1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* SCTP kernel implementation 3 * (C) Copyright IBM Corp. 2001, 2004 4 * Copyright (c) 1999-2000 Cisco, Inc. 5 * Copyright (c) 1999-2001 Motorola, Inc. 6 * Copyright (c) 2001 Intel Corp. 7 * Copyright (c) 2001 Nokia, Inc. 8 * Copyright (c) 2001 La Monte H.P. Yarroll 9 * 10 * This file is part of the SCTP kernel implementation 11 * 12 * Initialization/cleanup for SCTP protocol support. 13 * 14 * Please send any bug reports or fixes you make to the 15 * email address(es): 16 * lksctp developers <linux-sctp@vger.kernel.org> 17 * 18 * Written or modified by: 19 * La Monte H.P. Yarroll <piggy@acm.org> 20 * Karl Knutson <karl@athena.chicago.il.us> 21 * Jon Grimm <jgrimm@us.ibm.com> 22 * Sridhar Samudrala <sri@us.ibm.com> 23 * Daisy Chang <daisyc@us.ibm.com> 24 * Ardelle Fan <ardelle.fan@intel.com> 25 */ 26 27 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 28 29 #include <linux/module.h> 30 #include <linux/init.h> 31 #include <linux/netdevice.h> 32 #include <linux/inetdevice.h> 33 #include <linux/seq_file.h> 34 #include <linux/memblock.h> 35 #include <linux/highmem.h> 36 #include <linux/slab.h> 37 #include <net/net_namespace.h> 38 #include <net/protocol.h> 39 #include <net/ip.h> 40 #include <net/ipv6.h> 41 #include <net/route.h> 42 #include <net/sctp/sctp.h> 43 #include <net/addrconf.h> 44 #include <net/inet_common.h> 45 #include <net/inet_ecn.h> 46 #include <net/inet_sock.h> 47 #include <net/udp_tunnel.h> 48 #include <net/inet_dscp.h> 49 50 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024) 51 52 /* Global data structures. */ 53 struct sctp_globals sctp_globals __read_mostly; 54 55 struct idr sctp_assocs_id; 56 DEFINE_SPINLOCK(sctp_assocs_id_lock); 57 58 static struct sctp_pf *sctp_pf_inet6_specific; 59 static struct sctp_pf *sctp_pf_inet_specific; 60 static struct sctp_af *sctp_af_v4_specific; 61 static struct sctp_af *sctp_af_v6_specific; 62 63 struct kmem_cache *sctp_chunk_cachep __read_mostly; 64 struct kmem_cache *sctp_bucket_cachep __read_mostly; 65 66 long sysctl_sctp_mem[3]; 67 int sysctl_sctp_rmem[3]; 68 int sysctl_sctp_wmem[3]; 69 70 /* Private helper to extract ipv4 address and stash them in 71 * the protocol structure. 72 */ 73 static void sctp_v4_copy_addrlist(struct list_head *addrlist, 74 struct net_device *dev) 75 { 76 struct in_device *in_dev; 77 struct in_ifaddr *ifa; 78 struct sctp_sockaddr_entry *addr; 79 80 rcu_read_lock(); 81 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 82 rcu_read_unlock(); 83 return; 84 } 85 86 in_dev_for_each_ifa_rcu(ifa, in_dev) { 87 /* Add the address to the local list. */ 88 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 89 if (addr) { 90 addr->a.v4.sin_family = AF_INET; 91 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 92 addr->valid = 1; 93 INIT_LIST_HEAD(&addr->list); 94 list_add_tail(&addr->list, addrlist); 95 } 96 } 97 98 rcu_read_unlock(); 99 } 100 101 /* Extract our IP addresses from the system and stash them in the 102 * protocol structure. 103 */ 104 static void sctp_get_local_addr_list(struct net *net) 105 { 106 struct net_device *dev; 107 struct list_head *pos; 108 struct sctp_af *af; 109 110 rcu_read_lock(); 111 for_each_netdev_rcu(net, dev) { 112 list_for_each(pos, &sctp_address_families) { 113 af = list_entry(pos, struct sctp_af, list); 114 af->copy_addrlist(&net->sctp.local_addr_list, dev); 115 } 116 } 117 rcu_read_unlock(); 118 } 119 120 /* Free the existing local addresses. */ 121 static void sctp_free_local_addr_list(struct net *net) 122 { 123 struct sctp_sockaddr_entry *addr; 124 struct list_head *pos, *temp; 125 126 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 127 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 128 list_del(pos); 129 kfree(addr); 130 } 131 } 132 133 /* Copy the local addresses which are valid for 'scope' into 'bp'. */ 134 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 135 enum sctp_scope scope, gfp_t gfp, int copy_flags) 136 { 137 struct sctp_sockaddr_entry *addr; 138 union sctp_addr laddr; 139 int error = 0; 140 141 rcu_read_lock(); 142 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 143 if (!addr->valid) 144 continue; 145 if (!sctp_in_scope(net, &addr->a, scope)) 146 continue; 147 148 /* Now that the address is in scope, check to see if 149 * the address type is really supported by the local 150 * sock as well as the remote peer. 151 */ 152 if (addr->a.sa.sa_family == AF_INET && 153 (!(copy_flags & SCTP_ADDR4_ALLOWED) || 154 !(copy_flags & SCTP_ADDR4_PEERSUPP))) 155 continue; 156 if (addr->a.sa.sa_family == AF_INET6 && 157 (!(copy_flags & SCTP_ADDR6_ALLOWED) || 158 !(copy_flags & SCTP_ADDR6_PEERSUPP))) 159 continue; 160 161 laddr = addr->a; 162 /* also works for setting ipv6 address port */ 163 laddr.v4.sin_port = htons(bp->port); 164 if (sctp_bind_addr_state(bp, &laddr) != -1) 165 continue; 166 167 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a), 168 SCTP_ADDR_SRC, GFP_ATOMIC); 169 if (error) 170 break; 171 } 172 173 rcu_read_unlock(); 174 return error; 175 } 176 177 /* Copy over any ip options */ 178 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk) 179 { 180 struct inet_sock *newinet, *inet = inet_sk(sk); 181 struct ip_options_rcu *inet_opt, *newopt = NULL; 182 183 newinet = inet_sk(newsk); 184 185 rcu_read_lock(); 186 inet_opt = rcu_dereference(inet->inet_opt); 187 if (inet_opt) { 188 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) + 189 inet_opt->opt.optlen, GFP_ATOMIC); 190 if (newopt) 191 memcpy(newopt, inet_opt, sizeof(*inet_opt) + 192 inet_opt->opt.optlen); 193 else 194 pr_err("%s: Failed to copy ip options\n", __func__); 195 } 196 RCU_INIT_POINTER(newinet->inet_opt, newopt); 197 rcu_read_unlock(); 198 } 199 200 /* Account for the IP options */ 201 static int sctp_v4_ip_options_len(struct sock *sk) 202 { 203 struct inet_sock *inet = inet_sk(sk); 204 struct ip_options_rcu *inet_opt; 205 int len = 0; 206 207 rcu_read_lock(); 208 inet_opt = rcu_dereference(inet->inet_opt); 209 if (inet_opt) 210 len = inet_opt->opt.optlen; 211 212 rcu_read_unlock(); 213 return len; 214 } 215 216 /* Initialize a sctp_addr from in incoming skb. */ 217 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 218 int is_saddr) 219 { 220 /* Always called on head skb, so this is safe */ 221 struct sctphdr *sh = sctp_hdr(skb); 222 struct sockaddr_in *sa = &addr->v4; 223 224 addr->v4.sin_family = AF_INET; 225 226 if (is_saddr) { 227 sa->sin_port = sh->source; 228 