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, write to
  27  * the Free Software Foundation, 59 Temple Place - Suite 330,
  28  * Boston, MA 02111-1307, USA.
  29  *
  30  * Please send any bug reports or fixes you make to the
  31  * email address(es):
  32  *    lksctp developers <lksctp-developers@lists.sourceforge.net>
  33  *
  34  * Or submit a bug report through the following website:
  35  *    http://www.sf.net/projects/lksctp
  36  *
  37  * Written or modified by:
  38  *    La Monte H.P. Yarroll <piggy@acm.org>
  39  *    Karl Knutson <karl@athena.chicago.il.us>
  40  *    Jon Grimm <jgrimm@us.ibm.com>
  41  *    Sridhar Samudrala <sri@us.ibm.com>
  42  *    Daisy Chang <daisyc@us.ibm.com>
  43  *    Ardelle Fan <ardelle.fan@intel.com>
  44  *
  45  * Any bugs reported given to us we will try to fix... any fixes shared will
  46  * be incorporated into the next SCTP release.
  47  */
  48
  49 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  50
  51 #include <linux/module.h>
  52 #include <linux/init.h>
  53 #include <linux/netdevice.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/seq_file.h>
  56 #include <linux/bootmem.h>
  57 #include <linux/highmem.h>
  58 #include <linux/swap.h>
  59 #include <linux/slab.h>
  60 #include <net/net_namespace.h>
  61 #include <net/protocol.h>
  62 #include <net/ip.h>
  63 #include <net/ipv6.h>
  64 #include <net/route.h>
  65 #include <net/sctp/sctp.h>
  66 #include <net/addrconf.h>
  67 #include <net/inet_common.h>
  68 #include <net/inet_ecn.h>
  69
  70 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
  71
  72 /* Global data structures. */
  73 struct sctp_globals sctp_globals __read_mostly;
  74 DEFINE_SNMP_STAT(struct sctp_mib, sctp_statistics) __read_mostly;
  75
  76 #ifdef CONFIG_PROC_FS
  77 struct proc_dir_entry   *proc_net_sctp;
  78 #endif
  79
  80 struct idr sctp_assocs_id;
  81 DEFINE_SPINLOCK(sctp_assocs_id_lock);
  82
  83 /* This is the global socket data structure used for responding to
  84  * the Out-of-the-blue (OOTB) packets.  A control sock will be created
  85  * for this socket at the initialization time.
  86  */
  87 static struct sock *sctp_ctl_sock;
  88
  89 static struct sctp_pf *sctp_pf_inet6_specific;
  90 static struct sctp_pf *sctp_pf_inet_specific;
  91 static struct sctp_af *sctp_af_v4_specific;
  92 static struct sctp_af *sctp_af_v6_specific;
  93
  94 struct kmem_cache *sctp_chunk_cachep __read_mostly;
  95 struct kmem_cache *sctp_bucket_cachep __read_mostly;
  96
  97 long sysctl_sctp_mem[3];
  98 int sysctl_sctp_rmem[3];
  99 int sysctl_sctp_wmem[3];
 100
 101 /* Return the address of the control sock. */
 102 struct sock *sctp_get_ctl_sock(void)
 103 {
 104         return sctp_ctl_sock;
 105 }
 106
 107 /* Set up the proc fs entry for the SCTP protocol. */
 108 static __init int sctp_proc_init(void)
 109 {
 110         if (percpu_counter_init(&sctp_sockets_allocated, 0))
 111                 goto out_nomem;
 112 #ifdef CONFIG_PROC_FS
 113         if (!proc_net_sctp) {
 114                 proc_net_sctp = proc_mkdir("sctp", init_net.proc_net);
 115                 if (!proc_net_sctp)
 116                         goto out_free_percpu;
 117         }
 118
 119         if (sctp_snmp_proc_init())
 120                 goto out_snmp_proc_init;
 121         if (sctp_eps_proc_init())
 122                 goto out_eps_proc_init;
 123         if (sctp_assocs_proc_init())
 124                 goto out_assocs_proc_init;
 125         if (sctp_remaddr_proc_init())
 126                 goto out_remaddr_proc_init;
 127
 128         return 0;
 129
 130 out_remaddr_proc_init:
 131         sctp_assocs_proc_exit();
 132 out_assocs_proc_init:
 133         sctp_eps_proc_exit();
 134 out_eps_proc_init:
 135         sctp_snmp_proc_exit();
 136 out_snmp_proc_init:
 137         if (proc_net_sctp) {
 138                 proc_net_sctp = NULL;
 139                 remove_proc_entry("sctp", init_net.proc_net);
 140         }
 141 out_free_percpu:
 142         percpu_counter_destroy(&sctp_sockets_allocated);
 143 #else
 144         return 0;
 145 #endif /* CONFIG_PROC_FS */
 146
 147 out_nomem:
 148         return -ENOMEM;
 149 }
 150
 151 /* Clean up the proc fs entry for the SCTP protocol.
 152  * Note: Do not make this __exit as it is used in the init error
 153  * path.
 154  */
 155 static void sctp_proc_exit(void)
 156 {
 157 #ifdef CONFIG_PROC_FS
 158         sctp_snmp_proc_exit();
 159         sctp_eps_proc_exit();
 160         sctp_assocs_proc_exit();
 161         sctp_remaddr_proc_exit();
 162
 163         if (proc_net_sctp) {
 164                 proc_net_sctp = NULL;
 165                 remove_proc_entry("sctp", init_net.proc_net);
 166         }
 167 #endif
 168         percpu_counter_destroy(&sctp_sockets_allocated);
 169 }
 170
 171 /* Private helper to extract ipv4 address and stash them in
 172  * the protocol structure.
 173  */
 174 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
 175                                   struct net_device *dev)
 176 {
 177         struct in_device *in_dev;
 178         struct in_ifaddr *ifa;
 179         struct sctp_sockaddr_entry *addr;
 180
 181         rcu_read_lock();
 182         if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
 183                 rcu_read_unlock();
 184                 return;
 185         }
 186
 187         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
 188                 /* Add the address to the local list.  */
 189                 addr = t_new(struct sctp_sockaddr_entry, GFP_ATOMIC);
 190                 if (addr) {
 191                         addr->a.v4.sin_family = AF_INET;
 192                         addr->a.v4.sin_port = 0;
 193                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 194                         addr->valid = 1;
 195                         INIT_LIST_HEAD(&addr->list);
 196                         list_add_tail(&addr->list, addrlist);
 197                 }
 198         }
 199
 200         rcu_read_unlock();
 201 }
 202
 203 /* Extract our IP addresses from the system and stash them in the
 204  * protocol structure.
