net/sctp/protocol.c

   1 /* SCTP kernel implementation
   2  * (C) Copyright IBM Corp. 2001, 2004
   3  * Copyright (c) 1999-2000 Cisco, Inc.
   4  * Copyright (c) 1999-2001 Motorola, Inc.
   5  * Copyright (c) 2001 Intel Corp.
   6  * Copyright (c) 2001 Nokia, Inc.
   7  * Copyright (c) 2001 La Monte H.P. Yarroll
   8  *
   9  * This file is part of the SCTP kernel implementation
  10  *
  11  * Initialization/cleanup for SCTP protocol support.
  12  *
  13  * This SCTP implementation is free software;
  14  * you can redistribute it and/or modify it under the terms of
  15  * the GNU General Public License as published by
  16  * the Free Software Foundation; either version 2, or (at your option)
  17  * any later version.
  18  *
  19  * This SCTP implementation is distributed in the hope that it
  20  * will be useful, but WITHOUT ANY WARRANTY; without even the implied
  21  *                 ************************
  22  * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  23  * See the GNU General Public License for more details.
  24  *
  25  * You should have received a copy of the GNU General Public License
  26  * along with GNU CC; see the file COPYING.  If not, see
  27  * <http://www.gnu.org/licenses/>.
  28  *
  29  * Please send any bug reports or fixes you make to the
  30  * email address(es):
  31  *    lksctp developers <linux-sctp@vger.kernel.org>
  32  *
  33  * Written or modified by:
  34  *    La Monte H.P. Yarroll <piggy@acm.org>
  35  *    Karl Knutson <karl@athena.chicago.il.us>
  36  *    Jon Grimm <jgrimm@us.ibm.com>
  37  *    Sridhar Samudrala <sri@us.ibm.com>
  38  *    Daisy Chang <daisyc@us.ibm.com>
  39  *    Ardelle Fan <ardelle.fan@intel.com>
  40  */
  41
  42 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  43
  44 #include <linux/module.h>
  45 #include <linux/init.h>
  46 #include <linux/netdevice.h>
  47 #include <linux/inetdevice.h>
  48 #include <linux/seq_file.h>
  49 #include <linux/bootmem.h>
  50 #include <linux/highmem.h>
  51 #include <linux/swap.h>
  52 #include <linux/slab.h>
  53 #include <net/net_namespace.h>
  54 #include <net/protocol.h>
  55 #include <net/ip.h>
  56 #include <net/ipv6.h>
  57 #include <net/route.h>
  58 #include <net/sctp/sctp.h>
  59 #include <net/addrconf.h>
  60 #include <net/inet_common.h>
  61 #include <net/inet_ecn.h>
  62
  63 /* Global data structures. */
  64 struct sctp_globals sctp_globals __read_mostly;
  65
  66 struct idr sctp_assocs_id;
  67 DEFINE_SPINLOCK(sctp_assocs_id_lock);
  68
  69 static struct sctp_pf *sctp_pf_inet6_specific;
  70 static struct sctp_pf *sctp_pf_inet_specific;
  71 static struct sctp_af *sctp_af_v4_specific;
  72 static struct sctp_af *sctp_af_v6_specific;
  73
  74 struct kmem_cache *sctp_chunk_cachep __read_mostly;
  75 struct kmem_cache *sctp_bucket_cachep __read_mostly;
  76
  77 long sysctl_sctp_mem[3];
  78 int sysctl_sctp_rmem[3];
  79 int sysctl_sctp_wmem[3];
  80
  81 /* Set up the proc fs entry for the SCTP protocol. */
  82 static int __net_init sctp_proc_init(struct net *net)
  83 {
  84 #ifdef CONFIG_PROC_FS
  85         net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net);
  86         if (!net->sctp.proc_net_sctp)
  87                 goto out_proc_net_sctp;
  88         if (sctp_snmp_proc_init(net))
  89                 goto out_snmp_proc_init;
  90         if (sctp_eps_proc_init(net))
  91                 goto out_eps_proc_init;
  92         if (sctp_assocs_proc_init(net))
  93                 goto out_assocs_proc_init;
  94         if (sctp_remaddr_proc_init(net))
  95                 goto out_remaddr_proc_init;
  96
  97         return 0;
  98
  99 out_remaddr_proc_init:
 100         sctp_assocs_proc_exit(net);
 101 out_assocs_proc_init:
 102         sctp_eps_proc_exit(net);
 103 out_eps_proc_init:
 104         sctp_snmp_proc_exit(net);
 105 out_snmp_proc_init:
 106         remove_proc_entry("sctp", net->proc_net);
 107         net->sctp.proc_net_sctp = NULL;
 108 out_proc_net_sctp:
 109         return -ENOMEM;
 110 #endif /* CONFIG_PROC_FS */
 111         return 0;
 112 }
 113
 114 /* Clean up the proc fs entry for the SCTP protocol.
 115  * Note: Do not make this __exit as it is used in the init error
 116  * path.
 117  */
 118 static void sctp_proc_exit(struct net *net)
 119 {
 120 #ifdef CONFIG_PROC_FS
 121         sctp_snmp_proc_exit(net);
 122         sctp_eps_proc_exit(net);
 123         sctp_assocs_proc_exit(net);
 124         sctp_remaddr_proc_exit(net);
 125
 126         remove_proc_entry("sctp", net->proc_net);
 127         net->sctp.proc_net_sctp = NULL;
 128 #endif
 129 }
 130
 131 /* Private helper to extract ipv4 address and stash them in
 132  * the protocol structure.
 133  */
 134 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
 135                                   struct net_device *dev)
 136 {
 137         struct in_device *in_dev;
 138         struct in_ifaddr *ifa;
 139         struct sctp_sockaddr_entry *addr;
 140
 141         rcu_read_lock();
 142         if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
 143                 rcu_read_unlock();
 144                 return;
 145         }
 146
 147         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
 148                 /* Add the address to the local list.  */
 149                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
 150                 if (addr) {
 151                         addr->a.v4.sin_family = AF_INET;
 152                         addr->a.v4.sin_port = 0;
 153                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 154                         addr->valid = 1;
 155                         INIT_LIST_HEAD(&addr->list);
 156                         list_add_tail(&addr->list, addrlist);
 157                 }
 158         }
 159
 160         rcu_read_unlock();
 161 }
 162
 163 /* Extract our IP addresses from the system and stash them in the
 164  * protocol structure.