sa->sin_addr.s_addr = ip_hdr(skb)->saddr; 229 } else { 230 sa->sin_port = sh->dest; 231 sa->sin_addr.s_addr = ip_hdr(skb)->daddr; 232 } 233 memset(sa->sin_zero, 0, sizeof(sa->sin_zero)); 234 } 235 236 /* Initialize an sctp_addr from a socket. */ 237 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 238 { 239 addr->v4.sin_family = AF_INET; 240 addr->v4.sin_port = 0; 241 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 242 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 243 } 244 245 /* Initialize sk->sk_rcv_saddr from sctp_addr. */ 246 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 247 { 248 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 249 } 250 251 /* Initialize sk->sk_daddr from sctp_addr. */ 252 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 253 { 254 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 255 } 256 257 /* Initialize a sctp_addr from an address parameter. */ 258 static bool sctp_v4_from_addr_param(union sctp_addr *addr, 259 union sctp_addr_param *param, 260 __be16 port, int iif) 261 { 262 if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param)) 263 return false; 264 265 addr->v4.sin_family = AF_INET; 266 addr->v4.sin_port = port; 267 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 268 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 269 270 return true; 271 } 272 273 /* Initialize an address parameter from a sctp_addr and return the length 274 * of the address parameter. 275 */ 276 static int sctp_v4_to_addr_param(const union sctp_addr *addr, 277 union sctp_addr_param *param) 278 { 279 int length = sizeof(struct sctp_ipv4addr_param); 280 281 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 282 param->v4.param_hdr.length = htons(length); 283 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 284 285 return length; 286 } 287 288 /* Initialize a sctp_addr from a dst_entry. */ 289 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 290 __be16 port) 291 { 292 saddr->v4.sin_family = AF_INET; 293 saddr->v4.sin_port = port; 294 saddr->v4.sin_addr.s_addr = fl4->saddr; 295 memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero)); 296 } 297 298 /* Compare two addresses exactly. */ 299 static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 300 const union sctp_addr *addr2) 301 { 302 if (addr1->sa.sa_family != addr2->sa.sa_family) 303 return 0; 304 if (addr1->v4.sin_port != addr2->v4.sin_port) 305 return 0; 306 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 307 return 0; 308 309 return 1; 310 } 311 312 /* Initialize addr struct to INADDR_ANY. */ 313 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 314 { 315 addr->v4.sin_family = AF_INET; 316 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 317 addr->v4.sin_port = port; 318 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 319 } 320 321 /* Is this a wildcard address? */ 322 static int sctp_v4_is_any(const union sctp_addr *addr) 323 { 324 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 325 } 326 327 /* This function checks if the address is a valid address to be used for 328 * SCTP binding. 329 * 330 * Output: 331 * Return 0 - If the address is a non-unicast or an illegal address. 332 * Return 1 - If the address is a unicast. 333 */ 334 static int sctp_v4_addr_valid(union sctp_addr *addr, 335 struct sctp_sock *sp, 336 const struct sk_buff *skb) 337 { 338 /* IPv4 addresses not allowed */ 339 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 340 return 0; 341 342 /* Is this a non-unicast address or a unusable SCTP address? */ 343 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 344 return 0; 345 346 /* Is this a broadcast address? */ 347 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 348 return 0; 349 350 return 1; 351 } 352 353 /* Should this be available for binding? */ 354 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 355 { 356 struct sock *sk = &sp->inet.sk; 357 struct net *net = sock_net(sk); 358 int tb_id = RT_TABLE_LOCAL; 359 int ret; 360 361 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ?: tb_id; 362 ret = inet_addr_type_table(net, addr->v4.sin_addr.s_addr, tb_id); 363 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 364 ret != RTN_LOCAL && 365 !inet_test_bit(FREEBIND, sk) && 366 !READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind)) 367 return 0; 368 369 if (ipv6_only_sock(sctp_opt2sk(sp))) 370 return 0; 371 372 return 1; 373 } 374 375 /* Checking the loopback, private and other address scopes as defined in 376 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 377 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 378 * 379 * Level 0 - unusable SCTP addresses 380 * Level 1 - loopback address 381 * Level 2 - link-local addresses 382 * Level 3 - private addresses. 383 * Level 4 - global addresses 384 * For INIT and INIT-ACK address list, let L be the level of 385 * requested destination address, sender and receiver 386 * SHOULD include all of its addresses with level greater 387 * than or equal to L. 388 * 389 * IPv4 scoping can be controlled through sysctl option 390 * net.sctp.addr_scope_policy 391 */ 392 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr) 393 { 394 enum sctp_scope retval; 395 396 /* Check for unusable SCTP addresses. */ 397 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 398 retval = SCTP_SCOPE_UNUSABLE; 399 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 400 retval = SCTP_SCOPE_LOOPBACK; 401 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 402 retval = SCTP_SCOPE_LINK; 403 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 404 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 405 ipv4_is_private_192(addr->v4.sin_addr.s_addr) || 406 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) { 407 retval = SCTP_SCOPE_PRIVATE; 408 } else { 409 retval = SCTP_SCOPE_GLOBAL; 410 } 411 412 return retval; 413 } 414 415 /* Returns a valid dst cache entry for the given source and destination ip 416 * addresses. If an association is passed, trys to get a dst entry with a 417 * source address that matches an address in the bind address list. 418 */ 419 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 420 struct flowi *fl, struct sock *sk) 421 { 422 struct sctp_association *asoc = t->asoc; 423 struct rtable *rt; 424 struct flowi _fl; 425 struct flowi4 *fl4 = &_fl.u.ip4; 426 struct sctp_bind_addr *bp; 427 struct sctp_sockaddr_entry *laddr; 428 struct dst_entry *dst = NULL; 429 union sctp_addr *daddr = &t->ipaddr; 