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 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
  64
  65 /* Global data structures. */
  66 struct sctp_globals sctp_globals __read_mostly;
  67
  68 struct idr sctp_assocs_id;
  69 DEFINE_SPINLOCK(sctp_assocs_id_lock);
  70
  71 static struct sctp_pf *sctp_pf_inet6_specific;
  72 static struct sctp_pf *sctp_pf_inet_specific;
  73 static struct sctp_af *sctp_af_v4_specific;
  74 static struct sctp_af *sctp_af_v6_specific;
  75
  76 struct kmem_cache *sctp_chunk_cachep __read_mostly;
  77 struct kmem_cache *sctp_bucket_cachep __read_mostly;
  78
  79 long sysctl_sctp_mem[3];
  80 int sysctl_sctp_rmem[3];
  81 int sysctl_sctp_wmem[3];
  82
  83 /* Private helper to extract ipv4 address and stash them in
  84  * the protocol structure.
  85  */
  86 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
  87                                   struct net_device *dev)
  88 {
  89         struct in_device *in_dev;
  90         struct in_ifaddr *ifa;
  91         struct sctp_sockaddr_entry *addr;
  92
  93         rcu_read_lock();
  94         if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
  95                 rcu_read_unlock();
  96                 return;
  97         }
  98
  99         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
 100                 /* Add the address to the local list.  */
 101                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
 102                 if (addr) {
 103                         addr->a.v4.sin_family = AF_INET;
 104                         addr->a.v4.sin_port = 0;
 105                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 106                         addr->valid = 1;
 107                         INIT_LIST_HEAD(&addr->list);
 108                         list_add_tail(&addr->list, addrlist);
 109                 }
 110         }
 111
 112         rcu_read_unlock();
 113 }
 114
 115 /* Extract our IP addresses from the system and stash them in the
 116  * protocol structure.
 117  */
 118 static void sctp_get_local_addr_list(struct net *net)
 119 {
 120         struct net_device *dev;
 121         struct list_head *pos;
 122         struct sctp_af *af;
 123
 124         rcu_read_lock();
 125         for_each_netdev_rcu(net, dev) {
 126                 list_for_each(pos, &sctp_address_families) {
 127                         af = list_entry(pos, struct sctp_af, list);
 128                         af->copy_addrlist(&net->sctp.local_addr_list, dev);
 129                 }
 130         }
 131         rcu_read_unlock();
 132 }
 133
 134 /* Free the existing local addresses.  */
 135 static void sctp_free_local_addr_list(struct net *net)
 136 {
 137         struct sctp_sockaddr_entry *addr;
 138         struct list_head *pos, *temp;
 139
 140         list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
 141                 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
 142                 list_del(pos);
 143                 kfree(addr);
 144         }
 145 }
 146
 147 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
 148 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
 149                               enum sctp_scope scope, gfp_t gfp, int copy_flags)
 150 {
 151         struct sctp_sockaddr_entry *addr;
 152         union sctp_addr laddr;
 153         int error = 0;
 154
 155         rcu_read_lock();
 156         list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
 157                 if (!addr->valid)
 158                         continue;
 159                 if (!sctp_in_scope(net, &addr->a, scope))
 160                         continue;
 161
 162                 /* Now that the address is in scope, check to see if
 163                  * the address type is really supported by the local
 164                  * sock as well as the remote peer.
 165                  */
 166                 if (addr->a.sa.sa_family == AF_INET &&
 167                     !(copy_flags & SCTP_ADDR4_PEERSUPP))
 168                         continue;
 169                 if (addr->a.sa.sa_family == AF_INET6 &&
 170                     (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
 171                      !(copy_flags & SCTP_ADDR6_PEERSUPP)))
 172                         continue;
 173
 174                 laddr = addr->a;
 175                 /* also works for setting ipv6 address port */
 176                 laddr.v4.sin_port = htons(bp->port);
 177                 if (sctp_bind_addr_state(bp, &laddr) != -1)
 178                         continue;
 179
 180                 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
 181                                            SCTP_ADDR_SRC, GFP_ATOMIC);
 182                 if (error)
 183                         break;
 184         }
 185
 186         rcu_read_unlock();
 187         return error;
 188 }
 189
 190 /* Initialize a sctp_addr from in incoming skb.  */
 191 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
 192                              int is_saddr)
 193 {
 194         /* Always called on head skb, so this is safe */
 195         struct sctphdr *sh = sctp_hdr(skb);
 196         struct sockaddr_in *sa = &addr->v4;
 197
 198         addr->v4.sin_family = AF_INET;
 199
 200         if (is_saddr) {
 201                 sa->sin_port = sh->source;
 202                 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
 203         } else {
 204                 sa->sin_port = sh->dest;
 205                 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
 206         }
 207 }
 208
 209 /* Initialize an sctp_addr from a socket. */
 210 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
 211 {
 212         addr->v4.sin_family = AF_INET;
 213         addr->v4.sin_port = 0;
 214         addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
 215 }
 216
 217 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
 218 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
 219 {
 220         inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
 221 }
 222
 223 /* Initialize sk->sk_daddr from sctp_addr. */
 224 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
 225 {
 226         inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
 227 }
 228
 229 /* Initialize a sctp_addr from an address parameter. */
 230 static void sctp_v4_from_addr_param(union sctp_addr *addr,
 231                                     union sctp_addr_param *param,
 232                                     __be16 port, int iif)
 233 {
 234         addr->v4.sin_family = AF_INET;
 235         addr->v4.sin_port = port;
 236         addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
 237 }
 238
 239 /* Initialize an address parameter from a sctp_addr and return the length
 240  * of the address parameter.