 205  */
 206 static void sctp_get_local_addr_list(void)
 207 {
 208         struct net_device *dev;
 209         struct list_head *pos;
 210         struct sctp_af *af;
 211
 212         rcu_read_lock();
 213         for_each_netdev_rcu(&init_net, dev) {
 214                 __list_for_each(pos, &sctp_address_families) {
 215                         af = list_entry(pos, struct sctp_af, list);
 216                         af->copy_addrlist(&sctp_local_addr_list, dev);
 217                 }
 218         }
 219         rcu_read_unlock();
 220 }
 221
 222 /* Free the existing local addresses.  */
 223 static void sctp_free_local_addr_list(void)
 224 {
 225         struct sctp_sockaddr_entry *addr;
 226         struct list_head *pos, *temp;
 227
 228         list_for_each_safe(pos, temp, &sctp_local_addr_list) {
 229                 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
 230                 list_del(pos);
 231                 kfree(addr);
 232         }
 233 }
 234
 235 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
 236 int sctp_copy_local_addr_list(struct sctp_bind_addr *bp, sctp_scope_t scope,
 237                               gfp_t gfp, int copy_flags)
 238 {
 239         struct sctp_sockaddr_entry *addr;
 240         int error = 0;
 241
 242         rcu_read_lock();
 243         list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
 244                 if (!addr->valid)
 245                         continue;
 246                 if (sctp_in_scope(&addr->a, scope)) {
 247                         /* Now that the address is in scope, check to see if
 248                          * the address type is really supported by the local
 249                          * sock as well as the remote peer.
 250                          */
 251                         if ((((AF_INET == addr->a.sa.sa_family) &&
 252                               (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
 253                             (((AF_INET6 == addr->a.sa.sa_family) &&
 254                               (copy_flags & SCTP_ADDR6_ALLOWED) &&
 255                               (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
 256                                 error = sctp_add_bind_addr(bp, &addr->a,
 257                                                     SCTP_ADDR_SRC, GFP_ATOMIC);
 258                                 if (error)
 259                                         goto end_copy;
 260                         }
 261                 }
 262         }
 263
 264 end_copy:
 265         rcu_read_unlock();
 266         return error;
 267 }
 268
 269 /* Initialize a sctp_addr from in incoming skb.  */
 270 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
 271                              int is_saddr)
 272 {
 273         void *from;
 274         __be16 *port;
 275         struct sctphdr *sh;
 276
 277         port = &addr->v4.sin_port;
 278         addr->v4.sin_family = AF_INET;
 279
 280         sh = sctp_hdr(skb);
 281         if (is_saddr) {
 282                 *port  = sh->source;
 283                 from = &ip_hdr(skb)->saddr;
 284         } else {
 285                 *port = sh->dest;
 286                 from = &ip_hdr(skb)->daddr;
 287         }
 288         memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
 289 }
 290
 291 /* Initialize an sctp_addr from a socket. */
 292 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
 293 {
 294         addr->v4.sin_family = AF_INET;
 295         addr->v4.sin_port = 0;
 296         addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
 297 }
 298
 299 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
 300 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
 301 {
 302         inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
 303 }
 304
 305 /* Initialize sk->sk_daddr from sctp_addr. */
 306 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
 307 {
 308         inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
 309 }
 310
 311 /* Initialize a sctp_addr from an address parameter. */
 312 static void sctp_v4_from_addr_param(union sctp_addr *addr,
 313                                     union sctp_addr_param *param,
 314                                     __be16 port, int iif)
 315 {
 316         addr->v4.sin_family = AF_INET;
 317         addr->v4.sin_port = port;
 318         addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
 319 }
 320
 321 /* Initialize an address parameter from a sctp_addr and return the length
 322  * of the address parameter.
 323  */
 324 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
 325                                  union sctp_addr_param *param)
 326 {
 327         int length = sizeof(sctp_ipv4addr_param_t);
 328
 329         param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
 330         param->v4.param_hdr.length = htons(length);
 331         param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
 332
 333         return length;
 334 }
 335
 336 /* Initialize a sctp_addr from a dst_entry. */
 337 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
 338                               __be16 port)
 339 {
 340         saddr->v4.sin_family = AF_INET;
 341         saddr->v4.sin_port = port;
 342         saddr->v4.sin_addr.s_addr = fl4->saddr;
 343 }
 344
 345 /* Compare two addresses exactly. */
 346 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
 347                             const union sctp_addr *addr2)
 348 {
 349         if (addr1->sa.sa_family != addr2->sa.sa_family)
 350                 return 0;
 351         if (addr1->v4.sin_port != addr2->v4.sin_port)
 352                 return 0;
 353         if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
 354                 return 0;
 355
 356         return 1;
 357 }
 358
 359 /* Initialize addr struct to INADDR_ANY. */
 360 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
 361 {
 362         addr->v4.sin_family = AF_INET;
 363         addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
 364         addr->v4.sin_port = port;
 365 }
 366
 367 /* Is this a wildcard address? */
 368 static int sctp_v4_is_any(const union sctp_addr *addr)
 369 {
 370         return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
 371 }
 372
 373 /* This function checks if the address is a valid address to be used for
 374  * SCTP binding.
 375  *
 376  * Output:
 377  * Return 0 - If the address is a non-unicast or an illegal address.
 378  * Return 1 - If the address is a unicast.
 379  */
 380 static int sctp_v4_addr_valid(union sctp_addr *addr,
 381                               struct sctp_sock *sp,
 382                               const struct sk_buff *skb)
 383 {
 384         /* IPv4 addresses not allowed */
 385         if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
 386                 return 0;
 387
 388         /* Is this a non-unicast address or a unusable SCTP address? */
 389         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
 390                 return 0;
 391
 392         /* Is this a broadcast address? */
 393         if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
 394                 return 0;
 395
 396         return 1;
 397 }
 398
 399 /* Should this be available for binding?   */
 400 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
 401 {
 402         int ret = inet_addr_type(&init_net, addr->v4.sin_addr.s_addr);
 403
 404
 405         if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
 406            ret != RTN_LOCAL &&
 407            !sp->inet.freebind &&
 408            !sysctl_ip_nonlocal_bind)
 409                 return 0;
 410
 411         if (ipv6_only_sock(sctp_opt2sk(sp)))
 412                 return 0;
 413
 414         return 1;
 415 }
 416
 417 /* Checking the loopback, private and other address scopes as defined in
 418  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
 419  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
 420  *
 421  * Level 0 - unusable SCTP addresses
 422  * Level 1 - loopback address
 423  * Level 2 - link-local addresses
 424  * Level 3 - private addresses.
 425  * Level 4 - global addresses
 426  * For INIT and INIT-ACK address list, let L be the level of
 427  * of requested destination address, sender and receiver
 428  * SHOULD include all of its addresses with level greater
 429  * than or equal to L.