430 union sctp_addr dst_saddr; 431 dscp_t dscp; 432 433 if (t->dscp & SCTP_DSCP_SET_MASK) 434 dscp = inet_dsfield_to_dscp(t->dscp); 435 else 436 dscp = inet_sk_dscp(inet_sk(sk)); 437 438 memset(&_fl, 0x0, sizeof(_fl)); 439 fl4->daddr = daddr->v4.sin_addr.s_addr; 440 fl4->fl4_dport = daddr->v4.sin_port; 441 fl4->flowi4_proto = IPPROTO_SCTP; 442 if (asoc) { 443 fl4->flowi4_tos = inet_dscp_to_dsfield(dscp); 444 fl4->flowi4_scope = ip_sock_rt_scope(asoc->base.sk); 445 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 446 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 447 } 448 if (saddr) { 449 fl4->saddr = saddr->v4.sin_addr.s_addr; 450 if (!fl4->fl4_sport) 451 fl4->fl4_sport = saddr->v4.sin_port; 452 } 453 454 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 455 &fl4->saddr); 456 457 rt = ip_route_output_key(sock_net(sk), fl4); 458 if (!IS_ERR(rt)) { 459 dst = &rt->dst; 460 t->dst = dst; 461 memcpy(fl, &_fl, sizeof(_fl)); 462 } 463 464 /* If there is no association or if a source address is passed, no 465 * more validation is required. 466 */ 467 if (!asoc || saddr) 468 goto out; 469 470 bp = &asoc->base.bind_addr; 471 472 if (dst) { 473 /* Walk through the bind address list and look for a bind 474 * address that matches the source address of the returned dst. 475 */ 476 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 477 rcu_read_lock(); 478 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 479 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 480 (laddr->state != SCTP_ADDR_SRC && 481 !asoc->src_out_of_asoc_ok)) 482 continue; 483 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 484 goto out_unlock; 485 } 486 rcu_read_unlock(); 487 488 /* None of the bound addresses match the source address of the 489 * dst. So release it. 490 */ 491 dst_release(dst); 492 dst = NULL; 493 } 494 495 /* Walk through the bind address list and try to get a dst that 496 * matches a bind address as the source address. 497 */ 498 rcu_read_lock(); 499 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 500 struct net_device *odev; 501 502 if (!laddr->valid) 503 continue; 504 if (laddr->state != SCTP_ADDR_SRC || 505 AF_INET != laddr->a.sa.sa_family) 506 continue; 507 508 fl4->fl4_sport = laddr->a.v4.sin_port; 509 flowi4_update_output(fl4, asoc->base.sk->sk_bound_dev_if, 510 daddr->v4.sin_addr.s_addr, 511 laddr->a.v4.sin_addr.s_addr); 512 513 rt = ip_route_output_key(sock_net(sk), fl4); 514 if (IS_ERR(rt)) 515 continue; 516 517 /* Ensure the src address belongs to the output 518 * interface. 519 */ 520 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr, 521 false); 522 if (!odev || odev->ifindex != fl4->flowi4_oif) { 523 if (!dst) { 524 dst = &rt->dst; 525 t->dst = dst; 526 memcpy(fl, &_fl, sizeof(_fl)); 527 } else { 528 dst_release(&rt->dst); 529 } 530 continue; 531 } 532 533 dst_release(dst); 534 dst = &rt->dst; 535 t->dst = dst; 536 memcpy(fl, &_fl, sizeof(_fl)); 537 break; 538 } 539 540 out_unlock: 541 rcu_read_unlock(); 542 out: 543 if (dst) { 544 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 545 &fl->u.ip4.daddr, &fl->u.ip4.saddr); 546 } else { 547 t->dst = NULL; 548 pr_debug("no route\n"); 549 } 550 } 551 552 /* For v4, the source address is cached in the route entry(dst). So no need 553 * to cache it separately and hence this is an empty routine. 554 */ 555 static void sctp_v4_get_saddr(struct sctp_sock *sk, 556 struct sctp_transport *t, 557 struct flowi *fl) 558 { 559 union sctp_addr *saddr = &t->saddr; 560 struct rtable *rt = dst_rtable(t->dst); 561 562 if (rt) { 563 saddr->v4.sin_family = AF_INET; 564 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 565 } 566 } 567 568 /* What interface did this skb arrive on? */ 569 static int sctp_v4_skb_iif(const struct sk_buff *skb) 570 { 571 return inet_iif(skb); 572 } 573 574 static int sctp_v4_skb_sdif(const struct sk_buff *skb) 575 { 576 return inet_sdif(skb); 577 } 578 579 /* Was this packet marked by Explicit Congestion Notification? */ 580 static int sctp_v4_is_ce(const struct sk_buff *skb) 581 { 582 return INET_ECN_is_ce(ip_hdr(skb)->tos); 583 } 584 585 /* Create and initialize a new sk for the socket returned by accept(). */ 586 static struct sock *sctp_v4_create_accept_sk(struct sock *sk, 587 struct sctp_association *asoc, 588 bool kern) 589 { 590 struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL, 591 sk->sk_prot, kern); 592 struct inet_sock *newinet; 593 594 if (!newsk) 595 goto out; 596 597 sock_init_data(NULL, newsk); 598 599 sctp_copy_sock(newsk, sk, asoc); 600 sock_reset_flag(newsk, SOCK_ZAPPED); 601 602 sctp_v4_copy_ip_options(sk, newsk); 603 604 newinet = inet_sk(newsk); 605 606 newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr; 607 608 if (newsk->sk_prot->init(newsk)) { 609 sk_common_release(newsk); 610 newsk = NULL; 611 } 612 613 out: 614 return newsk; 615 } 616 617 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 618 { 619 /* No address mapping for V4 sockets */ 620 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 621 return sizeof(struct sockaddr_in); 622 } 623 624 /* Dump the v4 addr to the seq file. */ 625 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 626 { 627 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 628 } 629 630 static void sctp_v4_ecn_capable(struct sock *sk) 631 { 632 INET_ECN_xmit(sk); 633 } 634 635 static void sctp_addr_wq_timeout_handler(struct timer_list *t) 636 { 637 struct net *net = from_timer(net, t, sctp.addr_wq_timer); 638 struct sctp_sockaddr_entry *addrw, *temp; 639 struct sctp_sock *sp; 640 641 spin_lock_bh(&net->sctp.addr_wq_lock); 642 643 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 644 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 645 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 646 addrw->state, addrw); 647 648 #if IS_ENABLED(CONFIG_IPV6) 649 /* Now we send an ASCONF for each association */ 650 /* Note. we currently don't handle link local IPv6 addressees */ 651 if (addrw->a.sa.sa_family == AF_INET6) { 652 struct in6_addr *in6; 653 654 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 655 IPV6_ADDR_LINKLOCAL) 656 goto free_next; 657 658 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 659 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 660 addrw->state == SCTP_ADDR_NEW) { 661 unsigned long timeo_val; 662 663 pr_debug("%s: this is on DAD, trying %d sec " 664 "later\n", __func__, 665 SCTP_ADDRESS_TICK_DELAY); 666 667 timeo_val = jiffies; 668 