 165  */
 166 static void sctp_get_local_addr_list(struct net *net)
 167 {
 168         struct net_device *dev;
 169         struct list_head *pos;
 170         struct sctp_af *af;
 171
 172         rcu_read_lock();
 173         for_each_netdev_rcu(net, dev) {
 174                 list_for_each(pos, &sctp_address_families) {
 175                         af = list_entry(pos, struct sctp_af, list);
 176                         af->copy_addrlist(&net->sctp.local_addr_list, dev);
 177                 }
 178         }
 179         rcu_read_unlock();
 180 }
 181
 182 /* Free the existing local addresses.  */
 183 static void sctp_free_local_addr_list(struct net *net)
 184 {
 185         struct sctp_sockaddr_entry *addr;
 186         struct list_head *pos, *temp;
 187
 188         list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
 189                 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
 190                 list_del(pos);
 191                 kfree(addr);
 192         }
 193 }
 194
 195 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
 196 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
 197                               sctp_scope_t scope, gfp_t gfp, int copy_flags)
 198 {
 199         struct sctp_sockaddr_entry *addr;
 200         int error = 0;
 201
 202         rcu_read_lock();
 203         list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
 204                 if (!addr->valid)
 205                         continue;
 206                 if (sctp_in_scope(net, &addr->a, scope)) {
 207                         /* Now that the address is in scope, check to see if
 208                          * the address type is really supported by the local
 209                          * sock as well as the remote peer.
 210                          */
 211                         if ((((AF_INET == addr->a.sa.sa_family) &&
 212                               (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
 213                             (((AF_INET6 == addr->a.sa.sa_family) &&
 214                               (copy_flags & SCTP_ADDR6_ALLOWED) &&
 215                               (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
 216                                 error = sctp_add_bind_addr(bp, &addr->a,
 217                                                     SCTP_ADDR_SRC, GFP_ATOMIC);
 218                                 if (error)
 219                                         goto end_copy;
 220                         }
 221                 }
 222         }
 223
 224 end_copy:
 225         rcu_read_unlock();
 226         return error;
 227 }
 228
 229 /* Initialize a sctp_addr from in incoming skb.  */
 230 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
 231                              int is_saddr)
 232 {
 233         void *from;
 234         __be16 *port;
 235         struct sctphdr *sh;
 236
 237         port = &addr->v4.sin_port;
 238         addr->v4.sin_family = AF_INET;
 239
 240         sh = sctp_hdr(skb);
 241         if (is_saddr) {
 242                 *port  = sh->source;
 243                 from = &ip_hdr(skb)->saddr;
 244         } else {
 245                 *port = sh->dest;
 246                 from = &ip_hdr(skb)->daddr;
 247         }
 248         memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
 249 }
 250
 251 /* Initialize an sctp_addr from a socket. */
 252 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
 253 {
 254         addr->v4.sin_family = AF_INET;
 255         addr->v4.sin_port = 0;
 256         addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
 257 }
 258
 259 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
 260 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
 261 {
 262         inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
 263 }
 264
 265 /* Initialize sk->sk_daddr from sctp_addr. */
 266 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
 267 {
 268         inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
 269 }
 270
 271 /* Initialize a sctp_addr from an address parameter. */
 272 static void sctp_v4_from_addr_param(union sctp_addr *addr,
 273                                     union sctp_addr_param *param,
 274                                     __be16 port, int iif)
 275 {
 276         addr->v4.sin_family = AF_INET;
 277         addr->v4.sin_port = port;
 278         addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
 279 }
 280
 281 /* Initialize an address parameter from a sctp_addr and return the length
 282  * of the address parameter.
 283  */
 284 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
 285                                  union sctp_addr_param *param)
 286 {
 287         int length = sizeof(sctp_ipv4addr_param_t);
 288
 289         param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
 290         param->v4.param_hdr.length = htons(length);
 291         param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
 292
 293         return length;
 294 }
 295
 296 /* Initialize a sctp_addr from a dst_entry. */
 297 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
 298                               __be16 port)
 299 {
 300         saddr->v4.sin_family = AF_INET;
 301         saddr->v4.sin_port = port;
 302         saddr->v4.sin_addr.s_addr = fl4->saddr;
 303 }
 304
 305 /* Compare two addresses exactly. */
 306 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
 307                             const union sctp_addr *addr2)
 308 {
 309         if (addr1->sa.sa_family != addr2->sa.sa_family)
 310                 return 0;
 311         if (addr1->v4.sin_port != addr2->v4.sin_port)
 312                 return 0;
 313         if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
 314                 return 0;
 315
 316         return 1;
 317 }
 318
 319 /* Initialize addr struct to INADDR_ANY. */
 320 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
 321 {
 322         addr->v4.sin_family = AF_INET;
 323         addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
 324         addr->v4.sin_port = port;
 325 }
 326
 327 /* Is this a wildcard address? */
 328 static int sctp_v4_is_any(const union sctp_addr *addr)
 329 {
 330         return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
 331 }
 332
 333 /* This function checks if the address is a valid address to be used for
 334  * SCTP binding.
 335  *
 336  * Output:
 337  * Return 0 - If the address is a non-unicast or an illegal address.
 338  * Return 1 - If the address is a unicast.
 339  */
 340 static int sctp_v4_addr_valid(union sctp_addr *addr,
 341                               struct sctp_sock *sp,
 342                               const struct sk_buff *skb)
 343 {
 344         /* IPv4 addresses not allowed */
 345         if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
 346                 return 0;
 347
 348         /* Is this a non-unicast address or a unusable SCTP address? */
 349         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
 350                 return 0;
 351
 352         /* Is this a broadcast address? */
 353         if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
 354                 return 0;
 355
 356         return 1;
 357 }
 358
 359 /* Should this be available for binding?   */
 360 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
 361 {
 362         struct net *net = sock_net(&sp->inet.sk);
 363         int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
 364
 365
 366         if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
 367            ret != RTN_LOCAL &&
 368            !sp->inet.freebind &&
 369            !net->ipv4.sysctl_ip_nonlocal_bind)
 370                 return 0;
 371
 372         if (ipv6_only_sock(sctp_opt2sk(sp)))
 373                 return 0;
 374
 375         return 1;
 376 }
 377
 378 /* Checking the loopback, private and other address scopes as defined in
 379  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
 380  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
 381  *
 382  * Level 0 - unusable SCTP addresses
 383  * Level 1 - loopback address
 384  * Level 2 - link-local addresses
 385  * Level 3 - private addresses.
