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 /* Set up the proc fs entry for the SCTP protocol. */
  84 static int __net_init sctp_proc_init(struct net *net)
  85 {
  86 #ifdef CONFIG_PROC_FS
  87         net->sctp.proc_net_sctp = proc_net_mkdir(net, "sctp", net->proc_net);
  88         if (!net->sctp.proc_net_sctp)
  89                 goto out_proc_net_sctp;
  90         if (sctp_snmp_proc_init(net))
  91                 goto out_snmp_proc_init;
  92         if (sctp_eps_proc_init(net))
  93                 goto out_eps_proc_init;
  94         if (sctp_assocs_proc_init(net))
  95                 goto out_assocs_proc_init;
  96         if (sctp_remaddr_proc_init(net))
  97                 goto out_remaddr_proc_init;
  98
  99         return 0;
 100
 101 out_remaddr_proc_init:
 102         sctp_assocs_proc_exit(net);
 103 out_assocs_proc_init:
 104         sctp_eps_proc_exit(net);
 105 out_eps_proc_init:
 106         sctp_snmp_proc_exit(net);
 107 out_snmp_proc_init:
 108         remove_proc_entry("sctp", net->proc_net);
 109         net->sctp.proc_net_sctp = NULL;
 110 out_proc_net_sctp:
 111         return -ENOMEM;
 112 #endif /* CONFIG_PROC_FS */
 113         return 0;
 114 }
 115
 116 /* Clean up the proc fs entry for the SCTP protocol.
 117  * Note: Do not make this __exit as it is used in the init error
 118  * path.
 119  */
 120 static void sctp_proc_exit(struct net *net)
 121 {
 122 #ifdef CONFIG_PROC_FS
 123         sctp_snmp_proc_exit(net);
 124         sctp_eps_proc_exit(net);
 125         sctp_assocs_proc_exit(net);
 126         sctp_remaddr_proc_exit(net);
 127
 128         remove_proc_entry("sctp", net->proc_net);
 129         net->sctp.proc_net_sctp = NULL;
 130 #endif
 131 }
 132
 133 /* Private helper to extract ipv4 address and stash them in
 134  * the protocol structure.
 135  */
 136 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
 137                                   struct net_device *dev)
 138 {
 139         struct in_device *in_dev;
 140         struct in_ifaddr *ifa;
 141         struct sctp_sockaddr_entry *addr;
 142
 143         rcu_read_lock();
 144         if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
 145                 rcu_read_unlock();
 146                 return;
 147         }
 148
 149         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
 150                 /* Add the address to the local list.  */
 151                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
 152                 if (addr) {
 153                         addr->a.v4.sin_family = AF_INET;
 154                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 155                         addr->valid = 1;
 156                         INIT_LIST_HEAD(&addr->list);
 157                         list_add_tail(&addr->list, addrlist);
 158                 }
 159         }
 160
 161         rcu_read_unlock();
 162 }
 163
 164 /* Extract our IP addresses from the system and stash them in the
 165  * protocol structure.
 166  */
 167 static void sctp_get_local_addr_list(struct net *net)
 168 {
 169         struct net_device *dev;
 170         struct list_head *pos;
 171         struct sctp_af *af;
 172
 173         rcu_read_lock();
 174         for_each_netdev_rcu(net, dev) {
 175                 list_for_each(pos, &sctp_address_families) {
 176                         af = list_entry(pos, struct sctp_af, list);
 177                         af->copy_addrlist(&net->sctp.local_addr_list, dev);
 178                 }
 179         }
 180         rcu_read_unlock();
 181 }
 182
 183 /* Free the existing local addresses.  */
 184 static void sctp_free_local_addr_list(struct net *net)
 185 {
 186         struct sctp_sockaddr_entry *addr;
 187         struct list_head *pos, *temp;
 188
 189         list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
 190                 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
 191                 list_del(pos);
 192                 kfree(addr);
 193         }
 194 }
 195
 196 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
 197 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
 198                               sctp_scope_t scope, gfp_t gfp, int copy_flags)
 199 {
 200         struct sctp_sockaddr_entry *addr;
 201         int error = 0;
 202
 203         rcu_read_lock();
 204         list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
 205                 if (!addr->valid)
 206                         continue;
 207                 if (sctp_in_scope(net, &addr->a, scope)) {
 208                         /* Now that the address is in scope, check to see if
 209                          * the address type is really supported by the local
 210                          * sock as well as the remote peer.
 211                          */
 212                         if ((((AF_INET == addr->a.sa.sa_family) &&
 213                               (copy_flags & SCTP_ADDR4_ALLOWED) &&
 214                               (copy_flags & SCTP_ADDR4_PEERSUPP))) ||
 215                             (((AF_INET6 == addr->a.sa.sa_family) &&
 216                               (copy_flags & SCTP_ADDR6_ALLOWED) &&
 217                               (copy_flags & SCTP_ADDR6_PEERSUPP)))) {
 218                                 error = sctp_add_bind_addr(bp, &addr->a,
 219                                                     SCTP_ADDR_SRC, GFP_ATOMIC);
 220                                 if (error)
 221                                         goto end_copy;
 222                         }
 223                 }
 224         }
 225
 226 end_copy:
 227         rcu_read_unlock();
 228         return error;
 229 }
 230
 231 /* Initialize a sctp_addr from in incoming skb.  */
 232 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
 233                              int is_saddr)
 234 {
 235         void *from;
 236         __be16 *port;
 237         struct sctphdr *sh;
 238
 239         port = &addr->v4.sin_port;
 240         addr->v4.sin_family = AF_INET;
 241
 242         sh = sctp_hdr(skb);
 243         if (is_saddr) {
 244                 *port  = sh->source;
 245                 from = &ip_hdr(skb)->saddr;
 246         } else {
 247                 *port = sh->dest;
 248                 from = &ip_hdr(skb)->daddr;
 249         }
 250         memcpy(&addr->v4.sin_addr.s_addr, from, sizeof(struct in_addr));
 251 }
 252
 253 /* Initialize an sctp_addr from a socket. */
 254 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
 255 {
 256         addr->v4.sin_family = AF_INET;
 257         addr->v4.sin_port = 0;
 258         addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
 259         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 260 }
 261
 262 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
 263 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
 264 {
 265         inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
 266 }
 267
 268 /* Initialize sk->sk_daddr from sctp_addr. */
 269 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
 270 {
 271         inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
 272 }
 273
 274 /* Initialize a sctp_addr from an address parameter. */
 275 static bool sctp_v4_from_addr_param(union sctp_addr *addr,
 276                                     union sctp_addr_param *param,
 277                                     __be16 port, int iif)
 278 {
 279         if (ntohs(param->v4.param_hdr.length) < sizeof(struct sctp_ipv4addr_param))
 280                 return false;
 281
 282         addr->v4.sin_family = AF_INET;
 283         addr->v4.sin_port = port;
 284         addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
 285         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 286
 287         return true;
 288 }
 289
 290 /* Initialize an address parameter from a sctp_addr and return the length
 291  * of the address parameter.
