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/flow.h> 38 #include <net/net_namespace.h> 39 #include <net/protocol.h> 40 #include <net/ip.h> 41 #include <net/ipv6.h> 42 #include <net/route.h> 43 #include <net/sctp/sctp.h> 44 #include <net/addrconf.h> 45 #include <net/inet_common.h> 46 #include <net/inet_ecn.h> 47 #include <net/inet_sock.h> 48 #include <net/udp_tunnel.h> 49 #include <net/inet_dscp.h> 50 51 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024) 52 53 /* Global data structures. */ 54 struct sctp_globals sctp_globals __read_mostly; 55 56 struct idr sctp_assocs_id; 57 DEFINE_SPINLOCK(sctp_assocs_id_lock); 58 59 static struct sctp_pf *sctp_pf_inet6_specific; 60 static struct sctp_pf *sctp_pf_inet_specific; 61 static struct sctp_af *sctp_af_v4_specific; 62 static struct sctp_af *sctp_af_v6_specific; 63 64 struct kmem_cache *sctp_chunk_cachep __read_mostly; 65 struct kmem_cache *sctp_bucket_cachep __read_mostly; 66 67 long sysctl_sctp_mem[3]; 68 int sysctl_sctp_rmem[3]; 69 int sysctl_sctp_wmem[3]; 70 71 /* Private helper to extract ipv4 address and stash them in 72 * the protocol structure. 73 */ 74 static void sctp_v4_copy_addrlist(struct list_head *addrlist, 75 struct net_device *dev) 76 { 77 struct in_device *in_dev; 78 struct in_ifaddr *ifa; 79 struct sctp_sockaddr_entry *addr; 80 81 rcu_read_lock(); 82 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 83 rcu_read_unlock(); 84 return; 85 } 86 87 in_dev_for_each_ifa_rcu(ifa, in_dev) { 88 /* Add the address to the local list. */ 89 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 90 if (addr) { 91 addr->a.v4.sin_family = AF_INET; 92 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 93 addr->valid = 1; 94 INIT_LIST_HEAD(&addr->list); 95 list_add_tail(&addr->list, addrlist); 96 } 97 } 98 99 rcu_read_unlock(); 100 } 101 102 /* Extract our IP addresses from the system and stash them in the 103 * protocol structure. 104 */ 105 static void sctp_get_local_addr_list(struct net *net) 106 { 107 struct net_device *dev; 108 struct list_head *pos; 109 struct sctp_af *af; 110 111 rcu_read_lock(); 112 for_each_netdev_rcu(net, dev) { 113 list_for_each(pos, &sctp_address_families) { 114 af = list_entry(pos, struct sctp_af, list); 115 af->copy_addrlist(&net->sctp.local_addr_list, dev); 116 } 117 } 118 rcu_read_unlock(); 119 } 120 121 /* Free the existing local addresses. */ 122 static void sctp_free_local_addr_list(struct net *net) 123 { 124 struct sctp_sockaddr_entry *addr; 125 struct list_head *pos, *temp; 126 127 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 128 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 129 list_del(pos); 130 kfree(addr); 131 } 132 } 133 134 /* Copy the local addresses which are valid for 'scope' into 'bp'. */ 135 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 136 enum sctp_scope scope, gfp_t gfp, int copy_flags) 137 { 138 struct sctp_sockaddr_entry *addr; 139 union sctp_addr laddr; 140 int error = 0; 141 142 rcu_read_lock(); 143 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 144 if (!addr->valid) 145 continue; 146 if (!sctp_in_scope(net, &addr->a, scope)) 147 continue; 148 149 /* Now that the address is in scope, check to see if 150 * the address type is really supported by the local 151 * sock as well as the remote peer. 152 */ 153 if (addr->a.sa.sa_family == AF_INET && 154 (!(copy_flags & SCTP_ADDR4_ALLOWED) || 155 !(copy_flags & SCTP_ADDR4_PEERSUPP))) 156 continue; 157 if (addr->a.sa.sa_family == AF_INET6 && 158 (!(copy_flags & SCTP_ADDR6_ALLOWED) || 159 !(copy_flags & SCTP_ADDR6_PEERSUPP))) 160 continue; 161 162 laddr = addr->a; 163 /* also works for setting ipv6 address port */ 164 laddr.v4.sin_port = htons(bp->port); 165 if (sctp_bind_addr_state(bp, &laddr) != -1) 166 continue; 167 168 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a), 169 SCTP_ADDR_SRC, GFP_ATOMIC); 170 if (error) 171 break; 172 } 173 174 rcu_read_unlock(); 175 return error; 176 } 177 178 /* Copy over any ip options */ 179 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk) 180 { 181 struct inet_sock *newinet, *inet = inet_sk(sk); 182 struct ip_options_rcu *inet_opt, *newopt = NULL; 183 184 newinet = inet_sk(newsk); 185 186 rcu_read_lock(); 187 inet_opt = rcu_dereference(inet->inet_opt); 188 if (inet_opt) { 189 newopt = sock_kmemdup(newsk, inet_opt, sizeof(*inet_opt) + 190 inet_opt->opt.optlen, GFP_ATOMIC); 191 if (!newopt) 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 !inet_test_bit(FREEBIND, sk) && 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 dscp_t dscp; 430 431 if (t->dscp & SCTP_DSCP_SET_MASK) 432 dscp = inet_dsfield_to_dscp(t->dscp); 433 else 434 dscp = inet_sk_dscp(inet_sk(sk)); 435 436 memset(&_fl, 0x0, sizeof(_fl)); 437 fl4->daddr = daddr->v4.sin_addr.s_addr; 438 fl4->fl4_dport = daddr->v4.sin_port; 439 fl4->flowi4_proto = IPPROTO_SCTP; 440 if (asoc) { 441 fl4->flowi4_dscp = dscp; 442 fl4->flowi4_scope = ip_sock_rt_scope(asoc->base.sk); 443 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 444 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 445 } 446 if (saddr) { 447 fl4->saddr = saddr->v4.sin_addr.s_addr; 448 if (!fl4->fl4_sport) 449 fl4->fl4_sport = saddr->v4.sin_port; 450 } 451 452 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 453 &fl4->saddr); 454 455 rt = ip_route_output_key(sock_net(sk), fl4); 456 if (!IS_ERR(rt)) { 457 dst = &rt->dst; 458 t->dst = dst; 459 memcpy(fl, &_fl, sizeof(_fl)); 460 } 461 462 /* If there is no association or if a source address is passed, no 463 * more validation is required. 