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                 /* Ensure the src address belongs to the output
 510                  * interface.
 511                  */
 512                 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
 513                                      false);
 514                 if (!odev || odev->ifindex != fl4->flowi4_oif)
 515                         continue;
 516
 517                 dst = &rt->dst;
 518                 break;
 519         }
 520
 521 out_unlock:
 522         rcu_read_unlock();
 523 out:
 524         t->dst = dst;
 525         if (dst)
 526                 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
 527                          &fl4->daddr, &fl4->saddr);
 528         else
 529                 pr_debug("no route\n");
 530 }
 531
 532 /* For v4, the source address is cached in the route entry(dst). So no need
 533  * to cache it separately and hence this is an empty routine.
 534  */
 535 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 536                               struct sctp_transport *t,
 537                               struct flowi *fl)
 538 {
 539         union sctp_addr *saddr = &t->saddr;
 540         struct rtable *rt = (struct rtable *)t->dst;
 541
 542         if (rt) {
 543                 saddr->v4.sin_family = AF_INET;
 544                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 545         }
 546 }
 547
 548 /* What interface did this skb arrive on? */
 549 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 550 {
 551         return inet_iif(skb);
 552 }
 553
 554 /* Was this packet marked by Explicit Congestion Notification? */
 555 static int sctp_v4_is_ce(const struct sk_buff *skb)
 556 {
 557         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 558 }
 559
 560 /* Create and initialize a new sk for the socket returned by accept(). */
 561 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 562                                              struct sctp_association *asoc)
 563 {
 564         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 565                         sk->sk_prot, 0);
 566         struct inet_sock *newinet;
 567
 568         if (!newsk)
 569                 goto out;
 570
 571         sock_init_data(NULL, newsk);
 572
 573         sctp_copy_sock(newsk, sk, asoc);
 574         sock_reset_flag(newsk, SOCK_ZAPPED);
 575
 576         newinet = inet_sk(newsk);
 577
 578         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 579
 580         sk_refcnt_debug_inc(newsk);
 581
 582         if (newsk->sk_prot->init(newsk)) {
 583                 sk_common_release(newsk);
 584                 newsk = NULL;
 585         }
 586
 587 out:
 588         return newsk;
 589 }
 590
 591 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
 592 {
 593         /* No address mapping for V4 sockets */
 594         return sizeof(struct sockaddr_in);
 595 }
 596
 597 /* Dump the v4 addr to the seq file. */
 598 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 599 {
 600         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 601 }
 602
 603 static void sctp_v4_ecn_capable(struct sock *sk)
 604 {
 605         INET_ECN_xmit(sk);
 606 }
 607
 608 static void sctp_addr_wq_timeout_handler(unsigned long arg)
 609 {
 610         struct net *net = (struct net *)arg;
 611         struct sctp_sockaddr_entry *addrw, *temp;
 612         struct sctp_sock *sp;
 613
 614         spin_lock_bh(&net->sctp.addr_wq_lock);
 615
 616         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 617                 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
 618                          "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
 619                          addrw->state, addrw);
 620
 621 #if IS_ENABLED(CONFIG_IPV6)
 622                 /* Now we send an ASCONF for each association */
 623                 /* Note. we currently don't handle link local IPv6 addressees */
 624                 if (addrw->a.sa.sa_family == AF_INET6) {
 625                         struct in6_addr *in6;
 626
 627                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 628                             IPV6_ADDR_LINKLOCAL)
 629                                 goto free_next;
 630
 631                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 632                         if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
 633                             addrw->state == SCTP_ADDR_NEW) {
 634                                 unsigned long timeo_val;
 635
 636                                 pr_debug("%s: this is on DAD, trying %d sec "
 637                                          "later\n", __func__,
 638                                          SCTP_ADDRESS_TICK_DELAY);
 639
 640                                 timeo_val = jiffies;
 641                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 642                                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 643                                 break;
 644                         }
 645                 }
 646 #endif
 647                 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
 648                         struct sock *sk;
 649
 650                         sk = sctp_opt2sk(sp);
 651                         /* ignore bound-specific endpoints */
 652                         if (!sctp_is_ep_boundall(sk))
 653                                 continue;
 654                         bh_lock_sock(sk);
 655                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 656                                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
 657                         bh_unlock_sock(sk);
 658                 }
 659 #if IS_ENABLED(CONFIG_IPV6)
 660 free_next:
 661 #endif
 662                 list_del(&addrw->list);
 663                 kfree(addrw);
 664         }
 665         spin_unlock_bh(&net->sctp.addr_wq_lock);
 666 }
 667
 668 static void sctp_free_addr_wq(struct net *net)
 669 {
 670         struct sctp_sockaddr_entry *addrw;
 671         struct sctp_sockaddr_entry *temp;
 672
 673         spin_lock_bh(&net->sctp.addr_wq_lock);
 674         del_timer(&net->sctp.addr_wq_timer);
 675         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 676                 list_del(&addrw->list);
 677                 kfree(addrw);
 678         }
 679         spin_unlock_bh(&net->sctp.addr_wq_lock);
 680 }
 681
 682 /* lookup the entry for the same address in the addr_waitq
 683  * sctp_addr_wq MUST be locked
 684  */
 685 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
 686                                         struct sctp_sockaddr_entry *addr)
 687 {
 688         struct sctp_sockaddr_entry *addrw;
 689
 690         list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
 691                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 692                         continue;
 693                 if (addrw->a.sa.sa_family == AF_INET) {
 694                         if (addrw->a.v4.sin_addr.s_addr ==
 695                             addr->a.v4.sin_addr.s_addr)
 696                                 return addrw;
 697                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 698                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 699                             &addr->a.v6.sin6_addr))
 700                                 return addrw;
 701                 }
 702         }
 703         return NULL;
 704 }
 705
 706 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
 707 {
 708         struct sctp_sockaddr_entry *addrw;
 709         unsigned long timeo_val;
 710
 711         /* first, we check if an opposite message already exist in the queue.
 712          * If we found such message, it is removed.
 713          * This operation is a bit stupid, but the DHCP client attaches the
 714          * new address after a couple of addition and deletion of that address
 715          */
 716
 717         spin_lock_bh(&net->sctp.addr_wq_lock);
 718         /* Offsets existing events in addr_wq */
 719         addrw = sctp_addr_wq_lookup(net, addr);
 720         if (addrw) {
 721                 if (addrw->state != cmd) {
 722                         pr_debug("%s: offsets existing entry for %d, addr:%pISc "
 723                                  "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
 724                                  &net->sctp.addr_waitq);
 725
 726                         list_del(&addrw->list);
 727                         kfree(addrw);
 728                 }
 729                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 730                 return;
 731         }
 732
 733         /* OK, we have to add the new address to the wait queue */
 734         addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 735         if (addrw == NULL) {
 736                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 737                 return;
 738         }
 739         addrw->state = cmd;
 740         list_add_tail(&addrw->list, &net->sctp.addr_waitq);
 741
 742         pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
 743                  __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
 744
 745         if (!timer_pending(&net->sctp.addr_wq_timer)) {
 746                 timeo_val = jiffies;
 747                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 748                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 749         }
 750         spin_unlock_bh(&net->sctp.addr_wq_lock);
 751 }
 752
 753 /* Event handler for inet address addition/deletion events.
 754  * The sctp_local_addr_list needs to be protocted by a spin lock since
 755  * multiple notifiers (say IPv4 and IPv6) may be running at the same
 756  * time and thus corrupt the list.
 757  * The reader side is protected with RCU.
 758  */
 759 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
 760                                void *ptr)
 761 {
 762         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
 763         struct sctp_sockaddr_entry *addr = NULL;
 764         struct sctp_sockaddr_entry *temp;
 765         struct net *net = dev_net(ifa->ifa_dev->dev);
 766         int found = 0;
 767
 768         switch (ev) {
 769         case NETDEV_UP:
 770                 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 771                 if (addr) {
 772                         addr->a.v4.sin_family = AF_INET;
 773                         addr->a.v4.sin_port = 0;
 774                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 775                         addr->valid = 1;
 776                         spin_lock_bh(&net->sctp.local_addr_lock);
 777                         list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
 778                         sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
 779                         spin_unlock_bh(&net->sctp.local_addr_lock);
 780                 }
 781                 break;
 782         case NETDEV_DOWN:
 783                 spin_lock_bh(&net->sctp.local_addr_lock);
 784                 list_for_each_entry_safe(addr, temp,
 785                                         &net->sctp.local_addr_list, list) {
 786                         if (addr->a.sa.sa_family == AF_INET &&
 787                                         addr->a.v4.sin_addr.s_addr ==
 788                                         ifa->ifa_local) {
 789                                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
