1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* SCTP kernel implementation 3 * (C) Copyright IBM Corp. 2001, 2004 4 * Copyright (c) 1999-2000 Cisco, Inc. 5 * Copyright (c) 1999-2001 Motorola, Inc. 6 * Copyright (c) 2001 Intel Corp. 7 * Copyright (c) 2001 Nokia, Inc. 8 * Copyright (c) 2001 La Monte H.P. Yarroll 9 * 10 * This file is part of the SCTP kernel implementation 11 * 12 * Initialization/cleanup for SCTP protocol support. 13 * 14 * Please send any bug reports or fixes you make to the 15 * email address(es): 16 * lksctp developers <linux-sctp@vger.kernel.org> 17 * 18 * Written or modified by: 19 * La Monte H.P. Yarroll <piggy@acm.org> 20 * Karl Knutson <karl@athena.chicago.il.us> 21 * Jon Grimm <jgrimm@us.ibm.com> 22 * Sridhar Samudrala <sri@us.ibm.com> 23 * Daisy Chang <daisyc@us.ibm.com> 24 * Ardelle Fan <ardelle.fan@intel.com> 25 */ 26 27 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 28 29 #include <linux/module.h> 30 #include <linux/init.h> 31 #include <linux/netdevice.h> 32 #include <linux/inetdevice.h> 33 #include <linux/seq_file.h> 34 #include <linux/memblock.h> 35 #include <linux/highmem.h> 36 #include <linux/swap.h> 37 #include <linux/slab.h> 38 #include <net/net_namespace.h> 39 #include <net/protocol.h> 40 #include <net/ip.h> 41 #include <net/ipv6.h> 42 #include <net/route.h> 43 #include <net/sctp/sctp.h> 44 #include <net/addrconf.h> 45 #include <net/inet_common.h> 46 #include <net/inet_ecn.h> 47 48 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024) 49 50 /* Global data structures. */ 51 struct sctp_globals sctp_globals __read_mostly; 52 53 struct idr sctp_assocs_id; 54 DEFINE_SPINLOCK(sctp_assocs_id_lock); 55 56 static struct sctp_pf *sctp_pf_inet6_specific; 57 static struct sctp_pf *sctp_pf_inet_specific; 58 static struct sctp_af *sctp_af_v4_specific; 59 static struct sctp_af *sctp_af_v6_specific; 60 61 struct kmem_cache *sctp_chunk_cachep __read_mostly; 62 struct kmem_cache *sctp_bucket_cachep __read_mostly; 63 64 long sysctl_sctp_mem[3]; 65 int sysctl_sctp_rmem[3]; 66 int sysctl_sctp_wmem[3]; 67 68 /* Private helper to extract ipv4 address and stash them in 69 * the protocol structure. 70 */ 71 static void sctp_v4_copy_addrlist(struct list_head *addrlist, 72 struct net_device *dev) 73 { 74 struct in_device *in_dev; 75 struct in_ifaddr *ifa; 76 struct sctp_sockaddr_entry *addr; 77 78 rcu_read_lock(); 79 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 80 rcu_read_unlock(); 81 return; 82 } 83 84 in_dev_for_each_ifa_rcu(ifa, in_dev) { 85 /* Add the address to the local list. */ 86 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 87 if (addr) { 88 addr->a.v4.sin_family = AF_INET; 89 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 90 addr->valid = 1; 91 INIT_LIST_HEAD(&addr->list); 92 list_add_tail(&addr->list, addrlist); 93 } 94 } 95 96 rcu_read_unlock(); 97 } 98 99 /* Extract our IP addresses from the system and stash them in the 100 * protocol structure. 101 */ 102 static void sctp_get_local_addr_list(struct net *net) 103 { 104 struct net_device *dev; 105 struct list_head *pos; 106 struct sctp_af *af; 107 108 rcu_read_lock(); 109 for_each_netdev_rcu(net, dev) { 110 list_for_each(pos, &sctp_address_families) { 111 af = list_entry(pos, struct sctp_af, list); 112 af->copy_addrlist(&net->sctp.local_addr_list, dev); 113 } 114 } 115 rcu_read_unlock(); 116 } 117 118 /* Free the existing local addresses. */ 119 static void sctp_free_local_addr_list(struct net *net) 120 { 121 struct sctp_sockaddr_entry *addr; 122 struct list_head *pos, *temp; 123 124 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 125 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 126 list_del(pos); 127 kfree(addr); 128 } 129 } 130 131 /* Copy the local addresses which are valid for 'scope' into 'bp'. */ 132 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 133 enum sctp_scope scope, gfp_t gfp, int copy_flags) 134 { 135 struct sctp_sockaddr_entry *addr; 136 union sctp_addr laddr; 137 int error = 0; 138 139 rcu_read_lock(); 140 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 141 if (!addr->valid) 142 continue; 143 if (!sctp_in_scope(net, &addr->a, scope)) 144 continue; 145 146 /* Now that the address is in scope, check to see if 147 * the address type is really supported by the local 148 * sock as well as the remote peer. 149 */ 150 if (addr->a.sa.sa_family == AF_INET && 151 !(copy_flags & SCTP_ADDR4_PEERSUPP)) 152 continue; 153 if (addr->a.sa.sa_family == AF_INET6 && 154 (!(copy_flags & SCTP_ADDR6_ALLOWED) || 155 !(copy_flags & SCTP_ADDR6_PEERSUPP))) 156 continue; 157 158 laddr = addr->a; 159 /* also works for setting ipv6 address port */ 160 laddr.v4.sin_port = htons(bp->port); 161 if (sctp_bind_addr_state(bp, &laddr) != -1) 162 continue; 163 164 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a), 165 SCTP_ADDR_SRC, GFP_ATOMIC); 166 if (error) 167 break; 168 } 169 170 rcu_read_unlock(); 171 return error; 172 } 173 174 /* Copy over any ip options */ 175 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk) 176 { 177 struct inet_sock *newinet, *inet = inet_sk(sk); 178 struct ip_options_rcu *inet_opt, *newopt = NULL; 179 180 newinet = inet_sk(newsk); 181 182 rcu_read_lock(); 183 inet_opt = rcu_dereference(inet->inet_opt); 184 if (inet_opt) { 185 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) + 186 inet_opt->opt.optlen, GFP_ATOMIC); 187 if (newopt) 188 memcpy(newopt, inet_opt, sizeof(*inet_opt) + 189 inet_opt->opt.optlen); 190 else 191 pr_err("%s: Failed to copy ip options\n", __func__); 192 } 193 RCU_INIT_POINTER(newinet->inet_opt, newopt); 194 rcu_read_unlock(); 195 } 196 197 /* Account for the IP options */ 198 static int sctp_v4_ip_options_len(struct sock *sk) 199 { 200 struct inet_sock *inet = inet_sk(sk); 201 struct ip_options_rcu *inet_opt; 202 int len = 