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