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 sk_refcnt_debug_inc(newsk); 605 606 if (newsk->sk_prot->init(newsk)) { 607 sk_common_release(newsk); 608 newsk = NULL; 609 } 610 611 out: 612 return newsk; 613 } 614 615 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 616 { 617 /* No address mapping for V4 sockets */ 618 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 619 return sizeof(struct sockaddr_in); 620 } 621 622 /* Dump the v4 addr to the seq file. */ 623 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 624 { 625 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 626 } 627 628 static void sctp_v4_ecn_capable(struct sock *sk) 629 { 630 INET_ECN_xmit(sk); 631 } 632 633 static void sctp_addr_wq_timeout_handler(struct timer_list *t) 634 { 635 struct net *net = from_timer(net, t, sctp.addr_wq_timer); 636 struct sctp_sockaddr_entry *addrw, *temp; 637 struct sctp_sock *sp; 638 639 spin_lock_bh(&net->sctp.addr_wq_lock); 640 641 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 642 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 643 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 644 addrw->state, addrw); 645 646 #if IS_ENABLED(CONFIG_IPV6) 647 /* Now we send an ASCONF for each association */ 648 /* Note. we currently don't handle link local IPv6 addressees */ 649 if (addrw->a.sa.sa_family == AF_INET6) { 650 struct in6_addr *in6; 651 652 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 653 IPV6_ADDR_LINKLOCAL) 654 goto free_next; 655 656 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 657 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 658 addrw->state == SCTP_ADDR_NEW) { 659 unsigned long timeo_val; 660 661 pr_debug("%s: this is on DAD, trying %d sec " 662 "later\n", __func__, 663 SCTP_ADDRESS_TICK_DELAY); 664 665 timeo_val = jiffies; 666 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 667 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 668 break; 669 } 670 } 671 #endif 672 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 673 struct sock *sk; 674 675 sk = sctp_opt2sk(sp); 676 /* ignore bound-specific endpoints */ 677 if (!sctp_is_ep_boundall(sk)) 678 continue; 679 bh_lock_sock(sk); 680 if (sctp_asconf_mgmt(sp, addrw) < 0) 681 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 682 bh_unlock_sock(sk); 683 } 684 #if IS_ENABLED(CONFIG_IPV6) 685 free_next: 686 #endif 687 list_del(&addrw->list); 688 kfree(addrw); 689 } 690 spin_unlock_bh(&net->sctp.addr_wq_lock); 691 } 692 693 static void sctp_free_addr_wq(struct net *net) 694 { 695 struct sctp_sockaddr_entry *addrw; 696 struct sctp_sockaddr_entry *temp; 697 698 spin_lock_bh(&net->sctp.addr_wq_lock); 699 del_timer(&net->sctp.addr_wq_timer); 700 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 701 list_del(&addrw->list); 702 kfree(addrw); 703 } 704 spin_unlock_bh(&net->sctp.addr_wq_lock); 705 } 706 707 /* lookup the entry for the same address in the addr_waitq 708 * sctp_addr_wq MUST be locked 709 */ 710 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 711 struct sctp_sockaddr_entry *addr) 712 { 713 struct sctp_sockaddr_entry *addrw; 714 715 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 716 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 717 continue; 718 if (addrw->a.sa.sa_family == AF_INET) { 719 if (addrw->a.v4.sin_addr.s_addr == 720 addr->a.v4.sin_addr.s_addr) 721 return addrw; 722 } else if (addrw->a.sa.sa_family == AF_INET6) { 723 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 724 &addr->a.v6.sin6_addr)) 725 return addrw; 726 } 727 } 728 return NULL; 729 } 730 731 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 732 { 733 struct sctp_sockaddr_entry *addrw; 734 unsigned long timeo_val; 735 736 /* first, we check if an opposite message already exist in the queue. 737 * If we found such message, it is removed. 