 292  */
 293 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
 294                                  union sctp_addr_param *param)
 295 {
 296         int length = sizeof(sctp_ipv4addr_param_t);
 297
 298         param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
 299         param->v4.param_hdr.length = htons(length);
 300         param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
 301
 302         return length;
 303 }
 304
 305 /* Initialize a sctp_addr from a dst_entry. */
 306 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
 307                               __be16 port)
 308 {
 309         saddr->v4.sin_family = AF_INET;
 310         saddr->v4.sin_port = port;
 311         saddr->v4.sin_addr.s_addr = fl4->saddr;
 312         memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero));
 313 }
 314
 315 /* Compare two addresses exactly. */
 316 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
 317                             const union sctp_addr *addr2)
 318 {
 319         if (addr1->sa.sa_family != addr2->sa.sa_family)
 320                 return 0;
 321         if (addr1->v4.sin_port != addr2->v4.sin_port)
 322                 return 0;
 323         if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
 324                 return 0;
 325
 326         return 1;
 327 }
 328
 329 /* Initialize addr struct to INADDR_ANY. */
 330 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
 331 {
 332         addr->v4.sin_family = AF_INET;
 333         addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
 334         addr->v4.sin_port = port;
 335         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 336 }
 337
 338 /* Is this a wildcard address? */
 339 static int sctp_v4_is_any(const union sctp_addr *addr)
 340 {
 341         return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
 342 }
 343
 344 /* This function checks if the address is a valid address to be used for
 345  * SCTP binding.
 346  *
 347  * Output:
 348  * Return 0 - If the address is a non-unicast or an illegal address.
 349  * Return 1 - If the address is a unicast.
 350  */
 351 static int sctp_v4_addr_valid(union sctp_addr *addr,
 352                               struct sctp_sock *sp,
 353                               const struct sk_buff *skb)
 354 {
 355         /* IPv4 addresses not allowed */
 356         if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
 357                 return 0;
 358
 359         /* Is this a non-unicast address or a unusable SCTP address? */
 360         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
 361                 return 0;
 362
 363         /* Is this a broadcast address? */
 364         if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
 365                 return 0;
 366
 367         return 1;
 368 }
 369
 370 /* Should this be available for binding?   */
 371 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
 372 {
 373         struct net *net = sock_net(&sp->inet.sk);
 374         int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
 375
 376
 377         if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
 378            ret != RTN_LOCAL &&
 379            !sp->inet.freebind &&
 380            !net->ipv4.sysctl_ip_nonlocal_bind)
 381                 return 0;
 382
 383         if (ipv6_only_sock(sctp_opt2sk(sp)))
 384                 return 0;
 385
 386         return 1;
 387 }
 388
 389 /* Checking the loopback, private and other address scopes as defined in
 390  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
 391  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
 392  *
 393  * Level 0 - unusable SCTP addresses
 394  * Level 1 - loopback address
 395  * Level 2 - link-local addresses
 396  * Level 3 - private addresses.
 397  * Level 4 - global addresses
 398  * For INIT and INIT-ACK address list, let L be the level of
 399  * of requested destination address, sender and receiver
 400  * SHOULD include all of its addresses with level greater
 401  * than or equal to L.
 402  *
 403  * IPv4 scoping can be controlled through sysctl option
 404  * net.sctp.addr_scope_policy
 405  */
 406 static sctp_scope_t sctp_v4_scope(union sctp_addr *addr)
 407 {
 408         sctp_scope_t retval;
 409
 410         /* Check for unusable SCTP addresses. */
 411         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
 412                 retval =  SCTP_SCOPE_UNUSABLE;
 413         } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
 414                 retval = SCTP_SCOPE_LOOPBACK;
 415         } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
 416                 retval = SCTP_SCOPE_LINK;
 417         } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
 418                    ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
 419                    ipv4_is_private_192(addr->v4.sin_addr.s_addr) ||
 420                    ipv4_is_test_198(addr->v4.sin_addr.s_addr)) {
 421                 retval = SCTP_SCOPE_PRIVATE;
 422         } else {
 423                 retval = SCTP_SCOPE_GLOBAL;
 424         }
 425
 426         return retval;
 427 }
 428
 429 /* Returns a valid dst cache entry for the given source and destination ip
 430  * addresses. If an association is passed, trys to get a dst entry with a
 431  * source address that matches an address in the bind address list.
 432  */
 433 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
 434                                 struct flowi *fl, struct sock *sk)
 435 {
 436         struct sctp_association *asoc = t->asoc;
 437         struct rtable *rt;
 438         struct flowi _fl;
 439         struct flowi4 *fl4 = &_fl.u.ip4;
 440         struct sctp_bind_addr *bp;
 441         struct sctp_sockaddr_entry *laddr;
 442         struct dst_entry *dst = NULL;
 443         union sctp_addr *daddr = &t->ipaddr;
 444         union sctp_addr dst_saddr;
 445
 446         memset(&_fl, 0x0, sizeof(_fl));
 447         fl4->daddr  = daddr->v4.sin_addr.s_addr;
 448         fl4->fl4_dport = daddr->v4.sin_port;
 449         fl4->flowi4_proto = IPPROTO_SCTP;
 450         if (asoc) {
 451                 fl4->flowi4_tos = RT_CONN_FLAGS(asoc->base.sk);
 452                 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
 453                 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
 454         }
 455         if (saddr) {
 456                 fl4->saddr = saddr->v4.sin_addr.s_addr;
 457                 fl4->fl4_sport = saddr->v4.sin_port;
 458         }
 459
 460         pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
 461                  &fl4->saddr);
 462
 463         rt = ip_route_output_key(sock_net(sk), fl4);
 464         if (!IS_ERR(rt)) {
 465                 dst = &rt->dst;
 466                 t->dst = dst;
 467                 memcpy(fl, &_fl, sizeof(_fl));
 468         }
 469
 470         /* If there is no association or if a source address is passed, no
 471          * more validation is required.