 790                                 found = 1;
 791                                 addr->valid = 0;
 792                                 list_del_rcu(&addr->list);
 793                                 break;
 794                         }
 795                 }
 796                 spin_unlock_bh(&net->sctp.local_addr_lock);
 797                 if (found)
 798                         kfree_rcu(addr, rcu);
 799                 break;
 800         }
 801
 802         return NOTIFY_DONE;
 803 }
 804
 805 /*
 806  * Initialize the control inode/socket with a control endpoint data
 807  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 808  */
 809 static int sctp_ctl_sock_init(struct net *net)
 810 {
 811         int err;
 812         sa_family_t family = PF_INET;
 813
 814         if (sctp_get_pf_specific(PF_INET6))
 815                 family = PF_INET6;
 816
 817         err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
 818                                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
 819
 820         /* If IPv6 socket could not be created, try the IPv4 socket */
 821         if (err < 0 && family == PF_INET6)
 822                 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
 823                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 824                                            net);
 825
 826         if (err < 0) {
 827                 pr_err("Failed to create the SCTP control socket\n");
 828                 return err;
 829         }
 830         return 0;
 831 }
 832
 833 /* Register address family specific functions. */
 834 int sctp_register_af(struct sctp_af *af)
 835 {
 836         switch (af->sa_family) {
 837         case AF_INET:
 838                 if (sctp_af_v4_specific)
 839                         return 0;
 840                 sctp_af_v4_specific = af;
 841                 break;
 842         case AF_INET6:
 843                 if (sctp_af_v6_specific)
 844                         return 0;
 845                 sctp_af_v6_specific = af;
 846                 break;
 847         default:
 848                 return 0;
 849         }
 850
 851         INIT_LIST_HEAD(&af->list);
 852         list_add_tail(&af->list, &sctp_address_families);
 853         return 1;
 854 }
 855
 856 /* Get the table of functions for manipulating a particular address
 857  * family.
 858  */
 859 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 860 {
 861         switch (family) {
 862         case AF_INET:
 863                 return sctp_af_v4_specific;
 864         case AF_INET6:
 865                 return sctp_af_v6_specific;
 866         default:
 867                 return NULL;
 868         }
 869 }
 870
 871 /* Common code to initialize a AF_INET msg_name. */
 872 static void sctp_inet_msgname(char *msgname, int *addr_len)
 873 {
 874         struct sockaddr_in *sin;
 875
 876         sin = (struct sockaddr_in *)msgname;
 877         *addr_len = sizeof(struct sockaddr_in);
 878         sin->sin_family = AF_INET;
 879         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 880 }
 881
 882 /* Copy the primary address of the peer primary address as the msg_name. */
 883 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 884                                     int *addr_len)
 885 {
 886         struct sockaddr_in *sin, *sinfrom;
 887
 888         if (msgname) {
 889                 struct sctp_association *asoc;
 890
 891                 asoc = event->asoc;
 892                 sctp_inet_msgname(msgname, addr_len);
 893                 sin = (struct sockaddr_in *)msgname;
 894                 sinfrom = &asoc->peer.primary_addr.v4;
 895                 sin->sin_port = htons(asoc->peer.port);
 896                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 897         }
 898 }
 899
 900 /* Initialize and copy out a msgname from an inbound skb. */
 901 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 902 {
 903         if (msgname) {
 904                 struct sctphdr *sh = sctp_hdr(skb);
 905                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 906
 907                 sctp_inet_msgname(msgname, len);
 908                 sin->sin_port = sh->source;
 909                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 910         }
 911 }
 912
 913 /* Do we support this AF? */
 914 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 915 {
 916         /* PF_INET only supports AF_INET addresses. */
 917         return AF_INET == family;
 918 }
 919
 920 /* Address matching with wildcards allowed. */
 921 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 922                               const union sctp_addr *addr2,
 923                               struct sctp_sock *opt)
 924 {
 925         /* PF_INET only supports AF_INET addresses. */
 926         if (addr1->sa.sa_family != addr2->sa.sa_family)
 927                 return 0;
 928         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 929             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 930                 return 1;
 931         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 932                 return 1;
 933
 934         return 0;
 935 }
 936
 937 /* Verify that provided sockaddr looks bindable.  Common verification has
 938  * already been taken care of.
 939  */
 940 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 941 {
 942         return sctp_v4_available(addr, opt);
 943 }
 944
 945 /* Verify that sockaddr looks sendable.  Common verification has already
 946  * been taken care of.
 947  */
 948 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 949 {
 950         return 1;
 951 }
 952
 953 /* Fill in Supported Address Type information for INIT and INIT-ACK
 954  * chunks.  Returns number of addresses supported.