738 * This operation is a bit stupid, but the DHCP client attaches the 739 * new address after a couple of addition and deletion of that address 740 */ 741 742 spin_lock_bh(&net->sctp.addr_wq_lock); 743 /* Offsets existing events in addr_wq */ 744 addrw = sctp_addr_wq_lookup(net, addr); 745 if (addrw) { 746 if (addrw->state != cmd) { 747 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 748 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 749 &net->sctp.addr_waitq); 750 751 list_del(&addrw->list); 752 kfree(addrw); 753 } 754 spin_unlock_bh(&net->sctp.addr_wq_lock); 755 return; 756 } 757 758 /* OK, we have to add the new address to the wait queue */ 759 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 760 if (addrw == NULL) { 761 spin_unlock_bh(&net->sctp.addr_wq_lock); 762 return; 763 } 764 addrw->state = cmd; 765 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 766 767 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 768 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 769 770 if (!timer_pending(&net->sctp.addr_wq_timer)) { 771 timeo_val = jiffies; 772 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 773 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 774 } 775 spin_unlock_bh(&net->sctp.addr_wq_lock); 776 } 777 778 /* Event handler for inet address addition/deletion events. 779 * The sctp_local_addr_list needs to be protocted by a spin lock since 780 * multiple notifiers (say IPv4 and IPv6) may be running at the same 781 * time and thus corrupt the list. 782 * The reader side is protected with RCU. 783 */ 784 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 785 void *ptr) 786 { 787 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 788 struct sctp_sockaddr_entry *addr = NULL; 789 struct sctp_sockaddr_entry *temp; 790 struct net *net = dev_net(ifa->ifa_dev->dev); 791 int found = 0; 792 793 switch (ev) { 794 case NETDEV_UP: 795 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 796 if (addr) { 797 addr->a.v4.sin_family = AF_INET; 798 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 799 addr->valid = 1; 800 spin_lock_bh(&net->sctp.local_addr_lock); 801 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 802 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 803 spin_unlock_bh(&net->sctp.local_addr_lock); 804 } 805 break; 806 case NETDEV_DOWN: 807 spin_lock_bh(&net->sctp.local_addr_lock); 808 list_for_each_entry_safe(addr, temp, 809 &net->sctp.local_addr_list, list) { 810 if (addr->a.sa.sa_family == AF_INET && 811 addr->a.v4.sin_addr.s_addr == 812 ifa->ifa_local) { 813 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 814 found = 1; 815 addr->valid = 0; 816 list_del_rcu(&addr->list); 817 break; 818 } 819 } 820 spin_unlock_bh(&net->sctp.local_addr_lock); 821 if (found) 822 kfree_rcu(addr, rcu); 823 break; 824 } 825 826 return NOTIFY_DONE; 827 } 828 829 /* 830 * Initialize the control inode/socket with a control endpoint data 831 * structure. This endpoint is reserved exclusively for the OOTB processing. 832 */ 833 static int sctp_ctl_sock_init(struct net *net) 834 { 835 int err; 836 sa_family_t family = PF_INET; 837 838 if (sctp_get_pf_specific(PF_INET6)) 839 family = PF_INET6; 840 841 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 842 SOCK_SEQPACKET, IPPROTO_SCTP, net); 843 844 /* If IPv6 socket could not be created, try the IPv4 socket */ 845 if (err < 0 && family == PF_INET6) 846 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 847 SOCK_SEQPACKET, IPPROTO_SCTP, 848 net); 849 850 if (err < 0) { 851 pr_err("Failed to create the SCTP control socket\n"); 852 return err; 853 } 854 return 0; 855 } 856 857 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb) 858 { 859 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source; 860 861 skb_set_transport_header(skb, sizeof(struct udphdr)); 862 sctp_rcv(skb); 863 return 0; 864 } 865 866 int sctp_udp_sock_start(struct net *net) 867 { 868 struct udp_tunnel_sock_cfg tuncfg = {NULL}; 869 struct udp_port_cfg udp_conf = {0}; 870 struct socket *sock; 871 int err; 872 873 udp_conf.family = AF_INET; 874 udp_conf.local_ip.s_addr = htonl(INADDR_ANY); 875 udp_conf.local_udp_port = htons(net->sctp.udp_port); 876 err = udp_sock_create(net, &udp_conf, &sock); 877 if (err) { 878 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n"); 879 return err; 880 } 881 882 tuncfg.encap_type = 1; 883 tuncfg.encap_rcv = sctp_udp_rcv; 884 tuncfg.encap_err_lookup = sctp_udp_v4_err; 885 setup_udp_tunnel_sock(net, sock, &tuncfg); 886 net->sctp.udp4_sock = sock->sk; 887 888 #if IS_ENABLED(CONFIG_IPV6) 889 memset(&udp_conf, 0, sizeof(udp_conf)); 890 891 