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. */ 66 struct sctp_globals sctp_globals __read_mostly; 67 68 struct idr sctp_assocs_id; 69 DEFINE_SPINLOCK(sctp_assocs_id_lock); 70 71 static struct sctp_pf *sctp_pf_inet6_specific; 72 static struct sctp_pf *sctp_pf_inet_specific; 73 static struct sctp_af *sctp_af_v4_specific; 74 static struct sctp_af *sctp_af_v6_specific; 75 76 struct kmem_cache *sctp_chunk_cachep __read_mostly; 77 struct kmem_cache *sctp_bucket_cachep __read_mostly; 78 79 long sysctl_sctp_mem[3]; 80 int sysctl_sctp_rmem[3]; 81 int sysctl_sctp_wmem[3]; 82 83 /* Private helper to extract ipv4 address and stash them in 84 * the protocol structure. 85 */ 86 static void sctp_v4_copy_addrlist(struct list_head *addrlist, 87 struct net_device *dev) 88 { 89 struct in_device *in_dev; 90 struct in_ifaddr *ifa; 91 struct sctp_sockaddr_entry *addr; 92 93 rcu_read_lock(); 94 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 95 rcu_read_unlock(); 96 return; 97 } 98 99 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { 100 /* Add the address to the local list. */ 101 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 102 if (addr) { 103 addr->a.v4.sin_family = AF_INET; 104 addr->a.v4.sin_port = 0; 105 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 106 addr->valid = 1; 107 INIT_LIST_HEAD(&addr->list); 108 list_add_tail(&addr->list, addrlist); 109 } 110 } 111 112 rcu_read_unlock(); 113 } 114 115 /* Extract our IP addresses from the system and stash them in the 116 * protocol structure. 117 */ 118 static void sctp_get_local_addr_list(struct net *net) 119 { 120 struct net_device *dev; 121 struct list_head *pos; 122 struct sctp_af *af; 123 124 rcu_read_lock(); 125 for_each_netdev_rcu(net, dev) { 126 list_for_each(pos, &sctp_address_families) { 127 af = list_entry(pos, struct sctp_af, list); 128 af->copy_addrlist(&net->sctp.local_addr_list, dev); 129 } 130 } 131 rcu_read_unlock(); 132 } 133 134 /* Free the existing local addresses. */ 135 static void sctp_free_local_addr_list(struct net *net) 136 { 137 struct sctp_sockaddr_entry *addr; 138 struct list_head *pos, *temp; 139 140 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 141 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 142 list_del(pos); 143 kfree(addr); 144 } 145 } 146 147 /* Copy the local addresses which are valid for 'scope' into 'bp'. */ 148 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 149 enum sctp_scope scope, gfp_t gfp, int copy_flags) 150 { 151 struct sctp_sockaddr_entry *addr; 152 union sctp_addr laddr; 153 int error = 0; 154 155 rcu_read_lock(); 156 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 157 if (!addr->valid) 158 continue; 159 if (!sctp_in_scope(net, &addr->a, scope)) 160 continue; 161 162 /* Now that the address is in scope, check to see if 163 * the address type is really supported by the local 164 * sock as well as the remote peer. 165 */ 166 if (addr->a.sa.sa_family == AF_INET && 167 !(copy_flags & SCTP_ADDR4_PEERSUPP)) 168 continue; 169 if (addr->a.sa.sa_family == AF_INET6 && 170 (!(copy_flags & SCTP_ADDR6_ALLOWED) || 171 !(copy_flags & SCTP_ADDR6_PEERSUPP))) 172 continue; 173 174 laddr = addr->a; 175 /* also works for setting ipv6 address port */ 176 laddr.v4.sin_port = htons(bp->port); 177 if (sctp_bind_addr_state(bp, &laddr) != -1) 178 continue; 179 180 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a), 181 SCTP_ADDR_SRC, GFP_ATOMIC); 182 if (error) 183 break; 184 } 185 186 rcu_read_unlock(); 187 return error; 188 } 189 190 /* Copy over any ip options */ 191 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk) 192 { 193 struct inet_sock *newinet, *inet = inet_sk(sk); 194 struct ip_options_rcu *inet_opt, *newopt = NULL; 195 196 newinet = inet_sk(newsk); 197 198 rcu_read_lock(); 199 inet_opt = rcu_dereference(inet->inet_opt); 200 if (inet_opt) { 201 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) + 202 inet_opt->opt.optlen, GFP_ATOMIC); 203 if (newopt) 204 memcpy(newopt, inet_opt, sizeof(*inet_opt) + 205 inet_opt->opt.optlen); 206 else 207 pr_err("%s: Failed to copy ip options\n", __func__); 208 } 209 RCU_INIT_POINTER(newinet->inet_opt, newopt); 210 rcu_read_unlock(); 211 } 212 213 /* Account for the IP options */ 214 static int sctp_v4_ip_options_len(struct sock *sk) 215 { 216 struct inet_sock *inet = inet_sk(sk); 217 struct ip_options_rcu *inet_opt; 218 int len = 0; 219 220 rcu_read_lock(); 221 inet_opt = rcu_dereference(inet->inet_opt); 222 if (inet_opt) 223 len = inet_opt->opt.optlen; 224 225 rcu_read_unlock(); 226 return len; 227 } 228 229 /* Initialize a sctp_addr from in incoming skb. */ 230 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 231 int is_saddr) 232 { 233 /* Always called on head skb, so this is safe */ 234 struct sctphdr *sh = sctp_hdr(skb); 235 struct sockaddr_in *sa = &addr->v4; 236 237 addr->v4.sin_family = AF_INET; 238 239 if (is_saddr) { 240 sa->sin_port = sh->source; 241 sa->sin_addr.s_addr = ip_hdr(skb)->saddr; 242 } else { 243 sa->sin_port = sh->dest; 244 sa->sin_addr.s_addr = ip_hdr(skb)->daddr; 245 } 246 } 247 248 /* Initialize an sctp_addr from a socket. */ 249 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 250 { 251 addr->v4.sin_family = AF_INET; 252 addr->v4.sin_port = 0; 253 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 254 } 255 256 /* Initialize sk->sk_rcv_saddr from sctp_addr. */ 257 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 258 { 259 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 260 } 261 262 /* Initialize sk->sk_daddr from sctp_addr. */ 263 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 264 { 265 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 266 } 267 268 /* Initialize a sctp_addr from an address parameter. */ 269 static void sctp_v4_from_addr_param(union sctp_addr *addr, 270 union sctp_addr_param *param, 271 __be16 port, int iif) 272 { 273 addr->v4.sin_family = AF_INET; 274 addr->v4.sin_port = port; 275 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 276 } 277 278 /* Initialize an address parameter from a sctp_addr and return the length 279 * of the address parameter. 