464 */ 465 if (!asoc || saddr) 466 goto out; 467 468 bp = &asoc->base.bind_addr; 469 470 if (dst) { 471 /* Walk through the bind address list and look for a bind 472 * address that matches the source address of the returned dst. 473 */ 474 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 475 rcu_read_lock(); 476 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 477 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 478 (laddr->state != SCTP_ADDR_SRC && 479 !asoc->src_out_of_asoc_ok)) 480 continue; 481 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 482 goto out_unlock; 483 } 484 rcu_read_unlock(); 485 486 /* None of the bound addresses match the source address of the 487 * dst. So release it. 488 */ 489 dst_release(dst); 490 dst = NULL; 491 } 492 493 /* Walk through the bind address list and try to get a dst that 494 * matches a bind address as the source address. 495 */ 496 rcu_read_lock(); 497 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 498 struct net_device *odev; 499 500 if (!laddr->valid) 501 continue; 502 if (laddr->state != SCTP_ADDR_SRC || 503 AF_INET != laddr->a.sa.sa_family) 504 continue; 505 506 fl4->fl4_sport = laddr->a.v4.sin_port; 507 flowi4_update_output(fl4, asoc->base.sk->sk_bound_dev_if, 508 daddr->v4.sin_addr.s_addr, 509 laddr->a.v4.sin_addr.s_addr); 510 511 rt = ip_route_output_key(sock_net(sk), fl4); 512 if (IS_ERR(rt)) 513 continue; 514 515 /* Ensure the src address belongs to the output 516 * interface. 517 */ 518 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr, 519 false); 520 if (!odev || odev->ifindex != fl4->flowi4_oif) { 521 if (!dst) { 522 dst = &rt->dst; 523 t->dst = dst; 524 memcpy(fl, &_fl, sizeof(_fl)); 525 } else { 526 dst_release(&rt->dst); 527 } 528 continue; 529 } 530 531 dst_release(dst); 532 dst = &rt->dst; 533 t->dst = dst; 534 memcpy(fl, &_fl, sizeof(_fl)); 535 break; 536 } 537 538 out_unlock: 539 rcu_read_unlock(); 540 out: 541 if (dst) { 542 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 543 &fl->u.ip4.daddr, &fl->u.ip4.saddr); 544 } else { 545 t->dst = NULL; 546 pr_debug("no route\n"); 547 } 548 } 549 550 /* For v4, the source address is cached in the route entry(dst). So no need 551 * to cache it separately and hence this is an empty routine. 552 */ 553 static void sctp_v4_get_saddr(struct sctp_sock *sk, 554 struct sctp_transport *t, 555 struct flowi *fl) 556 { 557 union sctp_addr *saddr = &t->saddr; 558 struct rtable *rt = dst_rtable(t->dst); 559 560 if (rt) { 561 saddr->v4.sin_family = AF_INET; 562 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 563 } 564 } 565 566 /* What interface did this skb arrive on? */ 567 static int sctp_v4_skb_iif(const struct sk_buff *skb) 568 { 569 return inet_iif(skb); 570 } 571 572 static int sctp_v4_skb_sdif(const struct sk_buff *skb) 573 { 574 return inet_sdif(skb); 575 } 576 577 /* Was this packet marked by Explicit Congestion Notification? */ 578 static int sctp_v4_is_ce(const struct sk_buff *skb) 579 { 580 return INET_ECN_is_ce(ip_hdr(skb)->tos); 581 } 582 583 /* Create and initialize a new sk for the socket returned by accept(). */ 584 static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 585 struct sctp_association *asoc, 586 bool kern) 587 { 588 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 589 sk->sk_prot, kern); 590 struct inet_sock *newinet; 591 592 if (!newsk) 593 goto out; 594 595 sock_init_data(NULL, newsk); 596 597 sctp_copy_sock(newsk, sk, asoc); 598 sock_reset_flag(newsk, SOCK_ZAPPED); 599 600 sctp_v4_copy_ip_options(sk, newsk); 601 602 newinet = inet_sk(newsk); 603 604 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 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 = timer_container_of(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 timer_delete(&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); 743 744 /* Avoid searching the queue or modifying it if there are no consumers, 745 * as it can lead to performance degradation if addresses are modified 746 * en-masse. 747 * 748 * If the queue already contains some events, update it anyway to avoid 749 * ugly races between new sessions and new address events. 