V4 sockets */ 582 return sizeof(struct sockaddr_in); 583} 584 585/* Dump the v4 addr to the seq file. */ 586static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 587{ 588 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 589} 590 591static void sctp_v4_ecn_capable(struct sock *sk) 592{ 593 INET_ECN_xmit(sk); 594} 595 596static void sctp_addr_wq_timeout_handler(unsigned long arg) 597{ 598 struct net *net = (struct net *)arg; 599 struct sctp_sockaddr_entry *addrw, *temp; 600 struct sctp_sock *sp; 601 602 spin_lock_bh(&net->sctp.addr_wq_lock); 603 604 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 605 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 606 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 607 addrw->state, addrw); 608 609#if IS_ENABLED(CONFIG_IPV6) 610 /* Now we send an ASCONF for each association */ 611 /* Note. we currently don't handle link local IPv6 addressees */ 612 if (addrw->a.sa.sa_family == AF_INET6) { 613 struct in6_addr *in6; 614 615 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 616 IPV6_ADDR_LINKLOCAL) 617 goto free_next; 618 619 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 620 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 621 addrw->state == SCTP_ADDR_NEW) { 622 unsigned long timeo_val; 623 624 pr_debug("%s: this is on DAD, trying %d sec " 625 "later\n", __func__, 626 SCTP_ADDRESS_TICK_DELAY); 627 628 timeo_val = jiffies; 629 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 630 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 631 break; 632 } 633 } 634#endif 635 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 636 struct sock *sk; 637 638 sk = sctp_opt2sk(sp); 639 /* ignore bound-specific endpoints */ 640 if (!sctp_is_ep_boundall(sk)) 641 continue; 642 bh_lock_sock(sk); 643 if (sctp_asconf_mgmt(sp, addrw) < 0) 644 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 645 bh_unlock_sock(sk); 646 } 647#if IS_ENABLED(CONFIG_IPV6) 648free_next: 649#endif 650 list_del(&addrw->list); 651 kfree(addrw); 652 } 653 spin_unlock_bh(&net->sctp.addr_wq_lock); 654} 655 656static void sctp_free_addr_wq(struct net *net) 657{ 658 struct sctp_sockaddr_entry *addrw; 659 struct sctp_sockaddr_entry *temp; 660 661 spin_lock_bh(&net->sctp.addr_wq_lock); 662 del_timer(&net->sctp.addr_wq_timer); 663 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 664 list_del(&addrw->list); 665 kfree(addrw); 666 } 667 spin_unlock_bh(&net->sctp.addr_wq_lock); 668} 669 670/* lookup the entry for the same address in the addr_waitq 671 * sctp_addr_wq MUST be locked 672 */ 673static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 674 struct sctp_sockaddr_entry *addr) 675{ 676 struct sctp_sockaddr_entry *addrw; 677 678 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 679 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 680 continue; 681 if (addrw->a.sa.sa_family == AF_INET) { 682 if (addrw->a.v4.sin_addr.s_addr == 683 addr->a.v4.sin_addr.s_addr) 684 return addrw; 685 } else if (addrw->a.sa.sa_family == AF_INET6) { 686 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 687 &addr->a.v6.sin6_addr)) 688 return addrw; 689 } 690 } 691 return NULL; 692} 693 694void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 695{ 696 struct sctp_sockaddr_entry *addrw; 697 unsigned long timeo_val; 698 699 /* first, we check if an opposite message already exist in the queue. 700 * If we found such message, it is removed. 