 955  */
 956 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 957                                      __be16 *types)
 958 {
 959         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 960         return 1;
 961 }
 962
 963 /* Wrapper routine that calls the ip transmit routine. */
 964 static inline int sctp_v4_xmit(struct sk_buff *skb,
 965                                struct sctp_transport *transport)
 966 {
 967         struct inet_sock *inet = inet_sk(skb->sk);
 968
 969         pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
 970                  skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
 971
 972         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
 973                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 974
 975         SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
 976
 977         return ip_queue_xmit(&inet->sk, skb, &transport->fl);
 978 }
 979
 980 static struct sctp_af sctp_af_inet;
 981
 982 static struct sctp_pf sctp_pf_inet = {
 983         .event_msgname = sctp_inet_event_msgname,
 984         .skb_msgname   = sctp_inet_skb_msgname,
 985         .af_supported  = sctp_inet_af_supported,
 986         .cmp_addr      = sctp_inet_cmp_addr,
 987         .bind_verify   = sctp_inet_bind_verify,
 988         .send_verify   = sctp_inet_send_verify,
 989         .supported_addrs = sctp_inet_supported_addrs,
 990         .create_accept_sk = sctp_v4_create_accept_sk,
 991         .addr_to_user  = sctp_v4_addr_to_user,
 992         .to_sk_saddr   = sctp_v4_to_sk_saddr,
 993         .to_sk_daddr   = sctp_v4_to_sk_daddr,
 994         .af            = &sctp_af_inet
 995 };
 996
 997 /* Notifier for inetaddr addition/deletion events.  */
 998 static struct notifier_block sctp_inetaddr_notifier = {
 999         .notifier_call = sctp_inetaddr_event,
1000 };
1001
1002 /* Socket operations.  */
1003 static const struct proto_ops inet_seqpacket_ops = {
1004         .family            = PF_INET,
1005         .owner             = THIS_MODULE,
1006         .release           = inet_release,      /* Needs to be wrapped... */
1007         .bind              = inet_bind,
1008         .connect           = inet_dgram_connect,
1009         .socketpair        = sock_no_socketpair,
1010         .accept            = inet_accept,
1011         .getname           = inet_getname,      /* Semantics are different.  */
1012         .poll              = sctp_poll,
1013         .ioctl             = inet_ioctl,
1014         .listen            = sctp_inet_listen,
1015         .shutdown          = inet_shutdown,     /* Looks harmless.  */
1016         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1017         .getsockopt        = sock_common_getsockopt,
1018         .sendmsg           = inet_sendmsg,
1019         .recvmsg           = sock_common_recvmsg,
1020         .mmap              = sock_no_mmap,
1021         .sendpage          = sock_no_sendpage,
1022 #ifdef CONFIG_COMPAT
1023         .compat_setsockopt = compat_sock_common_setsockopt,
1024         .compat_getsockopt = compat_sock_common_getsockopt,
1025 #endif
1026 };
1027
1028 /* Registration with AF_INET family.  */
1029 static struct inet_protosw sctp_seqpacket_protosw = {
1030         .type       = SOCK_SEQPACKET,
1031         .protocol   = IPPROTO_SCTP,
1032         .prot       = &sctp_prot,
1033         .ops        = &inet_seqpacket_ops,
1034         .flags      = SCTP_PROTOSW_FLAG
1035 };
1036 static struct inet_protosw sctp_stream_protosw = {
1037         .type       = SOCK_STREAM,
1038         .protocol   = IPPROTO_SCTP,
1039         .prot       = &sctp_prot,
1040         .ops        = &inet_seqpacket_ops,
1041         .flags      = SCTP_PROTOSW_FLAG
1042 };
1043
1044 /* Register with IP layer.  */
1045 static const struct net_protocol sctp_protocol = {
1046         .handler     = sctp_rcv,
1047         .err_handler = sctp_v4_err,
1048         .no_policy   = 1,
1049         .netns_ok    = 1,
1050         .icmp_strict_tag_validation = 1,
1051 };
1052
1053 /* IPv4 address related functions.  */
1054 static struct sctp_af sctp_af_inet = {
1055         .sa_family         = AF_INET,
1056         .sctp_xmit         = sctp_v4_xmit,
1057         .setsockopt        = ip_setsockopt,
1058         .getsockopt        = ip_getsockopt,
1059         .get_dst           = sctp_v4_get_dst,
1060         .get_saddr         = sctp_v4_get_saddr,
1061         .copy_addrlist     = sctp_v4_copy_addrlist,
1062         .from_skb          = sctp_v4_from_skb,
1063         .from_sk           = sctp_v4_from_sk,
1064         .from_addr_param   = sctp_v4_from_addr_param,
1065         .to_addr_param     = sctp_v4_to_addr_param,
1066         .cmp_addr          = sctp_v4_cmp_addr,
