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/udp_tunnel.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_ALLOWED) || 152 !(copy_flags & SCTP_ADDR4_PEERSUPP))) 153 continue; 154 if (addr->a.sa.sa_family == AF_INET6 && 155 (!(copy_flags & SCTP_ADDR6_ALLOWED) || 156 !(copy_flags & SCTP_ADDR6_PEERSUPP))) 157 continue; 158 159 laddr = addr->a; 160 /* also works for setting ipv6 address port */ 161 laddr.v4.sin_port = htons(bp->port); 162 if (sctp_bind_addr_state(bp, &laddr) != -1) 163 continue; 164 165 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a), 166 SCTP_ADDR_SRC, GFP_ATOMIC); 167 if (error) 168 break; 169 } 170 171 rcu_read_unlock(); 172 return error; 173 } 174 175 /* Copy over any ip options */ 176 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk) 177 { 178 struct inet_sock *newinet, *inet = inet_sk(sk); 179 struct ip_options_rcu *inet_opt, *newopt = NULL; 180 181 newinet = inet_sk(newsk); 182 183 rcu_read_lock(); 184 inet_opt = rcu_dereference(inet->inet_opt); 185 if (inet_opt) { 186 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) + 187 inet_opt->opt.optlen, GFP_ATOMIC); 188 if (newopt) 189 memcpy(newopt, inet_opt, sizeof(*inet_opt) + 190 inet_opt->opt.optlen); 191 else 192 pr_err("%s: Failed to copy ip options\n", __func__); 193 } 194 RCU_INIT_POINTER(newinet->inet_opt, newopt); 195 rcu_read_unlock(); 196 } 197 198 /* Account for the IP options */ 199 static int sctp_v4_ip_options_len(struct sock *sk) 200 { 201 struct inet_sock *inet = inet_sk(sk); 202 struct ip_options_rcu *inet_opt; 203 int len = 0; 204 205 rcu_read_lock(); 206 inet_opt = rcu_dereference(inet->inet_opt); 207 if (inet_opt) 208 len = inet_opt->opt.optlen; 209 210 rcu_read_unlock(); 211 return len; 212 } 213 214 /* Initialize a sctp_addr from in incoming skb. */ 215 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 216 int is_saddr) 217 { 218 /* Always called on head skb, so this is safe */ 219 struct sctphdr *sh = sctp_hdr(skb); 220 struct sockaddr_in *sa = &addr->v4; 221 222 addr->v4.sin_family = AF_INET; 223 224 if (is_saddr) { 225 sa->sin_port = sh->source; 226 sa->sin_addr.s_addr = ip_hdr(skb)->saddr; 227 } else { 228 sa->sin_port = sh->dest; 229 sa->sin_addr.s_addr = ip_hdr(skb)->daddr; 230 } 231 memset(sa->sin_zero, 0, sizeof(sa->sin_zero)); 232 } 233 234 /* Initialize an sctp_addr from a socket. */ 235 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 236 { 237 addr->v4.sin_family = AF_INET; 238 addr->v4.sin_port = 0; 239 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 240 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 241 } 242 243 /* Initialize sk->sk_rcv_saddr from sctp_addr. */ 244 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 245 { 246 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 247 } 248 249 /* Initialize sk->sk_daddr from sctp_addr. */ 250 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 251 { 252 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 253 } 254 255 /* Initialize a sctp_addr from an address parameter. */ 256 static bool sctp_v4_from_addr_param(union sctp_addr *addr, 257 union sctp_addr_param *param, 258 __be16 port, int iif) 259 { 260 if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param)) 261 return false; 262 263 addr->v4.sin_family = AF_INET; 264 addr->v4.sin_port = port; 265 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 266 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 267 268 return true; 269 } 270 271 /* Initialize an address parameter from a sctp_addr and return the length 272 * of the address parameter. 273 */ 274 static int sctp_v4_to_addr_param(const union sctp_addr *addr, 275 union sctp_addr_param *param) 276 { 277 int length = sizeof(struct sctp_ipv4addr_param); 278 279 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 280 param->v4.param_hdr.length = htons(length); 281 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 282 283 return length; 284 } 285 286 /* Initialize a sctp_addr from a dst_entry. */ 287 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 288 __be16 port) 289 { 290 saddr->v4.sin_family = AF_INET; 291 saddr->v4.sin_port = port; 292 saddr->v4.sin_addr.s_addr = fl4->saddr; 293 memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero)); 294 } 295 296 /* Compare two addresses exactly. */ 297 static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 298 const union sctp_addr *addr2) 299 { 300 if (addr1->sa.sa_family != addr2->sa.sa_family) 301 return 0; 302 if (addr1->v4.sin_port != addr2->v4.sin_port) 303 return 0; 304 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 305 return 0; 306 307 return 1; 308 } 309 310 /* Initialize addr struct to INADDR_ANY. */ 311 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 312 { 313 addr->v4.sin_family = AF_INET; 314 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 315 addr->v4.sin_port = port; 316 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 317 } 318 319 /* Is this a wildcard address? */ 320 static int sctp_v4_is_any(const union sctp_addr *addr) 321 { 322 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 323 } 324 325 /* This function checks if the address is a valid address to be used for 326 * SCTP binding. 327 * 328 * Output: 329 * Return 0 - If the address is a non-unicast or an illegal address. 330 * Return 1 - If the address is a unicast. 