 241  */
 242 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
 243                                  union sctp_addr_param *param)
 244 {
 245         int length = sizeof(struct sctp_ipv4addr_param);
 246
 247         param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
 248         param->v4.param_hdr.length = htons(length);
 249         param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
 250
 251         return length;
 252 }
 253
 254 /* Initialize a sctp_addr from a dst_entry. */
 255 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
 256                               __be16 port)
 257 {
 258         saddr->v4.sin_family = AF_INET;
 259         saddr->v4.sin_port = port;
 260         saddr->v4.sin_addr.s_addr = fl4->saddr;
 261 }
 262
 263 /* Compare two addresses exactly. */
 264 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
 265                             const union sctp_addr *addr2)
 266 {
 267         if (addr1->sa.sa_family != addr2->sa.sa_family)
 268                 return 0;
 269         if (addr1->v4.sin_port != addr2->v4.sin_port)
 270                 return 0;
 271         if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
 272                 return 0;
 273
 274         return 1;
 275 }
 276
 277 /* Initialize addr struct to INADDR_ANY. */
 278 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
 279 {
 280         addr->v4.sin_family = AF_INET;
 281         addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
 282         addr->v4.sin_port = port;
 283 }
 284
 285 /* Is this a wildcard address? */
 286 static int sctp_v4_is_any(const union sctp_addr *addr)
 287 {
 288         return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
 289 }
 290
 291 /* This function checks if the address is a valid address to be used for
 292  * SCTP binding.
 293  *
 294  * Output:
 295  * Return 0 - If the address is a non-unicast or an illegal address.
 296  * Return 1 - If the address is a unicast.
 297  */
 298 static int sctp_v4_addr_valid(union sctp_addr *addr,
 299                               struct sctp_sock *sp,
 300                               const struct sk_buff *skb)
 301 {
 302         /* IPv4 addresses not allowed */
 303         if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
 304                 return 0;
 305
 306         /* Is this a non-unicast address or a unusable SCTP address? */
 307         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
 308                 return 0;
 309
 310         /* Is this a broadcast address? */
 311         if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
 312                 return 0;
 313
 314         return 1;
 315 }
 316
 317 /* Should this be available for binding?   */
 318 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
 319 {
 320         struct net *net = sock_net(&sp->inet.sk);
 321         int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
 322
 323
 324         if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
 325            ret != RTN_LOCAL &&
 326            !sp->inet.freebind &&
 327            !net->ipv4.sysctl_ip_nonlocal_bind)
 328                 return 0;
 329
 330         if (ipv6_only_sock(sctp_opt2sk(sp)))
 331                 return 0;
 332
 333         return 1;
 334 }
 335
 336 /* Checking the loopback, private and other address scopes as defined in
 337  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
 338  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
 339  *
 340  * Level 0 - unusable SCTP addresses
 341  * Level 1 - loopback address
 342  * Level 2 - link-local addresses
 343  * Level 3 - private addresses.
 344  * Level 4 - global addresses
 345  * For INIT and INIT-ACK address list, let L be the level of
 346  * of requested destination address, sender and receiver
 347  * SHOULD include all of its addresses with level greater
 348  * than or equal to L.
 349  *
 350  * IPv4 scoping can be controlled through sysctl option
 351  * net.sctp.addr_scope_policy
 352  */
 353 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
 354 {
 355         enum sctp_scope retval;
 356
 357         /* Check for unusable SCTP addresses. */
 358         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
 359                 retval =  SCTP_SCOPE_UNUSABLE;
 360         } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
 361                 retval = SCTP_SCOPE_LOOPBACK;
 362         } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
 363                 retval = SCTP_SCOPE_LINK;
 364         } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
 365                    ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
 366                    ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
 367                 retval = SCTP_SCOPE_PRIVATE;
 368         } else {
 369                 retval = SCTP_SCOPE_GLOBAL;
 370         }
 371
 372         return retval;
 373 }
 374
 375 /* Returns a valid dst cache entry for the given source and destination ip
 376  * addresses. If an association is passed, trys to get a dst entry with a
 377  * source address that matches an address in the bind address list.
 378  */
 379 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
 380                                 struct flowi *fl, struct sock *sk)
 381 {
 382         struct sctp_association *asoc = t->asoc;
 383         struct rtable *rt;
 384         struct flowi4 *fl4 = &fl->u.ip4;
 385         struct sctp_bind_addr *bp;
 386         struct sctp_sockaddr_entry *laddr;
 387         struct dst_entry *dst = NULL;
 388         union sctp_addr *daddr = &t->ipaddr;
 389         union sctp_addr dst_saddr;
 390
 391         memset(fl4, 0x0, sizeof(struct flowi4));
 392         fl4->daddr  = daddr->v4.sin_addr.s_addr;
 393         fl4->fl4_dport = daddr->v4.sin_port;
 394         fl4->flowi4_proto = IPPROTO_SCTP;
 395         if (asoc) {
 396                 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
 397                 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
 398                 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
 399         }
 400         if (saddr) {
 401                 fl4->saddr = saddr->v4.sin_addr.s_addr;
 402                 fl4->fl4_sport = saddr->v4.sin_port;
 403         }
 404
 405         pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
 406                  &fl4->saddr);
 407
 408         rt = ip_route_output_key(sock_net(sk), fl4);
 409         if (!IS_ERR(rt))
 410                 dst = &rt->dst;
 411
 412         /* If there is no association or if a source address is passed, no
 413          * more validation is required.
 414          */
 415         if (!asoc || saddr)
 416                 goto out;
 417
 418         bp = &asoc->base.bind_addr;
 419
 420         if (dst) {
 421                 /* Walk through the bind address list and look for a bind
 422                  * address that matches the source address of the returned dst.
 423                  */
 424                 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
 425                 rcu_read_lock();
 426                 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 427                         if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
 428                             (laddr->state != SCTP_ADDR_SRC &&
 429                             !asoc->src_out_of_asoc_ok))
 430                                 continue;
 431                         if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
 432                                 goto out_unlock;
 433                 }
 434                 rcu_read_unlock();
 435
 436                 /* None of the bound addresses match the source address of the
 437                  * dst. So release it.
 438                  */
 439                 dst_release(dst);
 440                 dst = NULL;
 441         }
 442
 443         /* Walk through the bind address list and try to get a dst that
 444          * matches a bind address as the source address.