 386  * Level 4 - global addresses
 387  * For INIT and INIT-ACK address list, let L be the level of
 388  * of requested destination address, sender and receiver
 389  * SHOULD include all of its addresses with level greater
 390  * than or equal to L.
 391  *
 392  * IPv4 scoping can be controlled through sysctl option
 393  * net.sctp.addr_scope_policy
 394  */
 395 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
 396 {
 397         sctp_scope_t retval;
 398
 399         /* Check for unusable SCTP addresses. */
 400         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
 401                 retval =  SCTP_SCOPE_UNUSABLE;
 402         } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
 403                 retval = SCTP_SCOPE_LOOPBACK;
 404         } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
 405                 retval = SCTP_SCOPE_LINK;
 406         } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
 407                    ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
 408                    ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
 409                 retval = SCTP_SCOPE_PRIVATE;
 410         } else {
 411                 retval = SCTP_SCOPE_GLOBAL;
 412         }
 413
 414         return retval;
 415 }
 416
 417 /* Returns a valid dst cache entry for the given source and destination ip
 418  * addresses. If an association is passed, trys to get a dst entry with a
 419  * source address that matches an address in the bind address list.
 420  */
 421 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
 422                                 struct flowi *fl, struct sock *sk)
 423 {
 424         struct sctp_association *asoc = t->asoc;
 425         struct rtable *rt;
 426         struct flowi4 *fl4 = &fl->u.ip4;
 427         struct sctp_bind_addr *bp;
 428         struct sctp_sockaddr_entry *laddr;
 429         struct dst_entry *dst = NULL;
 430         union sctp_addr *daddr = &t->ipaddr;
 431         union sctp_addr dst_saddr;
 432
 433         memset(fl4, 0x0, sizeof(struct flowi4));
 434         fl4->daddr  = daddr->v4.sin_addr.s_addr;
 435         fl4->fl4_dport = daddr->v4.sin_port;
 436         fl4->flowi4_proto = IPPROTO_SCTP;
 437         if (asoc) {
 438                 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
 439                 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
 440                 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
 441         }
 442         if (saddr) {
 443                 fl4->saddr = saddr->v4.sin_addr.s_addr;
 444                 fl4->fl4_sport = saddr->v4.sin_port;
 445         }
 446
 447         pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
 448                  &fl4->saddr);
 449
 450         rt = ip_route_output_key(sock_net(sk), fl4);
 451         if (!IS_ERR(rt))
 452                 dst = &rt->dst;
 453
 454         /* If there is no association or if a source address is passed, no
 455          * more validation is required.
 456          */
 457         if (!asoc || saddr)
 458                 goto out;
 459
 460         bp = &asoc->base.bind_addr;
 461
 462         if (dst) {
 463                 /* Walk through the bind address list and look for a bind
 464                  * address that matches the source address of the returned dst.
 465                  */
 466                 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
 467                 rcu_read_lock();
 468                 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 469                         if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
 470                             (laddr->state != SCTP_ADDR_SRC &&
 471                             !asoc->src_out_of_asoc_ok))
 472                                 continue;
 473                         if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
 474                                 goto out_unlock;
 475                 }
 476                 rcu_read_unlock();
 477
 478                 /* None of the bound addresses match the source address of the
 479                  * dst. So release it.
 480                  */
 481                 dst_release(dst);
 482                 dst = NULL;
 483         }
 484
 485         /* Walk through the bind address list and try to get a dst that
 486          * matches a bind address as the source address.
 487          */
 488         rcu_read_lock();
 489         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 490                 struct net_device *odev;
 491
 492                 if (!laddr->valid)
 493                         continue;
 494                 if (laddr->state != SCTP_ADDR_SRC ||
 495                     AF_INET != laddr->a.sa.sa_family)
 496                         continue;
 497
 498                 fl4->fl4_sport = laddr->a.v4.sin_port;
 499                 flowi4_update_output(fl4,
 500                                      asoc->base.sk->sk_bound_dev_if,
 501                                      RT_CONN_FLAGS(asoc->base.sk),
 502                                      daddr->v4.sin_addr.s_addr,
 503                                      laddr->a.v4.sin_addr.s_addr);
 504
 505                 rt = ip_route_output_key(sock_net(sk), fl4);
 506                 if (IS_ERR(rt))
 507                         continue;
 508
 509                 if (!dst)
 510                         dst = &rt->dst;
 511
 512                 /* Ensure the src address belongs to the output
 513                  * interface.
 514                  */
 515                 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
 516                                      false);
 517                 if (!odev || odev->ifindex != fl4->flowi4_oif) {
 518                         if (&rt->dst != dst)
 519                                 dst_release(&rt->dst);
 520                         continue;
 521                 }
 522
 523                 if (dst != &rt->dst)
 524                         dst_release(dst);
 525                 dst = &rt->dst;
 526                 break;
 527         }
 528
 529 out_unlock:
 530         rcu_read_unlock();
 531 out:
 532         t->dst = dst;
 533         if (dst)
 534                 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
 535                          &fl4->daddr, &fl4->saddr);
 536         else
 537                 pr_debug("no route\n");
 538 }
 539
 540 /* For v4, the source address is cached in the route entry(dst). So no need
 541  * to cache it separately and hence this is an empty routine.