701 * This operation is a bit stupid, but the DHCP client attaches the 702 * new address after a couple of addition and deletion of that address 703 */ 704 705 spin_lock_bh(&net->sctp.addr_wq_lock); 706 /* Offsets existing events in addr_wq */ 707 addrw = sctp_addr_wq_lookup(net, addr); 708 if (addrw) { 709 if (addrw->state != cmd) { 710 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 711 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 712 &net->sctp.addr_waitq); 713 714 list_del(&addrw->list); 715 kfree(addrw); 716 } 717 spin_unlock_bh(&net->sctp.addr_wq_lock); 718 return; 719 } 720 721 /* OK, we have to add the new address to the wait queue */ 722 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 723 if (addrw == NULL) { 724 spin_unlock_bh(&net->sctp.addr_wq_lock); 725 return; 726 } 727 addrw->state = cmd; 728 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 729 730 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 731 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 732 733 if (!timer_pending(&net->sctp.addr_wq_timer)) { 734 timeo_val = jiffies; 735 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 736 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 737 } 738 spin_unlock_bh(&net->sctp.addr_wq_lock); 739} 740 741/* Event handler for inet address addition/deletion events. 742 * The sctp_local_addr_list needs to be protocted by a spin lock since 743 * multiple notifiers (say IPv4 and IPv6) may be running at the same 744 * time and thus corrupt the list. 745 * The reader side is protected with RCU. 746 */ 747static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 748 void *ptr) 749{ 750 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 751 struct sctp_sockaddr_entry *addr = NULL; 752 struct sctp_sockaddr_entry *temp; 753 struct net *net = dev_net(ifa->ifa_dev->dev); 754 int found = 0; 755 756 switch (ev) { 757 case NETDEV_UP: 758 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 759 if (addr) { 760 addr->a.v4.sin_family = AF_INET; 761 addr->a.v4.sin_port = 0; 762 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 763 addr->valid = 1; 764 spin_lock_bh(&net->sctp.local_addr_lock); 765 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 766 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 767 spin_unlock_bh(&net->sctp.local_addr_lock); 768 } 769 break; 770 case NETDEV_DOWN: 771 spin_lock_bh(&net->sctp.local_addr_lock); 772 list_for_each_entry_safe(addr, temp, 773 &net->sctp.local_addr_list, list) { 774 if (addr->a.sa.sa_family == AF_INET && 775 addr->a.v4.sin_addr.s_addr == 776 ifa->ifa_local) { 777 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 778 found = 1; 779 addr->valid = 0; 780 list_del_rcu(&addr->list); 781 break; 782 } 783 } 784 spin_unlock_bh(&net->sctp.local_addr_lock); 785 if (found) 786 kfree_rcu(addr, rcu); 787 break; 788 } 789 790 return NOTIFY_DONE; 791} 792 793/* 794 * Initialize the control inode/socket with a control endpoint data 795 * structure. This endpoint is reserved exclusively for the OOTB processing. 796 */ 797static int sctp_ctl_sock_init(struct net *net) 798{ 799 int err; 800 sa_family_t family = PF_INET; 801 802 if (sctp_get_pf_specific(PF_INET6)) 803 family = PF_INET6; 804 805 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 806 SOCK_SEQPACKET, IPPROTO_SCTP, net); 807 808 /* If IPv6 socket could not be created, try the IPv4 socket */ 809 if (err < 0 && family == PF_INET6) 810 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 811 SOCK_SEQPACKET, IPPROTO_SCTP, 812 net); 813 814 if (err < 0) { 815 pr_err("Failed to create the SCTP control socket\n"); 816 return err; 817 } 818 return 0; 819} 820 821/* Register address family specific functions. */ 822int sctp_register_af(struct sctp_af *af) 823{ 824 switch (af->sa_family) { 825 case AF_INET: 826 if (sctp_af_v4_specific) 827 return 0; 828 sctp_af_v4_specific = af; 829 break; 830 case AF_INET6: 831 if (sctp_af_v6_specific) 832 return 0; 833 sctp_af_v6_specific = af; 834 break; 835 default: 836 return 0; 837 } 838 839 INIT_LIST_HEAD(&af->list); 840 list_add_tail(&af->list, &sctp_address_families); 841 return 1; 842} 843 844/* Get the table of functions for manipulating a particular address 845 * family. 