750 */ 751 if (list_empty(&net->sctp.auto_asconf_splist) && 752 list_empty(&net->sctp.addr_waitq)) { 753 spin_unlock_bh(&net->sctp.addr_wq_lock); 754 return; 755 } 756 757 /* Offsets existing events in addr_wq */ 758 addrw = sctp_addr_wq_lookup(net, addr); 759 if (addrw) { 760 if (addrw->state != cmd) { 761 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 762 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 763 &net->sctp.addr_waitq); 764 765 list_del(&addrw->list); 766 kfree(addrw); 767 } 768 spin_unlock_bh(&net->sctp.addr_wq_lock); 769 return; 770 } 771 772 /* OK, we have to add the new address to the wait queue */ 773 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 774 if (addrw == NULL) { 775 spin_unlock_bh(&net->sctp.addr_wq_lock); 776 return; 777 } 778 addrw->state = cmd; 779 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 780 781 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 782 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 783 784 if (!timer_pending(&net->sctp.addr_wq_timer)) { 785 timeo_val = jiffies; 786 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 787 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 788 } 789 spin_unlock_bh(&net->sctp.addr_wq_lock); 790 } 791 792 /* Event handler for inet address addition/deletion events. 793 * The sctp_local_addr_list needs to be protocted by a spin lock since 794 * multiple notifiers (say IPv4 and IPv6) may be running at the same 795 * time and thus corrupt the list. 796 * The reader side is protected with RCU. 797 */ 798 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 799 void *ptr) 800 { 801 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 802 struct sctp_sockaddr_entry *addr = NULL; 803 struct sctp_sockaddr_entry *temp; 804 struct net *net = dev_net(ifa->ifa_dev->dev); 805 int found = 0; 806 807 switch (ev) { 808 case NETDEV_UP: 809 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 810 if (addr) { 811 addr->a.v4.sin_family = AF_INET; 812 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 813 addr->valid = 1; 814 spin_lock_bh(&net->sctp.local_addr_lock); 815 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 816 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 817 spin_unlock_bh(&net->sctp.local_addr_lock); 818 } 819 break; 820 case NETDEV_DOWN: 821 spin_lock_bh(&net->sctp.local_addr_lock); 822 list_for_each_entry_safe(addr, temp, 823 &net->sctp.local_addr_list, list) { 824 if (addr->a.sa.sa_family == AF_INET && 825 addr->a.v4.sin_addr.s_addr == 826 ifa->ifa_local) { 827 found = 1; 828 addr->valid = 0; 829 list_del_rcu(&addr->list); 830 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 831 break; 832 } 833 } 834 spin_unlock_bh(&net->sctp.local_addr_lock); 835 if (found) 836 kfree_rcu(addr, rcu); 837 break; 838 } 839 840 return NOTIFY_DONE; 841 } 842 843 /* 844 * Initialize the control inode/socket with a control endpoint data 845 * structure. This endpoint is reserved exclusively for the OOTB processing. 846 */ 847 static int sctp_ctl_sock_init(struct net *net) 848 { 849 int err; 850 sa_family_t family = PF_INET; 851 852 if (sctp_get_pf_specific(PF_INET6)) 853 family = PF_INET6; 854 855 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 856 SOCK_SEQPACKET, IPPROTO_SCTP, net); 857 858 /* If IPv6 socket could not be created, try the IPv4 socket */ 859 if (err < 0 && family == PF_INET6) 860 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 861 SOCK_SEQPACKET, IPPROTO_SCTP, 862 net); 863 864 if (err < 0) { 865 pr_err("Failed to create the SCTP control socket\n"); 866 return err; 867 } 868 return 0; 869 } 870 871 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb) 872 { 873 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source; 874 875 skb_set_transport_header(skb, sizeof(struct udphdr)); 876 sctp_rcv(skb); 877 return 0; 878 } 879 880 int sctp_udp_sock_start(struct net *net) 881 { 882 struct udp_tunnel_sock_cfg tuncfg = {NULL}; 883 struct udp_port_cfg udp_conf = {0}; 884 struct socket *sock; 885 int err; 886 887 udp_conf.family = AF_INET; 888 udp_conf.local_ip.s_addr = htonl(INADDR_ANY); 889 udp_conf.local_udp_port = htons(net->sctp.udp_port); 890 err = udp_sock_create(net, &udp_conf, &sock); 891 if (err) { 892 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n"); 893 return err; 894 } 895 896 tuncfg.encap_type = 1; 897 tuncfg.encap_rcv = sctp_udp_rcv; 898 tuncfg.encap_err_lookup = sctp_udp_v4_err; 899 setup_udp_tunnel_sock(net, sock, &tuncfg); 900 net->sctp.udp4_sock = sock->sk; 901 902 #if IS_ENABLED(CONFIG_IPV6) 903 memset(&udp_conf, 0, sizeof(udp_conf)); 904 905 udp_conf.family = AF_INET6; 906 udp_conf.local_ip6 = in6addr_any; 907 udp_conf.local_udp_port = htons(net->sctp.udp_port); 908 udp_conf.use_udp6_rx_checksums = true; 909 udp_conf.ipv6_v6only = true; 910 err = udp_sock_create(net, &udp_conf, &sock); 911 if (err) { 912 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n"); 913 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 914 net->sctp.udp4_sock = NULL; 915 return err; 916 } 917 918 tuncfg.encap_type = 1; 919 