 445          */
 446         rcu_read_lock();
 447         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 448                 struct net_device *odev;
 449
 450                 if (!laddr->valid)
 451                         continue;
 452                 if (laddr->state != SCTP_ADDR_SRC ||
 453                     AF_INET != laddr->a.sa.sa_family)
 454                         continue;
 455
 456                 fl4->fl4_sport = laddr->a.v4.sin_port;
 457                 flowi4_update_output(fl4,
 458                                      asoc->base.sk->sk_bound_dev_if,
 459                                      RT_CONN_FLAGS(asoc->base.sk),
 460                                      daddr->v4.sin_addr.s_addr,
 461                                      laddr->a.v4.sin_addr.s_addr);
 462
 463                 rt = ip_route_output_key(sock_net(sk), fl4);
 464                 if (IS_ERR(rt))
 465                         continue;
 466
 467                 /* Ensure the src address belongs to the output
 468                  * interface.
 469                  */
 470                 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
 471                                      false);
 472                 if (!odev || odev->ifindex != fl4->flowi4_oif) {
 473                         if (!dst)
 474                                 dst = &rt->dst;
 475                         else
 476                                 dst_release(&rt->dst);
 477                         continue;
 478                 }
 479
 480                 dst_release(dst);
 481                 dst = &rt->dst;
 482                 break;
 483         }
 484
 485 out_unlock:
 486         rcu_read_unlock();
 487 out:
 488         t->dst = dst;
 489         if (dst)
 490                 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
 491                          &fl4->daddr, &fl4->saddr);
 492         else
 493                 pr_debug("no route\n");
 494 }
 495
 496 /* For v4, the source address is cached in the route entry(dst). So no need
 497  * to cache it separately and hence this is an empty routine.
 498  */
 499 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 500                               struct sctp_transport *t,
 501                               struct flowi *fl)
 502 {
 503         union sctp_addr *saddr = &t->saddr;
 504         struct rtable *rt = (struct rtable *)t->dst;
 505
 506         if (rt) {
 507                 saddr->v4.sin_family = AF_INET;
 508                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 509         }
 510 }
 511
 512 /* What interface did this skb arrive on? */
 513 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 514 {
 515         return inet_iif(skb);
 516 }
 517
 518 /* Was this packet marked by Explicit Congestion Notification? */
 519 static int sctp_v4_is_ce(const struct sk_buff *skb)
 520 {
 521         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 522 }
 523
 524 /* Create and initialize a new sk for the socket returned by accept(). */
 525 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 526                                              struct sctp_association *asoc,
 527                                              bool kern)
 528 {
 529         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 530                         sk->sk_prot, kern);
 531         struct inet_sock *newinet;
 532
 533         if (!newsk)
 534                 goto out;
 535
 536         sock_init_data(NULL, newsk);
 537
 538         sctp_copy_sock(newsk, sk, asoc);
 539         sock_reset_flag(newsk, SOCK_ZAPPED);
 540
 541         newinet = inet_sk(newsk);
 542
 543         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 544
 545         sk_refcnt_debug_inc(newsk);
 546
 547         if (newsk->sk_prot->init(newsk)) {
 548                 sk_common_release(newsk);
 549                 newsk = NULL;
 550         }
 551
 552 out:
 553         return newsk;
 554 }
 555
 556 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
 557 {
 558         /* No address mapping for V4 sockets */
 559         return sizeof(struct sockaddr_in);
 560 }
 561
 562 /* Dump the v4 addr to the seq file. */
 563 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 564 {
 565         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 566 }
 567
 568 static void sctp_v4_ecn_capable(struct sock *sk)
 569 {
 570         INET_ECN_xmit(sk);
 571 }
 572
 573 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
 574 {
 575         struct net *net = from_timer(net, t, sctp.addr_wq_timer);
 576         struct sctp_sockaddr_entry *addrw, *temp;
 577         struct sctp_sock *sp;
 578
 579         spin_lock_bh(&net->sctp.addr_wq_lock);
 580
 581         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 582                 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
 583                          "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
 584                          addrw->state, addrw);
 585
 586 #if IS_ENABLED(CONFIG_IPV6)
 587                 /* Now we send an ASCONF for each association */
 588                 /* Note. we currently don't handle link local IPv6 addressees */
 589                 if (addrw->a.sa.sa_family == AF_INET6) {
 590                         struct in6_addr *in6;
 591
 592                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 593                             IPV6_ADDR_LINKLOCAL)
 594                                 goto free_next;
 595
 596                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 597                         if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
 598                             addrw->state == SCTP_ADDR_NEW) {
 599                                 unsigned long timeo_val;
 600
 601                                 pr_debug("%s: this is on DAD, trying %d sec "
 602                                          "later\n", __func__,
 603                                          SCTP_ADDRESS_TICK_DELAY);
 604
 605                                 timeo_val = jiffies;
 606                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 607                                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 608                                 break;
 609                         }
 610                 }
 611 #endif
 612                 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
 613                         struct sock *sk;
 614
 615                         sk = sctp_opt2sk(sp);
 616                         /* ignore bound-specific endpoints */
 617                         if (!sctp_is_ep_boundall(sk))
 618                                 continue;
 619                         bh_lock_sock(sk);
 620                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 621                                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
 622                         bh_unlock_sock(sk);
 623                 }
 624 #if IS_ENABLED(CONFIG_IPV6)
 625 free_next:
 626 #endif
 627                 list_del(&addrw->list);
 628                 kfree(addrw);
 629         }
 630         spin_unlock_bh(&net->sctp.addr_wq_lock);
 631 }
 632
 633 static void sctp_free_addr_wq(struct net *net)
 634 {
 635         struct sctp_sockaddr_entry *addrw;
 636         struct sctp_sockaddr_entry *temp;
 637
 638         spin_lock_bh(&net->sctp.addr_wq_lock);
 639         del_timer(&net->sctp.addr_wq_timer);
 640         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 641                 list_del(&addrw->list);
 642                 kfree(addrw);
 643         }
 644         spin_unlock_bh(&net->sctp.addr_wq_lock);
 645 }
 646
 647 /* lookup the entry for the same address in the addr_waitq
 648  * sctp_addr_wq MUST be locked
 649  */
 650 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
 651                                         struct sctp_sockaddr_entry *addr)
 652 {
 653         struct sctp_sockaddr_entry *addrw;
 654
 655         list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
 656                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 657                         continue;
 658                 if (addrw->a.sa.sa_family == AF_INET) {
 659                         if (addrw->a.v4.sin_addr.s_addr ==
 660                             addr->a.v4.sin_addr.s_addr)
 661                                 return addrw;
 662                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 663                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 664                             &addr->a.v6.sin6_addr))
 665                                 return addrw;
 666                 }
 667         }
 668         return NULL;
 669 }
 670
 671 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
 672 {
 673         struct sctp_sockaddr_entry *addrw;
 674         unsigned long timeo_val;
 675
 676         /* first, we check if an opposite message already exist in the queue.
