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/memblock.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_addr.s_addr = ifa->ifa_local; 105 addr->valid = 1; 106 INIT_LIST_HEAD(&addr->list); 107 list_add_tail(&addr->list, addrlist); 108 } 109 } 110 111 rcu_read_unlock(); 112 } 113 114 /* Extract our IP addresses from the system and stash them in the 115 * protocol structure. 116 */ 117 static void sctp_get_local_addr_list(struct net *net) 118 { 119 struct net_device *dev; 120 struct list_head *pos; 121 struct sctp_af *af; 122 123 rcu_read_lock(); 124 for_each_netdev_rcu(net, dev) { 125 list_for_each(pos, &sctp_address_families) { 126 af = list_entry(pos, struct sctp_af, list); 127 af->copy_addrlist(&net->sctp.local_addr_list, dev); 128 } 129 } 130 rcu_read_unlock(); 131 } 132 133 /* Free the existing local addresses. */ 134 static void sctp_free_local_addr_list(struct net *net) 135 { 136 struct sctp_sockaddr_entry *addr; 137 struct list_head *pos, *temp; 138 139 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 140 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 141 list_del(pos); 142 kfree(addr); 143 } 144 } 145 146 /* Copy the local addresses which are valid for 'scope' into 'bp'. */ 147 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 148 enum sctp_scope scope, gfp_t gfp, int copy_flags) 149 { 150 struct sctp_sockaddr_entry *addr; 151 union sctp_addr laddr; 152 int error = 0; 153 154 rcu_read_lock(); 155 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 156 if (!addr->valid) 157 continue; 158 if (!sctp_in_scope(net, &addr->a, scope)) 159 continue; 160 161 /* Now that the address is in scope, check to see if 162 * the address type is really supported by the local 163 * sock as well as the remote peer. 164 */ 165 if (addr->a.sa.sa_family == AF_INET && 166 !(copy_flags & SCTP_ADDR4_PEERSUPP)) 167 continue; 168 if (addr->a.sa.sa_family == AF_INET6 && 169 (!(copy_flags & SCTP_ADDR6_ALLOWED) || 170 !(copy_flags & SCTP_ADDR6_PEERSUPP))) 171 continue; 172 173 laddr = addr->a; 174 /* also works for setting ipv6 address port */ 175 laddr.v4.sin_port = htons(bp->port); 176 if (sctp_bind_addr_state(bp, &laddr) != -1) 177 continue; 178 179 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a), 180 SCTP_ADDR_SRC, GFP_ATOMIC); 181 if (error) 182 break; 183 } 184 185 rcu_read_unlock(); 186 return error; 187 } 188 189 /* Copy over any ip options */ 190 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk) 191 { 192 struct inet_sock *newinet, *inet = inet_sk(sk); 193 struct ip_options_rcu *inet_opt, *newopt = NULL; 194 195 newinet = inet_sk(newsk); 196 197 rcu_read_lock(); 198 inet_opt = rcu_dereference(inet->inet_opt); 199 if (inet_opt) { 200 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) + 201 inet_opt->opt.optlen, GFP_ATOMIC); 202 if (newopt) 203 memcpy(newopt, inet_opt, sizeof(*inet_opt) + 204 inet_opt->opt.optlen); 205 else 206 pr_err("%s: Failed to copy ip options\n", __func__); 207 } 208 RCU_INIT_POINTER(newinet->inet_opt, newopt); 209 rcu_read_unlock(); 210 } 211 212 /* Account for the IP options */ 213 static int sctp_v4_ip_options_len(struct sock *sk) 214 { 215 struct inet_sock *inet = inet_sk(sk); 216 struct ip_options_rcu *inet_opt; 217 int len = 0; 218 219 rcu_read_lock(); 220 inet_opt = rcu_dereference(inet->inet_opt); 221 if (inet_opt) 222 len = inet_opt->opt.optlen; 223 224 rcu_read_unlock(); 225 return len; 226 } 227 228 /* Initialize a sctp_addr from in incoming skb. */ 229 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 230 int is_saddr) 231 { 232 /* Always called on head skb, so this is safe */ 233 struct sctphdr *sh = sctp_hdr(skb); 234 struct sockaddr_in *sa = &addr->v4; 235 236 addr->v4.sin_family = AF_INET; 237 238 if (is_saddr) { 239 sa->sin_port = sh->source; 240 sa->sin_addr.s_addr = ip_hdr(skb)->saddr; 241 } else { 242 sa->sin_port = sh->dest; 243 sa->sin_addr.s_addr = ip_hdr(skb)->daddr; 244 } 245 } 246 247 /* Initialize an sctp_addr from a socket. */ 248 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 249 { 250 addr->v4.sin_family = AF_INET; 251 addr->v4.sin_port = 0; 252 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 253 } 254 255 /* Initialize sk->sk_rcv_saddr from sctp_addr. */ 256 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 257 { 258 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 259 } 260 261 /* Initialize sk->sk_daddr from sctp_addr. */ 262 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 263 { 264 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 265 } 266 267 /* Initialize a sctp_addr from an address parameter. */ 268 static void sctp_v4_from_addr_param(union sctp_addr *addr, 269 union sctp_addr_param *param, 270 __be16 port, int iif) 271 { 272 addr->v4.sin_family = AF_INET; 273 addr->v4.sin_port = port; 274 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 275 } 276 277 /* Initialize an address parameter from a sctp_addr and return the length 278 * of the address parameter. 