 472          */
 473         if (!asoc || saddr)
 474                 goto out;
 475
 476         bp = &asoc->base.bind_addr;
 477
 478         if (dst) {
 479                 /* Walk through the bind address list and look for a bind
 480                  * address that matches the source address of the returned dst.
 481                  */
 482                 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
 483                 rcu_read_lock();
 484                 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 485                         if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
 486                             (laddr->state != SCTP_ADDR_SRC &&
 487                             !asoc->src_out_of_asoc_ok))
 488                                 continue;
 489                         if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
 490                                 goto out_unlock;
 491                 }
 492                 rcu_read_unlock();
 493
 494                 /* None of the bound addresses match the source address of the
 495                  * dst. So release it.
 496                  */
 497                 dst_release(dst);
 498                 dst = NULL;
 499         }
 500
 501         /* Walk through the bind address list and try to get a dst that
 502          * matches a bind address as the source address.
 503          */
 504         rcu_read_lock();
 505         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 506                 struct net_device *odev;
 507
 508                 if (!laddr->valid)
 509                         continue;
 510                 if (laddr->state != SCTP_ADDR_SRC ||
 511                     AF_INET != laddr->a.sa.sa_family)
 512                         continue;
 513
 514                 fl4->fl4_sport = laddr->a.v4.sin_port;
 515                 flowi4_update_output(fl4,
 516                                      asoc->base.sk->sk_bound_dev_if,
 517                                      RT_CONN_FLAGS(asoc->base.sk),
 518                                      daddr->v4.sin_addr.s_addr,
 519                                      laddr->a.v4.sin_addr.s_addr);
 520
 521                 rt = ip_route_output_key(sock_net(sk), fl4);
 522                 if (IS_ERR(rt))
 523                         continue;
 524
 525                 /* Ensure the src address belongs to the output
 526                  * interface.
 527                  */
 528                 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
 529                                      false);
 530                 if (!odev || odev->ifindex != fl4->flowi4_oif) {
 531                         if (!dst) {
 532                                 dst = &rt->dst;
 533                                 t->dst = dst;
 534                                 memcpy(fl, &_fl, sizeof(_fl));
 535                         } else {
 536                                 dst_release(&rt->dst);
 537                         }
 538                         continue;
 539                 }
 540
 541                 dst_release(dst);
 542                 dst = &rt->dst;
 543                 t->dst = dst;
 544                 memcpy(fl, &_fl, sizeof(_fl));
 545                 break;
 546         }
 547
 548 out_unlock:
 549         rcu_read_unlock();
 550 out:
 551         if (dst) {
 552                 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
 553                          &fl->u.ip4.daddr, &fl->u.ip4.saddr);
 554         } else {
 555                 t->dst = NULL;
 556                 pr_debug("no route\n");
 557         }
 558 }
 559
 560 /* For v4, the source address is cached in the route entry(dst). So no need
 561  * to cache it separately and hence this is an empty routine.
 562  */
 563 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 564                               struct sctp_transport *t,
 565                               struct flowi *fl)
 566 {
 567         union sctp_addr *saddr = &t->saddr;
 568         struct rtable *rt = (struct rtable *)t->dst;
 569
 570         if (rt) {
 571                 saddr->v4.sin_family = AF_INET;
 572                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 573         }
 574 }
 575
 576 /* What interface did this skb arrive on? */
 577 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 578 {
 579         return inet_iif(skb);
 580 }
 581
 582 /* Was this packet marked by Explicit Congestion Notification? */
 583 static int sctp_v4_is_ce(const struct sk_buff *skb)
 584 {
 585         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 586 }
 587
 588 /* Create and initialize a new sk for the socket returned by accept(). */
 589 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 590                                              struct sctp_association *asoc)
 591 {
 592         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 593                         sk->sk_prot, 0);
 594         struct inet_sock *newinet;
 595
 596         if (!newsk)
 597                 goto out;
 598
 599         sock_init_data(NULL, newsk);
 600
 601         sctp_copy_sock(newsk, sk, asoc);
 602         sock_reset_flag(newsk, SOCK_ZAPPED);
 603
 604         newinet = inet_sk(newsk);
 605
 606         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 607
 608         sk_refcnt_debug_inc(newsk);
 609
 610         if (newsk->sk_prot->init(newsk)) {
 611                 sk_common_release(newsk);
 612                 newsk = NULL;
 613         }
 614
 615 out:
 616         return newsk;
 617 }
 618
 619 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
 620 {
 621         /* No address mapping for V4 sockets */
 622         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 623         return sizeof(struct sockaddr_in);
 624 }
 625
 626 /* Dump the v4 addr to the seq file. */
 627 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 628 {
 629         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 630 }
 631
 632 static void sctp_v4_ecn_capable(struct sock *sk)
 633 {
 634         INET_ECN_xmit(sk);
 635 }
 636
 637 static void sctp_addr_wq_timeout_handler(unsigned long arg)
 638 {
 639         struct net *net = (struct net *)arg;
 640         struct sctp_sockaddr_entry *addrw, *temp;
 641         struct sctp_sock *sp;
 642
 643         spin_lock_bh(&net->sctp.addr_wq_lock);
 644
 645         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 646                 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
 647                          "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
 648                          addrw->state, addrw);
 649
 650 #if IS_ENABLED(CONFIG_IPV6)
 651                 /* Now we send an ASCONF for each association */
 652                 /* Note. we currently don't handle link local IPv6 addressees */
 653                 if (addrw->a.sa.sa_family == AF_INET6) {
 654                         struct in6_addr *in6;
 655
 656                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 657                             IPV6_ADDR_LINKLOCAL)
 658                                 goto free_next;
 659
 660                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 661                         if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
 662                             addrw->state == SCTP_ADDR_NEW) {
 663                                 unsigned long timeo_val;
 664
 665                                 pr_debug("%s: this is on DAD, trying %d sec "
 666                                          "later\n", __func__,
 667                                          SCTP_ADDRESS_TICK_DELAY);
 668
 669                                 timeo_val = jiffies;
 670                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 671                                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 672                                 break;
 673                         }
 674                 }
 675 #endif
 676                 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
 677                         struct sock *sk;
 678
 679                         sk = sctp_opt2sk(sp);
 680                         /* ignore bound-specific endpoints */
 681                         if (!sctp_is_ep_boundall(sk))
 682                                 continue;
 683                         bh_lock_sock(sk);
 684                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 685                                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
 686                         bh_unlock_sock(sk);
 687                 }
 688 #if IS_ENABLED(CONFIG_IPV6)
 689 free_next:
 690 #endif
 691                 list_del(&addrw->list);
 692                 kfree(addrw);
 693         }
 694         spin_unlock_bh(&net->sctp.addr_wq_lock);
 695 }
 696
 697 static void sctp_free_addr_wq(struct net *net)
 698 {
 699         struct sctp_sockaddr_entry *addrw;
 700         struct sctp_sockaddr_entry *temp;
 701
 702         spin_lock_bh(&net->sctp.addr_wq_lock);
 703         del_timer(&net->sctp.addr_wq_timer);
 704         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 705                 list_del(&addrw->list);
 706                 kfree(addrw);
 707         }
 708         spin_unlock_bh(&net->sctp.addr_wq_lock);
 709 }
 710
 711 /* lookup the entry for the same address in the addr_waitq
 712  * sctp_addr_wq MUST be locked
 713  */
 714 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
 715                                         struct sctp_sockaddr_entry *addr)
 716 {
 717         struct sctp_sockaddr_entry *addrw;
 718
 719         list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
 720                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 721                         continue;
 722                 if (addrw->a.sa.sa_family == AF_INET) {
 723                         if (addrw->a.v4.sin_addr.s_addr ==
 724                             addr->a.v4.sin_addr.s_addr)
 725                                 return addrw;
 726                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 727                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 728                             &addr->a.v6.sin6_addr))
 729                                 return addrw;
 730                 }
 731         }
 732         return NULL;
 733 }
 734
 735 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
 736 {
 737         struct sctp_sockaddr_entry *addrw;
 738         unsigned long timeo_val;
 739
 740         /* first, we check if an opposite message already exist in the queue.
