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 <[email protected]>
  32  *
  33  * Written or modified by:
  34  *    La Monte H.P. Yarroll <[email protected]>
  35  *    Karl Knutson <[email protected]>
  36  *    Jon Grimm <[email protected]>
  37  *    Sridhar Samudrala <[email protected]>
  38  *    Daisy Chang <[email protected]>
  39  *    Ardelle Fan <[email protected]>
  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/memblock.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_addr.s_addr = ifa->ifa_local;
 105                         addr->valid = 1;
 106                         INIT_LIST_HEAD(&addr->list);
 107                         list_add_tail(&addr->list, addrlist);
 108                 }
 109         }
 110
 111         rcu_read_unlock();
 112 }
 113
 114 /* Extract our IP addresses from the system and stash them in the
 115  * protocol structure.
 116  */
 117 static void sctp_get_local_addr_list(struct net *net)
 118 {
 119         struct net_device *dev;
 120         struct list_head *pos;
 121         struct sctp_af *af;
 122
 123         rcu_read_lock();
 124         for_each_netdev_rcu(net, dev) {
 125                 list_for_each(pos, &sctp_address_families) {
 126                         af = list_entry(pos, struct sctp_af, list);
 127                         af->copy_addrlist(&net->sctp.local_addr_list, dev);
 128                 }
 129         }
 130         rcu_read_unlock();
 131 }
 132
 133 /* Free the existing local addresses.  */
 134 static void sctp_free_local_addr_list(struct net *net)
 135 {
 136         struct sctp_sockaddr_entry *addr;
 137         struct list_head *pos, *temp;
 138
 139         list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
 140                 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
 141                 list_del(pos);
 142                 kfree(addr);
 143         }
 144 }
 145
 146 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
 147 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
 148                               enum sctp_scope scope, gfp_t gfp, int copy_flags)
 149 {
 150         struct sctp_sockaddr_entry *addr;
 151         union sctp_addr laddr;
 152         int error = 0;
 153
 154         rcu_read_lock();
 155         list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
 156                 if (!addr->valid)
 157                         continue;
 158                 if (!sctp_in_scope(net, &addr->a, scope))
 159                         continue;
 160
 161                 /* Now that the address is in scope, check to see if
 162                  * the address type is really supported by the local
 163                  * sock as well as the remote peer.
 164                  */
 165                 if (addr->a.sa.sa_family == AF_INET &&
 166                     !(copy_flags & SCTP_ADDR4_PEERSUPP))
 167                         continue;
 168                 if (addr->a.sa.sa_family == AF_INET6 &&
 169                     (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
 170                      !(copy_flags & SCTP_ADDR6_PEERSUPP)))
 171                         continue;
 172
 173                 laddr = addr->a;
 174                 /* also works for setting ipv6 address port */
 175                 laddr.v4.sin_port = htons(bp->port);
 176                 if (sctp_bind_addr_state(bp, &laddr) != -1)
 177                         continue;
 178
 179                 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
 180                                            SCTP_ADDR_SRC, GFP_ATOMIC);
 181                 if (error)
 182                         break;
 183         }
 184
 185         rcu_read_unlock();
 186         return error;
 187 }
 188
 189 /* Copy over any ip options */
 190 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
 191 {
 192         struct inet_sock *newinet, *inet = inet_sk(sk);
 193         struct ip_options_rcu *inet_opt, *newopt = NULL;
 194
 195         newinet = inet_sk(newsk);
 196
 197         rcu_read_lock();
 198         inet_opt = rcu_dereference(inet->inet_opt);
 199         if (inet_opt) {
 200                 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
 201                                       inet_opt->opt.optlen, GFP_ATOMIC);
 202                 if (newopt)
 203                         memcpy(newopt, inet_opt, sizeof(*inet_opt) +
 204                                inet_opt->opt.optlen);
 205                 else
 206                         pr_err("%s: Failed to copy ip options\n", __func__);
 207         }
 208         RCU_INIT_POINTER(newinet->inet_opt, newopt);
 209         rcu_read_unlock();
 210 }
 211
 212 /* Account for the IP options */
 213 static int sctp_v4_ip_options_len(struct sock *sk)
 214 {
 215         struct inet_sock *inet = inet_sk(sk);
 216         struct ip_options_rcu *inet_opt;
 217         int len = 0;
 218
 219         rcu_read_lock();
 220         inet_opt = rcu_dereference(inet->inet_opt);
 221         if (inet_opt)
 222                 len = inet_opt->opt.optlen;
 223
 224         rcu_read_unlock();
 225         return len;
 226 }
 227
 228 /* Initialize a sctp_addr from in incoming skb.  */
 229 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
 230                              int is_saddr)
 231 {
 232         /* Always called on head skb, so this is safe */
 233         struct sctphdr *sh = sctp_hdr(skb);
 234         struct sockaddr_in *sa = &addr->v4;
 235
 236         addr->v4.sin_family = AF_INET;
 237
 238         if (is_saddr) {
 239                 sa->sin_port = sh->source;
 240                 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
 241         } else {
 242                 sa->sin_port = sh->dest;
 243                 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
 244         }
 245 }
 246
 247 /* Initialize an sctp_addr from a socket. */
 248 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
 249 {
 250         addr->v4.sin_family = AF_INET;
 251         addr->v4.sin_port = 0;
 252         addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
 253 }
 254
 255 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
 256 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
 257 {
 258         inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
 259 }
 260
 261 /* Initialize sk->sk_daddr from sctp_addr. */
 262 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
 263 {
 264         inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
 265 }
 266
 267 /* Initialize a sctp_addr from an address parameter. */
 268 static void sctp_v4_from_addr_param(union sctp_addr *addr,
 269                                     union sctp_addr_param *param,
 270                                     __be16 port, int iif)
 271 {
 272         addr->v4.sin_family = AF_INET;
 273         addr->v4.sin_port = port;
 274         addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
 275 }
 276
 277 /* Initialize an address parameter from a sctp_addr and return the length
 278  * of the address parameter.