0; 203 204 rcu_read_lock(); 205 inet_opt = rcu_dereference(inet->inet_opt); 206 if (inet_opt) 207 len = inet_opt->opt.optlen; 208 209 rcu_read_unlock(); 210 return len; 211 } 212 213 /* Initialize a sctp_addr from in incoming skb. */ 214 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 215 int is_saddr) 216 { 217 /* Always called on head skb, so this is safe */ 218 struct sctphdr *sh = sctp_hdr(skb); 219 struct sockaddr_in *sa = &addr->v4; 220 221 addr->v4.sin_family = AF_INET; 222 223 if (is_saddr) { 224 sa->sin_port = sh->source; 225 sa->sin_addr.s_addr = ip_hdr(skb)->saddr; 226 } else { 227 sa->sin_port = sh->dest; 228 sa->sin_addr.s_addr = ip_hdr(skb)->daddr; 229 } 230 } 231 232 /* Initialize an sctp_addr from a socket. */ 233 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 234 { 235 addr->v4.sin_family = AF_INET; 236 addr->v4.sin_port = 0; 237 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 238 } 239 240 /* Initialize sk->sk_rcv_saddr from sctp_addr. */ 241 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 242 { 243 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 244 } 245 246 /* Initialize sk->sk_daddr from sctp_addr. */ 247 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 248 { 249 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 250 } 251 252 /* Initialize a sctp_addr from an address parameter. */ 253 static void sctp_v4_from_addr_param(union sctp_addr *addr, 254 union sctp_addr_param *param, 255 __be16 port, int iif) 256 { 257 addr->v4.sin_family = AF_INET; 258 addr->v4.sin_port = port; 259 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 260 } 261 262 /* Initialize an address parameter from a sctp_addr and return the length 263 * of the address parameter. 264 */ 265 static int sctp_v4_to_addr_param(const union sctp_addr *addr, 266 union sctp_addr_param *param) 267 { 268 int length = sizeof(struct sctp_ipv4addr_param); 269 270 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 271 param->v4.param_hdr.length = htons(length); 272 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 273 274 return length; 275 } 276 277 /* Initialize a sctp_addr from a dst_entry. */ 278 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 279 __be16 port) 280 { 281 saddr->v4.sin_family = AF_INET; 282 saddr->v4.sin_port = port; 283 saddr->v4.sin_addr.s_addr = fl4->saddr; 284 } 285 286 /* Compare two addresses exactly. */ 287 static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 288 const union sctp_addr *addr2) 289 { 290 if (addr1->sa.sa_family != addr2->sa.sa_family) 291 return 0; 292 if (addr1->v4.sin_port != addr2->v4.sin_port) 293 return 0; 294 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 295 return 0; 296 297 return 1; 298 } 299 300 /* Initialize addr struct to INADDR_ANY. */ 301 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 302 { 303 addr->v4.sin_family = AF_INET; 304 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 305 addr->v4.sin_port = port; 306 } 307 308 /* Is this a wildcard address? */ 309 static int sctp_v4_is_any(const union sctp_addr *addr) 310 { 311 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 312 } 313 314 /* This function checks if the address is a valid address to be used for 315 * SCTP binding. 316 * 317 * Output: 318 * Return 0 - If the address is a non-unicast or an illegal address. 319 * Return 1 - If the address is a unicast. 320 */ 321 static int sctp_v4_addr_valid(union sctp_addr *addr, 322 struct sctp_sock *sp, 323 const struct sk_buff *skb) 324 { 325 /* IPv4 addresses not allowed */ 326 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 327 return 0; 328 329 /* Is this a non-unicast address or a unusable SCTP address? */ 330 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 331 return 0; 332 333 /* Is this a broadcast address? */ 334 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 335 return 0; 336 337 return 1; 338 } 339 340 /* Should this be available for binding? */ 341 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 342 { 343 struct net *net = sock_net(&sp->inet.sk); 344 int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr); 345 346 347 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 348 ret != RTN_LOCAL && 349 !sp->inet.freebind && 350 !net->ipv4.sysctl_ip_nonlocal_bind) 351 return 0; 352 353 if (ipv6_only_sock(sctp_opt2sk(sp))) 354 return 0; 355 356 return 1; 357 } 358 359 /* Checking the loopback, private and other address scopes as defined in 360 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 361 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 362 * 363 * Level 0 - unusable SCTP addresses 364 * Level 1 - loopback address 365 * Level 2 - link-local addresses 366 * Level 3 - private addresses. 367 * Level 4 - global addresses 368 * For INIT and INIT-ACK address list, let L be the level of 369 * of requested destination address, sender and receiver 370 * SHOULD include all of its addresses with level greater 371 * than or equal to L. 372 * 373 * IPv4 scoping can be controlled through sysctl option 374 * net.sctp.addr_scope_policy 375 */ 376 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr) 377 { 378 enum sctp_scope retval; 379 380 /* Check for unusable SCTP addresses. */ 381 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 382 retval = SCTP_SCOPE_UNUSABLE; 383 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 384 retval = SCTP_SCOPE_LOOPBACK; 385 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 386 retval = SCTP_SCOPE_LINK; 387 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 388 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 389 ipv4_is_private_192(addr->v4.sin_addr.s_addr)) { 390 retval = SCTP_SCOPE_PRIVATE; 391 } else { 392 retval = SCTP_SCOPE_GLOBAL; 393 } 394 395 return retval; 396 } 397 398 /* Returns a valid dst cache entry for