 430  *
 431  * IPv4 scoping can be controlled through sysctl option
 432  * net.sctp.addr_scope_policy
 433  */
 434 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
 435 {
 436         sctp_scope_t retval;
 437
 438         /* Check for unusable SCTP addresses. */
 439         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
 440                 retval =  SCTP_SCOPE_UNUSABLE;
 441         } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
 442                 retval = SCTP_SCOPE_LOOPBACK;
 443         } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
 444                 retval = SCTP_SCOPE_LINK;
 445         } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
 446                    ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
 447                    ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
 448                 retval = SCTP_SCOPE_PRIVATE;
 449         } else {
 450                 retval = SCTP_SCOPE_GLOBAL;
 451         }
 452
 453         return retval;
 454 }
 455
 456 /* Returns a valid dst cache entry for the given source and destination ip
 457  * addresses. If an association is passed, trys to get a dst entry with a
 458  * source address that matches an address in the bind address list.
 459  */
 460 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
 461                                 struct flowi *fl, struct sock *sk)
 462 {
 463         struct sctp_association *asoc = t->asoc;
 464         struct rtable *rt;
 465         struct flowi4 *fl4 = &fl->u.ip4;
 466         struct sctp_bind_addr *bp;
 467         struct sctp_sockaddr_entry *laddr;
 468         struct dst_entry *dst = NULL;
 469         union sctp_addr *daddr = &t->ipaddr;
 470         union sctp_addr dst_saddr;
 471
 472         memset(fl4, 0x0, sizeof(struct flowi4));
 473         fl4->daddr  = daddr->v4.sin_addr.s_addr;
 474         fl4->fl4_dport = daddr->v4.sin_port;
 475         fl4->flowi4_proto = IPPROTO_SCTP;
 476         if (asoc) {
 477                 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
 478                 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
 479                 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
 480         }
 481         if (saddr) {
 482                 fl4->saddr = saddr->v4.sin_addr.s_addr;
 483                 fl4->fl4_sport = saddr->v4.sin_port;
 484         }
 485
 486         SCTP_DEBUG_PRINTK("%s: DST:%pI4, SRC:%pI4 - ",
 487                           __func__, &fl4->daddr, &fl4->saddr);
 488
 489         rt = ip_route_output_key(&init_net, fl4);
 490         if (!IS_ERR(rt))
 491                 dst = &rt->dst;
 492
 493         /* If there is no association or if a source address is passed, no
 494          * more validation is required.
 495          */
 496         if (!asoc || saddr)
 497                 goto out;
 498
 499         bp = &asoc->base.bind_addr;
 500
 501         if (dst) {
 502                 /* Walk through the bind address list and look for a bind
 503                  * address that matches the source address of the returned dst.
 504                  */
 505                 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
 506                 rcu_read_lock();
 507                 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 508                         if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
 509                             (laddr->state != SCTP_ADDR_SRC &&
 510                             !asoc->src_out_of_asoc_ok))
 511                                 continue;
 512                         if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
 513                                 goto out_unlock;
 514                 }
 515                 rcu_read_unlock();
 516
 517                 /* None of the bound addresses match the source address of the
 518                  * dst. So release it.
 519                  */
 520                 dst_release(dst);
 521                 dst = NULL;
 522         }
 523
 524         /* Walk through the bind address list and try to get a dst that
 525          * matches a bind address as the source address.
 526          */
 527         rcu_read_lock();
 528         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 529                 if (!laddr->valid)
 530                         continue;
 531                 if ((laddr->state == SCTP_ADDR_SRC) &&
 532                     (AF_INET == laddr->a.sa.sa_family)) {
 533                         fl4->fl4_sport = laddr->a.v4.sin_port;
 534                         flowi4_update_output(fl4,
 535                                              asoc->base.sk->sk_bound_dev_if,
 536                                              RT_CONN_FLAGS(asoc->base.sk),
 537                                              daddr->v4.sin_addr.s_addr,
 538                                              laddr->a.v4.sin_addr.s_addr);
 539
 540                         rt = ip_route_output_key(&init_net, fl4);
 541                         if (!IS_ERR(rt)) {
 542                                 dst = &rt->dst;
 543                                 goto out_unlock;
 544                         }
 545                 }
 546         }
 547
 548 out_unlock:
 549         rcu_read_unlock();
 550 out:
 551         t->dst = dst;
 552         if (dst)
 553                 SCTP_DEBUG_PRINTK("rt_dst:%pI4, rt_src:%pI4\n",
 554                                   &fl4->daddr, &fl4->saddr);
 555         else
 556                 SCTP_DEBUG_PRINTK("NO ROUTE\n");
 557 }
 558
 559 /* For v4, the source address is cached in the route entry(dst). So no need
 560  * to cache it separately and hence this is an empty routine.
 561  */
 562 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 563                               struct sctp_transport *t,
 564                               struct flowi *fl)
 565 {
 566         union sctp_addr *saddr = &t->saddr;
 567         struct rtable *rt = (struct rtable *)t->dst;
 568
 569         if (rt) {
 570                 saddr->v4.sin_family = AF_INET;
 571                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 572         }
 573 }
 574
 575 /* What interface did this skb arrive on? */
 576 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 577 {
 578         return skb_rtable(skb)->rt_iif;
 579 }
 580
 581 /* Was this packet marked by Explicit Congestion Notification? */
 582 static int sctp_v4_is_ce(const struct sk_buff *skb)
 583 {
 584         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 585 }
 586
 587 /* Create and initialize a new sk for the socket returned by accept(). */
 588 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 589                                              struct sctp_association *asoc)
 590 {
 591         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 592                         sk->sk_prot);
 593         struct inet_sock *newinet;
 594
 595         if (!newsk)
 596                 goto out;
 597
 598         sock_init_data(NULL, newsk);
 599
 600         sctp_copy_sock(newsk, sk, asoc);
 601         sock_reset_flag(newsk, SOCK_ZAPPED);
 602
 603         newinet = inet_sk(newsk);
 604
