1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* SCTP kernel implementation 3 * (C) Copyright IBM Corp. 2001, 2004 4 * Copyright (c) 1999-2000 Cisco, Inc. 5 * Copyright (c) 1999-2001 Motorola, Inc. 6 * Copyright (c) 2001 Intel Corp. 7 * Copyright (c) 2001 Nokia, Inc. 8 * Copyright (c) 2001 La Monte H.P. Yarroll 9 * 10 * This file is part of the SCTP kernel implementation 11 * 12 * Initialization/cleanup for SCTP protocol support. 13 * 14 * Please send any bug reports or fixes you make to the 15 * email address(es): 16 * lksctp developers <linux-sctp@vger.kernel.org> 17 * 18 * Written or modified by: 19 * La Monte H.P. Yarroll <piggy@acm.org> 20 * Karl Knutson <karl@athena.chicago.il.us> 21 * Jon Grimm <jgrimm@us.ibm.com> 22 * Sridhar Samudrala <sri@us.ibm.com> 23 * Daisy Chang <daisyc@us.ibm.com> 24 * Ardelle Fan <ardelle.fan@intel.com> 25 */ 26 27 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 28 29 #include <linux/module.h> 30 #include <linux/init.h> 31 #include <linux/netdevice.h> 32 #include <linux/inetdevice.h> 33 #include <linux/seq_file.h> 34 #include <linux/memblock.h> 35 #include <linux/highmem.h> 36 #include <linux/slab.h> 37 #include <net/flow.h> 38 #include <net/net_namespace.h> 39 #include <net/protocol.h> 40 #include <net/ip.h> 41 #include <net/ipv6.h> 42 #include <net/route.h> 43 #include <net/sctp/sctp.h> 44 #include <net/addrconf.h> 45 #include <net/inet_common.h> 46 #include <net/inet_ecn.h> 47 #include <net/inet_sock.h> 48 #include <net/udp_tunnel.h> 49 #include <net/inet_dscp.h> 50 51 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024) 52 53 /* Global data structures. */ 54 struct sctp_globals sctp_globals __read_mostly; 55 56 struct idr sctp_assocs_id; 57 DEFINE_SPINLOCK(sctp_assocs_id_lock); 58 59 static struct sctp_pf *sctp_pf_inet6_specific; 60 static struct sctp_pf *sctp_pf_inet_specific; 61 static struct sctp_af *sctp_af_v4_specific; 62 static struct sctp_af *sctp_af_v6_specific; 63 64 struct kmem_cache *sctp_chunk_cachep __read_mostly; 65 struct kmem_cache *sctp_bucket_cachep __read_mostly; 66 67 long sysctl_sctp_mem[3]; 68 int sysctl_sctp_rmem[3]; 69 int sysctl_sctp_wmem[3]; 70 71 /* Private helper to extract ipv4 address and stash them in 72 * the protocol structure. 73 */ 74 static void sctp_v4_copy_addrlist(struct list_head *addrlist, 75 struct net_device *dev) 76 { 77 struct in_device *in_dev; 78 struct in_ifaddr *ifa; 79 struct sctp_sockaddr_entry *addr; 80 81 rcu_read_lock(); 82 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) { 83 rcu_read_unlock(); 84 return; 85 } 86 87 in_dev_for_each_ifa_rcu(ifa, in_dev) { 88 /* Add the address to the local list. */ 89 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 90 if (addr) { 91 addr->a.v4.sin_family = AF_INET; 92 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 93 addr->valid = 1; 94 INIT_LIST_HEAD(&addr->list); 95 list_add_tail(&addr->list, addrlist); 96 } 97 } 98 99 rcu_read_unlock(); 100 } 101 102 /* Extract our IP addresses from the system and stash them in the 103 * protocol structure. 104 */ 105 static void sctp_get_local_addr_list(struct net *net) 106 { 107 struct net_device *dev; 108 struct list_head *pos; 109 struct sctp_af *af; 110 111 rcu_read_lock(); 112 for_each_netdev_rcu(net, dev) { 113 list_for_each(pos, &sctp_address_families) { 114 af = list_entry(pos, struct sctp_af, list); 115 af->copy_addrlist(&net->sctp.local_addr_list, dev); 116 } 117 } 118 rcu_read_unlock(); 119 } 120 121 /* Free the existing local addresses. */ 122 static void sctp_free_local_addr_list(struct net *net) 123 { 124 struct sctp_sockaddr_entry *addr; 125 struct list_head *pos, *temp; 126 127 list_for_each_safe(pos, temp, &net->sctp.local_addr_list) { 128 addr = list_entry(pos, struct sctp_sockaddr_entry, list); 129 list_del(pos); 130 kfree(addr); 131 } 132 } 133 134 /* Copy the local addresses which are valid for 'scope' into 'bp'. */ 135 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp, 136 enum sctp_scope scope, gfp_t gfp, int copy_flags) 137 { 138 struct sctp_sockaddr_entry *addr; 139 union sctp_addr laddr; 140 int error = 0; 141 142 rcu_read_lock(); 143 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) { 144 if (!addr->valid) 145 continue; 146 if (!sctp_in_scope(net, &addr->a, scope)) 147 continue; 148 149 /* Now that the address is in scope, check to see if 150 * the address type is really supported by the local 151 * sock as well as the remote peer. 152 */ 153 if (addr->a.sa.sa_family == AF_INET && 154 (!(copy_flags & SCTP_ADDR4_ALLOWED) || 155 !(copy_flags & SCTP_ADDR4_PEERSUPP))) 156 continue; 157 if (addr->a.sa.sa_family == AF_INET6 && 158 (!(copy_flags & SCTP_ADDR6_ALLOWED) || 159 !(copy_flags & SCTP_ADDR6_PEERSUPP))) 160 continue; 161 162 laddr = addr->a; 163 /* also works for setting ipv6 address port */ 164 laddr.v4.sin_port = htons(bp->port); 165 if (sctp_bind_addr_state(bp, &laddr) != -1) 166 continue; 167 168 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a), 169 SCTP_ADDR_SRC, GFP_ATOMIC); 170 if (error) 171 break; 172 } 173 174 rcu_read_unlock(); 175 return error; 176 } 177 178 /* Copy over any ip options */ 179 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk) 180 { 181 struct inet_sock *newinet, *inet = inet_sk(sk); 182 struct ip_options_rcu *inet_opt, *newopt = NULL; 183 184 newinet = inet_sk(newsk); 185 186 rcu_read_lock(); 187 inet_opt = rcu_dereference(inet->inet_opt); 188 if (inet_opt) { 189 newopt = sock_kmemdup(newsk, inet_opt, sizeof(*inet_opt) + 190 inet_opt->opt.optlen, GFP_ATOMIC); 191 if (!newopt) 192 pr_err("%s: Failed to copy ip options\n", __func__); 193 } 194 RCU_INIT_POINTER(newinet->inet_opt, newopt); 195 rcu_read_unlock(); 196 } 197 198 /* Account for the IP options */ 199 static int sctp_v4_ip_options_len(struct sock *sk) 200 { 201 struct inet_sock *inet = inet_sk(sk); 202 struct ip_options_rcu *inet_opt; 203 int len = 0; 204 205 rcu_read_lock(); 206 inet_opt = rcu_dereference(inet->inet_opt); 207 if (inet_opt) 208 len = inet_opt->opt.optlen; 209 210 rcu_read_unlock(); 211 return len; 212 } 213 214 /* Initialize a sctp_addr from in incoming skb. */ 215 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb, 216 int is_saddr) 217 { 218 /* Always called on head skb, so this is safe */ 219 struct sctphdr *sh = sctp_hdr(skb); 220 struct sockaddr_in *sa = &addr->v4; 221 222 addr->v4.sin_family = AF_INET; 223 