 542  */
 543 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 544                               struct sctp_transport *t,
 545                               struct flowi *fl)
 546 {
 547         union sctp_addr *saddr = &t->saddr;
 548         struct rtable *rt = (struct rtable *)t->dst;
 549
 550         if (rt) {
 551                 saddr->v4.sin_family = AF_INET;
 552                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 553         }
 554 }
 555
 556 /* What interface did this skb arrive on? */
 557 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 558 {
 559         return inet_iif(skb);
 560 }
 561
 562 /* Was this packet marked by Explicit Congestion Notification? */
 563 static int sctp_v4_is_ce(const struct sk_buff *skb)
 564 {
 565         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 566 }
 567
 568 /* Create and initialize a new sk for the socket returned by accept(). */
 569 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 570                                              struct sctp_association *asoc)
 571 {
 572         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 573                         sk->sk_prot, 0);
 574         struct inet_sock *newinet;
 575
 576         if (!newsk)
 577                 goto out;
 578
 579         sock_init_data(NULL, newsk);
 580
 581         sctp_copy_sock(newsk, sk, asoc);
 582         sock_reset_flag(newsk, SOCK_ZAPPED);
 583
 584         newinet = inet_sk(newsk);
 585
 586         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 587
 588         sk_refcnt_debug_inc(newsk);
 589
 590         if (newsk->sk_prot->init(newsk)) {
 591                 sk_common_release(newsk);
 592                 newsk = NULL;
 593         }
 594
 595 out:
 596         return newsk;
 597 }
 598
 599 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
 600 {
 601         /* No address mapping for V4 sockets */
 602         return sizeof(struct sockaddr_in);
 603 }
 604
 605 /* Dump the v4 addr to the seq file. */
 606 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 607 {
 608         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 609 }
 610
 611 static void sctp_v4_ecn_capable(struct sock *sk)
 612 {
 613         INET_ECN_xmit(sk);
 614 }
 615
 616 static void sctp_addr_wq_timeout_handler(unsigned long arg)
 617 {
 618         struct net *net = (struct net *)arg;
 619         struct sctp_sockaddr_entry *addrw, *temp;
 620         struct sctp_sock *sp;
 621
 622         spin_lock_bh(&net->sctp.addr_wq_lock);
 623
 624         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 625                 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
 626                          "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
 627                          addrw->state, addrw);
 628
 629 #if IS_ENABLED(CONFIG_IPV6)
 630                 /* Now we send an ASCONF for each association */
 631                 /* Note. we currently don't handle link local IPv6 addressees */
 632                 if (addrw->a.sa.sa_family == AF_INET6) {
 633                         struct in6_addr *in6;
 634
 635                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 636                             IPV6_ADDR_LINKLOCAL)
 637                                 goto free_next;
 638
 639                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 640                         if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
 641                             addrw->state == SCTP_ADDR_NEW) {
 642                                 unsigned long timeo_val;
 643
 644                                 pr_debug("%s: this is on DAD, trying %d sec "
 645                                          "later\n", __func__,
 646                                          SCTP_ADDRESS_TICK_DELAY);
 647
 648                                 timeo_val = jiffies;
 649                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 650                                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 651                                 break;
 652                         }
 653                 }
 654 #endif
 655                 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
 656                         struct sock *sk;
 657
 658                         sk = sctp_opt2sk(sp);
 659                         /* ignore bound-specific endpoints */
 660                         if (!sctp_is_ep_boundall(sk))
 661                                 continue;
 662                         bh_lock_sock(sk);
 663                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 664                                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
 665                         bh_unlock_sock(sk);
 666                 }
 667 #if IS_ENABLED(CONFIG_IPV6)
 668 free_next:
 669 #endif
 670                 list_del(&addrw->list);
 671                 kfree(addrw);
 672         }
 673         spin_unlock_bh(&net->sctp.addr_wq_lock);
 674 }
 675
 676 static void sctp_free_addr_wq(struct net *net)
 677 {
 678         struct sctp_sockaddr_entry *addrw;
 679         struct sctp_sockaddr_entry *temp;
 680
 681         spin_lock_bh(&net->sctp.addr_wq_lock);
 682         del_timer(&net->sctp.addr_wq_timer);
 683         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 684                 list_del(&addrw->list);
 685                 kfree(addrw);
 686         }
 687         spin_unlock_bh(&net->sctp.addr_wq_lock);
 688 }
 689
 690 /* lookup the entry for the same address in the addr_waitq
 691  * sctp_addr_wq MUST be locked
 692  */
 693 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
 694                                         struct sctp_sockaddr_entry *addr)
 695 {
 696         struct sctp_sockaddr_entry *addrw;
 697
 698         list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
 699                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 700                         continue;
 701                 if (addrw->a.sa.sa_family == AF_INET) {
 702                         if (addrw->a.v4.sin_addr.s_addr ==
 703                             addr->a.v4.sin_addr.s_addr)
 704                                 return addrw;
 705                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 706                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 707                             &addr->a.v6.sin6_addr))
 708                                 return addrw;
 709                 }
 710         }
 711         return NULL;
 712 }
 713
 714 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
 715 {
 716         struct sctp_sockaddr_entry *addrw;
 717         unsigned long timeo_val;
 718
 719         /* first, we check if an opposite message already exist in the queue.
 720          * If we found such message, it is removed.
 721          * This operation is a bit stupid, but the DHCP client attaches the
 722          * new address after a couple of addition and deletion of that address
 723          */
 724
 725         spin_lock_bh(&net->sctp.addr_wq_lock);
 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 = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 779                 if (addr) {
 780                         addr->a.v4.sin_family = AF_INET;
 781                         addr->a.v4.sin_port = 0;
 782                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 783                         addr->valid = 1;
 784                         spin_lock_bh(&net->sctp.local_addr_lock);
 785                         list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
 786                         sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
 787                         spin_unlock_bh(&net->sctp.local_addr_lock);
 788                 }
 789                 break;
 790         case NETDEV_DOWN:
 791                 spin_lock_bh(&net->sctp.local_addr_lock);
 792                 list_for_each_entry_safe(addr, temp,
 793                                         &net->sctp.local_addr_list, list) {
 794                         if (addr->a.sa.sa_family == AF_INET &&
 795                                         addr->a.v4.sin_addr.s_addr ==
 796                                         ifa->ifa_local) {
 797                                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
 798                                 found = 1;
 799                                 addr->valid = 0;
 800                                 list_del_rcu(&addr->list);
 801                                 break;
 802                         }
 803                 }
 804                 spin_unlock_bh(&net->sctp.local_addr_lock);
 805                 if (found)
 806                         kfree_rcu(addr, rcu);
 807                 break;
 808         }
 809
 810         return NOTIFY_DONE;
 811 }
 812
 813 /*
 814  * Initialize the control inode/socket with a control endpoint data
 815  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 816  */
 817 static int sctp_ctl_sock_init(struct net *net)
 818 {
 819         int err;
 820         sa_family_t family = PF_INET;
 821
 822         if (sctp_get_pf_specific(PF_INET6))
 823                 family = PF_INET6;
 824
 825         err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
 826                                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
 827
 828         /* If IPv6 socket could not be created, try the IPv4 socket */
 829         if (err < 0 && family == PF_INET6)
 830                 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
 831                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 832                                            net);
 833
 834         if (err < 0) {
 835                 pr_err("Failed to create the SCTP control socket\n");
 836                 return err;
 837         }
 838         return 0;
 839 }
 840
 841 /* Register address family specific functions. */
 842 int sctp_register_af(struct sctp_af *af)
 843 {
 844         switch (af->sa_family) {
 845         case AF_INET:
 846                 if (sctp_af_v4_specific)
 847                         return 0;
 848                 sctp_af_v4_specific = af;
 849                 break;
 850         case AF_INET6:
 851                 if (sctp_af_v6_specific)
 852                         return 0;
 853                 sctp_af_v6_specific = af;
 854                 break;
 855         default:
 856                 return 0;
 857         }
 858
 859         INIT_LIST_HEAD(&af->list);
 860         list_add_tail(&af->list, &sctp_address_families);
 861         return 1;
 862 }
 863
 864 /* Get the table of functions for manipulating a particular address
 865  * family.