 605         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 606
 607         sk_refcnt_debug_inc(newsk);
 608
 609         if (newsk->sk_prot->init(newsk)) {
 610                 sk_common_release(newsk);
 611                 newsk = NULL;
 612         }
 613
 614 out:
 615         return newsk;
 616 }
 617
 618 /* Map address, empty for v4 family */
 619 static void sctp_v4_addr_v4map(struct sctp_sock *sp, union sctp_addr *addr)
 620 {
 621         /* Empty */
 622 }
 623
 624 /* Dump the v4 addr to the seq file. */
 625 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 626 {
 627         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 628 }
 629
 630 static void sctp_v4_ecn_capable(struct sock *sk)
 631 {
 632         INET_ECN_xmit(sk);
 633 }
 634
 635 void sctp_addr_wq_timeout_handler(unsigned long arg)
 636 {
 637         struct sctp_sockaddr_entry *addrw, *temp;
 638         struct sctp_sock *sp;
 639
 640         spin_lock_bh(&sctp_addr_wq_lock);
 641
 642         list_for_each_entry_safe(addrw, temp, &sctp_addr_waitq, list) {
 643                 SCTP_DEBUG_PRINTK_IPADDR("sctp_addrwq_timo_handler: the first ent in wq %p is ",
 644                     " for cmd %d at entry %p\n", &sctp_addr_waitq, &addrw->a, addrw->state,
 645                     addrw);
 646
 647 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
 648                 /* Now we send an ASCONF for each association */
 649                 /* Note. we currently don't handle link local IPv6 addressees */
 650                 if (addrw->a.sa.sa_family == AF_INET6) {
 651                         struct in6_addr *in6;
 652
 653                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 654                             IPV6_ADDR_LINKLOCAL)
 655                                 goto free_next;
 656
 657                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 658                         if (ipv6_chk_addr(&init_net, in6, NULL, 0) == 0 &&
 659                             addrw->state == SCTP_ADDR_NEW) {
 660                                 unsigned long timeo_val;
 661
 662                                 SCTP_DEBUG_PRINTK("sctp_timo_handler: this is on DAD, trying %d sec later\n",
 663                                     SCTP_ADDRESS_TICK_DELAY);
 664                                 timeo_val = jiffies;
 665                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 666                                 mod_timer(&sctp_addr_wq_timer, timeo_val);
 667                                 break;
 668                         }
 669                 }
 670 #endif
 671                 list_for_each_entry(sp, &sctp_auto_asconf_splist, auto_asconf_list) {
 672                         struct sock *sk;
 673
 674                         sk = sctp_opt2sk(sp);
 675                         /* ignore bound-specific endpoints */
 676                         if (!sctp_is_ep_boundall(sk))
 677                                 continue;
 678                         sctp_bh_lock_sock(sk);
 679                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 680                                 SCTP_DEBUG_PRINTK("sctp_addrwq_timo_handler: sctp_asconf_mgmt failed\n");
 681                         sctp_bh_unlock_sock(sk);
 682                 }
 683 free_next:
 684                 list_del(&addrw->list);
 685                 kfree(addrw);
 686         }
 687         spin_unlock_bh(&sctp_addr_wq_lock);
 688 }
 689
 690 static void sctp_free_addr_wq(void)
 691 {
 692         struct sctp_sockaddr_entry *addrw;
 693         struct sctp_sockaddr_entry *temp;
 694
 695         spin_lock_bh(&sctp_addr_wq_lock);
 696         del_timer(&sctp_addr_wq_timer);
 697         list_for_each_entry_safe(addrw, temp, &sctp_addr_waitq, list) {
 698                 list_del(&addrw->list);
 699                 kfree(addrw);
 700         }
 701         spin_unlock_bh(&sctp_addr_wq_lock);
 702 }
 703
 704 /* lookup the entry for the same address in the addr_waitq
 705  * sctp_addr_wq MUST be locked
 706  */
 707 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct sctp_sockaddr_entry *addr)
 708 {
 709         struct sctp_sockaddr_entry *addrw;
 710
 711         list_for_each_entry(addrw, &sctp_addr_waitq, list) {
 712                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 713                         continue;
 714                 if (addrw->a.sa.sa_family == AF_INET) {
 715                         if (addrw->a.v4.sin_addr.s_addr ==
 716                             addr->a.v4.sin_addr.s_addr)
 717                                 return addrw;
 718                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 719                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 720                             &addr->a.v6.sin6_addr))
 721                                 return addrw;
 722                 }
 723         }
 724         return NULL;
 725 }
 726
 727 void sctp_addr_wq_mgmt(struct sctp_sockaddr_entry *addr, int cmd)
 728 {
 729         struct sctp_sockaddr_entry *addrw;
 730         unsigned long timeo_val;
 731
 732         /* first, we check if an opposite message already exist in the queue.
 733          * If we found such message, it is removed.
 734          * This operation is a bit stupid, but the DHCP client attaches the
 735          * new address after a couple of addition and deletion of that address
 736          */
 737
 738         spin_lock_bh(&sctp_addr_wq_lock);
 739         /* Offsets existing events in addr_wq */
 740         addrw = sctp_addr_wq_lookup(addr);
 741         if (addrw) {
 742                 if (addrw->state != cmd) {
 743                         SCTP_DEBUG_PRINTK_IPADDR("sctp_addr_wq_mgmt offsets existing entry for %d ",
 744                             " in wq %p\n", addrw->state, &addrw->a,
 745                             &sctp_addr_waitq);
 746                         list_del(&addrw->list);
 747                         kfree(addrw);
 748                 }
 749                 spin_unlock_bh(&sctp_addr_wq_lock);
 750                 return;
 751         }
 752
 753         /* OK, we have to add the new address to the wait queue */
 754         addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 755         if (addrw == NULL) {
 756                 spin_unlock_bh(&sctp_addr_wq_lock);
 757                 return;
 758         }
 759         addrw->state = cmd;
 760         list_add_tail(&addrw->list, &sctp_addr_waitq);
 761         SCTP_DEBUG_PRINTK_IPADDR("sctp_addr_wq_mgmt add new entry for cmd:%d ",
 762             " in wq %p\n", addrw->state, &addrw->a, &sctp_addr_waitq);
 763
 764         if (!timer_pending(&sctp_addr_wq_timer)) {
 765                 timeo_val = jiffies;
 766                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 767                 mod_timer(&sctp_addr_wq_timer, timeo_val);
 768         }
 769         spin_unlock_bh(&sctp_addr_wq_lock);
 770 }
 771
 772 /* Event handler for inet address addition/deletion events.
 773  * The sctp_local_addr_list needs to be protocted by a spin lock since
 774  * multiple notifiers (say IPv4 and IPv6) may be running at the same
 775  * time and thus corrupt the list.
 776  * The reader side is protected with RCU.