224 if (is_saddr) { 225 sa->sin_port = sh->source; 226 sa->sin_addr.s_addr = ip_hdr(skb)->saddr; 227 } else { 228 sa->sin_port = sh->dest; 229 sa->sin_addr.s_addr = ip_hdr(skb)->daddr; 230 } 231 memset(sa->sin_zero, 0, sizeof(sa->sin_zero)); 232 } 233 234 /* Initialize an sctp_addr from a socket. */ 235 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk) 236 { 237 addr->v4.sin_family = AF_INET; 238 addr->v4.sin_port = 0; 239 addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr; 240 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 241 } 242 243 /* Initialize sk->sk_rcv_saddr from sctp_addr. */ 244 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk) 245 { 246 inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr; 247 } 248 249 /* Initialize sk->sk_daddr from sctp_addr. */ 250 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk) 251 { 252 inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr; 253 } 254 255 /* Initialize a sctp_addr from an address parameter. */ 256 static bool sctp_v4_from_addr_param(union sctp_addr *addr, 257 union sctp_addr_param *param, 258 __be16 port, int iif) 259 { 260 if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param)) 261 return false; 262 263 addr->v4.sin_family = AF_INET; 264 addr->v4.sin_port = port; 265 addr->v4.sin_addr.s_addr = param->v4.addr.s_addr; 266 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 267 268 return true; 269 } 270 271 /* Initialize an address parameter from a sctp_addr and return the length 272 * of the address parameter. 273 */ 274 static int sctp_v4_to_addr_param(const union sctp_addr *addr, 275 union sctp_addr_param *param) 276 { 277 int length = sizeof(struct sctp_ipv4addr_param); 278 279 param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS; 280 param->v4.param_hdr.length = htons(length); 281 param->v4.addr.s_addr = addr->v4.sin_addr.s_addr; 282 283 return length; 284 } 285 286 /* Initialize a sctp_addr from a dst_entry. */ 287 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4, 288 __be16 port) 289 { 290 saddr->v4.sin_family = AF_INET; 291 saddr->v4.sin_port = port; 292 saddr->v4.sin_addr.s_addr = fl4->saddr; 293 memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero)); 294 } 295 296 /* Compare two addresses exactly. */ 297 static int sctp_v4_cmp_addr(const union sctp_addr *addr1, 298 const union sctp_addr *addr2) 299 { 300 if (addr1->sa.sa_family != addr2->sa.sa_family) 301 return 0; 302 if (addr1->v4.sin_port != addr2->v4.sin_port) 303 return 0; 304 if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr) 305 return 0; 306 307 return 1; 308 } 309 310 /* Initialize addr struct to INADDR_ANY. */ 311 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port) 312 { 313 addr->v4.sin_family = AF_INET; 314 addr->v4.sin_addr.s_addr = htonl(INADDR_ANY); 315 addr->v4.sin_port = port; 316 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 317 } 318 319 /* Is this a wildcard address? */ 320 static int sctp_v4_is_any(const union sctp_addr *addr) 321 { 322 return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr; 323 } 324 325 /* This function checks if the address is a valid address to be used for 326 * SCTP binding. 327 * 328 * Output: 329 * Return 0 - If the address is a non-unicast or an illegal address. 330 * Return 1 - If the address is a unicast. 331 */ 332 static int sctp_v4_addr_valid(union sctp_addr *addr, 333 struct sctp_sock *sp, 334 const struct sk_buff *skb) 335 { 336 /* IPv4 addresses not allowed */ 337 if (sp && ipv6_only_sock(sctp_opt2sk(sp))) 338 return 0; 339 340 /* Is this a non-unicast address or a unusable SCTP address? */ 341 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) 342 return 0; 343 344 /* Is this a broadcast address? */ 345 if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST) 346 return 0; 347 348 return 1; 349 } 350 351 /* Should this be available for binding? */ 352 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp) 353 { 354 struct sock *sk = &sp->inet.sk; 355 struct net *net = sock_net(sk); 356 int tb_id = RT_TABLE_LOCAL; 357 int ret; 358 359 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ?: tb_id; 360 ret = inet_addr_type_table(net, addr->v4.sin_addr.s_addr, tb_id); 361 if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) && 362 ret != RTN_LOCAL && 363 !inet_test_bit(FREEBIND, sk) && 364 !READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind)) 365 return 0; 366 367 if (ipv6_only_sock(sctp_opt2sk(sp))) 368 return 0; 369 370 return 1; 371 } 372 373 /* Checking the loopback, private and other address scopes as defined in 374 * RFC 1918. The IPv4 scoping is based on the draft for SCTP IPv4 375 * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>. 376 * 377 * Level 0 - unusable SCTP addresses 378 * Level 1 - loopback address 379 * Level 2 - link-local addresses 380 * Level 3 - private addresses. 381 * Level 4 - global addresses 382 * For INIT and INIT-ACK address list, let L be the level of 383 * requested destination address, sender and receiver 384 * SHOULD include all of its addresses with level greater 385 * than or equal to L. 386 * 387 * IPv4 scoping can be controlled through sysctl option 388 * net.sctp.addr_scope_policy 389 */ 390 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr) 391 { 392 enum sctp_scope retval; 393 394 /* Check for unusable SCTP addresses. */ 395 if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) { 396 retval = SCTP_SCOPE_UNUSABLE; 397 } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) { 398 retval = SCTP_SCOPE_LOOPBACK; 399 } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) { 400 retval = SCTP_SCOPE_LINK; 401 } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) || 402 ipv4_is_private_172(addr->v4.sin_addr.s_addr) || 403 ipv4_is_private_192(addr->v4.sin_addr.s_addr) || 404 ipv4_is_test_198(addr->v4.sin_addr.s_addr)) { 405 retval = SCTP_SCOPE_PRIVATE; 406 } else { 407 retval = SCTP_SCOPE_GLOBAL; 408 } 409 410 return retval; 411 } 412 413 /* Returns a valid dst cache entry for the given source and destination ip 414 * addresses. If an association is passed, trys to get a dst entry with a 415 * source address that matches an address in the bind address list. 416 */ 417 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr, 418 struct flowi *fl, struct sock *sk) 419 { 420 struct sctp_association *asoc = t->asoc; 421 struct rtable *rt; 422 