tuncfg.encap_rcv = sctp_udp_rcv; 920 tuncfg.encap_err_lookup = sctp_udp_v6_err; 921 setup_udp_tunnel_sock(net, sock, &tuncfg); 922 net->sctp.udp6_sock = sock->sk; 923 #endif 924 925 return 0; 926 } 927 928 void sctp_udp_sock_stop(struct net *net) 929 { 930 if (net->sctp.udp4_sock) { 931 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 932 net->sctp.udp4_sock = NULL; 933 } 934 if (net->sctp.udp6_sock) { 935 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket); 936 net->sctp.udp6_sock = NULL; 937 } 938 } 939 940 /* Register address family specific functions. */ 941 int sctp_register_af(struct sctp_af *af) 942 { 943 switch (af->sa_family) { 944 case AF_INET: 945 if (sctp_af_v4_specific) 946 return 0; 947 sctp_af_v4_specific = af; 948 break; 949 case AF_INET6: 950 if (sctp_af_v6_specific) 951 return 0; 952 sctp_af_v6_specific = af; 953 break; 954 default: 955 return 0; 956 } 957 958 INIT_LIST_HEAD(&af->list); 959 list_add_tail(&af->list, &sctp_address_families); 960 return 1; 961 } 962 963 /* Get the table of functions for manipulating a particular address 964 * family. 965 */ 966 struct sctp_af *sctp_get_af_specific(sa_family_t family) 967 { 968 switch (family) { 969 case AF_INET: 970 return sctp_af_v4_specific; 971 case AF_INET6: 972 return sctp_af_v6_specific; 973 default: 974 return NULL; 975 } 976 } 977 978 /* Common code to initialize a AF_INET msg_name. */ 979 static void sctp_inet_msgname(char *msgname, int *addr_len) 980 { 981 struct sockaddr_in *sin; 982 983 sin = (struct sockaddr_in *)msgname; 984 *addr_len = sizeof(struct sockaddr_in); 985 sin->sin_family = AF_INET; 986 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 987 } 988 989 /* Copy the primary address of the peer primary address as the msg_name. */ 990 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 991 int *addr_len) 992 { 993 struct sockaddr_in *sin, *sinfrom; 994 995 if (msgname) { 996 struct sctp_association *asoc; 997 998 asoc = event->asoc; 999 sctp_inet_msgname(msgname, addr_len); 1000 sin = (struct sockaddr_in *)msgname; 1001 sinfrom = &asoc->peer.primary_addr.v4; 1002 sin->sin_port = htons(asoc->peer.port); 1003 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 1004 } 1005 } 1006 1007 /* Initialize and copy out a msgname from an inbound skb. */ 1008 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 1009 { 1010 if (msgname) { 1011 struct sctphdr *sh = sctp_hdr(skb); 1012 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 1013 1014 sctp_inet_msgname(msgname, len); 1015 sin->sin_port = sh->source; 1016 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 1017 } 1018 } 1019 1020 /* Do we support this AF? */ 1021 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 1022 { 1023 /* PF_INET only supports AF_INET addresses. */ 1024 return AF_INET == family; 1025 } 1026 1027 /* Address matching with wildcards allowed. */ 1028 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 1029 const union sctp_addr *addr2, 1030 struct sctp_sock *opt) 1031 { 1032 /* PF_INET only supports AF_INET addresses. */ 1033 if (addr1->sa.sa_family != addr2->sa.sa_family) 1034 return 0; 1035 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 1036 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 1037 return 1; 1038 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 1039 return 1; 1040 1041 return 0; 1042 } 1043 1044 /* Verify that provided sockaddr looks bindable. Common verification has 1045 * already been taken care of. 1046 */ 1047 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 1048 { 1049 return sctp_v4_available(addr, opt); 1050 } 1051 1052 /* Verify that sockaddr looks sendable. Common verification has already 1053 * been taken care of. 1054 */ 1055 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 1056 { 1057 return 1; 1058 } 1059 1060 /* Fill in Supported Address Type information for INIT and INIT-ACK 1061 * chunks. Returns number of addresses supported. 1062 */ 1063 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 1064 __be16 *types) 1065 { 1066 types[0] = SCTP_PARAM_IPV4_ADDRESS; 1067 return 1; 1068 } 1069 1070 /* Wrapper routine that calls the ip transmit routine. */ 1071 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t) 1072 { 1073 struct dst_entry *dst = dst_clone(t->dst); 1074 struct flowi4 *fl4 = &t->fl.u.ip4; 1075 struct sock *sk = skb->sk; 1076 struct inet_sock *inet = inet_sk(sk); 1077 __u8 dscp = READ_ONCE(inet->tos); 1078 __be16 df = 0; 1079 1080 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 1081 skb->len, &fl4->saddr, &fl4->daddr); 1082 1083 if (t->dscp & SCTP_DSCP_SET_MASK) 1084 dscp = t->dscp & SCTP_DSCP_VAL_MASK; 1085 1086 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO 1087 : IP_PMTUDISC_DONT; 1088 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS); 1089 1090 if (!t->encap_port || !sctp_sk(sk)->udp_port) { 1091 skb_dst_set(skb, dst); 1092 return __ip_queue_xmit(sk, skb, &t->fl, dscp); 1093 } 1094 1095 if (skb_is_gso(skb)) 1096 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM; 1097 1098 if (ip_dont_fragment(sk, dst) && !skb->ignore_df) 1099 df = htons(IP_DF); 1100 1101 skb->encapsulation = 1; 1102 skb_reset_inner_mac_header(skb); 1103 skb_reset_inner_transport_header(skb); 1104 skb_set_inner_ipproto(skb, IPPROTO_SCTP); 1105 udp_tunnel_xmit_skb(dst_rtable(dst), sk, skb, fl4->saddr, 1106 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df, 1107 sctp_sk(sk)->udp_port, t->encap_port, false, false, 1108 0); 1109 return 0; 1110 } 1111 1112 static struct sctp_af sctp_af_inet; 1113 1114 static struct sctp_pf sctp_pf_inet = { 1115 .event_msgname = sctp_inet_event_msgname, 1116 .skb_msgname = sctp_inet_skb_msgname, 1117 .af_supported = sctp_inet_af_supported, 1118 .cmp_addr = sctp_inet_cmp_addr, 1119 .bind_verify = sctp_inet_bind_verify, 1120 .send_verify = sctp_inet_send_verify, 1121 .supported_addrs = sctp_inet_supported_addrs, 1122 .create_accept_sk = sctp_v4_create_accept_sk, 1123 .addr_to_user = sctp_v4_addr_to_user, 1124 .to_sk_saddr = sctp_v4_to_sk_saddr, 1125 .to_sk_daddr = sctp_v4_to_sk_daddr, 1126 .copy_ip_options = sctp_v4_copy_ip_options, 1127 .af = &sctp_af_inet 1128 }; 1129 1130 /* Notifier for inetaddr addition/deletion events. */ 1131 static struct notifier_block sctp_inetaddr_notifier = { 1132 .notifier_call = sctp_inetaddr_event, 1133 }; 1134 1135 /* Socket operations. */ 1136 static const struct proto_ops inet_seqpacket_ops = { 1137 .family = PF_INET, 1138 .owner = THIS_MODULE, 1139 .release = inet_release, /* Needs to be wrapped... */ 1140 .bind = inet_bind, 1141 .connect = sctp_inet_connect, 1142 .socketpair = sock_no_socketpair, 1143 .accept = inet_accept, 1144 .getname = inet_getname, /* Semantics are different. */ 1145 .poll = sctp_poll, 1146 .ioctl = inet_ioctl, 1147 .gettstamp = sock_gettstamp, 1148 .listen = sctp_inet_listen, 1149 .shutdown = inet_shutdown, /* Looks harmless. */ 1150 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1151 .getsockopt = sock_common_getsockopt, 1152 .sendmsg = inet_sendmsg, 1153 .recvmsg = inet_recvmsg, 1154 .mmap = sock_no_mmap, 1155 }; 1156 1157 /* Registration with AF_INET family. */ 1158 static struct inet_protosw sctp_seqpacket_protosw = { 1159 .type = SOCK_SEQPACKET, 1160 .protocol = IPPROTO_SCTP, 1161 .prot = &sctp_prot, 1162 .ops = &inet_seqpacket_ops, 1163 .flags = SCTP_PROTOSW_FLAG 1164 }; 1165 static struct inet_protosw sctp_stream_protosw = { 1166 .type = SOCK_STREAM, 1167 .protocol = IPPROTO_SCTP, 1168 .prot = &sctp_prot, 1169 .ops = &inet_seqpacket_ops, 1170 .flags = SCTP_PROTOSW_FLAG 1171 }; 1172 1173 static int sctp4_rcv(struct sk_buff *skb) 1174 { 1175 SCTP_INPUT_CB(skb)->encap_port = 0; 1176 return sctp_rcv(skb); 1177 } 1178 1179 /* Register with IP layer. */ 1180 static const struct net_protocol sctp_protocol = { 1181 .handler = sctp4_rcv, 1182 .err_handler = sctp_v4_err, 1183 .no_policy = 1, 1184 .icmp_strict_tag_validation = 1, 1185 }; 1186 1187 /* IPv4 address related functions. */ 1188 static struct sctp_af sctp_af_inet = { 1189 .sa_family = AF_INET, 1190 .sctp_xmit = sctp_v4_xmit, 1191 .setsockopt = ip_setsockopt, 1192 .getsockopt = ip_getsockopt, 1193 .get_dst = sctp_v4_get_dst, 1194 .get_saddr = sctp_v4_get_saddr, 1195 .copy_addrlist = sctp_v4_copy_addrlist, 1196 .from_skb = sctp_v4_from_skb, 1197 .from_sk = sctp_v4_from_sk, 1198 .from_addr_param = sctp_v4_from_addr_param, 1199 .to_addr_param = sctp_v4_to_addr_param, 1200 .cmp_addr = sctp_v4_cmp_addr, 1201 .addr_valid = sctp_v4_addr_valid, 1202 .inaddr_any = sctp_v4_inaddr_any, 1203 .is_any = sctp_v4_is_any, 1204 .available = sctp_v4_available, 1205 .scope = sctp_v4_scope, 1206 .skb_iif = sctp_v4_skb_iif, 1207 .skb_sdif = sctp_v4_skb_sdif, 1208 .is_ce = sctp_v4_is_ce, 1209 .seq_dump_addr = sctp_v4_seq_dump_addr, 1210 .ecn_capable = sctp_v4_ecn_capable, 1211 .net_header_len = sizeof(struct iphdr), 1212 .sockaddr_len = sizeof(struct sockaddr_in), 1213 .ip_options_len = sctp_v4_ip_options_len, 1214 }; 1215 1216 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1217 { 1218 switch (family) { 1219 case PF_INET: 1220 return sctp_pf_inet_specific; 1221 case PF_INET6: 1222 return sctp_pf_inet6_specific; 1223 default: 1224 return NULL; 1225 } 1226 } 1227 1228 /* Register the PF specific function table. */ 1229 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1230 { 1231 switch (family) { 1232 case PF_INET: 1233 if (sctp_pf_inet_specific) 1234 return 0; 1235 sctp_pf_inet_specific = pf; 1236 break; 1237 case PF_INET6: 1238 if (sctp_pf_inet6_specific) 1239 return 0; 1240 sctp_pf_inet6_specific = pf; 1241 break; 1242 default: 1243 return 0; 1244 } 1245 return 1; 1246 } 1247 1248 static inline int init_sctp_mibs(struct net *net) 1249 { 1250 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1251 if (!net->sctp.sctp_statistics) 1252 return -ENOMEM; 1253 return 0; 1254 } 1255 1256 static inline