280 */ 281 static int sctp_v4_to_addr_param(const union sctp_addr *addr, 282 union sctp_addr_param *param) 283 { 284 int length = sizeof(struct sctp_ipv4addr_param); 285 286 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 287 param->v4.param_hdr.length = htons(length); 288 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 289 290 return length; 291 } 292 293 /* Initialize a sctp_addr from a dst_entry. */ 294 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 295 __be16 port) 296 { 297 saddr->v4.sin_family = AF_INET; 298 saddr->v4.sin_port = port; 299 saddr->v4.sin_addr.s_addr = fl4->saddr; 300 } 301 302 /* Compare two addresses exactly. */ 303 static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 304 const union sctp_addr *addr2) 305 { 306 if (addr1->sa.sa_family != addr2->sa.sa_family) 307 return 0; 308 if (addr1->v4.sin_port != addr2->v4.sin_port) 309 return 0; 310 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 311 return 0; 312 313 return 1; 314 } 315 316 /* Initialize addr struct to INADDR_ANY. */ 317 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 318 { 319 addr->v4.sin_family = AF_INET; 320 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 321 addr->v4.sin_port = port; 322 } 323 324 /* Is this a wildcard address? */ 325 static int sctp_v4_is_any(const union sctp_addr *addr) 326 { 327 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 328 } 329 330 /* This function checks if the address is a valid address to be used for 331 * SCTP binding. 332 * 333 * Output: 334 * Return 0 - If the address is a non-unicast or an illegal address. 335 * Return 1 - If the address is a unicast. 336 */ 337 static int sctp_v4_addr_valid(union sctp_addr *addr, 338 struct sctp_sock *sp, 339 const struct sk_buff *skb) 340 { 341 /* IPv4 addresses not allowed */ 342 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 343 return 0; 344 345 /* Is this a non-unicast address or a unusable SCTP address? */ 346 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 347 return 0; 348 349 /* Is this a broadcast address? */ 350 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 351 return 0; 352 353 return 1; 354 } 355 356 /* Should this be available for binding? */ 357 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 358 { 359 struct net *net = sock_net(&sp->inet.sk); 360 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr); 361 362 363 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 364 ret != RTN_LOCAL && 365 !sp->inet.freebind && 366 !net->ipv4.sysctl_ip_nonlocal_bind) 367 return 0; 368 369 if (ipv6_only_sock(sctp_opt2sk(sp))) 370 return 0; 371 372 return 1; 373 } 374 375 /* Checking the loopback, private and other address scopes as defined in 376 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 377 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 378 * 379 * Level 0 - unusable SCTP addresses 380 * Level 1 - loopback address 381 * Level 2 - link-local addresses 382 * Level 3 - private addresses. 383 * Level 4 - global addresses 384 * For INIT and INIT-ACK address list, let L be the level of 385 * of requested destination address, sender and receiver 386 * SHOULD include all of its addresses with level greater 387 * than or equal to L. 388 * 389 * IPv4 scoping can be controlled through sysctl option 390 * net.sctp.addr_scope_policy 391 */ 392 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr) 393 { 394 enum sctp_scope retval; 395 396 /* Check for unusable SCTP addresses. */ 397 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 398 retval = SCTP_SCOPE_UNUSABLE; 399 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 400 retval = SCTP_SCOPE_LOOPBACK; 401 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 402 retval = SCTP_SCOPE_LINK; 403 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 404 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 405 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) { 406 retval = SCTP_SCOPE_PRIVATE; 407 } else { 408 retval = SCTP_SCOPE_GLOBAL; 409 } 410 411 return retval; 412 } 413 414 /* Returns a valid dst cache entry for the given source and destination ip 415 * addresses. If an association is passed, trys to get a dst entry with a 416 * source address that matches an address in the bind address list. 417 */ 418 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 419 struct flowi *fl, struct sock *sk) 420 { 421 struct sctp_association *asoc = t->asoc; 422 struct rtable *rt; 423 struct flowi4 *fl4 = &fl->u.ip4; 424 struct sctp_bind_addr *bp; 425 struct sctp_sockaddr_entry *laddr; 426 struct dst_entry *dst = NULL; 427 union sctp_addr *daddr = &t->ipaddr; 428 union sctp_addr dst_saddr; 429 430 memset(fl4, 0x0, sizeof(struct flowi4)); 431 fl4->daddr = daddr->v4.sin_addr.s_addr; 432 fl4->fl4_dport = daddr->v4.sin_port; 433 fl4->flowi4_proto = IPPROTO_SCTP; 434 if (asoc) { 435 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk); 436 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 437 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 438 } 439 if (saddr) { 440 fl4->saddr = saddr->v4.sin_addr.s_addr; 441 fl4->fl4_sport = saddr->v4.sin_port; 442 } 443 444 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 445 &fl4->saddr); 446 447 rt = ip_route_output_key(sock_net(sk), fl4); 448 if (!IS_ERR(rt)) 449 dst = &rt->dst; 450 451 /* If there is no association or if a source address is passed, no 452 * more validation is required. 