struct flowi _fl; 423 struct flowi4 *fl4 = &_fl.u.ip4; 424 struct sctp_bind_addr *bp; 425 struct sctp_sockaddr_entry *laddr; 426 struct dst_entry *dst = NULL; 427 union sctp_addr *daddr = &t->ipaddr; 428 union sctp_addr dst_saddr; 429 dscp_t dscp; 430 431 if (t->dscp & SCTP_DSCP_SET_MASK) 432 dscp = inet_dsfield_to_dscp(t->dscp); 433 else 434 dscp = inet_sk_dscp(inet_sk(sk)); 435 436 memset(&_fl, 0x0, sizeof(_fl)); 437 fl4->daddr = daddr->v4.sin_addr.s_addr; 438 fl4->fl4_dport = daddr->v4.sin_port; 439 fl4->flowi4_proto = IPPROTO_SCTP; 440 if (asoc) { 441 fl4->flowi4_dscp = dscp; 442 fl4->flowi4_scope = ip_sock_rt_scope(asoc->base.sk); 443 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if; 444 fl4->fl4_sport = htons(asoc->base.bind_addr.port); 445 } 446 if (saddr) { 447 fl4->saddr = saddr->v4.sin_addr.s_addr; 448 if (!fl4->fl4_sport) 449 fl4->fl4_sport = saddr->v4.sin_port; 450 } 451 452 pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr, 453 &fl4->saddr); 454 455 rt = ip_route_output_key(sock_net(sk), fl4); 456 if (!IS_ERR(rt)) { 457 dst = &rt->dst; 458 t->dst = dst; 459 memcpy(fl, &_fl, sizeof(_fl)); 460 } 461 462 /* If there is no association or if a source address is passed, no 463 * more validation is required. 464 */ 465 if (!asoc || saddr) 466 goto out; 467 468 bp = &asoc->base.bind_addr; 469 470 if (dst) { 471 /* Walk through the bind address list and look for a bind 472 * address that matches the source address of the returned dst. 473 */ 474 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port)); 475 rcu_read_lock(); 476 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 477 if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) || 478 (laddr->state != SCTP_ADDR_SRC && 479 !asoc->src_out_of_asoc_ok)) 480 continue; 481 if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a)) 482 goto out_unlock; 483 } 484 rcu_read_unlock(); 485 486 /* None of the bound addresses match the source address of the 487 * dst. So release it. 488 */ 489 dst_release(dst); 490 dst = NULL; 491 } 492 493 /* Walk through the bind address list and try to get a dst that 494 * matches a bind address as the source address. 495 */ 496 rcu_read_lock(); 497 list_for_each_entry_rcu(laddr, &bp->address_list, list) { 498 struct net_device *odev; 499 500 if (!laddr->valid) 501 continue; 502 if (laddr->state != SCTP_ADDR_SRC || 503 AF_INET != laddr->a.sa.sa_family) 504 continue; 505 506 fl4->fl4_sport = laddr->a.v4.sin_port; 507 flowi4_update_output(fl4, asoc->base.sk->sk_bound_dev_if, 508 daddr->v4.sin_addr.s_addr, 509 laddr->a.v4.sin_addr.s_addr); 510 511 rt = ip_route_output_key(sock_net(sk), fl4); 512 if (IS_ERR(rt)) 513 continue; 514 515 /* Ensure the src address belongs to the output 516 * interface. 517 */ 518 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr, 519 false); 520 if (!odev || odev->ifindex != fl4->flowi4_oif) { 521 if (!dst) { 522 dst = &rt->dst; 523 t->dst = dst; 524 memcpy(fl, &_fl, sizeof(_fl)); 525 } else { 526 dst_release(&rt->dst); 527 } 528 continue; 529 } 530 531 dst_release(dst); 532 dst = &rt->dst; 533 t->dst = dst; 534 memcpy(fl, &_fl, sizeof(_fl)); 535 break; 536 } 537 538 out_unlock: 539 rcu_read_unlock(); 540 out: 541 if (dst) { 542 pr_debug("rt_dst:%pI4, rt_src:%pI4\n", 543 &fl->u.ip4.daddr, &fl->u.ip4.saddr); 544 } else { 545 t->dst = NULL; 546 pr_debug("no route\n"); 547 } 548 } 549 550 /* For v4, the source address is cached in the route entry(dst). So no need 551 * to cache it separately and hence this is an empty routine. 552 */ 553 static void sctp_v4_get_saddr(struct sctp_sock *sk, 554 struct sctp_transport *t, 555 struct flowi *fl) 556 { 557 union sctp_addr *saddr = &t->saddr; 558 struct rtable *rt = dst_rtable(t->dst); 559 560 if (rt) { 561 saddr->v4.sin_family = AF_INET; 562 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr; 563 } 564 } 565 566 /* What interface did this skb arrive on? */ 567 static int sctp_v4_skb_iif(const struct sk_buff *skb) 568 { 569 return inet_iif(skb); 570 } 571 572 static int sctp_v4_skb_sdif(const struct sk_buff *skb) 573 { 574 return inet_sdif(skb); 575 } 576 577 /* Was this packet marked by Explicit Congestion Notification? */ 578 static int sctp_v4_is_ce(const struct sk_buff *skb) 579 { 580 return INET_ECN_is_ce(ip_hdr(skb)->tos); 581 } 582 583 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr) 584 { 585 /* No address mapping for V4 sockets */ 586 memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero)); 587 return sizeof(struct sockaddr_in); 588 } 589 590 /* Dump the v4 addr to the seq file. */ 591 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr) 592 { 593 seq_printf(seq, "%pI4 ", &addr->v4.sin_addr); 594 } 595 596 static void sctp_v4_ecn_capable(struct sock *sk) 597 { 598 INET_ECN_xmit(sk); 599 } 600 601 static void sctp_addr_wq_timeout_handler(struct timer_list *t) 602 { 603 struct net *net = timer_container_of(net, t, sctp.addr_wq_timer); 604 struct sctp_sockaddr_entry *addrw, *temp; 605 struct sctp_sock *sp; 606 607 spin_lock_bh(&net->sctp.addr_wq_lock); 608 609 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 610 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at " 611 "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa, 612 addrw->state, addrw); 613 614 #if IS_ENABLED(CONFIG_IPV6) 615 /* Now we send an ASCONF for each association */ 616 /* Note. we currently don't handle link local IPv6 addressees */ 617 if (addrw->a.sa.sa_family == AF_INET6) { 618 struct in6_addr *in6; 619 620 if (ipv6_addr_type(&addrw->a.v6.sin6_addr) & 621 IPV6_ADDR_LINKLOCAL) 622 goto free_next; 623 624 in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr; 625 if (ipv6_chk_addr(net, in6, NULL, 0) == 0 && 626 addrw->state == SCTP_ADDR_NEW) { 627 unsigned long timeo_val; 628 629 pr_debug("%s: this is on DAD, trying %d sec " 630 "later\n", __func__, 631 SCTP_ADDRESS_TICK_DELAY); 632 633 timeo_val = jiffies; 634 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 635 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 636 break; 637 } 638 } 639 #endif 640 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) { 641 struct sock *sk; 642 643 sk = sctp_opt2sk(sp); 644 /* ignore bound-specific endpoints */ 645 if (!sctp_is_ep_boundall(sk)) 646 continue; 647 bh_lock_sock(sk); 648 if (sctp_asconf_mgmt(sp, addrw) < 0) 649 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__); 650 