1067         .addr_valid        = sctp_v4_addr_valid,
1068         .inaddr_any        = sctp_v4_inaddr_any,
1069         .is_any            = sctp_v4_is_any,
1070         .available         = sctp_v4_available,
1071         .scope             = sctp_v4_scope,
1072         .skb_iif           = sctp_v4_skb_iif,
1073         .is_ce             = sctp_v4_is_ce,
1074         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1075         .ecn_capable       = sctp_v4_ecn_capable,
1076         .net_header_len    = sizeof(struct iphdr),
1077         .sockaddr_len      = sizeof(struct sockaddr_in),
1078 #ifdef CONFIG_COMPAT
1079         .compat_setsockopt = compat_ip_setsockopt,
1080         .compat_getsockopt = compat_ip_getsockopt,
1081 #endif
1082 };
1083
1084 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1085 {
1086         switch (family) {
1087         case PF_INET:
1088                 return sctp_pf_inet_specific;
1089         case PF_INET6:
1090                 return sctp_pf_inet6_specific;
1091         default:
1092                 return NULL;
1093         }
1094 }
1095
1096 /* Register the PF specific function table.  */
1097 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1098 {
1099         switch (family) {
1100         case PF_INET:
1101                 if (sctp_pf_inet_specific)
1102                         return 0;
1103                 sctp_pf_inet_specific = pf;
1104                 break;
1105         case PF_INET6:
1106                 if (sctp_pf_inet6_specific)
1107                         return 0;
1108                 sctp_pf_inet6_specific = pf;
1109                 break;
1110         default:
1111                 return 0;
1112         }
1113         return 1;
1114 }
1115
1116 static inline int init_sctp_mibs(struct net *net)
1117 {
1118         net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1119         if (!net->sctp.sctp_statistics)
1120                 return -ENOMEM;
1121         return 0;
1122 }
1123
1124 static inline void cleanup_sctp_mibs(struct net *net)
1125 {
1126         free_percpu(net->sctp.sctp_statistics);
1127 }
1128
1129 static void sctp_v4_pf_init(void)
1130 {
1131         /* Initialize the SCTP specific PF functions. */
1132         sctp_register_pf(&sctp_pf_inet, PF_INET);
1133         sctp_register_af(&sctp_af_inet);
1134 }
1135
1136 static void sctp_v4_pf_exit(void)
1137 {
1138         list_del(&sctp_af_inet.list);
1139 }
1140
1141 static int sctp_v4_protosw_init(void)
1142 {
1143         int rc;
1144
1145         rc = proto_register(&sctp_prot, 1);
1146         if (rc)
1147                 return rc;
1148
1149         /* Register SCTP(UDP and TCP style) with socket layer.  */
1150         inet_register_protosw(&sctp_seqpacket_protosw);
1151         inet_register_protosw(&sctp_stream_protosw);
1152
1153         return 0;
1154 }
1155
1156 static void sctp_v4_protosw_exit(void)
1157 {
1158         inet_unregister_protosw(&sctp_stream_protosw);
1159         inet_unregister_protosw(&sctp_seqpacket_protosw);
1160         proto_unregister(&sctp_prot);
1161 }
1162
1163 static int sctp_v4_add_protocol(void)
1164 {
1165         /* Register notifier for inet address additions/deletions. */
1166         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1167
1168         /* Register SCTP with inet layer.  */
1169         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1170                 return -EAGAIN;
1171
1172         return 0;
1173 }
1174
1175 static void sctp_v4_del_protocol(void)
1176 {
1177         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1178         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1179 }
1180
1181 static int __net_init sctp_net_init(struct net *net)
1182 {
1183         int status;
1184
1185         /*
1186          * 14. Suggested SCTP Protocol Parameter Values
1187          */
1188         /* The following protocol parameters are RECOMMENDED:  */
1189         /* RTO.Initial              - 3  seconds */
1190         net->sctp.rto_initial                   = SCTP_RTO_INITIAL;
1191         /* RTO.Min                  - 1  second */
1192         net->sctp.rto_min                       = SCTP_RTO_MIN;
1193         /* RTO.Max                 -  60 seconds */
1194         net->sctp.rto_max                       = SCTP_RTO_MAX;
1195         /* RTO.Alpha                - 1/8 */
1196         net->sctp.rto_alpha                     = SCTP_RTO_ALPHA;
1197         /* RTO.Beta                 - 1/4 */
1198         net->sctp.rto_beta                      = SCTP_RTO_BETA;
1199
1200         /* Valid.Cookie.Life        - 60  seconds */
1201         net->sctp.valid_cookie_life             = SCTP_DEFAULT_COOKIE_LIFE;
1202
1203         /* Whether Cookie Preservative is enabled(1) or not(0) */