331 */ 332 static int sctp_v4_addr_valid(union sctp_addr *addr, 333 struct sctp_sock *sp, 334 const struct sk_buff *skb) 335 { 336 /* IPv4 addresses not allowed */ 337 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 338 return 0; 339 340 /* Is this a non-unicast address or a unusable SCTP address? */ 341 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 342 return 0; 343 344 /* Is this a broadcast address? */ 345 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 346 return 0; 347 348 return 1; 349 } 350 351 /* Should this be available for binding? */ 352 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 353 { 354 struct sock *sk = &sp->inet.sk; 355 struct net *net = sock_net(sk); 356 int tb_id = RT_TABLE_LOCAL; 357 int ret; 358 359 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ?: tb_id; 360 ret = inet_addr_type_table(net, addr->v4.sin_addr.s_addr, tb_id); 361 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 362 ret != RTN_LOCAL && 363 !sp->inet.freebind && 364 !READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind)) 365 return 0; 366 367 if (ipv6_only_sock(sctp_opt2sk(sp))) 368 return 0; 369 370 return 1; 371 } 372 373 /* Checking the loopback, private and other address scopes as defined in 374 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 375 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 376 * 377 * Level 0 - unusable SCTP addresses 378 * Level 1 - loopback address 379 * Level 2 - link-local addresses 380 * Level 3 - private addresses. 381 * Level 4 - global addresses 382 * For INIT and INIT-ACK address list, let L be the level of 383 * requested destination address, sender and receiver 384 * SHOULD include all of its addresses with level greater 385 * than or equal to L. 386 * 387 * IPv4 scoping can be controlled through sysctl option 388 * net.sctp.addr_scope_policy 389 */ 390 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr) 391 { 392 enum sctp_scope retval; 393 394 /* Check for unusable SCTP addresses. */ 395 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 396 retval = SCTP_SCOPE_UNUSABLE; 397 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 398 retval = SCTP_SCOPE_LOOPBACK; 399 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 400 retval = SCTP_SCOPE_LINK; 401 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 402 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 403 ipv4_is_private_192(addr->v4.sin_addr.s_addr) || 404 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) { 405 retval = SCTP_SCOPE_PRIVATE; 406 } else { 407 retval = SCTP_SCOPE_GLOBAL; 408 } 409 410 return retval; 411 } 412 413 /* Returns a valid dst cache entry for the given source and destination ip 414 * addresses. If an association is passed, trys to get a dst entry with a 415 * source address that matches an address in the bind address list. 416 */ 417 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 418 struct flowi *fl, struct sock *sk) 419 { 420 struct sctp_association *asoc = t->asoc; 421 struct rtable *rt; 422 struct flowi _fl; 423 struct flowi4 *fl4 = &_fl.u.ip4; 424 struct sctp_bind_addr *bp; 425 struct sctp_sockaddr_entry *laddr; 426 struct dst_entry *dst = NULL; 427 union sctp_addr *daddr = &t->ipaddr; 428 union sctp_addr dst_saddr; 429 __u8 tos = inet_sk(sk)->tos; 430 431 if (t->dscp & SCTP_DSCP_SET_MASK) 432 tos = t->dscp & SCTP_DSCP_VAL_MASK; 433 memset(&_fl, 0x0, sizeof(_fl)); 434 fl4->daddr = daddr->v4.sin_addr.s_addr; 435 fl4->fl4_dport = daddr->v4.sin_port; 436 fl4->flowi4_proto = IPPROTO_SCTP; 437 if (asoc) { 438 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos); 439 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 440 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 441 } 442 if (saddr) { 443 fl4->saddr = saddr->v4.sin_addr.s_addr; 444 if (!fl4->fl4_sport) 445 fl4->fl4_sport = saddr->v4.sin_port; 446 } 447 448 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 449 &fl4->saddr); 450 451 rt = ip_route_output_key(sock_net(sk), fl4); 452 if (!IS_ERR(rt)) { 453 dst = &rt->dst; 454 t->dst = dst; 455 memcpy(fl, &_fl, sizeof(_fl)); 456 } 457 458 /* If there is no association or if a source address is passed, no 459 * more validation is required. 460 */ 461 if (!asoc || saddr) 462 goto out; 463 464 bp = &asoc->base.bind_addr; 465 466 if (dst) { 467 /* Walk through the bind address list and look for a bind 468 * address that matches the source address of the returned dst. 469 */ 470 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 471 rcu_read_lock(); 472 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 473 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 474 (laddr->state != SCTP_ADDR_SRC && 475 !asoc->src_out_of_asoc_ok)) 476 continue; 477 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 478 goto out_unlock; 479 } 480 rcu_read_unlock(); 481 482 /* None of the bound addresses match the source address of the 483 * dst. So release it. 484 */ 485 dst_release(dst); 486 dst = NULL; 487 } 488 489 /* Walk through the bind address list and try to get a dst that 490 * matches a bind address as the source address. 491 */ 492 rcu_read_lock(); 493 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 494 struct net_device *odev; 495 496 if (!laddr->valid) 497 continue; 498 if (laddr->state != SCTP_ADDR_SRC || 499 AF_INET != laddr->a.sa.sa_family) 500 continue; 501 502 fl4->fl4_sport = laddr->a.v4.sin_port; 503 flowi4_update_output(fl4, 504 asoc->base.sk->sk_bound_dev_if, 505 RT_CONN_FLAGS_TOS(asoc->base.sk, tos), 506 daddr->v4.sin_addr.s_addr, 507 laddr->a.v4.sin_addr.s_addr); 508 509 rt = ip_route_output_key(sock_net(sk), fl4); 510 if (IS_ERR(rt)) 511 continue; 512 513 /* Ensure the src address belongs to the output 514 * interface. 515 */ 516 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr, 517 false); 518 if (!odev || odev->ifindex != fl4->flowi4_oif) { 519 if (!dst) { 520 dst = &rt->dst; 521 t->dst = dst; 522 memcpy(fl, &_fl, sizeof(_fl)); 523 } else { 524 dst_release(&rt->dst); 525 } 526 continue; 527 } 528 529 dst_release(dst); 530 dst = &rt->dst; 531 t->dst = dst; 532 memcpy(fl, &_fl, sizeof(_fl)); 533 break; 534 } 535 536 out_unlock: 537 rcu_read_unlock(); 538 out: 539 if (dst) { 540 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 541 &fl->u.ip4.daddr, &fl->u.ip4.saddr); 542 } else { 543 t->dst = NULL; 544 pr_debug("no route\n"); 545 } 546 } 547 548 /* For v4, the source address is cached in the route entry(dst). So no need 549 * to cache it separately and hence this is an empty routine. 