the given source and destination ip 399 * addresses. If an association is passed, trys to get a dst entry with a 400 * source address that matches an address in the bind address list. 401 */ 402 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 403 struct flowi *fl, struct sock *sk) 404 { 405 struct sctp_association *asoc = t->asoc; 406 struct rtable *rt; 407 struct flowi4 *fl4 = &fl->u.ip4; 408 struct sctp_bind_addr *bp; 409 struct sctp_sockaddr_entry *laddr; 410 struct dst_entry *dst = NULL; 411 union sctp_addr *daddr = &t->ipaddr; 412 union sctp_addr dst_saddr; 413 __u8 tos = inet_sk(sk)->tos; 414 415 if (t->dscp & SCTP_DSCP_SET_MASK) 416 tos = t->dscp & SCTP_DSCP_VAL_MASK; 417 memset(fl4, 0x0, sizeof(struct flowi4)); 418 fl4->daddr = daddr->v4.sin_addr.s_addr; 419 fl4->fl4_dport = daddr->v4.sin_port; 420 fl4->flowi4_proto = IPPROTO_SCTP; 421 if (asoc) { 422 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos); 423 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 424 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 425 } 426 if (saddr) { 427 fl4->saddr = saddr->v4.sin_addr.s_addr; 428 if (!fl4->fl4_sport) 429 fl4->fl4_sport = saddr->v4.sin_port; 430 } 431 432 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 433 &fl4->saddr); 434 435 rt = ip_route_output_key(sock_net(sk), fl4); 436 if (!IS_ERR(rt)) 437 dst = &rt->dst; 438 439 /* If there is no association or if a source address is passed, no 440 * more validation is required. 441 */ 442 if (!asoc || saddr) 443 goto out; 444 445 bp = &asoc->base.bind_addr; 446 447 if (dst) { 448 /* Walk through the bind address list and look for a bind 449 * address that matches the source address of the returned dst. 450 */ 451 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 452 rcu_read_lock(); 453 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 454 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 455 (laddr->state != SCTP_ADDR_SRC && 456 !asoc->src_out_of_asoc_ok)) 457 continue; 458 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 459 goto out_unlock; 460 } 461 rcu_read_unlock(); 462 463 /* None of the bound addresses match the source address of the 464 * dst. So release it. 465 */ 466 dst_release(dst); 467 dst = NULL; 468 } 469 470 /* Walk through the bind address list and try to get a dst that 471 * matches a bind address as the source address. 472 */ 473 rcu_read_lock(); 474 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 475 struct net_device *odev; 476 477 if (!laddr->valid) 478 continue; 479 if (laddr->state != SCTP_ADDR_SRC || 480 AF_INET != laddr->a.sa.sa_family) 481 continue; 482 483 fl4->fl4_sport = laddr->a.v4.sin_port; 484 flowi4_update_output(fl4, 485 asoc->base.sk->sk_bound_dev_if, 486 RT_CONN_FLAGS_TOS(asoc->base.sk, tos), 487 daddr->v4.sin_addr.s_addr, 488 laddr->a.v4.sin_addr.s_addr); 489 490 rt = ip_route_output_key(sock_net(sk), fl4); 491 if (IS_ERR(rt)) 492 continue; 493 494 /* Ensure the src address belongs to the output 495 * interface. 496 */ 497 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr, 498 false); 499 if (!odev || odev->ifindex != fl4->flowi4_oif) { 500 if (!dst) 501 dst = &rt->dst; 502 else 503 dst_release(&rt->dst); 504 continue; 505 } 506 507 dst_release(dst); 508 dst = &rt->dst; 509 break; 510 } 511 512 out_unlock: 513 rcu_read_unlock(); 514 out: 515 t->dst = dst; 516 if (dst) 517 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 518 &fl4->daddr, &fl4->saddr); 519 else 520 pr_debug("no route\n"); 521 } 522 523 /* For v4, the source address is cached in the route entry(dst). So no need 524 * to cache it separately and hence this is an empty routine. 525 */ 526 static void sctp_v4_get_saddr(struct sctp_sock *sk, 527 struct sctp_transport *t, 528 struct flowi *fl) 529 { 530 union sctp_addr *saddr = &t->saddr; 531 struct rtable *rt = (struct rtable *)t->dst; 532 533 if (rt) { 534 saddr->v4.sin_family = AF_INET; 535 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 536 } 537 } 538 539 /* What interface did this skb arrive on? */ 540 static int sctp_v4_skb_iif(const struct sk_buff *skb) 541 { 542 return inet_iif(skb); 543 } 544 545 /* Was this packet marked by Explicit Congestion Notification? */ 546 static int sctp_v4_is_ce(const struct sk_buff *skb) 547 { 548 return INET_ECN_is_ce(ip_hdr(skb)->tos); 549 } 550 551 /* Create and initialize a new sk for the socket returned by accept(). */ 552 static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 553 struct sctp_association *asoc, 554 bool kern) 555 { 556 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 557 sk->sk_prot, kern); 558 struct inet_sock *newinet; 559 560 if (!newsk) 561 goto out; 562 563 sock_init_data(NULL, newsk); 564 565 sctp_copy_sock(newsk, sk, asoc); 566 sock_reset_flag(newsk, SOCK_ZAPPED); 567 568 sctp_v4_copy_ip_options(sk, newsk); 569 570 newinet = inet_sk(newsk); 571 572 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 573 574 sk_refcnt_debug_inc(newsk); 575 576 if (newsk->sk_prot->init(newsk)) { 577 sk_common_release(newsk); 578 newsk = NULL; 579 } 580 581 out: 582 return newsk; 583 } 584 585 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 586 { 587 /* No address mapping for V4 sockets */ 588 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 589 return sizeof(struct sockaddr_in); 590 } 591 592 /* Dump the v4 addr to the seq file. */ 593 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 594 { 595 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 596 } 597 598 static void sctp_v4_ecn_capable(struct sock *sk) 599 { 600 INET_ECN_xmit(sk); 601 } 602 603 static void sctp_addr_wq_timeout_handler(struct timer_list *t) 604 { 605 struct net *net = from_timer(net, t, sctp.addr_wq_timer); 606 struct sctp_sockaddr_entry *addrw, *temp; 607 struct sctp_sock *sp; 608 609 spin_lock_bh(&net->sctp.addr_wq_lock); 