net/sctp/protocol.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /* SCTP kernel implementation
   3  * (C) Copyright IBM Corp. 2001, 2004
   4  * Copyright (c) 1999-2000 Cisco, Inc.
   5  * Copyright (c) 1999-2001 Motorola, Inc.
   6  * Copyright (c) 2001 Intel Corp.
   7  * Copyright (c) 2001 Nokia, Inc.
   8  * Copyright (c) 2001 La Monte H.P. Yarroll
   9  *
  10  * This file is part of the SCTP kernel implementation
  11  *
  12  * Initialization/cleanup for SCTP protocol support.
  13  *
  14  * Please send any bug reports or fixes you make to the
  15  * email address(es):
  16  *    lksctp developers <linux-sctp@vger.kernel.org>
  17  *
  18  * Written or modified by:
  19  *    La Monte H.P. Yarroll <piggy@acm.org>
  20  *    Karl Knutson <karl@athena.chicago.il.us>
  21  *    Jon Grimm <jgrimm@us.ibm.com>
  22  *    Sridhar Samudrala <sri@us.ibm.com>
  23  *    Daisy Chang <daisyc@us.ibm.com>
  24  *    Ardelle Fan <ardelle.fan@intel.com>
  25  */
  26
  27 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  28
  29 #include <linux/module.h>
  30 #include <linux/init.h>
  31 #include <linux/netdevice.h>
  32 #include <linux/inetdevice.h>
  33 #include <linux/seq_file.h>
  34 #include <linux/memblock.h>
  35 #include <linux/highmem.h>
  36 #include <linux/swap.h>
  37 #include <linux/slab.h>
  38 #include <net/net_namespace.h>
  39 #include <net/protocol.h>
  40 #include <net/ip.h>
  41 #include <net/ipv6.h>
  42 #include <net/route.h>
  43 #include <net/sctp/sctp.h>
  44 #include <net/addrconf.h>
  45 #include <net/inet_common.h>
  46 #include <net/inet_ecn.h>
  47
  48 #define MAX_SCTP_PORT_HASH_ENTRIES (64 * 1024)
  49
  50 /* Global data structures. */
  51 struct sctp_globals sctp_globals __read_mostly;
  52
  53 struct idr sctp_assocs_id;
  54 DEFINE_SPINLOCK(sctp_assocs_id_lock);
  55
  56 static struct sctp_pf *sctp_pf_inet6_specific;
  57 static struct sctp_pf *sctp_pf_inet_specific;
  58 static struct sctp_af *sctp_af_v4_specific;
  59 static struct sctp_af *sctp_af_v6_specific;
  60
  61 struct kmem_cache *sctp_chunk_cachep __read_mostly;
  62 struct kmem_cache *sctp_bucket_cachep __read_mostly;
  63
  64 long sysctl_sctp_mem[3];
  65 int sysctl_sctp_rmem[3];
  66 int sysctl_sctp_wmem[3];
  67
  68 /* Private helper to extract ipv4 address and stash them in
  69  * the protocol structure.
  70  */
  71 static void sctp_v4_copy_addrlist(struct list_head *addrlist,
  72                                   struct net_device *dev)
  73 {
  74         struct in_device *in_dev;
  75         struct in_ifaddr *ifa;
  76         struct sctp_sockaddr_entry *addr;
  77
  78         rcu_read_lock();
  79         if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
  80                 rcu_read_unlock();
  81                 return;
  82         }
  83
  84         in_dev_for_each_ifa_rcu(ifa, in_dev) {
  85                 /* Add the address to the local list.  */
  86                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
  87                 if (addr) {
  88                         addr->a.v4.sin_family = AF_INET;
  89                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
  90                         addr->valid = 1;
  91                         INIT_LIST_HEAD(&addr->list);
  92                         list_add_tail(&addr->list, addrlist);
  93                 }
  94         }
  95
  96         rcu_read_unlock();
  97 }
  98
  99 /* Extract our IP addresses from the system and stash them in the
 100  * protocol structure.