550 */ 551 static void sctp_v4_get_saddr(struct sctp_sock *sk, 552 struct sctp_transport *t, 553 struct flowi *fl) 554 { 555 union sctp_addr *saddr = &t->saddr; 556 struct rtable *rt = (struct rtable *)t->dst; 557 558 if (rt) { 559 saddr->v4.sin_family = AF_INET; 560 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 561 } 562 } 563 564 /* What interface did this skb arrive on? */ 565 static int sctp_v4_skb_iif(const struct sk_buff *skb) 566 { 567 return inet_iif(skb); 568 } 569 570 static int sctp_v4_skb_sdif(const struct sk_buff *skb) 571 { 572 return inet_sdif(skb); 573 } 574 575 /* Was this packet marked by Explicit Congestion Notification? */ 576 static int sctp_v4_is_ce(const struct sk_buff *skb) 577 { 578 return INET_ECN_is_ce(ip_hdr(skb)->tos); 579 } 580 581 /* Create and initialize a new sk for the socket returned by accept(). */ 582 static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 583 struct sctp_association *asoc, 584 bool kern) 585 { 586 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 587 sk->sk_prot, kern); 588 struct inet_sock *newinet; 589 590 if (!newsk) 591 goto out; 592 593 sock_init_data(NULL, newsk); 594 595 sctp_copy_sock(newsk, sk, asoc); 596 sock_reset_flag(newsk, SOCK_ZAPPED); 597 598 sctp_v4_copy_ip_options(sk, newsk); 599 600 newinet = inet_sk(newsk); 601 602 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 603 604 if (newsk->sk_prot->init(newsk)) { 605 sk_common_release(newsk); 606 newsk = NULL; 607 } 608 609 out: 610 return newsk; 611 } 612 613 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 614 { 615 /* No address mapping for V4 sockets */ 616 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 617 return sizeof(struct sockaddr_in); 618 } 619 620 /* Dump the v4 addr to the seq file. */ 621 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 622 { 623 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 624 } 625 626 static void sctp_v4_ecn_capable(struct sock *sk) 627 { 628 INET_ECN_xmit(sk); 629 } 630 631 static void sctp_addr_wq_timeout_handler(struct timer_list *t) 632 { 633 struct net *net = from_timer(net, t, sctp.addr_wq_timer); 634 struct sctp_sockaddr_entry *addrw, *temp; 635 struct sctp_sock *sp; 636 637 spin_lock_bh(&net->sctp.addr_wq_lock); 638 639 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 640 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 641 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 642 addrw->state, addrw); 643 644 #if IS_ENABLED(CONFIG_IPV6) 645 /* Now we send an ASCONF for each association */ 646 /* Note. we currently don't handle link local IPv6 addressees */ 647 if (addrw->a.sa.sa_family == AF_INET6) { 648 struct in6_addr *in6; 649 650 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 651 IPV6_ADDR_LINKLOCAL) 652 goto free_next; 653 654 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 655 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 656 addrw->state == SCTP_ADDR_NEW) { 657 unsigned long timeo_val; 658 659 pr_debug("%s: this is on DAD, trying %d sec " 660 "later\n", __func__, 661 SCTP_ADDRESS_TICK_DELAY); 662 663 timeo_val = jiffies; 664 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 665 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 666 break; 667 } 668 } 669 #endif 670 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 671 struct sock *sk; 672 673 sk = sctp_opt2sk(sp); 674 /* ignore bound-specific endpoints */ 675 if (!sctp_is_ep_boundall(sk)) 676 continue; 677 bh_lock_sock(sk); 678 if (sctp_asconf_mgmt(sp, addrw) < 0) 679 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 680 bh_unlock_sock(sk); 681 } 682 #if IS_ENABLED(CONFIG_IPV6) 683 free_next: 684 #endif 685 list_del(&addrw->list); 686 kfree(addrw); 687 } 688 spin_unlock_bh(&net->sctp.addr_wq_lock); 689 } 690 691 static void sctp_free_addr_wq(struct net *net) 692 { 693 struct sctp_sockaddr_entry *addrw; 694 struct sctp_sockaddr_entry *temp; 695 696 spin_lock_bh(&net->sctp.addr_wq_lock); 697 del_timer(&net->sctp.addr_wq_timer); 698 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 699 list_del(&addrw->list); 700 kfree(addrw); 701 } 702 spin_unlock_bh(&net->sctp.addr_wq_lock); 703 } 704 705 /* lookup the entry for the same address in the addr_waitq 706 * sctp_addr_wq MUST be locked 707 */ 708 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 709 struct sctp_sockaddr_entry *addr) 710 { 711 struct sctp_sockaddr_entry *addrw; 712 713 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 714 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 715 continue; 716 if (addrw->a.sa.sa_family == AF_INET) { 717 if (addrw->a.v4.sin_addr.s_addr == 718 addr->a.v4.sin_addr.s_addr) 719 return addrw; 720 } else if (addrw->a.sa.sa_family == AF_INET6) { 721 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 722 &addr->a.v6.sin6_addr)) 723 return addrw; 724 } 725 } 726 return NULL; 727 } 728 729 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 730 { 731 struct sctp_sockaddr_entry *addrw; 732 unsigned long timeo_val; 733 734 /* first, we check if an opposite message already exist in the queue. 735 * If we found such message, it is removed. 