610 611 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 612 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 613 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 614 addrw->state, addrw); 615 616 #if IS_ENABLED(CONFIG_IPV6) 617 /* Now we send an ASCONF for each association */ 618 /* Note. we currently don't handle link local IPv6 addressees */ 619 if (addrw->a.sa.sa_family == AF_INET6) { 620 struct in6_addr *in6; 621 622 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 623 IPV6_ADDR_LINKLOCAL) 624 goto free_next; 625 626 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 627 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 628 addrw->state == SCTP_ADDR_NEW) { 629 unsigned long timeo_val; 630 631 pr_debug("%s: this is on DAD, trying %d sec " 632 "later\n", __func__, 633 SCTP_ADDRESS_TICK_DELAY); 634 635 timeo_val = jiffies; 636 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 637 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 638 break; 639 } 640 } 641 #endif 642 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 643 struct sock *sk; 644 645 sk = sctp_opt2sk(sp); 646 /* ignore bound-specific endpoints */ 647 if (!sctp_is_ep_boundall(sk)) 648 continue; 649 bh_lock_sock(sk); 650 if (sctp_asconf_mgmt(sp, addrw) < 0) 651 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 652 bh_unlock_sock(sk); 653 } 654 #if IS_ENABLED(CONFIG_IPV6) 655 free_next: 656 #endif 657 list_del(&addrw->list); 658 kfree(addrw); 659 } 660 spin_unlock_bh(&net->sctp.addr_wq_lock); 661 } 662 663 static void sctp_free_addr_wq(struct net *net) 664 { 665 struct sctp_sockaddr_entry *addrw; 666 struct sctp_sockaddr_entry *temp; 667 668 spin_lock_bh(&net->sctp.addr_wq_lock); 669 del_timer(&net->sctp.addr_wq_timer); 670 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 671 list_del(&addrw->list); 672 kfree(addrw); 673 } 674 spin_unlock_bh(&net->sctp.addr_wq_lock); 675 } 676 677 /* lookup the entry for the same address in the addr_waitq 678 * sctp_addr_wq MUST be locked 679 */ 680 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 681 struct sctp_sockaddr_entry *addr) 682 { 683 struct sctp_sockaddr_entry *addrw; 684 685 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 686 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 687 continue; 688 if (addrw->a.sa.sa_family == AF_INET) { 689 if (addrw->a.v4.sin_addr.s_addr == 690 addr->a.v4.sin_addr.s_addr) 691 return addrw; 692 } else if (addrw->a.sa.sa_family == AF_INET6) { 693 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 694 &addr->a.v6.sin6_addr)) 695 return addrw; 696 } 697 } 698 return NULL; 699 } 700 701 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 702 { 703 struct sctp_sockaddr_entry *addrw; 704 unsigned long timeo_val; 705 706 /* first, we check if an opposite message already exist in the queue. 707 * If we found such message, it is removed. 708 * This operation is a bit stupid, but the DHCP client attaches the 709 * new address after a couple of addition and deletion of that address 710 */ 711 712 spin_lock_bh(&net->sctp.addr_wq_lock); 713 /* Offsets existing events in addr_wq */ 714 addrw = sctp_addr_wq_lookup(net, addr); 715 if (addrw) { 716 if (addrw->state != cmd) { 717 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 718 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 719 &net->sctp.addr_waitq); 720 721 list_del(&addrw->list); 722 kfree(addrw); 723 } 724 spin_unlock_bh(&net->sctp.addr_wq_lock); 725 return; 726 } 727 728 /* OK, we have to add the new address to the wait queue */ 729 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 730 if (addrw == NULL) { 731 spin_unlock_bh(&net->sctp.addr_wq_lock); 732 return; 733 } 734 addrw->state = cmd; 735 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 736 737 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 738 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 739 740 if (!timer_pending(&net->sctp.addr_wq_timer)) { 741 timeo_val = jiffies; 742 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 743 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 744 } 745 spin_unlock_bh(&net->sctp.addr_wq_lock); 746 } 747 748 /* Event handler for inet address addition/deletion events. 749 * The sctp_local_addr_list needs to be protocted by a spin lock since 750 * multiple notifiers (say IPv4 and IPv6) may be running at the same 751 * time and thus corrupt the list. 752 * The reader side is protected with RCU. 753 */ 754 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 755 void *ptr) 756 { 757 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 758 struct sctp_sockaddr_entry *addr = NULL; 759 struct sctp_sockaddr_entry *temp; 760 struct net *net = dev_net(ifa->ifa_dev->dev); 761 int found = 0; 762 763 switch (ev) { 764 case NETDEV_UP: 765 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 766 if (addr) { 767 addr->a.v4.sin_family = AF_INET; 768 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 769 addr->valid = 1; 770 spin_lock_bh(&net->sctp.local_addr_lock); 771 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 772 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 773 spin_unlock_bh(&net->sctp.local_addr_lock); 774 } 775 break; 776 case NETDEV_DOWN: 777 spin_lock_bh(&net->sctp.local_addr_lock); 778 list_for_each_entry_safe(addr, temp, 779 &net->sctp.local_addr_list, list) { 780 if (addr->a.sa.sa_family == AF_INET && 781 addr->a.v4.sin_addr.s_addr == 782 ifa->ifa_local) { 783 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 784 found = 1; 785 addr->valid = 0; 786 list_del_rcu(&addr->list); 787 break; 788 } 789 } 790 spin_unlock_bh(&net->sctp.local_addr_lock); 791 if (found) 792 kfree_rcu(addr, rcu); 793 break; 794 } 795 796 return NOTIFY_DONE; 797 } 798 799 /* 800 * Initialize the control inode/socket with a control endpoint data 801 * structure. This endpoint is reserved exclusively for the OOTB processing. 