void cleanup_sctp_mibs(struct net *net) 1257 { 1258 free_percpu(net->sctp.sctp_statistics); 1259 } 1260 1261 static void sctp_v4_pf_init(void) 1262 { 1263 /* Initialize the SCTP specific PF functions. */ 1264 sctp_register_pf(&sctp_pf_inet, PF_INET); 1265 sctp_register_af(&sctp_af_inet); 1266 } 1267 1268 static void sctp_v4_pf_exit(void) 1269 { 1270 list_del(&sctp_af_inet.list); 1271 } 1272 1273 static int sctp_v4_protosw_init(void) 1274 { 1275 int rc; 1276 1277 rc = proto_register(&sctp_prot, 1); 1278 if (rc) 1279 return rc; 1280 1281 /* Register SCTP(UDP and TCP style) with socket layer. */ 1282 inet_register_protosw(&sctp_seqpacket_protosw); 1283 inet_register_protosw(&sctp_stream_protosw); 1284 1285 return 0; 1286 } 1287 1288 static void sctp_v4_protosw_exit(void) 1289 { 1290 inet_unregister_protosw(&sctp_stream_protosw); 1291 inet_unregister_protosw(&sctp_seqpacket_protosw); 1292 proto_unregister(&sctp_prot); 1293 } 1294 1295 static int sctp_v4_add_protocol(void) 1296 { 1297 /* Register notifier for inet address additions/deletions. */ 1298 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1299 1300 /* Register SCTP with inet layer. */ 1301 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1302 return -EAGAIN; 1303 1304 return 0; 1305 } 1306 1307 static void sctp_v4_del_protocol(void) 1308 { 1309 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1310 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1311 } 1312 1313 static int __net_init sctp_defaults_init(struct net *net) 1314 { 1315 int status; 1316 1317 /* 1318 * 14. Suggested SCTP Protocol Parameter Values 1319 */ 1320 /* The following protocol parameters are RECOMMENDED: */ 1321 /* RTO.Initial - 3 seconds */ 1322 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1323 /* RTO.Min - 1 second */ 1324 net->sctp.rto_min = SCTP_RTO_MIN; 1325 /* RTO.Max - 60 seconds */ 1326 net->sctp.rto_max = SCTP_RTO_MAX; 1327 /* RTO.Alpha - 1/8 */ 1328 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1329 /* RTO.Beta - 1/4 */ 1330 net->sctp.rto_beta = SCTP_RTO_BETA; 1331 1332 /* Valid.Cookie.Life - 60 seconds */ 1333 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1334 1335 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1336 net->sctp.cookie_preserve_enable = 1; 1337 1338 /* Whether cookie authentication is enabled(1) or not(0) */ 1339 net->sctp.cookie_auth_enable = 1340 !IS_ENABLED(CONFIG_SCTP_DEFAULT_COOKIE_HMAC_NONE); 1341 1342 /* Max.Burst - 4 */ 1343 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1344 1345 /* Disable of Primary Path Switchover by default */ 1346 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX; 1347 1348 /* Enable pf state by default */ 1349 net->sctp.pf_enable = 1; 1350 1351 /* Ignore pf exposure feature by default */ 1352 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET; 1353 1354 /* Association.Max.Retrans - 10 attempts 1355 * Path.Max.Retrans - 5 attempts (per destination address) 1356 * Max.Init.Retransmits - 8 attempts 1357 */ 1358 net->sctp.max_retrans_association = 10; 1359 net->sctp.max_retrans_path = 5; 1360 net->sctp.max_retrans_init = 8; 1361 1362 /* Sendbuffer growth - do per-socket accounting */ 1363 net->sctp.sndbuf_policy = 0; 1364 1365 /* Rcvbuffer growth - do per-socket accounting */ 1366 net->sctp.rcvbuf_policy = 0; 1367 1368 /* HB.interval - 30 seconds */ 1369 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1370 1371 /* delayed SACK timeout */ 1372 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1373 1374 /* Disable ADDIP by default. */ 1375 net->sctp.addip_enable = 0; 1376 net->sctp.addip_noauth = 0; 1377 net->sctp.default_auto_asconf = 0; 1378 1379 /* Enable PR-SCTP by default. */ 1380 net->sctp.prsctp_enable = 1; 1381 1382 /* Disable RECONF by default. */ 1383 net->sctp.reconf_enable = 0; 1384 1385 /* Disable AUTH by default. */ 1386 net->sctp.auth_enable = 0; 1387 1388 /* Enable ECN by default. */ 1389 net->sctp.ecn_enable = 1; 1390 1391 /* Set UDP tunneling listening port to 0 by default */ 1392 net->sctp.udp_port = 0; 1393 1394 /* Set remote encap port to 0 by default */ 1395 net->sctp.encap_port = 0; 1396 1397 /* Set SCOPE policy to enabled */ 1398 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1399 1400 /* Set the default rwnd update threshold */ 1401 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1402 1403 /* Initialize maximum autoclose timeout. */ 1404 net->sctp.max_autoclose = INT_MAX / HZ; 1405 1406 #ifdef CONFIG_NET_L3_MASTER_DEV 1407 net->sctp.l3mdev_accept = 1; 1408 #endif 1409 1410 status = sctp_sysctl_net_register(net); 1411 if (status) 1412 goto err_sysctl_register; 1413 1414 /* Allocate and initialise sctp mibs. */ 1415 status = init_sctp_mibs(net); 1416 if (status) 1417 goto err_init_mibs; 1418 1419 #ifdef CONFIG_PROC_FS 1420 /* Initialize proc fs directory. */ 1421 status = sctp_proc_init(net); 1422 if (status) 1423 goto err_init_proc; 1424 #endif 1425 1426 