846 */ 847struct sctp_af *sctp_get_af_specific(sa_family_t family) 848{ 849 switch (family) { 850 case AF_INET: 851 return sctp_af_v4_specific; 852 case AF_INET6: 853 return sctp_af_v6_specific; 854 default: 855 return NULL; 856 } 857} 858 859/* Common code to initialize a AF_INET msg_name. */ 860static void sctp_inet_msgname(char *msgname, int *addr_len) 861{ 862 struct sockaddr_in *sin; 863 864 sin = (struct sockaddr_in *)msgname; 865 *addr_len = sizeof(struct sockaddr_in); 866 sin->sin_family = AF_INET; 867 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 868} 869 870/* Copy the primary address of the peer primary address as the msg_name. */ 871static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 872 int *addr_len) 873{ 874 struct sockaddr_in *sin, *sinfrom; 875 876 if (msgname) { 877 struct sctp_association *asoc; 878 879 asoc = event->asoc; 880 sctp_inet_msgname(msgname, addr_len); 881 sin = (struct sockaddr_in *)msgname; 882 sinfrom = &asoc->peer.primary_addr.v4; 883 sin->sin_port = htons(asoc->peer.port); 884 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 885 } 886} 887 888/* Initialize and copy out a msgname from an inbound skb. */ 889static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 890{ 891 if (msgname) { 892 struct sctphdr *sh = sctp_hdr(skb); 893 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 894 895 sctp_inet_msgname(msgname, len); 896 sin->sin_port = sh->source; 897 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 898 } 899} 900 901/* Do we support this AF? */ 902static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 903{ 904 /* PF_INET only supports AF_INET addresses. */ 905 return AF_INET == family; 906} 907 908/* Address matching with wildcards allowed. */ 909static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 910 const union sctp_addr *addr2, 911 struct sctp_sock *opt) 912{ 913 /* PF_INET only supports AF_INET addresses. */ 914 if (addr1->sa.sa_family != addr2->sa.sa_family) 915 return 0; 916 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 917 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 918 return 1; 919 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 920 return 1; 921 922 return 0; 923} 924 925/* Verify that provided sockaddr looks bindable. Common verification has 926 * already been taken care of. 927 */ 928static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 929{ 930 return sctp_v4_available(addr, opt); 931} 932 933/* Verify that sockaddr looks sendable. Common verification has already 934 * been taken care of. 935 */ 936static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 937{ 938 return 1; 939} 940 941/* Fill in Supported Address Type information for INIT and INIT-ACK 942 * chunks. Returns number of addresses supported. 943 */ 944static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 945 __be16 *types) 946{ 947 types[0] = SCTP_PARAM_IPV4_ADDRESS; 948 return 1; 949} 950 951/* Wrapper routine that calls the ip transmit routine. */ 952static inline int sctp_v4_xmit(struct sk_buff *skb, 953 struct sctp_transport *transport) 954{ 955 struct inet_sock *inet = inet_sk(skb->sk); 956 957 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 958 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr); 959 960 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ? 