 279  */
 280 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
 281                                  union sctp_addr_param *param)
 282 {
 283         int length = sizeof(struct sctp_ipv4addr_param);
 284
 285         param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
 286         param->v4.param_hdr.length = htons(length);
 287         param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
 288
 289         return length;
 290 }
 291
 292 /* Initialize a sctp_addr from a dst_entry. */
 293 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
 294                               __be16 port)
 295 {
 296         saddr->v4.sin_family = AF_INET;
 297         saddr->v4.sin_port = port;
 298         saddr->v4.sin_addr.s_addr = fl4->saddr;
 299 }
 300
 301 /* Compare two addresses exactly. */
 302 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
 303                             const union sctp_addr *addr2)
 304 {
 305         if (addr1->sa.sa_family != addr2->sa.sa_family)
 306                 return 0;
 307         if (addr1->v4.sin_port != addr2->v4.sin_port)
 308                 return 0;
 309         if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
 310                 return 0;
 311
 312         return 1;
 313 }
 314
 315 /* Initialize addr struct to INADDR_ANY. */
 316 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
 317 {
 318         addr->v4.sin_family = AF_INET;
 319         addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
 320         addr->v4.sin_port = port;
 321 }
 322
 323 /* Is this a wildcard address? */
 324 static int sctp_v4_is_any(const union sctp_addr *addr)
 325 {
 326         return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
 327 }
 328
 329 /* This function checks if the address is a valid address to be used for
 330  * SCTP binding.
 331  *
 332  * Output:
 333  * Return 0 - If the address is a non-unicast or an illegal address.
 334  * Return 1 - If the address is a unicast.
 335  */
 336 static int sctp_v4_addr_valid(union sctp_addr *addr,
 337                               struct sctp_sock *sp,
 338                               const struct sk_buff *skb)
 339 {
 340         /* IPv4 addresses not allowed */
 341         if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
 342                 return 0;
 343
 344         /* Is this a non-unicast address or a unusable SCTP address? */
 345         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
 346                 return 0;
 347
 348         /* Is this a broadcast address? */
 349         if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
 350                 return 0;
 351
 352         return 1;
 353 }
 354
 355 /* Should this be available for binding?   */
 356 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
 357 {
 358         struct net *net = sock_net(&sp->inet.sk);
 359         int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
 360
 361
 362         if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
 363            ret != RTN_LOCAL &&
 364            !sp->inet.freebind &&
 365            !net->ipv4.sysctl_ip_nonlocal_bind)
 366                 return 0;
 367
 368         if (ipv6_only_sock(sctp_opt2sk(sp)))
 369                 return 0;
 370
 371         return 1;
 372 }
 373
 374 /* Checking the loopback, private and other address scopes as defined in
 375  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
 376  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
 377  *
 378  * Level 0 - unusable SCTP addresses
 379  * Level 1 - loopback address
 380  * Level 2 - link-local addresses
 381  * Level 3 - private addresses.
 382  * Level 4 - global addresses
 383  * For INIT and INIT-ACK address list, let L be the level of
 384  * of requested destination address, sender and receiver
 385  * SHOULD include all of its addresses with level greater
 386  * than or equal to L.
 387  *
 388  * IPv4 scoping can be controlled through sysctl option
 389  * net.sctp.addr_scope_policy
 390  */
 391 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
 392 {
 393         enum sctp_scope retval;
 394
 395         /* Check for unusable SCTP addresses. */
 396         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
 397                 retval =  SCTP_SCOPE_UNUSABLE;
 398         } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
 399                 retval = SCTP_SCOPE_LOOPBACK;
 400         } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
 401                 retval = SCTP_SCOPE_LINK;
 402         } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
 403                    ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
 404                    ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
 405                 retval = SCTP_SCOPE_PRIVATE;
 406         } else {
 407                 retval = SCTP_SCOPE_GLOBAL;
 408         }
 409
 410         return retval;
 411 }
 412
 413 /* Returns a valid dst cache entry for the given source and destination ip
 414  * addresses. If an association is passed, trys to get a dst entry with a
 415  * source address that matches an address in the bind address list.
 416  */
 417 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
 418                                 struct flowi *fl, struct sock *sk)
 419 {
 420         struct sctp_association *asoc = t->asoc;
 421         struct rtable *rt;
 422         struct flowi4 *fl4 = &fl->u.ip4;
 423         struct sctp_bind_addr *bp;
 424         struct sctp_sockaddr_entry *laddr;
 425         struct dst_entry *dst = NULL;
 426         union sctp_addr *daddr = &t->ipaddr;
 427         union sctp_addr dst_saddr;
 428         __u8 tos = inet_sk(sk)->tos;
 429
 430         if (t->dscp & SCTP_DSCP_SET_MASK)
 431                 tos = t->dscp & SCTP_DSCP_VAL_MASK;
 432         memset(fl4, 0x0, sizeof(struct flowi4));
 433         fl4->daddr  = daddr->v4.sin_addr.s_addr;
 434         fl4->fl4_dport = daddr->v4.sin_port;
 435         fl4->flowi4_proto = IPPROTO_SCTP;
 436         if (asoc) {
 437                 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
 438                 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
 439                 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
 440         }
 441         if (saddr) {
 442                 fl4->saddr = saddr->v4.sin_addr.s_addr;
 443                 if (!fl4->fl4_sport)
 444                         fl4->fl4_sport = saddr->v4.sin_port;
 445         }
 446
 447         pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
 448                  &fl4->saddr);
 449
 450         rt = ip_route_output_key(sock_net(sk), fl4);
 451         if (!IS_ERR(rt))
 452                 dst = &rt->dst;
 453
 454         /* If there is no association or if a source address is passed, no
 455          * more validation is required.
 456          */
 457         if (!asoc || saddr)
 458                 goto out;
 459
 460         bp = &asoc->base.bind_addr;
 461
 462         if (dst) {
 463                 /* Walk through the bind address list and look for a bind
 464                  * address that matches the source address of the returned dst.
 465                  */
 466                 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
 467                 rcu_read_lock();
 468                 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 469                         if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
 470                             (laddr->state != SCTP_ADDR_SRC &&
 471                             !asoc->src_out_of_asoc_ok))
 472                                 continue;
 473                         if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
 474                                 goto out_unlock;
 475                 }
 476                 rcu_read_unlock();
 477
 478                 /* None of the bound addresses match the source address of the
 479                  * dst. So release it.
 480                  */
 481                 dst_release(dst);
 482                 dst = NULL;
 483         }
 484
 485         /* Walk through the bind address list and try to get a dst that
 486          * matches a bind address as the source address.