 677          * If we found such message, it is removed.
 678          * This operation is a bit stupid, but the DHCP client attaches the
 679          * new address after a couple of addition and deletion of that address
 680          */
 681
 682         spin_lock_bh(&net->sctp.addr_wq_lock);
 683         /* Offsets existing events in addr_wq */
 684         addrw = sctp_addr_wq_lookup(net, addr);
 685         if (addrw) {
 686                 if (addrw->state != cmd) {
 687                         pr_debug("%s: offsets existing entry for %d, addr:%pISc "
 688                                  "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
 689                                  &net->sctp.addr_waitq);
 690
 691                         list_del(&addrw->list);
 692                         kfree(addrw);
 693                 }
 694                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 695                 return;
 696         }
 697
 698         /* OK, we have to add the new address to the wait queue */
 699         addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 700         if (addrw == NULL) {
 701                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 702                 return;
 703         }
 704         addrw->state = cmd;
 705         list_add_tail(&addrw->list, &net->sctp.addr_waitq);
 706
 707         pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
 708                  __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
 709
 710         if (!timer_pending(&net->sctp.addr_wq_timer)) {
 711                 timeo_val = jiffies;
 712                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 713                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 714         }
 715         spin_unlock_bh(&net->sctp.addr_wq_lock);
 716 }
 717
 718 /* Event handler for inet address addition/deletion events.
 719  * The sctp_local_addr_list needs to be protocted by a spin lock since
 720  * multiple notifiers (say IPv4 and IPv6) may be running at the same
 721  * time and thus corrupt the list.
 722  * The reader side is protected with RCU.
 723  */
 724 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
 725                                void *ptr)
 726 {
 727         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
 728         struct sctp_sockaddr_entry *addr = NULL;
 729         struct sctp_sockaddr_entry *temp;
 730         struct net *net = dev_net(ifa->ifa_dev->dev);
 731         int found = 0;
 732
 733         switch (ev) {
 734         case NETDEV_UP:
 735                 addr = kmalloc(sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 736                 if (addr) {
 737                         addr->a.v4.sin_family = AF_INET;
 738                         addr->a.v4.sin_port = 0;
 739                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 740                         addr->valid = 1;
 741                         spin_lock_bh(&net->sctp.local_addr_lock);
 742                         list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
 743                         sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
 744                         spin_unlock_bh(&net->sctp.local_addr_lock);
 745                 }
 746                 break;
 747         case NETDEV_DOWN:
 748                 spin_lock_bh(&net->sctp.local_addr_lock);
 749                 list_for_each_entry_safe(addr, temp,
 750                                         &net->sctp.local_addr_list, list) {
 751                         if (addr->a.sa.sa_family == AF_INET &&
 752                                         addr->a.v4.sin_addr.s_addr ==
 753                                         ifa->ifa_local) {
 754                                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
 755                                 found = 1;
 756                                 addr->valid = 0;
 757                                 list_del_rcu(&addr->list);
 758                                 break;
 759                         }
 760                 }
 761                 spin_unlock_bh(&net->sctp.local_addr_lock);
 762                 if (found)
 763                         kfree_rcu(addr, rcu);
 764                 break;
 765         }
 766
 767         return NOTIFY_DONE;
 768 }
 769
 770 /*
 771  * Initialize the control inode/socket with a control endpoint data
 772  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 773  */
 774 static int sctp_ctl_sock_init(struct net *net)
 775 {
 776         int err;
 777         sa_family_t family = PF_INET;
 778
 779         if (sctp_get_pf_specific(PF_INET6))
 780                 family = PF_INET6;
 781
 782         err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
 783                                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
 784
 785         /* If IPv6 socket could not be created, try the IPv4 socket */
 786         if (err < 0 && family == PF_INET6)
 787                 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
 788                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 789                                            net);
 790
 791         if (err < 0) {
 792                 pr_err("Failed to create the SCTP control socket\n");
 793                 return err;
 794         }
 795         return 0;
 796 }
 797
 798 /* Register address family specific functions. */
 799 int sctp_register_af(struct sctp_af *af)
 800 {
 801         switch (af->sa_family) {
 802         case AF_INET:
 803                 if (sctp_af_v4_specific)
 804                         return 0;
 805                 sctp_af_v4_specific = af;
 806                 break;
 807         case AF_INET6:
 808                 if (sctp_af_v6_specific)
 809                         return 0;
 810                 sctp_af_v6_specific = af;
 811                 break;
 812         default:
 813                 return 0;
 814         }
 815
 816         INIT_LIST_HEAD(&af->list);
 817         list_add_tail(&af->list, &sctp_address_families);
 818         return 1;
 819 }
 820
 821 /* Get the table of functions for manipulating a particular address
 822  * family.