453 */ 454 if (!asoc || saddr) 455 goto out; 456 457 bp = &asoc->base.bind_addr; 458 459 if (dst) { 460 /* Walk through the bind address list and look for a bind 461 * address that matches the source address of the returned dst. 462 */ 463 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 464 rcu_read_lock(); 465 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 466 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 467 (laddr->state != SCTP_ADDR_SRC && 468 !asoc->src_out_of_asoc_ok)) 469 continue; 470 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 471 goto out_unlock; 472 } 473 rcu_read_unlock(); 474 475 /* None of the bound addresses match the source address of the 476 * dst. So release it. 477 */ 478 dst_release(dst); 479 dst = NULL; 480 } 481 482 /* Walk through the bind address list and try to get a dst that 483 * matches a bind address as the source address. 484 */ 485 rcu_read_lock(); 486 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 487 struct net_device *odev; 488 489 if (!laddr->valid) 490 continue; 491 if (laddr->state != SCTP_ADDR_SRC || 492 AF_INET != laddr->a.sa.sa_family) 493 continue; 494 495 fl4->fl4_sport = laddr->a.v4.sin_port; 496 flowi4_update_output(fl4, 497 asoc->base.sk->sk_bound_dev_if, 498 RT_CONN_FLAGS(asoc->base.sk), 499 daddr->v4.sin_addr.s_addr, 500 laddr->a.v4.sin_addr.s_addr); 501 502 rt = ip_route_output_key(sock_net(sk), fl4); 503 if (IS_ERR(rt)) 504 continue; 505 506 /* Ensure the src address belongs to the output 507 * interface. 508 */ 509 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr, 510 false); 511 if (!odev || odev->ifindex != fl4->flowi4_oif) { 512 if (!dst) 513 dst = &rt->dst; 514 else 515 dst_release(&rt->dst); 516 continue; 517 } 518 519 dst_release(dst); 520 dst = &rt->dst; 521 break; 522 } 523 524 out_unlock: 525 rcu_read_unlock(); 526 out: 527 t->dst = dst; 528 if (dst) 529 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 530 &fl4->daddr, &fl4->saddr); 531 else 532 pr_debug("no route\n"); 533 } 534 535 /* For v4, the source address is cached in the route entry(dst). So no need 536 * to cache it separately and hence this is an empty routine. 537 */ 538 static void sctp_v4_get_saddr(struct sctp_sock *sk, 539 struct sctp_transport *t, 540 struct flowi *fl) 541 { 542 union sctp_addr *saddr = &t->saddr; 543 struct rtable *rt = (struct rtable *)t->dst; 544 545 if (rt) { 546 saddr->v4.sin_family = AF_INET; 547 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 548 } 549 } 550 551 /* What interface did this skb arrive on? */ 552 static int sctp_v4_skb_iif(const struct sk_buff *skb) 553 { 554 return inet_iif(skb); 555 } 556 557 /* Was this packet marked by Explicit Congestion Notification? */ 558 static int sctp_v4_is_ce(const struct sk_buff *skb) 559 { 560 return INET_ECN_is_ce(ip_hdr(skb)->tos); 561 } 562 563 /* Create and initialize a new sk for the socket returned by accept(). */ 564 static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 565 struct sctp_association *asoc, 566 bool kern) 567 { 568 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 569 sk->sk_prot, kern); 570 struct inet_sock *newinet; 571 572 if (!newsk) 573 goto out; 574 575 sock_init_data(NULL, newsk); 576 577 sctp_copy_sock(newsk, sk, asoc); 578 sock_reset_flag(newsk, SOCK_ZAPPED); 579 580 sctp_v4_copy_ip_options(sk, newsk); 581 582 newinet = inet_sk(newsk); 583 584 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 585 586 sk_refcnt_debug_inc(newsk); 587 588 if (newsk->sk_prot->init(newsk)) { 589 sk_common_release(newsk); 590 newsk = NULL; 591 } 592 593 out: 594 return newsk; 595 } 596 597 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 598 { 599 /* No address mapping for V4 sockets */ 600 return sizeof(struct sockaddr_in); 601 } 602 603 /* Dump the v4 addr to the seq file. */ 604 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 605 { 606 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 607 } 608 609 static void sctp_v4_ecn_capable(struct sock *sk) 610 { 611 INET_ECN_xmit(sk); 612 } 613 614 static void sctp_addr_wq_timeout_handler(struct timer_list *t) 615 { 616 struct net *net = from_timer(net, t, sctp.addr_wq_timer); 617 struct sctp_sockaddr_entry *addrw, *temp; 618 struct sctp_sock *sp; 619 620 spin_lock_bh(&net->sctp.addr_wq_lock); 621 622 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 623 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 624 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 625 addrw->state, addrw); 626 627 #if IS_ENABLED(CONFIG_IPV6) 628 /* Now we send an ASCONF for each association */ 629 /* Note. we currently don't handle link local IPv6 addressees */ 630 if (addrw->a.sa.sa_family == AF_INET6) { 631 struct in6_addr *in6; 632 633 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 634 IPV6_ADDR_LINKLOCAL) 635 goto free_next; 636 637 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 638 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 639 addrw->state == SCTP_ADDR_NEW) { 640 unsigned long timeo_val; 641 642 pr_debug("%s: this is on DAD, trying %d sec " 643 "later\n", __func__, 644 SCTP_ADDRESS_TICK_DELAY); 645 646 timeo_val = jiffies; 647 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 648 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 649 break; 650 } 651 } 652 #endif 653 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 654 struct sock *sk; 655 656 sk = sctp_opt2sk(sp); 657 /* ignore bound-specific endpoints */ 658 if (!sctp_is_ep_boundall(sk)) 659 