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 669 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 670 break; 671 } 672 } 673 #endif 674 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 675 struct sock *sk; 676 677 sk = sctp_opt2sk(sp); 678 /* ignore bound-specific endpoints */ 679 if (!sctp_is_ep_boundall(sk)) 680 continue; 681 bh_lock_sock(sk); 682 if (sctp_asconf_mgmt(sp, addrw) < 0) 683 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 684 bh_unlock_sock(sk); 685 } 686 #if IS_ENABLED(CONFIG_IPV6) 687 free_next: 688 #endif 689 list_del(&addrw->list); 690 kfree(addrw); 691 } 692 spin_unlock_bh(&net->sctp.addr_wq_lock); 693 } 694 695 static void sctp_free_addr_wq(struct net *net) 696 { 697 struct sctp_sockaddr_entry *addrw; 698 struct sctp_sockaddr_entry *temp; 699 700 spin_lock_bh(&net->sctp.addr_wq_lock); 701 del_timer(&net->sctp.addr_wq_timer); 702 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 703 list_del(&addrw->list); 704 kfree(addrw); 705 } 706 spin_unlock_bh(&net->sctp.addr_wq_lock); 707 } 708 709 /* lookup the entry for the same address in the addr_waitq 710 * sctp_addr_wq MUST be locked 711 */ 712 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 713 struct sctp_sockaddr_entry *addr) 714 { 715 struct sctp_sockaddr_entry *addrw; 716 717 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 718 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 719 continue; 720 if (addrw->a.sa.sa_family == AF_INET) { 721 if (addrw->a.v4.sin_addr.s_addr == 722 addr->a.v4.sin_addr.s_addr) 723 return addrw; 724 } else if (addrw->a.sa.sa_family == AF_INET6) { 725 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 726 &addr->a.v6.sin6_addr)) 727 return addrw; 728 } 729 } 730 return NULL; 731 } 732 733 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 734 { 735 struct sctp_sockaddr_entry *addrw; 736 unsigned long timeo_val; 737 738 /* first, we check if an opposite message already exist in the queue. 739 * If we found such message, it is removed. 740 * This operation is a bit stupid, but the DHCP client attaches the 741 * new address after a couple of addition and deletion of that address 742 */ 743 744 spin_lock_bh(&net->sctp.addr_wq_lock); 745 746 /* Avoid searching the queue or modifying it if there are no consumers, 747 * as it can lead to performance degradation if addresses are modified 748 * en-masse. 749 * 750 * If the queue already contains some events, update it anyway to avoid 751 * ugly races between new sessions and new address events. 752 */ 753 if (list_empty(&net->sctp.auto_asconf_splist) && 754 list_empty(&net->sctp.addr_waitq)) { 755 spin_unlock_bh(&net->sctp.addr_wq_lock); 756 return; 757 } 758 759 /* Offsets existing events in addr_wq */ 760 addrw = sctp_addr_wq_lookup(net, addr); 761 if (addrw) { 762 if (addrw->state != cmd) { 763 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 764 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 765 &net->sctp.addr_waitq); 766 767 list_del(&addrw->list); 768 kfree(addrw); 769 } 770 spin_unlock_bh(&net->sctp.addr_wq_lock); 771 return; 772 } 773 774 /* OK, we have to add the new address to the wait queue */ 775 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 776 if (addrw == NULL) { 777 spin_unlock_bh(&net->sctp.addr_wq_lock); 778 return; 779 } 780 addrw->state = cmd; 781 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 782 783 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 784 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 785 786 if (!timer_pending(&net->sctp.addr_wq_timer)) { 787 timeo_val = jiffies; 788 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 789 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 790 } 791 spin_unlock_bh(&net->sctp.addr_wq_lock); 792 } 793 794 /* Event handler for inet address addition/deletion events. 795 * The sctp_local_addr_list needs to be protocted by a spin lock since 796 * multiple notifiers (say IPv4 and IPv6) may be running at the same 797 * time and thus corrupt the list. 798 * The reader side is protected with RCU. 799 */ 800 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 801 void *ptr) 802 { 803 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 804 struct sctp_sockaddr_entry *addr = NULL; 805 struct sctp_sockaddr_entry *temp; 806 struct net *net = dev_net(ifa->ifa_dev->dev); 807 int found = 0; 808 809 switch (ev) { 810 case NETDEV_UP: 811 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 812 if (addr) { 813 addr->a.v4.sin_family = AF_INET; 814 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 815 addr->valid = 1; 816 spin_lock_bh(&net->sctp.local_addr_lock); 817 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 818 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 819 spin_unlock_bh(&net->sctp.local_addr_lock); 820 } 821 break; 822 case NETDEV_DOWN: 823 spin_lock_bh(&net->sctp.local_addr_lock); 824 list_for_each_entry_safe(addr, temp, 825 &net->sctp.local_addr_list, list) { 826 if (addr->a.sa.sa_family == AF_INET && 827 addr->a.v4.sin_addr.s_addr == 828 ifa->ifa_local) { 829 found = 1; 830 addr->valid = 0; 831 list_del_rcu(&addr->list); 832 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 833 break; 834 } 835 } 836 spin_unlock_bh(&net->sctp.local_addr_lock); 837 if (found) 838 kfree_rcu(addr, rcu); 839 break; 840 } 841 842 return NOTIFY_DONE; 843 } 844 845 /* 846 * Initialize the control inode/socket with a control endpoint data 847 * structure. This endpoint is reserved exclusively for the OOTB processing. 848 */ 849 static int sctp_ctl_sock_init(struct net *net) 850 { 851 int err; 852 sa_family_t family = PF_INET; 853 854 if (sctp_get_pf_specific(PF_INET6)) 855 family = PF_INET6; 856 857 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 858 SOCK_SEQPACKET, IPPROTO_SCTP, net); 859 860 /* If IPv6 socket could not be created, try the IPv4 socket */ 861 if (err < 0 && family == PF_INET6) 862 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 863 SOCK_SEQPACKET, IPPROTO_SCTP, 864 net); 865 866 if (err < 0) { 867 pr_err("Failed to create the SCTP control socket\n"); 868 return err; 869 } 870 return 0; 871 } 872 873 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb) 874 { 875 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source; 876 877 skb_set_transport_header(skb, sizeof(struct udphdr)); 878 sctp_rcv(skb); 879 return 0; 880 } 881 882 int sctp_udp_sock_start(struct net *net) 883 { 884 struct udp_tunnel_sock_cfg tuncfg = {NULL}; 885 struct udp_port_cfg