 823  */
 824 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 825 {
 826         switch (family) {
 827         case AF_INET:
 828                 return sctp_af_v4_specific;
 829         case AF_INET6:
 830                 return sctp_af_v6_specific;
 831         default:
 832                 return NULL;
 833         }
 834 }
 835
 836 /* Common code to initialize a AF_INET msg_name. */
 837 static void sctp_inet_msgname(char *msgname, int *addr_len)
 838 {
 839         struct sockaddr_in *sin;
 840
 841         sin = (struct sockaddr_in *)msgname;
 842         *addr_len = sizeof(struct sockaddr_in);
 843         sin->sin_family = AF_INET;
 844         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 845 }
 846
 847 /* Copy the primary address of the peer primary address as the msg_name. */
 848 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 849                                     int *addr_len)
 850 {
 851         struct sockaddr_in *sin, *sinfrom;
 852
 853         if (msgname) {
 854                 struct sctp_association *asoc;
 855
 856                 asoc = event->asoc;
 857                 sctp_inet_msgname(msgname, addr_len);
 858                 sin = (struct sockaddr_in *)msgname;
 859                 sinfrom = &asoc->peer.primary_addr.v4;
 860                 sin->sin_port = htons(asoc->peer.port);
 861                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 862         }
 863 }
 864
 865 /* Initialize and copy out a msgname from an inbound skb. */
 866 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 867 {
 868         if (msgname) {
 869                 struct sctphdr *sh = sctp_hdr(skb);
 870                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 871
 872                 sctp_inet_msgname(msgname, len);
 873                 sin->sin_port = sh->source;
 874                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 875         }
 876 }
 877
 878 /* Do we support this AF? */
 879 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 880 {
 881         /* PF_INET only supports AF_INET addresses. */
 882         return AF_INET == family;
 883 }
 884
 885 /* Address matching with wildcards allowed. */
 886 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 887                               const union sctp_addr *addr2,
 888                               struct sctp_sock *opt)
 889 {
 890         /* PF_INET only supports AF_INET addresses. */
 891         if (addr1->sa.sa_family != addr2->sa.sa_family)
 892                 return 0;
 893         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 894             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 895                 return 1;
 896         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 897                 return 1;
 898
 899         return 0;
 900 }
 901
 902 /* Verify that provided sockaddr looks bindable.  Common verification has
 903  * already been taken care of.
 904  */
 905 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 906 {
 907         return sctp_v4_available(addr, opt);
 908 }
 909
 910 /* Verify that sockaddr looks sendable.  Common verification has already
 911  * been taken care of.
 912  */
 913 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 914 {
 915         return 1;
 916 }
 917
 918 /* Fill in Supported Address Type information for INIT and INIT-ACK
 919  * chunks.  Returns number of addresses supported.
 920  */
 921 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 922                                      __be16 *types)
 923 {
 924         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 925         return 1;
 926 }
 927
 928 /* Wrapper routine that calls the ip transmit routine. */
 929 static inline int sctp_v4_xmit(struct sk_buff *skb,
 930                                struct sctp_transport *transport)
 931 {
 932         struct inet_sock *inet = inet_sk(skb->sk);
 933
 934         pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
 935                  skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
 936
 937         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
 938                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 939
 940         SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
 941
 942         return ip_queue_xmit(&inet->sk, skb, &transport->fl);
 943 }
 944
 945 static struct sctp_af sctp_af_inet;
 946
 947 static struct sctp_pf sctp_pf_inet = {
 948         .event_msgname = sctp_inet_event_msgname,
 949         .skb_msgname   = sctp_inet_skb_msgname,
 950         .af_supported  = sctp_inet_af_supported,
 951         .cmp_addr      = sctp_inet_cmp_addr,
 952         .bind_verify   = sctp_inet_bind_verify,
 953         .send_verify   = sctp_inet_send_verify,
 954         .supported_addrs = sctp_inet_supported_addrs,
 955         .create_accept_sk = sctp_v4_create_accept_sk,
 956         .addr_to_user  = sctp_v4_addr_to_user,
 957         .to_sk_saddr   = sctp_v4_to_sk_saddr,
 958         .to_sk_daddr   = sctp_v4_to_sk_daddr,
 959         .af            = &sctp_af_inet
 960 };
 961
 962 /* Notifier for inetaddr addition/deletion events.  */
 963 static struct notifier_block sctp_inetaddr_notifier = {
 964         .notifier_call = sctp_inetaddr_event,
 965 };
 966
 967 /* Socket operations.  */
 968 static const struct proto_ops inet_seqpacket_ops = {
 969         .family            = PF_INET,
 970         .owner             = THIS_MODULE,
 971         .release           = inet_release,      /* Needs to be wrapped... */
 972         .bind              = inet_bind,
 973         .connect           = inet_dgram_connect,
 974         .socketpair        = sock_no_socketpair,
 975         .accept            = inet_accept,
 976         .getname           = inet_getname,      /* Semantics are different.  */
 977         .poll              = sctp_poll,
 978         .ioctl             = inet_ioctl,
 979         .listen            = sctp_inet_listen,
 980         .shutdown          = inet_shutdown,     /* Looks harmless.  */
 981         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
 982         .getsockopt        = sock_common_getsockopt,
 983         .sendmsg           = inet_sendmsg,
 984         .recvmsg           = inet_recvmsg,
 985         .mmap              = sock_no_mmap,
 986         .sendpage          = sock_no_sendpage,
 987 #ifdef CONFIG_COMPAT
 988         .compat_setsockopt = compat_sock_common_setsockopt,
 989         .compat_getsockopt = compat_sock_common_getsockopt,
 990 #endif
 991 };
 992
 993 /* Registration with AF_INET family.  */
 994 static struct inet_protosw sctp_seqpacket_protosw = {
 995         .type       = SOCK_SEQPACKET,
 996         .protocol   = IPPROTO_SCTP,
 997         .prot       = &sctp_prot,
 998         .ops        = &inet_seqpacket_ops,
 999         .flags      = SCTP_PROTOSW_FLAG
1000 };
1001 static struct inet_protosw sctp_stream_protosw = {