continue; 660 bh_lock_sock(sk); 661 if (sctp_asconf_mgmt(sp, addrw) < 0) 662 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 663 bh_unlock_sock(sk); 664 } 665 #if IS_ENABLED(CONFIG_IPV6) 666 free_next: 667 #endif 668 list_del(&addrw->list); 669 kfree(addrw); 670 } 671 spin_unlock_bh(&net->sctp.addr_wq_lock); 672 } 673 674 static void sctp_free_addr_wq(struct net *net) 675 { 676 struct sctp_sockaddr_entry *addrw; 677 struct sctp_sockaddr_entry *temp; 678 679 spin_lock_bh(&net->sctp.addr_wq_lock); 680 del_timer(&net->sctp.addr_wq_timer); 681 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 682 list_del(&addrw->list); 683 kfree(addrw); 684 } 685 spin_unlock_bh(&net->sctp.addr_wq_lock); 686 } 687 688 /* lookup the entry for the same address in the addr_waitq 689 * sctp_addr_wq MUST be locked 690 */ 691 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 692 struct sctp_sockaddr_entry *addr) 693 { 694 struct sctp_sockaddr_entry *addrw; 695 696 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 697 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 698 continue; 699 if (addrw->a.sa.sa_family == AF_INET) { 700 if (addrw->a.v4.sin_addr.s_addr == 701 addr->a.v4.sin_addr.s_addr) 702 return addrw; 703 } else if (addrw->a.sa.sa_family == AF_INET6) { 704 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 705 &addr->a.v6.sin6_addr)) 706 return addrw; 707 } 708 } 709 return NULL; 710 } 711 712 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 713 { 714 struct sctp_sockaddr_entry *addrw; 715 unsigned long timeo_val; 716 717 /* first, we check if an opposite message already exist in the queue. 718 * If we found such message, it is removed. 719 * This operation is a bit stupid, but the DHCP client attaches the 720 * new address after a couple of addition and deletion of that address 721 */ 722 723 spin_lock_bh(&net->sctp.addr_wq_lock); 724 /* Offsets existing events in addr_wq */ 725 addrw = sctp_addr_wq_lookup(net, addr); 726 if (addrw) { 727 if (addrw->state != cmd) { 728 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 729 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 730 &net->sctp.addr_waitq); 731 732 list_del(&addrw->list); 733 kfree(addrw); 734 } 735 spin_unlock_bh(&net->sctp.addr_wq_lock); 736 return; 737 } 738 739 /* OK, we have to add the new address to the wait queue */ 740 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 741 if (addrw == NULL) { 742 spin_unlock_bh(&net->sctp.addr_wq_lock); 743 return; 744 } 745 addrw->state = cmd; 746 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 747 748 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 749 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 750 751 if (!timer_pending(&net->sctp.addr_wq_timer)) { 752 timeo_val = jiffies; 753 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 754 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 755 } 756 spin_unlock_bh(&net->sctp.addr_wq_lock); 757 } 758 759 /* Event handler for inet address addition/deletion events. 760 * The sctp_local_addr_list needs to be protocted by a spin lock since 761 * multiple notifiers (say IPv4 and IPv6) may be running at the same 762 * time and thus corrupt the list. 763 * The reader side is protected with RCU. 764 */ 765 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 766 void *ptr) 767 { 768 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 769 struct sctp_sockaddr_entry *addr = NULL; 770 struct sctp_sockaddr_entry *temp; 771 struct net *net = dev_net(ifa->ifa_dev->dev); 772 int found = 0; 773 774 switch (ev) { 775 case NETDEV_UP: 776 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 777 if (addr) { 778 addr->a.v4.sin_family = AF_INET; 779 addr->a.v4.sin_port = 0; 780 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 781 addr->valid = 1; 782 spin_lock_bh(&net->sctp.local_addr_lock); 783 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 784 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 785 spin_unlock_bh(&net->sctp.local_addr_lock); 786 } 787 break; 788 case NETDEV_DOWN: 789 spin_lock_bh(&net->sctp.local_addr_lock); 790 list_for_each_entry_safe(addr, temp, 791 &net->sctp.local_addr_list, list) { 792 if (addr->a.sa.sa_family == AF_INET && 793 addr->a.v4.sin_addr.s_addr == 794 ifa->ifa_local) { 795 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 796 found = 1; 797 addr->valid = 0; 798 list_del_rcu(&addr->list); 799 break; 800 } 801 } 802 spin_unlock_bh(&net->sctp.local_addr_lock); 803 if (found) 804 kfree_rcu(addr, rcu); 805 break; 806 } 807 808 return NOTIFY_DONE; 809 } 810 811 /* 812 * Initialize the control inode/socket with a control endpoint data 813 * structure. This endpoint is reserved exclusively for the OOTB processing. 814 */ 815 static int sctp_ctl_sock_init(struct net *net) 816 { 817 int err; 818 sa_family_t family = PF_INET; 819 820 if (sctp_get_pf_specific(PF_INET6)) 821 family = PF_INET6; 822 823 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 824 SOCK_SEQPACKET, IPPROTO_SCTP, net); 825 826 /* If IPv6 socket could not be created, try the IPv4 socket */ 827 if (err < 0 && family == PF_INET6) 828 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 829 SOCK_SEQPACKET, IPPROTO_SCTP, 830 net); 831 832 if (err < 0) { 833 pr_err("Failed to create the SCTP control socket\n"); 834 return err; 835 } 836 return 0; 837 } 838 839 /* Register address family specific functions. */ 840 int sctp_register_af(struct sctp_af *af) 841 { 842 switch (af->sa_family) { 843 case AF_INET: 844 if (sctp_af_v4_specific) 845 return 0; 846 sctp_af_v4_specific = af; 847 break; 848 case AF_INET6: 849 if (sctp_af_v6_specific) 850 return 0; 851 sctp_af_v6_specific = af; 852 break; 853 default: 854 return 0; 855 } 856 857 INIT_LIST_HEAD(&af->list); 858 list_add_tail(&af->list, &sctp_address_families); 859 return 1; 860 } 861 862 /* Get the table of functions for manipulating a particular address 863 * family. 