 741          * If we found such message, it is removed.
 742          * This operation is a bit stupid, but the DHCP client attaches the
 743          * new address after a couple of addition and deletion of that address
 744          */
 745
 746         spin_lock_bh(&net->sctp.addr_wq_lock);
 747         /* Offsets existing events in addr_wq */
 748         addrw = sctp_addr_wq_lookup(net, addr);
 749         if (addrw) {
 750                 if (addrw->state != cmd) {
 751                         pr_debug("%s: offsets existing entry for %d, addr:%pISc "
 752                                  "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
 753                                  &net->sctp.addr_waitq);
 754
 755                         list_del(&addrw->list);
 756                         kfree(addrw);
 757                 }
 758                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 759                 return;
 760         }
 761
 762         /* OK, we have to add the new address to the wait queue */
 763         addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 764         if (addrw == NULL) {
 765                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 766                 return;
 767         }
 768         addrw->state = cmd;
 769         list_add_tail(&addrw->list, &net->sctp.addr_waitq);
 770
 771         pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
 772                  __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
 773
 774         if (!timer_pending(&net->sctp.addr_wq_timer)) {
 775                 timeo_val = jiffies;
 776                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 777                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 778         }
 779         spin_unlock_bh(&net->sctp.addr_wq_lock);
 780 }
 781
 782 /* Event handler for inet address addition/deletion events.
 783  * The sctp_local_addr_list needs to be protocted by a spin lock since
 784  * multiple notifiers (say IPv4 and IPv6) may be running at the same
 785  * time and thus corrupt the list.
 786  * The reader side is protected with RCU.
 787  */
 788 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
 789                                void *ptr)
 790 {
 791         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
 792         struct sctp_sockaddr_entry *addr = NULL;
 793         struct sctp_sockaddr_entry *temp;
 794         struct net *net = dev_net(ifa->ifa_dev->dev);
 795         int found = 0;
 796
 797         switch (ev) {
 798         case NETDEV_UP:
 799                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
 800                 if (addr) {
 801                         addr->a.v4.sin_family = AF_INET;
 802                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 803                         addr->valid = 1;
 804                         spin_lock_bh(&net->sctp.local_addr_lock);
 805                         list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
 806                         sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
 807                         spin_unlock_bh(&net->sctp.local_addr_lock);
 808                 }
 809                 break;
 810         case NETDEV_DOWN:
 811                 spin_lock_bh(&net->sctp.local_addr_lock);
 812                 list_for_each_entry_safe(addr, temp,
 813                                         &net->sctp.local_addr_list, list) {
 814                         if (addr->a.sa.sa_family == AF_INET &&
 815                                         addr->a.v4.sin_addr.s_addr ==
 816                                         ifa->ifa_local) {
 817                                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
 818                                 found = 1;
 819                                 addr->valid = 0;
 820                                 list_del_rcu(&addr->list);
 821                                 break;
 822                         }
 823                 }
 824                 spin_unlock_bh(&net->sctp.local_addr_lock);
 825                 if (found)
 826                         kfree_rcu(addr, rcu);
 827                 break;
 828         }
 829
 830         return NOTIFY_DONE;
 831 }
 832
 833 /*
 834  * Initialize the control inode/socket with a control endpoint data
 835  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 836  */
 837 static int sctp_ctl_sock_init(struct net *net)
 838 {
 839         int err;
 840         sa_family_t family = PF_INET;
 841
 842         if (sctp_get_pf_specific(PF_INET6))
 843                 family = PF_INET6;
 844
 845         err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
 846                                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
 847
 848         /* If IPv6 socket could not be created, try the IPv4 socket */
 849         if (err < 0 && family == PF_INET6)
 850                 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
 851                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 852                                            net);
 853
 854         if (err < 0) {
 855                 pr_err("Failed to create the SCTP control socket\n");
 856                 return err;
 857         }
 858         return 0;
 859 }
 860
 861 /* Register address family specific functions. */
 862 int sctp_register_af(struct sctp_af *af)
 863 {
 864         switch (af->sa_family) {
 865         case AF_INET:
 866                 if (sctp_af_v4_specific)
 867                         return 0;
 868                 sctp_af_v4_specific = af;
 869                 break;
 870         case AF_INET6:
 871                 if (sctp_af_v6_specific)
 872                         return 0;
 873                 sctp_af_v6_specific = af;
 874                 break;
 875         default:
 876                 return 0;
 877         }
 878
 879         INIT_LIST_HEAD(&af->list);
 880         list_add_tail(&af->list, &sctp_address_families);
 881         return 1;
 882 }
 883
 884 /* Get the table of functions for manipulating a particular address
 885  * family.