1002         .type       = SOCK_STREAM,
1003         .protocol   = IPPROTO_SCTP,
1004         .prot       = &sctp_prot,
1005         .ops        = &inet_seqpacket_ops,
1006         .flags      = SCTP_PROTOSW_FLAG
1007 };
1008
1009 /* Register with IP layer.  */
1010 static const struct net_protocol sctp_protocol = {
1011         .handler     = sctp_rcv,
1012         .err_handler = sctp_v4_err,
1013         .no_policy   = 1,
1014         .netns_ok    = 1,
1015         .icmp_strict_tag_validation = 1,
1016 };
1017
1018 /* IPv4 address related functions.  */
1019 static struct sctp_af sctp_af_inet = {
1020         .sa_family         = AF_INET,
1021         .sctp_xmit         = sctp_v4_xmit,
1022         .setsockopt        = ip_setsockopt,
1023         .getsockopt        = ip_getsockopt,
1024         .get_dst           = sctp_v4_get_dst,
1025         .get_saddr         = sctp_v4_get_saddr,
1026         .copy_addrlist     = sctp_v4_copy_addrlist,
1027         .from_skb          = sctp_v4_from_skb,
1028         .from_sk           = sctp_v4_from_sk,
1029         .from_addr_param   = sctp_v4_from_addr_param,
1030         .to_addr_param     = sctp_v4_to_addr_param,
1031         .cmp_addr          = sctp_v4_cmp_addr,
1032         .addr_valid        = sctp_v4_addr_valid,
1033         .inaddr_any        = sctp_v4_inaddr_any,
1034         .is_any            = sctp_v4_is_any,
1035         .available         = sctp_v4_available,
1036         .scope             = sctp_v4_scope,
1037         .skb_iif           = sctp_v4_skb_iif,
1038         .is_ce             = sctp_v4_is_ce,
1039         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1040         .ecn_capable       = sctp_v4_ecn_capable,
1041         .net_header_len    = sizeof(struct iphdr),
1042         .sockaddr_len      = sizeof(struct sockaddr_in),
1043 #ifdef CONFIG_COMPAT
1044         .compat_setsockopt = compat_ip_setsockopt,
1045         .compat_getsockopt = compat_ip_getsockopt,
1046 #endif
1047 };
1048
1049 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1050 {
1051         switch (family) {
1052         case PF_INET:
1053                 return sctp_pf_inet_specific;
1054         case PF_INET6:
1055                 return sctp_pf_inet6_specific;
1056         default:
1057                 return NULL;
1058         }
1059 }
1060
1061 /* Register the PF specific function table.  */
1062 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1063 {
1064         switch (family) {
1065         case PF_INET:
1066                 if (sctp_pf_inet_specific)
1067                         return 0;
1068                 sctp_pf_inet_specific = pf;
1069                 break;
1070         case PF_INET6:
1071                 if (sctp_pf_inet6_specific)
1072                         return 0;
1073                 sctp_pf_inet6_specific = pf;
1074                 break;
1075         default:
1076                 return 0;
1077         }
1078         return 1;
1079 }
1080
1081 static inline int init_sctp_mibs(struct net *net)
1082 {
1083         net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1084         if (!net->sctp.sctp_statistics)
1085                 return -ENOMEM;
1086         return 0;
1087 }
1088
1089 static inline void cleanup_sctp_mibs(struct net *net)
1090 {
1091         free_percpu(net->sctp.sctp_statistics);
1092 }
1093
1094 static void sctp_v4_pf_init(void)
1095 {
1096         /* Initialize the SCTP specific PF functions. */
1097         sctp_register_pf(&sctp_pf_inet, PF_INET);
1098         sctp_register_af(&sctp_af_inet);
1099 }
1100
1101 static void sctp_v4_pf_exit(void)
1102 {
1103         list_del(&sctp_af_inet.list);
1104 }
1105
1106 static int sctp_v4_protosw_init(void)
1107 {
1108         int rc;
1109
1110         rc = proto_register(&sctp_prot, 1);
1111         if (rc)
1112                 return rc;
1113
1114         /* Register SCTP(UDP and TCP style) with socket layer.  */
1115         inet_register_protosw(&sctp_seqpacket_protosw);
1116         inet_register_protosw(&sctp_stream_protosw);
1117
1118         return 0;
1119 }
1120
1121 static void sctp_v4_protosw_exit(void)
1122 {
1123         inet_unregister_protosw(&sctp_stream_protosw);
1124         inet_unregister_protosw(&sctp_seqpacket_protosw);
1125         proto_unregister(&sctp_prot);
1126 }
1127
1128 static int sctp_v4_add_protocol(void)
1129 {
1130         /* Register notifier for inet address additions/deletions. */
1131         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1132
1133         /* Register SCTP with inet layer.  */
1134         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1135                 return -EAGAIN;
1136
1137         return 0;
1138 }
1139
1140 static void sctp_v4_del_protocol(void)
1141 {
1142         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1143         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1144 }
1145
1146 static int __net_init sctp_defaults_init(struct net *net)
1147 {
1148         int status;
1149
1150         /*
1151          * 14. Suggested SCTP Protocol Parameter Values
1152          */
1153         /* The following protocol parameters are RECOMMENDED:  */
1154         /* RTO.Initial              - 3  seconds */
1155         net->sctp.rto_initial                   = SCTP_RTO_INITIAL;
1156         /* RTO.Min                  - 1  second */
1157         net->sctp.rto_min                       = SCTP_RTO_MIN;
1158         /* RTO.Max                 -  60 seconds */
1159         net->sctp.rto_max                       = SCTP_RTO_MAX;
1160         /* RTO.Alpha                - 1/8 */
1161         net->sctp.rto_alpha                     = SCTP_RTO_ALPHA;
1162         /* RTO.Beta                 - 1/4 */
1163         net->sctp.rto_beta                      = SCTP_RTO_BETA;
1164
1165         /* Valid.Cookie.Life        - 60  seconds */
1166         net->sctp.valid_cookie_life             = SCTP_DEFAULT_COOKIE_LIFE;
1167
1168         /* Whether Cookie Preservative is enabled(1) or not(0) */
1169         net->sctp.cookie_preserve_enable        = 1;
1170
1171         /* Default sctp sockets to use md5 as their hmac alg */
1172 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1173         net->sctp.sctp_hmac_alg                 = "md5";
1174 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1175         net->sctp.sctp_hmac_alg                 = "sha1";
1176 #else
1177         net->sctp.sctp_hmac_alg                 = NULL;
1178 #endif
1179
1180         /* Max.Burst                - 4 */
1181         net->sctp.max_burst                     = SCTP_DEFAULT_MAX_BURST;
1182
1183         /* Enable pf state by default */
1184         net->sctp.pf_enable = 1;
1185
1186         /* Association.Max.Retrans  - 10 attempts
1187          * Path.Max.Retrans         - 5  attempts (per destination address)
1188          * Max.Init.Retransmits     - 8  attempts
1189          */
1190         net->sctp.max_retrans_association       = 10;