 866  */
 867 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 868 {
 869         switch (family) {
 870         case AF_INET:
 871                 return sctp_af_v4_specific;
 872         case AF_INET6:
 873                 return sctp_af_v6_specific;
 874         default:
 875                 return NULL;
 876         }
 877 }
 878
 879 /* Common code to initialize a AF_INET msg_name. */
 880 static void sctp_inet_msgname(char *msgname, int *addr_len)
 881 {
 882         struct sockaddr_in *sin;
 883
 884         sin = (struct sockaddr_in *)msgname;
 885         *addr_len = sizeof(struct sockaddr_in);
 886         sin->sin_family = AF_INET;
 887         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 888 }
 889
 890 /* Copy the primary address of the peer primary address as the msg_name. */
 891 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 892                                     int *addr_len)
 893 {
 894         struct sockaddr_in *sin, *sinfrom;
 895
 896         if (msgname) {
 897                 struct sctp_association *asoc;
 898
 899                 asoc = event->asoc;
 900                 sctp_inet_msgname(msgname, addr_len);
 901                 sin = (struct sockaddr_in *)msgname;
 902                 sinfrom = &asoc->peer.primary_addr.v4;
 903                 sin->sin_port = htons(asoc->peer.port);
 904                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 905         }
 906 }
 907
 908 /* Initialize and copy out a msgname from an inbound skb. */
 909 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 910 {
 911         if (msgname) {
 912                 struct sctphdr *sh = sctp_hdr(skb);
 913                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 914
 915                 sctp_inet_msgname(msgname, len);
 916                 sin->sin_port = sh->source;
 917                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 918         }
 919 }
 920
 921 /* Do we support this AF? */
 922 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 923 {
 924         /* PF_INET only supports AF_INET addresses. */
 925         return AF_INET == family;
 926 }
 927
 928 /* Address matching with wildcards allowed. */
 929 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 930                               const union sctp_addr *addr2,
 931                               struct sctp_sock *opt)
 932 {
 933         /* PF_INET only supports AF_INET addresses. */
 934         if (addr1->sa.sa_family != addr2->sa.sa_family)
 935                 return 0;
 936         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 937             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 938                 return 1;
 939         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 940                 return 1;
 941
 942         return 0;
 943 }
 944
 945 /* Verify that provided sockaddr looks bindable.  Common verification has
 946  * already been taken care of.
 947  */
 948 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 949 {
 950         return sctp_v4_available(addr, opt);
 951 }
 952
 953 /* Verify that sockaddr looks sendable.  Common verification has already
 954  * been taken care of.
 955  */
 956 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 957 {
 958         return 1;
 959 }
 960
 961 /* Fill in Supported Address Type information for INIT and INIT-ACK
 962  * chunks.  Returns number of addresses supported.
 963  */
 964 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 965                                      __be16 *types)
 966 {
 967         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 968         return 1;
 969 }
 970
 971 /* Wrapper routine that calls the ip transmit routine. */
 972 static inline int sctp_v4_xmit(struct sk_buff *skb,
 973                                struct sctp_transport *transport)
 974 {
 975         struct inet_sock *inet = inet_sk(skb->sk);
 976
 977         pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
 978                  skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
 979
 980         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
 981                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 982
 983         SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
 984
 985         return ip_queue_xmit(&inet->sk, skb, &transport->fl);
 986 }
 987
 988 static struct sctp_af sctp_af_inet;
 989
 990 static struct sctp_pf sctp_pf_inet = {
 991         .event_msgname = sctp_inet_event_msgname,
 992         .skb_msgname   = sctp_inet_skb_msgname,
 993         .af_supported  = sctp_inet_af_supported,
 994         .cmp_addr      = sctp_inet_cmp_addr,
 995         .bind_verify   = sctp_inet_bind_verify,
 996         .send_verify   = sctp_inet_send_verify,
 997         .supported_addrs = sctp_inet_supported_addrs,
 998         .create_accept_sk = sctp_v4_create_accept_sk,
 999         .addr_to_user  = sctp_v4_addr_to_user,
1000         .to_sk_saddr   = sctp_v4_to_sk_saddr,
1001         .to_sk_daddr   = sctp_v4_to_sk_daddr,
1002         .af            = &sctp_af_inet
1003 };
1004
1005 /* Notifier for inetaddr addition/deletion events.  */
1006 static struct notifier_block sctp_inetaddr_notifier = {
1007         .notifier_call = sctp_inetaddr_event,
1008 };
1009
1010 /* Socket operations.  */
1011 static const struct proto_ops inet_seqpacket_ops = {
1012         .family            = PF_INET,
1013         .owner             = THIS_MODULE,
1014         .release           = inet_release,      /* Needs to be wrapped... */
1015         .bind              = inet_bind,
1016         .connect           = inet_dgram_connect,
1017         .socketpair        = sock_no_socketpair,
1018         .accept            = inet_accept,
1019         .getname           = inet_getname,      /* Semantics are different.  */
1020         .poll              = sctp_poll,
1021         .ioctl             = inet_ioctl,
1022         .listen            = sctp_inet_listen,
1023         .shutdown          = inet_shutdown,     /* Looks harmless.  */
1024         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1025         .getsockopt        = sock_common_getsockopt,
1026         .sendmsg           = inet_sendmsg,
1027         .recvmsg           = sock_common_recvmsg,
1028         .mmap              = sock_no_mmap,
1029         .sendpage          = sock_no_sendpage,
1030 #ifdef CONFIG_COMPAT
1031         .compat_setsockopt = compat_sock_common_setsockopt,
1032         .compat_getsockopt = compat_sock_common_getsockopt,
1033 #endif
1034 };
1035
1036 /* Registration with AF_INET family.  */
1037 static struct inet_protosw sctp_seqpacket_protosw = {
1038         .type       = SOCK_SEQPACKET,
1039         .protocol   = IPPROTO_SCTP,
1040         .prot       = &sctp_prot,