bh_unlock_sock(sk); 651 } 652 #if IS_ENABLED(CONFIG_IPV6) 653 free_next: 654 #endif 655 list_del(&addrw->list); 656 kfree(addrw); 657 } 658 spin_unlock_bh(&net->sctp.addr_wq_lock); 659 } 660 661 static void sctp_free_addr_wq(struct net *net) 662 { 663 struct sctp_sockaddr_entry *addrw; 664 struct sctp_sockaddr_entry *temp; 665 666 spin_lock_bh(&net->sctp.addr_wq_lock); 667 timer_delete(&net->sctp.addr_wq_timer); 668 list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) { 669 list_del(&addrw->list); 670 kfree(addrw); 671 } 672 spin_unlock_bh(&net->sctp.addr_wq_lock); 673 } 674 675 /* lookup the entry for the same address in the addr_waitq 676 * sctp_addr_wq MUST be locked 677 */ 678 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net, 679 struct sctp_sockaddr_entry *addr) 680 { 681 struct sctp_sockaddr_entry *addrw; 682 683 list_for_each_entry(addrw, &net->sctp.addr_waitq, list) { 684 if (addrw->a.sa.sa_family != addr->a.sa.sa_family) 685 continue; 686 if (addrw->a.sa.sa_family == AF_INET) { 687 if (addrw->a.v4.sin_addr.s_addr == 688 addr->a.v4.sin_addr.s_addr) 689 return addrw; 690 } else if (addrw->a.sa.sa_family == AF_INET6) { 691 if (ipv6_addr_equal(&addrw->a.v6.sin6_addr, 692 &addr->a.v6.sin6_addr)) 693 return addrw; 694 } 695 } 696 return NULL; 697 } 698 699 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd) 700 { 701 struct sctp_sockaddr_entry *addrw; 702 unsigned long timeo_val; 703 704 /* first, we check if an opposite message already exist in the queue. 705 * If we found such message, it is removed. 706 * This operation is a bit stupid, but the DHCP client attaches the 707 * new address after a couple of addition and deletion of that address 708 */ 709 710 spin_lock_bh(&net->sctp.addr_wq_lock); 711 712 /* Avoid searching the queue or modifying it if there are no consumers, 713 * as it can lead to performance degradation if addresses are modified 714 * en-masse. 715 * 716 * If the queue already contains some events, update it anyway to avoid 717 * ugly races between new sessions and new address events. 718 */ 719 if (list_empty(&net->sctp.auto_asconf_splist) && 720 list_empty(&net->sctp.addr_waitq)) { 721 spin_unlock_bh(&net->sctp.addr_wq_lock); 722 return; 723 } 724 725 /* Offsets existing events in addr_wq */ 726 addrw = sctp_addr_wq_lookup(net, addr); 727 if (addrw) { 728 if (addrw->state != cmd) { 729 pr_debug("%s: offsets existing entry for %d, addr:%pISc " 730 "in wq:%p\n", __func__, addrw->state, &addrw->a.sa, 731 &net->sctp.addr_waitq); 732 733 list_del(&addrw->list); 734 kfree(addrw); 735 } 736 spin_unlock_bh(&net->sctp.addr_wq_lock); 737 return; 738 } 739 740 /* OK, we have to add the new address to the wait queue */ 741 addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC); 742 if (addrw == NULL) { 743 spin_unlock_bh(&net->sctp.addr_wq_lock); 744 return; 745 } 746 addrw->state = cmd; 747 list_add_tail(&addrw->list, &net->sctp.addr_waitq); 748 749 pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n", 750 __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq); 751 752 if (!timer_pending(&net->sctp.addr_wq_timer)) { 753 timeo_val = jiffies; 754 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY); 755 mod_timer(&net->sctp.addr_wq_timer, timeo_val); 756 } 757 spin_unlock_bh(&net->sctp.addr_wq_lock); 758 } 759 760 /* Event handler for inet address addition/deletion events. 761 * The sctp_local_addr_list needs to be protocted by a spin lock since 762 * multiple notifiers (say IPv4 and IPv6) may be running at the same 763 * time and thus corrupt the list. 764 * The reader side is protected with RCU. 765 */ 766 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev, 767 void *ptr) 768 { 769 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; 770 struct sctp_sockaddr_entry *addr = NULL; 771 struct sctp_sockaddr_entry *temp; 772 struct net *net = dev_net(ifa->ifa_dev->dev); 773 int found = 0; 774 775 switch (ev) { 776 case NETDEV_UP: 777 addr = kzalloc(sizeof(*addr), GFP_ATOMIC); 778 if (addr) { 779 addr->a.v4.sin_family = AF_INET; 780 addr->a.v4.sin_addr.s_addr = ifa->ifa_local; 781 addr->valid = 1; 782 spin_lock_bh(&net->sctp.local_addr_lock); 783 list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list); 784 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW); 785 spin_unlock_bh(&net->sctp.local_addr_lock); 786 } 787 break; 788 case NETDEV_DOWN: 789 spin_lock_bh(&net->sctp.local_addr_lock); 790 list_for_each_entry_safe(addr, temp, 791 &net->sctp.local_addr_list, list) { 792 if (addr->a.sa.sa_family == AF_INET && 793 addr->a.v4.sin_addr.s_addr == 794 ifa->ifa_local) { 795 found = 1; 796 addr->valid = 0; 797 list_del_rcu(&addr->list); 798 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL); 799 break; 800 } 801 } 802 spin_unlock_bh(&net->sctp.local_addr_lock); 803 if (found) 804 kfree_rcu(addr, rcu); 805 break; 806 } 807 808 return NOTIFY_DONE; 809 } 810 811 /* 812 * Initialize the control inode/socket with a control endpoint data 813 * structure. This endpoint is reserved exclusively for the OOTB processing. 814 */ 815 static int sctp_ctl_sock_init(struct net *net) 816 { 817 int err; 818 sa_family_t family = PF_INET; 819 820 if (sctp_get_pf_specific(PF_INET6)) 821 family = PF_INET6; 822 823 err = inet_ctl_sock_create(&net->sctp.ctl_sock, family, 824 SOCK_SEQPACKET, IPPROTO_SCTP, net); 825 826 /* If IPv6 socket could not be created, try the IPv4 socket */ 827 if (err < 0 && family == PF_INET6) 828 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET, 829 SOCK_SEQPACKET, IPPROTO_SCTP, 830 net); 831 832 if (err < 0) { 833 pr_err("Failed to create the SCTP control socket\n"); 834 return err; 835 } 836 return 0; 837 } 838 839 static int sctp_udp_rcv(struct sock *sk, struct sk_buff *skb) 840 { 841 SCTP_INPUT_CB(skb)->encap_port = udp_hdr(skb)->source; 842 843 skb_set_transport_header(skb, sizeof(struct udphdr)); 844 sctp_rcv(skb); 845 return 0; 846 } 847 848 int sctp_udp_sock_start(struct net *net) 849 { 850 struct udp_tunnel_sock_cfg tuncfg = {NULL}; 851 struct udp_port_cfg udp_conf = {0}; 852 struct socket *sock; 853 int err; 854 855 udp_conf.family = AF_INET; 856 udp_conf.local_ip.s_addr = htonl(INADDR_ANY); 857 udp_conf.local_udp_port = htons(net->sctp.udp_port); 858 err = udp_sock_create(net, &udp_conf, &sock); 859 if (err) { 860 pr_err("Failed to create the SCTP UDP tunneling v4 sock\n"); 861 return err; 