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/slab.h> 37 #include <net/net_namespace.h> 38 #include <net/protocol.h> 39 #include <net/ip.h> 40 #include <net/ipv6.h> 41 #include <net/route.h> 42 #include <net/sctp/sctp.h> 43 #include <net/addrconf.h> 44 #include <net/inet_common.h> 45 #include <net/inet_ecn.h> 46 #include <net/inet_sock.h> 47 #include <net/udp_tunnel.h> 48 #include <net/inet_dscp.h> 49 50 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024) 51 52 /* Global data structures. */ 53 struct sctp_globals sctp_globals __read_mostly; 54 55 struct idr sctp_assocs_id; 56 DEFINE_SPINLOCK(sctp_assocs_id_lock); 57 58 static struct sctp_pf *sctp_pf_inet6_specific; 59 static struct sctp_pf *sctp_pf_inet_specific; 60 static struct sctp_af *sctp_af_v4_specific; 61 static struct sctp_af *sctp_af_v6_specific; 62 63 struct kmem_cache *sctp_chunk_cachep __read_mostly; 64 struct kmem_cache *sctp_bucket_cachep __read_mostly; 65 66 long sysctl_sctp_mem[3]; 67 int sysctl_sctp_rmem[3]; 68 int sysctl_sctp_wmem[3]; 69 70 /* Private helper to extract ipv4 address and stash them in 71 * the protocol structure. 72 */ 73 static void sctp_v4_copy_addrlist(struct list_head *addrlist, 74 struct net_device *dev) 75 { 76 struct in_device *in_dev; 77 struct in_ifaddr *ifa; 78 struct sctp_sockaddr_entry *addr; 79 80 rcu_read_lock(); 81 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 82 rcu_read_unlock(); 83 return; 84 } 85 86 in_dev_for_each_ifa_rcu(ifa, in_dev) { 87 /* Add the address to the local list. */ 88 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 89 if (addr) { 90 addr->a.v4.sin_family = AF_INET; 91 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 92 addr->valid = 1; 93 INIT_LIST_HEAD(&addr->list); 94 list_add_tail(&addr->list, addrlist); 95 } 96 } 97 98 rcu_read_unlock(); 99 } 100 101 /* Extract our IP addresses from the system and stash them in the 102 * protocol structure. 103 */ 104 static void sctp_get_local_addr_list(struct net *net) 105 { 106 struct net_device *dev; 107 struct list_head *pos; 108 struct sctp_af *af; 109 110 rcu_read_lock(); 111 for_each_netdev_rcu(net, dev) { 112 list_for_each(pos, &sctp_address_families) { 113 af = list_entry(pos, struct sctp_af, list); 114 af->copy_addrlist(&net->sctp.local_addr_list, dev); 115 } 116 } 117 rcu_read_unlock(); 118 } 119 120 /* Free the existing local addresses. */ 121 static void sctp_free_local_addr_list(struct net *net) 122 { 123 struct sctp_sockaddr_entry *addr; 124 struct list_head *pos, *temp; 125 126 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 127 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 128 list_del(pos); 129 kfree(addr); 130 } 131 } 132 133 /* Copy the local addresses which are valid for 'scope' into 'bp'. */ 134 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 135 enum sctp_scope scope, gfp_t gfp, int copy_flags) 136 { 137 struct sctp_sockaddr_entry *addr; 138 union sctp_addr laddr; 139 int error = 0; 140 141 rcu_read_lock(); 142 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 143 if (!addr->valid) 144 continue; 145 if (!sctp_in_scope(net, &addr->a, scope)) 146 continue; 147 148 /* Now that the address is in scope, check to see if 149 * the address type is really supported by the local 150 * sock as well as the remote peer. 151 */ 152 if (addr->a.sa.sa_family == AF_INET && 153 (!(copy_flags & SCTP_ADDR4_ALLOWED) || 154 !(copy_flags & SCTP_ADDR4_PEERSUPP))) 155 continue; 156 if (addr->a.sa.sa_family == AF_INET6 && 157 (!(copy_flags & SCTP_ADDR6_ALLOWED) || 158 !(copy_flags & SCTP_ADDR6_PEERSUPP))) 159 continue; 160 161 laddr = addr->a; 162 /* also works for setting ipv6 address port */ 163 laddr.v4.sin_port = htons(bp->port); 164 if (sctp_bind_addr_state(bp, &laddr) != -1) 165 continue; 166 167 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a), 168 SCTP_ADDR_SRC, GFP_ATOMIC); 169 if (error) 170 break; 171 } 172 173 rcu_read_unlock(); 174 return error; 175 } 176 177 /* Copy over any ip options */ 178 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk) 179 { 180 struct inet_sock *newinet, *inet = inet_sk(sk); 181 struct ip_options_rcu *inet_opt, *newopt = NULL; 182 183 newinet = inet_sk(newsk); 184 185 rcu_read_lock(); 186 inet_opt = rcu_dereference(inet->inet_opt); 187 if (inet_opt) { 188 newopt = sock_kmemdup(newsk, inet_opt, sizeof(*inet_opt) + 189 inet_opt->opt.optlen, GFP_ATOMIC); 190 if (!newopt) 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 memset(sa->sin_zero, 0, sizeof(sa->sin_zero)); 231 } 232 233 /* Initialize an sctp_addr from a socket. */ 234 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 235 { 236 addr->v4.sin_family = AF_INET; 237 addr->v4.sin_port = 0; 238 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 239 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 240 } 241 242 /* Initialize sk->sk_rcv_saddr from sctp_addr. */ 243 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 244 { 245 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 246 } 247 248 /* Initialize sk->sk_daddr from sctp_addr. */ 249 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 250 { 251 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 252 } 253 254 /* Initialize a sctp_addr from an address parameter. */ 255 static bool sctp_v4_from_addr_param(union sctp_addr *addr, 256 union sctp_addr_param *param, 257 __be16 port, int iif) 258 { 259 if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param)) 260 return false; 261 262 addr->v4.sin_family = AF_INET; 263 addr->v4.sin_port = port; 264 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 265 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 266 267 return true; 268 } 269 270 /* Initialize an address parameter from a sctp_addr and return the length 271 * of the address parameter. 