udp_conf.family = AF_INET6; 892 udp_conf.local_ip6 = in6addr_any; 893 udp_conf.local_udp_port = htons(net->sctp.udp_port); 894 udp_conf.use_udp6_rx_checksums = true; 895 udp_conf.ipv6_v6only = true; 896 err = udp_sock_create(net, &udp_conf, &sock); 897 if (err) { 898 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n"); 899 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 900 net->sctp.udp4_sock = NULL; 901 return err; 902 } 903 904 tuncfg.encap_type = 1; 905 tuncfg.encap_rcv = sctp_udp_rcv; 906 tuncfg.encap_err_lookup = sctp_udp_v6_err; 907 setup_udp_tunnel_sock(net, sock, &tuncfg); 908 net->sctp.udp6_sock = sock->sk; 909 #endif 910 911 return 0; 912 } 913 914 void sctp_udp_sock_stop(struct net *net) 915 { 916 if (net->sctp.udp4_sock) { 917 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 918 net->sctp.udp4_sock = NULL; 919 } 920 if (net->sctp.udp6_sock) { 921 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket); 922 net->sctp.udp6_sock = NULL; 923 } 924 } 925 926 /* Register address family specific functions. */ 927 int sctp_register_af(struct sctp_af *af) 928 { 929 switch (af->sa_family) { 930 case AF_INET: 931 if (sctp_af_v4_specific) 932 return 0; 933 sctp_af_v4_specific = af; 934 break; 935 case AF_INET6: 936 if (sctp_af_v6_specific) 937 return 0; 938 sctp_af_v6_specific = af; 939 break; 940 default: 941 return 0; 942 } 943 944 INIT_LIST_HEAD(&af->list); 945 list_add_tail(&af->list, &sctp_address_families); 946 return 1; 947 } 948 949 /* Get the table of functions for manipulating a particular address 950 * family. 951 */ 952 struct sctp_af *sctp_get_af_specific(sa_family_t family) 953 { 954 switch (family) { 955 case AF_INET: 956 return sctp_af_v4_specific; 957 case AF_INET6: 958 return sctp_af_v6_specific; 959 default: 960 return NULL; 961 } 962 } 963 964 /* Common code to initialize a AF_INET msg_name. */ 965 static void sctp_inet_msgname(char *msgname, int *addr_len) 966 { 967 struct sockaddr_in *sin; 968 969 sin = (struct sockaddr_in *)msgname; 970 *addr_len = sizeof(struct sockaddr_in); 971 sin->sin_family = AF_INET; 972 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 973 } 974 975 /* Copy the primary address of the peer primary address as the msg_name. */ 976 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 977 int *addr_len) 978 { 979 struct sockaddr_in *sin, *sinfrom; 980 981 if (msgname) { 982 struct sctp_association *asoc; 983 984 asoc = event->asoc; 985 sctp_inet_msgname(msgname, addr_len); 986 sin = (struct sockaddr_in *)msgname; 987 sinfrom = &asoc->peer.primary_addr.v4; 988 sin->sin_port = htons(asoc->peer.port); 989 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 990 } 991 } 992 993 /* Initialize and copy out a msgname from an inbound skb. */ 994 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 995 { 996 if (msgname) { 997 struct sctphdr *sh = sctp_hdr(skb); 998 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 999 1000 sctp_inet_msgname(msgname, len); 1001 sin->sin_port = sh->source; 1002 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 1003 } 1004 } 1005 1006 /* Do we support this AF? */ 1007 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 1008 { 1009 /* PF_INET only supports AF_INET addresses. */ 1010 return AF_INET == family; 1011 } 1012 1013 /* Address matching with wildcards allowed. */ 1014 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 1015 const union sctp_addr *addr2, 1016 struct sctp_sock *opt) 1017 { 1018 /* PF_INET only supports AF_INET addresses. */ 1019 if (addr1->sa.sa_family != addr2->sa.sa_family) 1020 return 0; 1021 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 1022 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 1023 return 1; 1024 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 1025 return 1; 1026 1027 return 0; 1028 } 1029 1030 /* Verify that provided sockaddr looks bindable. Common verification has 1031 * already been taken care of. 1032 */ 1033 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 1034 { 1035 return sctp_v4_available(addr, opt); 1036 } 1037 1038 /* Verify that sockaddr looks sendable. Common verification has already 1039 * been taken care of. 1040 */ 1041 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 1042 { 1043 return 1; 1044 } 1045 1046 /* Fill in Supported Address Type information for INIT and INIT-ACK 1047 * chunks. Returns number of addresses supported. 