862 } 863 864 tuncfg.encap_type = 1; 865 tuncfg.encap_rcv = sctp_udp_rcv; 866 tuncfg.encap_err_lookup = sctp_udp_v4_err; 867 setup_udp_tunnel_sock(net, sock, &tuncfg); 868 net->sctp.udp4_sock = sock->sk; 869 870 #if IS_ENABLED(CONFIG_IPV6) 871 memset(&udp_conf, 0, sizeof(udp_conf)); 872 873 udp_conf.family = AF_INET6; 874 udp_conf.local_ip6 = in6addr_any; 875 udp_conf.local_udp_port = htons(net->sctp.udp_port); 876 udp_conf.use_udp6_rx_checksums = true; 877 udp_conf.ipv6_v6only = true; 878 err = udp_sock_create(net, &udp_conf, &sock); 879 if (err) { 880 pr_err("Failed to create the SCTP UDP tunneling v6 sock\n"); 881 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 882 net->sctp.udp4_sock = NULL; 883 return err; 884 } 885 886 tuncfg.encap_type = 1; 887 tuncfg.encap_rcv = sctp_udp_rcv; 888 tuncfg.encap_err_lookup = sctp_udp_v6_err; 889 setup_udp_tunnel_sock(net, sock, &tuncfg); 890 net->sctp.udp6_sock = sock->sk; 891 #endif 892 893 return 0; 894 } 895 896 void sctp_udp_sock_stop(struct net *net) 897 { 898 if (net->sctp.udp4_sock) { 899 udp_tunnel_sock_release(net->sctp.udp4_sock->sk_socket); 900 net->sctp.udp4_sock = NULL; 901 } 902 if (net->sctp.udp6_sock) { 903 udp_tunnel_sock_release(net->sctp.udp6_sock->sk_socket); 904 net->sctp.udp6_sock = NULL; 905 } 906 } 907 908 /* Register address family specific functions. */ 909 int sctp_register_af(struct sctp_af *af) 910 { 911 switch (af->sa_family) { 912 case AF_INET: 913 if (sctp_af_v4_specific) 914 return 0; 915 sctp_af_v4_specific = af; 916 break; 917 case AF_INET6: 918 if (sctp_af_v6_specific) 919 return 0; 920 sctp_af_v6_specific = af; 921 break; 922 default: 923 return 0; 924 } 925 926 INIT_LIST_HEAD(&af->list); 927 list_add_tail(&af->list, &sctp_address_families); 928 return 1; 929 } 930 931 /* Get the table of functions for manipulating a particular address 932 * family. 933 */ 934 struct sctp_af *sctp_get_af_specific(sa_family_t family) 935 { 936 switch (family) { 937 case AF_INET: 938 return sctp_af_v4_specific; 939 case AF_INET6: 940 return sctp_af_v6_specific; 941 default: 942 return NULL; 943 } 944 } 945 946 /* Common code to initialize a AF_INET msg_name. */ 947 static void sctp_inet_msgname(char *msgname, int *addr_len) 948 { 949 struct sockaddr_in *sin; 950 951 sin = (struct sockaddr_in *)msgname; 952 *addr_len = sizeof(struct sockaddr_in); 953 sin->sin_family = AF_INET; 954 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 955 } 956 957 /* Copy the primary address of the peer primary address as the msg_name. */ 958 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname, 959 int *addr_len) 960 { 961 struct sockaddr_in *sin, *sinfrom; 962 963 if (msgname) { 964 struct sctp_association *asoc; 965 966 asoc = event->asoc; 967 sctp_inet_msgname(msgname, addr_len); 968 sin = (struct sockaddr_in *)msgname; 969 sinfrom = &asoc->peer.primary_addr.v4; 970 sin->sin_port = htons(asoc->peer.port); 971 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr; 972 } 973 } 974 975 /* Initialize and copy out a msgname from an inbound skb. */ 976 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len) 977 { 978 if (msgname) { 979 struct sctphdr *sh = sctp_hdr(skb); 980 struct sockaddr_in *sin = (struct sockaddr_in *)msgname; 981 982 sctp_inet_msgname(msgname, len); 983 sin->sin_port = sh->source; 984 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 985 } 986 } 987 988 /* Do we support this AF? */ 989 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp) 990 { 991 /* PF_INET only supports AF_INET addresses. */ 992 return AF_INET == family; 993 } 994 995 /* Address matching with wildcards allowed. */ 996 static int sctp_inet_cmp_addr(const union sctp_addr *addr1, 997 const union sctp_addr *addr2, 998 struct sctp_sock *opt) 999 { 1000 /* PF_INET only supports AF_INET addresses. */ 1001 if (addr1->sa.sa_family != addr2->sa.sa_family) 1002 return 0; 1003 if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr || 1004 htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr) 1005 return 1; 1006 if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr) 1007 return 1; 1008 1009 return 0; 1010 } 1011 1012 /* Verify that provided sockaddr looks bindable. Common verification has 1013 * already been taken care of. 1014 */ 1015 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr) 1016 { 1017 return sctp_v4_available(addr, opt); 1018 } 1019 1020 /* Verify that sockaddr looks sendable. Common verification has already 1021 * been taken care of. 1022 */ 1023 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr) 1024 { 1025 return 1; 1026 } 1027 1028 /* Fill in Supported Address Type information for INIT and INIT-ACK 1029 * chunks. Returns number of addresses supported. 1030 */ 1031 static int sctp_inet_supported_addrs(const struct sctp_sock *opt, 1032 __be16 *types) 1033 { 1034 types[0] = SCTP_PARAM_IPV4_ADDRESS; 1035 return 1; 1036 } 1037 1038 /* Wrapper routine that calls the ip transmit routine. */ 1039 static inline int sctp_v4_xmit(struct sk_buff *skb, struct sctp_transport *t) 1040 { 1041 struct dst_entry *dst = dst_clone(t->dst); 1042 struct flowi4 *fl4 = &t->fl.u.ip4; 1043 struct sock *sk = skb->sk; 1044 struct inet_sock *inet = inet_sk(sk); 1045 __u8 dscp = READ_ONCE(inet->tos); 1046 __be16 df = 0; 1047 1048 pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb, 1049 skb->len, &fl4->saddr, &fl4->daddr); 1050 1051 if (t->dscp & SCTP_DSCP_SET_MASK) 1052 dscp = t->dscp & SCTP_DSCP_VAL_MASK; 1053 1054 inet->pmtudisc = t->param_flags & SPP_PMTUD_ENABLE ? IP_PMTUDISC_DO 1055 : IP_PMTUDISC_DONT; 1056 SCTP_INC_STATS(sock_net(sk), SCTP_MIB_OUTSCTPPACKS); 1057 1058 if (!t->encap_port || !sctp_sk(sk)->udp_port) { 1059 skb_dst_set(skb, dst); 1060 return __ip_queue_xmit(sk, skb, &t->fl, dscp); 1061 } 1062 1063 if (skb_is_gso(skb)) 1064 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM; 1065 1066 if (ip_dont_fragment(sk, dst) && !skb->ignore_df) 1067 df = htons(IP_DF); 1068 1069 skb->encapsulation = 1; 1070 skb_reset_inner_mac_header(skb); 1071 skb_reset_inner_transport_header(skb); 1072 skb_set_inner_ipproto(skb, IPPROTO_SCTP); 1073 udp_tunnel_xmit_skb(dst_rtable(dst), sk, skb, fl4->saddr, 1074 fl4->daddr, dscp, ip4_dst_hoplimit(dst), df, 1075 sctp_sk(sk)->udp_port, t->encap_port, false, false, 1076 0); 1077 return 0; 1078 } 1079 1080 static struct sctp_af sctp_af_inet; 1081 1082 static struct sctp_pf