1191         net->sctp.max_retrans_path              = 5;
1192         net->sctp.max_retrans_init              = 8;
1193
1194         /* Sendbuffer growth        - do per-socket accounting */
1195         net->sctp.sndbuf_policy                 = 0;
1196
1197         /* Rcvbuffer growth         - do per-socket accounting */
1198         net->sctp.rcvbuf_policy                 = 0;
1199
1200         /* HB.interval              - 30 seconds */
1201         net->sctp.hb_interval                   = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1202
1203         /* delayed SACK timeout */
1204         net->sctp.sack_timeout                  = SCTP_DEFAULT_TIMEOUT_SACK;
1205
1206         /* Disable ADDIP by default. */
1207         net->sctp.addip_enable = 0;
1208         net->sctp.addip_noauth = 0;
1209         net->sctp.default_auto_asconf = 0;
1210
1211         /* Enable PR-SCTP by default. */
1212         net->sctp.prsctp_enable = 1;
1213
1214         /* Disable RECONF by default. */
1215         net->sctp.reconf_enable = 0;
1216
1217         /* Disable AUTH by default. */
1218         net->sctp.auth_enable = 0;
1219
1220         /* Set SCOPE policy to enabled */
1221         net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1222
1223         /* Set the default rwnd update threshold */
1224         net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1225
1226         /* Initialize maximum autoclose timeout. */
1227         net->sctp.max_autoclose         = INT_MAX / HZ;
1228
1229         status = sctp_sysctl_net_register(net);
1230         if (status)
1231                 goto err_sysctl_register;
1232
1233         /* Allocate and initialise sctp mibs.  */
1234         status = init_sctp_mibs(net);
1235         if (status)
1236                 goto err_init_mibs;
1237
1238 #ifdef CONFIG_PROC_FS
1239         /* Initialize proc fs directory.  */
1240         status = sctp_proc_init(net);
1241         if (status)
1242                 goto err_init_proc;
1243 #endif
1244
1245         sctp_dbg_objcnt_init(net);
1246
1247         /* Initialize the local address list. */
1248         INIT_LIST_HEAD(&net->sctp.local_addr_list);
1249         spin_lock_init(&net->sctp.local_addr_lock);
1250         sctp_get_local_addr_list(net);
1251
1252         /* Initialize the address event list */
1253         INIT_LIST_HEAD(&net->sctp.addr_waitq);
1254         INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1255         spin_lock_init(&net->sctp.addr_wq_lock);
1256         net->sctp.addr_wq_timer.expires = 0;
1257         timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1258
1259         return 0;
1260
1261 #ifdef CONFIG_PROC_FS
1262 err_init_proc:
1263         cleanup_sctp_mibs(net);
1264 #endif
1265 err_init_mibs:
1266         sctp_sysctl_net_unregister(net);
1267 err_sysctl_register:
1268         return status;
1269 }
1270
1271 static void __net_exit sctp_defaults_exit(struct net *net)
1272 {
1273         /* Free the local address list */
1274         sctp_free_addr_wq(net);
1275         sctp_free_local_addr_list(net);
1276
1277 #ifdef CONFIG_PROC_FS
1278         remove_proc_subtree("sctp", net->proc_net);
1279         net->sctp.proc_net_sctp = NULL;
1280 #endif
1281         cleanup_sctp_mibs(net);
1282         sctp_sysctl_net_unregister(net);
1283 }
1284
1285 static struct pernet_operations sctp_defaults_ops = {
1286         .init = sctp_defaults_init,
1287         .exit = sctp_defaults_exit,
1288 };
1289
1290 static int __net_init sctp_ctrlsock_init(struct net *net)
1291 {
1292         int status;
1293
1294         /* Initialize the control inode/socket for handling OOTB packets.  */
1295         status = sctp_ctl_sock_init(net);
1296         if (status)
1297                 pr_err("Failed to initialize the SCTP control sock\n");
1298
1299         return status;
1300 }
1301
1302 static void __net_init sctp_ctrlsock_exit(struct net *net)
1303 {
1304         /* Free the control endpoint.  */
1305         inet_ctl_sock_destroy(net->sctp.ctl_sock);
1306 }
1307
1308 static struct pernet_operations sctp_ctrlsock_ops = {
1309         .init = sctp_ctrlsock_init,
1310         .exit = sctp_ctrlsock_exit,
1311 };
1312
1313 /* Initialize the universe into something sensible.  */
1314 static __init int sctp_init(void)
1315 {
1316         int i;
1317         int status = -EINVAL;
1318         unsigned long goal;
1319         unsigned long limit;
1320         int max_share;
1321         int order;
1322         int num_entries;
1323         int max_entry_order;
1324
1325         sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1326
1327         /* Allocate bind_bucket and chunk caches. */
1328         status = -ENOBUFS;
1329         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1330                                                sizeof(struct sctp_bind_bucket),
1331                                                0, SLAB_HWCACHE_ALIGN,
1332                                                NULL);
1333         if (!sctp_bucket_cachep)
1334                 goto out;
1335
1336         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1337                                                sizeof(struct sctp_chunk),
1338                                                0, SLAB_HWCACHE_ALIGN,
1339                                                NULL);
1340         if (!sctp_chunk_cachep)
1341                 goto err_chunk_cachep;
1342
1343         status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1344         if (status)
1345                 goto err_percpu_counter_init;
1346
1347         /* Implementation specific variables. */
1348
1349         /* Initialize default stream count setup information. */
1350         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1351         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1352
1353         /* Initialize handle used for association ids. */
1354         idr_init(&sctp_assocs_id);
1355
1356         limit = nr_free_buffer_pages() / 8;
1357         limit = max(limit, 128UL);
1358         sysctl_sctp_mem[0] = limit / 4 * 3;
1359         sysctl_sctp_mem[1] = limit;
1360         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1361
1362         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1363         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1364         max_share = min(4UL*1024*1024, limit);
1365
1366         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1367         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1368         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1369
1370         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1371         sysctl_sctp_wmem[1] = 16*1024;
1372         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1373
1374         /* Size and allocate the association hash table.
1375          * The methodology is similar to that of the tcp hash tables.