961 IP_PMTUDISC_DO : IP_PMTUDISC_DONT; 962 963 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS); 964 965 return ip_queue_xmit(&inet->sk, skb, &transport->fl); 966} 967 968static struct sctp_af sctp_af_inet; 969 970static struct sctp_pf sctp_pf_inet = { 971 .event_msgname = sctp_inet_event_msgname, 972 .skb_msgname = sctp_inet_skb_msgname, 973 .af_supported = sctp_inet_af_supported, 974 .cmp_addr = sctp_inet_cmp_addr, 975 .bind_verify = sctp_inet_bind_verify, 976 .send_verify = sctp_inet_send_verify, 977 .supported_addrs = sctp_inet_supported_addrs, 978 .create_accept_sk = sctp_v4_create_accept_sk, 979 .addr_to_user = sctp_v4_addr_to_user, 980 .to_sk_saddr = sctp_v4_to_sk_saddr, 981 .to_sk_daddr = sctp_v4_to_sk_daddr, 982 .af = &sctp_af_inet 983}; 984 985/* Notifier for inetaddr addition/deletion events. */ 986static struct notifier_block sctp_inetaddr_notifier = { 987 .notifier_call = sctp_inetaddr_event, 988}; 989 990/* Socket operations. */ 991static const struct proto_ops inet_seqpacket_ops = { 992 .family = PF_INET, 993 .owner = THIS_MODULE, 994 .release = inet_release, /* Needs to be wrapped... */ 995 .bind = inet_bind, 996 .connect = inet_dgram_connect, 997 .socketpair = sock_no_socketpair, 998 .accept = inet_accept, 999 .getname = inet_getname, /* Semantics are different. */ 1000 .poll = sctp_poll, 1001 .ioctl = inet_ioctl, 1002 .listen = sctp_inet_listen, 1003 .shutdown = inet_shutdown, /* Looks harmless. */ 1004 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1005 .getsockopt = sock_common_getsockopt, 1006 .sendmsg = inet_sendmsg, 1007 .recvmsg = sock_common_recvmsg, 1008 .mmap = sock_no_mmap, 1009 .sendpage = sock_no_sendpage, 1010#ifdef CONFIG_COMPAT 1011 .compat_setsockopt = compat_sock_common_setsockopt, 1012 .compat_getsockopt = compat_sock_common_getsockopt, 1013#endif 1014}; 1015 1016/* Registration with AF_INET family. */ 1017static struct inet_protosw sctp_seqpacket_protosw = { 1018 .type = SOCK_SEQPACKET, 1019 .protocol = IPPROTO_SCTP, 1020 .prot = &sctp_prot, 1021 .ops = &inet_seqpacket_ops, 1022 .flags = SCTP_PROTOSW_FLAG 1023}; 1024static struct inet_protosw sctp_stream_protosw = { 1025 .type = SOCK_STREAM, 1026 .protocol = IPPROTO_SCTP, 1027 .prot = &sctp_prot, 1028 .ops = &inet_seqpacket_ops, 1029 .flags = SCTP_PROTOSW_FLAG 1030}; 1031 1032/* Register with IP layer. */ 1033static const struct net_protocol sctp_protocol = { 1034 .handler = sctp_rcv, 1035 .err_handler = sctp_v4_err, 1036 .no_policy = 1, 1037 .netns_ok = 1, 1038 .icmp_strict_tag_validation = 1, 1039}; 1040 1041/* IPv4 address related functions. */ 1042static struct sctp_af sctp_af_inet = { 1043 .sa_family = AF_INET, 1044 .sctp_xmit = sctp_v4_xmit, 1045 .setsockopt = ip_setsockopt, 1046 .getsockopt = ip_getsockopt, 1047 .get_dst = sctp_v4_get_dst, 1048 .get_saddr = sctp_v4_get_saddr, 1049 .copy_addrlist = sctp_v4_copy_addrlist, 1050 .from_skb = sctp_v4_from_skb, 1051 .from_sk = sctp_v4_from_sk, 1052 .from_addr_param = sctp_v4_from_addr_param, 1053 .to_addr_param = sctp_v4_to_addr_param, 1054 .cmp_addr = sctp_v4_cmp_addr, 1055 .addr_valid = sctp_v4_addr_valid, 1056 .inaddr_any = sctp_v4_inaddr_any, 1057 .is_any = sctp_v4_is_any, 1058 .available = sctp_v4_available, 1059 .scope = sctp_v4_scope, 1060 .skb_iif = sctp_v4_skb_iif, 1061 .is_ce = sctp_v4_is_ce, 1062 .seq_dump_addr = sctp_v4_seq_dump_addr, 1063 .ecn_capable = sctp_v4_ecn_capable, 1064 .net_header_len = sizeof(struct iphdr), 1065 .sockaddr_len = sizeof(struct