 101  */
 102 static void sctp_get_local_addr_list(struct net *net)
 103 {
 104         struct net_device *dev;
 105         struct list_head *pos;
 106         struct sctp_af *af;
 107
 108         rcu_read_lock();
 109         for_each_netdev_rcu(net, dev) {
 110                 list_for_each(pos, &sctp_address_families) {
 111                         af = list_entry(pos, struct sctp_af, list);
 112                         af->copy_addrlist(&net->sctp.local_addr_list, dev);
 113                 }
 114         }
 115         rcu_read_unlock();
 116 }
 117
 118 /* Free the existing local addresses.  */
 119 static void sctp_free_local_addr_list(struct net *net)
 120 {
 121         struct sctp_sockaddr_entry *addr;
 122         struct list_head *pos, *temp;
 123
 124         list_for_each_safe(pos, temp, &net->sctp.local_addr_list) {
 125                 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
 126                 list_del(pos);
 127                 kfree(addr);
 128         }
 129 }
 130
 131 /* Copy the local addresses which are valid for 'scope' into 'bp'.  */
 132 int sctp_copy_local_addr_list(struct net *net, struct sctp_bind_addr *bp,
 133                               enum sctp_scope scope, gfp_t gfp, int copy_flags)
 134 {
 135         struct sctp_sockaddr_entry *addr;
 136         union sctp_addr laddr;
 137         int error = 0;
 138
 139         rcu_read_lock();
 140         list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
 141                 if (!addr->valid)
 142                         continue;
 143                 if (!sctp_in_scope(net, &addr->a, scope))
 144                         continue;
 145
 146                 /* Now that the address is in scope, check to see if
 147                  * the address type is really supported by the local
 148                  * sock as well as the remote peer.
 149                  */
 150                 if (addr->a.sa.sa_family == AF_INET &&
 151                     (!(copy_flags & SCTP_ADDR4_ALLOWED) ||
 152                      !(copy_flags & SCTP_ADDR4_PEERSUPP)))
 153                         continue;
 154                 if (addr->a.sa.sa_family == AF_INET6 &&
 155                     (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
 156                      !(copy_flags & SCTP_ADDR6_PEERSUPP)))
 157                         continue;
 158
 159                 laddr = addr->a;
 160                 /* also works for setting ipv6 address port */
 161                 laddr.v4.sin_port = htons(bp->port);
 162                 if (sctp_bind_addr_state(bp, &laddr) != -1)
 163                         continue;
 164
 165                 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
 166                                            SCTP_ADDR_SRC, GFP_ATOMIC);
 167                 if (error)
 168                         break;
 169         }
 170
 171         rcu_read_unlock();
 172         return error;
 173 }
 174
 175 /* Copy over any ip options */
 176 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
 177 {
 178         struct inet_sock *newinet, *inet = inet_sk(sk);
 179         struct ip_options_rcu *inet_opt, *newopt = NULL;
 180
 181         newinet = inet_sk(newsk);
 182
 183         rcu_read_lock();
 184         inet_opt = rcu_dereference(inet->inet_opt);
 185         if (inet_opt) {
 186                 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
 187                                       inet_opt->opt.optlen, GFP_ATOMIC);
 188                 if (newopt)
 189                         memcpy(newopt, inet_opt, sizeof(*inet_opt) +
 190                                inet_opt->opt.optlen);
 191                 else
 192                         pr_err("%s: Failed to copy ip options\n", __func__);
 193         }
 194         RCU_INIT_POINTER(newinet->inet_opt, newopt);
 195         rcu_read_unlock();
 196 }
 197
 198 /* Account for the IP options */
 199 static int sctp_v4_ip_options_len(struct sock *sk)
 200 {
 201         struct inet_sock *inet = inet_sk(sk);
 202         struct ip_options_rcu *inet_opt;
 203         int len = 0;
 204
 205         rcu_read_lock();
 206         inet_opt = rcu_dereference(inet->inet_opt);
 207         if (inet_opt)
 208                 len = inet_opt->opt.optlen;
 209
 210         rcu_read_unlock();
 211         return len;
 212 }
 213
 214 /* Initialize a sctp_addr from in incoming skb.  */
 215 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
 216                              int is_saddr)
 217 {
 218         /* Always called on head skb, so this is safe */
 219         struct sctphdr *sh = sctp_hdr(skb);
 220         struct sockaddr_in *sa = &addr->v4;
 221
 222         addr->v4.sin_family = AF_INET;
 223
 224         if (is_saddr) {
 225                 sa->sin_port = sh->source;
 226                 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
 227         } else {
 228                 sa->sin_port = sh->dest;
 229                 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
 230         }
 231         memset(sa->sin_zero, 0, sizeof(sa->sin_zero));
 232 }
 233
 234 /* Initialize an sctp_addr from a socket. */
 235 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
 236 {
 237         addr->v4.sin_family = AF_INET;
 238         addr->v4.sin_port = 0;
 239         addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
 240         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 241 }
 242
 243 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
 244 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
 245 {
 246         inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
 247 }
 248
 249 /* Initialize sk->sk_daddr from sctp_addr. */
 250 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
 251 {
 252         inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
 253 }
 254
 255 /* Initialize a sctp_addr from an address parameter. */
 256 static void sctp_v4_from_addr_param(union sctp_addr *addr,
 257                                     union sctp_addr_param *param,
 258                                     __be16 port, int iif)
 259 {
 260         addr->v4.sin_family = AF_INET;
 261         addr->v4.sin_port = port;
 262         addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
 263         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 264 }
 265
 266 /* Initialize an address parameter from a sctp_addr and return the length
 267  * of the address parameter.