 487          */
 488         rcu_read_lock();
 489         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 490                 struct net_device *odev;
 491
 492                 if (!laddr->valid)
 493                         continue;
 494                 if (laddr->state != SCTP_ADDR_SRC ||
 495                     AF_INET != laddr->a.sa.sa_family)
 496                         continue;
 497
 498                 fl4->fl4_sport = laddr->a.v4.sin_port;
 499                 flowi4_update_output(fl4,
 500                                      asoc->base.sk->sk_bound_dev_if,
 501                                      RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
 502                                      daddr->v4.sin_addr.s_addr,
 503                                      laddr->a.v4.sin_addr.s_addr);
 504
 505                 rt = ip_route_output_key(sock_net(sk), fl4);
 506                 if (IS_ERR(rt))
 507                         continue;
 508
 509                 /* Ensure the src address belongs to the output
 510                  * interface.
 511                  */
 512                 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
 513                                      false);
 514                 if (!odev || odev->ifindex != fl4->flowi4_oif) {
 515                         if (!dst)
 516                                 dst = &rt->dst;
 517                         else
 518                                 dst_release(&rt->dst);
 519                         continue;
 520                 }
 521
 522                 dst_release(dst);
 523                 dst = &rt->dst;
 524                 break;
 525         }
 526
 527 out_unlock:
 528         rcu_read_unlock();
 529 out:
 530         t->dst = dst;
 531         if (dst)
 532                 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
 533                          &fl4->daddr, &fl4->saddr);
 534         else
 535                 pr_debug("no route\n");
 536 }
 537
 538 /* For v4, the source address is cached in the route entry(dst). So no need
 539  * to cache it separately and hence this is an empty routine.
 540  */
 541 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 542                               struct sctp_transport *t,
 543                               struct flowi *fl)
 544 {
 545         union sctp_addr *saddr = &t->saddr;
 546         struct rtable *rt = (struct rtable *)t->dst;
 547
 548         if (rt) {
 549                 saddr->v4.sin_family = AF_INET;
 550                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 551         }
 552 }
 553
 554 /* What interface did this skb arrive on? */
 555 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 556 {
 557         return inet_iif(skb);
 558 }
 559
 560 /* Was this packet marked by Explicit Congestion Notification? */
 561 static int sctp_v4_is_ce(const struct sk_buff *skb)
 562 {
 563         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 564 }
 565
 566 /* Create and initialize a new sk for the socket returned by accept(). */
 567 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 568                                              struct sctp_association *asoc,
 569                                              bool kern)
 570 {
 571         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 572                         sk->sk_prot, kern);
 573         struct inet_sock *newinet;
 574
 575         if (!newsk)
 576                 goto out;
 577
 578         sock_init_data(NULL, newsk);
 579
 580         sctp_copy_sock(newsk, sk, asoc);
 581         sock_reset_flag(newsk, SOCK_ZAPPED);
 582
 583         sctp_v4_copy_ip_options(sk, newsk);
 584
 585         newinet = inet_sk(newsk);
 586
 587         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 588
 589         sk_refcnt_debug_inc(newsk);
 590
 591         if (newsk->sk_prot->init(newsk)) {
 592                 sk_common_release(newsk);
 593                 newsk = NULL;
 594         }
 595
 596 out:
 597         return newsk;
 598 }
 599
 600 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
 601 {
 602         /* No address mapping for V4 sockets */
 603         return sizeof(struct sockaddr_in);
 604 }
 605
 606 /* Dump the v4 addr to the seq file. */
 607 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 608 {
 609         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 610 }
 611
 612 static void sctp_v4_ecn_capable(struct sock *sk)
 613 {
 614         INET_ECN_xmit(sk);
 615 }
 616
 617 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
 618 {
 619         struct net *net = from_timer(net, t, sctp.addr_wq_timer);
 620         struct sctp_sockaddr_entry *addrw, *temp;
 621         struct sctp_sock *sp;
 622
 623         spin_lock_bh(&net->sctp.addr_wq_lock);
 624
 625         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 626                 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
 627                          "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
 628                          addrw->state, addrw);
 629
 630 #if IS_ENABLED(CONFIG_IPV6)
 631                 /* Now we send an ASCONF for each association */
 632                 /* Note. we currently don't handle link local IPv6 addressees */
 633                 if (addrw->a.sa.sa_family == AF_INET6) {
 634                         struct in6_addr *in6;
 635
 636                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 637                             IPV6_ADDR_LINKLOCAL)
 638                                 goto free_next;
 639
 640                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 641                         if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
 642                             addrw->state == SCTP_ADDR_NEW) {
 643                                 unsigned long timeo_val;
 644
 645                                 pr_debug("%s: this is on DAD, trying %d sec "
 646                                          "later\n", __func__,
 647                                          SCTP_ADDRESS_TICK_DELAY);
 648
 649                                 timeo_val = jiffies;
 650                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 651                                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 652                                 break;
 653                         }
 654                 }
 655 #endif
 656                 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
 657                         struct sock *sk;
 658
 659                         sk = sctp_opt2sk(sp);
 660                         /* ignore bound-specific endpoints */
 661                         if (!sctp_is_ep_boundall(sk))
 662                                 continue;
 663                         bh_lock_sock(sk);
 664                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 665                                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
 666                         bh_unlock_sock(sk);
 667                 }
 668 #if IS_ENABLED(CONFIG_IPV6)
 669 free_next:
 670 #endif
 671                 list_del(&addrw->list);
 672                 kfree(addrw);
 673         }
 674         spin_unlock_bh(&net->sctp.addr_wq_lock);
 675 }
 676
 677 static void sctp_free_addr_wq(struct net *net)
 678 {
 679         struct sctp_sockaddr_entry *addrw;
 680         struct sctp_sockaddr_entry *temp;
 681
 682         spin_lock_bh(&net->sctp.addr_wq_lock);
 683         del_timer(&net->sctp.addr_wq_timer);
 684         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 685                 list_del(&addrw->list);
 686                 kfree(addrw);
 687         }
 688         spin_unlock_bh(&net->sctp.addr_wq_lock);
 689 }
 690
 691 /* lookup the entry for the same address in the addr_waitq
 692  * sctp_addr_wq MUST be locked
 693  */
 694 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
 695                                         struct sctp_sockaddr_entry *addr)
 696 {
 697         struct sctp_sockaddr_entry *addrw;
 698
 699         list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
 700                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 701                         continue;
 702                 if (addrw->a.sa.sa_family == AF_INET) {
 703                         if (addrw->a.v4.sin_addr.s_addr ==
 704                             addr->a.v4.sin_addr.s_addr)
 705                                 return addrw;
 706                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 707                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 708                             &addr->a.v6.sin6_addr))
 709                                 return addrw;
 710                 }
 711         }
 712         return NULL;
 713 }
 714
 715 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
 716 {
 717         struct sctp_sockaddr_entry *addrw;
 718         unsigned long timeo_val;
 719
 720         /* first, we check if an opposite message already exist in the queue.