 886  */
 887 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 888 {
 889         switch (family) {
 890         case AF_INET:
 891                 return sctp_af_v4_specific;
 892         case AF_INET6:
 893                 return sctp_af_v6_specific;
 894         default:
 895                 return NULL;
 896         }
 897 }
 898
 899 /* Common code to initialize a AF_INET msg_name. */
 900 static void sctp_inet_msgname(char *msgname, int *addr_len)
 901 {
 902         struct sockaddr_in *sin;
 903
 904         sin = (struct sockaddr_in *)msgname;
 905         *addr_len = sizeof(struct sockaddr_in);
 906         sin->sin_family = AF_INET;
 907         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 908 }
 909
 910 /* Copy the primary address of the peer primary address as the msg_name. */
 911 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 912                                     int *addr_len)
 913 {
 914         struct sockaddr_in *sin, *sinfrom;
 915
 916         if (msgname) {
 917                 struct sctp_association *asoc;
 918
 919                 asoc = event->asoc;
 920                 sctp_inet_msgname(msgname, addr_len);
 921                 sin = (struct sockaddr_in *)msgname;
 922                 sinfrom = &asoc->peer.primary_addr.v4;
 923                 sin->sin_port = htons(asoc->peer.port);
 924                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 925         }
 926 }
 927
 928 /* Initialize and copy out a msgname from an inbound skb. */
 929 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 930 {
 931         if (msgname) {
 932                 struct sctphdr *sh = sctp_hdr(skb);
 933                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 934
 935                 sctp_inet_msgname(msgname, len);
 936                 sin->sin_port = sh->source;
 937                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 938         }
 939 }
 940
 941 /* Do we support this AF? */
 942 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 943 {
 944         /* PF_INET only supports AF_INET addresses. */
 945         return AF_INET == family;
 946 }
 947
 948 /* Address matching with wildcards allowed. */
 949 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 950                               const union sctp_addr *addr2,
 951                               struct sctp_sock *opt)
 952 {
 953         /* PF_INET only supports AF_INET addresses. */
 954         if (addr1->sa.sa_family != addr2->sa.sa_family)
 955                 return 0;
 956         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 957             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 958                 return 1;
 959         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 960                 return 1;
 961
 962         return 0;
 963 }
 964
 965 /* Verify that provided sockaddr looks bindable.  Common verification has
 966  * already been taken care of.
 967  */
 968 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 969 {
 970         return sctp_v4_available(addr, opt);
 971 }
 972
 973 /* Verify that sockaddr looks sendable.  Common verification has already
 974  * been taken care of.
 975  */
 976 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 977 {
 978         return 1;
 979 }
 980
 981 /* Fill in Supported Address Type information for INIT and INIT-ACK
 982  * chunks.  Returns number of addresses supported.
 983  */
 984 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 985                                      __be16 *types)
 986 {
 987         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 988         return 1;
 989 }
 990
 991 /* Wrapper routine that calls the ip transmit routine. */
 992 static inline int sctp_v4_xmit(struct sk_buff *skb,
 993                                struct sctp_transport *transport)
 994 {
 995         struct inet_sock *inet = inet_sk(skb->sk);
 996
 997         pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
 998                  skb->len, &transport->fl.u.ip4.saddr, &transport->fl.u.ip4.daddr);
 999
1000         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
1001                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
1002
1003         SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
1004
1005         return ip_queue_xmit(&inet->sk, skb, &transport->fl);
1006 }
1007
1008 static struct sctp_af sctp_af_inet;
1009
1010 static struct sctp_pf sctp_pf_inet = {
1011         .event_msgname = sctp_inet_event_msgname,
1012         .skb_msgname   = sctp_inet_skb_msgname,
1013         .af_supported  = sctp_inet_af_supported,
1014         .cmp_addr      = sctp_inet_cmp_addr,
1015         .bind_verify   = sctp_inet_bind_verify,
1016         .send_verify   = sctp_inet_send_verify,
1017         .supported_addrs = sctp_inet_supported_addrs,
1018         .create_accept_sk = sctp_v4_create_accept_sk,
1019         .addr_to_user  = sctp_v4_addr_to_user,
1020         .to_sk_saddr   = sctp_v4_to_sk_saddr,
1021         .to_sk_daddr   = sctp_v4_to_sk_daddr,
1022         .af            = &sctp_af_inet
1023 };
1024
1025 /* Notifier for inetaddr addition/deletion events.  */
1026 static struct notifier_block sctp_inetaddr_notifier = {
1027         .notifier_call = sctp_inetaddr_event,
1028 };
1029
1030 /* Socket operations.  */
1031 static const struct proto_ops inet_seqpacket_ops = {
1032         .family            = PF_INET,
1033         .owner             = THIS_MODULE,
1034         .release           = inet_release,      /* Needs to be wrapped... */
1035         .bind              = inet_bind,
1036         .connect           = sctp_inet_connect,
1037         .socketpair        = sock_no_socketpair,
1038         .accept            = inet_accept,
1039         .getname           = inet_getname,      /* Semantics are different.  */
1040         .poll              = sctp_poll,
1041         .ioctl             = inet_ioctl,
1042         .listen            = sctp_inet_listen,
1043         .shutdown          = inet_shutdown,     /* Looks harmless.  */
1044         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1045         .getsockopt        = sock_common_getsockopt,
1046         .sendmsg           = inet_sendmsg,
1047         .recvmsg           = sock_common_recvmsg,
1048         .mmap              = sock_no_mmap,
1049         .sendpage          = sock_no_sendpage,
1050 #ifdef CONFIG_COMPAT
1051         .compat_setsockopt = compat_sock_common_setsockopt,
1052         .compat_getsockopt = compat_sock_common_getsockopt,
1053 #endif
1054 };
1055
1056 /* Registration with AF_INET family.  */
1057 static struct inet_protosw sctp_seqpacket_protosw = {
1058         .type       = SOCK_SEQPACKET,
1059         .protocol   = IPPROTO_SCTP,
1060         .prot       = &sctp_prot,
1061         .ops        = &inet_seqpacket_ops,
1062         .flags      = SCTP_PROTOSW_FLAG
1063 };
1064 static struct inet_protosw sctp_stream_protosw = {
1065         .type       = SOCK_STREAM,
1066         .protocol   = IPPROTO_SCTP,
1067         .prot       = &sctp_prot,
1068         .ops        = &inet_seqpacket_ops,
1069         .flags      = SCTP_PROTOSW_FLAG
1070 };
1071
1072 /* Register with IP layer.  */
1073 static const struct net_protocol sctp_protocol = {
1074         .handler     = sctp_rcv,
1075         .err_handler = sctp_v4_err,
1076         .no_policy   = 1,
1077         .netns_ok    = 1,
1078         .icmp_strict_tag_validation = 1,
1079 };
1080
1081 /* IPv4 address related functions.  */
1082 static struct sctp_af sctp_af_inet = {
1083         .sa_family         = AF_INET,
1084         .sctp_xmit         = sctp_v4_xmit,