 268  */
 269 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
 270                                  union sctp_addr_param *param)
 271 {
 272         int length = sizeof(struct sctp_ipv4addr_param);
 273
 274         param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
 275         param->v4.param_hdr.length = htons(length);
 276         param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
 277
 278         return length;
 279 }
 280
 281 /* Initialize a sctp_addr from a dst_entry. */
 282 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
 283                               __be16 port)
 284 {
 285         saddr->v4.sin_family = AF_INET;
 286         saddr->v4.sin_port = port;
 287         saddr->v4.sin_addr.s_addr = fl4->saddr;
 288         memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero));
 289 }
 290
 291 /* Compare two addresses exactly. */
 292 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
 293                             const union sctp_addr *addr2)
 294 {
 295         if (addr1->sa.sa_family != addr2->sa.sa_family)
 296                 return 0;
 297         if (addr1->v4.sin_port != addr2->v4.sin_port)
 298                 return 0;
 299         if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
 300                 return 0;
 301
 302         return 1;
 303 }
 304
 305 /* Initialize addr struct to INADDR_ANY. */
 306 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
 307 {
 308         addr->v4.sin_family = AF_INET;
 309         addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
 310         addr->v4.sin_port = port;
 311         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 312 }
 313
 314 /* Is this a wildcard address? */
 315 static int sctp_v4_is_any(const union sctp_addr *addr)
 316 {
 317         return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
 318 }
 319
 320 /* This function checks if the address is a valid address to be used for
 321  * SCTP binding.
 322  *
 323  * Output:
 324  * Return 0 - If the address is a non-unicast or an illegal address.
 325  * Return 1 - If the address is a unicast.
 326  */
 327 static int sctp_v4_addr_valid(union sctp_addr *addr,
 328                               struct sctp_sock *sp,
 329                               const struct sk_buff *skb)
 330 {
 331         /* IPv4 addresses not allowed */
 332         if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
 333                 return 0;
 334
 335         /* Is this a non-unicast address or a unusable SCTP address? */
 336         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
 337                 return 0;
 338
 339         /* Is this a broadcast address? */
 340         if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
 341                 return 0;
 342
 343         return 1;
 344 }
 345
 346 /* Should this be available for binding?   */
 347 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
 348 {
 349         struct net *net = sock_net(&sp->inet.sk);
 350         int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
 351
 352
 353         if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
 354            ret != RTN_LOCAL &&
 355            !sp->inet.freebind &&
 356            !net->ipv4.sysctl_ip_nonlocal_bind)
 357                 return 0;
 358
 359         if (ipv6_only_sock(sctp_opt2sk(sp)))
 360                 return 0;
 361
 362         return 1;
 363 }
 364
 365 /* Checking the loopback, private and other address scopes as defined in
 366  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
 367  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
 368  *
 369  * Level 0 - unusable SCTP addresses
 370  * Level 1 - loopback address
 371  * Level 2 - link-local addresses
 372  * Level 3 - private addresses.
 373  * Level 4 - global addresses
 374  * For INIT and INIT-ACK address list, let L be the level of
 375  * requested destination address, sender and receiver
 376  * SHOULD include all of its addresses with level greater
 377  * than or equal to L.
 378  *
 379  * IPv4 scoping can be controlled through sysctl option
 380  * net.sctp.addr_scope_policy
 381  */
 382 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
 383 {
 384         enum sctp_scope retval;
 385
 386         /* Check for unusable SCTP addresses. */
 387         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
 388                 retval =  SCTP_SCOPE_UNUSABLE;
 389         } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
 390                 retval = SCTP_SCOPE_LOOPBACK;
 391         } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
 392                 retval = SCTP_SCOPE_LINK;
 393         } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
 394                    ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
 395                    ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
 396                 retval = SCTP_SCOPE_PRIVATE;
 397         } else {
 398                 retval = SCTP_SCOPE_GLOBAL;
 399         }
 400
 401         return retval;
 402 }
 403
 404 /* Returns a valid dst cache entry for the given source and destination ip
 405  * addresses. If an association is passed, trys to get a dst entry with a
 406  * source address that matches an address in the bind address list.