272 */ 273 static int sctp_v4_to_addr_param(const union sctp_addr *addr, 274 union sctp_addr_param *param) 275 { 276 int length = sizeof(struct sctp_ipv4addr_param); 277 278 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 279 param->v4.param_hdr.length = htons(length); 280 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 281 282 return length; 283 } 284 285 /* Initialize a sctp_addr from a dst_entry. */ 286 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 287 __be16 port) 288 { 289 saddr->v4.sin_family = AF_INET; 290 saddr->v4.sin_port = port; 291 saddr->v4.sin_addr.s_addr = fl4->saddr; 292 memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero)); 293 } 294 295 /* Compare two addresses exactly. */ 296 static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 297 const union sctp_addr *addr2) 298 { 299 if (addr1->sa.sa_family != addr2->sa.sa_family) 300 return 0; 301 if (addr1->v4.sin_port != addr2->v4.sin_port) 302 return 0; 303 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 304 return 0; 305 306 return 1; 307 } 308 309 /* Initialize addr struct to INADDR_ANY. */ 310 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 311 { 312 addr->v4.sin_family = AF_INET; 313 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 314 addr->v4.sin_port = port; 315 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 316 } 317 318 /* Is this a wildcard address? */ 319 static int sctp_v4_is_any(const union sctp_addr *addr) 320 { 321 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 322 } 323 324 /* This function checks if the address is a valid address to be used for 325 * SCTP binding. 326 * 327 * Output: 328 * Return 0 - If the address is a non-unicast or an illegal address. 329 * Return 1 - If the address is a unicast. 330 */ 331 static int sctp_v4_addr_valid(union sctp_addr *addr, 332 struct sctp_sock *sp, 333 const struct sk_buff *skb) 334 { 335 /* IPv4 addresses not allowed */ 336 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 337 return 0; 338 339 /* Is this a non-unicast address or a unusable SCTP address? */ 340 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 341 return 0; 342 343 /* Is this a broadcast address? */ 344 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 345 return 0; 346 347 return 1; 348 } 349 350 /* Should this be available for binding? */ 351 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 352 { 353 struct sock *sk = &sp->inet.sk; 354 struct net *net = sock_net(sk); 355 int tb_id = RT_TABLE_LOCAL; 356 int ret; 357 358 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ?: tb_id; 359 ret = inet_addr_type_table(net, addr->v4.sin_addr.s_addr, tb_id); 360 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 361 ret != RTN_LOCAL && 362 !inet_test_bit(FREEBIND, sk) && 363 !READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind)) 364 return 0; 365 366 if (ipv6_only_sock(sctp_opt2sk(sp))) 367 return 0; 368 369 return 1; 370 } 371 372 /* Checking the loopback, private and other address scopes as defined in 373 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 374 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 375 * 376 * Level 0 - unusable SCTP addresses 377 * Level 1 - loopback address 378 * Level 2 - link-local addresses 379 * Level 3 - private addresses. 380 * Level 4 - global addresses 381 * For INIT and INIT-ACK address list, let L be the level of 382 * requested destination address, sender and receiver 383 * SHOULD include all of its addresses with level greater 384 * than or equal to L. 385 * 386 * IPv4 scoping can be controlled through sysctl option 387 * net.sctp.addr_scope_policy 388 */ 389 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr) 390 { 391 enum sctp_scope retval; 392 393 /* Check for unusable SCTP addresses. */ 394 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 395 retval = SCTP_SCOPE_UNUSABLE; 396 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 397 retval = SCTP_SCOPE_LOOPBACK; 398 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 399 retval = SCTP_SCOPE_LINK; 400 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 401 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 402 ipv4_is_private_192(addr->v4.sin_addr.s_addr) || 403 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) { 404 retval = SCTP_SCOPE_PRIVATE; 405 } else { 406 retval = SCTP_SCOPE_GLOBAL; 407 } 408 409 return retval; 410 } 411 412 /* Returns a valid dst cache entry for the given source and destination ip 413 * addresses. If an association is passed, trys to get a dst entry with a 414 * source address that matches an address in the bind address list. 415 */ 416 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 417 struct flowi *fl, struct sock *sk) 418 { 419 struct sctp_association *asoc = t->asoc; 420 struct rtable *rt; 421 struct flowi _fl; 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 dscp_t dscp; 429 430 if (t->dscp & SCTP_DSCP_SET_MASK) 431 dscp = inet_dsfield_to_dscp(t->dscp); 432 else 433 dscp = inet_sk_dscp(inet_sk(sk)); 434 435 memset(&_fl, 0x0, sizeof(_fl)); 436 fl4->daddr = daddr->v4.sin_addr.s_addr; 437 fl4->fl4_dport = daddr->v4.sin_port; 438 fl4->flowi4_proto = IPPROTO_SCTP; 439 if (asoc) { 440 fl4->flowi4_tos = inet_dscp_to_dsfield(dscp); 441 fl4->flowi4_scope = ip_sock_rt_scope(asoc->base.sk); 442 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 443 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 444 } 445 if (saddr) { 446 fl4->saddr = saddr->v4.sin_addr.s_addr; 447 if (!fl4->fl4_sport) 448 fl4->fl4_sport = saddr->v4.sin_port; 449 } 450 451 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 452 &fl4->saddr); 453 454 rt = ip_route_output_key(sock_net(sk), fl4); 455 if (!IS_ERR(rt)) { 456 dst = &rt->dst; 457 t->dst = dst; 458 memcpy(fl, &_fl, sizeof(_fl)); 459 } 460 461 /* If there is no association or if a source address is passed, no 462 * more validation is required. 463 */ 464 if (!asoc || saddr) 465 goto out; 466 467 bp = &asoc->base.bind_addr; 468 469 if (dst) { 470 /* Walk through the bind address list and look for a bind 471 * address that matches the source address of the returned dst. 472 */ 473 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 474 rcu_read_lock(); 475 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 476 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 477 (laddr->state != SCTP_ADDR_SRC && 478 !asoc->src_out_of_asoc_ok)) 479 continue; 480 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 481 goto out_unlock; 482 } 483 rcu_read_unlock(); 484 485 /* None of the bound addresses match the source address of the 486 * dst. So release it. 487 */ 488 dst_release(dst); 489 dst = NULL; 490 } 491 492 /* Walk through the bind address list and try to get a dst that 493 * matches a bind address as the source address. 