1204         net->sctp.cookie_preserve_enable        = 1;
1205
1206         /* Default sctp sockets to use md5 as their hmac alg */
1207 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1208         net->sctp.sctp_hmac_alg                 = "md5";
1209 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1210         net->sctp.sctp_hmac_alg                 = "sha1";
1211 #else
1212         net->sctp.sctp_hmac_alg                 = NULL;
1213 #endif
1214
1215         /* Max.Burst                - 4 */
1216         net->sctp.max_burst                     = SCTP_DEFAULT_MAX_BURST;
1217
1218         /* Association.Max.Retrans  - 10 attempts
1219          * Path.Max.Retrans         - 5  attempts (per destination address)
1220          * Max.Init.Retransmits     - 8  attempts
1221          */
1222         net->sctp.max_retrans_association       = 10;
1223         net->sctp.max_retrans_path              = 5;
1224         net->sctp.max_retrans_init              = 8;
1225
1226         /* Sendbuffer growth        - do per-socket accounting */
1227         net->sctp.sndbuf_policy                 = 0;
1228
1229         /* Rcvbuffer growth         - do per-socket accounting */
1230         net->sctp.rcvbuf_policy                 = 0;
1231
1232         /* HB.interval              - 30 seconds */
1233         net->sctp.hb_interval                   = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1234
1235         /* delayed SACK timeout */
1236         net->sctp.sack_timeout                  = SCTP_DEFAULT_TIMEOUT_SACK;
1237
1238         /* Disable ADDIP by default. */
1239         net->sctp.addip_enable = 0;
1240         net->sctp.addip_noauth = 0;
1241         net->sctp.default_auto_asconf = 0;
1242
1243         /* Enable PR-SCTP by default. */
1244         net->sctp.prsctp_enable = 1;
1245
1246         /* Disable AUTH by default. */
1247         net->sctp.auth_enable = 0;
1248
1249         /* Set SCOPE policy to enabled */
1250         net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1251
1252         /* Set the default rwnd update threshold */
1253         net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1254
1255         /* Initialize maximum autoclose timeout. */
1256         net->sctp.max_autoclose         = INT_MAX / HZ;
1257
1258         status = sctp_sysctl_net_register(net);
1259         if (status)
1260                 goto err_sysctl_register;
1261
1262         /* Allocate and initialise sctp mibs.  */
1263         status = init_sctp_mibs(net);
1264         if (status)
1265                 goto err_init_mibs;
1266
1267         /* Initialize proc fs directory.  */
1268         status = sctp_proc_init(net);
1269         if (status)
1270                 goto err_init_proc;
1271
1272         sctp_dbg_objcnt_init(net);
1273
1274         /* Initialize the control inode/socket for handling OOTB packets.  */
1275         if ((status = sctp_ctl_sock_init(net))) {
1276                 pr_err("Failed to initialize the SCTP control sock\n");
1277                 goto err_ctl_sock_init;
1278         }
1279
1280         /* Initialize the local address list. */
1281         INIT_LIST_HEAD(&net->sctp.local_addr_list);
1282         spin_lock_init(&net->sctp.local_addr_lock);
1283         sctp_get_local_addr_list(net);
1284
1285         /* Initialize the address event list */
1286         INIT_LIST_HEAD(&net->sctp.addr_waitq);
1287         INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1288         spin_lock_init(&net->sctp.addr_wq_lock);
1289         net->sctp.addr_wq_timer.expires = 0;
1290         setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1291                     (unsigned long)net);
1292
1293         return 0;
1294
1295 err_ctl_sock_init:
1296         sctp_dbg_objcnt_exit(net);
1297         sctp_proc_exit(net);
1298 err_init_proc:
1299         cleanup_sctp_mibs(net);
1300 err_init_mibs:
1301         sctp_sysctl_net_unregister(net);
1302 err_sysctl_register:
1303         return status;
1304 }
1305
1306 static void __net_exit sctp_net_exit(struct net *net)
1307 {
1308         /* Free the local address list */
1309         sctp_free_addr_wq(net);
1310         sctp_free_local_addr_list(net);
1311
1312         /* Free the control endpoint.  */
1313         inet_ctl_sock_destroy(net->sctp.ctl_sock);
1314
1315         sctp_dbg_objcnt_exit(net);
1316
1317         sctp_proc_exit(net);
1318         cleanup_sctp_mibs(net);
1319         sctp_sysctl_net_unregister(net);
1320 }
1321
1322 static struct pernet_operations sctp_net_ops = {
1323         .init = sctp_net_init,
1324         .exit = sctp_net_exit,
1325 };
1326
1327 /* Initialize the universe into something sensible.  */
1328 static __init int sctp_init(void)