279 */ 280 static int sctp_v4_to_addr_param(const union sctp_addr *addr, 281 union sctp_addr_param *param) 282 { 283 int length = sizeof(struct sctp_ipv4addr_param); 284 285 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 286 param->v4.param_hdr.length = htons(length); 287 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 288 289 return length; 290 } 291 292 /* Initialize a sctp_addr from a dst_entry. */ 293 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 294 __be16 port) 295 { 296 saddr->v4.sin_family = AF_INET; 297 saddr->v4.sin_port = port; 298 saddr->v4.sin_addr.s_addr = fl4->saddr; 299 } 300 301 /* Compare two addresses exactly. */ 302 static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 303 const union sctp_addr *addr2) 304 { 305 if (addr1->sa.sa_family != addr2->sa.sa_family) 306 return 0; 307 if (addr1->v4.sin_port != addr2->v4.sin_port) 308 return 0; 309 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 310 return 0; 311 312 return 1; 313 } 314 315 /* Initialize addr struct to INADDR_ANY. */ 316 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 317 { 318 addr->v4.sin_family = AF_INET; 319 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 320 addr->v4.sin_port = port; 321 } 322 323 /* Is this a wildcard address? */ 324 static int sctp_v4_is_any(const union sctp_addr *addr) 325 { 326 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 327 } 328 329 /* This function checks if the address is a valid address to be used for 330 * SCTP binding. 331 * 332 * Output: 333 * Return 0 - If the address is a non-unicast or an illegal address. 334 * Return 1 - If the address is a unicast. 335 */ 336 static int sctp_v4_addr_valid(union sctp_addr *addr, 337 struct sctp_sock *sp, 338 const struct sk_buff *skb) 339 { 340 /* IPv4 addresses not allowed */ 341 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 342 return 0; 343 344 /* Is this a non-unicast address or a unusable SCTP address? */ 345 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 346 return 0; 347 348 /* Is this a broadcast address? */ 349 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 350 return 0; 351 352 return 1; 353 } 354 355 /* Should this be available for binding? */ 356 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 357 { 358 struct net *net = sock_net(&sp->inet.sk); 359 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr); 360 361 362 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 363 ret != RTN_LOCAL && 364 !sp->inet.freebind && 365 !net->ipv4.sysctl_ip_nonlocal_bind) 366 return 0; 367 368 if (ipv6_only_sock(sctp_opt2sk(sp))) 369 return 0; 370 371 return 1; 372 } 373 374 /* Checking the loopback, private and other address scopes as defined in 375 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 376 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 377 * 378 * Level 0 - unusable SCTP addresses 379 * Level 1 - loopback address 380 * Level 2 - link-local addresses 381 * Level 3 - private addresses. 382 * Level 4 - global addresses 383 * For INIT and INIT-ACK address list, let L be the level of 384 * of requested destination address, sender and receiver 385 * SHOULD include all of its addresses with level greater 386 * than or equal to L. 387 * 388 * IPv4 scoping can be controlled through sysctl option 389 * net.sctp.addr_scope_policy 390 */ 391 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr) 392 { 393 enum sctp_scope retval; 394 395 /* Check for unusable SCTP addresses. */ 396 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 397 retval = SCTP_SCOPE_UNUSABLE; 398 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 399 retval = SCTP_SCOPE_LOOPBACK; 400 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 401 retval = SCTP_SCOPE_LINK; 402 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 403 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 404 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) { 405 retval = SCTP_SCOPE_PRIVATE; 406 } else { 407 retval = SCTP_SCOPE_GLOBAL; 408 } 409 410 return retval; 411 } 412 413 /* Returns a valid dst cache entry for the given source and destination ip 414 * addresses. If an association is passed, trys to get a dst entry with a 415 * source address that matches an address in the bind address list. 416 */ 417 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 418 struct flowi *fl, struct sock *sk) 419 { 420 struct sctp_association *asoc = t->asoc; 421 struct rtable *rt; 422 struct flowi4 *fl4 = &fl->u.ip4; 423 struct sctp_bind_addr *bp; 424 struct sctp_sockaddr_entry *laddr; 425 struct dst_entry *dst = NULL; 426 union sctp_addr *daddr = &t->ipaddr; 427 union sctp_addr dst_saddr; 428 __u8 tos = inet_sk(sk)->tos; 429 430 if (t->dscp & SCTP_DSCP_SET_MASK) 431 tos = t->dscp & SCTP_DSCP_VAL_MASK; 432 memset(fl4, 0x0, sizeof(struct flowi4)); 433 fl4->daddr = daddr->v4.sin_addr.s_addr; 434 fl4->fl4_dport = daddr->v4.sin_port; 435 fl4->flowi4_proto = IPPROTO_SCTP; 436 if (asoc) { 437 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos); 438 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 439 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 440 } 441 if (saddr) { 442 fl4->saddr = saddr->v4.sin_addr.s_addr; 443 if (!fl4->fl4_sport) 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 struct net_device *odev; 491 492 if (!laddr->valid) 493 continue; 494 if (laddr->state != SCTP_ADDR_SRC || 495 AF_INET != laddr->a.sa.sa_family) 496 continue; 497 498 fl4->fl4_sport = laddr->a.v4.sin_port; 499 flowi4_update_output(fl4, 500 asoc->base.sk->sk_bound_dev_if, 501 RT_CONN_FLAGS_TOS(asoc->base.sk, tos), 502 daddr->v4.sin_addr.s_addr, 503 laddr->a.v4.sin_addr.s_addr); 504 505 rt = ip_route_output_key(sock_net(sk), fl4); 506 if (IS_ERR(rt)) 507 continue; 508 509 /* Ensure the src address belongs to the output 510 * interface. 