494 */ 495 rcu_read_lock(); 496 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 497 struct net_device *odev; 498 499 if (!laddr->valid) 500 continue; 501 if (laddr->state != SCTP_ADDR_SRC || 502 AF_INET != laddr->a.sa.sa_family) 503 continue; 504 505 fl4->fl4_sport = laddr->a.v4.sin_port; 506 flowi4_update_output(fl4, asoc->base.sk->sk_bound_dev_if, 507 daddr->v4.sin_addr.s_addr, 508 laddr->a.v4.sin_addr.s_addr); 509 510 rt = ip_route_output_key(sock_net(sk), fl4); 511 if (IS_ERR(rt)) 512 continue; 513 514 /* Ensure the src address belongs to the output 515 * interface. 516 */ 517 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr, 518 false); 519 if (!odev || odev->ifindex != fl4->flowi4_oif) { 520 if (!dst) { 521 dst = &rt->dst; 522 t->dst = dst; 523 memcpy(fl, &_fl, sizeof(_fl)); 524 } else { 525 dst_release(&rt->dst); 526 } 527 continue; 528 } 529 530 dst_release(dst); 531 dst = &rt->dst; 532 t->dst = dst; 533 memcpy(fl, &_fl, sizeof(_fl)); 534 break; 535 } 536 537 out_unlock: 538 rcu_read_unlock(); 539 out: 540 if (dst) { 541 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 542 &fl->u.ip4.daddr, &fl->u.ip4.saddr); 543 } else { 544 t->dst = NULL; 545 pr_debug("no route\n"); 546 } 547 } 548 549 /* For v4, the source address is cached in the route entry(dst). So no need 550 * to cache it separately and hence this is an empty routine. 551 */ 552 static void sctp_v4_get_saddr(struct sctp_sock *sk, 553 struct sctp_transport *t, 554 struct flowi *fl) 555 { 556 union sctp_addr *saddr = &t->saddr; 557 struct rtable *rt = dst_rtable(t->dst); 558 559 if (rt) { 560 saddr->v4.sin_family = AF_INET; 561 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 562 } 563 } 564 565 /* What interface did this skb arrive on? */ 566 static int sctp_v4_skb_iif(const struct sk_buff *skb) 567 { 568 return inet_iif(skb); 569 } 570 571 static int sctp_v4_skb_sdif(const struct sk_buff *skb) 572 { 573 return inet_sdif(skb); 574 } 575 576 /* Was this packet marked by Explicit Congestion Notification? */ 577 static int sctp_v4_is_ce(const struct sk_buff *skb) 578 { 579 return INET_ECN_is_ce(ip_hdr(skb)->tos); 580 } 581 582 /* Create and initialize a new sk for the socket returned by accept(). */ 583 static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 584 struct sctp_association *asoc, 585 bool kern) 586 { 587 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 588 sk->sk_prot, kern); 589 struct inet_sock *newinet; 590 591 if (!newsk) 592 goto out; 593 594 sock_init_data(NULL, newsk); 595 596 sctp_copy_sock(newsk, sk, asoc); 597 sock_reset_flag(newsk, SOCK_ZAPPED); 598 599 sctp_v4_copy_ip_options(sk, newsk); 600 601 newinet = inet_sk(newsk); 602 603 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 604 605 if (newsk->sk_prot->init(newsk)) { 606 sk_common_release(newsk); 607 newsk = NULL; 608 } 609 610 out: 611 return newsk; 612 } 613 614 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 615 { 616 /* No address mapping for V4 sockets */ 617 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 618 return sizeof(struct sockaddr_in); 619 } 620 621 /* Dump the v4 addr to the seq file. */ 622 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 623 { 624 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 625 } 626 627 static void sctp_v4_ecn_capable(struct sock *sk) 628 { 629 INET_ECN_xmit(sk); 630 } 631 632 static void sctp_addr_wq_timeout_handler(struct timer_list *t) 633 { 634 struct net *net = timer_container_of(net, t, sctp.addr_wq_timer); 635 struct sctp_sockaddr_entry *addrw, *temp; 636 struct sctp_sock *sp; 637 638 spin_lock_bh(&net->sctp.addr_wq_lock); 639 640 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 641 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 642 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 643 addrw->state, addrw); 644 645 #if IS_ENABLED(CONFIG_IPV6) 646 /* Now we send an ASCONF for each association */ 647 /* Note. we currently don't handle link local IPv6 addressees */ 648 if (addrw->a.sa.sa_family == AF_INET6) { 649 struct in6_addr *in6; 650 651 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 652 IPV6_ADDR_LINKLOCAL) 653 goto free_next; 654 655 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 656 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 657 addrw->state == SCTP_ADDR_NEW) { 658 unsigned long timeo_val; 659 660 pr_debug("%s: this is on DAD, trying %d sec " 661 "later\n", __func__, 662 SCTP_ADDRESS_TICK_DELAY); 663 664 timeo_val = jiffies; 665 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 666 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 667 break; 668 } 669 } 670 #endif 671 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 672 struct sock *sk; 673 674 sk = sctp_opt2sk(sp); 675 /* ignore bound-specific endpoints */ 676 if (!sctp_is_ep_boundall(sk)) 677 continue; 678 bh_lock_sock(sk); 679 if (sctp_asconf_mgmt(sp, addrw) < 0) 680 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 681 bh_unlock_sock(sk); 682 } 683 #if IS_ENABLED(CONFIG_IPV6) 684 free_next: 685 #endif 686 list_del(&addrw->list); 687 kfree(addrw); 688 } 689 spin_unlock_bh(&net->sctp.addr_wq_lock); 690 } 691 692 static void sctp_free_addr_wq(struct net *net) 693 { 694 struct sctp_sockaddr_entry *addrw; 695 struct sctp_sockaddr_entry *temp; 696 697 spin_lock_bh(&net->sctp.addr_wq_lock); 698 timer_delete(&net->sctp.addr_wq_timer); 699 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 700 list_del(&addrw->list); 701 kfree(addrw); 702 } 703 spin_unlock_bh(&net->sctp.addr_wq_lock); 704 } 705 706 /* lookup the entry for the same address in the addr_waitq 707 * sctp_addr_wq MUST be locked 708 */ 709 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 710 struct sctp_sockaddr_entry *addr) 711 { 712 struct sctp_sockaddr_entry *addrw; 713 714 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 715 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 716 continue; 717 if (addrw->a.sa.sa_family == AF_INET) { 718 if (addrw->a.v4.sin_addr.s_addr == 719 addr->a.v4.sin_addr.s_addr) 720 return addrw; 721 } else if (addrw->a.sa.sa_family == AF_INET6) { 722 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 723 &addr->a.v6.sin6_addr)) 724 return addrw; 725 } 726 } 727 return NULL; 728 } 729 730 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 731 { 732 struct sctp_sockaddr_entry *addrw; 733 unsigned long timeo_val; 734 735 /* first, we check if an opposite message already exist in the queue. 