736 * This operation is a bit stupid, but the DHCP client attaches the 737 * new address after a couple of addition and deletion of that address 738 */ 739 740 spin_lock_bh(&net->sctp.addr_wq_lock); 741 /* Offsets existing events in addr_wq */ 742 addrw = sctp_addr_wq_lookup(net, addr); 743 if (addrw) { 744 if (addrw->state != cmd) { 745 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 746 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 747 &net->sctp.addr_waitq); 748 749 list_del(&addrw->list); 750 kfree(addrw); 751 } 752 spin_unlock_bh(&net->sctp.addr_wq_lock); 753 return; 754 } 755 756 /* OK, we have to add the new address to the wait queue */ 757 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 758 if (addrw == NULL) { 759 spin_unlock_bh(&net->sctp.addr_wq_lock); 760 return; 761 } 762 addrw->state = cmd; 763 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 764 765 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 766 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 767 768 if (!timer_pending(&net->sctp.addr_wq_timer)) { 769 timeo_val = jiffies; 770 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 771 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 772 } 773 spin_unlock_bh(&net->sctp.addr_wq_lock); 774 } 775 776 /* Event handler for inet address addition/deletion events. 777 * The sctp_local_addr_list needs to be protocted by a spin lock since 778 * multiple notifiers (say IPv4 and IPv6) may be running at the same 779 * time and thus corrupt the list. 780 * The reader side is protected with RCU. 781 */ 782 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 783 void *ptr) 784 { 785 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 786 struct sctp_sockaddr_entry *addr = NULL; 787 struct sctp_sockaddr_entry *temp; 788 struct net *net = dev_net(ifa->ifa_dev->dev); 789 int found = 0; 790 791 switch (ev) { 792 case NETDEV_UP: 793 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 794 if (addr) { 795 addr->a.v4.sin_family = AF_INET; 796 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 797 addr->valid = 1; 798 spin_lock_bh(&net->sctp.local_addr_lock); 799 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 800 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 801 spin_unlock_bh(&net->sctp.local_addr_lock); 802 } 803 break; 804 case NETDEV_DOWN: 805 spin_lock_bh(&net->sctp.local_addr_lock); 806 list_for_each_entry_safe(addr, temp, 807 &net->sctp.local_addr_list, list) { 808 if (addr->a.sa.sa_family == AF_INET && 809 addr->a.v4.sin_addr.s_addr == 810 ifa->ifa_local) { 811 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 812 found = 1; 813 addr->valid = 0; 814 list_del_rcu(&addr->list); 815 break; 816 } 817 } 818 spin_unlock_bh(&net->sctp.local_addr_lock); 819 if (found) 820 kfree_rcu(addr, rcu); 821 break; 822 } 823 824 return NOTIFY_DONE; 825 } 826 827 /* 828 * Initialize the control inode/socket with a control endpoint data 829 * structure. This endpoint is reserved exclusively for the OOTB processing. 830 */ 831 static int sctp_ctl_sock_init(struct net *net) 832 { 833 int err; 834 sa_family_t family = PF_INET; 835 836 if (sctp_get_pf_specific(PF_INET6)) 837 family = PF_INET6; 838 839 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 840 SOCK_SEQPACKET, IPPROTO_SCTP, net); 841 842 /* If IPv6 socket could not be created, try the IPv4 socket */ 843 if (err < 0 && family == PF_INET6) 844 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 845 SOCK_SEQPACKET, IPPROTO_SCTP, 846 net); 847 848 if (err < 0) { 849 pr_err("Failed to create the SCTP control socket\n"); 850 return err; 851 } 852 return 0; 853 } 854 855 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb) 856 { 857 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source; 858 859 skb_set_transport_header(skb, sizeof(struct udphdr)); 860 sctp_rcv(skb); 861 return 0; 862 } 863 864 int sctp_udp_sock_start(struct net *net) 865 { 866 struct udp_tunnel_sock_cfg tuncfg = {NULL}; 867 struct udp_port_cfg udp_conf = {0}; 868 struct socket *sock; 869 int err; 870 871 udp_conf.family = AF_INET; 872 udp_conf.local_ip.s_addr = htonl(INADDR_ANY); 873 udp_conf.local_udp_port = htons(net->sctp.udp_port); 874 err = udp_sock_create(net, &udp_conf, &sock); 875 if (err) { 876 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n"); 877 return err; 878 } 879 880 tuncfg.encap_type = 1; 881 tuncfg.encap_rcv = sctp_udp_rcv; 882 tuncfg.encap_err_lookup = sctp_udp_v4_err; 883 setup_udp_tunnel_sock(net, sock, &tuncfg); 884 net->sctp.udp4_sock = sock->sk; 885 886 #if IS_ENABLED(CONFIG_IPV6) 887 memset(&udp_conf, 0, sizeof(udp_conf)); 888 889 udp_conf.family = AF_INET6; 890 udp_conf.local_ip6 = in6addr_any; 891 udp_conf.local_udp_port = htons(net->sctp.udp_port); 892 udp_conf.use_udp6_rx_checksums = true; 893 udp_conf.ipv6_v6only = true; 894 err = udp_sock_create(net, &udp_conf, &sock); 895 if (err) { 896 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n"); 897 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 898 net->sctp.udp4_sock = NULL; 899 return err; 900 } 901 902 tuncfg.encap_type = 1; 903 tuncfg.encap_rcv = sctp_udp_rcv; 904 tuncfg.encap_err_lookup = sctp_udp_v6_err; 905 setup_udp_tunnel_sock(net, sock, &tuncfg); 906 net->sctp.udp6_sock = sock->sk; 907 #endif 908 909 return 0; 910 } 911 912 void sctp_udp_sock_stop(struct net *net) 913 { 914 if (net->sctp.udp4_sock) { 915 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 916 net->sctp.udp4_sock = NULL; 917 } 918 if (net->sctp.udp6_sock) { 919 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket); 920 net->sctp.udp6_sock = NULL; 921 } 922 } 923 924 /* Register address family specific functions. */ 925 int sctp_register_af(struct sctp_af *af) 926 { 927 switch (af->sa_family) { 928 case AF_INET: 929 if (sctp_af_v4_specific) 930 return 0; 931 sctp_af_v4_specific = af; 932 break; 933 case AF_INET6: 934 if (sctp_af_v6_specific) 935 return 0; 936 sctp_af_v6_specific = af; 937 break; 938 default: 939 return 0; 940 } 941 942 INIT_LIST_HEAD(&af->list); 943 list_add_tail(&af->list, &sctp_address_families); 944 return 1; 945 } 946 947 /* Get the table of functions for manipulating a particular address 948 * family. 