1376          * Though not identical.  Start by getting a goal size
1377          */
1378         if (totalram_pages >= (128 * 1024))
1379                 goal = totalram_pages >> (22 - PAGE_SHIFT);
1380         else
1381                 goal = totalram_pages >> (24 - PAGE_SHIFT);
1382
1383         /* Then compute the page order for said goal */
1384         order = get_order(goal);
1385
1386         /* Now compute the required page order for the maximum sized table we
1387          * want to create
1388          */
1389         max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1390                                     sizeof(struct sctp_bind_hashbucket));
1391
1392         /* Limit the page order by that maximum hash table size */
1393         order = min(order, max_entry_order);
1394
1395         /* Allocate and initialize the endpoint hash table.  */
1396         sctp_ep_hashsize = 64;
1397         sctp_ep_hashtable =
1398                 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1399         if (!sctp_ep_hashtable) {
1400                 pr_err("Failed endpoint_hash alloc\n");
1401                 status = -ENOMEM;
1402                 goto err_ehash_alloc;
1403         }
1404         for (i = 0; i < sctp_ep_hashsize; i++) {
1405                 rwlock_init(&sctp_ep_hashtable[i].lock);
1406                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1407         }
1408
1409         /* Allocate and initialize the SCTP port hash table.
1410          * Note that order is initalized to start at the max sized
1411          * table we want to support.  If we can't get that many pages
1412          * reduce the order and try again
1413          */
1414         do {
1415                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1416                         __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1417         } while (!sctp_port_hashtable && --order > 0);
1418
1419         if (!sctp_port_hashtable) {
1420                 pr_err("Failed bind hash alloc\n");
1421                 status = -ENOMEM;
1422                 goto err_bhash_alloc;
1423         }
1424
1425         /* Now compute the number of entries that will fit in the
1426          * port hash space we allocated
1427          */
1428         num_entries = (1UL << order) * PAGE_SIZE /
1429                       sizeof(struct sctp_bind_hashbucket);
1430
1431         /* And finish by rounding it down to the nearest power of two
1432          * this wastes some memory of course, but its needed because
1433          * the hash function operates based on the assumption that
1434          * that the number of entries is a power of two
1435          */
1436         sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1437
1438         for (i = 0; i < sctp_port_hashsize; i++) {
1439                 spin_lock_init(&sctp_port_hashtable[i].lock);
1440                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1441         }
1442
1443         status = sctp_transport_hashtable_init();
1444         if (status)
1445                 goto err_thash_alloc;
1446
1447         pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1448                 num_entries);
1449
1450         sctp_sysctl_register();
1451
1452         INIT_LIST_HEAD(&sctp_address_families);
1453         sctp_v4_pf_init();
1454         sctp_v6_pf_init();
1455         sctp_sched_ops_init();
1456
1457         status = register_pernet_subsys(&sctp_defaults_ops);
1458         if (status)
1459                 goto err_register_defaults;
1460
1461         status = sctp_v4_protosw_init();
1462         if (status)
1463                 goto err_protosw_init;
1464
1465         status = sctp_v6_protosw_init();
1466         if (status)
1467                 goto err_v6_protosw_init;
1468
1469         status = register_pernet_subsys(&sctp_ctrlsock_ops);
1470         if (status)
1471                 goto err_register_ctrlsock;
1472
1473         status = sctp_v4_add_protocol();
1474         if (status)
1475                 goto err_add_protocol;
1476
1477         /* Register SCTP with inet6 layer.  */
1478         status = sctp_v6_add_protocol();
1479         if (status)
1480                 goto err_v6_add_protocol;
1481
1482         if (sctp_offload_init() < 0)
1483                 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1484
1485 out:
1486         return status;
1487 err_v6_add_protocol:
1488         sctp_v4_del_protocol();
1489 err_add_protocol:
1490         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1491 err_register_ctrlsock:
1492         sctp_v6_protosw_exit();
1493 err_v6_protosw_init:
1494         sctp_v4_protosw_exit();
1495 err_protosw_init:
1496         unregister_pernet_subsys(&sctp_defaults_ops);
1497 err_register_defaults:
1498         sctp_v4_pf_exit();
1499         sctp_v6_pf_exit();
1500         sctp_sysctl_unregister();
1501         free_pages((unsigned long)sctp_port_hashtable,
1502                    get_order(sctp_port_hashsize *
1503                              sizeof(struct sctp_bind_hashbucket)));
1504 err_bhash_alloc:
1505         sctp_transport_hashtable_destroy();
1506 err_thash_alloc:
1507         kfree(sctp_ep_hashtable);
1508 err_ehash_alloc:
1509         percpu_counter_destroy(&sctp_sockets_allocated);
1510 err_percpu_counter_init:
1511         kmem_cache_destroy(sctp_chunk_cachep);
1512 err_chunk_cachep:
1513         kmem_cache_destroy(sctp_bucket_cachep);
1514         goto out;
1515 }
1516
1517 /* Exit handler for the SCTP protocol.  */
1518 static __exit void sctp_exit(void)
1519 {
1520         /* BUG.  This should probably do something useful like clean
1521          * up all the remaining associations and all that memory.
1522          */
1523
1524         /* Unregister with inet6/inet layers. */
1525         sctp_v6_del_protocol();
1526         sctp_v4_del_protocol();
1527
1528         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1529
1530         /* Free protosw registrations */
1531         sctp_v6_protosw_exit();
1532         sctp_v4_protosw_exit();
1533
1534         unregister_pernet_subsys(&sctp_defaults_ops);
1535
1536         /* Unregister with socket layer. */
1537         sctp_v6_pf_exit();
1538         sctp_v4_pf_exit();
1539
1540         sctp_sysctl_unregister();
1541
1542         free_pages((unsigned long)sctp_port_hashtable,
1543                    get_order(sctp_port_hashsize *
1544                              sizeof(struct sctp_bind_hashbucket)));
1545         kfree(sctp_ep_hashtable);
1546         sctp_transport_hashtable_destroy();
1547
1548         percpu_counter_destroy(&sctp_sockets_allocated);
1549
1550         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1551
1552         kmem_cache_destroy(sctp_chunk_cachep);
1553         kmem_cache_destroy(sctp_bucket_cachep);
1554 }
1555
1556 module_init(sctp_init);
1557 module_exit(sctp_exit);
1558
1559 /*
1560  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1561  */
1562 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1563 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1564 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1565 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1566 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1567 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1568 MODULE_LICENSE("GPL");