736 * If we found such message, it is removed. 737 * This operation is a bit stupid, but the DHCP client attaches the 738 * new address after a couple of addition and deletion of that address 739 */ 740 741 spin_lock_bh(&net->sctp.addr_wq_lock); 742 743 /* Avoid searching the queue or modifying it if there are no consumers, 744 * as it can lead to performance degradation if addresses are modified 745 * en-masse. 746 * 747 * If the queue already contains some events, update it anyway to avoid 748 * ugly races between new sessions and new address events. 749 */ 750 if (list_empty(&net->sctp.auto_asconf_splist) && 751 list_empty(&net->sctp.addr_waitq)) { 752 spin_unlock_bh(&net->sctp.addr_wq_lock); 753 return; 754 } 755 756 /* Offsets existing events in addr_wq */ 757 addrw = sctp_addr_wq_lookup(net, addr); 758 if (addrw) { 759 if (addrw->state != cmd) { 760 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 761 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 762 &net->sctp.addr_waitq); 763 764 list_del(&addrw->list); 765 kfree(addrw); 766 } 767 spin_unlock_bh(&net->sctp.addr_wq_lock); 768 return; 769 } 770 771 /* OK, we have to add the new address to the wait queue */ 772 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 773 if (addrw == NULL) { 774 spin_unlock_bh(&net->sctp.addr_wq_lock); 775 return; 776 } 777 addrw->state = cmd; 778 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 779 780 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 781 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 782 783 if (!timer_pending(&net->sctp.addr_wq_timer)) { 784 timeo_val = jiffies; 785 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 786 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 787 } 788 spin_unlock_bh(&net->sctp.addr_wq_lock); 789 } 790 791 /* Event handler for inet address addition/deletion events. 792 * The sctp_local_addr_list needs to be protocted by a spin lock since 793 * multiple notifiers (say IPv4 and IPv6) may be running at the same 794 * time and thus corrupt the list. 795 * The reader side is protected with RCU. 796 */ 797 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 798 void *ptr) 799 { 800 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 801 struct sctp_sockaddr_entry *addr = NULL; 802 struct sctp_sockaddr_entry *temp; 803 struct net *net = dev_net(ifa->ifa_dev->dev); 804 int found = 0; 805 806 switch (ev) { 807 case NETDEV_UP: 808 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 809 if (addr) { 810 addr->a.v4.sin_family = AF_INET; 811 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 812 addr->valid = 1; 813 spin_lock_bh(&net->sctp.local_addr_lock); 814 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 815 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 816 spin_unlock_bh(&net->sctp.local_addr_lock); 817 } 818 break; 819 case NETDEV_DOWN: 820 spin_lock_bh(&net->sctp.local_addr_lock); 821 list_for_each_entry_safe(addr, temp, 822 &net->sctp.local_addr_list, list) { 823 if (addr->a.sa.sa_family == AF_INET && 824 addr->a.v4.sin_addr.s_addr == 825 ifa->ifa_local) { 826 found = 1; 827 addr->valid = 0; 828 list_del_rcu(&addr->list); 829 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 830 break; 831 } 832 } 833 spin_unlock_bh(&net->sctp.local_addr_lock); 834 if (found) 835 kfree_rcu(addr, rcu); 836 break; 837 } 838 839 return NOTIFY_DONE; 840 } 841 842 /* 843 * Initialize the control inode/socket with a control endpoint data 844 * structure. This endpoint is reserved exclusively for the OOTB processing. 845 */ 846 static int sctp_ctl_sock_init(struct net *net) 847 { 848 int err; 849 sa_family_t family = PF_INET; 850 851 if (sctp_get_pf_specific(PF_INET6)) 852 family = PF_INET6; 853 854 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 855 SOCK_SEQPACKET, IPPROTO_SCTP, net); 856 857 /* If IPv6 socket could not be created, try the IPv4 socket */ 858 if (err < 0 && family == PF_INET6) 859 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 860 SOCK_SEQPACKET, IPPROTO_SCTP, 861 net); 862 863 if (err < 0) { 864 pr_err("Failed to create the SCTP control socket\n"); 865 return err; 866 } 867 return 0; 868 } 869 870 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb) 871 { 872 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source; 873 874 skb_set_transport_header(skb, sizeof(struct udphdr)); 875 sctp_rcv(skb); 876 return 0; 877 } 878 879 int sctp_udp_sock_start(struct net *net) 880 { 881 struct udp_tunnel_sock_cfg tuncfg = {NULL}; 882 struct udp_port_cfg udp_conf = {0}; 883 struct socket *sock; 884 int err; 885 886 udp_conf.family = AF_INET; 887 udp_conf.local_ip.s_addr = htonl(INADDR_ANY); 888 udp_conf.local_udp_port = htons(net->sctp.udp_port); 889 err = udp_sock_create(net, &udp_conf, &sock); 890 if (err) { 891 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n"); 892 return err; 893 } 894 895 tuncfg.encap_type = 1; 896 tuncfg.encap_rcv = sctp_udp_rcv; 897 tuncfg.encap_err_lookup = sctp_udp_v4_err; 898 setup_udp_tunnel_sock(net, sock, &tuncfg); 899 net->sctp.udp4_sock = sock->sk; 900 901 #if IS_ENABLED(CONFIG_IPV6) 902 memset(&udp_conf, 0, sizeof(udp_conf)); 903 904 udp_conf.family = AF_INET6; 905 udp_conf.local_ip6 = in6addr_any; 906 udp_conf.local_udp_port = htons(net->sctp.udp_port); 907 udp_conf.use_udp6_rx_checksums = true; 908 udp_conf.ipv6_v6only = true; 909 err = udp_sock_create(net, &udp_conf, &sock); 910 if (err) { 911 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n"); 912 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 913 net->sctp.udp4_sock = NULL; 914 return err; 915 } 916 917 tuncfg.encap_type = 1; 918 tuncfg.encap_rcv = sctp_udp_rcv; 919 tuncfg.encap_err_lookup = sctp_udp_v6_err; 920 setup_udp_tunnel_sock(net, sock, &tuncfg); 921 net->sctp.udp6_sock = sock->sk; 922 #endif 923 924 return 0; 925 } 926 927 void sctp_udp_sock_stop(struct net *net) 928 { 929 if (net->sctp.udp4_sock) { 930 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 931 net->sctp.udp4_sock = NULL; 932 } 933 if (net->sctp.udp6_sock) { 934 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket); 935 net->sctp.udp6_sock = NULL; 936 } 937 } 938 939 /* Register address family specific functions. */ 940 int sctp_register_af(struct sctp_af *af) 941 { 942 switch (af->sa_family) { 943 case AF_INET: 944 if (sctp_af_v4_specific) 945 return 0; 946 sctp_af_v4_specific = af; 947 break; 948 case AF_INET6: 949 if (sctp_af_v6_specific) 950 return 0; 951 sctp_af_v6_specific = af; 952 break; 953 default: 954 return 0; 955 } 956 957 INIT_LIST_HEAD(&af->list); 958 list_add_tail(&af->list, &sctp_address_families); 959 return 1; 960 } 961 962 /* Get the table of functions for manipulating a particular address 963 * family. 964 */ 965 struct sctp_af *sctp_get_af_specific(sa_family_t family) 966 { 967 switch (family) { 968 case AF_INET: 969 return sctp_af_v4_specific; 970 case AF_INET6: 971 return sctp_af_v6_specific; 972 default: 973 return NULL; 974 } 975 } 976 977 /* Common code to initialize a AF_INET msg_name. */ 978 static void sctp_inet_msgname(char *msgname, int *addr_len) 979 { 980 struct sockaddr_in *sin; 981 982 sin = (struct sockaddr_in *)msgname; 983 *addr_len = sizeof(struct sockaddr_in); 984 sin->sin_family = AF_INET; 985 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 986 } 987 988 /* Copy the primary address of the peer primary address as the msg_name. */ 989 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 990 int *addr_len) 991 { 992 struct sockaddr_in *sin, *sinfrom; 993 994 if (msgname) { 995 struct sctp_association *asoc; 996 997 asoc = event->asoc; 998 sctp_inet_msgname(msgname, addr_len); 999 sin = (struct sockaddr_in *)msgname; 1000 sinfrom = &asoc->peer.primary_addr.v4; 1001 sin->sin_port = htons(asoc->peer.port); 1002 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 1003 } 1004 } 1005 1006 /* Initialize and copy out a msgname from an inbound skb. */ 1007 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 1008 { 1009 if (msgname) { 1010 struct sctphdr *sh = sctp_hdr(skb); 1011 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 1012 1013 sctp_inet_msgname(msgname, len); 1014 sin->sin_port = sh->source; 1015 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 1016 } 1017 } 1018 1019 /* Do we support this AF? */ 1020 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 1021 { 1022 /* PF_INET only supports AF_INET addresses. */ 1023 return AF_INET == family; 1024 } 1025 1026 /* Address matching with wildcards allowed. */ 1027 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 1028 const union sctp_addr *addr2, 1029 struct sctp_sock *opt) 1030 { 1031 /* PF_INET only supports AF_INET addresses. */ 1032 if (addr1->sa.sa_family != addr2->sa.sa_family) 1033 return 0; 1034 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 1035 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 1036 return 1; 1037 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 1038 return 1; 1039 1040 return 0; 1041 } 1042 1043 /* Verify that provided sockaddr looks bindable. Common verification has 1044 * already been taken care of. 1045 */ 1046 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 1047 { 1048 return sctp_v4_available(addr, opt); 1049 } 1050 1051 /* Verify that sockaddr looks sendable. Common verification has already 1052 * been taken care of. 1053 */ 1054 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 1055 { 1056 return 1; 1057 } 1058 1059 /* Fill in Supported Address Type information for INIT and INIT-ACK 1060 * chunks. Returns number of addresses supported. 1061 */ 1062 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 1063 __be16 *types) 1064 { 1065 types[0] = SCTP_PARAM_IPV4_ADDRESS; 1066 return 1; 1067 } 1068 1069 /* Wrapper routine that calls the ip transmit routine. */ 1070 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t) 1071 { 1072 struct dst_entry *dst = dst_clone(t->dst); 1073 struct flowi4 *fl4 = &t->fl.u.ip4; 1074 struct sock *sk = skb->sk; 1075 struct inet_sock *inet = inet_sk(sk); 1076 __u8 dscp = READ_ONCE(inet->tos); 1077 __be16 df = 0; 1078 1079 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 1080 skb->len, &fl4->saddr, &fl4->daddr); 1081 1082 if (t->dscp & SCTP_DSCP_SET_MASK) 1083 dscp = t->dscp & SCTP_DSCP_VAL_MASK; 1084 1085 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO 1086 : IP_PMTUDISC_DONT; 1087 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS); 1088 1089 if (!t->encap_port || !sctp_sk(sk)->udp_port) { 1090 skb_dst_set(skb, dst); 1091 return __ip_queue_xmit(sk, skb, &t->fl, dscp); 1092 } 1093 1094 if (skb_is_gso(skb)) 1095 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM; 1096 1097 if (ip_dont_fragment(sk, dst) && !skb->ignore_df) 1098 df = htons(IP_DF); 1099 1100 skb->encapsulation = 1; 1101 skb_reset_inner_mac_header(skb); 1102 skb_reset_inner_transport_header(skb); 1103 skb_set_inner_ipproto(skb, IPPROTO_SCTP); 1104 udp_tunnel_xmit_skb(dst_rtable(dst), sk, skb, fl4->saddr, 1105 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df, 1106 sctp_sk(sk)->udp_port, t->encap_port, false, false); 1107 return 0; 1108 } 1109 1110 static struct sctp_af sctp_af_inet; 1111 1112 static struct sctp_pf sctp_pf_inet = { 1113 .event_msgname = sctp_inet_event_msgname, 1114 .skb_msgname = sctp_inet_skb_msgname, 1115 .af_supported = sctp_inet_af_supported, 1116 .cmp_addr = sctp_inet_cmp_addr, 1117 .bind_verify = sctp_inet_bind_verify, 1118 .send_verify = sctp_inet_send_verify, 1119 .supported_addrs = sctp_inet_supported_addrs, 1120 .create_accept_sk = sctp_v4_create_accept_sk, 1121 .addr_to_user = sctp_v4_addr_to_user, 1122 .to_sk_saddr = sctp_v4_to_sk_saddr, 1123 .to_sk_daddr = sctp_v4_to_sk_daddr, 1124 .copy_ip_options = sctp_v4_copy_ip_options, 1125 .af = &sctp_af_inet 1126 }; 1127 1128 /* Notifier for inetaddr addition/deletion events. */ 1129 static struct notifier_block