1048 */ 1049 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 1050 __be16 *types) 1051 { 1052 types[0] = SCTP_PARAM_IPV4_ADDRESS; 1053 return 1; 1054 } 1055 1056 /* Wrapper routine that calls the ip transmit routine. */ 1057 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t) 1058 { 1059 struct dst_entry *dst = dst_clone(t->dst); 1060 struct flowi4 *fl4 = &t->fl.u.ip4; 1061 struct sock *sk = skb->sk; 1062 struct inet_sock *inet = inet_sk(sk); 1063 __u8 dscp = inet->tos; 1064 __be16 df = 0; 1065 1066 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 1067 skb->len, &fl4->saddr, &fl4->daddr); 1068 1069 if (t->dscp & SCTP_DSCP_SET_MASK) 1070 dscp = t->dscp & SCTP_DSCP_VAL_MASK; 1071 1072 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO 1073 : IP_PMTUDISC_DONT; 1074 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS); 1075 1076 if (!t->encap_port || !sctp_sk(sk)->udp_port) { 1077 skb_dst_set(skb, dst); 1078 return __ip_queue_xmit(sk, skb, &t->fl, dscp); 1079 } 1080 1081 if (skb_is_gso(skb)) 1082 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM; 1083 1084 if (ip_dont_fragment(sk, dst) && !skb->ignore_df) 1085 df = htons(IP_DF); 1086 1087 skb->encapsulation = 1; 1088 skb_reset_inner_mac_header(skb); 1089 skb_reset_inner_transport_header(skb); 1090 skb_set_inner_ipproto(skb, IPPROTO_SCTP); 1091 udp_tunnel_xmit_skb((struct rtable *)dst, sk, skb, fl4->saddr, 1092 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df, 1093 sctp_sk(sk)->udp_port, t->encap_port, false, false); 1094 return 0; 1095 } 1096 1097 static struct sctp_af sctp_af_inet; 1098 1099 static struct sctp_pf sctp_pf_inet = { 1100 .event_msgname = sctp_inet_event_msgname, 1101 .skb_msgname = sctp_inet_skb_msgname, 1102 .af_supported = sctp_inet_af_supported, 1103 .cmp_addr = sctp_inet_cmp_addr, 1104 .bind_verify = sctp_inet_bind_verify, 1105 .send_verify = sctp_inet_send_verify, 1106 .supported_addrs = sctp_inet_supported_addrs, 1107 .create_accept_sk = sctp_v4_create_accept_sk, 1108 .addr_to_user = sctp_v4_addr_to_user, 1109 .to_sk_saddr = sctp_v4_to_sk_saddr, 1110 .to_sk_daddr = sctp_v4_to_sk_daddr, 1111 .copy_ip_options = sctp_v4_copy_ip_options, 1112 .af = &sctp_af_inet 1113 }; 1114 1115 /* Notifier for inetaddr addition/deletion events. */ 1116 static struct notifier_block sctp_inetaddr_notifier = { 1117 .notifier_call = sctp_inetaddr_event, 1118 }; 1119 1120 /* Socket operations. */ 1121 static const struct proto_ops inet_seqpacket_ops = { 1122 .family = PF_INET, 1123 .owner = THIS_MODULE, 1124 .release = inet_release, /* Needs to be wrapped... */ 1125 .bind = inet_bind, 1126 .connect = sctp_inet_connect, 1127 .socketpair = sock_no_socketpair, 1128 .accept = inet_accept, 1129 .getname = inet_getname, /* Semantics are different. */ 1130 .poll = sctp_poll, 1131 .ioctl = inet_ioctl, 1132 .gettstamp = sock_gettstamp, 1133 .listen = sctp_inet_listen, 1134 .shutdown = inet_shutdown, /* Looks harmless. */ 1135 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1136 .getsockopt = sock_common_getsockopt, 1137 .sendmsg = inet_sendmsg, 1138 .recvmsg = inet_recvmsg, 1139 .mmap = sock_no_mmap, 1140 .sendpage = sock_no_sendpage, 1141 }; 1142 1143 /* Registration with AF_INET family. */ 1144 static struct inet_protosw sctp_seqpacket_protosw = { 1145 .type = SOCK_SEQPACKET, 1146 .protocol = IPPROTO_SCTP, 1147 .prot = &sctp_prot, 1148 .ops = &inet_seqpacket_ops, 1149 .flags = SCTP_PROTOSW_FLAG 1150 }; 1151 static struct inet_protosw sctp_stream_protosw = { 1152 .type = SOCK_STREAM, 1153 .protocol = IPPROTO_SCTP, 1154 .prot = &sctp_prot, 1155 .ops = &inet_seqpacket_ops, 1156 .flags = SCTP_PROTOSW_FLAG 1157 }; 1158 1159 static int sctp4_rcv(struct sk_buff *skb) 1160 { 1161 SCTP_INPUT_CB(skb)->encap_port = 0; 1162 return sctp_rcv(skb); 1163 } 1164 1165 /* Register with IP layer. */ 1166 static const struct net_protocol sctp_protocol = { 1167 .handler = sctp4_rcv, 1168 .err_handler = sctp_v4_err, 1169 .no_policy = 1, 1170 .icmp_strict_tag_validation = 1, 1171 }; 1172 1173 /* IPv4 address related functions. */ 1174 static struct sctp_af