1041         .ops        = &inet_seqpacket_ops,
1042         .flags      = SCTP_PROTOSW_FLAG
1043 };
1044 static struct inet_protosw sctp_stream_protosw = {
1045         .type       = SOCK_STREAM,
1046         .protocol   = IPPROTO_SCTP,
1047         .prot       = &sctp_prot,
1048         .ops        = &inet_seqpacket_ops,
1049         .flags      = SCTP_PROTOSW_FLAG
1050 };
1051
1052 /* Register with IP layer.  */
1053 static const struct net_protocol sctp_protocol = {
1054         .handler     = sctp_rcv,
1055         .err_handler = sctp_v4_err,
1056         .no_policy   = 1,
1057         .netns_ok    = 1,
1058         .icmp_strict_tag_validation = 1,
1059 };
1060
1061 /* IPv4 address related functions.  */
1062 static struct sctp_af sctp_af_inet = {
1063         .sa_family         = AF_INET,
1064         .sctp_xmit         = sctp_v4_xmit,
1065         .setsockopt        = ip_setsockopt,
1066         .getsockopt        = ip_getsockopt,
1067         .get_dst           = sctp_v4_get_dst,
1068         .get_saddr         = sctp_v4_get_saddr,
1069         .copy_addrlist     = sctp_v4_copy_addrlist,
1070         .from_skb          = sctp_v4_from_skb,
1071         .from_sk           = sctp_v4_from_sk,
1072         .from_addr_param   = sctp_v4_from_addr_param,
1073         .to_addr_param     = sctp_v4_to_addr_param,
1074         .cmp_addr          = sctp_v4_cmp_addr,
1075         .addr_valid        = sctp_v4_addr_valid,
1076         .inaddr_any        = sctp_v4_inaddr_any,
1077         .is_any            = sctp_v4_is_any,
1078         .available         = sctp_v4_available,
1079         .scope             = sctp_v4_scope,
1080         .skb_iif           = sctp_v4_skb_iif,
1081         .is_ce             = sctp_v4_is_ce,
1082         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1083         .ecn_capable       = sctp_v4_ecn_capable,
1084         .net_header_len    = sizeof(struct iphdr),
1085         .sockaddr_len      = sizeof(struct sockaddr_in),
1086 #ifdef CONFIG_COMPAT
1087         .compat_setsockopt = compat_ip_setsockopt,
1088         .compat_getsockopt = compat_ip_getsockopt,
1089 #endif
1090 };
1091
1092 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1093 {
1094         switch (family) {
1095         case PF_INET:
1096                 return sctp_pf_inet_specific;
1097         case PF_INET6:
1098                 return sctp_pf_inet6_specific;
1099         default:
1100                 return NULL;
1101         }
1102 }
1103
1104 /* Register the PF specific function table.  */
1105 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1106 {
1107         switch (family) {
1108         case PF_INET:
1109                 if (sctp_pf_inet_specific)
1110                         return 0;
1111                 sctp_pf_inet_specific = pf;
1112                 break;
1113         case PF_INET6:
1114                 if (sctp_pf_inet6_specific)
1115                         return 0;
1116                 sctp_pf_inet6_specific = pf;
1117                 break;
1118         default:
1119                 return 0;
1120         }
1121         return 1;
1122 }
1123
1124 static inline int init_sctp_mibs(struct net *net)
1125 {
1126         net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1127         if (!net->sctp.sctp_statistics)
1128                 return -ENOMEM;
1129         return 0;
1130 }
1131
1132 static inline void cleanup_sctp_mibs(struct net *net)
1133 {
1134         free_percpu(net->sctp.sctp_statistics);
1135 }
1136
1137 static void sctp_v4_pf_init(void)
1138 {
1139         /* Initialize the SCTP specific PF functions. */
1140         sctp_register_pf(&sctp_pf_inet, PF_INET);
1141         sctp_register_af(&sctp_af_inet);
1142 }
1143
1144 static void sctp_v4_pf_exit(void)
1145 {
1146         list_del(&sctp_af_inet.list);
1147 }
1148
1149 static int sctp_v4_protosw_init(void)
1150 {
1151         int rc;
1152
1153         rc = proto_register(&sctp_prot, 1);
1154         if (rc)
1155                 return rc;
1156
1157         /* Register SCTP(UDP and TCP style) with socket layer.  */
1158         inet_register_protosw(&sctp_seqpacket_protosw);
1159         inet_register_protosw(&sctp_stream_protosw);
1160
1161         return 0;
1162 }
1163
1164 static void sctp_v4_protosw_exit(void)
1165 {
1166         inet_unregister_protosw(&sctp_stream_protosw);
1167         inet_unregister_protosw(&sctp_seqpacket_protosw);
1168         proto_unregister(&sctp_prot);
1169 }
1170
1171 static int sctp_v4_add_protocol(void)
1172 {
1173         /* Register notifier for inet address additions/deletions. */
1174         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1175
1176         /* Register SCTP with inet layer.  */
1177         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1178                 return -EAGAIN;
1179
1180         return 0;
1181 }
1182
1183 static void sctp_v4_del_protocol(void)
1184 {
1185         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1186         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1187 }
1188
1189 static int __net_init sctp_defaults_init(struct net *net)
1190 {
1191         int status;
1192
1193         /*
1194          * 14. Suggested SCTP Protocol Parameter Values
1195          */
1196         /* The following protocol parameters are RECOMMENDED:  */
1197         /* RTO.Initial              - 3  seconds */
1198         net->sctp.rto_initial                   = SCTP_RTO_INITIAL;
1199         /* RTO.Min                  - 1  second */
1200         net->sctp.rto_min                       = SCTP_RTO_MIN;
1201         /* RTO.Max                 -  60 seconds */
1202         net->sctp.rto_max                       = SCTP_RTO_MAX;
1203         /* RTO.Alpha                - 1/8 */
1204         net->sctp.rto_alpha                     = SCTP_RTO_ALPHA;
1205         /* RTO.Beta                 - 1/4 */
1206         net->sctp.rto_beta                      = SCTP_RTO_BETA;
1207
1208         /* Valid.Cookie.Life        - 60  seconds */
1209         net->sctp.valid_cookie_life             = SCTP_DEFAULT_COOKIE_LIFE;
1210
1211         /* Whether Cookie Preservative is enabled(1) or not(0) */
1212         net->sctp.cookie_preserve_enable        = 1;
1213
1214         /* Default sctp sockets to use md5 as their hmac alg */
1215 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1216         net->sctp.sctp_hmac_alg                 = "md5";
1217 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1218         net->sctp.sctp_hmac_alg                 = "sha1";
1219 #else
1220         net->sctp.sctp_hmac_alg                 = NULL;
1221 #endif
1222
1223         /* Max.Burst                - 4 */
1224         net->sctp.max_burst                     = SCTP_DEFAULT_MAX_BURST;
1225
1226         /* Association.Max.Retrans  - 10 attempts