511 */ 512 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr, 513 false); 514 if (!odev || odev->ifindex != fl4->flowi4_oif) { 515 if (!dst) 516 dst = &rt->dst; 517 else 518 dst_release(&rt->dst); 519 continue; 520 } 521 522 dst_release(dst); 523 dst = &rt->dst; 524 break; 525 } 526 527 out_unlock: 528 rcu_read_unlock(); 529 out: 530 t->dst = dst; 531 if (dst) 532 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 533 &fl4->daddr, &fl4->saddr); 534 else 535 pr_debug("no route\n"); 536 } 537 538 /* For v4, the source address is cached in the route entry(dst). So no need 539 * to cache it separately and hence this is an empty routine. 540 */ 541 static void sctp_v4_get_saddr(struct sctp_sock *sk, 542 struct sctp_transport *t, 543 struct flowi *fl) 544 { 545 union sctp_addr *saddr = &t->saddr; 546 struct rtable *rt = (struct rtable *)t->dst; 547 548 if (rt) { 549 saddr->v4.sin_family = AF_INET; 550 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 551 } 552 } 553 554 /* What interface did this skb arrive on? */ 555 static int sctp_v4_skb_iif(const struct sk_buff *skb) 556 { 557 return inet_iif(skb); 558 } 559 560 /* Was this packet marked by Explicit Congestion Notification? */ 561 static int sctp_v4_is_ce(const struct sk_buff *skb) 562 { 563 return INET_ECN_is_ce(ip_hdr(skb)->tos); 564 } 565 566 /* Create and initialize a new sk for the socket returned by accept(). */ 567 static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 568 struct sctp_association *asoc, 569 bool kern) 570 { 571 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 572 sk->sk_prot, kern); 573 struct inet_sock *newinet; 574 575 if (!newsk) 576 goto out; 577 578 sock_init_data(NULL, newsk); 579 580 sctp_copy_sock(newsk, sk, asoc); 581 sock_reset_flag(newsk, SOCK_ZAPPED); 582 583 sctp_v4_copy_ip_options(sk, newsk); 584 585 newinet = inet_sk(newsk); 586 587 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 588 589 sk_refcnt_debug_inc(newsk); 590 591 if (newsk->sk_prot->init(newsk)) { 592 sk_common_release(newsk); 593 newsk = NULL; 594 } 595 596 out: 597 return newsk; 598 } 599 600 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 601 { 602 /* No address mapping for V4 sockets */ 603 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 604 return sizeof(struct sockaddr_in); 605 } 606 607 /* Dump the v4 addr to the seq file. */ 608 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 609 { 610 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 611 } 612 613 static void sctp_v4_ecn_capable(struct sock *sk) 614 { 615 INET_ECN_xmit(sk); 616 } 617 618 static void sctp_addr_wq_timeout_handler(struct timer_list *t) 619 { 620 struct net *net = from_timer(net, t, sctp.addr_wq_timer); 621 struct sctp_sockaddr_entry *addrw, *temp; 622 struct sctp_sock *sp; 623 624 spin_lock_bh(&net->sctp.addr_wq_lock); 625 626 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 627 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 628 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 629 addrw->state, addrw); 630 631 #if IS_ENABLED(CONFIG_IPV6) 632 /* Now we send an ASCONF for each association */ 633 /* Note. we currently don't handle link local IPv6 addressees */ 634 if (addrw->a.sa.sa_family == AF_INET6) { 635 struct in6_addr *in6; 636 637 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 638 IPV6_ADDR_LINKLOCAL) 639 goto free_next; 640 641 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 642 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 643 addrw->state == SCTP_ADDR_NEW) { 644 unsigned long timeo_val; 645 646 pr_debug("%s: this is on DAD, trying %d sec " 647 "later\n", __func__, 648 SCTP_ADDRESS_TICK_DELAY); 649 650 timeo_val = jiffies; 651 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 652 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 653 break; 654 } 655 } 656 #endif 657 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 658 struct sock *sk; 659 660 sk = sctp_opt2sk(sp); 661 /* ignore bound-specific endpoints */ 662 if (!sctp_is_ep_boundall(sk)) 663 continue; 664 bh_lock_sock(sk); 665 if (sctp_asconf_mgmt(sp, addrw) < 0) 666 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 667 bh_unlock_sock(sk); 668 } 669 #if IS_ENABLED(CONFIG_IPV6) 670 free_next: 671 #endif 672 list_del(&addrw->list); 673 kfree(addrw); 674 } 675 spin_unlock_bh(&net->sctp.addr_wq_lock); 676 } 677 678 static void sctp_free_addr_wq(struct net *net) 679 { 680 struct sctp_sockaddr_entry *addrw; 681 struct sctp_sockaddr_entry *temp; 682 683 spin_lock_bh(&net->sctp.addr_wq_lock); 684 del_timer(&net->sctp.addr_wq_timer); 685 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 686 list_del(&addrw->list); 687 kfree(addrw); 688 } 689 spin_unlock_bh(&net->sctp.addr_wq_lock); 690 } 691 692 /* lookup the entry for the same address in the addr_waitq 693 * sctp_addr_wq MUST be locked 694 */ 695 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 696 struct sctp_sockaddr_entry *addr) 697 { 698 struct sctp_sockaddr_entry *addrw; 699 700 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 701 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 702 continue; 703 if (addrw->a.sa.sa_family == AF_INET) { 704 if (addrw->a.v4.sin_addr.s_addr == 705 addr->a.v4.sin_addr.s_addr) 706 return addrw; 707 } else if (addrw->a.sa.sa_family == AF_INET6) { 708 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 709 &addr->a.v6.sin6_addr)) 710 return addrw; 711 } 712 } 713 return NULL; 714 } 715 716 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 717 { 718 struct sctp_sockaddr_entry *addrw; 719 unsigned long timeo_val; 720 721 /* first, we check if an opposite message already exist in the queue. 