949 */ 950 struct sctp_af *sctp_get_af_specific(sa_family_t family) 951 { 952 switch (family) { 953 case AF_INET: 954 return sctp_af_v4_specific; 955 case AF_INET6: 956 return sctp_af_v6_specific; 957 default: 958 return NULL; 959 } 960 } 961 962 /* Common code to initialize a AF_INET msg_name. */ 963 static void sctp_inet_msgname(char *msgname, int *addr_len) 964 { 965 struct sockaddr_in *sin; 966 967 sin = (struct sockaddr_in *)msgname; 968 *addr_len = sizeof(struct sockaddr_in); 969 sin->sin_family = AF_INET; 970 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 971 } 972 973 /* Copy the primary address of the peer primary address as the msg_name. */ 974 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 975 int *addr_len) 976 { 977 struct sockaddr_in *sin, *sinfrom; 978 979 if (msgname) { 980 struct sctp_association *asoc; 981 982 asoc = event->asoc; 983 sctp_inet_msgname(msgname, addr_len); 984 sin = (struct sockaddr_in *)msgname; 985 sinfrom = &asoc->peer.primary_addr.v4; 986 sin->sin_port = htons(asoc->peer.port); 987 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 988 } 989 } 990 991 /* Initialize and copy out a msgname from an inbound skb. */ 992 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 993 { 994 if (msgname) { 995 struct sctphdr *sh = sctp_hdr(skb); 996 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 997 998 sctp_inet_msgname(msgname, len); 999 sin->sin_port = sh->source; 1000 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 1001 } 1002 } 1003 1004 /* Do we support this AF? */ 1005 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 1006 { 1007 /* PF_INET only supports AF_INET addresses. */ 1008 return AF_INET == family; 1009 } 1010 1011 /* Address matching with wildcards allowed. */ 1012 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 1013 const union sctp_addr *addr2, 1014 struct sctp_sock *opt) 1015 { 1016 /* PF_INET only supports AF_INET addresses. */ 1017 if (addr1->sa.sa_family != addr2->sa.sa_family) 1018 return 0; 1019 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 1020 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 1021 return 1; 1022 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 1023 return 1; 1024 1025 return 0; 1026 } 1027 1028 /* Verify that provided sockaddr looks bindable. Common verification has 1029 * already been taken care of. 1030 */ 1031 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 1032 { 1033 return sctp_v4_available(addr, opt); 1034 } 1035 1036 /* Verify that sockaddr looks sendable. Common verification has already 1037 * been taken care of. 1038 */ 1039 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 1040 { 1041 return 1; 1042 } 1043 1044 /* Fill in Supported Address Type information for INIT and INIT-ACK 1045 * chunks. Returns number of addresses supported. 1046 */ 1047 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 1048 __be16 *types) 1049 { 1050 types[0] = SCTP_PARAM_IPV4_ADDRESS; 1051 return 1; 1052 } 1053 1054 /* Wrapper routine that calls the ip transmit routine. */ 1055 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t) 1056 { 1057 struct dst_entry *dst = dst_clone(t->dst); 1058 struct flowi4 *fl4 = &t->fl.u.ip4; 1059 struct sock *sk = skb->sk; 1060 struct inet_sock *inet = inet_sk(sk); 1061 __u8 dscp = inet->tos; 1062 __be16 df = 0; 1063 1064 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 1065 skb->len, &fl4->saddr, &fl4->daddr); 1066 1067 if (t->dscp & SCTP_DSCP_SET_MASK) 1068 dscp = t->dscp & SCTP_DSCP_VAL_MASK; 1069 1070 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO 1071 : IP_PMTUDISC_DONT; 1072 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS); 1073 1074 if (!t->encap_port || !sctp_sk(sk)->udp_port) { 1075 skb_dst_set(skb, dst); 1076 return __ip_queue_xmit(sk, skb, &t->fl, dscp); 1077 } 1078 1079 if (skb_is_gso(skb)) 1080 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM; 1081 1082 if (ip_dont_fragment(sk, dst) && !skb->ignore_df) 1083 df = htons(IP_DF); 1084 1085 skb->encapsulation = 1; 1086 skb_reset_inner_mac_header(skb); 1087 skb_reset_inner_transport_header(skb); 1088 skb_set_inner_ipproto(skb, IPPROTO_SCTP); 1089 udp_tunnel_xmit_skb((struct rtable *)dst, sk, skb, fl4->saddr, 1090 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df, 1091 sctp_sk(sk)->udp_port, t->encap_port, false, false); 1092 return 0; 1093 } 1094 1095 static struct sctp_af sctp_af_inet; 1096 1097 static struct sctp_pf sctp_pf_inet = { 1098 .event_msgname = sctp_inet_event_msgname, 1099 .skb_msgname = sctp_inet_skb_msgname, 1100 .af_supported = sctp_inet_af_supported, 1101 .cmp_addr = sctp_inet_cmp_addr, 1102 .bind_verify = sctp_inet_bind_verify, 1103 .send_verify = sctp_inet_send_verify, 1104 .supported_addrs = sctp_inet_supported_addrs, 1105 .create_accept_sk = sctp_v4_create_accept_sk, 1106 .addr_to_user = sctp_v4_addr_to_user, 1107 .to_sk_saddr = sctp_v4_to_sk_saddr, 1108 .to_sk_daddr = sctp_v4_to_sk_daddr, 1109 .copy_ip_options = sctp_v4_copy_ip_options, 1110 .af = &sctp_af_inet 1111 }; 1112 1113 /* Notifier for inetaddr addition/deletion events. */ 1114 static struct notifier_block sctp_inetaddr_notifier = { 1115 .notifier_call = sctp_inetaddr_event, 1116 }; 1117 1118 /* Socket operations. */ 1119 static const struct proto_ops inet_seqpacket_ops = { 1120 .family = PF_INET, 1121 .owner = THIS_MODULE, 1122 .release = inet_release, /* Needs to be wrapped... */ 1123 .bind = inet_bind, 1124 .connect = sctp_inet_connect, 1125 .socketpair = sock_no_socketpair, 1126 .accept = inet_accept, 1127 .getname = inet_getname, /* Semantics are different. */ 1128 .poll = sctp_poll, 1129 .ioctl = inet_ioctl, 1130 .gettstamp = sock_gettstamp, 1131 .listen = sctp_inet_listen, 1132 .shutdown = inet_shutdown, /* Looks harmless. */ 1133 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1134 .getsockopt = sock_common_getsockopt, 1135 .sendmsg = inet_sendmsg, 1136 .recvmsg = inet_recvmsg, 1137 .mmap = sock_no_mmap, 1138 .sendpage = sock_no_sendpage, 1139 }; 1140 1141 /* Registration with AF_INET family. */ 1142 static struct inet_protosw sctp_seqpacket_protosw = { 1143 .type = SOCK_SEQPACKET, 1144 .protocol = IPPROTO_SCTP, 1145 .prot = &sctp_prot, 1146 .ops = &inet_seqpacket_ops, 1147 .flags = SCTP_PROTOSW_FLAG 1148 }; 1149 static struct inet_protosw sctp_stream_protosw = { 1150 .type = SOCK_STREAM, 1151 .protocol = IPPROTO_SCTP, 1152 .prot = &sctp_prot, 1153 .ops = &inet_seqpacket_ops, 1154 .flags = SCTP_PROTOSW_FLAG 1155 }; 1156 1157 static int sctp4_rcv(struct sk_buff *skb) 1158 { 1159 SCTP_INPUT_CB(skb)->encap_port = 0; 1160 return sctp_rcv(skb); 1161 } 1162 1163 /* Register with IP layer. */ 1164 static const struct net_protocol sctp_protocol = { 1165 .handler = sctp4_rcv, 1166 .err_handler = sctp_v4_err, 1167 .no_policy = 1, 1168 .icmp_strict_tag_validation = 1, 1169 }; 1170 1171 /* IPv4 address related functions. */ 1172 static struct sctp_af sctp_af_inet = { 1173 .sa_family = AF_INET, 1174 .sctp_xmit = sctp_v4_xmit, 1175 .setsockopt = ip_setsockopt, 1176 .getsockopt = ip_getsockopt, 1177 .get_dst = sctp_v4_get_dst, 1178 .get_saddr = sctp_v4_get_saddr, 1179 .copy_addrlist = sctp_v4_copy_addrlist, 1180 .from_skb = sctp_v4_from_skb, 1181 .from_sk = sctp_v4_from_sk, 1182 .from_addr_param = sctp_v4_from_addr_param, 1183 .to_addr_param = sctp_v4_to_addr_param, 1184 .cmp_addr = sctp_v4_cmp_addr, 1185 .addr_valid = sctp_v4_addr_valid, 1186 .inaddr_any = sctp_v4_inaddr_any, 1187 .is_any = sctp_v4_is_any, 1188 .available = sctp_v4_available, 1189 .scope = sctp_v4_scope, 1190 .skb_iif = sctp_v4_skb_iif, 1191 .skb_sdif = sctp_v4_skb_sdif, 1192 .is_ce = sctp_v4_is_ce, 1193 .seq_dump_addr = sctp_v4_seq_dump_addr, 1194 .ecn_capable = sctp_v4_ecn_capable, 1195 .net_header_len = sizeof(struct iphdr), 1196 .sockaddr_len = sizeof(struct sockaddr_in), 1197 .ip_options_len = sctp_v4_ip_options_len, 1198 }; 1199 1200 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1201 { 1202 switch (family) { 1203 case PF_INET: 1204 return sctp_pf_inet_specific; 1205 case PF_INET6: 1206 return sctp_pf_inet6_specific; 1207 default: 1208 return NULL; 1209 } 1210 } 1211 1212 /* Register the PF specific function table. */ 1213 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1214 { 1215 switch (family) { 1216 case PF_INET: 1217 if (sctp_pf_inet_specific) 1218 return 0; 1219 sctp_pf_inet_specific = pf; 1220 break; 1221 case PF_INET6: 1222 if (sctp_pf_inet6_specific) 1223 return 0; 1224 sctp_pf_inet6_specific = pf; 1225 break; 1226 default: 1227 return 0; 1228 } 1229 return 1; 1230 } 1231 1232 static inline int init_sctp_mibs(struct net *net) 1233 { 1234 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1235 if (!net->sctp.sctp_statistics) 1236 return -ENOMEM; 1237 return 0; 1238 } 1239 1240 static inline void cleanup_sctp_mibs(struct net *net) 1241 { 1242 free_percpu(net->sctp.sctp_statistics); 1243 } 1244 1245 static void sctp_v4_pf_init(void) 1246 { 1247 /* Initialize the SCTP specific PF functions. */ 1248 sctp_register_pf(&sctp_pf_inet, PF_INET); 1249 sctp_register_af(&sctp_af_inet); 1250 } 1251 1252 static void sctp_v4_pf_exit(void) 1253 { 1254 list_del(&sctp_af_inet.list); 1255 } 1256 1257 static int sctp_v4_protosw_init(void) 1258 { 1259 int rc; 1260 1261 rc = proto_register(&sctp_prot, 1); 1262 if (rc) 1263 return rc; 1264 1265 /* Register SCTP(UDP and TCP style) with socket layer. */ 1266 inet_register_protosw(&sctp_seqpacket_protosw); 1267 inet_register_protosw(&sctp_stream_protosw); 1268 1269 return 0; 1270 } 1271 1272 static void sctp_v4_protosw_exit(void) 1273 { 1274 inet_unregister_protosw(&sctp_stream_protosw); 1275 inet_unregister_protosw(&sctp_seqpacket_protosw); 1276 proto_unregister(&sctp_prot); 1277 } 1278 1279 static int sctp_v4_add_protocol(void) 1280 { 1281 /* Register notifier for inet address additions/deletions. */ 1282 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1283 1284 /* Register SCTP with inet layer. */ 1285 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1286 return -EAGAIN; 1287 1288 return 0; 1289 } 1290 1291 static void sctp_v4_del_protocol(void) 1292 { 1293 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1294 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1295 } 1296 1297 static int __net_init sctp_defaults_init(struct net *net) 1298 { 1299 int status; 1300 1301 /* 1302 * 14. Suggested SCTP Protocol Parameter Values 1303 */ 1304 /* The following protocol parameters are RECOMMENDED: */ 1305 /* RTO.Initial - 3 seconds */ 1306 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1307 /* RTO.Min - 1 second */ 1308 net->sctp.rto_min = SCTP_RTO_MIN; 1309 /* RTO.Max - 60 seconds */ 1310 net->sctp.rto_max = SCTP_RTO_MAX; 1311 /* RTO.Alpha - 1/8 */ 1312 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1313 /* RTO.Beta - 1/4 */ 1314 net->sctp.rto_beta = SCTP_RTO_BETA; 1315 1316 /* Valid.Cookie.Life - 60 seconds */ 1317 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1318 1319 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1320 net->sctp.cookie_preserve_enable = 1; 1321 1322 /* Default sctp sockets to use md5 as their hmac alg */ 1323 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1324 net->sctp.sctp_hmac_alg = "md5"; 1325 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1326 net->sctp.sctp_hmac_alg = "sha1"; 1327 #else 1328 net->sctp.sctp_hmac_alg = NULL; 1329 #endif 1330 1331 /* Max.Burst - 4 */ 1332 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1333 1334 /* Disable of Primary Path Switchover by default */ 1335 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX; 1336 1337 /* Enable pf state by default */ 1338 net->sctp.pf_enable = 1; 1339 1340 /* Ignore pf exposure feature by default */ 1341 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET; 1342 1343 /* Association.Max.Retrans - 10 attempts 1344 * Path.Max.Retrans - 5 attempts (per destination address) 1345 * Max.Init.Retransmits - 8 attempts 1346 */ 1347 net->sctp.max_retrans_association = 10; 1348 net->sctp.max_retrans_path = 5; 1349 net->sctp.max_retrans_init = 8; 1350 1351 /* Sendbuffer growth - do