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