sctp_af_inet = { 1175 .sa_family = AF_INET, 1176 .sctp_xmit = sctp_v4_xmit, 1177 .setsockopt = ip_setsockopt, 1178 .getsockopt = ip_getsockopt, 1179 .get_dst = sctp_v4_get_dst, 1180 .get_saddr = sctp_v4_get_saddr, 1181 .copy_addrlist = sctp_v4_copy_addrlist, 1182 .from_skb = sctp_v4_from_skb, 1183 .from_sk = sctp_v4_from_sk, 1184 .from_addr_param = sctp_v4_from_addr_param, 1185 .to_addr_param = sctp_v4_to_addr_param, 1186 .cmp_addr = sctp_v4_cmp_addr, 1187 .addr_valid = sctp_v4_addr_valid, 1188 .inaddr_any = sctp_v4_inaddr_any, 1189 .is_any = sctp_v4_is_any, 1190 .available = sctp_v4_available, 1191 .scope = sctp_v4_scope, 1192 .skb_iif = sctp_v4_skb_iif, 1193 .skb_sdif = sctp_v4_skb_sdif, 1194 .is_ce = sctp_v4_is_ce, 1195 .seq_dump_addr = sctp_v4_seq_dump_addr, 1196 .ecn_capable = sctp_v4_ecn_capable, 1197 .net_header_len = sizeof(struct iphdr), 1198 .sockaddr_len = sizeof(struct sockaddr_in), 1199 .ip_options_len = sctp_v4_ip_options_len, 1200 }; 1201 1202 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1203 { 1204 switch (family) { 1205 case PF_INET: 1206 return sctp_pf_inet_specific; 1207 case PF_INET6: 1208 return sctp_pf_inet6_specific; 1209 default: 1210 return NULL; 1211 } 1212 } 1213 1214 /* Register the PF specific function table. */ 1215 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1216 { 1217 switch (family) { 1218 case PF_INET: 1219 if (sctp_pf_inet_specific) 1220 return 0; 1221 sctp_pf_inet_specific = pf; 1222 break; 1223 case PF_INET6: 1224 if (sctp_pf_inet6_specific) 1225 return 0; 1226 sctp_pf_inet6_specific = pf; 1227 break; 1228 default: 1229 return 0; 1230 } 1231 return 1; 1232 } 1233 1234 static inline int init_sctp_mibs(struct net *net) 1235 { 1236 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1237 if (!net->sctp.sctp_statistics) 1238 return -ENOMEM; 1239 return 0; 1240 } 1241 1242 static inline void cleanup_sctp_mibs(struct net *net) 1243 { 1244 free_percpu(net->sctp.sctp_statistics); 1245 } 1246 1247 static void sctp_v4_pf_init(void) 1248 { 1249 /* Initialize the SCTP specific PF functions. */ 1250 sctp_register_pf(&sctp_pf_inet, PF_INET); 1251 sctp_register_af(&sctp_af_inet); 1252 } 1253 1254 static void sctp_v4_pf_exit(void) 1255 { 1256 list_del(&sctp_af_inet.list); 1257 } 1258 1259 static int sctp_v4_protosw_init(void) 1260 { 1261 int rc; 1262 1263 rc = proto_register(&sctp_prot, 1); 1264 if (rc) 1265 return rc; 1266 1267 /* Register SCTP(UDP and TCP style) with socket layer. */ 1268 inet_register_protosw(&sctp_seqpacket_protosw); 1269 inet_register_protosw(&sctp_stream_protosw); 1270 1271 return 0; 1272 } 1273 1274 static void sctp_v4_protosw_exit(void) 1275 { 1276 inet_unregister_protosw(&sctp_stream_protosw); 1277 inet_unregister_protosw(&sctp_seqpacket_protosw); 1278 proto_unregister(&sctp_prot); 1279 } 1280 1281 static int sctp_v4_add_protocol(void) 1282 { 1283 /* Register notifier for inet address additions/deletions. */ 1284 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1285 1286 /* Register SCTP with inet layer. */ 1287 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1288 return -EAGAIN; 1289 1290 return 0; 1291 } 1292 1293 static void sctp_v4_del_protocol(void) 1294 { 1295 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1296 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1297 } 1298 1299 static int __net_init sctp_defaults_init(struct net *net) 1300 { 1301 int status; 1302 1303 /* 1304 * 14. Suggested SCTP Protocol Parameter Values 1305 */ 1306 /* The following protocol parameters are RECOMMENDED: */ 1307 /* RTO.Initial - 3 seconds */ 1308 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1309 /* RTO.Min - 1 second */ 1310 net->sctp.rto_min = SCTP_RTO_MIN; 1311 /* RTO.Max - 60 seconds */ 1312 net->sctp.rto_max = SCTP_RTO_MAX; 1313 /* RTO.Alpha - 1/8 */ 1314 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1315 /* RTO.Beta - 1/4 */ 1316 net->sctp.rto_beta = SCTP_RTO_BETA; 1317 1318 /* Valid.Cookie.Life - 60 seconds */ 1319 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1320 1321 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1322 net->sctp.cookie_preserve_enable = 1; 1323 1324 /* Default sctp sockets to use md5 as their hmac alg */ 1325 