udp_conf = {0}; 886 struct socket *sock; 887 int err; 888 889 udp_conf.family = AF_INET; 890 udp_conf.local_ip.s_addr = htonl(INADDR_ANY); 891 udp_conf.local_udp_port = htons(net->sctp.udp_port); 892 err = udp_sock_create(net, &udp_conf, &sock); 893 if (err) { 894 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n"); 895 return err; 896 } 897 898 tuncfg.encap_type = 1; 899 tuncfg.encap_rcv = sctp_udp_rcv; 900 tuncfg.encap_err_lookup = sctp_udp_v4_err; 901 setup_udp_tunnel_sock(net, sock, &tuncfg); 902 net->sctp.udp4_sock = sock->sk; 903 904 #if IS_ENABLED(CONFIG_IPV6) 905 memset(&udp_conf, 0, sizeof(udp_conf)); 906 907 udp_conf.family = AF_INET6; 908 udp_conf.local_ip6 = in6addr_any; 909 udp_conf.local_udp_port = htons(net->sctp.udp_port); 910 udp_conf.use_udp6_rx_checksums = true; 911 udp_conf.ipv6_v6only = true; 912 err = udp_sock_create(net, &udp_conf, &sock); 913 if (err) { 914 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n"); 915 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 916 net->sctp.udp4_sock = NULL; 917 return err; 918 } 919 920 tuncfg.encap_type = 1; 921 tuncfg.encap_rcv = sctp_udp_rcv; 922 tuncfg.encap_err_lookup = sctp_udp_v6_err; 923 setup_udp_tunnel_sock(net, sock, &tuncfg); 924 net->sctp.udp6_sock = sock->sk; 925 #endif 926 927 return 0; 928 } 929 930 void sctp_udp_sock_stop(struct net *net) 931 { 932 if (net->sctp.udp4_sock) { 933 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 934 net->sctp.udp4_sock = NULL; 935 } 936 if (net->sctp.udp6_sock) { 937 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket); 938 net->sctp.udp6_sock = NULL; 939 } 940 } 941 942 /* Register address family specific functions. */ 943 int sctp_register_af(struct sctp_af *af) 944 { 945 switch (af->sa_family) { 946 case AF_INET: 947 if (sctp_af_v4_specific) 948 return 0; 949 sctp_af_v4_specific = af; 950 break; 951 case AF_INET6: 952 if (sctp_af_v6_specific) 953 return 0; 954 sctp_af_v6_specific = af; 955 break; 956 default: 957 return 0; 958 } 959 960 INIT_LIST_HEAD(&af->list); 961 list_add_tail(&af->list, &sctp_address_families); 962 return 1; 963 } 964 965 /* Get the table of functions for manipulating a particular address 966 * family. 967 */ 968 struct sctp_af *sctp_get_af_specific(sa_family_t family) 969 { 970 switch (family) { 971 case AF_INET: 972 return sctp_af_v4_specific; 973 case AF_INET6: 974 return sctp_af_v6_specific; 975 default: 976 return NULL; 977 } 978 } 979 980 /* Common code to initialize a AF_INET msg_name. */ 981 static void sctp_inet_msgname(char *msgname, int *addr_len) 982 { 983 struct sockaddr_in *sin; 984 985 sin = (struct sockaddr_in *)msgname; 986 *addr_len = sizeof(struct sockaddr_in); 987 sin->sin_family = AF_INET; 988 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 989 } 990 991 /* Copy the primary address of the peer primary address as the msg_name. */ 992 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 993 int *addr_len) 994 { 995 struct sockaddr_in *sin, *sinfrom; 996 997 if (msgname) { 998 struct sctp_association *asoc; 999 1000 asoc = event->asoc; 1001 sctp_inet_msgname(msgname, addr_len); 1002 sin = (struct sockaddr_in *)msgname; 1003 sinfrom = &asoc->peer.primary_addr.v4; 1004 sin->sin_port = htons(asoc->peer.port); 1005 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 1006 } 1007 } 1008 1009 /* Initialize and copy out a msgname from an inbound skb. */ 1010 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 1011 { 1012 if (msgname) { 1013 struct sctphdr *sh = sctp_hdr(skb); 1014 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 1015 1016 sctp_inet_msgname(msgname, len); 1017 sin->sin_port = sh->source; 1018 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 1019 } 1020 } 1021 1022 /* Do we support this AF? */ 1023 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 1024 { 1025 /* PF_INET only supports AF_INET addresses. */ 1026 return AF_INET == family; 1027 } 1028 1029 /* Address matching with wildcards allowed. */ 1030 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 1031 const union sctp_addr *addr2, 1032 struct sctp_sock *opt) 1033 { 1034 /* PF_INET only supports AF_INET addresses. */ 1035 if (addr1->sa.sa_family != addr2->sa.sa_family) 1036 return 0; 1037 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 1038 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 1039 return 1; 1040 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 1041 return 1; 1042 1043 return 0; 1044 } 1045 1046 /* Verify that provided sockaddr looks bindable. Common verification has 1047 * already been taken care of. 1048 */ 1049 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 1050 { 1051 return sctp_v4_available(addr, opt); 1052 } 1053 1054 /* Verify that sockaddr looks sendable. Common verification has already 1055 * been taken care of. 1056 */ 1057 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 1058 { 1059 return 1; 1060 } 1061 1062 /* Fill in Supported Address Type information for INIT and INIT-ACK 1063 * chunks. Returns number of addresses supported. 1064 */ 1065 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 1066 __be16 *types) 1067 { 1068 types[0] = SCTP_PARAM_IPV4_ADDRESS; 1069 return 1; 1070 } 1071 1072 /* Wrapper routine that calls the ip transmit routine. */ 1073 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t) 1074 { 1075 struct dst_entry *dst = dst_clone(t->dst); 1076 struct flowi4 *fl4 = &t->fl.u.ip4; 1077 struct sock *sk = skb->sk; 1078 struct inet_sock *inet = inet_sk(sk); 1079 __u8 dscp = READ_ONCE(inet->tos); 1080 __be16 df = 0; 1081 1082 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 1083 skb->len, &fl4->saddr, &fl4->daddr); 1084 1085 if (t->dscp & SCTP_DSCP_SET_MASK) 1086 dscp = t->dscp & SCTP_DSCP_VAL_MASK; 1087 1088 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO 1089 : IP_PMTUDISC_DONT; 1090 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS); 1091 1092 if (!t->encap_port || !sctp_sk(sk)->udp_port) { 1093 skb_dst_set(skb, dst); 1094 return __ip_queue_xmit(sk, skb, &t->fl, dscp); 1095 } 1096 1097 if (skb_is_gso(skb)) 1098 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM; 1099 1100 if (ip_dont_fragment(sk, dst) && !skb->ignore_df) 1101 df = htons(IP_DF); 1102 1103 skb->encapsulation = 1; 1104 skb_reset_inner_mac_header(skb); 1105 skb_reset_inner_transport_header(skb); 1106 skb_set_inner_ipproto(skb, IPPROTO_SCTP); 1107 udp_tunnel_xmit_skb(dst_rtable(dst), sk, skb, fl4->saddr, 1108 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df, 