722 * If we found such message, it is removed. 723 * This operation is a bit stupid, but the DHCP client attaches the 724 * new address after a couple of addition and deletion of that address 725 */ 726 727 spin_lock_bh(&net->sctp.addr_wq_lock); 728 /* Offsets existing events in addr_wq */ 729 addrw = sctp_addr_wq_lookup(net, addr); 730 if (addrw) { 731 if (addrw->state != cmd) { 732 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 733 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 734 &net->sctp.addr_waitq); 735 736 list_del(&addrw->list); 737 kfree(addrw); 738 } 739 spin_unlock_bh(&net->sctp.addr_wq_lock); 740 return; 741 } 742 743 /* OK, we have to add the new address to the wait queue */ 744 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 745 if (addrw == NULL) { 746 spin_unlock_bh(&net->sctp.addr_wq_lock); 747 return; 748 } 749 addrw->state = cmd; 750 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 751 752 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 753 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 754 755 if (!timer_pending(&net->sctp.addr_wq_timer)) { 756 timeo_val = jiffies; 757 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 758 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 759 } 760 spin_unlock_bh(&net->sctp.addr_wq_lock); 761 } 762 763 /* Event handler for inet address addition/deletion events. 764 * The sctp_local_addr_list needs to be protocted by a spin lock since 765 * multiple notifiers (say IPv4 and IPv6) may be running at the same 766 * time and thus corrupt the list. 767 * The reader side is protected with RCU. 768 */ 769 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 770 void *ptr) 771 { 772 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 773 struct sctp_sockaddr_entry *addr = NULL; 774 struct sctp_sockaddr_entry *temp; 775 struct net *net = dev_net(ifa->ifa_dev->dev); 776 int found = 0; 777 778 switch (ev) { 779 case NETDEV_UP: 780 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 781 if (addr) { 782 addr->a.v4.sin_family = AF_INET; 783 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 784 addr->valid = 1; 785 spin_lock_bh(&net->sctp.local_addr_lock); 786 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 787 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 788 spin_unlock_bh(&net->sctp.local_addr_lock); 789 } 790 break; 791 case NETDEV_DOWN: 792 spin_lock_bh(&net->sctp.local_addr_lock); 793 list_for_each_entry_safe(addr, temp, 794 &net->sctp.local_addr_list, list) { 795 if (addr->a.sa.sa_family == AF_INET && 796 addr->a.v4.sin_addr.s_addr == 797 ifa->ifa_local) { 798 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 799 found = 1; 800 addr->valid = 0; 801 list_del_rcu(&addr->list); 802 break; 803 } 804 } 805 spin_unlock_bh(&net->sctp.local_addr_lock); 806 if (found) 807 kfree_rcu(addr, rcu); 808 break; 809 } 810 811 return NOTIFY_DONE; 812 } 813 814 /* 815 * Initialize the control inode/socket with a control endpoint data 816 * structure. This endpoint is reserved exclusively for the OOTB processing. 817 */ 818 static int sctp_ctl_sock_init(struct net *net) 819 { 820 int err; 821 sa_family_t family = PF_INET; 822 823 if (sctp_get_pf_specific(PF_INET6)) 824 family = PF_INET6; 825 826 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 827 SOCK_SEQPACKET, IPPROTO_SCTP, net); 828 829 /* If IPv6 socket could not be created, try the IPv4 socket */ 830 if (err < 0 && family == PF_INET6) 831 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 832 SOCK_SEQPACKET, IPPROTO_SCTP, 833 net); 834 835 if (err < 0) { 836 pr_err("Failed to create the SCTP control socket\n"); 837 return err; 838 } 839 return 0; 840 } 841 842 /* Register address family specific functions. */ 843 int sctp_register_af(struct sctp_af *af) 844 { 845 switch (af->sa_family) { 846 case AF_INET: 847 if (sctp_af_v4_specific) 848 return 0; 849 sctp_af_v4_specific = af; 850 break; 851 case AF_INET6: 852 if (sctp_af_v6_specific) 853 return 0; 854 sctp_af_v6_specific = af; 855 break; 856 default: 857 return 0; 858 } 859 860 INIT_LIST_HEAD(&af->list); 861 list_add_tail(&af->list, &sctp_address_families); 862 return 1; 863 } 864 865 /* Get the table of functions for manipulating a particular address 866 * family. 867 */ 868 struct sctp_af *sctp_get_af_specific(sa_family_t family) 869 { 870 switch (family) { 871 case AF_INET: 872 return sctp_af_v4_specific; 873 case AF_INET6: 874 return sctp_af_v6_specific; 875 default: 876 return NULL; 877 } 878 } 879 880 /* Common code to initialize a AF_INET msg_name. */ 881 static void sctp_inet_msgname(char *msgname, int *addr_len) 882 { 883 struct sockaddr_in *sin; 884 885 sin = (struct sockaddr_in *)msgname; 886 *addr_len = sizeof(struct sockaddr_in); 887 sin->sin_family = AF_INET; 888 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 889 } 890 891 /* Copy the primary address of the peer primary address as the msg_name. */ 892 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 893 int *addr_len) 894 { 895 struct sockaddr_in *sin, *sinfrom; 896 897 if (msgname) { 898 struct sctp_association *asoc; 899 900 asoc = event->asoc; 901 sctp_inet_msgname(msgname, addr_len); 902 sin = (struct sockaddr_in *)msgname; 903 sinfrom = &asoc->peer.primary_addr.v4; 