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