sctp_inetaddr_notifier = { 1130 .notifier_call = sctp_inetaddr_event, 1131 }; 1132 1133 /* Socket operations. */ 1134 static const struct proto_ops inet_seqpacket_ops = { 1135 .family = PF_INET, 1136 .owner = THIS_MODULE, 1137 .release = inet_release, /* Needs to be wrapped... */ 1138 .bind = inet_bind, 1139 .connect = sctp_inet_connect, 1140 .socketpair = sock_no_socketpair, 1141 .accept = inet_accept, 1142 .getname = inet_getname, /* Semantics are different. */ 1143 .poll = sctp_poll, 1144 .ioctl = inet_ioctl, 1145 .gettstamp = sock_gettstamp, 1146 .listen = sctp_inet_listen, 1147 .shutdown = inet_shutdown, /* Looks harmless. */ 1148 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1149 .getsockopt = sock_common_getsockopt, 1150 .sendmsg = inet_sendmsg, 1151 .recvmsg = inet_recvmsg, 1152 .mmap = sock_no_mmap, 1153 }; 1154 1155 /* Registration with AF_INET family. */ 1156 static struct inet_protosw sctp_seqpacket_protosw = { 1157 .type = SOCK_SEQPACKET, 1158 .protocol = IPPROTO_SCTP, 1159 .prot = &sctp_prot, 1160 .ops = &inet_seqpacket_ops, 1161 .flags = SCTP_PROTOSW_FLAG 1162 }; 1163 static struct inet_protosw sctp_stream_protosw = { 1164 .type = SOCK_STREAM, 1165 .protocol = IPPROTO_SCTP, 1166 .prot = &sctp_prot, 1167 .ops = &inet_seqpacket_ops, 1168 .flags = SCTP_PROTOSW_FLAG 1169 }; 1170 1171 static int sctp4_rcv(struct sk_buff *skb) 1172 { 1173 SCTP_INPUT_CB(skb)->encap_port = 0; 1174 return sctp_rcv(skb); 1175 } 1176 1177 /* Register with IP layer. */ 1178 static const struct net_protocol sctp_protocol = { 1179 .handler = sctp4_rcv, 1180 .err_handler = sctp_v4_err, 1181 .no_policy = 1, 1182 .icmp_strict_tag_validation = 1, 1183 }; 1184 1185 /* IPv4 address related functions. */ 1186 static struct sctp_af sctp_af_inet = { 1187 .sa_family = AF_INET, 1188 .sctp_xmit = sctp_v4_xmit, 1189 .setsockopt = ip_setsockopt, 1190 .getsockopt = ip_getsockopt, 1191 .get_dst = sctp_v4_get_dst, 1192 .get_saddr = sctp_v4_get_saddr, 1193 .copy_addrlist = sctp_v4_copy_addrlist, 1194 .from_skb = sctp_v4_from_skb, 1195 .from_sk = sctp_v4_from_sk, 1196 .from_addr_param = sctp_v4_from_addr_param, 1197 .to_addr_param = sctp_v4_to_addr_param, 1198 .cmp_addr = sctp_v4_cmp_addr, 1199 .addr_valid = sctp_v4_addr_valid, 1200 .inaddr_any = sctp_v4_inaddr_any, 1201 .is_any = sctp_v4_is_any, 1202 .available = sctp_v4_available, 1203 .scope = sctp_v4_scope, 1204 .skb_iif = sctp_v4_skb_iif, 1205 .skb_sdif = sctp_v4_skb_sdif, 1206 .is_ce = sctp_v4_is_ce, 1207 .seq_dump_addr = sctp_v4_seq_dump_addr, 1208 .ecn_capable = sctp_v4_ecn_capable, 1209 .net_header_len = sizeof(struct iphdr), 1210 .sockaddr_len = sizeof(struct sockaddr_in), 1211 .ip_options_len = sctp_v4_ip_options_len, 1212 }; 1213 1214 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1215 { 1216 switch (family) { 1217 case PF_INET: 1218 return sctp_pf_inet_specific; 1219 case PF_INET6: 1220 return sctp_pf_inet6_specific; 1221 default: 1222 return NULL; 1223 } 1224 } 1225 1226 /* Register the PF specific function table. */ 1227 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1228 { 1229 switch (family) { 1230 case PF_INET: 1231 if (sctp_pf_inet_specific) 1232 return 0; 1233 sctp_pf_inet_specific = pf; 1234 break; 1235 case PF_INET6: 1236 if (sctp_pf_inet6_specific) 1237 return 0; 1238 sctp_pf_inet6_specific = pf; 1239 break; 1240 default: 1241 return 0; 1242 } 1243 return 1; 1244 } 1245 1246 static inline int init_sctp_mibs(struct net *net) 1247 { 1248 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1249 if (!net->sctp.sctp_statistics) 1250 return -ENOMEM; 1251 return 0; 1252 } 1253 1254 static inline void cleanup_sctp_mibs(struct net *net) 1255 { 1256 free_percpu(net->sctp.sctp_statistics); 1257 } 1258 1259 static void sctp_v4_pf_init(void) 1260 { 1261 /* Initialize the SCTP specific PF functions. */ 1262 sctp_register_pf(&sctp_pf_inet, PF_INET); 1263 sctp_register_af(&sctp_af_inet); 1264 } 1265 1266 static void sctp_v4_pf_exit(void) 1267 { 1268 list_del(&sctp_af_inet.list); 1269 } 1270 1271 static int sctp_v4_protosw_init(void) 1272 { 1273 int rc; 1274 1275 rc = proto_register(&sctp_prot, 1); 1276 if (rc) 1277 return rc; 1278 1279 /* Register SCTP(UDP and TCP style) with socket layer. */ 1280 inet_register_protosw(&sctp_seqpacket_protosw); 1281 inet_register_protosw(&sctp_stream_protosw); 1282 1283 return 0; 1284 } 1285 1286 static void sctp_v4_protosw_exit(void) 1287 { 1288 inet_unregister_protosw(&sctp_stream_protosw); 1289 inet_unregister_protosw(&sctp_seqpacket_protosw); 1290 proto_unregister(&sctp_prot); 1291 } 1292 1293 static int sctp_v4_add_protocol(void) 1294 { 1295 /* Register notifier for inet address additions/deletions. */ 1296 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1297 1298 /* Register SCTP with inet layer. */ 1299 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1300 return -EAGAIN; 1301 1302 return 0; 1303 } 1304 1305 static void sctp_v4_del_protocol(void) 1306 { 1307 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1308 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1309 } 1310 1311 static int __net_init sctp_defaults_init(struct net *net) 1312 { 1313 int status; 1314 1315 /* 1316 * 14. Suggested SCTP Protocol Parameter Values 1317 */ 1318 /* The following protocol parameters are RECOMMENDED: */ 1319 /* RTO.Initial - 3 seconds */ 1320 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1321 /* RTO.Min - 1 second */ 1322 net->sctp.rto_min = SCTP_RTO_MIN; 1323 /* RTO.Max - 60 seconds */ 1324 net->sctp.rto_max = SCTP_RTO_MAX; 1325 /* RTO.Alpha - 1/8 */ 1326 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1327 /* RTO.Beta - 1/4 */ 1328 net->sctp.rto_beta = SCTP_RTO_BETA; 1329 1330 /* Valid.Cookie.Life - 60 seconds */ 1331 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1332 1333 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1334 net->sctp.cookie_preserve_enable = 1; 1335 1336 /* Default sctp sockets to use md5 as their hmac alg */ 1337 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1338 net->sctp.sctp_hmac_alg = "md5"; 1339 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1340 net->sctp.sctp_hmac_alg = "sha1"; 1341 #else 1342 net->sctp.sctp_hmac_alg = NULL; 1343 #endif 1344 1345 /* Max.Burst - 4 */ 1346 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1347 1348 /* Disable of Primary Path Switchover by default */ 1349 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX; 1350 1351 /* Enable pf state by default */ 1352 net->sctp.pf_enable = 1; 1353 1354 /* Ignore pf exposure feature by default */ 1355 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET; 1356 1357 /* Association.Max.Retrans - 10 attempts 1358 * Path.Max.Retrans - 5 attempts (per destination address) 1359 * Max.Init.Retransmits - 8 attempts 1360 */ 1361 net->sctp.max_retrans_association = 10; 1362 net->sctp.max_retrans_path = 5; 1363 net->sctp.max_retrans_init = 8; 1364 1365 /* Sendbuffer growth - do per-socket accounting */ 1366 net->sctp.sndbuf_policy = 0; 1367 1368 /* Rcvbuffer growth - do per-socket accounting */ 1369 net->sctp.rcvbuf_policy = 0; 1370 1371 /* HB.interval - 30 seconds */ 1372 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1373 1374 /* delayed SACK timeout */ 1375 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1376 1377 /* Disable ADDIP by default. */ 1378 net->sctp.addip_enable = 0; 1379 net->sctp.addip_noauth = 0; 1380 net->sctp.default_auto_asconf = 0; 1381 1382 /* Enable PR-SCTP by default. */ 1383 net->sctp.prsctp_enable = 1; 1384 1385 /* Disable RECONF by default. */ 1386 net->sctp.reconf_enable = 0; 1387 1388 /* Disable AUTH by default. */ 1389 net->sctp.auth_enable = 0; 1390 1391 /* Enable ECN by default. */ 1392 net->sctp.ecn_enable = 1; 1393 1394 /* Set UDP tunneling listening port to 0 by default */ 1395 net->sctp.udp_port = 0; 1396 1397 /* Set remote encap port to 0 by default */ 1398 net->sctp.encap_port = 0; 1399 1400 /* Set SCOPE policy to enabled */ 1401 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1402 1403 /* Set the default rwnd update threshold */ 1404 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1405 1406 /* Initialize maximum autoclose timeout. */ 1407 net->sctp.max_autoclose = INT_MAX / HZ; 1408 1409 #ifdef CONFIG_NET_L3_MASTER_DEV 1410 net->sctp.l3mdev_accept = 1; 1411 #endif 1412 1413 status = sctp_sysctl_net_register(net); 1414 if (status) 1415 goto err_sysctl_register; 1416 1417 /* Allocate and initialise sctp mibs. */ 1418 status = init_sctp_mibs(net); 1419 if (status) 1420 goto err_init_mibs; 1421 1422 #ifdef CONFIG_PROC_FS 1423 /* Initialize proc fs directory. */ 1424 status = sctp_proc_init(net); 1425 if (status) 1426 goto err_init_proc; 1427 #endif 1428 1429 sctp_dbg_objcnt_init(net); 1430 1431 /* Initialize the local address list. */ 1432 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1433 spin_lock_init(&net->sctp.local_addr_lock); 1434 sctp_get_local_addr_list(net); 1435 1436 /* Initialize the address event list */ 1437 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1438 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1439 spin_lock_init(&net->sctp.addr_wq_lock); 1440 net->sctp.addr_wq_timer.expires = 0; 1441 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1442 1443 return 0; 1444 1445 #ifdef CONFIG_PROC_FS 1446 err_init_proc: 1447 cleanup_sctp_mibs(net); 1448 #endif 1449 err_init_mibs: 1450 sctp_sysctl_net_unregister(net); 1451 err_sysctl_register: 1452 return status; 1453 } 1454 1455 static void __net_exit sctp_defaults_exit(struct net *net) 1456 { 1457 /* Free the local address list */ 1458 sctp_free_addr_wq(net); 1459 sctp_free_local_addr_list(net); 1460 1461 #ifdef CONFIG_PROC_FS 1462 remove_proc_subtree("sctp", net->proc_net); 1463 net->sctp.proc_net_sctp = NULL; 1464 #endif 1465 cleanup_sctp_mibs(net); 1466 sctp_sysctl_net_unregister(net); 1467 } 1468 1469 static struct pernet_operations sctp_defaults_ops = { 1470 .init = sctp_defaults_init, 1471 .exit = sctp_defaults_exit, 1472 }; 1473 1474 static int __net_init sctp_ctrlsock_init(struct net *net) 1475 { 1476 int status; 1477 1478 /* Initialize the control inode/socket for handling OOTB packets. */ 1479 status = sctp_ctl_sock_init(net); 1480 if (status) 1481 pr_err("Failed to initialize the SCTP control sock\n"); 1482 1483 return status; 1484 } 1485 1486 static void __net_exit sctp_ctrlsock_exit(struct net *net) 1487 { 1488 /* Free the control endpoint. */ 1489 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1490 } 1491 1492 static struct pernet_operations sctp_ctrlsock_ops = { 1493 .init = sctp_ctrlsock_init, 1494 .exit = sctp_ctrlsock_exit, 1495 }; 1496 1497 /* Initialize the universe into something sensible. */ 1498 static __init int sctp_init(void) 1499 { 1500 unsigned long nr_pages = totalram_pages(); 1501 unsigned long limit; 1502 unsigned long goal; 1503 int max_entry_order; 1504 int num_entries; 1505 int max_share; 1506 int status; 1507 int order; 1508 int i; 1509 1510 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1511 1512 /* Allocate bind_bucket and chunk caches. */ 1513 status = -ENOBUFS; 1514 sctp_bucket_cachep = KMEM_CACHE(sctp_bind_bucket, SLAB_HWCACHE_ALIGN); 1515 if (!sctp_bucket_cachep) 1516 goto out; 1517 1518 sctp_chunk_cachep = KMEM_CACHE(sctp_chunk, SLAB_HWCACHE_ALIGN); 1519 if (!sctp_chunk_cachep) 1520 goto err_chunk_cachep; 1521 1522 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1523 if (status) 1524 goto err_percpu_counter_init; 1525 1526 /* Implementation specific variables. */ 1527 1528 /* Initialize default stream count setup information. */ 1529 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1530 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1531 1532 /* Initialize handle used for association ids. */ 1533 idr_init(&sctp_assocs_id); 1534 1535 limit = nr_free_buffer_pages() / 8; 1536 limit = max(limit, 128UL); 1537 sysctl_sctp_mem[0] = limit / 4 * 3; 1538 sysctl_sctp_mem[1] = limit; 1539 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1540 1541 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1542 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1543 max_share = min(4UL*1024*1024, limit); 1544 1545 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */ 1546 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1547 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1548 1549 sysctl_sctp_wmem[0] = PAGE_SIZE; 1550 sysctl_sctp_wmem[1] = 16*1024; 1551 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1552 1553 /* Size and allocate the association hash table. 