1109 sctp_sk(sk)->udp_port, t->encap_port, false, false); 1110 return 0; 1111 } 1112 1113 static struct sctp_af sctp_af_inet; 1114 1115 static struct sctp_pf sctp_pf_inet = { 1116 .event_msgname = sctp_inet_event_msgname, 1117 .skb_msgname = sctp_inet_skb_msgname, 1118 .af_supported = sctp_inet_af_supported, 1119 .cmp_addr = sctp_inet_cmp_addr, 1120 .bind_verify = sctp_inet_bind_verify, 1121 .send_verify = sctp_inet_send_verify, 1122 .supported_addrs = sctp_inet_supported_addrs, 1123 .create_accept_sk = sctp_v4_create_accept_sk, 1124 .addr_to_user = sctp_v4_addr_to_user, 1125 .to_sk_saddr = sctp_v4_to_sk_saddr, 1126 .to_sk_daddr = sctp_v4_to_sk_daddr, 1127 .copy_ip_options = sctp_v4_copy_ip_options, 1128 .af = &sctp_af_inet 1129 }; 1130 1131 /* Notifier for inetaddr addition/deletion events. */ 1132 static struct notifier_block sctp_inetaddr_notifier = { 1133 .notifier_call = sctp_inetaddr_event, 1134 }; 1135 1136 /* Socket operations. */ 1137 static const struct proto_ops inet_seqpacket_ops = { 1138 .family = PF_INET, 1139 .owner = THIS_MODULE, 1140 .release = inet_release, /* Needs to be wrapped... */ 1141 .bind = inet_bind, 1142 .connect = sctp_inet_connect, 1143 .socketpair = sock_no_socketpair, 1144 .accept = inet_accept, 1145 .getname = inet_getname, /* Semantics are different. */ 1146 .poll = sctp_poll, 1147 .ioctl = inet_ioctl, 1148 .gettstamp = sock_gettstamp, 1149 .listen = sctp_inet_listen, 1150 .shutdown = inet_shutdown, /* Looks harmless. */ 1151 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1152 .getsockopt = sock_common_getsockopt, 1153 .sendmsg = inet_sendmsg, 1154 .recvmsg = inet_recvmsg, 1155 .mmap = sock_no_mmap, 1156 }; 1157 1158 /* Registration with AF_INET family. */ 1159 static struct inet_protosw sctp_seqpacket_protosw = { 1160 .type = SOCK_SEQPACKET, 1161 .protocol = IPPROTO_SCTP, 1162 .prot = &sctp_prot, 1163 .ops = &inet_seqpacket_ops, 1164 .flags = SCTP_PROTOSW_FLAG 1165 }; 1166 static struct inet_protosw sctp_stream_protosw = { 1167 .type = SOCK_STREAM, 1168 .protocol = IPPROTO_SCTP, 1169 .prot = &sctp_prot, 1170 .ops = &inet_seqpacket_ops, 1171 .flags = SCTP_PROTOSW_FLAG 1172 }; 1173 1174 static int sctp4_rcv(struct sk_buff *skb) 1175 { 1176 SCTP_INPUT_CB(skb)->encap_port = 0; 1177 return sctp_rcv(skb); 1178 } 1179 1180 /* Register with IP layer. */ 1181 static const struct net_protocol sctp_protocol = { 1182 .handler = sctp4_rcv, 1183 .err_handler = sctp_v4_err, 1184 .no_policy = 1, 1185 .icmp_strict_tag_validation = 1, 1186 }; 1187 1188 /* IPv4 address related functions. */ 1189 static struct sctp_af sctp_af_inet = { 1190 .sa_family = AF_INET, 1191 .sctp_xmit = sctp_v4_xmit, 1192 .setsockopt = ip_setsockopt, 1193 .getsockopt = ip_getsockopt, 1194 .get_dst = sctp_v4_get_dst, 1195 .get_saddr = sctp_v4_get_saddr, 1196 .copy_addrlist = sctp_v4_copy_addrlist, 1197 .from_skb = sctp_v4_from_skb, 1198 .from_sk = sctp_v4_from_sk, 1199 .from_addr_param = sctp_v4_from_addr_param, 1200 .to_addr_param = sctp_v4_to_addr_param, 1201 .cmp_addr = sctp_v4_cmp_addr, 1202 .addr_valid = sctp_v4_addr_valid, 1203 .inaddr_any = sctp_v4_inaddr_any, 1204 .is_any = sctp_v4_is_any, 1205 .available = sctp_v4_available, 1206 .scope = sctp_v4_scope, 1207 .skb_iif = sctp_v4_skb_iif, 1208 .skb_sdif = sctp_v4_skb_sdif, 1209 .is_ce = sctp_v4_is_ce, 1210 .seq_dump_addr = sctp_v4_seq_dump_addr, 1211 .ecn_capable = sctp_v4_ecn_capable, 1212 .net_header_len = sizeof(struct iphdr), 1213 .sockaddr_len = sizeof(struct sockaddr_in), 1214 .ip_options_len = sctp_v4_ip_options_len, 1215 }; 1216 1217 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1218 { 1219 switch (family) { 1220 case PF_INET: 1221 return sctp_pf_inet_specific; 1222 case PF_INET6: 1223 return sctp_pf_inet6_specific; 1224 default: 1225 return NULL; 1226 } 1227 } 1228 1229 /* Register the PF specific function table. */ 1230 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1231 { 1232 switch (family) { 1233 case PF_INET: 1234 if (sctp_pf_inet_specific) 1235 return 0; 1236 sctp_pf_inet_specific = pf; 1237 break; 1238 case PF_INET6: 1239 if (sctp_pf_inet6_specific) 1240 return 0; 1241 sctp_pf_inet6_specific = pf; 1242 break; 1243 default: 1244 return 0; 1245 } 1246 return 1; 1247 } 1248 1249 static inline int init_sctp_mibs(struct net *net) 1250 { 1251 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1252 if (!net->sctp.sctp_statistics) 1253 return -ENOMEM; 1254 return 0; 1255 } 1256 1257 static inline void cleanup_sctp_mibs(struct net *net) 1258 { 1259 free_percpu(net->sctp.sctp_statistics); 1260 } 1261 1262 static void sctp_v4_pf_init(void) 1263 { 1264 /* Initialize the SCTP specific PF functions. */ 1265 sctp_register_pf(&sctp_pf_inet, PF_INET); 1266 sctp_register_af(&sctp_af_inet); 1267 } 1268 1269 static void sctp_v4_pf_exit(void) 1270 { 1271 list_del(&sctp_af_inet.list); 1272 } 1273 1274 static int sctp_v4_protosw_init(void) 1275 { 1276 int rc; 1277 1278 rc = proto_register(&sctp_prot, 1); 1279 if (rc) 1280 return rc; 1281 1282 /* Register SCTP(UDP and TCP style) with socket layer. */ 1283 inet_register_protosw(&sctp_seqpacket_protosw); 1284 inet_register_protosw(&sctp_stream_protosw); 1285 1286 return 0; 1287 } 1288 1289 static void sctp_v4_protosw_exit(void) 1290 { 1291 inet_unregister_protosw(&sctp_stream_protosw); 1292 inet_unregister_protosw(&sctp_seqpacket_protosw); 1293 proto_unregister(&sctp_prot); 1294 } 1295 1296 static int sctp_v4_add_protocol(void) 1297 { 1298 /* Register notifier for inet address additions/deletions. */ 1299 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1300 1301 /* Register SCTP with inet layer. */ 1302 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1303 return -EAGAIN; 1304 1305 return 0; 1306 } 1307 1308 static void sctp_v4_del_protocol(void) 1309 { 1310 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1311 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1312 } 1313 1314 static int __net_init sctp_defaults_init(struct net *net) 1315 { 1316 int status; 1317 1318 /* 1319 * 14. Suggested SCTP Protocol Parameter Values 1320 */ 1321 /* The following protocol parameters are RECOMMENDED: */ 1322 /* RTO.Initial - 3 seconds */ 1323 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1324 /* RTO.Min - 1 second */ 1325 net->sctp.rto_min = SCTP_RTO_MIN; 1326 /* RTO.Max - 60 seconds */ 1327 net->sctp.rto_max = SCTP_RTO_MAX; 1328 /* RTO.Alpha - 1/8 */ 1329 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1330 /* RTO.Beta - 1/4 */ 1331 net->sctp.rto_beta = SCTP_RTO_BETA; 1332 1333 /* Valid.Cookie.Life - 60 seconds */ 1334 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1335 1336 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1337 net->sctp.cookie_preserve_enable = 1; 1338 1339 /* Default sctp sockets to use md5 as their hmac alg */ 1340 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5) 1341 net->sctp.sctp_hmac_alg = "md5"; 1342 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1) 1343 net->sctp.sctp_hmac_alg = "sha1"; 1344 #else 1345 net->sctp.sctp_hmac_alg = NULL; 1346 #endif 1347 1348 /* Max.Burst - 4 */ 1349 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1350 1351 /* Disable of Primary Path Switchover by default */ 1352 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX; 1353 1354 /* Enable pf state by default */ 1355 net->sctp.pf_enable = 1; 1356 1357 /* Ignore pf exposure feature by default */ 1358 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET; 1359 1360 /* Association.Max.Retrans - 10 attempts 1361 * Path.Max.Retrans - 5 attempts (per destination address) 1362 * Max.Init.Retransmits - 8 attempts 1363 */ 1364 net->sctp.max_retrans_association = 10; 1365 net->sctp.max_retrans_path = 5; 1366 net->sctp.max_retrans_init = 8; 1367 1368 /* Sendbuffer growth - do per-socket accounting */ 1369 net->sctp.sndbuf_policy = 0; 1370 1371 /* Rcvbuffer growth - do per-socket accounting */ 1372 net->sctp.rcvbuf_policy = 0; 1373 1374 /* HB.interval - 30 seconds */ 1375 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1376 1377 /* delayed SACK timeout */ 1378 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1379 1380 /* Disable ADDIP by default. */ 1381 net->sctp.addip_enable = 0; 1382 net->sctp.addip_noauth = 0; 1383 net->sctp.default_auto_asconf = 0; 1384 1385 /* Enable PR-SCTP by default. */ 1386 net->sctp.prsctp_enable = 1; 1387 1388 /* Disable RECONF by default. */ 1389 net->sctp.reconf_enable = 0; 1390 1391 /* Disable AUTH by default. */ 1392 net->sctp.auth_enable = 0; 1393 1394 /* Enable ECN by default. */ 1395 net->sctp.ecn_enable = 1; 1396 1397 /* Set UDP tunneling listening port to 0 by default */ 1398 net->sctp.udp_port = 0; 1399 1400 /* Set remote encap port to 0 by default */ 1401 net->sctp.encap_port = 0; 1402 1403 /* Set SCOPE policy to enabled */ 1404 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1405 1406 /* Set the default rwnd update threshold */ 1407 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1408 1409 /* Initialize maximum autoclose timeout. */ 1410 net->sctp.max_autoclose = INT_MAX / HZ; 1411 1412 #ifdef CONFIG_NET_L3_MASTER_DEV 1413 net->sctp.l3mdev_accept = 1; 1414 #endif 1415 1416 status = sctp_sysctl_net_register(net); 1417 if (status) 1418 goto err_sysctl_register; 1419 1420 /* Allocate and initialise sctp mibs. */ 1421 status = init_sctp_mibs(net); 1422 if (status) 1423 goto err_init_mibs; 1424 1425 #ifdef CONFIG_PROC_FS 1426 /* Initialize proc fs directory. */ 1427 status = sctp_proc_init(net); 1428 if (status) 1429 goto err_init_proc; 1430 #endif 1431 1432 sctp_dbg_objcnt_init(net); 1433 1434 /* Initialize the local address list. */ 1435 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1436 spin_lock_init(&net->sctp.local_addr_lock); 1437 sctp_get_local_addr_list(net); 1438 1439 /* Initialize the address event list */ 1440 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1441 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1442 spin_lock_init(&net->sctp.addr_wq_lock); 1443 net->sctp.addr_wq_timer.expires = 0; 1444 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1445 1446 return 0; 1447 1448 #ifdef CONFIG_PROC_FS 1449 err_init_proc: 1450 cleanup_sctp_mibs(net); 1451 #endif 1452 err_init_mibs: 1453 sctp_sysctl_net_unregister(net); 1454 err_sysctl_register: 1455 return status; 1456 } 1457 1458 static void __net_exit sctp_defaults_exit(struct net *net) 1459 { 1460 /* Free the local address list */ 1461 sctp_free_addr_wq(net); 1462 sctp_free_local_addr_list(net); 1463 1464 #ifdef CONFIG_PROC_FS 1465 remove_proc_subtree("sctp", net->proc_net); 1466 net->sctp.proc_net_sctp = NULL; 1467 #endif 1468 cleanup_sctp_mibs(net); 1469 sctp_sysctl_net_unregister(net); 1470 } 1471 1472 static struct pernet_operations sctp_defaults_ops = { 1473 .init = sctp_defaults_init, 1474 .exit = sctp_defaults_exit, 1475 }; 1476 1477 static int __net_init sctp_ctrlsock_init(struct net *net) 1478 { 1479 int status; 1480 1481 /* Initialize the control inode/socket for handling OOTB packets. */ 1482 status = sctp_ctl_sock_init(net); 1483 if (status) 1484 pr_err("Failed to initialize the SCTP control sock\n"); 1485 1486 return status; 1487 } 1488 1489 static void __net_exit sctp_ctrlsock_exit(struct net *net) 1490 { 1491 /* Free the control endpoint. */ 1492 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1493 } 1494 1495 static struct pernet_operations sctp_ctrlsock_ops = { 1496 .init = sctp_ctrlsock_init, 1497 .exit = sctp_ctrlsock_exit, 1498 }; 1499 1500 /* Initialize the universe into something sensible. */ 1501 static __init int sctp_init(void) 1502 { 1503 unsigned long nr_pages = totalram_pages(); 1504 unsigned long limit; 1505 unsigned long goal; 1506 int max_entry_order; 1507 int num_entries; 1508 int max_share; 1509 int status; 1510 int order; 1511 int i; 1512 1513 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1514 1515 /* Allocate bind_bucket and chunk caches. */ 1516 status = -ENOBUFS; 1517 sctp_bucket_cachep = KMEM_CACHE(sctp_bind_bucket, SLAB_HWCACHE_ALIGN); 1518 if (!sctp_bucket_cachep) 1519 goto out; 1520 1521 sctp_chunk_cachep = KMEM_CACHE(sctp_chunk, SLAB_HWCACHE_ALIGN); 1522 if (!sctp_chunk_cachep) 1523 goto err_chunk_cachep; 1524 1525 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1526 if (status) 1527 goto err_percpu_counter_init; 1528 1529 /* Implementation specific variables. */ 1530 1531 /* Initialize default stream count setup information. */ 1532 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1533 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1534 1535 /* Initialize handle used for association ids. */ 1536 idr_init(&sctp_assocs_id); 1537 1538 limit = nr_free_buffer_pages() / 8; 1539 limit = max(limit, 128UL); 1540 sysctl_sctp_mem[0] = limit / 4 * 3; 1541 sysctl_sctp_mem[1] = limit; 1542 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1543 1544 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1545 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1546 max_share = min(4UL*1024*1024, limit); 1547 1548 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */ 1549 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1550 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1551 1552 sysctl_sctp_wmem[0] = PAGE_SIZE; 1553 sysctl_sctp_wmem[1] = 16*1024; 1554 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1555 1556 /* Size and allocate the association hash table. 