802 */ 803 static int sctp_ctl_sock_init(struct net *net) 804 { 805 int err; 806 sa_family_t family = PF_INET; 807 808 if (sctp_get_pf_specific(PF_INET6)) 809 family = PF_INET6; 810 811 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 812 SOCK_SEQPACKET, IPPROTO_SCTP, net); 813 814 /* If IPv6 socket could not be created, try the IPv4 socket */ 815 if (err < 0 && family == PF_INET6) 816 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 817 SOCK_SEQPACKET, IPPROTO_SCTP, 818 net); 819 820 if (err < 0) { 821 pr_err("Failed to create the SCTP control socket\n"); 822 return err; 823 } 824 return 0; 825 } 826 827 /* Register address family specific functions. */ 828 int sctp_register_af(struct sctp_af *af) 829 { 830 switch (af->sa_family) { 831 case AF_INET: 832 if (sctp_af_v4_specific) 833 return 0; 834 sctp_af_v4_specific = af; 835 break; 836 case AF_INET6: 837 if (sctp_af_v6_specific) 838 return 0; 839 sctp_af_v6_specific = af; 840 break; 841 default: 842 return 0; 843 } 844 845 INIT_LIST_HEAD(&af->list); 846 list_add_tail(&af->list, &sctp_address_families); 847 return 1; 848 } 849 850 /* Get the table of functions for manipulating a particular address 851 * family. 852 */ 853 struct sctp_af *sctp_get_af_specific(sa_family_t family) 854 { 855 switch (family) { 856 case AF_INET: 857 return sctp_af_v4_specific; 858 case AF_INET6: 859 return sctp_af_v6_specific; 860 default: 861 return NULL; 862 } 863 } 864 865 /* Common code to initialize a AF_INET msg_name. */ 866 static void sctp_inet_msgname(char *msgname, int *addr_len) 867 { 868 struct sockaddr_in *sin; 869 870 sin = (struct sockaddr_in *)msgname; 871 *addr_len = sizeof(struct sockaddr_in); 872 sin->sin_family = AF_INET; 873 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 874 } 875 876 /* Copy the primary address of the peer primary address as the msg_name. */ 877 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 878 int *addr_len) 879 { 880 struct sockaddr_in *sin, *sinfrom; 881 882 if (msgname) { 883 struct sctp_association *asoc; 884 885 asoc = event->asoc; 886 sctp_inet_msgname(msgname, addr_len); 887 sin = (struct sockaddr_in *)msgname; 888 sinfrom = &asoc->peer.primary_addr.v4; 889 sin->sin_port = htons(asoc->peer.port); 890 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 891 } 892 } 893 894 /* Initialize and copy out a msgname from an inbound skb. */ 895 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 896 { 897 if (msgname) { 898 struct sctphdr *sh = sctp_hdr(skb); 899 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 900 901 sctp_inet_msgname(msgname, len); 902 sin->sin_port = sh->source; 903 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 904 } 905 } 906 907 /* Do we support this AF? */ 908 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 909 { 910 /* PF_INET only supports AF_INET addresses. */ 911 return AF_INET == family; 912 } 913 914 /* Address matching with wildcards allowed. */ 915 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 916 const union sctp_addr *addr2, 917 struct sctp_sock *opt) 918 { 919 /* PF_INET only supports AF_INET addresses. */ 920 if (addr1->sa.sa_family != addr2->sa.sa_family) 921 return 0; 922 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 923 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 924 return 1; 925 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 926 return 1; 927 928 return 0; 929 } 930 931 /* Verify that provided sockaddr looks bindable. Common verification has 932 * already been taken care of. 933 */ 934 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 935 { 936 return sctp_v4_available(addr, opt); 937 } 938 939 /* Verify that sockaddr looks sendable. Common verification has already 940 * been taken care of. 941 */ 942 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 943 { 944 return 1; 945 } 946 947 /* Fill in Supported Address Type information for INIT and INIT-ACK 948 * chunks. Returns number of addresses supported. 949 */ 950 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 951 __be16 *types) 952 { 953 types[0] = SCTP_PARAM_IPV4_ADDRESS; 954 return 1; 955 } 956 957 /* Wrapper routine that calls the ip transmit routine. */ 958 static inline int sctp_v4_xmit(struct sk_buff *skb, 959 struct sctp_transport *transport) 960 { 961 struct inet_sock *inet = inet_sk(skb->sk); 962 __u8 dscp = inet->tos; 963 964 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 965 skb->len, &transport->fl.u.ip4.saddr, 966 &transport->fl.u.ip4.daddr); 967 968 if (transport->dscp & SCTP_DSCP_SET_MASK) 969 dscp = transport->dscp & SCTP_DSCP_VAL_MASK; 970 971 inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ? 972 IP_PMTUDISC_DO : IP_PMTUDISC_DONT; 973 974 SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS); 975 976 return __ip_queue_xmit(&inet->sk, skb, &transport->fl, dscp); 977 } 978 979 static struct sctp_af sctp_af_inet; 980 981 static struct sctp_pf sctp_pf_inet = { 982 .event_msgname = sctp_inet_event_msgname, 983 .skb_msgname = sctp_inet_skb_msgname, 984 .af_supported = sctp_inet_af_supported, 985 .cmp_addr = sctp_inet_cmp_addr, 986 .bind_verify = sctp_inet_bind_verify, 987 .send_verify = sctp_inet_send_verify, 988 .supported_addrs = sctp_inet_supported_addrs, 989 .create_accept_sk = sctp_v4_create_accept_sk, 990 .addr_to_user = sctp_v4_addr_to_user, 991 .to_sk_saddr = sctp_v4_to_sk_saddr, 992 .to_sk_daddr = sctp_v4_to_sk_daddr, 993 .copy_ip_options = sctp_v4_copy_ip_options, 994 .af = &sctp_af_inet 995 }; 996 997 /* Notifier for inetaddr addition/deletion events. */ 998 static struct notifier_block sctp_inetaddr_notifier = { 999 .notifier_call = sctp_inetaddr_event, 1000 }; 1001 1002 /* Socket operations. */ 1003 static const struct proto_ops