864 */ 865 struct sctp_af *sctp_get_af_specific(sa_family_t family) 866 { 867 switch (family) { 868 case AF_INET: 869 return sctp_af_v4_specific; 870 case AF_INET6: 871 return sctp_af_v6_specific; 872 default: 873 return NULL; 874 } 875 } 876 877 /* Common code to initialize a AF_INET msg_name. */ 878 static void sctp_inet_msgname(char *msgname, int *addr_len) 879 { 880 struct sockaddr_in *sin; 881 882 sin = (struct sockaddr_in *)msgname; 883 *addr_len = sizeof(struct sockaddr_in); 884 sin->sin_family = AF_INET; 885 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 886 } 887 888 /* Copy the primary address of the peer primary address as the msg_name. */ 889 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 890 int *addr_len) 891 { 892 struct sockaddr_in *sin, *sinfrom; 893 894 if (msgname) { 895 struct sctp_association *asoc; 896 897 asoc = event->asoc; 898 sctp_inet_msgname(msgname, addr_len); 899 sin = (struct sockaddr_in *)msgname; 900 sinfrom = &asoc->peer.primary_addr.v4; 901 sin->sin_port = htons(asoc->peer.port); 902 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 903 } 904 } 905 906 /* Initialize and copy out a msgname from an inbound skb. */ 907 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 908 { 909 if (msgname) { 910 struct sctphdr *sh = sctp_hdr(skb); 911 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 912 913 sctp_inet_msgname(msgname, len); 914 sin->sin_port = sh->source; 915 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 916 } 917 } 918 919 /* Do we support this AF? */ 920 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 921 { 922 /* PF_INET only supports AF_INET addresses. */ 923 return AF_INET == family; 924 } 925 926 /* Address matching with wildcards allowed. */ 927 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 928 const union sctp_addr *addr2, 929 struct sctp_sock *opt) 930 { 931 /* PF_INET only supports AF_INET addresses. */ 932 if (addr1->sa.sa_family != addr2->sa.sa_family) 933 return 0; 934 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 935 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 936 return 1; 937 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 938 return 1; 939 940 return 0; 941 } 942 943 /* Verify that provided sockaddr looks bindable. Common verification has 944 * already been taken care of. 945 */ 946 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 947 { 948 return sctp_v4_available(addr, opt); 949 } 950 951 /* Verify that sockaddr looks sendable. Common verification has already 952 * been taken care of. 953 */ 954 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 955 { 956 return 1; 957 } 958 959 /* Fill in Supported Address Type information for INIT and INIT-ACK 960 * chunks. Returns number of addresses supported. 961 */ 962 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 963 __be16 *types) 964 { 965 types[0] = SCTP_PARAM_IPV4_ADDRESS; 966 return 1; 967 } 968 969 /* Wrapper routine that calls the ip transmit routine. */ 970 static inline int sctp_v4_xmit(struct sk_buff *skb, 971 struct sctp_transport *transport) 972 { 973 struct inet_sock *inet = inet_sk(skb->sk); 974 975 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 976 skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr); 977 978 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ? 979 IP_PMTUDISC_DO : IP_PMTUDISC_DONT; 980 981 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS); 982 983 return ip_queue_xmit(&inet->sk, skb, &transport->fl); 984 } 985 986 static struct sctp_af sctp_af_inet; 987 988 static struct sctp_pf sctp_pf_inet = { 989 .event_msgname = sctp_inet_event_msgname, 990 .skb_msgname = sctp_inet_skb_msgname, 991 .af_supported = sctp_inet_af_supported, 992 .cmp_addr = sctp_inet_cmp_addr, 993 .bind_verify = sctp_inet_bind_verify, 994 .send_verify = sctp_inet_send_verify, 995 .supported_addrs = sctp_inet_supported_addrs, 996 .create_accept_sk = sctp_v4_create_accept_sk, 997 .addr_to_user = sctp_v4_addr_to_user, 998 .to_sk_saddr = sctp_v4_to_sk_saddr, 999 .to_sk_daddr = sctp_v4_to_sk_daddr, 1000 .copy_ip_options = sctp_v4_copy_ip_options, 1001 .af = &sctp_af_inet 1002 }; 1003 1004 /* Notifier for inetaddr addition/deletion events. */ 1005 static struct notifier_block sctp_inetaddr_notifier = { 1006 .notifier_call = sctp_inetaddr_event, 1007 }; 1008 1009 /* Socket operations. */ 1010 static const struct proto_ops inet_seqpacket_ops = { 1011 .family = PF_INET, 1012 .owner = THIS_MODULE, 1013 .release = inet_release, /* Needs to be wrapped... */ 1014 .bind = inet_bind, 1015 .connect = sctp_inet_connect, 1016 .socketpair = sock_no_socketpair, 1017 .accept = inet_accept, 1018 .getname = inet_getname, /* Semantics are different. */ 1019 .poll_mask = sctp_poll_mask, 1020 .ioctl = inet_ioctl, 1021 .listen = sctp_inet_listen, 1022 .shutdown = inet_shutdown, /* Looks harmless. */ 1023 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1024 .getsockopt = sock_common_getsockopt, 1025 .sendmsg = inet_sendmsg, 1026 .recvmsg = inet_recvmsg, 1027 .mmap = sock_no_mmap, 1028 .sendpage = sock_no_sendpage, 1029 #ifdef CONFIG_COMPAT 1030 .compat_setsockopt = compat_sock_common_setsockopt, 1031 .compat_getsockopt = compat_sock_common_getsockopt, 1032 #endif 1033 }; 1034 1035 /* Registration with AF_INET family. */ 1036 static struct inet_protosw sctp_seqpacket_protosw = { 1037 .type = SOCK_SEQPACKET, 1038 .protocol = IPPROTO_SCTP, 1039 .prot = &sctp_prot, 1040 .ops = &inet_seqpacket_ops, 1041 .flags = SCTP_PROTOSW_FLAG 1042 }; 1043 static struct inet_protosw sctp_stream_protosw = { 1044 .type = SOCK_STREAM, 1045 .protocol = IPPROTO_SCTP, 1046 .prot = &sctp_prot, 1047 .ops = &inet_seqpacket_ops, 1048 .flags = SCTP_PROTOSW_FLAG 1049 }; 1050 1051 /* Register with IP layer. */ 1052 static const struct net_protocol sctp_protocol = { 1053 .handler = sctp_rcv, 1054 .err_handler = sctp_v4_err, 1055 .no_policy = 1, 1056 .netns_ok = 1, 1057 .icmp_strict_tag_validation = 1, 1058 }; 1059 1060 /* IPv4 address related functions. */ 1061 static struct sctp_af sctp_af_inet = { 1062 .sa_family = AF_INET, 1063 .sctp_xmit = sctp_v4_xmit, 1064 .setsockopt = ip_setsockopt, 1065 .getsockopt = ip_getsockopt, 1066 .get_dst = sctp_v4_get_dst, 1067 .get_saddr = sctp_v4_get_saddr, 1068 .copy_addrlist = sctp_v4_copy_addrlist, 1069 .from_skb = sctp_v4_from_skb, 1070 .from_sk = sctp_v4_from_sk, 1071 .from_addr_param = sctp_v4_from_addr_param, 1072 .to_addr_param = sctp_v4_to_addr_param, 1073 .cmp_addr = sctp_v4_cmp_addr, 1074 .addr_valid = sctp_v4_addr_valid, 1075 .inaddr_any = sctp_v4_inaddr_any, 1076 .is_any = sctp_v4_is_any, 1077 .available = sctp_v4_available, 1078 .scope = sctp_v4_scope, 1079 .skb_iif = sctp_v4_skb_iif, 1080 .is_ce = sctp_v4_is_ce, 1081 .seq_dump_addr = sctp_v4_seq_dump_addr, 1082 .ecn_capable = sctp_v4_ecn_capable, 1083 .net_header_len = sizeof(struct iphdr), 1084 .sockaddr_len = sizeof(struct sockaddr_in), 1085 .ip_options_len = sctp_v4_ip_options_len, 1086 #ifdef CONFIG_COMPAT 1087 .compat_setsockopt = compat_ip_setsockopt, 1088 .compat_getsockopt = compat_ip_getsockopt, 1089 #endif 1090 }; 1091 1092 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1093 { 1094 switch (family) { 1095 case PF_INET: 1096 return sctp_pf_inet_specific; 1097 case PF_INET6: 1098 return sctp_pf_inet6_specific; 1099 default: 1100 return NULL; 1101 } 1102 } 1103 1104 /* Register the PF specific function table. */ 1105 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1106 { 1107 switch (family) { 1108 case PF_INET: 1109 if (sctp_pf_inet_specific) 1110 return 0; 1111 sctp_pf_inet_specific = pf; 1112 break; 1113 case PF_INET6: 1114 if (sctp_pf_inet6_specific) 1115 return 0; 1116 sctp_pf_inet6_specific = pf; 1117 break; 1118 default: 1119 return 0; 1120 } 1121 return 1; 1122 } 1123 1124 static inline int init_sctp_mibs(struct net *net) 1125 { 1126 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1127 if (!net->sctp.sctp_statistics) 1128 return -ENOMEM; 1129 return 0; 1130 } 1131 1132 static inline void cleanup_sctp_mibs(struct net *net) 1133 { 1134 free_percpu(net->sctp.sctp_statistics); 1135 } 1136 1137 static void sctp_v4_pf_init(void) 1138 { 1139 /* Initialize the SCTP specific PF functions. */ 1140 sctp_register_pf(&sctp_pf_inet, PF_INET); 1141 sctp_register_af(&sctp_af_inet); 1142 } 1143 1144 static void sctp_v4_pf_exit(void) 1145 { 1146 list_del(&sctp_af_inet.list); 1147 } 1148 1149 static int sctp_v4_protosw_init(void) 1150 { 1151 int rc; 1152 1153 rc = proto_register(&sctp_prot, 1); 1154 if (rc) 1155 return rc; 1156 1157 /* Register SCTP(UDP and TCP style) with socket layer. */ 1158 inet_register_protosw(&sctp_seqpacket_protosw); 1159 inet_register_protosw(&sctp_stream_protosw); 1160 1161 return 0; 1162 } 1163 1164 static void sctp_v4_protosw_exit(void) 1165 { 1166 inet_unregister_protosw(&sctp_stream_protosw); 1167 inet_unregister_protosw(&sctp_seqpacket_protosw); 1168 proto_unregister(&sctp_prot); 1169 } 1170 1171 static int sctp_v4_add_protocol(void) 1172 { 1173 /* Register notifier for inet address additions/deletions. */ 1174 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1175 1176 /* Register SCTP with inet layer. */ 1177 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1178 return -EAGAIN; 1179 1180 return 0; 1181 } 1182 1183 static void sctp_v4_del_protocol(void) 1184 { 1185 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1186 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1187 } 1188 1189 static int __net_init sctp_defaults_init(struct net *net) 1190 { 1191 int status; 1192 1193 /* 1194 * 14. Suggested SCTP Protocol Parameter Values 1195 */ 1196 /* The following protocol parameters are RECOMMENDED: */ 1197 /* RTO.Initial - 3 seconds */ 1198 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1199 /* RTO.Min - 1 second */ 1200 net->sctp.rto_min = SCTP_RTO_MIN; 1201 /* RTO.Max - 60 seconds */ 1202 net->sctp.rto_max = SCTP_RTO_MAX; 1203 /* RTO.Alpha - 1/8 */ 1204 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1205 /* RTO.Beta - 1/4 */ 1206 net->sctp.rto_beta = SCTP_RTO_BETA; 1207 1208 /* Valid.Cookie.Life - 60 seconds */ 1209 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1210 1211 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1212 net->sctp.cookie_preserve_enable = 1; 1213 1214 /* Default sctp sockets to use md5 as their hmac alg */ 1215 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1216 net->sctp.sctp_hmac_alg = "md5"; 1217 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1218 net->sctp.sctp_hmac_alg = "sha1"; 1219 #else 1220 net->sctp.sctp_hmac_alg = NULL; 1221 #endif 1222 1223 /* Max.Burst - 4 */ 1224 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1225 1226 /* Enable pf state by default */ 1227 net->sctp.pf_enable = 1; 1228 1229 /* Association.Max.Retrans - 10 attempts 1230 * Path.Max.Retrans - 5 attempts (per destination address) 1231 * Max.Init.Retransmits - 8 attempts 1232 */ 1233 net->sctp.max_retrans_association = 10; 1234 net->sctp.max_retrans_path = 5; 1235 net->sctp.max_retrans_init = 8; 1236 1237 /* Sendbuffer growth - do per-socket accounting */ 1238 net->sctp.sndbuf_policy = 0; 1239 