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1326 net->sctp.sctp_hmac_alg = "md5"; 1327 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1328 net->sctp.sctp_hmac_alg = "sha1"; 1329 #else 1330 net->sctp.sctp_hmac_alg = NULL; 1331 #endif 1332 1333 /* Max.Burst - 4 */ 1334 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1335 1336 /* Disable of Primary Path Switchover by default */ 1337 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX; 1338 1339 /* Enable pf state by default */ 1340 net->sctp.pf_enable = 1; 1341 1342 /* Ignore pf exposure feature by default */ 1343 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET; 1344 1345 /* Association.Max.Retrans - 10 attempts 1346 * Path.Max.Retrans - 5 attempts (per destination address) 1347 * Max.Init.Retransmits - 8 attempts 1348 */ 1349 net->sctp.max_retrans_association = 10; 1350 net->sctp.max_retrans_path = 5; 1351 net->sctp.max_retrans_init = 8; 1352 1353 /* Sendbuffer growth - do per-socket accounting */ 1354 net->sctp.sndbuf_policy = 0; 1355 1356 /* Rcvbuffer growth - do per-socket accounting */ 1357 net->sctp.rcvbuf_policy = 0; 1358 1359 /* HB.interval - 30 seconds */ 1360 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1361 1362 /* delayed SACK timeout */ 1363 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1364 1365 /* Disable ADDIP by default. */ 1366 net->sctp.addip_enable = 0; 1367 net->sctp.addip_noauth = 0; 1368 net->sctp.default_auto_asconf = 0; 1369 1370 /* Enable PR-SCTP by default. */ 1371 net->sctp.prsctp_enable = 1; 1372 1373 /* Disable RECONF by default. */ 1374 net->sctp.reconf_enable = 0; 1375 1376 /* Disable AUTH by default. */ 1377 net->sctp.auth_enable = 0; 1378 1379 /* Enable ECN by default. */ 1380 net->sctp.ecn_enable = 1; 1381 1382 /* Set UDP tunneling listening port to 0 by default */ 1383 net->sctp.udp_port = 0; 1384 1385 /* Set remote encap port to 0 by default */ 1386 net->sctp.encap_port = 0; 1387 1388 /* Set SCOPE policy to enabled */ 1389 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1390 1391 /* Set the default rwnd update threshold */ 1392 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1393 1394 /* Initialize maximum autoclose timeout. */ 1395 net->sctp.max_autoclose = INT_MAX / HZ; 1396 1397 #ifdef CONFIG_NET_L3_MASTER_DEV 1398 net->sctp.l3mdev_accept = 1; 1399 #endif 1400 1401 status = sctp_sysctl_net_register(net); 1402 if (status) 1403 goto err_sysctl_register; 1404 1405 /* Allocate and initialise sctp mibs. */ 1406 status = init_sctp_mibs(net); 1407 if (status) 1408 goto err_init_mibs; 1409 1410 #ifdef CONFIG_PROC_FS 1411 /* Initialize proc fs directory. */ 1412 status = sctp_proc_init(net); 1413 if (status) 1414 goto err_init_proc; 1415 #endif 1416 1417 sctp_dbg_objcnt_init(net); 1418 1419 /* Initialize the local address list. */ 1420 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1421 spin_lock_init(&net->sctp.local_addr_lock); 1422 sctp_get_local_addr_list(net); 1423 1424 /* Initialize the address event list */ 1425 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1426 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1427 spin_lock_init(&net->sctp.addr_wq_lock); 1428 net->sctp.addr_wq_timer.expires = 0; 1429 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1430 1431 return 0; 1432 1433 #ifdef CONFIG_PROC_FS 1434 err_init_proc: 1435 cleanup_sctp_mibs(net); 1436 #endif 1437 err_init_mibs: 1438 sctp_sysctl_net_unregister(net); 1439 err_sysctl_register: 1440 return status; 1441 } 1442 1443 static void __net_exit sctp_defaults_exit(struct net *net) 1444 { 1445 /* Free the local address list */ 1446 sctp_free_addr_wq(net); 1447 sctp_free_local_addr_list(net); 1448 1449 #ifdef CONFIG_PROC_FS 1450 remove_proc_subtree("sctp", net->proc_net); 1451 net->sctp.proc_net_sctp = NULL; 1452 #endif 1453 cleanup_sctp_mibs(net); 1454 sctp_sysctl_net_unregister(net); 1455 } 1456 1457 static struct pernet_operations sctp_defaults_ops = { 1458 .init = sctp_defaults_init, 1459 .exit = sctp_defaults_exit, 1460 }; 1461 1462 static int __net_init sctp_ctrlsock_init(struct net *net) 1463 { 1464 int status; 1465 1466 /* Initialize the control inode/socket for handling OOTB packets. */ 1467 status = sctp_ctl_sock_init(net); 1468 if (status) 1469 pr_err("Failed