904 sin->sin_port = htons(asoc->peer.port); 905 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 906 } 907 } 908 909 /* Initialize and copy out a msgname from an inbound skb. */ 910 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 911 { 912 if (msgname) { 913 struct sctphdr *sh = sctp_hdr(skb); 914 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 915 916 sctp_inet_msgname(msgname, len); 917 sin->sin_port = sh->source; 918 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 919 } 920 } 921 922 /* Do we support this AF? */ 923 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 924 { 925 /* PF_INET only supports AF_INET addresses. */ 926 return AF_INET == family; 927 } 928 929 /* Address matching with wildcards allowed. */ 930 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 931 const union sctp_addr *addr2, 932 struct sctp_sock *opt) 933 { 934 /* PF_INET only supports AF_INET addresses. */ 935 if (addr1->sa.sa_family != addr2->sa.sa_family) 936 return 0; 937 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 938 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 939 return 1; 940 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 941 return 1; 942 943 return 0; 944 } 945 946 /* Verify that provided sockaddr looks bindable. Common verification has 947 * already been taken care of. 948 */ 949 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 950 { 951 return sctp_v4_available(addr, opt); 952 } 953 954 /* Verify that sockaddr looks sendable. Common verification has already 955 * been taken care of. 956 */ 957 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 958 { 959 return 1; 960 } 961 962 /* Fill in Supported Address Type information for INIT and INIT-ACK 963 * chunks. Returns number of addresses supported. 964 */ 965 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 966 __be16 *types) 967 { 968 types[0] = SCTP_PARAM_IPV4_ADDRESS; 969 return 1; 970 } 971 972 /* Wrapper routine that calls the ip transmit routine. */ 973 static inline int sctp_v4_xmit(struct sk_buff *skb, 974 struct sctp_transport *transport) 975 { 976 struct inet_sock *inet = inet_sk(skb->sk); 977 __u8 dscp = inet->tos; 978 979 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 980 skb->len, &transport->fl.u.ip4.saddr, 981 &transport->fl.u.ip4.daddr); 982 983 if (transport->dscp & SCTP_DSCP_SET_MASK) 984 dscp = transport->dscp & SCTP_DSCP_VAL_MASK; 985 986 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ? 987 IP_PMTUDISC_DO : IP_PMTUDISC_DONT; 988 989 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS); 990 991 return __ip_queue_xmit(&inet->sk, skb, &transport->fl, dscp); 992 } 993 994 static struct sctp_af sctp_af_inet; 995 996 static struct sctp_pf sctp_pf_inet = { 997 .event_msgname = sctp_inet_event_msgname, 998 .skb_msgname = sctp_inet_skb_msgname, 999 .af_supported = sctp_inet_af_supported, 1000 .cmp_addr = sctp_inet_cmp_addr, 1001 .bind_verify = sctp_inet_bind_verify, 1002 .send_verify = sctp_inet_send_verify, 1003 .supported_addrs = sctp_inet_supported_addrs, 1004 .create_accept_sk = sctp_v4_create_accept_sk, 1005 .addr_to_user = sctp_v4_addr_to_user, 1006 .to_sk_saddr = sctp_v4_to_sk_saddr, 1007 .to_sk_daddr = sctp_v4_to_sk_daddr, 1008 .copy_ip_options = sctp_v4_copy_ip_options, 1009 .af = &sctp_af_inet 1010 }; 1011 1012 /* Notifier for inetaddr addition/deletion events. */ 1013 static struct notifier_block sctp_inetaddr_notifier = { 1014 .notifier_call = sctp_inetaddr_event, 1015 }; 1016 1017 /* Socket operations. */ 1018 static const struct proto_ops inet_seqpacket_ops = { 1019 .family = PF_INET, 1020 .owner = THIS_MODULE, 1021 .release = inet_release, /* Needs to be wrapped... */ 1022 .bind = inet_bind, 1023 .connect = sctp_inet_connect, 1024 .socketpair = sock_no_socketpair, 1025 .accept = inet_accept, 1026 .getname = inet_getname, /* Semantics are different. */ 1027 .poll = sctp_poll, 1028 .ioctl = inet_ioctl, 1029 .gettstamp = sock_gettstamp, 1030 .listen = sctp_inet_listen, 1031 .shutdown = inet_shutdown, /* Looks harmless. */ 1032 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1033 .getsockopt = sock_common_getsockopt, 1034 .sendmsg = inet_sendmsg, 1035 .recvmsg = inet_recvmsg, 1036 .mmap = sock_no_mmap, 1037 .sendpage = sock_no_sendpage, 1038 #ifdef CONFIG_COMPAT 1039 .compat_setsockopt = compat_sock_common_setsockopt, 1040 .compat_getsockopt = compat_sock_common_getsockopt, 1041 #endif 1042 }; 1043 1044 /* Registration with AF_INET family. */ 1045 static struct inet_protosw sctp_seqpacket_protosw = { 1046 .type = SOCK_SEQPACKET, 1047 .protocol = IPPROTO_SCTP, 1048 .prot = &sctp_prot, 1049 .ops = &inet_seqpacket_ops, 1050 .flags = SCTP_PROTOSW_FLAG 1051 }; 1052 static struct inet_protosw sctp_stream_protosw = { 1053 .type = SOCK_STREAM, 1054 .protocol = IPPROTO_SCTP, 1055 .prot = &sctp_prot, 1056 .ops = &inet_seqpacket_ops, 1057 .flags = SCTP_PROTOSW_FLAG 1058 }; 1059 1060 /* Register with IP layer. */ 1061 static const struct net_protocol sctp_protocol = { 1062 .handler = sctp_rcv, 1063 .err_handler = sctp_v4_err, 1064 .no_policy = 1, 1065 .netns_ok = 1, 1066 .icmp_strict_tag_validation = 1, 1067 }; 1068 1069 /* IPv4 address related functions. */ 1070 static struct sctp_af sctp_af_inet = { 1071 .sa_family = AF_INET, 1072 .sctp_xmit = sctp_v4_xmit, 1073 .setsockopt = ip_setsockopt, 1074 .getsockopt = ip_getsockopt, 1075 .get_dst = sctp_v4_get_dst, 1076 .get_saddr = sctp_v4_get_saddr, 