sctp_pf_inet = { 1083 .event_msgname = sctp_inet_event_msgname, 1084 .skb_msgname = sctp_inet_skb_msgname, 1085 .af_supported = sctp_inet_af_supported, 1086 .cmp_addr = sctp_inet_cmp_addr, 1087 .bind_verify = sctp_inet_bind_verify, 1088 .send_verify = sctp_inet_send_verify, 1089 .supported_addrs = sctp_inet_supported_addrs, 1090 .addr_to_user = sctp_v4_addr_to_user, 1091 .to_sk_saddr = sctp_v4_to_sk_saddr, 1092 .to_sk_daddr = sctp_v4_to_sk_daddr, 1093 .copy_ip_options = sctp_v4_copy_ip_options, 1094 .af = &sctp_af_inet 1095 }; 1096 1097 /* Notifier for inetaddr addition/deletion events. */ 1098 static struct notifier_block sctp_inetaddr_notifier = { 1099 .notifier_call = sctp_inetaddr_event, 1100 }; 1101 1102 /* Socket operations. */ 1103 static const struct proto_ops inet_seqpacket_ops = { 1104 .family = PF_INET, 1105 .owner = THIS_MODULE, 1106 .release = inet_release, /* Needs to be wrapped... */ 1107 .bind = inet_bind, 1108 .connect = sctp_inet_connect, 1109 .socketpair = sock_no_socketpair, 1110 .accept = inet_accept, 1111 .getname = inet_getname, /* Semantics are different. */ 1112 .poll = sctp_poll, 1113 .ioctl = inet_ioctl, 1114 .gettstamp = sock_gettstamp, 1115 .listen = sctp_inet_listen, 1116 .shutdown = inet_shutdown, /* Looks harmless. */ 1117 .setsockopt = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */ 1118 .getsockopt = sock_common_getsockopt, 1119 .sendmsg = inet_sendmsg, 1120 .recvmsg = inet_recvmsg, 1121 .mmap = sock_no_mmap, 1122 }; 1123 1124 /* Registration with AF_INET family. */ 1125 static struct inet_protosw sctp_seqpacket_protosw = { 1126 .type = SOCK_SEQPACKET, 1127 .protocol = IPPROTO_SCTP, 1128 .prot = &sctp_prot, 1129 .ops = &inet_seqpacket_ops, 1130 .flags = SCTP_PROTOSW_FLAG 1131 }; 1132 static struct inet_protosw sctp_stream_protosw = { 1133 .type = SOCK_STREAM, 1134 .protocol = IPPROTO_SCTP, 1135 .prot = &sctp_prot, 1136 .ops = &inet_seqpacket_ops, 1137 .flags = SCTP_PROTOSW_FLAG 1138 }; 1139 1140 static int sctp4_rcv(struct sk_buff *skb) 1141 { 1142 SCTP_INPUT_CB(skb)->encap_port = 0; 1143 return sctp_rcv(skb); 1144 } 1145 1146 /* Register with IP layer. */ 1147 static const struct net_protocol sctp_protocol = { 1148 .handler = sctp4_rcv, 1149 .err_handler = sctp_v4_err, 1150 .no_policy = 1, 1151 .icmp_strict_tag_validation = 1, 1152 }; 1153 1154 /* IPv4 address related functions. */ 1155 static struct sctp_af sctp_af_inet = { 1156 .sa_family = AF_INET, 1157 .sctp_xmit = sctp_v4_xmit, 1158 .setsockopt = ip_setsockopt, 1159 .getsockopt = ip_getsockopt, 1160 .get_dst = sctp_v4_get_dst, 1161 .get_saddr = sctp_v4_get_saddr, 1162 .copy_addrlist = sctp_v4_copy_addrlist, 1163 .from_skb = sctp_v4_from_skb, 1164 .from_sk = sctp_v4_from_sk, 1165 .from_addr_param = sctp_v4_from_addr_param, 1166 .to_addr_param = sctp_v4_to_addr_param, 1167 .cmp_addr = sctp_v4_cmp_addr, 1168 .addr_valid = sctp_v4_addr_valid, 1169 .inaddr_any = sctp_v4_inaddr_any, 1170 .is_any = sctp_v4_is_any, 1171 .available = sctp_v4_available, 1172 .scope = sctp_v4_scope, 1173 .skb_iif = sctp_v4_skb_iif, 1174 .skb_sdif = sctp_v4_skb_sdif, 1175 .is_ce = sctp_v4_is_ce, 1176 .seq_dump_addr = sctp_v4_seq_dump_addr, 1177 .ecn_capable = sctp_v4_ecn_capable, 1178 .net_header_len = sizeof(struct iphdr), 1179 .sockaddr_len = sizeof(struct sockaddr_in), 1180 .ip_options_len = sctp_v4_ip_options_len, 1181 }; 1182 1183 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) 1184 { 1185 switch (family) { 1186 case PF_INET: 1187 return sctp_pf_inet_specific; 1188 case PF_INET6: 1189 return sctp_pf_inet6_specific; 1190 default: 1191 return NULL; 1192 } 1193 } 1194 1195 /* Register the PF specific function table. */ 1196 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family) 1197 { 1198 switch (family) { 1199 case PF_INET: 1200 if (sctp_pf_inet_specific) 1201 return 0; 1202 sctp_pf_inet_specific = pf; 1203 break; 1204 case PF_INET6: 1205 if (sctp_pf_inet6_specific) 1206 return 0; 1207 sctp_pf_inet6_specific = pf; 1208 break; 1209 default: 1210 return 0; 1211 } 1212 return 1; 1213 } 1214 1215 static inline int init_sctp_mibs(struct net *net) 1216 { 1217 net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib); 1218 if (!net->sctp.sctp_statistics) 1219 return -ENOMEM; 1220 return 0; 1221 } 1222 1223 static inline void cleanup_sctp_mibs(struct net *net) 1224 { 1225 free_percpu(net->sctp.sctp_statistics); 1226 } 1227 1228 static void sctp_v4_pf_init(void) 1229 { 1230 /* Initialize the SCTP specific PF functions. */ 1231 sctp_register_pf(&sctp_pf_inet, PF_INET); 1232 sctp_register_af(&sctp_af_inet); 1233 } 1234 1235 static void sctp_v4_pf_exit(void) 1236 { 1237 list_del(&sctp_af_inet.list); 1238 } 1239 1240 static int sctp_v4_protosw_init(void) 1241 { 1242 int rc; 1243 1244 rc = proto_register(&sctp_prot, 1); 1245 if (rc) 1246 return rc; 1247 1248 /* Register SCTP(UDP and TCP style) with socket layer. */ 1249 inet_register_protosw(&sctp_seqpacket_protosw); 1250 inet_register_protosw(&sctp_stream_protosw); 1251 1252 return 0; 1253 } 1254 1255 static void sctp_v4_protosw_exit(void) 1256 { 1257 inet_unregister_protosw(&sctp_stream_protosw); 1258 inet_unregister_protosw(&sctp_seqpacket_protosw); 1259 proto_unregister(&sctp_prot); 1260 } 1261 1262 static int sctp_v4_add_protocol(void) 1263 { 1264 /* Register notifier for inet address additions/deletions. */ 1265 register_inetaddr_notifier(&sctp_inetaddr_notifier); 1266 1267 /* Register SCTP with inet layer. */ 1268 if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0) 1269 return -EAGAIN; 1270 1271 return 0; 1272 } 1273 1274 static void sctp_v4_del_protocol(void) 1275 { 1276 inet_del_protocol(&sctp_protocol, IPPROTO_SCTP); 1277 unregister_inetaddr_notifier(&sctp_inetaddr_notifier); 1278 } 1279 1280 static int __net_init sctp_defaults_init(struct net *net) 1281 { 1282 int status; 1283 1284 /* 1285 * 14. Suggested SCTP Protocol Parameter Values 1286 */ 1287 /* The following protocol parameters are RECOMMENDED: */ 1288 /* RTO.Initial - 3 seconds */ 1289 net->sctp.rto_initial = SCTP_RTO_INITIAL; 1290 /* RTO.Min - 1 second */ 1291 net->sctp.rto_min = SCTP_RTO_MIN; 1292 /* RTO.Max - 60 seconds */ 1293 net->sctp.rto_max = SCTP_RTO_MAX; 1294 /* RTO.Alpha - 1/8 */ 1295 net->sctp.rto_alpha = SCTP_RTO_ALPHA; 1296 /* RTO.Beta - 1/4 */ 1297 net->sctp.rto_beta = SCTP_RTO_BETA; 1298 1299 /* Valid.Cookie.Life - 60 seconds */ 1300 net->sctp.valid_cookie_life = SCTP_DEFAULT_COOKIE_LIFE; 1301 1302 /* Whether Cookie Preservative is enabled(1) or not(0) */ 1303 net->sctp.cookie_preserve_enable = 1; 1304 1305 /* Whether cookie authentication is enabled(1) or not(0) */ 1306 net->sctp.cookie_auth_enable = 1307 !IS_ENABLED(CONFIG_SCTP_DEFAULT_COOKIE_HMAC_NONE); 1308 1309 /* Max.Burst - 4 */ 1310 net->sctp.max_burst = SCTP_DEFAULT_MAX_BURST; 1311 1312 /* Disable of Primary Path Switchover by default */ 1313 net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX; 1314 1315 /* Enable pf state by default */ 1316 net->sctp.pf_enable = 1; 1317 1318 /* Ignore pf exposure feature by default */ 1319 net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET; 1320 1321 /* Association.Max.Retrans - 10 attempts 1322 * Path.Max.Retrans - 5 attempts (per destination address) 1323 * Max.Init.Retransmits - 8 attempts 1324 */ 1325 net->sctp.max_retrans_association = 10; 1326 net->sctp.max_retrans_path = 5; 1327 net->sctp.max_retrans_init = 8; 1328 1329 /* Sendbuffer growth - do per-socket accounting */ 1330 net->sctp.sndbuf_policy = 0; 1331 1332 /* Rcvbuffer growth - do per-socket accounting */ 1333 net->sctp.rcvbuf_policy = 0; 1334 1335 /* HB.interval - 30 seconds */ 1336 net->sctp.hb_interval = SCTP_DEFAULT_TIMEOUT_HEARTBEAT; 1337 1338 /* delayed SACK timeout */ 1339 net->sctp.sack_timeout = SCTP_DEFAULT_TIMEOUT_SACK; 1340 1341 /* Disable ADDIP by default. */ 1342 net->sctp.addip_enable = 0; 1343 net->sctp.addip_noauth = 0; 1344 net->sctp.default_auto_asconf = 0; 1345 1346 /* Enable PR-SCTP by default. */ 1347 net->sctp.prsctp_enable = 1; 1348 1349 /* Disable RECONF by default. */ 1350 net->sctp.reconf_enable = 0; 1351 1352 /* Disable AUTH by default. */ 1353 net->sctp.auth_enable = 0; 1354 1355 /* Enable ECN by default. */ 1356 net->sctp.ecn_enable = 1; 1357 1358 /* Set UDP tunneling listening port to 0 by default */ 1359 net->sctp.udp_port = 0; 1360 1361 /* Set remote encap port to 0 by default */ 1362 net->sctp.encap_port = 0; 1363 1364 /* Set SCOPE policy to enabled */ 1365 net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE; 1366 1367 /* Set the default rwnd update threshold */ 1368 net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT; 1369 1370 /* Initialize maximum autoclose timeout. */ 1371 net->sctp.max_autoclose = INT_MAX / HZ; 1372 1373 #ifdef CONFIG_NET_L3_MASTER_DEV 1374 net->sctp.l3mdev_accept = 1; 1375 #endif 1376 1377 status = sctp_sysctl_net_register(net); 1378 if (status) 1379 goto err_sysctl_register; 1380 1381 /* Allocate and initialise sctp mibs. */ 1382 status = init_sctp_mibs(net); 1383 if (status) 1384 goto err_init_mibs; 1385 1386 #ifdef CONFIG_PROC_FS 1387 /* Initialize proc fs directory. */ 1388 status = sctp_proc_init(net); 1389 if (status) 1390 goto err_init_proc; 1391 #endif 1392 1393 sctp_dbg_objcnt_init(net); 1394 1395 /* Initialize the local address list. */ 1396 INIT_LIST_HEAD(&net->sctp.local_addr_list); 1397 spin_lock_init(&net->sctp.local_addr_lock); 1398 sctp_get_local_addr_list(net); 1399 1400 /* Initialize the address event list */ 1401 INIT_LIST_HEAD(&net->sctp.addr_waitq); 1402 INIT_LIST_HEAD(&net->sctp.auto_asconf_splist); 1403 spin_lock_init(&net->sctp.addr_wq_lock); 1404 net->sctp.addr_wq_timer.expires = 0; 1405 timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0); 1406 1407 return 0; 1408 1409 #ifdef CONFIG_PROC_FS 1410 err_init_proc: 1411 cleanup_sctp_mibs(net); 1412 #endif 1413 err_init_mibs: 1414 sctp_sysctl_net_unregister(net); 1415 err_sysctl_register: 1416 return status; 1417 } 1418 1419 static void __net_exit sctp_defaults_exit(struct net *net) 1420 { 1421 /* Free the local address list */ 1422 sctp_free_addr_wq(net); 1423 sctp_free_local_addr_list(net); 1424 1425 #ifdef CONFIG_PROC_FS 1426 remove_proc_subtree("sctp", net->proc_net); 1427 net->sctp.proc_net_sctp = NULL; 1428 #endif 1429 cleanup_sctp_mibs(net); 1430 sctp_sysctl_net_unregister(net); 1431 } 1432 1433 static struct pernet_operations sctp_defaults_ops = { 1434 .init = sctp_defaults_init, 1435 .exit = sctp_defaults_exit, 1436 }; 1437 1438 static int __net_init sctp_ctrlsock_init(struct net *net) 1439 { 1440 int status; 1441 1442 /* Initialize the control inode/socket for handling OOTB packets. */ 1443 status = sctp_ctl_sock_init(net); 1444 if (status) 1445 pr_err("Failed to initialize the SCTP control sock\n"); 1446 1447 return status; 1448 } 1449 1450 static void __net_exit sctp_ctrlsock_exit(struct net *net) 1451 { 1452 /* Free the control endpoint. */ 1453 inet_ctl_sock_destroy(net->sctp.ctl_sock); 1454 } 1455 1456 static struct pernet_operations sctp_ctrlsock_ops = { 1457 .init = sctp_ctrlsock_init, 1458 .exit = sctp_ctrlsock_exit, 1459 }; 1460 1461 /* Initialize the universe into something sensible. */ 1462 static __init int sctp_init(void) 1463 { 1464 unsigned long nr_pages = totalram_pages(); 1465 unsigned long limit; 1466 unsigned long goal; 1467 int max_entry_order; 1468 int num_entries; 1469 int max_share; 1470 int status; 1471 int order; 1472 int i; 1473 1474 sock_skb_cb_check_size(sizeof(struct sctp_ulpevent)); 1475 1476 /* Allocate bind_bucket and chunk caches. */ 1477 status = -ENOBUFS; 1478 sctp_bucket_cachep = KMEM_CACHE(sctp_bind_bucket, SLAB_HWCACHE_ALIGN); 1479 if (!sctp_bucket_cachep) 1480 goto out; 1481 1482 sctp_chunk_cachep = KMEM_CACHE(sctp_chunk, SLAB_HWCACHE_ALIGN); 1483 if (!sctp_chunk_cachep) 1484 goto err_chunk_cachep; 1485 1486 status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL); 1487 if (status) 1488 goto err_percpu_counter_init; 1489 1490 /* Implementation specific variables. */ 1491 1492 /* Initialize default stream count setup information. */ 1493 sctp_max_instreams = SCTP_DEFAULT_INSTREAMS; 1494 sctp_max_outstreams = SCTP_DEFAULT_OUTSTREAMS; 1495 1496 /* Initialize handle used for association ids. */ 1497 idr_init(&sctp_assocs_id); 1498 1499 limit = nr_free_buffer_pages() / 8; 1500 limit = max(limit, 128UL); 1501 sysctl_sctp_mem[0] = limit / 4 * 3; 1502 sysctl_sctp_mem[1] = limit; 1503 sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2; 1504 1505 /* Set per-socket limits to no more than 1/128 the pressure threshold*/ 1506 limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7); 1507 max_share = min(4UL*1024*1024, limit); 1508 1509 sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */ 1510 sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1); 1511 sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share); 1512 1513 sysctl_sctp_wmem[0] = PAGE_SIZE; 1514 sysctl_sctp_wmem[1] = 16*1024; 1515 sysctl_sctp_wmem[2] = max(64*1024, max_share); 1516 1517 /* Size and allocate the association hash table. 