inet_seqpacket_ops = { 1004 .family = PF_INET, 1005 .owner = THIS_MODULE, 1006 .release = inet_release, /* Needs to be wrapped... */ 1007 .bind = inet_bind, 1008 .connect = sctp_inet_connect, 1009 .socketpair = sock_no_socketpair, 1010 .accept = inet_accept, 1011 .getname = inet_getname, /* Semantics are different. */ 1012 .poll = sctp_poll, 1013 .ioctl = inet_ioctl, 1014 .gettstamp = sock_gettstamp, 1015 .listen = sctp_inet_listen, 1016 .shutdown = inet_shutdown, /* Looks harmless. */ 1017 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1018 .getsockopt = sock_common_getsockopt, 1019 .sendmsg = inet_sendmsg, 1020 .recvmsg = inet_recvmsg, 1021 .mmap = sock_no_mmap, 1022 .sendpage = sock_no_sendpage, 1023 #ifdef CONFIG_COMPAT 1024 .compat_setsockopt = compat_sock_common_setsockopt, 1025 .compat_getsockopt = compat_sock_common_getsockopt, 1026 #endif 1027 }; 1028 1029 /* Registration with AF_INET family. */ 1030 static struct inet_protosw sctp_seqpacket_protosw = { 1031 .type = SOCK_SEQPACKET, 1032 .protocol = IPPROTO_SCTP, 1033 .prot = &sctp_prot, 1034 .ops = &inet_seqpacket_ops, 1035 .flags = SCTP_PROTOSW_FLAG 1036 }; 1037 static struct inet_protosw sctp_stream_protosw = { 1038 .type = SOCK_STREAM, 1039 .protocol = IPPROTO_SCTP, 1040 .prot = &sctp_prot, 1041 .ops = &inet_seqpacket_ops, 1042 .flags = SCTP_PROTOSW_FLAG 1043 }; 1044 1045 /* Register with IP layer. */ 1046 static const struct net_protocol sctp_protocol = { 1047 .handler = sctp_rcv, 1048 .err_handler = sctp_v4_err, 1049 .no_policy = 1, 1050 .netns_ok = 1, 1051 .icmp_strict_tag_validation = 1, 1052 }; 1053 1054 /* IPv4 address related functions. */ 1055 static struct sctp_af sctp_af_inet = { 1056 .sa_family = AF_INET, 1057 .sctp_xmit = sctp_v4_xmit, 1058 .setsockopt = ip_setsockopt, 1059 .getsockopt = ip_getsockopt, 1060 .get_dst = sctp_v4_get_dst, 1061 .get_saddr = sctp_v4_get_saddr, 1062 .copy_addrlist = sctp_v4_copy_addrlist, 1063 .from_skb = sctp_v4_from_skb, 1064 .from_sk = sctp_v4_from_sk, 1065 .from_addr_param = sctp_v4_from_addr_param, 1066 .to_addr_param = sctp_v4_to_addr_param, 1067 .cmp_addr = sctp_v4_cmp_addr, 1068 .addr_valid = sctp_v4_addr_valid, 1069 .inaddr_any = sctp_v4_inaddr_any, 1070 .is_any = sctp_v4_is_any, 1071 .available = sctp_v4_available, 1072 .scope = sctp_v4_scope, 1073 .skb_iif = sctp_v4_skb_iif, 1074 .is_ce = sctp_v4_is_ce, 1075 .seq_dump_addr = sctp_v4_seq_dump_addr, 1076 .ecn_capable = sctp_v4_ecn_capable, 1077 .net_header_len = sizeof(struct iphdr), 1078 .sockaddr_len = sizeof(struct sockaddr_in), 1079 .ip_options_len = sctp_v4_ip_options_len, 1080 #ifdef CONFIG_COMPAT 1081 .compat_setsockopt = compat_ip_setsockopt, 1082 .compat_getsockopt = compat_ip_getsockopt, 1083 #endif 1084 }; 1085 1086 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1087 { 1088 switch (family) { 1089 case PF_INET: 1090 return sctp_pf_inet_specific; 1091 case PF_INET6: 1092 return sctp_pf_inet6_specific; 1093 default: 1094 return NULL; 1095 } 1096 } 1097 1098 /* Register the PF specific function table. */ 1099 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1100 { 1101 switch (family) { 1102 case PF_INET: 1103 if (sctp_pf_inet_specific) 1104 return 0; 1105 sctp_pf_inet_specific = pf; 1106 break; 1107 case PF_INET6: 1108 if (sctp_pf_inet6_specific) 1109 return 0; 1110 sctp_pf_inet6_specific = pf; 1111 break; 1112 default: 1113 return 0; 1114 } 1115 return 1; 1116 } 1117 1118 static inline int init_sctp_mibs(struct net *net) 1119 { 1120 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1121 if (!net->sctp.sctp_statistics) 1122 return -ENOMEM; 1123 return 0; 1124 } 1125 1126 static inline void cleanup_sctp_mibs(struct net *net) 1127 { 1128 free_percpu(net->sctp.sctp_statistics); 1129 } 1130 1131 static void sctp_v4_pf_init(void) 1132 { 1133 /* Initialize the SCTP specific PF functions. */ 1134 sctp_register_pf(&sctp_pf_inet, PF_INET); 1135 sctp_register_af(&sctp_af_inet); 1136 } 1137 1138 static void sctp_v4_pf_exit(void) 1139 { 1140 list_del(&sctp_af_inet.list); 1141 } 1142 1143 static int sctp_v4_protosw_init(void) 1144 { 1145 int rc; 1146 1147 rc = proto_register(&sctp_prot, 1); 1148 if (rc) 1149 return rc; 1150 1151 /* Register SCTP(UDP and TCP style) with socket layer. */ 1152 inet_register_protosw(&sctp_seqpacket_protosw); 1153 inet_register_protosw(&sctp_stream_protosw); 1154 1155 return 0; 1156 } 1157 1158 static void sctp_v4_protosw_exit(void) 1159 { 1160 inet_unregister_protosw(&sctp_stream_protosw); 1161 inet_unregister_protosw(&sctp_seqpacket_protosw); 1162 proto_unregister(&sctp_prot); 1163 } 1164 1165 static int sctp_v4_add_protocol(void) 1166 { 1167 /* Register notifier for inet address additions/deletions. */ 1168 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1169 1170 /* Register SCTP with inet layer. */ 1171 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1172 return -EAGAIN; 1173 1174 return 0; 1175 } 1176 1177 static void sctp_v4_del_protocol(void) 1178 { 1179 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1180 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1181 } 1182 1183 static int __net_init sctp_defaults_init(struct net *net) 1184 { 1185 int status; 1186 1187 /* 1188 * 14. Suggested SCTP Protocol Parameter Values 1189 */ 1190 /* The following protocol parameters are RECOMMENDED: */ 1191 /* RTO.Initial - 3 seconds */ 1192 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1193 /* RTO.Min - 1 second */ 1194 net->sctp.rto_min = SCTP_RTO_MIN; 1195 /* RTO.Max - 60 seconds */ 1196 net->sctp.rto_max = SCTP_RTO_MAX; 1197 /* RTO.Alpha - 1/8 */ 1198 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1199 /* RTO.Beta - 1/4 */ 1200 net->sctp.rto_beta = SCTP_RTO_BETA; 1201 1202 /* Valid.Cookie.Life - 60 seconds */ 1203 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1204 1205 