1077 .copy_addrlist = sctp_v4_copy_addrlist, 1078 .from_skb = sctp_v4_from_skb, 1079 .from_sk = sctp_v4_from_sk, 1080 .from_addr_param = sctp_v4_from_addr_param, 1081 .to_addr_param = sctp_v4_to_addr_param, 1082 .cmp_addr = sctp_v4_cmp_addr, 1083 .addr_valid = sctp_v4_addr_valid, 1084 .inaddr_any = sctp_v4_inaddr_any, 1085 .is_any = sctp_v4_is_any, 1086 .available = sctp_v4_available, 1087 .scope = sctp_v4_scope, 1088 .skb_iif = sctp_v4_skb_iif, 1089 .is_ce = sctp_v4_is_ce, 1090 .seq_dump_addr = sctp_v4_seq_dump_addr, 1091 .ecn_capable = sctp_v4_ecn_capable, 1092 .net_header_len = sizeof(struct iphdr), 1093 .sockaddr_len = sizeof(struct sockaddr_in), 1094 .ip_options_len = sctp_v4_ip_options_len, 1095 #ifdef CONFIG_COMPAT 1096 .compat_setsockopt = compat_ip_setsockopt, 1097 .compat_getsockopt = compat_ip_getsockopt, 1098 #endif 1099 }; 1100 1101 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1102 { 1103 switch (family) { 1104 case PF_INET: 1105 return sctp_pf_inet_specific; 1106 case PF_INET6: 1107 return sctp_pf_inet6_specific; 1108 default: 1109 return NULL; 1110 } 1111 } 1112 1113 /* Register the PF specific function table. */ 1114 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1115 { 1116 switch (family) { 1117 case PF_INET: 1118 if (sctp_pf_inet_specific) 1119 return 0; 1120 sctp_pf_inet_specific = pf; 1121 break; 1122 case PF_INET6: 1123 if (sctp_pf_inet6_specific) 1124 return 0; 1125 sctp_pf_inet6_specific = pf; 1126 break; 1127 default: 1128 return 0; 1129 } 1130 return 1; 1131 } 1132 1133 static inline int init_sctp_mibs(struct net *net) 1134 { 1135 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1136 if (!net->sctp.sctp_statistics) 1137 return -ENOMEM; 1138 return 0; 1139 } 1140 1141 static inline void cleanup_sctp_mibs(struct net *net) 1142 { 1143 free_percpu(net->sctp.sctp_statistics); 1144 } 1145 1146 static void sctp_v4_pf_init(void) 1147 { 1148 /* Initialize the SCTP specific PF functions. */ 1149 sctp_register_pf(&sctp_pf_inet, PF_INET); 1150 sctp_register_af(&sctp_af_inet); 1151 } 1152 1153 static void sctp_v4_pf_exit(void) 1154 { 1155 list_del(&sctp_af_inet.list); 1156 } 1157 1158 static int sctp_v4_protosw_init(void) 1159 { 1160 int rc; 1161 1162 rc = proto_register(&sctp_prot, 1); 1163 if (rc) 1164 return rc; 1165 1166 /* Register SCTP(UDP and TCP style) with socket layer. */ 1167 inet_register_protosw(&sctp_seqpacket_protosw); 1168 inet_register_protosw(&sctp_stream_protosw); 1169 1170 return 0; 1171 } 1172 1173 static void sctp_v4_protosw_exit(void) 1174 { 1175 inet_unregister_protosw(&sctp_stream_protosw); 1176 inet_unregister_protosw(&sctp_seqpacket_protosw); 1177 proto_unregister(&sctp_prot); 1178 } 1179 1180 static int sctp_v4_add_protocol(void) 1181 { 1182 /* Register notifier for inet address additions/deletions. */ 1183 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1184 1185 /* Register SCTP with inet layer. */ 1186 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1187 return -EAGAIN; 1188 1189 return 0; 1190 } 1191 1192 static void sctp_v4_del_protocol(void) 1193 { 1194 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1195 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1196 } 1197 1198 static int __net_init sctp_defaults_init(struct net *net) 1199 { 1200 int status; 1201 1202 /* 1203 * 14. Suggested SCTP Protocol Parameter Values 1204 */ 1205 /* The following protocol parameters are RECOMMENDED: */ 1206 /* RTO.Initial - 3 seconds */ 1207 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1208 /* RTO.Min - 1 second */ 1209 net->sctp.rto_min = SCTP_RTO_MIN; 1210 /* RTO.Max - 60 seconds */ 1211 net->sctp.rto_max = SCTP_RTO_MAX; 1212 /* RTO.Alpha - 1/8 */ 1213 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1214 /* RTO.Beta - 1/4 */ 1215 net->sctp.rto_beta = SCTP_RTO_BETA; 1216 1217 /* Valid.Cookie.Life - 60 seconds */ 1218 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1219 1220 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1221 net->sctp.cookie_preserve_enable = 1; 1222 1223 /* Default sctp sockets to use md5 as their hmac alg */ 1224 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1225 net->sctp.sctp_hmac_alg = "md5"; 1226 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1227 net->sctp.sctp_hmac_alg = "sha1"; 1228 #else 1229 net->sctp.sctp_hmac_alg = NULL; 1230 #endif 1231 1232 /* Max.Burst - 4 */ 1233 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1234 1235 /* Enable pf state by default */ 1236 net->sctp.pf_enable = 1; 1237 1238 /* Association.Max.Retrans - 10 attempts 1239 * Path.Max.Retrans - 5 attempts (per destination address) 1240 * Max.Init.Retransmits - 8 attempts 1241 */ 1242 net->sctp.max_retrans_association = 10; 1243 net->sctp.max_retrans_path = 5; 1244 net->sctp.max_retrans_init = 8; 1245 1246 /* Sendbuffer growth - do per-socket accounting */ 1247 net->sctp.sndbuf_policy = 0; 1248 1249 /* Rcvbuffer growth - do per-socket accounting */ 1250 net->sctp.rcvbuf_policy = 0; 1251 1252 /* HB.interval - 30 seconds */ 1253 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1254 1255 /* delayed SACK timeout */ 1256 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1257 1258 /* Disable ADDIP by default. */ 1259 net->sctp.addip_enable = 0; 1260 net->sctp.addip_noauth = 0; 1261 net->sctp.default_auto_asconf = 0; 1262 1263 /* Enable PR-SCTP by default. */ 1264 net->sctp.prsctp_enable = 1; 1265 1266 /* Disable RECONF