 777  */
 778 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
 779                                void *ptr)
 780 {
 781         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
 782         struct sctp_sockaddr_entry *addr = NULL;
 783         struct sctp_sockaddr_entry *temp;
 784         int found = 0;
 785
 786         if (!net_eq(dev_net(ifa->ifa_dev->dev), &init_net))
 787                 return NOTIFY_DONE;
 788
 789         switch (ev) {
 790         case NETDEV_UP:
 791                 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 792                 if (addr) {
 793                         addr->a.v4.sin_family = AF_INET;
 794                         addr->a.v4.sin_port = 0;
 795                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 796                         addr->valid = 1;
 797                         spin_lock_bh(&sctp_local_addr_lock);
 798                         list_add_tail_rcu(&addr->list, &sctp_local_addr_list);
 799                         sctp_addr_wq_mgmt(addr, SCTP_ADDR_NEW);
 800                         spin_unlock_bh(&sctp_local_addr_lock);
 801                 }
 802                 break;
 803         case NETDEV_DOWN:
 804                 spin_lock_bh(&sctp_local_addr_lock);
 805                 list_for_each_entry_safe(addr, temp,
 806                                         &sctp_local_addr_list, list) {
 807                         if (addr->a.sa.sa_family == AF_INET &&
 808                                         addr->a.v4.sin_addr.s_addr ==
 809                                         ifa->ifa_local) {
 810                                 sctp_addr_wq_mgmt(addr, SCTP_ADDR_DEL);
 811                                 found = 1;
 812                                 addr->valid = 0;
 813                                 list_del_rcu(&addr->list);
 814                                 break;
 815                         }
 816                 }
 817                 spin_unlock_bh(&sctp_local_addr_lock);
 818                 if (found)
 819                         kfree_rcu(addr, rcu);
 820                 break;
 821         }
 822
 823         return NOTIFY_DONE;
 824 }
 825
 826 /*
 827  * Initialize the control inode/socket with a control endpoint data
 828  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 829  */
 830 static int sctp_ctl_sock_init(void)
 831 {
 832         int err;
 833         sa_family_t family = PF_INET;
 834
 835         if (sctp_get_pf_specific(PF_INET6))
 836                 family = PF_INET6;
 837
 838         err = inet_ctl_sock_create(&sctp_ctl_sock, family,
 839                                    SOCK_SEQPACKET, IPPROTO_SCTP, &init_net);
 840
 841         /* If IPv6 socket could not be created, try the IPv4 socket */
 842         if (err < 0 && family == PF_INET6)
 843                 err = inet_ctl_sock_create(&sctp_ctl_sock, AF_INET,
 844                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 845                                            &init_net);
 846
 847         if (err < 0) {
 848                 pr_err("Failed to create the SCTP control socket\n");
 849                 return err;
 850         }
 851         return 0;
 852 }
 853
 854 /* Register address family specific functions. */
 855 int sctp_register_af(struct sctp_af *af)
 856 {
 857         switch (af->sa_family) {
 858         case AF_INET:
 859                 if (sctp_af_v4_specific)
 860                         return 0;
 861                 sctp_af_v4_specific = af;
 862                 break;
 863         case AF_INET6:
 864                 if (sctp_af_v6_specific)
 865                         return 0;
 866                 sctp_af_v6_specific = af;
 867                 break;
 868         default:
 869                 return 0;
 870         }
 871
 872         INIT_LIST_HEAD(&af->list);
 873         list_add_tail(&af->list, &sctp_address_families);
 874         return 1;
 875 }
 876
 877 /* Get the table of functions for manipulating a particular address
 878  * family.
 879  */
 880 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 881 {
 882         switch (family) {
 883         case AF_INET:
 884                 return sctp_af_v4_specific;
 885         case AF_INET6:
 886                 return sctp_af_v6_specific;
 887         default:
 888                 return NULL;
 889         }
 890 }
 891
 892 /* Common code to initialize a AF_INET msg_name. */
 893 static void sctp_inet_msgname(char *msgname, int *addr_len)
 894 {
 895         struct sockaddr_in *sin;
 896
 897         sin = (struct sockaddr_in *)msgname;
 898         *addr_len = sizeof(struct sockaddr_in);
 899         sin->sin_family = AF_INET;
 900         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 901 }
 902
 903 /* Copy the primary address of the peer primary address as the msg_name. */
 904 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 905                                     int *addr_len)
 906 {
 907         struct sockaddr_in *sin, *sinfrom;
 908
 909         if (msgname) {
 910                 struct sctp_association *asoc;
 911
 912                 asoc = event->asoc;
 913                 sctp_inet_msgname(msgname, addr_len);
 914                 sin = (struct sockaddr_in *)msgname;
 915                 sinfrom = &asoc->peer.primary_addr.v4;
 916                 sin->sin_port = htons(asoc->peer.port);
 917                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 918         }
 919 }
 920
 921 /* Initialize and copy out a msgname from an inbound skb. */
 922 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 923 {
 924         if (msgname) {
 925                 struct sctphdr *sh = sctp_hdr(skb);
 926                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 927
 928                 sctp_inet_msgname(msgname, len);
 929                 sin->sin_port = sh->source;
 930                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 931         }
 932 }
 933
 934 /* Do we support this AF? */
 935 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 936 {
 937         /* PF_INET only supports AF_INET addresses. */
 938         return AF_INET == family;
 939 }
 940
 941 /* Address matching with wildcards allowed. */
 942 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 943                               const union sctp_addr *addr2,
 944                               struct sctp_sock *opt)
 945 {
 946         /* PF_INET only supports AF_INET addresses. */
 947         if (addr1->sa.sa_family != addr2->sa.sa_family)
 948                 return 0;
 949         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 950             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 951                 return 1;
 952         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 953                 return 1;
 954
 955         return 0;
 956 }
 957
 958 /* Verify that provided sockaddr looks bindable.  Common verification has
 959  * already been taken care of.
 960  */
 961 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 962 {
 963         return sctp_v4_available(addr, opt);
 964 }
 965
 966 /* Verify that sockaddr looks sendable.  Common verification has already
 967  * been taken care of.
 968  */
 969 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 970 {
 971         return 1;
 972 }
 973
 974 /* Fill in Supported Address Type information for INIT and INIT-ACK
 975  * chunks.  Returns number of addresses supported.