sockaddr_in), 1066#ifdef CONFIG_COMPAT 1067 .compat_setsockopt = compat_ip_setsockopt, 1068 .compat_getsockopt = compat_ip_getsockopt, 1069#endif 1070}; 1071 1072struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1073{ 1074 switch (family) { 1075 case PF_INET: 1076 return sctp_pf_inet_specific; 1077 case PF_INET6: 1078 return sctp_pf_inet6_specific; 1079 default: 1080 return NULL; 1081 } 1082} 1083 1084/* Register the PF specific function table. */ 1085int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1086{ 1087 switch (family) { 1088 case PF_INET: 1089 if (sctp_pf_inet_specific) 1090 return 0; 1091 sctp_pf_inet_specific = pf; 1092 break; 1093 case PF_INET6: 1094 if (sctp_pf_inet6_specific) 1095 return 0; 1096 sctp_pf_inet6_specific = pf; 1097 break; 1098 default: 1099 return 0; 1100 } 1101 return 1; 1102} 1103 1104static inline int init_sctp_mibs(struct net *net) 1105{ 1106 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1107 if (!net->sctp.sctp_statistics) 1108 return -ENOMEM; 1109 return 0; 1110} 1111 1112static inline void cleanup_sctp_mibs(struct net *net) 1113{ 1114 free_percpu(net->sctp.sctp_statistics); 1115} 1116 1117static void sctp_v4_pf_init(void) 1118{ 1119 /* Initialize the SCTP specific PF functions. */ 1120 sctp_register_pf(&sctp_pf_inet, PF_INET); 1121 sctp_register_af(&sctp_af_inet); 1122} 1123 1124static void sctp_v4_pf_exit(void) 1125{ 1126 list_del(&sctp_af_inet.list); 1127} 1128 1129static int sctp_v4_protosw_init(void) 1130{ 1131 int rc; 1132 1133 rc = proto_register(&sctp_prot, 1); 1134 if (rc) 1135 return rc; 1136 1137 /* Register SCTP(UDP and TCP style) with socket layer. */ 1138 inet_register_protosw(&sctp_seqpacket_protosw); 1139 inet_register_protosw(&sctp_stream_protosw); 1140 1141 return 0; 1142} 1143 1144static void sctp_v4_protosw_exit(void) 1145{ 1146 inet_unregister_protosw(&sctp_stream_protosw); 1147 inet_unregister_protosw(&sctp_seqpacket_protosw); 1148 proto_unregister(&sctp_prot); 1149} 1150 1151static int sctp_v4_add_protocol(void) 1152{ 1153 /* Register notifier for inet address additions/deletions. */ 1154 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1155 1156 /* Register SCTP with inet layer. */ 1157 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1158 return -EAGAIN; 1159 1160 return 0; 1161} 1162 1163static void sctp_v4_del_protocol(void) 1164{ 1165 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1166 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1167} 1168 1169static int __net_init sctp_net_init(struct net *net) 1170{ 1171 int status; 1172 1173 /* 1174 * 14. Suggested SCTP Protocol Parameter Values 1175 */ 1176 /* The following protocol parameters are RECOMMENDED: */ 1177 /* RTO.Initial - 3 seconds */ 1178 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1179 /* RTO.Min - 1 second */ 1180 net->sctp.rto_min = SCTP_RTO_MIN; 1181 /* RTO.Max - 60 seconds */ 1182 net->sctp.rto_max = SCTP_RTO_MAX; 1183 /* RTO.Alpha - 1/8 */ 1184 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1185 /* RTO.Beta - 1/4 */ 1186 net->sctp.rto_beta = SCTP_RTO_BETA; 1187 1188 /* Valid.Cookie.Life - 60 seconds */ 1189 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1190 1191 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1192 net->sctp.cookie_preserve_enable = 1; 1193 1194 /* Default sctp sockets to use md5 as their hmac alg */ 1195#if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1196 net->sctp.sctp_hmac_alg = "md5"; 1197#elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1198 net->sctp.sctp_hmac_alg = "sha1"; 1199#else 1200 net->sctp.sctp_hmac_alg = NULL; 