sctp_dbg_objcnt_init(net); 1427 1428 /* Initialize the local address list. */ 1429 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1430 spin_lock_init(&net->sctp.local_addr_lock); 1431 sctp_get_local_addr_list(net); 1432 1433 /* Initialize the address event list */ 1434 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1435 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1436 spin_lock_init(&net->sctp.addr_wq_lock); 1437 net->sctp.addr_wq_timer.expires = 0; 1438 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1439 1440 return 0; 1441 1442 #ifdef CONFIG_PROC_FS 1443 err_init_proc: 1444 cleanup_sctp_mibs(net); 1445 #endif 1446 err_init_mibs: 1447 sctp_sysctl_net_unregister(net); 1448 err_sysctl_register: 1449 return status; 1450 } 1451 1452 static void __net_exit sctp_defaults_exit(struct net *net) 1453 { 1454 /* Free the local address list */ 1455 sctp_free_addr_wq(net); 1456 sctp_free_local_addr_list(net); 1457 1458 #ifdef CONFIG_PROC_FS 1459 remove_proc_subtree("sctp", net->proc_net); 1460 net->sctp.proc_net_sctp = NULL; 1461 #endif 1462 cleanup_sctp_mibs(net); 1463 sctp_sysctl_net_unregister(net); 1464 } 1465 1466 static struct pernet_operations sctp_defaults_ops = { 1467 .init = sctp_defaults_init, 1468 .exit = sctp_defaults_exit, 1469 }; 1470 1471 static int __net_init sctp_ctrlsock_init(struct net *net) 1472 { 1473 int status; 1474 1475 /* Initialize the control inode/socket for handling OOTB packets. */ 1476 status = sctp_ctl_sock_init(net); 1477 if (status) 1478 pr_err("Failed to initialize the SCTP control sock\n"); 1479 1480 return status; 1481 } 1482 1483 static void __net_exit sctp_ctrlsock_exit(struct net *net) 1484 { 1485 /* Free the control endpoint. */ 1486 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1487 } 1488 1489 static struct pernet_operations sctp_ctrlsock_ops = { 1490 .init = sctp_ctrlsock_init, 1491 .exit = sctp_ctrlsock_exit, 1492 }; 1493 1494 /* Initialize the universe into something sensible. */ 1495 static __init int sctp_init(void) 1496 { 1497 unsigned long nr_pages = totalram_pages(); 1498 unsigned long limit; 1499 unsigned long goal; 1500 int max_entry_order; 1501 int num_entries; 1502 int max_share; 1503 int status; 1504 int order; 1505 int i; 1506 1507 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1508 1509 /* Allocate bind_bucket and chunk caches. */ 1510 status = -ENOBUFS; 1511 sctp_bucket_cachep = KMEM_CACHE(sctp_bind_bucket, SLAB_HWCACHE_ALIGN); 1512 if (!sctp_bucket_cachep) 1513 goto out; 1514 1515 sctp_chunk_cachep = KMEM_CACHE(sctp_chunk, SLAB_HWCACHE_ALIGN); 1516 if (!sctp_chunk_cachep) 1517 goto err_chunk_cachep; 1518 1519 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1520 if (status) 1521 goto err_percpu_counter_init; 1522 1523 /* Implementation specific variables. */ 1524 1525 /* Initialize default stream count setup information. */ 1526 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1527 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1528 1529 /* Initialize handle used for association ids. */ 1530 idr_init(&sctp_assocs_id); 1531 1532 limit = nr_free_buffer_pages() / 8; 1533 limit = max(limit, 128UL); 1534 sysctl_sctp_mem[0] = limit / 4 * 3; 1535 sysctl_sctp_mem[1] = limit; 1536 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1537 1538 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1539 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1540 max_share = min(4UL*1024*1024, limit); 1541 1542 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */ 1543 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1544 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1545 1546 sysctl_sctp_wmem[0] = PAGE_SIZE; 1547 sysctl_sctp_wmem[1] = 16*1024; 1548 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1549 1550 /* Size and allocate the association hash table. 1551 * The methodology is similar to that of the tcp hash tables. 1552 * Though not identical. Start by getting a goal size 1553 */ 1554 if (nr_pages >= (128 * 1024)) 1555 goal = nr_pages >> (22 - PAGE_SHIFT); 1556 else 1557 goal = nr_pages >> (24 - PAGE_SHIFT); 1558 1559 /* Then compute the page order for said goal */ 1560 order = get_order(goal); 1561 1562 /* Now compute the required page order for the maximum sized table we 1563 * want to create 1564 */ 1565 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1566 sizeof(struct sctp_bind_hashbucket)); 1567 1568 /* Limit the page order by that maximum hash table size */ 1569 order = min(order, max_entry_order); 1570 1571 /* Allocate and initialize the endpoint hash table. */ 1572 sctp_ep_hashsize = 64; 1573 sctp_ep_hashtable = 1574 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1575 if (!sctp_ep_hashtable) { 1576 pr_err("Failed endpoint_hash alloc\n"); 1577 status = -ENOMEM; 1578 goto err_ehash_alloc; 1579 } 1580 for (i = 0; i < sctp_ep_hashsize; i++) { 1581 rwlock_init(&sctp_ep_hashtable[i].lock); 1582 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1583 } 1584 1585 /* Allocate and initialize the SCTP port hash table. 