by default. */ 1267 net->sctp.reconf_enable = 0; 1268 1269 /* Disable AUTH by default. */ 1270 net->sctp.auth_enable = 0; 1271 1272 /* Set SCOPE policy to enabled */ 1273 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1274 1275 /* Set the default rwnd update threshold */ 1276 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1277 1278 /* Initialize maximum autoclose timeout. */ 1279 net->sctp.max_autoclose = INT_MAX / HZ; 1280 1281 status = sctp_sysctl_net_register(net); 1282 if (status) 1283 goto err_sysctl_register; 1284 1285 /* Allocate and initialise sctp mibs. */ 1286 status = init_sctp_mibs(net); 1287 if (status) 1288 goto err_init_mibs; 1289 1290 #ifdef CONFIG_PROC_FS 1291 /* Initialize proc fs directory. */ 1292 status = sctp_proc_init(net); 1293 if (status) 1294 goto err_init_proc; 1295 #endif 1296 1297 sctp_dbg_objcnt_init(net); 1298 1299 /* Initialize the local address list. */ 1300 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1301 spin_lock_init(&net->sctp.local_addr_lock); 1302 sctp_get_local_addr_list(net); 1303 1304 /* Initialize the address event list */ 1305 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1306 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1307 spin_lock_init(&net->sctp.addr_wq_lock); 1308 net->sctp.addr_wq_timer.expires = 0; 1309 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1310 1311 return 0; 1312 1313 #ifdef CONFIG_PROC_FS 1314 err_init_proc: 1315 cleanup_sctp_mibs(net); 1316 #endif 1317 err_init_mibs: 1318 sctp_sysctl_net_unregister(net); 1319 err_sysctl_register: 1320 return status; 1321 } 1322 1323 static void __net_exit sctp_defaults_exit(struct net *net) 1324 { 1325 /* Free the local address list */ 1326 sctp_free_addr_wq(net); 1327 sctp_free_local_addr_list(net); 1328 1329 #ifdef CONFIG_PROC_FS 1330 remove_proc_subtree("sctp", net->proc_net); 1331 net->sctp.proc_net_sctp = NULL; 1332 #endif 1333 cleanup_sctp_mibs(net); 1334 sctp_sysctl_net_unregister(net); 1335 } 1336 1337 static struct pernet_operations sctp_defaults_ops = { 1338 .init = sctp_defaults_init, 1339 .exit = sctp_defaults_exit, 1340 }; 1341 1342 static int __net_init sctp_ctrlsock_init(struct net *net) 1343 { 1344 int status; 1345 1346 /* Initialize the control inode/socket for handling OOTB packets. */ 1347 status = sctp_ctl_sock_init(net); 1348 if (status) 1349 pr_err("Failed to initialize the SCTP control sock\n"); 1350 1351 return status; 1352 } 1353 1354 static void __net_init sctp_ctrlsock_exit(struct net *net) 1355 { 1356 /* Free the control endpoint. */ 1357 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1358 } 1359 1360 static struct pernet_operations sctp_ctrlsock_ops = { 1361 .init = sctp_ctrlsock_init, 1362 .exit = sctp_ctrlsock_exit, 1363 }; 1364 1365 /* Initialize the universe into something sensible. */ 1366 static __init int sctp_init(void) 1367 { 1368 int i; 1369 int status = -EINVAL; 1370 unsigned long goal; 1371 unsigned long limit; 1372 unsigned long nr_pages = totalram_pages(); 1373 int max_share; 1374 int order; 1375 int num_entries; 1376 int max_entry_order; 1377 1378 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1379 1380 /* Allocate bind_bucket and chunk caches. */ 1381 status = -ENOBUFS; 1382 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket", 1383 sizeof(struct sctp_bind_bucket), 1384 0, SLAB_HWCACHE_ALIGN, 1385 NULL); 1386 if (!sctp_bucket_cachep) 1387 goto out; 1388 1389 sctp_chunk_cachep = kmem_cache_create("sctp_chunk", 1390 sizeof(struct sctp_chunk), 1391 0, SLAB_HWCACHE_ALIGN, 1392 NULL); 1393 if (!sctp_chunk_cachep) 1394 goto err_chunk_cachep; 1395 1396 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1397 if (status) 1398 goto err_percpu_counter_init; 1399 1400 /* Implementation specific variables. */ 1401 1402 /* Initialize default stream count setup information. */ 1403 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1404 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1405 1406 /* Initialize handle used for association ids. */ 1407 idr_init(&sctp_assocs_id); 1408 1409 limit = nr_free_buffer_pages() / 8; 1410 limit = max(limit, 128UL); 1411 sysctl_sctp_mem[0] = limit / 4 * 3; 1412 sysctl_sctp_mem[1] = limit; 1413 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1414 1415 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1416 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1417 max_share = min(4UL*1024*1024, limit); 1418 1419 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */ 1420 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1421 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1422 1423 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM; 1424 sysctl_sctp_wmem[1] = 16*1024; 1425 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1426 1427 /* Size and allocate the association hash table. 1428 * The methodology is similar to that of the tcp hash tables. 