1240 /* Rcvbuffer growth - do per-socket accounting */ 1241 net->sctp.rcvbuf_policy = 0; 1242 1243 /* HB.interval - 30 seconds */ 1244 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1245 1246 /* delayed SACK timeout */ 1247 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1248 1249 /* Disable ADDIP by default. */ 1250 net->sctp.addip_enable = 0; 1251 net->sctp.addip_noauth = 0; 1252 net->sctp.default_auto_asconf = 0; 1253 1254 /* Enable PR-SCTP by default. */ 1255 net->sctp.prsctp_enable = 1; 1256 1257 /* Disable RECONF by default. */ 1258 net->sctp.reconf_enable = 0; 1259 1260 /* Disable AUTH by default. */ 1261 net->sctp.auth_enable = 0; 1262 1263 /* Set SCOPE policy to enabled */ 1264 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1265 1266 /* Set the default rwnd update threshold */ 1267 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1268 1269 /* Initialize maximum autoclose timeout. */ 1270 net->sctp.max_autoclose = INT_MAX / HZ; 1271 1272 status = sctp_sysctl_net_register(net); 1273 if (status) 1274 goto err_sysctl_register; 1275 1276 /* Allocate and initialise sctp mibs. */ 1277 status = init_sctp_mibs(net); 1278 if (status) 1279 goto err_init_mibs; 1280 1281 #ifdef CONFIG_PROC_FS 1282 /* Initialize proc fs directory. */ 1283 status = sctp_proc_init(net); 1284 if (status) 1285 goto err_init_proc; 1286 #endif 1287 1288 sctp_dbg_objcnt_init(net); 1289 1290 /* Initialize the local address list. */ 1291 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1292 spin_lock_init(&net->sctp.local_addr_lock); 1293 sctp_get_local_addr_list(net); 1294 1295 /* Initialize the address event list */ 1296 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1297 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1298 spin_lock_init(&net->sctp.addr_wq_lock); 1299 net->sctp.addr_wq_timer.expires = 0; 1300 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1301 1302 return 0; 1303 1304 #ifdef CONFIG_PROC_FS 1305 err_init_proc: 1306 cleanup_sctp_mibs(net); 1307 #endif 1308 err_init_mibs: 1309 sctp_sysctl_net_unregister(net); 1310 err_sysctl_register: 1311 return status; 1312 } 1313 1314 static void __net_exit sctp_defaults_exit(struct net *net) 1315 { 1316 /* Free the local address list */ 1317 sctp_free_addr_wq(net); 1318 sctp_free_local_addr_list(net); 1319 1320 #ifdef CONFIG_PROC_FS 1321 remove_proc_subtree("sctp", net->proc_net); 1322 net->sctp.proc_net_sctp = NULL; 1323 #endif 1324 cleanup_sctp_mibs(net); 1325 sctp_sysctl_net_unregister(net); 1326 } 1327 1328 static struct pernet_operations sctp_defaults_ops = { 1329 .init = sctp_defaults_init, 1330 .exit = sctp_defaults_exit, 1331 }; 1332 1333 static int __net_init sctp_ctrlsock_init(struct net *net) 1334 { 1335 int status; 1336 1337 /* Initialize the control inode/socket for handling OOTB packets. */ 1338 status = sctp_ctl_sock_init(net); 1339 if (status) 1340 pr_err("Failed to initialize the SCTP control sock\n"); 1341 1342 return status; 1343 } 1344 1345 static void __net_init sctp_ctrlsock_exit(struct net *net) 1346 { 1347 /* Free the control endpoint. */ 1348 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1349 } 1350 1351 static struct pernet_operations sctp_ctrlsock_ops = { 1352 .init = sctp_ctrlsock_init, 1353 .exit = sctp_ctrlsock_exit, 1354 }; 1355 1356 /* Initialize the universe into something sensible. */ 1357 static __init int sctp_init(void) 1358 { 1359 int i; 1360 int status = -EINVAL; 1361 unsigned long goal; 1362 unsigned long limit; 1363 int max_share; 1364 int order; 1365 int num_entries; 1366 int max_entry_order; 1367 1368 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1369 1370 /* Allocate bind_bucket and chunk caches. */ 1371 status = -ENOBUFS; 1372 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket", 1373 sizeof(struct sctp_bind_bucket), 1374 0, SLAB_HWCACHE_ALIGN, 1375 NULL); 1376 if (!sctp_bucket_cachep) 1377 goto out; 1378 1379 sctp_chunk_cachep = kmem_cache_create("sctp_chunk", 1380 sizeof(struct sctp_chunk), 1381 0, SLAB_HWCACHE_ALIGN, 1382 NULL); 1383 if (!sctp_chunk_cachep) 1384 goto err_chunk_cachep; 1385 1386 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1387 if (status) 1388 goto err_percpu_counter_init; 1389 1390 /* Implementation specific variables. */ 1391 1392 /* Initialize default stream count setup information. */ 1393 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1394 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1395 1396 /* Initialize handle used for association ids. */ 1397 idr_init(&sctp_assocs_id); 1398 1399 limit = nr_free_buffer_pages() / 8; 1400 limit = max(limit, 128UL); 1401 sysctl_sctp_mem[0] = limit / 4 * 3; 1402 sysctl_sctp_mem[1] = limit; 1403 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1404 1405 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1406 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1407 max_share = min(4UL*1024*1024, limit); 1408 1409 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */ 1410 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1411 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1412 1413 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM; 1414 sysctl_sctp_wmem[1] = 16*1024; 1415 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1416 1417 /* Size and allocate the association hash table. 1418 * The methodology is similar to that of the tcp hash tables. 