1201#endif 1202 1203 /* Max.Burst - 4 */ 1204 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1205 1206 /* Association.Max.Retrans - 10 attempts 1207 * Path.Max.Retrans - 5 attempts (per destination address) 1208 * Max.Init.Retransmits - 8 attempts 1209 */ 1210 net->sctp.max_retrans_association = 10; 1211 net->sctp.max_retrans_path = 5; 1212 net->sctp.max_retrans_init = 8; 1213 1214 /* Sendbuffer growth - do per-socket accounting */ 1215 net->sctp.sndbuf_policy = 0; 1216 1217 /* Rcvbuffer growth - do per-socket accounting */ 1218 net->sctp.rcvbuf_policy = 0; 1219 1220 /* HB.interval - 30 seconds */ 1221 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1222 1223 /* delayed SACK timeout */ 1224 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1225 1226 /* Disable ADDIP by default. */ 1227 net->sctp.addip_enable = 0; 1228 net->sctp.addip_noauth = 0; 1229 net->sctp.default_auto_asconf = 0; 1230 1231 /* Enable PR-SCTP by default. */ 1232 net->sctp.prsctp_enable = 1; 1233 1234 /* Disable AUTH by default. */ 1235 net->sctp.auth_enable = 0; 1236 1237 /* Set SCOPE policy to enabled */ 1238 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1239 1240 /* Set the default rwnd update threshold */ 1241 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1242 1243 /* Initialize maximum autoclose timeout. */ 1244 net->sctp.max_autoclose = INT_MAX / HZ; 1245 1246 status = sctp_sysctl_net_register(net); 1247 if (status) 1248 goto err_sysctl_register; 1249 1250 /* Allocate and initialise sctp mibs. */ 1251 status = init_sctp_mibs(net); 1252 if (status) 1253 goto err_init_mibs; 1254 1255 /* Initialize proc fs directory. */ 1256 status = sctp_proc_init(net); 1257 if (status) 1258 goto err_init_proc; 1259 1260 sctp_dbg_objcnt_init(net); 1261 1262 /* Initialize the control inode/socket for handling OOTB packets. */ 1263 if ((status = sctp_ctl_sock_init(net))) { 1264 pr_err("Failed to initialize the SCTP control sock\n"); 1265 goto err_ctl_sock_init; 1266 } 1267 1268 /* Initialize the local address list. */ 1269 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1270 spin_lock_init(&net->sctp.local_addr_lock); 1271 sctp_get_local_addr_list(net); 1272 1273 /* Initialize the address event list */ 1274 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1275 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1276 spin_lock_init(&net->sctp.addr_wq_lock); 1277 net->sctp.addr_wq_timer.expires = 0; 1278 setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 1279 (unsigned long)net); 1280 1281 return 0; 1282 1283err_ctl_sock_init: 1284 sctp_dbg_objcnt_exit(net); 1285 sctp_proc_exit(net); 1286err_init_proc: 1287 cleanup_sctp_mibs(net); 1288err_init_mibs: 1289 sctp_sysctl_net_unregister(net); 1290err_sysctl_register: 1291 return status; 1292} 1293 1294static void __net_exit sctp_net_exit(struct net *net) 1295{ 1296 /* Free the local address list */ 1297 sctp_free_addr_wq(net); 1298 sctp_free_local_addr_list(net); 1299 1300 /* Free the control endpoint. */ 1301 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1302 1303 sctp_dbg_objcnt_exit(net); 1304 1305 sctp_proc_exit(net); 1306 cleanup_sctp_mibs(net); 1307 sctp_sysctl_net_unregister(net); 1308} 1309 1310static struct pernet_operations sctp_net_ops = { 1311 .init = sctp_net_init, 1312 .exit = sctp_net_exit, 1313}; 1314 1315/* Initialize the universe into something sensible. */ 1316static __init int sctp_init(void) 1317{ 1318 int i; 1319 int status = -EINVAL; 1320 unsigned long goal; 1321 unsigned long limit; 1322 int max_share; 1323 int order; 1324 1325 BUILD_BUG_ON(sizeof(struct sctp_ulpevent) > 1326 