1429 * Though not identical. Start by getting a goal size 1430 */ 1431 if (nr_pages >= (128 * 1024)) 1432 goal = nr_pages >> (22 - PAGE_SHIFT); 1433 else 1434 goal = nr_pages >> (24 - PAGE_SHIFT); 1435 1436 /* Then compute the page order for said goal */ 1437 order = get_order(goal); 1438 1439 /* Now compute the required page order for the maximum sized table we 1440 * want to create 1441 */ 1442 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1443 sizeof(struct sctp_bind_hashbucket)); 1444 1445 /* Limit the page order by that maximum hash table size */ 1446 order = min(order, max_entry_order); 1447 1448 /* Allocate and initialize the endpoint hash table. */ 1449 sctp_ep_hashsize = 64; 1450 sctp_ep_hashtable = 1451 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1452 if (!sctp_ep_hashtable) { 1453 pr_err("Failed endpoint_hash alloc\n"); 1454 status = -ENOMEM; 1455 goto err_ehash_alloc; 1456 } 1457 for (i = 0; i < sctp_ep_hashsize; i++) { 1458 rwlock_init(&sctp_ep_hashtable[i].lock); 1459 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1460 } 1461 1462 /* Allocate and initialize the SCTP port hash table. 1463 * Note that order is initalized to start at the max sized 1464 * table we want to support. If we can't get that many pages 1465 * reduce the order and try again 1466 */ 1467 do { 1468 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1469 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1470 } while (!sctp_port_hashtable && --order > 0); 1471 1472 if (!sctp_port_hashtable) { 1473 pr_err("Failed bind hash alloc\n"); 1474 status = -ENOMEM; 1475 goto err_bhash_alloc; 1476 } 1477 1478 /* Now compute the number of entries that will fit in the 1479 * port hash space we allocated 1480 */ 1481 num_entries = (1UL << order) * PAGE_SIZE / 1482 sizeof(struct sctp_bind_hashbucket); 1483 1484 /* And finish by rounding it down to the nearest power of two 1485 * this wastes some memory of course, but its needed because 1486 * the hash function operates based on the assumption that 1487 * that the number of entries is a power of two 1488 */ 1489 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1490 1491 for (i = 0; i < sctp_port_hashsize; i++) { 1492 spin_lock_init(&sctp_port_hashtable[i].lock); 1493 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1494 } 1495 1496 status = sctp_transport_hashtable_init(); 1497 if (status) 1498 goto err_thash_alloc; 1499 1500 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1501 num_entries); 1502 1503 sctp_sysctl_register(); 1504 1505 INIT_LIST_HEAD(&sctp_address_families); 1506 sctp_v4_pf_init(); 1507 sctp_v6_pf_init(); 1508 sctp_sched_ops_init(); 1509 1510 status = register_pernet_subsys(&sctp_defaults_ops); 1511 if (status) 1512 goto err_register_defaults; 1513 1514 status = sctp_v4_protosw_init(); 1515 if (status) 1516 goto err_protosw_init; 1517 1518 status = sctp_v6_protosw_init(); 1519 if (status) 1520 goto err_v6_protosw_init; 1521 1522 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1523 if (status) 1524 goto err_register_ctrlsock; 1525 1526 status = sctp_v4_add_protocol(); 1527 if (status) 1528 goto err_add_protocol; 1529 1530 /* Register SCTP with inet6 layer. */ 1531 status = sctp_v6_add_protocol(); 1532 if (status) 1533 goto err_v6_add_protocol; 1534 1535 if (sctp_offload_init() < 0) 1536 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1537 1538 out: 1539 return status; 1540 err_v6_add_protocol: 1541 sctp_v4_del_protocol(); 1542 err_add_protocol: 1543 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1544 err_register_ctrlsock: 1545 sctp_v6_protosw_exit(); 1546 err_v6_protosw_init: 1547 sctp_v4_protosw_exit(); 1548 err_protosw_init: 1549 unregister_pernet_subsys(&sctp_defaults_ops); 1550 err_register_defaults: 1551 sctp_v4_pf_exit(); 1552 sctp_v6_pf_exit(); 1553 sctp_sysctl_unregister(); 1554 free_pages((unsigned long)sctp_port_hashtable, 1555 get_order(sctp_port_hashsize * 1556 sizeof(struct sctp_bind_hashbucket))); 1557 err_bhash_alloc: 1558 sctp_transport_hashtable_destroy(); 1559 err_thash_alloc: 1560 kfree(sctp_ep_hashtable); 1561 err_ehash_alloc: 1562 percpu_counter_destroy(&sctp_sockets_allocated); 1563 err_percpu_counter_init: 1564 kmem_cache_destroy(sctp_chunk_cachep); 1565 err_chunk_cachep: 1566 kmem_cache_destroy(sctp_bucket_cachep); 1567 goto out; 1568 } 1569 1570 /* Exit handler for the SCTP protocol. */ 1571 static __exit void sctp_exit(void) 1572 { 1573 /* BUG. This should probably do something useful like clean 1574 * up all the remaining associations and all that memory. 1575 */ 1576 1577 /* Unregister with inet6/inet layers. */ 1578 sctp_v6_del_protocol(); 1579 sctp_v4_del_protocol(); 1580 1581 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1582 1583 /* Free protosw registrations */ 1584 sctp_v6_protosw_exit(); 1585 sctp_v4_protosw_exit(); 1586 1587 unregister_pernet_subsys(&sctp_defaults_ops); 1588 1589 /* Unregister with socket layer. */ 1590 sctp_v6_pf_exit(); 1591 sctp_v4_pf_exit(); 1592 1593 sctp_sysctl_unregister(); 1594 1595 free_pages((unsigned long)sctp_port_hashtable, 1596 get_order(sctp_port_hashsize * 1597 sizeof(struct sctp_bind_hashbucket))); 1598 kfree(sctp_ep_hashtable); 1599 sctp_transport_hashtable_destroy(); 1600 1601 percpu_counter_destroy(&sctp_sockets_allocated); 1602 1603 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1604 1605 kmem_cache_destroy(sctp_chunk_cachep); 1606 kmem_cache_destroy(sctp_bucket_cachep); 1607 } 1608 1609 module_init(sctp_init); 1610 module_exit(sctp_exit); 1611 1612 /* 1613 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1614 */ 1615 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1616 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1617 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1618 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1619 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1620 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1621 MODULE_LICENSE("GPL"); 1622