 721          * If we found such message, it is removed.
 722          * This operation is a bit stupid, but the DHCP client attaches the
 723          * new address after a couple of addition and deletion of that address
 724          */
 725
 726         spin_lock_bh(&net->sctp.addr_wq_lock);
 727         /* Offsets existing events in addr_wq */
 728         addrw = sctp_addr_wq_lookup(net, addr);
 729         if (addrw) {
 730                 if (addrw->state != cmd) {
 731                         pr_debug("%s: offsets existing entry for %d, addr:%pISc "
 732                                  "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
 733                                  &net->sctp.addr_waitq);
 734
 735                         list_del(&addrw->list);
 736                         kfree(addrw);
 737                 }
 738                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 739                 return;
 740         }
 741
 742         /* OK, we have to add the new address to the wait queue */
 743         addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 744         if (addrw == NULL) {
 745                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 746                 return;
 747         }
 748         addrw->state = cmd;
 749         list_add_tail(&addrw->list, &net->sctp.addr_waitq);
 750
 751         pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
 752                  __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
 753
 754         if (!timer_pending(&net->sctp.addr_wq_timer)) {
 755                 timeo_val = jiffies;
 756                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 757                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 758         }
 759         spin_unlock_bh(&net->sctp.addr_wq_lock);
 760 }
 761
 762 /* Event handler for inet address addition/deletion events.
 763  * The sctp_local_addr_list needs to be protocted by a spin lock since
 764  * multiple notifiers (say IPv4 and IPv6) may be running at the same
 765  * time and thus corrupt the list.
 766  * The reader side is protected with RCU.
 767  */
 768 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
 769                                void *ptr)
 770 {
 771         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
 772         struct sctp_sockaddr_entry *addr = NULL;
 773         struct sctp_sockaddr_entry *temp;
 774         struct net *net = dev_net(ifa->ifa_dev->dev);
 775         int found = 0;
 776
 777         switch (ev) {
 778         case NETDEV_UP:
 779                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
 780                 if (addr) {
 781                         addr->a.v4.sin_family = AF_INET;
 782                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 783                         addr->valid = 1;
 784                         spin_lock_bh(&net->sctp.local_addr_lock);
 785                         list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
 786                         sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
 787                         spin_unlock_bh(&net->sctp.local_addr_lock);
 788                 }
 789                 break;
 790         case NETDEV_DOWN:
 791                 spin_lock_bh(&net->sctp.local_addr_lock);
 792                 list_for_each_entry_safe(addr, temp,
 793                                         &net->sctp.local_addr_list, list) {
 794                         if (addr->a.sa.sa_family == AF_INET &&
 795                                         addr->a.v4.sin_addr.s_addr ==
 796                                         ifa->ifa_local) {
 797                                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
 798                                 found = 1;
 799                                 addr->valid = 0;
 800                                 list_del_rcu(&addr->list);
 801                                 break;
 802                         }
 803                 }
 804                 spin_unlock_bh(&net->sctp.local_addr_lock);
 805                 if (found)
 806                         kfree_rcu(addr, rcu);
 807                 break;
 808         }
 809
 810         return NOTIFY_DONE;
 811 }
 812
 813 /*
 814  * Initialize the control inode/socket with a control endpoint data
 815  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 816  */
 817 static int sctp_ctl_sock_init(struct net *net)
 818 {
 819         int err;
 820         sa_family_t family = PF_INET;
 821
 822         if (sctp_get_pf_specific(PF_INET6))
 823                 family = PF_INET6;
 824
 825         err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
 826                                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
 827
 828         /* If IPv6 socket could not be created, try the IPv4 socket */
 829         if (err < 0 && family == PF_INET6)
 830                 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
 831                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 832                                            net);
 833
 834         if (err < 0) {
 835                 pr_err("Failed to create the SCTP control socket\n");
 836                 return err;
 837         }
 838         return 0;
 839 }
 840
 841 /* Register address family specific functions. */
 842 int sctp_register_af(struct sctp_af *af)
 843 {
 844         switch (af->sa_family) {
 845         case AF_INET:
 846                 if (sctp_af_v4_specific)
 847                         return 0;
 848                 sctp_af_v4_specific = af;
 849                 break;
 850         case AF_INET6:
 851                 if (sctp_af_v6_specific)
 852                         return 0;
 853                 sctp_af_v6_specific = af;
 854                 break;
 855         default:
 856                 return 0;
 857         }
 858
 859         INIT_LIST_HEAD(&af->list);
 860         list_add_tail(&af->list, &sctp_address_families);
 861         return 1;
 862 }
 863
 864 /* Get the table of functions for manipulating a particular address
 865  * family.