1518 * The methodology is similar to that of the tcp hash tables. 1519 * Though not identical. Start by getting a goal size 1520 */ 1521 if (nr_pages >= (128 * 1024)) 1522 goal = nr_pages >> (22 - PAGE_SHIFT); 1523 else 1524 goal = nr_pages >> (24 - PAGE_SHIFT); 1525 1526 /* Then compute the page order for said goal */ 1527 order = get_order(goal); 1528 1529 /* Now compute the required page order for the maximum sized table we 1530 * want to create 1531 */ 1532 max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES * 1533 sizeof(struct sctp_bind_hashbucket)); 1534 1535 /* Limit the page order by that maximum hash table size */ 1536 order = min(order, max_entry_order); 1537 1538 /* Allocate and initialize the endpoint hash table. */ 1539 sctp_ep_hashsize = 64; 1540 sctp_ep_hashtable = 1541 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL); 1542 if (!sctp_ep_hashtable) { 1543 pr_err("Failed endpoint_hash alloc\n"); 1544 status = -ENOMEM; 1545 goto err_ehash_alloc; 1546 } 1547 for (i = 0; i < sctp_ep_hashsize; i++) { 1548 rwlock_init(&sctp_ep_hashtable[i].lock); 1549 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain); 1550 } 1551 1552 /* Allocate and initialize the SCTP port hash table. 1553 * Note that order is initalized to start at the max sized 1554 * table we want to support. If we can't get that many pages 1555 * reduce the order and try again 1556 */ 1557 do { 1558 sctp_port_hashtable = (struct sctp_bind_hashbucket *) 1559 __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order); 1560 } while (!sctp_port_hashtable && --order > 0); 1561 1562 if (!sctp_port_hashtable) { 1563 pr_err("Failed bind hash alloc\n"); 1564 status = -ENOMEM; 1565 goto err_bhash_alloc; 1566 } 1567 1568 /* Now compute the number of entries that will fit in the 1569 * port hash space we allocated 1570 */ 1571 num_entries = (1UL << order) * PAGE_SIZE / 1572 sizeof(struct sctp_bind_hashbucket); 1573 1574 /* And finish by rounding it down to the nearest power of two. 1575 * This wastes some memory of course, but it's needed because 1576 * the hash function operates based on the assumption that 1577 * the number of entries is a power of two. 1578 */ 1579 sctp_port_hashsize = rounddown_pow_of_two(num_entries); 1580 1581 for (i = 0; i < sctp_port_hashsize; i++) { 1582 spin_lock_init(&sctp_port_hashtable[i].lock); 1583 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain); 1584 } 1585 1586 status = sctp_transport_hashtable_init(); 1587 if (status) 1588 goto err_thash_alloc; 1589 1590 pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize, 1591 num_entries); 1592 1593 sctp_sysctl_register(); 1594 1595 INIT_LIST_HEAD(&sctp_address_families); 1596 sctp_v4_pf_init(); 1597 sctp_v6_pf_init(); 1598 sctp_sched_ops_init(); 1599 1600 status = register_pernet_subsys(&sctp_defaults_ops); 1601 if (status) 1602 goto err_register_defaults; 1603 1604 status = sctp_v4_protosw_init(); 1605 if (status) 1606 goto err_protosw_init; 1607 1608 status = sctp_v6_protosw_init(); 1609 if (status) 1610 goto err_v6_protosw_init; 1611 1612 status = register_pernet_subsys(&sctp_ctrlsock_ops); 1613 if (status) 1614 goto err_register_ctrlsock; 1615 1616 status = sctp_v4_add_protocol(); 1617 if (status) 1618 goto err_add_protocol; 1619 1620 /* Register SCTP with inet6 layer. */ 1621 status = sctp_v6_add_protocol(); 1622 if (status) 1623 goto err_v6_add_protocol; 1624 1625 if (sctp_offload_init() < 0) 1626 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__); 1627 1628 out: 1629 return status; 1630 err_v6_add_protocol: 1631 sctp_v4_del_protocol(); 1632 err_add_protocol: 1633 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1634 err_register_ctrlsock: 1635 sctp_v6_protosw_exit(); 1636 err_v6_protosw_init: 1637 sctp_v4_protosw_exit(); 1638 err_protosw_init: 1639 unregister_pernet_subsys(&sctp_defaults_ops); 1640 err_register_defaults: 1641 sctp_v4_pf_exit(); 1642 sctp_v6_pf_exit(); 1643 sctp_sysctl_unregister(); 1644 free_pages((unsigned long)sctp_port_hashtable, 1645 get_order(sctp_port_hashsize * 1646 sizeof(struct sctp_bind_hashbucket))); 1647 err_bhash_alloc: 1648 sctp_transport_hashtable_destroy(); 1649 err_thash_alloc: 1650 kfree(sctp_ep_hashtable); 1651 err_ehash_alloc: 1652 percpu_counter_destroy(&sctp_sockets_allocated); 1653 err_percpu_counter_init: 1654 kmem_cache_destroy(sctp_chunk_cachep); 1655 err_chunk_cachep: 1656 kmem_cache_destroy(sctp_bucket_cachep); 1657 goto out; 1658 } 1659 1660 /* Exit handler for the SCTP protocol. */ 1661 static __exit void sctp_exit(void) 1662 { 1663 /* BUG. This should probably do something useful like clean 1664 * up all the remaining associations and all that memory. 1665 */ 1666 1667 /* Unregister with inet6/inet layers. */ 1668 sctp_v6_del_protocol(); 1669 sctp_v4_del_protocol(); 1670 1671 unregister_pernet_subsys(&sctp_ctrlsock_ops); 1672 1673 /* Free protosw registrations */ 1674 sctp_v6_protosw_exit(); 1675 sctp_v4_protosw_exit(); 1676 1677 unregister_pernet_subsys(&sctp_defaults_ops); 1678 1679 /* Unregister with socket layer. */ 1680 sctp_v6_pf_exit(); 1681 sctp_v4_pf_exit(); 1682 1683 sctp_sysctl_unregister(); 1684 1685 free_pages((unsigned long)sctp_port_hashtable, 1686 get_order(sctp_port_hashsize * 1687 sizeof(struct sctp_bind_hashbucket))); 1688 kfree(sctp_ep_hashtable); 1689 sctp_transport_hashtable_destroy(); 1690 1691 percpu_counter_destroy(&sctp_sockets_allocated); 1692 1693 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1694 1695 kmem_cache_destroy(sctp_chunk_cachep); 1696 kmem_cache_destroy(sctp_bucket_cachep); 1697 } 1698 1699 module_init(sctp_init); 1700 module_exit(sctp_exit); 1701 1702 /* 1703 * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly. 1704 */ 1705 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132"); 1706 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132"); 1707 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>"); 1708 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)"); 1709 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644); 1710 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification"); 1711 MODULE_LICENSE("GPL"); 1712