per-socket accounting */ 1352 net->sctp.sndbuf_policy = 0; 1353 1354 /* Rcvbuffer growth - do per-socket accounting */ 1355 net->sctp.rcvbuf_policy = 0; 1356 1357 /* HB.interval - 30 seconds */ 1358 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1359 1360 /* delayed SACK timeout */ 1361 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1362 1363 /* Disable ADDIP by default. */ 1364 net->sctp.addip_enable = 0; 1365 net->sctp.addip_noauth = 0; 1366 net->sctp.default_auto_asconf = 0; 1367 1368 /* Enable PR-SCTP by default. */ 1369 net->sctp.prsctp_enable = 1; 1370 1371 /* Disable RECONF by default. */ 1372 net->sctp.reconf_enable = 0; 1373 1374 /* Disable AUTH by default. */ 1375 net->sctp.auth_enable = 0; 1376 1377 /* Enable ECN by default. */ 1378 net->sctp.ecn_enable = 1; 1379 1380 /* Set UDP tunneling listening port to 0 by default */ 1381 net->sctp.udp_port = 0; 1382 1383 /* Set remote encap port to 0 by default */ 1384 net->sctp.encap_port = 0; 1385 1386 /* Set SCOPE policy to enabled */ 1387 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1388 1389 /* Set the default rwnd update threshold */ 1390 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1391 1392 /* Initialize maximum autoclose timeout. */ 1393 net->sctp.max_autoclose = INT_MAX / HZ; 1394 1395 #ifdef CONFIG_NET_L3_MASTER_DEV 1396 net->sctp.l3mdev_accept = 1; 1397 #endif 1398 1399 status = sctp_sysctl_net_register(net); 1400 if (status) 1401 goto err_sysctl_register; 1402 1403 /* Allocate and initialise sctp mibs. */ 1404 status = init_sctp_mibs(net); 1405 if (status) 1406 goto err_init_mibs; 1407 1408 #ifdef CONFIG_PROC_FS 1409 /* Initialize proc fs directory. */ 1410 status = sctp_proc_init(net); 1411 if (status) 1412 goto err_init_proc; 1413 #endif 1414 1415 sctp_dbg_objcnt_init(net); 1416 1417 /* Initialize the local address list. */ 1418 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1419 spin_lock_init(&net->sctp.local_addr_lock); 1420 sctp_get_local_addr_list(net); 1421 1422 /* Initialize the address event list */ 1423 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1424 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1425 spin_lock_init(&net->sctp.addr_wq_lock); 1426 net->sctp.addr_wq_timer.expires = 0; 1427 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1428 1429 return 0; 1430 1431 #ifdef CONFIG_PROC_FS 1432 err_init_proc: 1433 cleanup_sctp_mibs(net); 1434 #endif 1435 err_init_mibs: 1436 sctp_sysctl_net_unregister(net); 1437 err_sysctl_register: 1438 return status; 1439 } 1440 1441 static void __net_exit sctp_defaults_exit(struct net *net) 1442 { 1443 /* Free the local address list */ 1444 sctp_free_addr_wq(net); 1445 sctp_free_local_addr_list(net); 1446 1447 #ifdef CONFIG_PROC_FS 1448 remove_proc_subtree("sctp", net->proc_net); 1449 net->sctp.proc_net_sctp = NULL; 1450 #endif 1451 cleanup_sctp_mibs(net); 1452 sctp_sysctl_net_unregister(net); 1453 } 1454 1455 static struct pernet_operations sctp_defaults_ops = { 1456 .init = sctp_defaults_init, 1457 .exit = sctp_defaults_exit, 1458 }; 1459 1460 static int __net_init sctp_ctrlsock_init(struct net *net) 1461 { 1462 int status; 1463 1464 /* Initialize the control inode/socket for handling OOTB packets. */ 1465 status = sctp_ctl_sock_init(net); 1466 if (status) 1467 pr_err("Failed to initialize the SCTP control sock\n"); 1468 1469 return status; 1470 } 1471 1472 static void __net_exit sctp_ctrlsock_exit(struct net *net) 1473 { 1474 /* Free the control endpoint. */ 1475 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1476 } 1477 1478 static struct pernet_operations sctp_ctrlsock_ops = { 1479 .init = sctp_ctrlsock_init, 1480 .exit = sctp_ctrlsock_exit, 1481 }; 1482 1483 /* Initialize the universe into something sensible. */ 1484 static __init int sctp_init(void) 1485 { 1486 unsigned long nr_pages = totalram_pages(); 1487 unsigned long limit; 1488 unsigned long goal; 1489 int max_entry_order; 1490 int num_entries; 1491 int max_share; 1492 int status; 1493 int order; 1494 int i; 1495 1496 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1497 1498 /* Allocate bind_bucket and chunk caches. */ 1499 status = -ENOBUFS; 1500 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket", 1501 sizeof(struct sctp_bind_bucket), 1502 0, SLAB_HWCACHE_ALIGN, 1503 NULL); 1504 if (!sctp_bucket_cachep) 1505 goto out; 1506 1507 sctp_chunk_cachep = kmem_cache_create("sctp_chunk", 1508 sizeof(struct sctp_chunk), 1509 0, SLAB_HWCACHE_ALIGN, 1510 NULL); 1511 if (!sctp_chunk_cachep) 1512 goto err_chunk_cachep; 1513 1514 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1515 if (status) 1516 goto err_percpu_counter_init; 1517 1518 /* Implementation specific variables. */ 1519 1520 /* Initialize default stream count setup information. */ 1521 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1522 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1523 1524 /* Initialize handle used for association ids. */ 1525 idr_init(&sctp_assocs_id); 1526 1527 limit = nr_free_buffer_pages() / 8; 1528 limit = max(limit, 128UL); 1529 sysctl_sctp_mem[0] = limit / 4 * 3; 1530 sysctl_sctp_mem[1] = limit; 1531 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1532 1533 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1534 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1535 max_share = min(4UL*1024*1024, limit); 1536 1537 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */ 1538 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1539 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1540 1541 sysctl_sctp_wmem[0] = PAGE_SIZE; 1542 sysctl_sctp_wmem[1] = 16*1024; 1543 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1544 1545 /* Size and allocate the association hash table. 1546 * The methodology is similar to that of the tcp hash tables. 