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1206 net->sctp.cookie_preserve_enable = 1; 1207 1208 /* Default sctp sockets to use md5 as their hmac alg */ 1209 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1210 net->sctp.sctp_hmac_alg = "md5"; 1211 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1212 net->sctp.sctp_hmac_alg = "sha1"; 1213 #else 1214 net->sctp.sctp_hmac_alg = NULL; 1215 #endif 1216 1217 /* Max.Burst - 4 */ 1218 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1219 1220 /* Disable of Primary Path Switchover by default */ 1221 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX; 1222 1223 /* Enable pf state by default */ 1224 net->sctp.pf_enable = 1; 1225 1226 /* Ignore pf exposure feature by default */ 1227 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET; 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 /* Enable ECN by default. */ 1264 net->sctp.ecn_enable = 1; 1265 1266 /* Set SCOPE policy to enabled */ 1267 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1268 1269 /* Set the default rwnd update threshold */ 1270 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1271 1272 /* Initialize maximum autoclose timeout. */ 1273 net->sctp.max_autoclose = INT_MAX / HZ; 1274 1275 status = sctp_sysctl_net_register(net); 1276 if (status) 1277 goto err_sysctl_register; 1278 1279 /* Allocate and initialise sctp mibs. */ 1280 status = init_sctp_mibs(net); 1281 if (status) 1282 goto err_init_mibs; 1283 1284 #ifdef CONFIG_PROC_FS 1285 /* Initialize proc fs directory. */ 1286 status = sctp_proc_init(net); 1287 if (status) 1288 goto err_init_proc; 1289 #endif 1290 1291 sctp_dbg_objcnt_init(net); 1292 1293 /* Initialize the local address list. */ 1294 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1295 spin_lock_init(&net->sctp.local_addr_lock); 1296 sctp_get_local_addr_list(net); 1297 1298 /* Initialize the address event list */ 1299 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1300 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1301 spin_lock_init(&net->sctp.addr_wq_lock); 1302 net->sctp.addr_wq_timer.expires = 0; 1303 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1304 1305 return 0; 1306 1307 #ifdef CONFIG_PROC_FS 1308 err_init_proc: 1309 cleanup_sctp_mibs(net); 1310 #endif 1311 err_init_mibs: 1312 sctp_sysctl_net_unregister(net); 1313 err_sysctl_register: 1314 return status; 1315 } 1316 1317 static void __net_exit sctp_defaults_exit(struct net *net) 1318 { 1319 /* Free the local address list */ 1320 sctp_free_addr_wq(net); 1321 sctp_free_local_addr_list(net); 1322 1323 #ifdef CONFIG_PROC_FS 1324 remove_proc_subtree("sctp", net->proc_net); 1325 net->sctp.proc_net_sctp = NULL; 1326 #endif 1327 cleanup_sctp_mibs(net); 1328 sctp_sysctl_net_unregister(net); 1329 } 1330 1331 static struct pernet_operations sctp_defaults_ops = { 1332 .init = sctp_defaults_init, 1333 .exit = sctp_defaults_exit, 1334 }; 1335 1336 static int __net_init sctp_ctrlsock_init(struct net *net) 1337 { 1338 int status; 1339 1340 /* Initialize the control inode/socket for handling OOTB packets. */ 1341 status = sctp_ctl_sock_init(net); 1342 if (status) 1343 pr_err("Failed to initialize the SCTP control sock\n"); 1344 1345 return status; 1346 } 1347 1348 static void __net_exit sctp_ctrlsock_exit(struct net *net) 1349 { 1350 /* Free the control endpoint. */ 1351 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1352 } 1353 1354 static struct pernet_operations sctp_ctrlsock_ops = { 1355 .init = sctp_ctrlsock_init, 1356 .exit = sctp_ctrlsock_exit, 1357 }; 1358 1359 /* Initialize the universe into something sensible. */ 1360 static __init int sctp_init(void) 1361 { 1362 int i; 1363 int status = -EINVAL; 1364 unsigned long goal; 1365 unsigned long limit; 1366 unsigned long nr_pages = totalram_pages(); 1367 int max_share; 1368 int order; 1369 int num_entries; 1370 int max_entry_order; 1371 1372 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1373 1374 /* Allocate bind_bucket and chunk caches. */ 1375 status = -ENOBUFS; 1376 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket", 1377 sizeof(struct sctp_bind_bucket), 1378 0, SLAB_HWCACHE_ALIGN, 1379 NULL); 1380 if (!sctp_bucket_cachep) 1381 goto out; 1382 1383 sctp_chunk_cachep = kmem_cache_create("sctp_chunk", 1384 sizeof(struct sctp_chunk), 1385 0, SLAB_HWCACHE_ALIGN, 1386 NULL); 1387 if (!sctp_chunk_cachep) 1388 goto err_chunk_cachep; 1389 1390 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1391 if (status) 1392 goto err_percpu_counter_init; 1393 1394 /* Implementation specific variables. */ 1395 1396 /* Initialize default stream count setup information. */ 1397 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1398 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1399 1400 /* Initialize handle used for association ids. */ 1401 idr_init(&sctp_assocs_id); 1402 1403 limit = nr_free_buffer_pages() / 8; 1404 limit = max(limit, 128UL); 1405 sysctl_sctp_mem[0] = limit / 4 * 3; 1406 sysctl_sctp_mem[1] = limit; 1407 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1408 1409 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1410 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1411 max_share = min(4UL*1024*1024, limit); 1412 1413 sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */ 1414 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1415 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1416 1417 sysctl_sctp_wmem[0] = SK_MEM_QUANTUM; 1418 sysctl_sctp_wmem[1] = 16*1024; 1419 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1420 1421 /* Size and allocate the association hash table. 1422 * The methodology is similar to that of the tcp hash tables. 