1586 * Note that order is initalized to start at the max sized 1587 * table we want to support. If we can't get that many pages 1588 * reduce the order and try again 1589 */ 1590 do { 1591 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1592 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1593 } while (!sctp_port_hashtable && --order > 0); 1594 1595 if (!sctp_port_hashtable) { 1596 pr_err("Failed bind hash alloc\n"); 1597 status = -ENOMEM; 1598 goto err_bhash_alloc; 1599 } 1600 1601 /* Now compute the number of entries that will fit in the 1602 * port hash space we allocated 1603 */ 1604 num_entries = (1UL << order) * PAGE_SIZE / 1605 sizeof(struct sctp_bind_hashbucket); 1606 1607 /* And finish by rounding it down to the nearest power of two. 1608 * This wastes some memory of course, but it's needed because 1609 * the hash function operates based on the assumption that 1610 * the number of entries is a power of two. 1611 */ 1612 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1613 1614 for (i = 0; i < sctp_port_hashsize; i++) { 1615 spin_lock_init(&sctp_port_hashtable[i].lock); 1616 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1617 } 1618 1619 status = sctp_transport_hashtable_init(); 1620 if (status) 1621 goto err_thash_alloc; 1622 1623 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1624 num_entries); 1625 1626 sctp_sysctl_register(); 1627 1628 INIT_LIST_HEAD(&sctp_address_families); 1629 sctp_v4_pf_init(); 1630 sctp_v6_pf_init(); 1631 sctp_sched_ops_init(); 1632 1633 status = register_pernet_subsys(&sctp_defaults_ops); 1634 if (status) 1635 goto err_register_defaults; 1636 1637 status = sctp_v4_protosw_init(); 1638 if (status) 1639 goto err_protosw_init; 1640 1641 status = sctp_v6_protosw_init(); 1642 if (status) 1643 goto err_v6_protosw_init; 1644 1645 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1646 if (status) 1647 goto err_register_ctrlsock; 1648 1649 status = sctp_v4_add_protocol(); 1650 if (status) 1651 goto err_add_protocol; 1652 1653 /* Register SCTP with inet6 layer. */ 1654 status = sctp_v6_add_protocol(); 1655 if (status) 1656 goto err_v6_add_protocol; 1657 1658 if (sctp_offload_init() < 0) 1659 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1660 1661 out: 1662 return status; 1663 err_v6_add_protocol: 1664 sctp_v4_del_protocol(); 1665 err_add_protocol: 1666 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1667 err_register_ctrlsock: 1668 sctp_v6_protosw_exit(); 1669 err_v6_protosw_init: 1670 sctp_v4_protosw_exit(); 1671 err_protosw_init: 1672 unregister_pernet_subsys(&sctp_defaults_ops); 1673 err_register_defaults: 1674 sctp_v4_pf_exit(); 1675 sctp_v6_pf_exit(); 1676 sctp_sysctl_unregister(); 1677 free_pages((unsigned long)sctp_port_hashtable, 1678 get_order(sctp_port_hashsize * 1679 sizeof(struct sctp_bind_hashbucket))); 1680 err_bhash_alloc: 1681 sctp_transport_hashtable_destroy(); 1682 err_thash_alloc: 1683 kfree(sctp_ep_hashtable); 1684 err_ehash_alloc: 1685 percpu_counter_destroy(&sctp_sockets_allocated); 1686 err_percpu_counter_init: 1687 kmem_cache_destroy(sctp_chunk_cachep); 1688 err_chunk_cachep: 1689 kmem_cache_destroy(sctp_bucket_cachep); 1690 goto out; 1691 } 1692 1693 /* Exit handler for the SCTP protocol. */ 1694 static __exit void sctp_exit(void) 1695 { 1696 /* BUG. This should probably do something useful like clean 1697 * up all the remaining associations and all that memory. 1698 */ 1699 1700 /* Unregister with inet6/inet layers. */ 1701 sctp_v6_del_protocol(); 1702 sctp_v4_del_protocol(); 1703 1704 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1705 1706 /* Free protosw registrations */ 1707 sctp_v6_protosw_exit(); 1708 sctp_v4_protosw_exit(); 1709 1710 unregister_pernet_subsys(&sctp_defaults_ops); 1711 1712 /* Unregister with socket layer. */ 1713 sctp_v6_pf_exit(); 1714 sctp_v4_pf_exit(); 1715 1716 sctp_sysctl_unregister(); 1717 1718 free_pages((unsigned long)sctp_port_hashtable, 1719 get_order(sctp_port_hashsize * 1720 sizeof(struct sctp_bind_hashbucket))); 1721 kfree(sctp_ep_hashtable); 1722 sctp_transport_hashtable_destroy(); 1723 1724 percpu_counter_destroy(&sctp_sockets_allocated); 1725 1726 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1727 1728 kmem_cache_destroy(sctp_chunk_cachep); 1729 kmem_cache_destroy(sctp_bucket_cachep); 1730 } 1731 1732 module_init(sctp_init); 1733 module_exit(sctp_exit); 1734 1735 /* 1736 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1737 */ 1738 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1739 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1740 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1741 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1742 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1743 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1744 MODULE_LICENSE("GPL"); 1745