to initialize the SCTP control sock\n"); 1470 1471 return status; 1472 } 1473 1474 static void __net_exit sctp_ctrlsock_exit(struct net *net) 1475 { 1476 /* Free the control endpoint. */ 1477 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1478 } 1479 1480 static struct pernet_operations sctp_ctrlsock_ops = { 1481 .init = sctp_ctrlsock_init, 1482 .exit = sctp_ctrlsock_exit, 1483 }; 1484 1485 /* Initialize the universe into something sensible. */ 1486 static __init int sctp_init(void) 1487 { 1488 unsigned long nr_pages = totalram_pages(); 1489 unsigned long limit; 1490 unsigned long goal; 1491 int max_entry_order; 1492 int num_entries; 1493 int max_share; 1494 int status; 1495 int order; 1496 int i; 1497 1498 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1499 1500 /* Allocate bind_bucket and chunk caches. */ 1501 status = -ENOBUFS; 1502 sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket", 1503 sizeof(struct sctp_bind_bucket), 1504 0, SLAB_HWCACHE_ALIGN, 1505 NULL); 1506 if (!sctp_bucket_cachep) 1507 goto out; 1508 1509 sctp_chunk_cachep = kmem_cache_create("sctp_chunk", 1510 sizeof(struct sctp_chunk), 1511 0, SLAB_HWCACHE_ALIGN, 1512 NULL); 1513 if (!sctp_chunk_cachep) 1514 goto err_chunk_cachep; 1515 1516 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1517 if (status) 1518 goto err_percpu_counter_init; 1519 1520 /* Implementation specific variables. */ 1521 1522 /* Initialize default stream count setup information. */ 1523 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1524 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1525 1526 /* Initialize handle used for association ids. */ 1527 idr_init(&sctp_assocs_id); 1528 1529 limit = nr_free_buffer_pages() / 8; 1530 limit = max(limit, 128UL); 1531 sysctl_sctp_mem[0] = limit / 4 * 3; 1532 sysctl_sctp_mem[1] = limit; 1533 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1534 1535 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1536 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1537 max_share = min(4UL*1024*1024, limit); 1538 1539 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */ 1540 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1541 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1542 1543 sysctl_sctp_wmem[0] = PAGE_SIZE; 1544 sysctl_sctp_wmem[1] = 16*1024; 1545 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1546 1547 /* Size and allocate the association hash table. 1548 * The methodology is similar to that of the tcp hash tables. 1549 * Though not identical. Start by getting a goal size 1550 */ 1551 if (nr_pages >= (128 * 1024)) 1552 goal = nr_pages >> (22 - PAGE_SHIFT); 1553 else 1554 goal = nr_pages >> (24 - PAGE_SHIFT); 1555 1556 /* Then compute the page order for said goal */ 1557 order = get_order(goal); 1558 1559 /* Now compute the required page order for the maximum sized table we 1560 * want to create 1561 */ 1562 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1563 sizeof(struct sctp_bind_hashbucket)); 1564 1565 /* Limit the page order by that maximum hash table size */ 1566 order = min(order, max_entry_order); 1567 1568 /* Allocate and initialize the endpoint hash table. */ 1569 sctp_ep_hashsize = 64; 1570 sctp_ep_hashtable = 1571 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1572 if (!sctp_ep_hashtable) { 1573 pr_err("Failed endpoint_hash alloc\n"); 1574 status = -ENOMEM; 1575 goto err_ehash_alloc; 1576 } 1577 for (i = 0; i < sctp_ep_hashsize; i++) { 1578 rwlock_init(&sctp_ep_hashtable[i].lock); 1579 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1580 } 1581 1582 /* Allocate and initialize the SCTP port hash table. 