sizeof(((struct sk_buff *) 0)->cb)); 1327 1328 /* Allocate bind_bucket and chunk caches. */ 1329 status = -ENOBUFS; 1330 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket", 1331 sizeof(struct sctp_bind_bucket), 1332 0, SLAB_HWCACHE_ALIGN, 1333 NULL); 1334 if (!sctp_bucket_cachep) 1335 goto out; 1336 1337 sctp_chunk_cachep = kmem_cache_create("sctp_chunk", 1338 sizeof(struct sctp_chunk), 1339 0, SLAB_HWCACHE_ALIGN, 1340 NULL); 1341 if (!sctp_chunk_cachep) 1342 goto err_chunk_cachep; 1343 1344 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1345 if (status) 1346 goto err_percpu_counter_init; 1347 1348 /* Implementation specific variables. */ 1349 1350 /* Initialize default stream count setup information. */ 1351 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1352 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1353 1354 /* Initialize handle used for association ids. */ 1355 idr_init(&sctp_assocs_id); 1356 1357 limit = nr_free_buffer_pages() / 8; 1358 limit = max(limit, 128UL); 1359 sysctl_sctp_mem[0] = limit / 4 * 3; 1360 sysctl_sctp_mem[1] = limit; 1361 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1362 1363 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1364 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1365 max_share = min(4UL*1024*1024, limit); 1366 1367 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */ 1368 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1369 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1370 1371 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM; 1372 sysctl_sctp_wmem[1] = 16*1024; 1373 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1374 1375 /* Size and allocate the association hash table. 1376 * The methodology is similar to that of the tcp hash tables. 1377 */ 1378 if (totalram_pages >= (128 * 1024)) 1379 goal = totalram_pages >> (22 - PAGE_SHIFT); 1380 else 1381 goal = totalram_pages >> (24 - PAGE_SHIFT); 1382 1383 for (order = 0; (1UL << order) < goal; order++) 1384 ; 1385 1386 do { 1387 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE / 1388 sizeof(struct sctp_hashbucket); 1389 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0) 1390 continue; 1391 sctp_assoc_hashtable = (struct sctp_hashbucket *) 1392 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order); 1393 } while (!sctp_assoc_hashtable && --order > 0); 1394 if (!sctp_assoc_hashtable) { 1395 pr_err("Failed association hash alloc\n"); 1396 status = -ENOMEM; 1397 goto err_ahash_alloc; 1398 } 1399 for (i = 0; i < sctp_assoc_hashsize; i++) { 1400 rwlock_init(&sctp_assoc_hashtable[i].lock); 1401 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain); 1402 } 1403 1404 /* Allocate and initialize the endpoint hash table. */ 1405 sctp_ep_hashsize = 64; 1406 sctp_ep_hashtable = 1407 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL); 1408 if (!sctp_ep_hashtable) { 1409 pr_err("Failed endpoint_hash alloc\n"); 1410 status = -ENOMEM; 1411 goto err_ehash_alloc; 1412 } 1413 for (i = 0; i < sctp_ep_hashsize; i++) { 1414 rwlock_init(&sctp_ep_hashtable[i].lock); 1415 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1416 } 1417 1418 /* Allocate and initialize the SCTP port hash table. */ 1419 do { 1420 sctp_port_hashsize = (1UL << order) * PAGE_SIZE / 1421 sizeof(struct sctp_bind_hashbucket); 1422 if ((sctp_port_hashsize > (64 * 1024)) && order > 0) 1423 continue; 1424 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1425 __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order); 1426 } while (!sctp_port_hashtable && --order > 0); 1427 if (!sctp_port_hashtable) { 1428 pr_err("Failed