1419 * Though not identical. Start by getting a goal size 1420 */ 1421 if (totalram_pages >= (128 * 1024)) 1422 goal = totalram_pages >> (22 - PAGE_SHIFT); 1423 else 1424 goal = totalram_pages >> (24 - PAGE_SHIFT); 1425 1426 /* Then compute the page order for said goal */ 1427 order = get_order(goal); 1428 1429 /* Now compute the required page order for the maximum sized table we 1430 * want to create 1431 */ 1432 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1433 sizeof(struct sctp_bind_hashbucket)); 1434 1435 /* Limit the page order by that maximum hash table size */ 1436 order = min(order, max_entry_order); 1437 1438 /* Allocate and initialize the endpoint hash table. */ 1439 sctp_ep_hashsize = 64; 1440 sctp_ep_hashtable = 1441 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1442 if (!sctp_ep_hashtable) { 1443 pr_err("Failed endpoint_hash alloc\n"); 1444 status = -ENOMEM; 1445 goto err_ehash_alloc; 1446 } 1447 for (i = 0; i < sctp_ep_hashsize; i++) { 1448 rwlock_init(&sctp_ep_hashtable[i].lock); 1449 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1450 } 1451 1452 /* Allocate and initialize the SCTP port hash table. 1453 * Note that order is initalized to start at the max sized 1454 * table we want to support. If we can't get that many pages 1455 * reduce the order and try again 1456 */ 1457 do { 1458 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1459 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1460 } while (!sctp_port_hashtable && --order > 0); 1461 1462 if (!sctp_port_hashtable) { 1463 pr_err("Failed bind hash alloc\n"); 1464 status = -ENOMEM; 1465 goto err_bhash_alloc; 1466 } 1467 1468 /* Now compute the number of entries that will fit in the 1469 * port hash space we allocated 1470 */ 1471 num_entries = (1UL << order) * PAGE_SIZE / 1472 sizeof(struct sctp_bind_hashbucket); 1473 1474 /* And finish by rounding it down to the nearest power of two 1475 * this wastes some memory of course, but its needed because 1476 * the hash function operates based on the assumption that 1477 * that the number of entries is a power of two 1478 */ 1479 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1480 1481 for (i = 0; i < sctp_port_hashsize; i++) { 1482 spin_lock_init(&sctp_port_hashtable[i].lock); 1483 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1484 } 1485 1486 status = sctp_transport_hashtable_init(); 1487 if (status) 1488 goto err_thash_alloc; 1489 1490 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1491 num_entries); 1492 1493 sctp_sysctl_register(); 1494 1495 INIT_LIST_HEAD(&sctp_address_families); 1496 sctp_v4_pf_init(); 1497 sctp_v6_pf_init(); 1498 sctp_sched_ops_init(); 1499 1500 status = register_pernet_subsys(&sctp_defaults_ops); 1501 if (status) 1502 goto err_register_defaults; 1503 1504 status = sctp_v4_protosw_init(); 1505 if (status) 1506 goto err_protosw_init; 1507 1508 status = sctp_v6_protosw_init(); 1509 if (status) 1510 goto err_v6_protosw_init; 1511 1512 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1513 if (status) 1514 goto err_register_ctrlsock; 1515 1516 status = sctp_v4_add_protocol(); 1517 if (status) 1518 goto err_add_protocol; 1519 1520 /* Register SCTP with inet6 layer. */ 1521 status = sctp_v6_add_protocol(); 1522 if (status) 1523 goto err_v6_add_protocol; 1524 1525 if (sctp_offload_init() < 0) 1526 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1527 1528 out: 1529 return status; 1530 err_v6_add_protocol: 1531 sctp_v4_del_protocol(); 1532 err_add_protocol: 1533 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1534 err_register_ctrlsock: 1535 sctp_v6_protosw_exit(); 1536 err_v6_protosw_init: 1537 sctp_v4_protosw_exit(); 1538 err_protosw_init: 1539 unregister_pernet_subsys(&sctp_defaults_ops); 1540 err_register_defaults: 1541 sctp_v4_pf_exit(); 1542 sctp_v6_pf_exit(); 1543 sctp_sysctl_unregister(); 1544 free_pages((unsigned long)sctp_port_hashtable, 1545 get_order(sctp_port_hashsize * 1546 sizeof(struct sctp_bind_hashbucket))); 1547 err_bhash_alloc: 1548 sctp_transport_hashtable_destroy(); 1549 err_thash_alloc: 1550 kfree(sctp_ep_hashtable); 1551 err_ehash_alloc: 1552 percpu_counter_destroy(&sctp_sockets_allocated); 1553 err_percpu_counter_init: 1554 kmem_cache_destroy(sctp_chunk_cachep); 1555 err_chunk_cachep: 1556 kmem_cache_destroy(sctp_bucket_cachep); 1557 goto out; 1558 } 1559 1560 /* Exit handler for the SCTP protocol. */ 1561 static __exit void sctp_exit(void) 1562 { 1563 /* BUG. This should probably do something useful like clean 1564 * up all the remaining associations and all that memory. 1565 */ 1566 1567 /* Unregister with inet6/inet layers. */ 1568 sctp_v6_del_protocol(); 1569 sctp_v4_del_protocol(); 1570 1571 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1572 1573 /* Free protosw registrations */ 1574 sctp_v6_protosw_exit(); 1575 sctp_v4_protosw_exit(); 1576 1577 unregister_pernet_subsys(&sctp_defaults_ops); 1578 1579 /* Unregister with socket layer. */ 1580 sctp_v6_pf_exit(); 1581 sctp_v4_pf_exit(); 1582 1583 sctp_sysctl_unregister(); 1584 1585 free_pages((unsigned long)sctp_port_hashtable, 1586 get_order(sctp_port_hashsize * 1587 sizeof(struct sctp_bind_hashbucket))); 1588 kfree(sctp_ep_hashtable); 1589 sctp_transport_hashtable_destroy(); 1590 1591 percpu_counter_destroy(&sctp_sockets_allocated); 1592 1593 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1594 1595 kmem_cache_destroy(sctp_chunk_cachep); 1596 kmem_cache_destroy(sctp_bucket_cachep); 1597 } 1598 1599 module_init(sctp_init); 1600 module_exit(sctp_exit); 1601 1602 /* 1603 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1604 */ 1605 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1606 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1607 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1608 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1609 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1610 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1611 MODULE_LICENSE("GPL"); 1612