1085         .setsockopt        = ip_setsockopt,
1086         .getsockopt        = ip_getsockopt,
1087         .get_dst           = sctp_v4_get_dst,
1088         .get_saddr         = sctp_v4_get_saddr,
1089         .copy_addrlist     = sctp_v4_copy_addrlist,
1090         .from_skb          = sctp_v4_from_skb,
1091         .from_sk           = sctp_v4_from_sk,
1092         .from_addr_param   = sctp_v4_from_addr_param,
1093         .to_addr_param     = sctp_v4_to_addr_param,
1094         .cmp_addr          = sctp_v4_cmp_addr,
1095         .addr_valid        = sctp_v4_addr_valid,
1096         .inaddr_any        = sctp_v4_inaddr_any,
1097         .is_any            = sctp_v4_is_any,
1098         .available         = sctp_v4_available,
1099         .scope             = sctp_v4_scope,
1100         .skb_iif           = sctp_v4_skb_iif,
1101         .is_ce             = sctp_v4_is_ce,
1102         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1103         .ecn_capable       = sctp_v4_ecn_capable,
1104         .net_header_len    = sizeof(struct iphdr),
1105         .sockaddr_len      = sizeof(struct sockaddr_in),
1106 #ifdef CONFIG_COMPAT
1107         .compat_setsockopt = compat_ip_setsockopt,
1108         .compat_getsockopt = compat_ip_getsockopt,
1109 #endif
1110 };
1111
1112 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1113 {
1114         switch (family) {
1115         case PF_INET:
1116                 return sctp_pf_inet_specific;
1117         case PF_INET6:
1118                 return sctp_pf_inet6_specific;
1119         default:
1120                 return NULL;
1121         }
1122 }
1123
1124 /* Register the PF specific function table.  */
1125 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1126 {
1127         switch (family) {
1128         case PF_INET:
1129                 if (sctp_pf_inet_specific)
1130                         return 0;
1131                 sctp_pf_inet_specific = pf;
1132                 break;
1133         case PF_INET6:
1134                 if (sctp_pf_inet6_specific)
1135                         return 0;
1136                 sctp_pf_inet6_specific = pf;
1137                 break;
1138         default:
1139                 return 0;
1140         }
1141         return 1;
1142 }
1143
1144 static inline int init_sctp_mibs(struct net *net)
1145 {
1146         net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1147         if (!net->sctp.sctp_statistics)
1148                 return -ENOMEM;
1149         return 0;
1150 }
1151
1152 static inline void cleanup_sctp_mibs(struct net *net)
1153 {
1154         free_percpu(net->sctp.sctp_statistics);
1155 }
1156
1157 static void sctp_v4_pf_init(void)
1158 {
1159         /* Initialize the SCTP specific PF functions. */
1160         sctp_register_pf(&sctp_pf_inet, PF_INET);
1161         sctp_register_af(&sctp_af_inet);
1162 }
1163
1164 static void sctp_v4_pf_exit(void)
1165 {
1166         list_del(&sctp_af_inet.list);
1167 }
1168
1169 static int sctp_v4_protosw_init(void)
1170 {
1171         int rc;
1172
1173         rc = proto_register(&sctp_prot, 1);
1174         if (rc)
1175                 return rc;
1176
1177         /* Register SCTP(UDP and TCP style) with socket layer.  */
1178         inet_register_protosw(&sctp_seqpacket_protosw);
1179         inet_register_protosw(&sctp_stream_protosw);
1180
1181         return 0;
1182 }
1183
1184 static void sctp_v4_protosw_exit(void)
1185 {
1186         inet_unregister_protosw(&sctp_stream_protosw);
1187         inet_unregister_protosw(&sctp_seqpacket_protosw);
1188         proto_unregister(&sctp_prot);
1189 }
1190
1191 static int sctp_v4_add_protocol(void)
1192 {
1193         /* Register notifier for inet address additions/deletions. */
1194         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1195
1196         /* Register SCTP with inet layer.  */
1197         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1198                 return -EAGAIN;
1199
1200         return 0;
1201 }
1202
1203 static void sctp_v4_del_protocol(void)
1204 {
1205         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1206         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1207 }
1208
1209 static int __net_init sctp_defaults_init(struct net *net)
1210 {
1211         int status;
1212
1213         /*
1214          * 14. Suggested SCTP Protocol Parameter Values
1215          */
1216         /* The following protocol parameters are RECOMMENDED:  */
1217         /* RTO.Initial              - 3  seconds */
1218         net->sctp.rto_initial                   = SCTP_RTO_INITIAL;
1219         /* RTO.Min                  - 1  second */
1220         net->sctp.rto_min                       = SCTP_RTO_MIN;
1221         /* RTO.Max                 -  60 seconds */
1222         net->sctp.rto_max                       = SCTP_RTO_MAX;
1223         /* RTO.Alpha                - 1/8 */
1224         net->sctp.rto_alpha                     = SCTP_RTO_ALPHA;
1225         /* RTO.Beta                 - 1/4 */
1226         net->sctp.rto_beta                      = SCTP_RTO_BETA;
1227
1228         /* Valid.Cookie.Life        - 60  seconds */
1229         net->sctp.valid_cookie_life             = SCTP_DEFAULT_COOKIE_LIFE;
1230
1231         /* Whether Cookie Preservative is enabled(1) or not(0) */
1232         net->sctp.cookie_preserve_enable        = 1;
1233
1234         /* Default sctp sockets to use md5 as their hmac alg */
1235 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1236         net->sctp.sctp_hmac_alg                 = "md5";
1237 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1238         net->sctp.sctp_hmac_alg                 = "sha1";
1239 #else
1240         net->sctp.sctp_hmac_alg                 = NULL;
1241 #endif
1242
1243         /* Max.Burst                - 4 */
1244         net->sctp.max_burst                     = SCTP_DEFAULT_MAX_BURST;
1245
1246         /* Association.Max.Retrans  - 10 attempts
1247          * Path.Max.Retrans         - 5  attempts (per destination address)
1248          * Max.Init.Retransmits     - 8  attempts
1249          */
1250         net->sctp.max_retrans_association       = 10;
1251         net->sctp.max_retrans_path              = 5;
1252         net->sctp.max_retrans_init              = 8;
1253
1254         /* Sendbuffer growth        - do per-socket accounting */
1255         net->sctp.sndbuf_policy                 = 0;
1256
1257         /* Rcvbuffer growth         - do per-socket accounting */
1258         net->sctp.rcvbuf_policy                 = 0;
1259
1260         /* HB.interval              - 30 seconds */
1261         net->sctp.hb_interval                   = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1262
1263         /* delayed SACK timeout */
1264         net->sctp.sack_timeout                  = SCTP_DEFAULT_TIMEOUT_SACK;
1265
1266         /* Disable ADDIP by default. */
1267         net->sctp.addip_enable = 0;
1268         net->sctp.addip_noauth = 0;
1269         net->sctp.default_auto_asconf = 0;
1270
1271         /* Enable PR-SCTP by default. */
1272         net->sctp.prsctp_enable = 1;
1273
1274         /* Disable AUTH by default. */
1275         net->sctp.auth_enable = 0;
1276
1277         /* Set SCOPE policy to enabled */
1278         net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1279
1280         /* Set the default rwnd update threshold */
1281         net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1282
1283         /* Initialize maximum autoclose timeout. */
1284         net->sctp.max_autoclose         = INT_MAX / HZ;
1285
1286         status = sctp_sysctl_net_register(net);