bind hash alloc\n"); 1429 status = -ENOMEM; 1430 goto err_bhash_alloc; 1431 } 1432 for (i = 0; i < sctp_port_hashsize; i++) { 1433 spin_lock_init(&sctp_port_hashtable[i].lock); 1434 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1435 } 1436 1437 pr_info("Hash tables configured (established %d bind %d)\n", 1438 sctp_assoc_hashsize, sctp_port_hashsize); 1439 1440 sctp_sysctl_register(); 1441 1442 INIT_LIST_HEAD(&sctp_address_families); 1443 sctp_v4_pf_init(); 1444 sctp_v6_pf_init(); 1445 1446 status = sctp_v4_protosw_init(); 1447 1448 if (status) 1449 goto err_protosw_init; 1450 1451 status = sctp_v6_protosw_init(); 1452 if (status) 1453 goto err_v6_protosw_init; 1454 1455 status = register_pernet_subsys(&sctp_net_ops); 1456 if (status) 1457 goto err_register_pernet_subsys; 1458 1459 status = sctp_v4_add_protocol(); 1460 if (status) 1461 goto err_add_protocol; 1462 1463 /* Register SCTP with inet6 layer. */ 1464 status = sctp_v6_add_protocol(); 1465 if (status) 1466 goto err_v6_add_protocol; 1467 1468out: 1469 return status; 1470err_v6_add_protocol: 1471 sctp_v4_del_protocol(); 1472err_add_protocol: 1473 unregister_pernet_subsys(&sctp_net_ops); 1474err_register_pernet_subsys: 1475 sctp_v6_protosw_exit(); 1476err_v6_protosw_init: 1477 sctp_v4_protosw_exit(); 1478err_protosw_init: 1479 sctp_v4_pf_exit(); 1480 sctp_v6_pf_exit(); 1481 sctp_sysctl_unregister(); 1482 free_pages((unsigned long)sctp_port_hashtable, 1483 get_order(sctp_port_hashsize * 1484 sizeof(struct sctp_bind_hashbucket))); 1485err_bhash_alloc: 1486 kfree(sctp_ep_hashtable); 1487err_ehash_alloc: 1488 free_pages((unsigned long)sctp_assoc_hashtable, 1489 get_order(sctp_assoc_hashsize * 1490 sizeof(struct sctp_hashbucket))); 1491err_ahash_alloc: 1492 percpu_counter_destroy(&sctp_sockets_allocated); 1493err_percpu_counter_init: 1494 kmem_cache_destroy(sctp_chunk_cachep); 1495err_chunk_cachep: 1496 kmem_cache_destroy(sctp_bucket_cachep); 1497 goto out; 1498} 1499 1500/* Exit handler for the SCTP protocol. */ 1501static __exit void sctp_exit(void) 1502{ 1503 /* BUG. This should probably do something useful like clean 1504 * up all the remaining associations and all that memory. 1505 */ 1506 1507 /* Unregister with inet6/inet layers. */ 1508 sctp_v6_del_protocol(); 1509 sctp_v4_del_protocol(); 1510 1511 unregister_pernet_subsys(&sctp_net_ops); 1512 1513 /* Free protosw registrations */ 1514 sctp_v6_protosw_exit(); 1515 sctp_v4_protosw_exit(); 1516 1517 /* Unregister with socket layer. */ 1518 sctp_v6_pf_exit(); 1519 sctp_v4_pf_exit(); 1520 1521 sctp_sysctl_unregister(); 1522 1523 free_pages((unsigned long)sctp_assoc_hashtable, 1524 get_order(sctp_assoc_hashsize * 1525 sizeof(struct sctp_hashbucket))); 1526 kfree(sctp_ep_hashtable); 1527 free_pages((unsigned long)sctp_port_hashtable, 1528 get_order(sctp_port_hashsize * 1529 sizeof(struct sctp_bind_hashbucket))); 1530 1531 percpu_counter_destroy(&sctp_sockets_allocated); 1532 1533 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1534 1535 kmem_cache_destroy(sctp_chunk_cachep); 1536 kmem_cache_destroy(sctp_bucket_cachep); 1537} 1538 1539module_init(sctp_init); 1540module_exit(sctp_exit); 1541 1542/* 1543 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1544 */ 1545MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1546MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1547MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1548MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1549module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1550MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1551MODULE_LICENSE("GPL"); 1552