1557 * The methodology is similar to that of the tcp hash tables. 1558 * Though not identical. Start by getting a goal size 1559 */ 1560 if (nr_pages >= (128 * 1024)) 1561 goal = nr_pages >> (22 - PAGE_SHIFT); 1562 else 1563 goal = nr_pages >> (24 - PAGE_SHIFT); 1564 1565 /* Then compute the page order for said goal */ 1566 order = get_order(goal); 1567 1568 /* Now compute the required page order for the maximum sized table we 1569 * want to create 1570 */ 1571 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1572 sizeof(struct sctp_bind_hashbucket)); 1573 1574 /* Limit the page order by that maximum hash table size */ 1575 order = min(order, max_entry_order); 1576 1577 /* Allocate and initialize the endpoint hash table. */ 1578 sctp_ep_hashsize = 64; 1579 sctp_ep_hashtable = 1580 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1581 if (!sctp_ep_hashtable) { 1582 pr_err("Failed endpoint_hash alloc\n"); 1583 status = -ENOMEM; 1584 goto err_ehash_alloc; 1585 } 1586 for (i = 0; i < sctp_ep_hashsize; i++) { 1587 rwlock_init(&sctp_ep_hashtable[i].lock); 1588 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1589 } 1590 1591 /* Allocate and initialize the SCTP port hash table. 1592 * Note that order is initalized to start at the max sized 1593 * table we want to support. If we can't get that many pages 1594 * reduce the order and try again 1595 */ 1596 do { 1597 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1598 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1599 } while (!sctp_port_hashtable && --order > 0); 1600 1601 if (!sctp_port_hashtable) { 1602 pr_err("Failed bind hash alloc\n"); 1603 status = -ENOMEM; 1604 goto err_bhash_alloc; 1605 } 1606 1607 /* Now compute the number of entries that will fit in the 1608 * port hash space we allocated 1609 */ 1610 num_entries = (1UL << order) * PAGE_SIZE / 1611 sizeof(struct sctp_bind_hashbucket); 1612 1613 /* And finish by rounding it down to the nearest power of two. 1614 * This wastes some memory of course, but it's needed because 1615 * the hash function operates based on the assumption that 1616 * the number of entries is a power of two. 1617 */ 1618 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1619 1620 for (i = 0; i < sctp_port_hashsize; i++) { 1621 spin_lock_init(&sctp_port_hashtable[i].lock); 1622 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1623 } 1624 1625 status = sctp_transport_hashtable_init(); 1626 if (status) 1627 goto err_thash_alloc; 1628 1629 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1630 num_entries); 1631 1632 sctp_sysctl_register(); 1633 1634 INIT_LIST_HEAD(&sctp_address_families); 1635 sctp_v4_pf_init(); 1636 sctp_v6_pf_init(); 1637 sctp_sched_ops_init(); 1638 1639 status = register_pernet_subsys(&sctp_defaults_ops); 1640 if (status) 1641 goto err_register_defaults; 1642 1643 status = sctp_v4_protosw_init(); 1644 if (status) 1645 goto err_protosw_init; 1646 1647 status = sctp_v6_protosw_init(); 1648 if (status) 1649 goto err_v6_protosw_init; 1650 1651 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1652 if (status) 1653 goto err_register_ctrlsock; 1654 1655 status = sctp_v4_add_protocol(); 1656 if (status) 1657 goto err_add_protocol; 1658 1659 /* Register SCTP with inet6 layer. */ 1660 status = sctp_v6_add_protocol(); 1661 if (status) 1662 goto err_v6_add_protocol; 1663 1664 if (sctp_offload_init() < 0) 1665 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1666 1667 out: 1668 return status; 1669 err_v6_add_protocol: 1670 sctp_v4_del_protocol(); 1671 err_add_protocol: 1672 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1673 err_register_ctrlsock: 1674 sctp_v6_protosw_exit(); 1675 err_v6_protosw_init: 1676 sctp_v4_protosw_exit(); 1677 err_protosw_init: 1678 unregister_pernet_subsys(&sctp_defaults_ops); 1679 err_register_defaults: 1680 sctp_v4_pf_exit(); 1681 sctp_v6_pf_exit(); 1682 sctp_sysctl_unregister(); 1683 free_pages((unsigned long)sctp_port_hashtable, 1684 get_order(sctp_port_hashsize * 1685 sizeof(struct sctp_bind_hashbucket))); 1686 err_bhash_alloc: 1687 sctp_transport_hashtable_destroy(); 1688 err_thash_alloc: 1689 kfree(sctp_ep_hashtable); 1690 err_ehash_alloc: 1691 percpu_counter_destroy(&sctp_sockets_allocated); 1692 err_percpu_counter_init: 1693 kmem_cache_destroy(sctp_chunk_cachep); 1694 err_chunk_cachep: 1695 kmem_cache_destroy(sctp_bucket_cachep); 1696 goto out; 1697 } 1698 1699 /* Exit handler for the SCTP protocol. */ 1700 static __exit void sctp_exit(void) 1701 { 1702 /* BUG. This should probably do something useful like clean 1703 * up all the remaining associations and all that memory. 1704 */ 1705 1706 /* Unregister with inet6/inet layers. */ 1707 sctp_v6_del_protocol(); 1708 sctp_v4_del_protocol(); 1709 1710 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1711 1712 /* Free protosw registrations */ 1713 sctp_v6_protosw_exit(); 1714 sctp_v4_protosw_exit(); 1715 1716 unregister_pernet_subsys(&sctp_defaults_ops); 1717 1718 /* Unregister with socket layer. */ 1719 sctp_v6_pf_exit(); 1720 sctp_v4_pf_exit(); 1721 1722 sctp_sysctl_unregister(); 1723 1724 free_pages((unsigned long)sctp_port_hashtable, 1725 get_order(sctp_port_hashsize * 1726 sizeof(struct sctp_bind_hashbucket))); 1727 kfree(sctp_ep_hashtable); 1728 sctp_transport_hashtable_destroy(); 1729 1730 percpu_counter_destroy(&sctp_sockets_allocated); 1731 1732 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1733 1734 kmem_cache_destroy(sctp_chunk_cachep); 1735 kmem_cache_destroy(sctp_bucket_cachep); 1736 } 1737 1738 module_init(sctp_init); 1739 module_exit(sctp_exit); 1740 1741 /* 1742 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1743 */ 1744 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1745 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1746 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1747 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1748 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1749 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1750 MODULE_LICENSE("GPL"); 1751