1287         if (status)
1288                 goto err_sysctl_register;
1289
1290         /* Allocate and initialise sctp mibs.  */
1291         status = init_sctp_mibs(net);
1292         if (status)
1293                 goto err_init_mibs;
1294
1295         /* Initialize proc fs directory.  */
1296         status = sctp_proc_init(net);
1297         if (status)
1298                 goto err_init_proc;
1299
1300         sctp_dbg_objcnt_init(net);
1301
1302         /* Initialize the local address list. */
1303         INIT_LIST_HEAD(&net->sctp.local_addr_list);
1304         spin_lock_init(&net->sctp.local_addr_lock);
1305         sctp_get_local_addr_list(net);
1306
1307         /* Initialize the address event list */
1308         INIT_LIST_HEAD(&net->sctp.addr_waitq);
1309         INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1310         spin_lock_init(&net->sctp.addr_wq_lock);
1311         net->sctp.addr_wq_timer.expires = 0;
1312         setup_timer(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler,
1313                     (unsigned long)net);
1314
1315         return 0;
1316
1317 err_init_proc:
1318         cleanup_sctp_mibs(net);
1319 err_init_mibs:
1320         sctp_sysctl_net_unregister(net);
1321 err_sysctl_register:
1322         return status;
1323 }
1324
1325 static void __net_exit sctp_defaults_exit(struct net *net)
1326 {
1327         /* Free the local address list */
1328         sctp_free_addr_wq(net);
1329         sctp_free_local_addr_list(net);
1330
1331         sctp_dbg_objcnt_exit(net);
1332
1333         sctp_proc_exit(net);
1334         cleanup_sctp_mibs(net);
1335         sctp_sysctl_net_unregister(net);
1336 }
1337
1338 static struct pernet_operations sctp_defaults_ops = {
1339         .init = sctp_defaults_init,
1340         .exit = sctp_defaults_exit,
1341 };
1342
1343 static int __net_init sctp_ctrlsock_init(struct net *net)
1344 {
1345         int status;
1346
1347         /* Initialize the control inode/socket for handling OOTB packets.  */
1348         status = sctp_ctl_sock_init(net);
1349         if (status)
1350                 pr_err("Failed to initialize the SCTP control sock\n");
1351
1352         return status;
1353 }
1354
1355 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1356 {
1357         /* Free the control endpoint.  */
1358         inet_ctl_sock_destroy(net->sctp.ctl_sock);
1359 }
1360
1361 static struct pernet_operations sctp_ctrlsock_ops = {
1362         .init = sctp_ctrlsock_init,
1363         .exit = sctp_ctrlsock_exit,
1364 };
1365
1366 /* Initialize the universe into something sensible.  */
1367 static __init int sctp_init(void)
1368 {
1369         int i;
1370         int status = -EINVAL;
1371         unsigned long goal;
1372         unsigned long limit;
1373         int max_share;
1374         int order;
1375         int num_entries;
1376         int max_entry_order;
1377
1378         sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1379
1380         /* Allocate bind_bucket and chunk caches. */
1381         status = -ENOBUFS;
1382         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1383                                                sizeof(struct sctp_bind_bucket),
1384                                                0, SLAB_HWCACHE_ALIGN,
1385                                                NULL);
1386         if (!sctp_bucket_cachep)
1387                 goto out;
1388
1389         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1390                                                sizeof(struct sctp_chunk),
1391                                                0, SLAB_HWCACHE_ALIGN,
1392                                                NULL);
1393         if (!sctp_chunk_cachep)
1394                 goto err_chunk_cachep;
1395
1396         status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1397         if (status)
1398                 goto err_percpu_counter_init;
1399
1400         /* Implementation specific variables. */
1401
1402         /* Initialize default stream count setup information. */
1403         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1404         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1405
1406         /* Initialize handle used for association ids. */
1407         idr_init(&sctp_assocs_id);
1408
1409         limit = nr_free_buffer_pages() / 8;
1410         limit = max(limit, 128UL);
1411         sysctl_sctp_mem[0] = limit / 4 * 3;
1412         sysctl_sctp_mem[1] = limit;
1413         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1414
1415         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1416         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1417         max_share = min(4UL*1024*1024, limit);
1418
1419         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1420         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1421         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1422
1423         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1424         sysctl_sctp_wmem[1] = 16*1024;
1425         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1426
1427         /* Size and allocate the association hash table.
1428          * The methodology is similar to that of the tcp hash tables.
1429          * Though not identical.  Start by getting a goal size
1430          */
1431         if (totalram_pages >= (128 * 1024))
1432                 goal = totalram_pages >> (22 - PAGE_SHIFT);
1433         else
1434                 goal = totalram_pages >> (24 - PAGE_SHIFT);
1435
1436         /* Then compute the page order for said goal */
1437         order = get_order(goal);
1438
1439         /* Now compute the required page order for the maximum sized table we
1440          * want to create
1441          */
1442         max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1443                                     sizeof(struct sctp_bind_hashbucket));
1444
1445         /* Limit the page order by that maximum hash table size */
1446         order = min(order, max_entry_order);
1447
1448         do {
1449                 sctp_assoc_hashsize = (1UL << order) * PAGE_SIZE /
1450                                         sizeof(struct sctp_hashbucket);
1451                 if ((sctp_assoc_hashsize > (64 * 1024)) && order > 0)
1452                         continue;
1453                 sctp_assoc_hashtable = (struct sctp_hashbucket *)
1454                         __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1455         } while (!sctp_assoc_hashtable && --order > 0);
1456         if (!sctp_assoc_hashtable) {
1457                 pr_err("Failed association hash alloc\n");
1458                 status = -ENOMEM;
1459                 goto err_ahash_alloc;
1460         }
1461         for (i = 0; i < sctp_assoc_hashsize; i++) {
1462                 rwlock_init(&sctp_assoc_hashtable[i].lock);
1463                 INIT_HLIST_HEAD(&sctp_assoc_hashtable[i].chain);
1464         }
1465
1466         /* Allocate and initialize the endpoint hash table.  */
1467         sctp_ep_hashsize = 64;
1468         sctp_ep_hashtable =
1469                 kmalloc(64 * sizeof(struct sctp_hashbucket), GFP_KERNEL);
1470         if (!sctp_ep_hashtable) {
1471                 pr_err("Failed endpoint_hash alloc\n");
1472                 status = -ENOMEM;
1473                 goto err_ehash_alloc;
1474         }
1475         for (i = 0; i < sctp_ep_hashsize; i++) {
1476                 rwlock_init(&sctp_ep_hashtable[i].lock);
1477                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1478         }
1479
1480         /* Allocate and initialize the SCTP port hash table.