1227          * Path.Max.Retrans         - 5  attempts (per destination address)
1228          * Max.Init.Retransmits     - 8  attempts
1229          */
1230         net->sctp.max_retrans_association       = 10;
1231         net->sctp.max_retrans_path              = 5;
1232         net->sctp.max_retrans_init              = 8;
1233
1234         /* Sendbuffer growth        - do per-socket accounting */
1235         net->sctp.sndbuf_policy                 = 0;
1236
1237         /* Rcvbuffer growth         - do per-socket accounting */
1238         net->sctp.rcvbuf_policy                 = 0;
1239
1240         /* HB.interval              - 30 seconds */
1241         net->sctp.hb_interval                   = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1242
1243         /* delayed SACK timeout */
1244         net->sctp.sack_timeout                  = SCTP_DEFAULT_TIMEOUT_SACK;
1245
1246         /* Disable ADDIP by default. */
1247         net->sctp.addip_enable = 0;
1248         net->sctp.addip_noauth = 0;
1249         net->sctp.default_auto_asconf = 0;
1250
1251         /* Enable PR-SCTP by default. */
1252         net->sctp.prsctp_enable = 1;
1253
1254         /* Disable AUTH by default. */
1255         net->sctp.auth_enable = 0;
1256
1257         /* Set SCOPE policy to enabled */
1258         net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1259
1260         /* Set the default rwnd update threshold */
1261         net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1262
1263         /* Initialize maximum autoclose timeout. */
1264         net->sctp.max_autoclose         = INT_MAX / HZ;
1265
1266         status = sctp_sysctl_net_register(net);
1267         if (status)
1268                 goto err_sysctl_register;
1269
1270         /* Allocate and initialise sctp mibs.  */
1271         status = init_sctp_mibs(net);
1272         if (status)
1273                 goto err_init_mibs;
1274
1275         /* Initialize proc fs directory.  */
1276         status = sctp_proc_init(net);
1277         if (status)
1278                 goto err_init_proc;
1279
1280         sctp_dbg_objcnt_init(net);
1281
1282         /* Initialize the local address list. */
1283         INIT_LIST_HEAD(&net->sctp.local_addr_list);
1284         spin_lock_init(&net->sctp.local_addr_lock);
1285         sctp_get_local_addr_list(net);
1286
1287         /* Initialize the address event list */
1288         INIT_LIST_HEAD(&net->sctp.addr_waitq);
1289         INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1290         spin_lock_init(&net->sctp.addr_wq_lock);
1291         net->sctp.addr_wq_timer.expires = 0;
1292         setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1293                     (unsigned long)net);
1294
1295         return 0;
1296
1297 err_init_proc:
1298         cleanup_sctp_mibs(net);
1299 err_init_mibs:
1300         sctp_sysctl_net_unregister(net);
1301 err_sysctl_register:
1302         return status;
1303 }
1304
1305 static void __net_exit sctp_defaults_exit(struct net *net)
1306 {
1307         /* Free the local address list */
1308         sctp_free_addr_wq(net);
1309         sctp_free_local_addr_list(net);
1310
1311         sctp_dbg_objcnt_exit(net);
1312
1313         sctp_proc_exit(net);
1314         cleanup_sctp_mibs(net);
1315         sctp_sysctl_net_unregister(net);
1316 }
1317
1318 static struct pernet_operations sctp_defaults_ops = {
1319         .init = sctp_defaults_init,
1320         .exit = sctp_defaults_exit,
1321 };
1322
1323 static int __net_init sctp_ctrlsock_init(struct net *net)
1324 {
1325         int status;
1326
1327         /* Initialize the control inode/socket for handling OOTB packets.  */
1328         status = sctp_ctl_sock_init(net);
1329         if (status)
1330                 pr_err("Failed to initialize the SCTP control sock\n");
1331
1332         return status;
1333 }
1334
1335 static void __net_init sctp_ctrlsock_exit(struct net *net)
1336 {
1337         /* Free the control endpoint.  */
1338         inet_ctl_sock_destroy(net->sctp.ctl_sock);
1339 }
1340
1341 static struct pernet_operations sctp_ctrlsock_ops = {
1342         .init = sctp_ctrlsock_init,
1343         .exit = sctp_ctrlsock_exit,
1344 };
1345
1346 /* Initialize the universe into something sensible.  */
1347 static __init int sctp_init(void)
1348 {
1349         int i;
1350         int status = -EINVAL;
1351         unsigned long goal;
1352         unsigned long limit;
1353         int max_share;
1354         int order;
1355
1356         sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1357
1358         /* Allocate bind_bucket and chunk caches. */
1359         status = -ENOBUFS;
1360         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1361                                                sizeof(struct sctp_bind_bucket),
1362                                                0, SLAB_HWCACHE_ALIGN,
1363                                                NULL);
1364         if (!sctp_bucket_cachep)
1365                 goto out;
1366
1367         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1368                                                sizeof(struct sctp_chunk),
1369                                                0, SLAB_HWCACHE_ALIGN,
1370                                                NULL);
1371         if (!sctp_chunk_cachep)
1372                 goto err_chunk_cachep;
1373
1374         status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1375         if (status)
1376                 goto err_percpu_counter_init;
1377
1378         /* Implementation specific variables. */
1379
1380         /* Initialize default stream count setup information. */
1381         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1382         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1383
1384         /* Initialize handle used for association ids. */
1385         idr_init(&sctp_assocs_id);
1386
1387         limit = nr_free_buffer_pages() / 8;
1388         limit = max(limit, 128UL);
1389         sysctl_sctp_mem[0] = limit / 4 * 3;
1390         sysctl_sctp_mem[1] = limit;
1391         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1392
1393         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1394         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1395         max_share = min(4UL*1024*1024, limit);
1396
1397         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1398         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1399         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1400
1401         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1402         sysctl_sctp_wmem[1] = 16*1024;
1403         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1404
1405         /* Size and allocate the association hash table.