 976  */
 977 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 978                                      __be16 *types)
 979 {
 980         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 981         return 1;
 982 }
 983
 984 /* Wrapper routine that calls the ip transmit routine. */
 985 static inline int sctp_v4_xmit(struct sk_buff *skb,
 986                                struct sctp_transport *transport)
 987 {
 988         struct inet_sock *inet = inet_sk(skb->sk);
 989
 990         SCTP_DEBUG_PRINTK("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n",
 991                           __func__, skb, skb->len,
 992                           &transport->fl.u.ip4.saddr,
 993                           &transport->fl.u.ip4.daddr);
 994
 995         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
 996                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 997
 998         SCTP_INC_STATS(SCTP_MIB_OUTSCTPPACKS);
 999         return ip_queue_xmit(skb, &transport->fl);
1000 }
1001
1002 static struct sctp_af sctp_af_inet;
1003
1004 static struct sctp_pf sctp_pf_inet = {
1005         .event_msgname = sctp_inet_event_msgname,
1006         .skb_msgname   = sctp_inet_skb_msgname,
1007         .af_supported  = sctp_inet_af_supported,
1008         .cmp_addr      = sctp_inet_cmp_addr,
1009         .bind_verify   = sctp_inet_bind_verify,
1010         .send_verify   = sctp_inet_send_verify,
1011         .supported_addrs = sctp_inet_supported_addrs,
1012         .create_accept_sk = sctp_v4_create_accept_sk,
1013         .addr_v4map     = sctp_v4_addr_v4map,
1014         .af            = &sctp_af_inet
1015 };
1016
1017 /* Notifier for inetaddr addition/deletion events.  */
1018 static struct notifier_block sctp_inetaddr_notifier = {
1019         .notifier_call = sctp_inetaddr_event,
1020 };
1021
1022 /* Socket operations.  */
1023 static const struct proto_ops inet_seqpacket_ops = {
1024         .family            = PF_INET,
1025         .owner             = THIS_MODULE,
1026         .release           = inet_release,      /* Needs to be wrapped... */
1027         .bind              = inet_bind,
1028         .connect           = inet_dgram_connect,
1029         .socketpair        = sock_no_socketpair,
1030         .accept            = inet_accept,
1031         .getname           = inet_getname,      /* Semantics are different.  */
1032         .poll              = sctp_poll,
1033         .ioctl             = inet_ioctl,
1034         .listen            = sctp_inet_listen,
1035         .shutdown          = inet_shutdown,     /* Looks harmless.  */
1036         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1037         .getsockopt        = sock_common_getsockopt,
1038         .sendmsg           = inet_sendmsg,
1039         .recvmsg           = sock_common_recvmsg,
1040         .mmap              = sock_no_mmap,
1041         .sendpage          = sock_no_sendpage,
1042 #ifdef CONFIG_COMPAT
1043         .compat_setsockopt = compat_sock_common_setsockopt,
1044         .compat_getsockopt = compat_sock_common_getsockopt,
1045 #endif
1046 };
1047
1048 /* Registration with AF_INET family.  */
1049 static struct inet_protosw sctp_seqpacket_protosw = {
1050         .type       = SOCK_SEQPACKET,
1051         .protocol   = IPPROTO_SCTP,
1052         .prot       = &sctp_prot,
1053         .ops        = &inet_seqpacket_ops,
1054         .no_check   = 0,
1055         .flags      = SCTP_PROTOSW_FLAG
1056 };
1057 static struct inet_protosw sctp_stream_protosw = {
1058         .type       = SOCK_STREAM,
1059         .protocol   = IPPROTO_SCTP,
1060         .prot       = &sctp_prot,
1061         .ops        = &inet_seqpacket_ops,
1062         .no_check   = 0,
1063         .flags      = SCTP_PROTOSW_FLAG
1064 };
1065
1066 /* Register with IP layer.  */
1067 static const struct net_protocol sctp_protocol = {
1068         .handler     = sctp_rcv,
1069         .err_handler = sctp_v4_err,
1070         .no_policy   = 1,
1071 };
1072
1073 /* IPv4 address related functions.  */
1074 static struct sctp_af sctp_af_inet = {
1075         .sa_family         = AF_INET,
1076         .sctp_xmit         = sctp_v4_xmit,
1077         .setsockopt        = ip_setsockopt,
1078         .getsockopt        = ip_getsockopt,
1079         .get_dst           = sctp_v4_get_dst,
1080         .get_saddr         = sctp_v4_get_saddr,
1081         .copy_addrlist     = sctp_v4_copy_addrlist,
1082         .from_skb          = sctp_v4_from_skb,
1083         .from_sk           = sctp_v4_from_sk,
1084         .to_sk_saddr       = sctp_v4_to_sk_saddr,
1085         .to_sk_daddr       = sctp_v4_to_sk_daddr,
1086         .from_addr_param   = sctp_v4_from_addr_param,
1087         .to_addr_param     = sctp_v4_to_addr_param,
1088         .cmp_addr          = sctp_v4_cmp_addr,
1089         .addr_valid        = sctp_v4_addr_valid,
1090         .inaddr_any        = sctp_v4_inaddr_any,
1091         .is_any            = sctp_v4_is_any,
1092         .available         = sctp_v4_available,
1093         .scope             = sctp_v4_scope,
1094         .skb_iif           = sctp_v4_skb_iif,
1095         .is_ce             = sctp_v4_is_ce,
1096         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1097         .ecn_capable       = sctp_v4_ecn_capable,
1098         .net_header_len    = sizeof(struct iphdr),
1099         .sockaddr_len      = sizeof(struct sockaddr_in),
1100 #ifdef CONFIG_COMPAT
1101         .compat_setsockopt = compat_ip_setsockopt,
1102         .compat_getsockopt = compat_ip_getsockopt,
1103 #endif
1104 };
1105
1106 struct sctp_pf *sctp_get_pf_specific(sa_family_t family) {
1107
1108         switch (family) {
1109         case PF_INET:
1110                 return sctp_pf_inet_specific;
1111         case PF_INET6:
1112                 return sctp_pf_inet6_specific;
1113         default:
1114                 return NULL;
1115         }
1116 }
1117
1118 /* Register the PF specific function table.  */
1119 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1120 {
1121         switch (family) {
1122         case PF_INET:
1123                 if (sctp_pf_inet_specific)
1124                         return 0;
1125                 sctp_pf_inet_specific = pf;
1126                 break;
1127         case PF_INET6:
1128                 if (sctp_pf_inet6_specific)
1129                         return 0;
1130                 sctp_pf_inet6_specific = pf;
1131                 break;
1132         default:
1133                 return 0;
1134         }
1135         return 1;
1136 }
1137
1138 static inline int init_sctp_mibs(void)
1139 {
1140         return snmp_mib_init((void __percpu **)sctp_statistics,
1141                              sizeof(struct sctp_mib),
1142                              __alignof__(struct sctp_mib));
1143 }
1144
1145 static inline void cleanup_sctp_mibs(void)
1146 {
1147         snmp_mib_free((void __percpu **)sctp_statistics);
1148 }
1149
1150 static void sctp_v4_pf_init(void)
1151 {
1152         /* Initialize the SCTP specific PF functions. */
1153         sctp_register_pf(&sctp_pf_inet, PF_INET);
1154         sctp_register_af(&sctp_af_inet);
1155 }
1156
1157 static void sctp_v4_pf_exit(void)
1158 {
1159         list_del(&sctp_af_inet.list);
1160 }
1161
1162 static int sctp_v4_protosw_init(void)
1163 {
1164         int rc;
1165
1166         rc = proto_register(&sctp_prot, 1);
1167         if (rc)
1168                 return rc;
1169
1170         /* Register SCTP(UDP and TCP style) with socket layer.  */
1171         inet_register_protosw(&sctp_seqpacket_protosw);
1172         inet_register_protosw(&sctp_stream_protosw);
1173
1174         return 0;
1175 }
1176
1177 static void sctp_v4_protosw_exit(void)
1178 {
1179         inet_unregister_protosw(&sctp_stream_protosw);
1180         inet_unregister_protosw(&sctp_seqpacket_protosw);
1181         proto_unregister(&sctp_prot);
1182 }
1183
1184 static int sctp_v4_add_protocol(void)
1185 {
1186         /* Register notifier for inet address additions/deletions. */