1329 {
1330         int i;
1331         int status = -EINVAL;
1332         unsigned long goal;
1333         unsigned long limit;
1334         int max_share;
1335         int order;
1336
1337         sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1338
1339         /* Allocate bind_bucket and chunk caches. */
1340         status = -ENOBUFS;
1341         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1342                                                sizeof(struct sctp_bind_bucket),
1343                                                0, SLAB_HWCACHE_ALIGN,
1344                                                NULL);
1345         if (!sctp_bucket_cachep)
1346                 goto out;
1347
1348         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1349                                                sizeof(struct sctp_chunk),
1350                                                0, SLAB_HWCACHE_ALIGN,
1351                                                NULL);
1352         if (!sctp_chunk_cachep)
1353                 goto err_chunk_cachep;
1354
1355         status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1356         if (status)
1357                 goto err_percpu_counter_init;
1358
1359         /* Implementation specific variables. */
1360
1361         /* Initialize default stream count setup information. */
1362         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1363         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1364
1365         /* Initialize handle used for association ids. */
1366         idr_init(&sctp_assocs_id);
1367
1368         limit = nr_free_buffer_pages() / 8;
1369         limit = max(limit, 128UL);
1370         sysctl_sctp_mem[0] = limit / 4 * 3;
1371         sysctl_sctp_mem[1] = limit;
1372         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1373
1374         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1375         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1376         max_share = min(4UL*1024*1024, limit);
1377
1378         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1379         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1380         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1381
1382         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1383         sysctl_sctp_wmem[1] = 16*1024;
1384         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1385
1386         /* Size and allocate the association hash table.
1387          * The methodology is similar to that of the tcp hash tables.
1388          */
1389         if (totalram_pages >= (128 * 1024))
1390                 goal = totalram_pages >> (22 - PAGE_SHIFT);
1391         else
1392                 goal = totalram_pages >> (24 - PAGE_SHIFT);
1393
1394         for (order = 0; (1UL << order) < goal; order++)
1395                 ;
1396
1397         do {
1398                 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1399                                         sizeof(struct sctp_hashbucket);
1400                 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1401                         continue;
1402                 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1403                         __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1404         } while (!sctp_assoc_hashtable && --order > 0);
1405         if (!sctp_assoc_hashtable) {
1406                 pr_err("Failed association hash alloc\n");
1407                 status = -ENOMEM;
1408                 goto err_ahash_alloc;
1409         }
1410         for (i = 0; i < sctp_assoc_hashsize; i++) {
1411                 rwlock_init(&sctp_assoc_hashtable[i].lock);
1412                 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1413         }
1414
1415         /* Allocate and initialize the endpoint hash table.  */
1416         sctp_ep_hashsize = 64;
1417         sctp_ep_hashtable =
1418                 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1419         if (!sctp_ep_hashtable) {
1420                 pr_err("Failed endpoint_hash alloc\n");
1421                 status = -ENOMEM;
1422                 goto err_ehash_alloc;
1423         }
1424         for (i = 0; i < sctp_ep_hashsize; i++) {
1425                 rwlock_init(&sctp_ep_hashtable[i].lock);
1426                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1427         }
1428
1429         /* Allocate and initialize the SCTP port hash table.  */
1430         do {
1431                 sctp_port_hashsize = (1UL << order) * PAGE_SIZE /
1432                                         sizeof(struct sctp_bind_hashbucket);
1433                 if ((sctp_port_hashsize > (64 * 1024)) && order > 0)