 866  */
 867 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 868 {
 869         switch (family) {
 870         case AF_INET:
 871                 return sctp_af_v4_specific;
 872         case AF_INET6:
 873                 return sctp_af_v6_specific;
 874         default:
 875                 return NULL;
 876         }
 877 }
 878
 879 /* Common code to initialize a AF_INET msg_name. */
 880 static void sctp_inet_msgname(char *msgname, int *addr_len)
 881 {
 882         struct sockaddr_in *sin;
 883
 884         sin = (struct sockaddr_in *)msgname;
 885         *addr_len = sizeof(struct sockaddr_in);
 886         sin->sin_family = AF_INET;
 887         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 888 }
 889
 890 /* Copy the primary address of the peer primary address as the msg_name. */
 891 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 892                                     int *addr_len)
 893 {
 894         struct sockaddr_in *sin, *sinfrom;
 895
 896         if (msgname) {
 897                 struct sctp_association *asoc;
 898
 899                 asoc = event->asoc;
 900                 sctp_inet_msgname(msgname, addr_len);
 901                 sin = (struct sockaddr_in *)msgname;
 902                 sinfrom = &asoc->peer.primary_addr.v4;
 903                 sin->sin_port = htons(asoc->peer.port);
 904                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 905         }
 906 }
 907
 908 /* Initialize and copy out a msgname from an inbound skb. */
 909 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 910 {
 911         if (msgname) {
 912                 struct sctphdr *sh = sctp_hdr(skb);
 913                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 914
 915                 sctp_inet_msgname(msgname, len);
 916                 sin->sin_port = sh->source;
 917                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 918         }
 919 }
 920
 921 /* Do we support this AF? */
 922 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 923 {
 924         /* PF_INET only supports AF_INET addresses. */
 925         return AF_INET == family;
 926 }
 927
 928 /* Address matching with wildcards allowed. */
 929 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 930                               const union sctp_addr *addr2,
 931                               struct sctp_sock *opt)
 932 {
 933         /* PF_INET only supports AF_INET addresses. */
 934         if (addr1->sa.sa_family != addr2->sa.sa_family)
 935                 return 0;
 936         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 937             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 938                 return 1;
 939         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 940                 return 1;
 941
 942         return 0;
 943 }
 944
 945 /* Verify that provided sockaddr looks bindable.  Common verification has
 946  * already been taken care of.
 947  */
 948 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 949 {
 950         return sctp_v4_available(addr, opt);
 951 }
 952
 953 /* Verify that sockaddr looks sendable.  Common verification has already
 954  * been taken care of.
 955  */
 956 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 957 {
 958         return 1;
 959 }
 960
 961 /* Fill in Supported Address Type information for INIT and INIT-ACK
 962  * chunks.  Returns number of addresses supported.
 963  */
 964 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 965                                      __be16 *types)
 966 {
 967         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 968         return 1;
 969 }
 970
 971 /* Wrapper routine that calls the ip transmit routine. */
 972 static inline int sctp_v4_xmit(struct sk_buff *skb,
 973                                struct sctp_transport *transport)
 974 {
 975         struct inet_sock *inet = inet_sk(skb->sk);
 976         __u8 dscp = inet->tos;
 977
 978         pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
 979                  skb->len, &transport->fl.u.ip4.saddr,
 980                  &transport->fl.u.ip4.daddr);
 981
 982         if (transport->dscp & SCTP_DSCP_SET_MASK)
 983                 dscp = transport->dscp & SCTP_DSCP_VAL_MASK;
 984
 985         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
 986                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 987
 988         SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
 989
 990         return __ip_queue_xmit(&inet->sk, skb, &transport->fl, dscp);
 991 }
 992
 993 static struct sctp_af sctp_af_inet;
 994
 995 static struct sctp_pf sctp_pf_inet = {
 996         .event_msgname = sctp_inet_event_msgname,
 997         .skb_msgname   = sctp_inet_skb_msgname,
 998         .af_supported  = sctp_inet_af_supported,
 999         .cmp_addr      = sctp_inet_cmp_addr,
1000         .bind_verify   = sctp_inet_bind_verify,
1001         .send_verify   = sctp_inet_send_verify,
1002         .supported_addrs = sctp_inet_supported_addrs,
1003         .create_accept_sk = sctp_v4_create_accept_sk,
1004         .addr_to_user  = sctp_v4_addr_to_user,
1005         .to_sk_saddr   = sctp_v4_to_sk_saddr,
1006         .to_sk_daddr   = sctp_v4_to_sk_daddr,
1007         .copy_ip_options = sctp_v4_copy_ip_options,
1008         .af            = &sctp_af_inet
1009 };
1010
1011 /* Notifier for inetaddr addition/deletion events.  */
1012 static struct notifier_block sctp_inetaddr_notifier = {
1013         .notifier_call = sctp_inetaddr_event,
1014 };
1015
1016 /* Socket operations.  */
1017 static const struct proto_ops inet_seqpacket_ops = {
1018         .family            = PF_INET,
1019         .owner             = THIS_MODULE,
1020         .release           = inet_release,      /* Needs to be wrapped... */
1021         .bind              = inet_bind,
1022         .connect           = sctp_inet_connect,
1023         .socketpair        = sock_no_socketpair,
1024         .accept            = inet_accept,
1025         .getname           = inet_getname,      /* Semantics are different.  */
1026         .poll              = sctp_poll,
1027         .ioctl             = inet_ioctl,
1028         .listen            = sctp_inet_listen,
1029         .shutdown          = inet_shutdown,     /* Looks harmless.  */
1030         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1031         .getsockopt        = sock_common_getsockopt,
1032         .sendmsg           = inet_sendmsg,
1033         .recvmsg           = inet_recvmsg,
1034         .mmap              = sock_no_mmap,
1035         .sendpage          = sock_no_sendpage,
1036 #ifdef CONFIG_COMPAT
1037         .compat_setsockopt = compat_sock_common_setsockopt,
1038         .compat_getsockopt = compat_sock_common_getsockopt,
1039 #endif
1040 };
1041
1042 /* Registration with AF_INET family.  */
1043 static struct inet_protosw sctp_seqpacket_protosw = {
1044         .type       = SOCK_SEQPACKET,
1045         .protocol   = IPPROTO_SCTP,
1046         .prot       = &sctp_prot,