1583 * Note that order is initalized to start at the max sized 1584 * table we want to support. If we can't get that many pages 1585 * reduce the order and try again 1586 */ 1587 do { 1588 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1589 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1590 } while (!sctp_port_hashtable && --order > 0); 1591 1592 if (!sctp_port_hashtable) { 1593 pr_err("Failed bind hash alloc\n"); 1594 status = -ENOMEM; 1595 goto err_bhash_alloc; 1596 } 1597 1598 /* Now compute the number of entries that will fit in the 1599 * port hash space we allocated 1600 */ 1601 num_entries = (1UL << order) * PAGE_SIZE / 1602 sizeof(struct sctp_bind_hashbucket); 1603 1604 /* And finish by rounding it down to the nearest power of two. 1605 * This wastes some memory of course, but it's needed because 1606 * the hash function operates based on the assumption that 1607 * the number of entries is a power of two. 1608 */ 1609 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1610 1611 for (i = 0; i < sctp_port_hashsize; i++) { 1612 spin_lock_init(&sctp_port_hashtable[i].lock); 1613 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1614 } 1615 1616 status = sctp_transport_hashtable_init(); 1617 if (status) 1618 goto err_thash_alloc; 1619 1620 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1621 num_entries); 1622 1623 sctp_sysctl_register(); 1624 1625 INIT_LIST_HEAD(&sctp_address_families); 1626 sctp_v4_pf_init(); 1627 sctp_v6_pf_init(); 1628 sctp_sched_ops_init(); 1629 1630 status = register_pernet_subsys(&sctp_defaults_ops); 1631 if (status) 1632 goto err_register_defaults; 1633 1634 status = sctp_v4_protosw_init(); 1635 if (status) 1636 goto err_protosw_init; 1637 1638 status = sctp_v6_protosw_init(); 1639 if (status) 1640 goto err_v6_protosw_init; 1641 1642 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1643 if (status) 1644 goto err_register_ctrlsock; 1645 1646 status = sctp_v4_add_protocol(); 1647 if (status) 1648 goto err_add_protocol; 1649 1650 /* Register SCTP with inet6 layer. */ 1651 status = sctp_v6_add_protocol(); 1652 if (status) 1653 goto err_v6_add_protocol; 1654 1655 if (sctp_offload_init() < 0) 1656 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1657 1658 out: 1659 return status; 1660 err_v6_add_protocol: 1661 sctp_v4_del_protocol(); 1662 err_add_protocol: 1663 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1664 err_register_ctrlsock: 1665 sctp_v6_protosw_exit(); 1666 err_v6_protosw_init: 1667 sctp_v4_protosw_exit(); 1668 err_protosw_init: 1669 unregister_pernet_subsys(&sctp_defaults_ops); 1670 err_register_defaults: 1671 sctp_v4_pf_exit(); 1672 sctp_v6_pf_exit(); 1673 sctp_sysctl_unregister(); 1674 free_pages((unsigned long)sctp_port_hashtable, 1675 get_order(sctp_port_hashsize * 1676 sizeof(struct sctp_bind_hashbucket))); 1677 err_bhash_alloc: 1678 sctp_transport_hashtable_destroy(); 1679 err_thash_alloc: 1680 kfree(sctp_ep_hashtable); 1681 err_ehash_alloc: 1682 percpu_counter_destroy(&sctp_sockets_allocated); 1683 err_percpu_counter_init: 1684 kmem_cache_destroy(sctp_chunk_cachep); 1685 err_chunk_cachep: 1686 kmem_cache_destroy(sctp_bucket_cachep); 1687 goto out; 1688 } 1689 1690 /* Exit handler for the SCTP protocol. */ 1691 static __exit void sctp_exit(void) 1692 { 1693 /* BUG. This should probably do something useful like clean 1694 * up all the remaining associations and all that memory. 1695 */ 1696 1697 /* Unregister with inet6/inet layers. */ 1698 sctp_v6_del_protocol(); 1699 sctp_v4_del_protocol(); 1700 1701 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1702 1703 /* Free protosw registrations */ 1704 sctp_v6_protosw_exit(); 1705 sctp_v4_protosw_exit(); 1706 1707 unregister_pernet_subsys(&sctp_defaults_ops); 1708 1709 /* Unregister with socket layer. */ 1710 sctp_v6_pf_exit(); 1711 sctp_v4_pf_exit(); 1712 1713 sctp_sysctl_unregister(); 1714 1715 free_pages((unsigned long)sctp_port_hashtable, 1716 get_order(sctp_port_hashsize * 1717 sizeof(struct sctp_bind_hashbucket))); 1718 kfree(sctp_ep_hashtable); 1719 sctp_transport_hashtable_destroy(); 1720 1721 percpu_counter_destroy(&sctp_sockets_allocated); 1722 1723 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1724 1725 kmem_cache_destroy(sctp_chunk_cachep); 1726 kmem_cache_destroy(sctp_bucket_cachep); 1727 } 1728 1729 module_init(sctp_init); 1730 module_exit(sctp_exit); 1731 1732 /* 1733 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1734 */ 1735 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1736 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1737 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1738 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1739 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1740 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1741 MODULE_LICENSE("GPL"); 1742