1547 * Though not identical. Start by getting a goal size 1548 */ 1549 if (nr_pages >= (128 * 1024)) 1550 goal = nr_pages >> (22 - PAGE_SHIFT); 1551 else 1552 goal = nr_pages >> (24 - PAGE_SHIFT); 1553 1554 /* Then compute the page order for said goal */ 1555 order = get_order(goal); 1556 1557 /* Now compute the required page order for the maximum sized table we 1558 * want to create 1559 */ 1560 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1561 sizeof(struct sctp_bind_hashbucket)); 1562 1563 /* Limit the page order by that maximum hash table size */ 1564 order = min(order, max_entry_order); 1565 1566 /* Allocate and initialize the endpoint hash table. */ 1567 sctp_ep_hashsize = 64; 1568 sctp_ep_hashtable = 1569 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1570 if (!sctp_ep_hashtable) { 1571 pr_err("Failed endpoint_hash alloc\n"); 1572 status = -ENOMEM; 1573 goto err_ehash_alloc; 1574 } 1575 for (i = 0; i < sctp_ep_hashsize; i++) { 1576 rwlock_init(&sctp_ep_hashtable[i].lock); 1577 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1578 } 1579 1580 /* Allocate and initialize the SCTP port hash table. 1581 * Note that order is initalized to start at the max sized 1582 * table we want to support. If we can't get that many pages 1583 * reduce the order and try again 1584 */ 1585 do { 1586 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1587 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1588 } while (!sctp_port_hashtable && --order > 0); 1589 1590 if (!sctp_port_hashtable) { 1591 pr_err("Failed bind hash alloc\n"); 1592 status = -ENOMEM; 1593 goto err_bhash_alloc; 1594 } 1595 1596 /* Now compute the number of entries that will fit in the 1597 * port hash space we allocated 1598 */ 1599 num_entries = (1UL << order) * PAGE_SIZE / 1600 sizeof(struct sctp_bind_hashbucket); 1601 1602 /* And finish by rounding it down to the nearest power of two. 1603 * This wastes some memory of course, but it's needed because 1604 * the hash function operates based on the assumption that 1605 * the number of entries is a power of two. 1606 */ 1607 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1608 1609 for (i = 0; i < sctp_port_hashsize; i++) { 1610 spin_lock_init(&sctp_port_hashtable[i].lock); 1611 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1612 } 1613 1614 status = sctp_transport_hashtable_init(); 1615 if (status) 1616 goto err_thash_alloc; 1617 1618 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1619 num_entries); 1620 1621 sctp_sysctl_register(); 1622 1623 INIT_LIST_HEAD(&sctp_address_families); 1624 sctp_v4_pf_init(); 1625 sctp_v6_pf_init(); 1626 sctp_sched_ops_init(); 1627 1628 status = register_pernet_subsys(&sctp_defaults_ops); 1629 if (status) 1630 goto err_register_defaults; 1631 1632 status = sctp_v4_protosw_init(); 1633 if (status) 1634 goto err_protosw_init; 1635 1636 status = sctp_v6_protosw_init(); 1637 if (status) 1638 goto err_v6_protosw_init; 1639 1640 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1641 if (status) 1642 goto err_register_ctrlsock; 1643 1644 status = sctp_v4_add_protocol(); 1645 if (status) 1646 goto err_add_protocol; 1647 1648 /* Register SCTP with inet6 layer. */ 1649 status = sctp_v6_add_protocol(); 1650 if (status) 1651 goto err_v6_add_protocol; 1652 1653 if (sctp_offload_init() < 0) 1654 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1655 1656 out: 1657 return status; 1658 err_v6_add_protocol: 1659 sctp_v4_del_protocol(); 1660 err_add_protocol: 1661 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1662 err_register_ctrlsock: 1663 sctp_v6_protosw_exit(); 1664 err_v6_protosw_init: 1665 sctp_v4_protosw_exit(); 1666 err_protosw_init: 1667 unregister_pernet_subsys(&sctp_defaults_ops); 1668 err_register_defaults: 1669 sctp_v4_pf_exit(); 1670 sctp_v6_pf_exit(); 1671 sctp_sysctl_unregister(); 1672 free_pages((unsigned long)sctp_port_hashtable, 1673 get_order(sctp_port_hashsize * 1674 sizeof(struct sctp_bind_hashbucket))); 1675 err_bhash_alloc: 1676 sctp_transport_hashtable_destroy(); 1677 err_thash_alloc: 1678 kfree(sctp_ep_hashtable); 1679 err_ehash_alloc: 1680 percpu_counter_destroy(&sctp_sockets_allocated); 1681 err_percpu_counter_init: 1682 kmem_cache_destroy(sctp_chunk_cachep); 1683 err_chunk_cachep: 1684 kmem_cache_destroy(sctp_bucket_cachep); 1685 goto out; 1686 } 1687 1688 /* Exit handler for the SCTP protocol. */ 1689 static __exit void sctp_exit(void) 1690 { 1691 /* BUG. This should probably do something useful like clean 1692 * up all the remaining associations and all that memory. 1693 */ 1694 1695 /* Unregister with inet6/inet layers. */ 1696 sctp_v6_del_protocol(); 1697 sctp_v4_del_protocol(); 1698 1699 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1700 1701 /* Free protosw registrations */ 1702 sctp_v6_protosw_exit(); 1703 sctp_v4_protosw_exit(); 1704 1705 unregister_pernet_subsys(&sctp_defaults_ops); 1706 1707 /* Unregister with socket layer. */ 1708 sctp_v6_pf_exit(); 1709 sctp_v4_pf_exit(); 1710 1711 sctp_sysctl_unregister(); 1712 1713 free_pages((unsigned long)sctp_port_hashtable, 1714 get_order(sctp_port_hashsize * 1715 sizeof(struct sctp_bind_hashbucket))); 1716 kfree(sctp_ep_hashtable); 1717 sctp_transport_hashtable_destroy(); 1718 1719 percpu_counter_destroy(&sctp_sockets_allocated); 1720 1721 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1722 1723 kmem_cache_destroy(sctp_chunk_cachep); 1724 kmem_cache_destroy(sctp_bucket_cachep); 1725 } 1726 1727 module_init(sctp_init); 1728 module_exit(sctp_exit); 1729 1730 /* 1731 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1732 */ 1733 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1734 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1735 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1736 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1737 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1738 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1739 MODULE_LICENSE("GPL"); 1740