1047         .ops        = &inet_seqpacket_ops,
1048         .flags      = SCTP_PROTOSW_FLAG
1049 };
1050 static struct inet_protosw sctp_stream_protosw = {
1051         .type       = SOCK_STREAM,
1052         .protocol   = IPPROTO_SCTP,
1053         .prot       = &sctp_prot,
1054         .ops        = &inet_seqpacket_ops,
1055         .flags      = SCTP_PROTOSW_FLAG
1056 };
1057
1058 /* Register with IP layer.  */
1059 static const struct net_protocol sctp_protocol = {
1060         .handler     = sctp_rcv,
1061         .err_handler = sctp_v4_err,
1062         .no_policy   = 1,
1063         .netns_ok    = 1,
1064         .icmp_strict_tag_validation = 1,
1065 };
1066
1067 /* IPv4 address related functions.  */
1068 static struct sctp_af sctp_af_inet = {
1069         .sa_family         = AF_INET,
1070         .sctp_xmit         = sctp_v4_xmit,
1071         .setsockopt        = ip_setsockopt,
1072         .getsockopt        = ip_getsockopt,
1073         .get_dst           = sctp_v4_get_dst,
1074         .get_saddr         = sctp_v4_get_saddr,
1075         .copy_addrlist     = sctp_v4_copy_addrlist,
1076         .from_skb          = sctp_v4_from_skb,
1077         .from_sk           = sctp_v4_from_sk,
1078         .from_addr_param   = sctp_v4_from_addr_param,
1079         .to_addr_param     = sctp_v4_to_addr_param,
1080         .cmp_addr          = sctp_v4_cmp_addr,
1081         .addr_valid        = sctp_v4_addr_valid,
1082         .inaddr_any        = sctp_v4_inaddr_any,
1083         .is_any            = sctp_v4_is_any,
1084         .available         = sctp_v4_available,
1085         .scope             = sctp_v4_scope,
1086         .skb_iif           = sctp_v4_skb_iif,
1087         .is_ce             = sctp_v4_is_ce,
1088         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1089         .ecn_capable       = sctp_v4_ecn_capable,
1090         .net_header_len    = sizeof(struct iphdr),
1091         .sockaddr_len      = sizeof(struct sockaddr_in),
1092         .ip_options_len    = sctp_v4_ip_options_len,
1093 #ifdef CONFIG_COMPAT
1094         .compat_setsockopt = compat_ip_setsockopt,
1095         .compat_getsockopt = compat_ip_getsockopt,
1096 #endif
1097 };
1098
1099 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1100 {
1101         switch (family) {
1102         case PF_INET:
1103                 return sctp_pf_inet_specific;
1104         case PF_INET6:
1105                 return sctp_pf_inet6_specific;
1106         default:
1107                 return NULL;
1108         }
1109 }
1110
1111 /* Register the PF specific function table.  */
1112 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1113 {
1114         switch (family) {
1115         case PF_INET:
1116                 if (sctp_pf_inet_specific)
1117                         return 0;
1118                 sctp_pf_inet_specific = pf;
1119                 break;
1120         case PF_INET6:
1121                 if (sctp_pf_inet6_specific)
1122                         return 0;
1123                 sctp_pf_inet6_specific = pf;
1124                 break;
1125         default:
1126                 return 0;
1127         }
1128         return 1;
1129 }
1130
1131 static inline int init_sctp_mibs(struct net *net)
1132 {
1133         net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1134         if (!net->sctp.sctp_statistics)
1135                 return -ENOMEM;
1136         return 0;
1137 }
1138
1139 static inline void cleanup_sctp_mibs(struct net *net)
1140 {
1141         free_percpu(net->sctp.sctp_statistics);
1142 }
1143
1144 static void sctp_v4_pf_init(void)
1145 {
1146         /* Initialize the SCTP specific PF functions. */
1147         sctp_register_pf(&sctp_pf_inet, PF_INET);
1148         sctp_register_af(&sctp_af_inet);
1149 }
1150
1151 static void sctp_v4_pf_exit(void)
1152 {
1153         list_del(&sctp_af_inet.list);
1154 }
1155
1156 static int sctp_v4_protosw_init(void)
1157 {
1158         int rc;
1159
1160         rc = proto_register(&sctp_prot, 1);
1161         if (rc)
1162                 return rc;
1163
1164         /* Register SCTP(UDP and TCP style) with socket layer.  */
1165         inet_register_protosw(&sctp_seqpacket_protosw);
1166         inet_register_protosw(&sctp_stream_protosw);
1167
1168         return 0;
1169 }
1170
1171 static void sctp_v4_protosw_exit(void)
1172 {
1173         inet_unregister_protosw(&sctp_stream_protosw);
1174         inet_unregister_protosw(&sctp_seqpacket_protosw);
1175         proto_unregister(&sctp_prot);
1176 }
1177
1178 static int sctp_v4_add_protocol(void)
1179 {
1180         /* Register notifier for inet address additions/deletions. */
1181         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1182
1183         /* Register SCTP with inet layer.  */
1184         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1185                 return -EAGAIN;
1186
1187         return 0;
1188 }
1189
1190 static void sctp_v4_del_protocol(void)
1191 {
1192         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1193         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1194 }
1195
1196 static int __net_init sctp_defaults_init(struct net *net)
1197 {
1198         int status;
1199
1200         /*
1201          * 14. Suggested SCTP Protocol Parameter Values
1202          */
1203         /* The following protocol parameters are RECOMMENDED:  */
1204         /* RTO.Initial              - 3  seconds */
1205         net->sctp.rto_initial                   = SCTP_RTO_INITIAL;
1206         /* RTO.Min                  - 1  second */
1207         net->sctp.rto_min                       = SCTP_RTO_MIN;
1208         /* RTO.Max                 -  60 seconds */
1209         net->sctp.rto_max                       = SCTP_RTO_MAX;
1210         /* RTO.Alpha                - 1/8 */
1211         net->sctp.rto_alpha                     = SCTP_RTO_ALPHA;
1212         /* RTO.Beta                 - 1/4 */
1213         net->sctp.rto_beta                      = SCTP_RTO_BETA;
1214
1215         /* Valid.Cookie.Life        - 60  seconds */
1216         net->sctp.valid_cookie_life             = SCTP_DEFAULT_COOKIE_LIFE;
1217
1218         /* Whether Cookie Preservative is enabled(1) or not(0) */
1219         net->sctp.cookie_preserve_enable        = 1;
1220
1221         /* Default sctp sockets to use md5 as their hmac alg */
1222 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1223         net->sctp.sctp_hmac_alg                 = "md5";
1224 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1225         net->sctp.sctp_hmac_alg                 = "sha1";
1226 #else
1227         net->sctp.sctp_hmac_alg                 = NULL;
1228 #endif
1229
1230         /* Max.Burst                - 4 */
1231         net->sctp.max_burst                     = SCTP_DEFAULT_MAX_BURST;
1232
1233         /* Enable pf state by default */
1234         net->sctp.pf_enable = 1;
1235
1236         /* Association.Max.Retrans  - 10 attempts
1237          * Path.Max.Retrans         - 5  attempts (per destination address)
1238          * Max.Init.Retransmits     - 8  attempts