1554 * The methodology is similar to that of the tcp hash tables. 1555 * Though not identical. Start by getting a goal size 1556 */ 1557 if (nr_pages >= (128 * 1024)) 1558 goal = nr_pages >> (22 - PAGE_SHIFT); 1559 else 1560 goal = nr_pages >> (24 - PAGE_SHIFT); 1561 1562 /* Then compute the page order for said goal */ 1563 order = get_order(goal); 1564 1565 /* Now compute the required page order for the maximum sized table we 1566 * want to create 1567 */ 1568 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1569 sizeof(struct sctp_bind_hashbucket)); 1570 1571 /* Limit the page order by that maximum hash table size */ 1572 order = min(order, max_entry_order); 1573 1574 /* Allocate and initialize the endpoint hash table. */ 1575 sctp_ep_hashsize = 64; 1576 sctp_ep_hashtable = 1577 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1578 if (!sctp_ep_hashtable) { 1579 pr_err("Failed endpoint_hash alloc\n"); 1580 status = -ENOMEM; 1581 goto err_ehash_alloc; 1582 } 1583 for (i = 0; i < sctp_ep_hashsize; i++) { 1584 rwlock_init(&sctp_ep_hashtable[i].lock); 1585 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1586 } 1587 1588 /* Allocate and initialize the SCTP port hash table. 1589 * Note that order is initalized to start at the max sized 1590 * table we want to support. If we can't get that many pages 1591 * reduce the order and try again 1592 */ 1593 do { 1594 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1595 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1596 } while (!sctp_port_hashtable && --order > 0); 1597 1598 if (!sctp_port_hashtable) { 1599 pr_err("Failed bind hash alloc\n"); 1600 status = -ENOMEM; 1601 goto err_bhash_alloc; 1602 } 1603 1604 /* Now compute the number of entries that will fit in the 1605 * port hash space we allocated 1606 */ 1607 num_entries = (1UL << order) * PAGE_SIZE / 1608 sizeof(struct sctp_bind_hashbucket); 1609 1610 /* And finish by rounding it down to the nearest power of two. 1611 * This wastes some memory of course, but it's needed because 1612 * the hash function operates based on the assumption that 1613 * the number of entries is a power of two. 1614 */ 1615 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1616 1617 for (i = 0; i < sctp_port_hashsize; i++) { 1618 spin_lock_init(&sctp_port_hashtable[i].lock); 1619 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1620 } 1621 1622 status = sctp_transport_hashtable_init(); 1623 if (status) 1624 goto err_thash_alloc; 1625 1626 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1627 num_entries); 1628 1629 sctp_sysctl_register(); 1630 1631 INIT_LIST_HEAD(&sctp_address_families); 1632 sctp_v4_pf_init(); 1633 sctp_v6_pf_init(); 1634 sctp_sched_ops_init(); 1635 1636 status = register_pernet_subsys(&sctp_defaults_ops); 1637 if (status) 1638 goto err_register_defaults; 1639 1640 status = sctp_v4_protosw_init(); 1641 if (status) 1642 goto err_protosw_init; 1643 1644 status = sctp_v6_protosw_init(); 1645 if (status) 1646 goto err_v6_protosw_init; 1647 1648 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1649 if (status) 1650 goto err_register_ctrlsock; 1651 1652 status = sctp_v4_add_protocol(); 1653 if (status) 1654 goto err_add_protocol; 1655 1656 /* Register SCTP with inet6 layer. */ 1657 status = sctp_v6_add_protocol(); 1658 if (status) 1659 goto err_v6_add_protocol; 1660 1661 if (sctp_offload_init() < 0) 1662 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1663 1664 out: 1665 return status; 1666 err_v6_add_protocol: 1667 sctp_v4_del_protocol(); 1668 err_add_protocol: 1669 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1670 err_register_ctrlsock: 1671 sctp_v6_protosw_exit(); 1672 err_v6_protosw_init: 1673 sctp_v4_protosw_exit(); 1674 err_protosw_init: 1675 unregister_pernet_subsys(&sctp_defaults_ops); 1676 err_register_defaults: 1677 sctp_v4_pf_exit(); 1678 sctp_v6_pf_exit(); 1679 sctp_sysctl_unregister(); 1680 free_pages((unsigned long)sctp_port_hashtable, 1681 get_order(sctp_port_hashsize * 1682 sizeof(struct sctp_bind_hashbucket))); 1683 err_bhash_alloc: 1684 sctp_transport_hashtable_destroy(); 1685 err_thash_alloc: 1686 kfree(sctp_ep_hashtable); 1687 err_ehash_alloc: 1688 percpu_counter_destroy(&sctp_sockets_allocated); 1689 err_percpu_counter_init: 1690 kmem_cache_destroy(sctp_chunk_cachep); 1691 err_chunk_cachep: 1692 kmem_cache_destroy(sctp_bucket_cachep); 1693 goto out; 1694 } 1695 1696 /* Exit handler for the SCTP protocol. */ 1697 static __exit void sctp_exit(void) 1698 { 1699 /* BUG. This should probably do something useful like clean 1700 * up all the remaining associations and all that memory. 1701 */ 1702 1703 /* Unregister with inet6/inet layers. */ 1704 sctp_v6_del_protocol(); 1705 sctp_v4_del_protocol(); 1706 1707 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1708 1709 /* Free protosw registrations */ 1710 sctp_v6_protosw_exit(); 1711 sctp_v4_protosw_exit(); 1712 1713 unregister_pernet_subsys(&sctp_defaults_ops); 1714 1715 /* Unregister with socket layer. */ 1716 sctp_v6_pf_exit(); 1717 sctp_v4_pf_exit(); 1718 1719 sctp_sysctl_unregister(); 1720 1721 free_pages((unsigned long)sctp_port_hashtable, 1722 get_order(sctp_port_hashsize * 1723 sizeof(struct sctp_bind_hashbucket))); 1724 kfree(sctp_ep_hashtable); 1725 sctp_transport_hashtable_destroy(); 1726 1727 percpu_counter_destroy(&sctp_sockets_allocated); 1728 1729 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1730 1731 kmem_cache_destroy(sctp_chunk_cachep); 1732 kmem_cache_destroy(sctp_bucket_cachep); 1733 } 1734 1735 module_init(sctp_init); 1736 module_exit(sctp_exit); 1737 1738 /* 1739 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1740 */ 1741 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1742 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1743 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1744 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1745 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1746 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1747 MODULE_LICENSE("GPL"); 1748