1481          * Note that order is initalized to start at the max sized
1482          * table we want to support.  If we can't get that many pages
1483          * reduce the order and try again
1484          */
1485         do {
1486                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1487                         __get_free_pages(GFP_ATOMIC|__GFP_NOWARN, order);
1488         } while (!sctp_port_hashtable && --order > 0);
1489
1490         if (!sctp_port_hashtable) {
1491                 pr_err("Failed bind hash alloc\n");
1492                 status = -ENOMEM;
1493                 goto err_bhash_alloc;
1494         }
1495
1496         /* Now compute the number of entries that will fit in the
1497          * port hash space we allocated
1498          */
1499         num_entries = (1UL << order) * PAGE_SIZE /
1500                       sizeof(struct sctp_bind_hashbucket);
1501
1502         /* And finish by rounding it down to the nearest power of two
1503          * this wastes some memory of course, but its needed because
1504          * the hash function operates based on the assumption that
1505          * that the number of entries is a power of two
1506          */
1507         sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1508
1509         for (i = 0; i < sctp_port_hashsize; i++) {
1510                 spin_lock_init(&sctp_port_hashtable[i].lock);
1511                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1512         }
1513
1514         pr_info("Hash tables configured (established %d bind %d)\n",
1515                 sctp_assoc_hashsize, sctp_port_hashsize);
1516
1517         sctp_sysctl_register();
1518
1519         INIT_LIST_HEAD(&sctp_address_families);
1520         sctp_v4_pf_init();
1521         sctp_v6_pf_init();
1522
1523         status = register_pernet_subsys(&sctp_defaults_ops);
1524         if (status)
1525                 goto err_register_defaults;
1526
1527         status = sctp_v4_protosw_init();
1528         if (status)
1529                 goto err_protosw_init;
1530
1531         status = sctp_v6_protosw_init();
1532         if (status)
1533                 goto err_v6_protosw_init;
1534
1535         status = register_pernet_subsys(&sctp_ctrlsock_ops);
1536         if (status)
1537                 goto err_register_ctrlsock;
1538
1539         status = sctp_v4_add_protocol();
1540         if (status)
1541                 goto err_add_protocol;
1542
1543         /* Register SCTP with inet6 layer.  */
1544         status = sctp_v6_add_protocol();
1545         if (status)
1546                 goto err_v6_add_protocol;
1547
1548 out:
1549         return status;
1550 err_v6_add_protocol:
1551         sctp_v4_del_protocol();
1552 err_add_protocol:
1553         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1554 err_register_ctrlsock:
1555         sctp_v6_protosw_exit();
1556 err_v6_protosw_init:
1557         sctp_v4_protosw_exit();
1558 err_protosw_init:
1559         unregister_pernet_subsys(&sctp_defaults_ops);
1560 err_register_defaults:
1561         sctp_v4_pf_exit();
1562         sctp_v6_pf_exit();
1563         sctp_sysctl_unregister();
1564         free_pages((unsigned long)sctp_port_hashtable,
1565                    get_order(sctp_port_hashsize *
1566                              sizeof(struct sctp_bind_hashbucket)));
1567 err_bhash_alloc:
1568         kfree(sctp_ep_hashtable);
1569 err_ehash_alloc:
1570         free_pages((unsigned long)sctp_assoc_hashtable,
1571                    get_order(sctp_assoc_hashsize *
1572                              sizeof(struct sctp_hashbucket)));
1573 err_ahash_alloc:
1574         percpu_counter_destroy(&sctp_sockets_allocated);
1575 err_percpu_counter_init:
1576         kmem_cache_destroy(sctp_chunk_cachep);
1577 err_chunk_cachep:
1578         kmem_cache_destroy(sctp_bucket_cachep);
1579         goto out;
1580 }
1581
1582 /* Exit handler for the SCTP protocol.  */
1583 static __exit void sctp_exit(void)
1584 {
1585         /* BUG.  This should probably do something useful like clean
1586          * up all the remaining associations and all that memory.
1587          */
1588
1589         /* Unregister with inet6/inet layers. */
1590         sctp_v6_del_protocol();
1591         sctp_v4_del_protocol();
1592
1593         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1594
1595         /* Free protosw registrations */
1596         sctp_v6_protosw_exit();
1597         sctp_v4_protosw_exit();
1598
1599         unregister_pernet_subsys(&sctp_defaults_ops);
1600
1601         /* Unregister with socket layer. */
1602         sctp_v6_pf_exit();
1603         sctp_v4_pf_exit();
1604
1605         sctp_sysctl_unregister();
1606
1607         free_pages((unsigned long)sctp_assoc_hashtable,
1608                    get_order(sctp_assoc_hashsize *
1609                              sizeof(struct sctp_hashbucket)));
1610         kfree(sctp_ep_hashtable);
1611         free_pages((unsigned long)sctp_port_hashtable,
1612                    get_order(sctp_port_hashsize *
1613                              sizeof(struct sctp_bind_hashbucket)));
1614
1615         percpu_counter_destroy(&sctp_sockets_allocated);
1616
1617         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1618
1619         kmem_cache_destroy(sctp_chunk_cachep);
1620         kmem_cache_destroy(sctp_bucket_cachep);
1621 }
1622
1623 module_init(sctp_init);
1624 module_exit(sctp_exit);
1625
1626 /*
1627  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1628  */
1629 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1630 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1631 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1632 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1633 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1634 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1635 MODULE_LICENSE("GPL");