1239          */
1240         net->sctp.max_retrans_association       = 10;
1241         net->sctp.max_retrans_path              = 5;
1242         net->sctp.max_retrans_init              = 8;
1243
1244         /* Sendbuffer growth        - do per-socket accounting */
1245         net->sctp.sndbuf_policy                 = 0;
1246
1247         /* Rcvbuffer growth         - do per-socket accounting */
1248         net->sctp.rcvbuf_policy                 = 0;
1249
1250         /* HB.interval              - 30 seconds */
1251         net->sctp.hb_interval                   = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1252
1253         /* delayed SACK timeout */
1254         net->sctp.sack_timeout                  = SCTP_DEFAULT_TIMEOUT_SACK;
1255
1256         /* Disable ADDIP by default. */
1257         net->sctp.addip_enable = 0;
1258         net->sctp.addip_noauth = 0;
1259         net->sctp.default_auto_asconf = 0;
1260
1261         /* Enable PR-SCTP by default. */
1262         net->sctp.prsctp_enable = 1;
1263
1264         /* Disable RECONF by default. */
1265         net->sctp.reconf_enable = 0;
1266
1267         /* Disable AUTH by default. */
1268         net->sctp.auth_enable = 0;
1269
1270         /* Set SCOPE policy to enabled */
1271         net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1272
1273         /* Set the default rwnd update threshold */
1274         net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1275
1276         /* Initialize maximum autoclose timeout. */
1277         net->sctp.max_autoclose         = INT_MAX / HZ;
1278
1279         status = sctp_sysctl_net_register(net);
1280         if (status)
1281                 goto err_sysctl_register;
1282
1283         /* Allocate and initialise sctp mibs.  */
1284         status = init_sctp_mibs(net);
1285         if (status)
1286                 goto err_init_mibs;
1287
1288 #ifdef CONFIG_PROC_FS
1289         /* Initialize proc fs directory.  */
1290         status = sctp_proc_init(net);
1291         if (status)
1292                 goto err_init_proc;
1293 #endif
1294
1295         sctp_dbg_objcnt_init(net);
1296
1297         /* Initialize the local address list. */
1298         INIT_LIST_HEAD(&net->sctp.local_addr_list);
1299         spin_lock_init(&net->sctp.local_addr_lock);
1300         sctp_get_local_addr_list(net);
1301
1302         /* Initialize the address event list */
1303         INIT_LIST_HEAD(&net->sctp.addr_waitq);
1304         INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1305         spin_lock_init(&net->sctp.addr_wq_lock);
1306         net->sctp.addr_wq_timer.expires = 0;
1307         timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1308
1309         return 0;
1310
1311 #ifdef CONFIG_PROC_FS
1312 err_init_proc:
1313         cleanup_sctp_mibs(net);
1314 #endif
1315 err_init_mibs:
1316         sctp_sysctl_net_unregister(net);
1317 err_sysctl_register:
1318         return status;
1319 }
1320
1321 static void __net_exit sctp_defaults_exit(struct net *net)
1322 {
1323         /* Free the local address list */
1324         sctp_free_addr_wq(net);
1325         sctp_free_local_addr_list(net);
1326
1327 #ifdef CONFIG_PROC_FS
1328         remove_proc_subtree("sctp", net->proc_net);
1329         net->sctp.proc_net_sctp = NULL;
1330 #endif
1331         cleanup_sctp_mibs(net);
1332         sctp_sysctl_net_unregister(net);
1333 }
1334
1335 static struct pernet_operations sctp_defaults_ops = {
1336         .init = sctp_defaults_init,
1337         .exit = sctp_defaults_exit,
1338 };
1339
1340 static int __net_init sctp_ctrlsock_init(struct net *net)
1341 {
1342         int status;
1343
1344         /* Initialize the control inode/socket for handling OOTB packets.  */
1345         status = sctp_ctl_sock_init(net);
1346         if (status)
1347                 pr_err("Failed to initialize the SCTP control sock\n");
1348
1349         return status;
1350 }
1351
1352 static void __net_init sctp_ctrlsock_exit(struct net *net)
1353 {
1354         /* Free the control endpoint.  */
1355         inet_ctl_sock_destroy(net->sctp.ctl_sock);
1356 }
1357
1358 static struct pernet_operations sctp_ctrlsock_ops = {
1359         .init = sctp_ctrlsock_init,
1360         .exit = sctp_ctrlsock_exit,
1361 };
1362
1363 /* Initialize the universe into something sensible.  */
1364 static __init int sctp_init(void)
1365 {
1366         int i;
1367         int status = -EINVAL;
1368         unsigned long goal;
1369         unsigned long limit;
1370         unsigned long nr_pages = totalram_pages();
1371         int max_share;
1372         int order;
1373         int num_entries;
1374         int max_entry_order;
1375
1376         sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1377
1378         /* Allocate bind_bucket and chunk caches. */
1379         status = -ENOBUFS;
1380         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1381                                                sizeof(struct sctp_bind_bucket),
1382                                                0, SLAB_HWCACHE_ALIGN,
1383                                                NULL);
1384         if (!sctp_bucket_cachep)
1385                 goto out;
1386
1387         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1388                                                sizeof(struct sctp_chunk),
1389                                                0, SLAB_HWCACHE_ALIGN,
1390                                                NULL);
1391         if (!sctp_chunk_cachep)
1392                 goto err_chunk_cachep;
1393
1394         status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1395         if (status)
1396                 goto err_percpu_counter_init;
1397
1398         /* Implementation specific variables. */
1399
1400         /* Initialize default stream count setup information. */
1401         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1402         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1403
1404         /* Initialize handle used for association ids. */
1405         idr_init(&sctp_assocs_id);
1406
1407         limit = nr_free_buffer_pages() / 8;
1408         limit = max(limit, 128UL);
1409         sysctl_sctp_mem[0] = limit / 4 * 3;
1410         sysctl_sctp_mem[1] = limit;
1411         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1412
1413         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1414         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1415         max_share = min(4UL*1024*1024, limit);
1416
1417         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1418         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1419         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1420
1421         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1422         sysctl_sctp_wmem[1] = 16*1024;
1423         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1424
1425         /* Size and allocate the association hash table.
1426          * The methodology is similar to that of the tcp hash tables.