1187         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1188
1189         /* Register SCTP with inet layer.  */
1190         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1191                 return -EAGAIN;
1192
1193         return 0;
1194 }
1195
1196 static void sctp_v4_del_protocol(void)
1197 {
1198         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1199         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1200 }
1201
1202 /* Initialize the universe into something sensible.  */
1203 SCTP_STATIC __init int sctp_init(void)
1204 {
1205         int i;
1206         int status = -EINVAL;
1207         unsigned long goal;
1208         unsigned long limit;
1209         int max_share;
1210         int order;
1211         int num_entries;
1212         int max_entry_order;
1213
1214         /* SCTP_DEBUG sanity check. */
1215         if (!sctp_sanity_check())
1216                 goto out;
1217
1218         /* Allocate bind_bucket and chunk caches. */
1219         status = -ENOBUFS;
1220         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1221                                                sizeof(struct sctp_bind_bucket),
1222                                                0, SLAB_HWCACHE_ALIGN,
1223                                                NULL);
1224         if (!sctp_bucket_cachep)
1225                 goto out;
1226
1227         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1228                                                sizeof(struct sctp_chunk),
1229                                                0, SLAB_HWCACHE_ALIGN,
1230                                                NULL);
1231         if (!sctp_chunk_cachep)
1232                 goto err_chunk_cachep;
1233
1234         /* Allocate and initialise sctp mibs.  */
1235         status = init_sctp_mibs();
1236         if (status)
1237                 goto err_init_mibs;
1238
1239         /* Initialize proc fs directory.  */
1240         status = sctp_proc_init();
1241         if (status)
1242                 goto err_init_proc;
1243
1244         /* Initialize object count debugging.  */
1245         sctp_dbg_objcnt_init();
1246
1247         /*
1248          * 14. Suggested SCTP Protocol Parameter Values
1249          */
1250         /* The following protocol parameters are RECOMMENDED:  */
1251         /* RTO.Initial              - 3  seconds */
1252         sctp_rto_initial                = SCTP_RTO_INITIAL;
1253         /* RTO.Min                  - 1  second */
1254         sctp_rto_min                    = SCTP_RTO_MIN;
1255         /* RTO.Max                 -  60 seconds */
1256         sctp_rto_max                    = SCTP_RTO_MAX;
1257         /* RTO.Alpha                - 1/8 */
1258         sctp_rto_alpha                  = SCTP_RTO_ALPHA;
1259         /* RTO.Beta                 - 1/4 */
1260         sctp_rto_beta                   = SCTP_RTO_BETA;
1261
1262         /* Valid.Cookie.Life        - 60  seconds */
1263         sctp_valid_cookie_life          = SCTP_DEFAULT_COOKIE_LIFE;
1264
1265         /* Whether Cookie Preservative is enabled(1) or not(0) */
1266         sctp_cookie_preserve_enable     = 1;
1267
1268         /* Max.Burst                - 4 */
1269         sctp_max_burst                  = SCTP_DEFAULT_MAX_BURST;
1270
1271         /* Association.Max.Retrans  - 10 attempts
1272          * Path.Max.Retrans         - 5  attempts (per destination address)
1273          * Max.Init.Retransmits     - 8  attempts
1274          */
1275         sctp_max_retrans_association    = 10;
1276         sctp_max_retrans_path           = 5;
1277         sctp_max_retrans_init           = 8;
1278
1279         /* Sendbuffer growth        - do per-socket accounting */
1280         sctp_sndbuf_policy              = 0;
1281
1282         /* Rcvbuffer growth         - do per-socket accounting */
1283         sctp_rcvbuf_policy              = 0;
1284
1285         /* HB.interval              - 30 seconds */
1286         sctp_hb_interval                = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1287
1288         /* delayed SACK timeout */
1289         sctp_sack_timeout               = SCTP_DEFAULT_TIMEOUT_SACK;
1290
1291         /* Implementation specific variables. */
1292
1293         /* Initialize default stream count setup information. */
1294         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1295         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1296
1297         /* Initialize maximum autoclose timeout. */
1298         sctp_max_autoclose              = INT_MAX / HZ;
1299
1300         /* Initialize handle used for association ids. */
1301         idr_init(&sctp_assocs_id);
1302
1303         limit = nr_free_buffer_pages() / 8;
1304         limit = max(limit, 128UL);
1305         sysctl_sctp_mem[0] = limit / 4 * 3;
1306         sysctl_sctp_mem[1] = limit;
1307         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1308
1309         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1310         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1311         max_share = min(4UL*1024*1024, limit);
1312
1313         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1314         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1315         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1316
1317         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1318         sysctl_sctp_wmem[1] = 16*1024;
1319         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1320
1321         /* Size and allocate the association hash table.
1322          * The methodology is similar to that of the tcp hash tables.
1323          * Though not identical.  Start by getting a goal size
1324          */
1325         if (totalram_pages >= (128 * 1024))
1326                 goal = totalram_pages >> (22 - PAGE_SHIFT);
1327         else
1328                 goal = totalram_pages >> (24 - PAGE_SHIFT);
1329
1330         /* Then compute the page order for said goal */
1331         order = get_order(goal);
1332
1333         /* Now compute the required page order for the maximum sized table we
1334          * want to create
1335          */
1336         max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1337                                     sizeof(struct sctp_bind_hashbucket));
1338
1339         /* Limit the page order by that maximum hash table size */
1340         order = min(order, max_entry_order);
1341
1342         do {
1343                 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1344                                         sizeof(struct sctp_hashbucket);
1345                 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1346                         continue;
1347                 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1348                         __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1349         } while (!sctp_assoc_hashtable && --order > 0);
1350         if (!sctp_assoc_hashtable) {
1351                 pr_err("Failed association hash alloc\n");
1352                 status = -ENOMEM;
1353                 goto err_ahash_alloc;
1354         }
1355         for (i = 0; i < sctp_assoc_hashsize; i++) {
1356                 rwlock_init(&sctp_assoc_hashtable[i].lock);
1357                 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1358         }
1359
1360         /* Allocate and initialize the endpoint hash table.  */
1361         sctp_ep_hashsize = 64;
1362         sctp_ep_hashtable = (struct sctp_hashbucket *)
1363                 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1364         if (!sctp_ep_hashtable) {
1365                 pr_err("Failed endpoint_hash alloc\n");
1366                 status = -ENOMEM;
1367                 goto err_ehash_alloc;
1368         }
1369         for (i = 0; i < sctp_ep_hashsize; i++) {
1370                 rwlock_init(&sctp_ep_hashtable[i].lock);
1371                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1372         }
1373
1374         /* Allocate and initialize the SCTP port hash table.