1434                         continue;
1435                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1436                         __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1437         } while (!sctp_port_hashtable && --order > 0);
1438         if (!sctp_port_hashtable) {
1439                 pr_err("Failed bind hash alloc\n");
1440                 status = -ENOMEM;
1441                 goto err_bhash_alloc;
1442         }
1443         for (i = 0; i < sctp_port_hashsize; i++) {
1444                 spin_lock_init(&sctp_port_hashtable[i].lock);
1445                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1446         }
1447
1448         pr_info("Hash tables configured (established %d bind %d)\n",
1449                 sctp_assoc_hashsize, sctp_port_hashsize);
1450
1451         sctp_sysctl_register();
1452
1453         INIT_LIST_HEAD(&sctp_address_families);
1454         sctp_v4_pf_init();
1455         sctp_v6_pf_init();
1456
1457         status = sctp_v4_protosw_init();
1458
1459         if (status)
1460                 goto err_protosw_init;
1461
1462         status = sctp_v6_protosw_init();
1463         if (status)
1464                 goto err_v6_protosw_init;
1465
1466         status = register_pernet_subsys(&sctp_net_ops);
1467         if (status)
1468                 goto err_register_pernet_subsys;
1469
1470         status = sctp_v4_add_protocol();
1471         if (status)
1472                 goto err_add_protocol;
1473
1474         /* Register SCTP with inet6 layer.  */
1475         status = sctp_v6_add_protocol();
1476         if (status)
1477                 goto err_v6_add_protocol;
1478
1479 out:
1480         return status;
1481 err_v6_add_protocol:
1482         sctp_v4_del_protocol();
1483 err_add_protocol:
1484         unregister_pernet_subsys(&sctp_net_ops);
1485 err_register_pernet_subsys:
1486         sctp_v6_protosw_exit();
1487 err_v6_protosw_init:
1488         sctp_v4_protosw_exit();
1489 err_protosw_init:
1490         sctp_v4_pf_exit();
1491         sctp_v6_pf_exit();
1492         sctp_sysctl_unregister();
1493         free_pages((unsigned long)sctp_port_hashtable,
1494                    get_order(sctp_port_hashsize *
1495                              sizeof(struct sctp_bind_hashbucket)));
1496 err_bhash_alloc:
1497         kfree(sctp_ep_hashtable);
1498 err_ehash_alloc:
1499         free_pages((unsigned long)sctp_assoc_hashtable,
1500                    get_order(sctp_assoc_hashsize *
1501                              sizeof(struct sctp_hashbucket)));
1502 err_ahash_alloc:
1503         percpu_counter_destroy(&sctp_sockets_allocated);
1504 err_percpu_counter_init:
1505         kmem_cache_destroy(sctp_chunk_cachep);
1506 err_chunk_cachep:
1507         kmem_cache_destroy(sctp_bucket_cachep);
1508         goto out;
1509 }
1510
1511 /* Exit handler for the SCTP protocol.  */
1512 static __exit void sctp_exit(void)
1513 {
1514         /* BUG.  This should probably do something useful like clean
1515          * up all the remaining associations and all that memory.
1516          */
1517
1518         /* Unregister with inet6/inet layers. */
1519         sctp_v6_del_protocol();
1520         sctp_v4_del_protocol();
1521
1522         unregister_pernet_subsys(&sctp_net_ops);
1523
1524         /* Free protosw registrations */
1525         sctp_v6_protosw_exit();
1526         sctp_v4_protosw_exit();
1527
1528         /* Unregister with socket layer. */
1529         sctp_v6_pf_exit();
1530         sctp_v4_pf_exit();
1531
1532         sctp_sysctl_unregister();
1533
1534         free_pages((unsigned long)sctp_assoc_hashtable,
1535                    get_order(sctp_assoc_hashsize *
1536                              sizeof(struct sctp_hashbucket)));
1537         kfree(sctp_ep_hashtable);
1538         free_pages((unsigned long)sctp_port_hashtable,
1539                    get_order(sctp_port_hashsize *
1540                              sizeof(struct sctp_bind_hashbucket)));
1541
1542         percpu_counter_destroy(&sctp_sockets_allocated);
1543
1544         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1545
1546         kmem_cache_destroy(sctp_chunk_cachep);
1547         kmem_cache_destroy(sctp_bucket_cachep);
1548 }
1549
1550 module_init(sctp_init);
1551 module_exit(sctp_exit);
1552
1553 /*
1554  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1555  */
1556 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1557 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1558 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1559 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1560 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1561 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1562 MODULE_LICENSE("GPL");