1406          * The methodology is similar to that of the tcp hash tables.
1407          */
1408         if (totalram_pages >= (128 * 1024))
1409                 goal = totalram_pages >> (22 - PAGE_SHIFT);
1410         else
1411                 goal = totalram_pages >> (24 - PAGE_SHIFT);
1412
1413         for (order = 0; (1UL << order) < goal; order++)
1414                 ;
1415
1416         do {
1417                 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1418                                         sizeof(struct sctp_hashbucket);
1419                 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1420                         continue;
1421                 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1422                         __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1423         } while (!sctp_assoc_hashtable && --order > 0);
1424         if (!sctp_assoc_hashtable) {
1425                 pr_err("Failed association hash alloc\n");
1426                 status = -ENOMEM;
1427                 goto err_ahash_alloc;
1428         }
1429         for (i = 0; i < sctp_assoc_hashsize; i++) {
1430                 rwlock_init(&sctp_assoc_hashtable[i].lock);
1431                 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1432         }
1433
1434         /* Allocate and initialize the endpoint hash table.  */
1435         sctp_ep_hashsize = 64;
1436         sctp_ep_hashtable =
1437                 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1438         if (!sctp_ep_hashtable) {
1439                 pr_err("Failed endpoint_hash alloc\n");
1440                 status = -ENOMEM;
1441                 goto err_ehash_alloc;
1442         }
1443         for (i = 0; i < sctp_ep_hashsize; i++) {
1444                 rwlock_init(&sctp_ep_hashtable[i].lock);
1445                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1446         }
1447
1448         /* Allocate and initialize the SCTP port hash table.  */
1449         do {
1450                 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1451                                         sizeof(struct sctp_bind_hashbucket);
1452                 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1453                         continue;
1454                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1455                         __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1456         } while (!sctp_port_hashtable && --order > 0);
1457         if (!sctp_port_hashtable) {
1458                 pr_err("Failed bind hash alloc\n");
1459                 status = -ENOMEM;
1460                 goto err_bhash_alloc;
1461         }
1462         for (i = 0; i < sctp_port_hashsize; i++) {
1463                 spin_lock_init(&sctp_port_hashtable[i].lock);
1464                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1465         }
1466
1467         pr_info("Hash tables configured (established %d bind %d)\n",
1468                 sctp_assoc_hashsize, sctp_port_hashsize);
1469
1470         sctp_sysctl_register();
1471
1472         INIT_LIST_HEAD(&sctp_address_families);
1473         sctp_v4_pf_init();
1474         sctp_v6_pf_init();
1475
1476         status = register_pernet_subsys(&sctp_defaults_ops);
1477         if (status)
1478                 goto err_register_defaults;
1479
1480         status = sctp_v4_protosw_init();
1481         if (status)
1482                 goto err_protosw_init;
1483
1484         status = sctp_v6_protosw_init();
1485         if (status)
1486                 goto err_v6_protosw_init;
1487
1488         status = register_pernet_subsys(&sctp_ctrlsock_ops);
1489         if (status)
1490                 goto err_register_ctrlsock;
1491
1492         status = sctp_v4_add_protocol();
1493         if (status)
1494                 goto err_add_protocol;
1495
1496         /* Register SCTP with inet6 layer.  */
1497         status = sctp_v6_add_protocol();
1498         if (status)
1499                 goto err_v6_add_protocol;
1500
1501 out:
1502         return status;
1503 err_v6_add_protocol:
1504         sctp_v4_del_protocol();
1505 err_add_protocol:
1506         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1507 err_register_ctrlsock:
1508         sctp_v6_protosw_exit();
1509 err_v6_protosw_init:
1510         sctp_v4_protosw_exit();
1511 err_protosw_init:
1512         unregister_pernet_subsys(&sctp_defaults_ops);
1513 err_register_defaults:
1514         sctp_v4_pf_exit();
1515         sctp_v6_pf_exit();
1516         sctp_sysctl_unregister();
1517         free_pages((unsigned long)sctp_port_hashtable,
1518                    get_order(sctp_port_hashsize *
1519                              sizeof(struct sctp_bind_hashbucket)));
1520 err_bhash_alloc:
1521         kfree(sctp_ep_hashtable);
1522 err_ehash_alloc:
1523         free_pages((unsigned long)sctp_assoc_hashtable,
1524                    get_order(sctp_assoc_hashsize *
1525                              sizeof(struct sctp_hashbucket)));
1526 err_ahash_alloc:
1527         percpu_counter_destroy(&sctp_sockets_allocated);
1528 err_percpu_counter_init:
1529         kmem_cache_destroy(sctp_chunk_cachep);
1530 err_chunk_cachep:
1531         kmem_cache_destroy(sctp_bucket_cachep);
1532         goto out;
1533 }
1534
1535 /* Exit handler for the SCTP protocol.  */
1536 static __exit void sctp_exit(void)
1537 {
1538         /* BUG.  This should probably do something useful like clean
1539          * up all the remaining associations and all that memory.
1540          */
1541
1542         /* Unregister with inet6/inet layers. */
1543         sctp_v6_del_protocol();
1544         sctp_v4_del_protocol();
1545
1546         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1547
1548         /* Free protosw registrations */
1549         sctp_v6_protosw_exit();
1550         sctp_v4_protosw_exit();
1551
1552         unregister_pernet_subsys(&sctp_defaults_ops);
1553
1554         /* Unregister with socket layer. */
1555         sctp_v6_pf_exit();
1556         sctp_v4_pf_exit();
1557
1558         sctp_sysctl_unregister();
1559
1560         free_pages((unsigned long)sctp_assoc_hashtable,
1561                    get_order(sctp_assoc_hashsize *
1562                              sizeof(struct sctp_hashbucket)));
1563         kfree(sctp_ep_hashtable);
1564         free_pages((unsigned long)sctp_port_hashtable,
1565                    get_order(sctp_port_hashsize *
1566                              sizeof(struct sctp_bind_hashbucket)));
1567
1568         percpu_counter_destroy(&sctp_sockets_allocated);
1569
1570         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1571
1572         kmem_cache_destroy(sctp_chunk_cachep);
1573         kmem_cache_destroy(sctp_bucket_cachep);
1574 }
1575
1576 module_init(sctp_init);
1577 module_exit(sctp_exit);
1578
1579 /*
1580  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1581  */
1582 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1583 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1584 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1585 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1586 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1587 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1588 MODULE_LICENSE("GPL");