1423 * Though not identical. Start by getting a goal size 1424 */ 1425 if (nr_pages >= (128 * 1024)) 1426 goal = nr_pages >> (22 - PAGE_SHIFT); 1427 else 1428 goal = nr_pages >> (24 - PAGE_SHIFT); 1429 1430 /* Then compute the page order for said goal */ 1431 order = get_order(goal); 1432 1433 /* Now compute the required page order for the maximum sized table we 1434 * want to create 1435 */ 1436 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1437 sizeof(struct sctp_bind_hashbucket)); 1438 1439 /* Limit the page order by that maximum hash table size */ 1440 order = min(order, max_entry_order); 1441 1442 /* Allocate and initialize the endpoint hash table. */ 1443 sctp_ep_hashsize = 64; 1444 sctp_ep_hashtable = 1445 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1446 if (!sctp_ep_hashtable) { 1447 pr_err("Failed endpoint_hash alloc\n"); 1448 status = -ENOMEM; 1449 goto err_ehash_alloc; 1450 } 1451 for (i = 0; i < sctp_ep_hashsize; i++) { 1452 rwlock_init(&sctp_ep_hashtable[i].lock); 1453 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1454 } 1455 1456 /* Allocate and initialize the SCTP port hash table. 1457 * Note that order is initalized to start at the max sized 1458 * table we want to support. If we can't get that many pages 1459 * reduce the order and try again 1460 */ 1461 do { 1462 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1463 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1464 } while (!sctp_port_hashtable && --order > 0); 1465 1466 if (!sctp_port_hashtable) { 1467 pr_err("Failed bind hash alloc\n"); 1468 status = -ENOMEM; 1469 goto err_bhash_alloc; 1470 } 1471 1472 /* Now compute the number of entries that will fit in the 1473 * port hash space we allocated 1474 */ 1475 num_entries = (1UL << order) * PAGE_SIZE / 1476 sizeof(struct sctp_bind_hashbucket); 1477 1478 /* And finish by rounding it down to the nearest power of two 1479 * this wastes some memory of course, but its needed because 1480 * the hash function operates based on the assumption that 1481 * that the number of entries is a power of two 1482 */ 1483 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1484 1485 for (i = 0; i < sctp_port_hashsize; i++) { 1486 spin_lock_init(&sctp_port_hashtable[i].lock); 1487 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1488 } 1489 1490 status = sctp_transport_hashtable_init(); 1491 if (status) 1492 goto err_thash_alloc; 1493 1494 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1495 num_entries); 1496 1497 sctp_sysctl_register(); 1498 1499 INIT_LIST_HEAD(&sctp_address_families); 1500 sctp_v4_pf_init(); 1501 sctp_v6_pf_init(); 1502 sctp_sched_ops_init(); 1503 1504 status = register_pernet_subsys(&sctp_defaults_ops); 1505 if (status) 1506 goto err_register_defaults; 1507 1508 status = sctp_v4_protosw_init(); 1509 if (status) 1510 goto err_protosw_init; 1511 1512 status = sctp_v6_protosw_init(); 1513 if (status) 1514 goto err_v6_protosw_init; 1515 1516 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1517 if (status) 1518 goto err_register_ctrlsock; 1519 1520 status = sctp_v4_add_protocol(); 1521 if (status) 1522 goto err_add_protocol; 1523 1524 /* Register SCTP with inet6 layer. */ 1525 status = sctp_v6_add_protocol(); 1526 if (status) 1527 goto err_v6_add_protocol; 1528 1529 if (sctp_offload_init() < 0) 1530 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1531 1532 out: 1533 return status; 1534 err_v6_add_protocol: 1535 sctp_v4_del_protocol(); 1536 err_add_protocol: 1537 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1538 err_register_ctrlsock: 1539 sctp_v6_protosw_exit(); 1540 err_v6_protosw_init: 1541 sctp_v4_protosw_exit(); 1542 err_protosw_init: 1543 unregister_pernet_subsys(&sctp_defaults_ops); 1544 err_register_defaults: 1545 sctp_v4_pf_exit(); 1546 sctp_v6_pf_exit(); 1547 sctp_sysctl_unregister(); 1548 free_pages((unsigned long)sctp_port_hashtable, 1549 get_order(sctp_port_hashsize * 1550 sizeof(struct sctp_bind_hashbucket))); 1551 err_bhash_alloc: 1552 sctp_transport_hashtable_destroy(); 1553 err_thash_alloc: 1554 kfree(sctp_ep_hashtable); 1555 err_ehash_alloc: 1556 percpu_counter_destroy(&sctp_sockets_allocated); 1557 err_percpu_counter_init: 1558 kmem_cache_destroy(sctp_chunk_cachep); 1559 err_chunk_cachep: 1560 kmem_cache_destroy(sctp_bucket_cachep); 1561 goto out; 1562 } 1563 1564 /* Exit handler for the SCTP protocol. */ 1565 static __exit void sctp_exit(void) 1566 { 1567 /* BUG. This should probably do something useful like clean 1568 * up all the remaining associations and all that memory. 1569 */ 1570 1571 /* Unregister with inet6/inet layers. */ 1572 sctp_v6_del_protocol(); 1573 sctp_v4_del_protocol(); 1574 1575 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1576 1577 /* Free protosw registrations */ 1578 sctp_v6_protosw_exit(); 1579 sctp_v4_protosw_exit(); 1580 1581 unregister_pernet_subsys(&sctp_defaults_ops); 1582 1583 /* Unregister with socket layer. */ 1584 sctp_v6_pf_exit(); 1585 sctp_v4_pf_exit(); 1586 1587 sctp_sysctl_unregister(); 1588 1589 free_pages((unsigned long)sctp_port_hashtable, 1590 get_order(sctp_port_hashsize * 1591 sizeof(struct sctp_bind_hashbucket))); 1592 kfree(sctp_ep_hashtable); 1593 sctp_transport_hashtable_destroy(); 1594 1595 percpu_counter_destroy(&sctp_sockets_allocated); 1596 1597 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1598 1599 kmem_cache_destroy(sctp_chunk_cachep); 1600 kmem_cache_destroy(sctp_bucket_cachep); 1601 } 1602 1603 module_init(sctp_init); 1604 module_exit(sctp_exit); 1605 1606 /* 1607 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1608 */ 1609 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1610 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1611 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1612 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1613 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1614 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1615 MODULE_LICENSE("GPL"); 1616