1427          * Though not identical.  Start by getting a goal size
1428          */
1429         if (nr_pages >= (128 * 1024))
1430                 goal = nr_pages >> (22 - PAGE_SHIFT);
1431         else
1432                 goal = nr_pages >> (24 - PAGE_SHIFT);
1433
1434         /* Then compute the page order for said goal */
1435         order = get_order(goal);
1436
1437         /* Now compute the required page order for the maximum sized table we
1438          * want to create
1439          */
1440         max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1441                                     sizeof(struct sctp_bind_hashbucket));
1442
1443         /* Limit the page order by that maximum hash table size */
1444         order = min(order, max_entry_order);
1445
1446         /* Allocate and initialize the endpoint hash table.  */
1447         sctp_ep_hashsize = 64;
1448         sctp_ep_hashtable =
1449                 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1450         if (!sctp_ep_hashtable) {
1451                 pr_err("Failed endpoint_hash alloc\n");
1452                 status = -ENOMEM;
1453                 goto err_ehash_alloc;
1454         }
1455         for (i = 0; i < sctp_ep_hashsize; i++) {
1456                 rwlock_init(&sctp_ep_hashtable[i].lock);
1457                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1458         }
1459
1460         /* Allocate and initialize the SCTP port hash table.
1461          * Note that order is initalized to start at the max sized
1462          * table we want to support.  If we can't get that many pages
1463          * reduce the order and try again
1464          */
1465         do {
1466                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1467                         __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1468         } while (!sctp_port_hashtable && --order > 0);
1469
1470         if (!sctp_port_hashtable) {
1471                 pr_err("Failed bind hash alloc\n");
1472                 status = -ENOMEM;
1473                 goto err_bhash_alloc;
1474         }
1475
1476         /* Now compute the number of entries that will fit in the
1477          * port hash space we allocated
1478          */
1479         num_entries = (1UL << order) * PAGE_SIZE /
1480                       sizeof(struct sctp_bind_hashbucket);
1481
1482         /* And finish by rounding it down to the nearest power of two
1483          * this wastes some memory of course, but its needed because
1484          * the hash function operates based on the assumption that
1485          * that the number of entries is a power of two
1486          */
1487         sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1488
1489         for (i = 0; i < sctp_port_hashsize; i++) {
1490                 spin_lock_init(&sctp_port_hashtable[i].lock);
1491                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1492         }
1493
1494         status = sctp_transport_hashtable_init();
1495         if (status)
1496                 goto err_thash_alloc;
1497
1498         pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1499                 num_entries);
1500
1501         sctp_sysctl_register();
1502
1503         INIT_LIST_HEAD(&sctp_address_families);
1504         sctp_v4_pf_init();
1505         sctp_v6_pf_init();
1506         sctp_sched_ops_init();
1507
1508         status = register_pernet_subsys(&sctp_defaults_ops);
1509         if (status)
1510                 goto err_register_defaults;
1511
1512         status = sctp_v4_protosw_init();
1513         if (status)
1514                 goto err_protosw_init;
1515
1516         status = sctp_v6_protosw_init();
1517         if (status)
1518                 goto err_v6_protosw_init;
1519
1520         status = register_pernet_subsys(&sctp_ctrlsock_ops);
1521         if (status)
1522                 goto err_register_ctrlsock;
1523
1524         status = sctp_v4_add_protocol();
1525         if (status)
1526                 goto err_add_protocol;
1527
1528         /* Register SCTP with inet6 layer.  */
1529         status = sctp_v6_add_protocol();
1530         if (status)
1531                 goto err_v6_add_protocol;
1532
1533         if (sctp_offload_init() < 0)
1534                 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1535
1536 out:
1537         return status;
1538 err_v6_add_protocol:
1539         sctp_v4_del_protocol();
1540 err_add_protocol:
1541         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1542 err_register_ctrlsock:
1543         sctp_v6_protosw_exit();
1544 err_v6_protosw_init:
1545         sctp_v4_protosw_exit();
1546 err_protosw_init:
1547         unregister_pernet_subsys(&sctp_defaults_ops);
1548 err_register_defaults:
1549         sctp_v4_pf_exit();
1550         sctp_v6_pf_exit();
1551         sctp_sysctl_unregister();
1552         free_pages((unsigned long)sctp_port_hashtable,
1553                    get_order(sctp_port_hashsize *
1554                              sizeof(struct sctp_bind_hashbucket)));
1555 err_bhash_alloc:
1556         sctp_transport_hashtable_destroy();
1557 err_thash_alloc:
1558         kfree(sctp_ep_hashtable);
1559 err_ehash_alloc:
1560         percpu_counter_destroy(&sctp_sockets_allocated);
1561 err_percpu_counter_init:
1562         kmem_cache_destroy(sctp_chunk_cachep);
1563 err_chunk_cachep:
1564         kmem_cache_destroy(sctp_bucket_cachep);
1565         goto out;
1566 }
1567
1568 /* Exit handler for the SCTP protocol.  */
1569 static __exit void sctp_exit(void)
1570 {
1571         /* BUG.  This should probably do something useful like clean
1572          * up all the remaining associations and all that memory.
1573          */
1574
1575         /* Unregister with inet6/inet layers. */
1576         sctp_v6_del_protocol();
1577         sctp_v4_del_protocol();
1578
1579         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1580
1581         /* Free protosw registrations */
1582         sctp_v6_protosw_exit();
1583         sctp_v4_protosw_exit();
1584
1585         unregister_pernet_subsys(&sctp_defaults_ops);
1586
1587         /* Unregister with socket layer. */
1588         sctp_v6_pf_exit();
1589         sctp_v4_pf_exit();
1590
1591         sctp_sysctl_unregister();
1592
1593         free_pages((unsigned long)sctp_port_hashtable,
1594                    get_order(sctp_port_hashsize *
1595                              sizeof(struct sctp_bind_hashbucket)));
1596         kfree(sctp_ep_hashtable);
1597         sctp_transport_hashtable_destroy();
1598
1599         percpu_counter_destroy(&sctp_sockets_allocated);
1600
1601         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1602
1603         kmem_cache_destroy(sctp_chunk_cachep);
1604         kmem_cache_destroy(sctp_bucket_cachep);
1605 }
1606
1607 module_init(sctp_init);
1608 module_exit(sctp_exit);
1609
1610 /*
1611  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1612  */
1613 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1614 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1615 MODULE_AUTHOR("Linux Kernel SCTP developers <[email protected]>");
1616 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1617 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1618 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1619 MODULE_LICENSE("GPL");