1375          * Note that order is initalized to start at the max sized
1376          * table we want to support.  If we can't get that many pages
1377          * reduce the order and try again
1378          */
1379         do {
1380                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1381                         __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1382         } while (!sctp_port_hashtable && --order > 0);
1383
1384         if (!sctp_port_hashtable) {
1385                 pr_err("Failed bind hash alloc\n");
1386                 status = -ENOMEM;
1387                 goto err_bhash_alloc;
1388         }
1389
1390         /* Now compute the number of entries that will fit in the
1391          * port hash space we allocated
1392          */
1393         num_entries = (1UL << order) * PAGE_SIZE /
1394                       sizeof(struct sctp_bind_hashbucket);
1395
1396         /* And finish by rounding it down to the nearest power of two
1397          * this wastes some memory of course, but its needed because
1398          * the hash function operates based on the assumption that
1399          * that the number of entries is a power of two
1400          */
1401         sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1402
1403         for (i = 0; i < sctp_port_hashsize; i++) {
1404                 spin_lock_init(&sctp_port_hashtable[i].lock);
1405                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1406         }
1407
1408         pr_info("Hash tables configured (established %d bind %d/%d)\n",
1409                 sctp_assoc_hashsize, sctp_port_hashsize, num_entries);
1410
1411         /* Disable ADDIP by default. */
1412         sctp_addip_enable = 0;
1413         sctp_addip_noauth = 0;
1414         sctp_default_auto_asconf = 0;
1415
1416         /* Enable PR-SCTP by default. */
1417         sctp_prsctp_enable = 1;
1418
1419         /* Disable AUTH by default. */
1420         sctp_auth_enable = 0;
1421
1422         /* Set SCOPE policy to enabled */
1423         sctp_scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1424
1425         /* Set the default rwnd update threshold */
1426         sctp_rwnd_upd_shift             = SCTP_DEFAULT_RWND_SHIFT;
1427
1428         sctp_sysctl_register();
1429
1430         INIT_LIST_HEAD(&sctp_address_families);
1431         sctp_v4_pf_init();
1432         sctp_v6_pf_init();
1433
1434         /* Initialize the local address list. */
1435         INIT_LIST_HEAD(&sctp_local_addr_list);
1436         spin_lock_init(&sctp_local_addr_lock);
1437         sctp_get_local_addr_list();
1438
1439         /* Initialize the address event list */
1440         INIT_LIST_HEAD(&sctp_addr_waitq);
1441         INIT_LIST_HEAD(&sctp_auto_asconf_splist);
1442         spin_lock_init(&sctp_addr_wq_lock);
1443         sctp_addr_wq_timer.expires = 0;
1444         setup_timer(&sctp_addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1445
1446         status = sctp_v4_protosw_init();
1447
1448         if (status)
1449                 goto err_protosw_init;
1450
1451         status = sctp_v6_protosw_init();
1452         if (status)
1453                 goto err_v6_protosw_init;
1454
1455         /* Initialize the control inode/socket for handling OOTB packets.  */
1456         if ((status = sctp_ctl_sock_init())) {
1457                 pr_err("Failed to initialize the SCTP control sock\n");
1458                 goto err_ctl_sock_init;
1459         }
1460
1461         status = sctp_v4_add_protocol();
1462         if (status)
1463                 goto err_add_protocol;
1464
1465         /* Register SCTP with inet6 layer.  */
1466         status = sctp_v6_add_protocol();
1467         if (status)
1468                 goto err_v6_add_protocol;
1469
1470         status = 0;
1471 out:
1472         return status;
1473 err_v6_add_protocol:
1474         sctp_v4_del_protocol();
1475 err_add_protocol:
1476         inet_ctl_sock_destroy(sctp_ctl_sock);
1477 err_ctl_sock_init:
1478         sctp_v6_protosw_exit();
1479 err_v6_protosw_init:
1480         sctp_v4_protosw_exit();
1481 err_protosw_init:
1482         sctp_free_local_addr_list();
1483         sctp_v4_pf_exit();
1484         sctp_v6_pf_exit();
1485         sctp_sysctl_unregister();
1486         free_pages((unsigned long)sctp_port_hashtable,
1487                    get_order(sctp_port_hashsize *
1488                              sizeof(struct sctp_bind_hashbucket)));
1489 err_bhash_alloc:
1490         kfree(sctp_ep_hashtable);
1491 err_ehash_alloc:
1492         free_pages((unsigned long)sctp_assoc_hashtable,
1493                    get_order(sctp_assoc_hashsize *
1494                              sizeof(struct sctp_hashbucket)));
1495 err_ahash_alloc:
1496         sctp_dbg_objcnt_exit();
1497         sctp_proc_exit();
1498 err_init_proc:
1499         cleanup_sctp_mibs();
1500 err_init_mibs:
1501         kmem_cache_destroy(sctp_chunk_cachep);
1502 err_chunk_cachep:
1503         kmem_cache_destroy(sctp_bucket_cachep);
1504         goto out;
1505 }
1506
1507 /* Exit handler for the SCTP protocol.  */
1508 SCTP_STATIC __exit void sctp_exit(void)
1509 {
1510         /* BUG.  This should probably do something useful like clean
1511          * up all the remaining associations and all that memory.
1512          */
1513
1514         /* Unregister with inet6/inet layers. */
1515         sctp_v6_del_protocol();
1516         sctp_v4_del_protocol();
1517         sctp_free_addr_wq();
1518
1519         /* Free the control endpoint.  */
1520         inet_ctl_sock_destroy(sctp_ctl_sock);
1521
1522         /* Free protosw registrations */
1523         sctp_v6_protosw_exit();
1524         sctp_v4_protosw_exit();
1525
1526         /* Free the local address list.  */
1527         sctp_free_local_addr_list();
1528
1529         /* Unregister with socket layer. */
1530         sctp_v6_pf_exit();
1531         sctp_v4_pf_exit();
1532
1533         sctp_sysctl_unregister();
1534
1535         free_pages((unsigned long)sctp_assoc_hashtable,
1536                    get_order(sctp_assoc_hashsize *
1537                              sizeof(struct sctp_hashbucket)));
1538         kfree(sctp_ep_hashtable);
1539         free_pages((unsigned long)sctp_port_hashtable,
1540                    get_order(sctp_port_hashsize *
1541                              sizeof(struct sctp_bind_hashbucket)));
1542
1543         sctp_dbg_objcnt_exit();
1544         sctp_proc_exit();
1545         cleanup_sctp_mibs();
1546
1547         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1548
1549         kmem_cache_destroy(sctp_chunk_cachep);
1550         kmem_cache_destroy(sctp_bucket_cachep);
1551 }
1552
1553 module_init(sctp_init);
1554 module_exit(sctp_exit);
1555
1556 /*
1557  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1558  */
1559 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1560 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1561 MODULE_AUTHOR("Linux Kernel SCTP developers <lksctp-developers@lists.sourceforge.net>");
1562 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1563 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1564 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1565 MODULE_LICENSE("GPL");