 407  */
 408 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
 409                                 struct flowi *fl, struct sock *sk)
 410 {
 411         struct sctp_association *asoc = t->asoc;
 412         struct rtable *rt;
 413         struct flowi _fl;
 414         struct flowi4 *fl4 = &_fl.u.ip4;
 415         struct sctp_bind_addr *bp;
 416         struct sctp_sockaddr_entry *laddr;
 417         struct dst_entry *dst = NULL;
 418         union sctp_addr *daddr = &t->ipaddr;
 419         union sctp_addr dst_saddr;
 420         __u8 tos = inet_sk(sk)->tos;
 421
 422         if (t->dscp & SCTP_DSCP_SET_MASK)
 423                 tos = t->dscp & SCTP_DSCP_VAL_MASK;
 424         memset(&_fl, 0x0, sizeof(_fl));
 425         fl4->daddr  = daddr->v4.sin_addr.s_addr;
 426         fl4->fl4_dport = daddr->v4.sin_port;
 427         fl4->flowi4_proto = IPPROTO_SCTP;
 428         if (asoc) {
 429                 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
 430                 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
 431                 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
 432         }
 433         if (saddr) {
 434                 fl4->saddr = saddr->v4.sin_addr.s_addr;
 435                 if (!fl4->fl4_sport)
 436                         fl4->fl4_sport = saddr->v4.sin_port;
 437         }
 438
 439         pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
 440                  &fl4->saddr);
 441
 442         rt = ip_route_output_key(sock_net(sk), fl4);
 443         if (!IS_ERR(rt)) {
 444                 dst = &rt->dst;
 445                 t->dst = dst;
 446                 memcpy(fl, &_fl, sizeof(_fl));
 447         }
 448
 449         /* If there is no association or if a source address is passed, no
 450          * more validation is required.
 451          */
 452         if (!asoc || saddr)
 453                 goto out;
 454
 455         bp = &asoc->base.bind_addr;
 456
 457         if (dst) {
 458                 /* Walk through the bind address list and look for a bind
 459                  * address that matches the source address of the returned dst.
 460                  */
 461                 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
 462                 rcu_read_lock();
 463                 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 464                         if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
 465                             (laddr->state != SCTP_ADDR_SRC &&
 466                             !asoc->src_out_of_asoc_ok))
 467                                 continue;
 468                         if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
 469                                 goto out_unlock;
 470                 }
 471                 rcu_read_unlock();
 472
 473                 /* None of the bound addresses match the source address of the
 474                  * dst. So release it.
 475                  */
 476                 dst_release(dst);
 477                 dst = NULL;
 478         }
 479
 480         /* Walk through the bind address list and try to get a dst that
 481          * matches a bind address as the source address.
 482          */
 483         rcu_read_lock();
 484         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 485                 struct net_device *odev;
 486
 487                 if (!laddr->valid)
 488                         continue;
 489                 if (laddr->state != SCTP_ADDR_SRC ||
 490                     AF_INET != laddr->a.sa.sa_family)
 491                         continue;
 492
 493                 fl4->fl4_sport = laddr->a.v4.sin_port;
 494                 flowi4_update_output(fl4,
 495                                      asoc->base.sk->sk_bound_dev_if,
 496                                      RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
 497                                      daddr->v4.sin_addr.s_addr,
 498                                      laddr->a.v4.sin_addr.s_addr);
 499
 500                 rt = ip_route_output_key(sock_net(sk), fl4);
 501                 if (IS_ERR(rt))
 502                         continue;
 503
 504                 /* Ensure the src address belongs to the output
 505                  * interface.
 506                  */
 507                 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
 508                                      false);
 509                 if (!odev || odev->ifindex != fl4->flowi4_oif) {
 510                         if (!dst) {
 511                                 dst = &rt->dst;
 512                                 t->dst = dst;
 513                                 memcpy(fl, &_fl, sizeof(_fl));
 514                         } else {
 515                                 dst_release(&rt->dst);
 516                         }
 517                         continue;
 518                 }
 519
 520                 dst_release(dst);
 521                 dst = &rt->dst;
 522                 t->dst = dst;
 523                 memcpy(fl, &_fl, sizeof(_fl));
 524                 break;
 525         }
 526
 527 out_unlock:
 528         rcu_read_unlock();
 529 out:
 530         if (dst) {
 531                 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
 532                          &fl->u.ip4.daddr, &fl->u.ip4.saddr);
 533         } else {
 534                 t->dst = NULL;
 535                 pr_debug("no route\n");
 536         }
 537 }
 538
 539 /* For v4, the source address is cached in the route entry(dst). So no need
 540  * to cache it separately and hence this is an empty routine.
 541  */
 542 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 543                               struct sctp_transport *t,
 544                               struct flowi *fl)
 545 {
 546         union sctp_addr *saddr = &t->saddr;
 547         struct rtable *rt = (struct rtable *)t->dst;
 548
 549         if (rt) {
 550                 saddr->v4.sin_family = AF_INET;
 551                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 552         }
 553 }
 554
 555 /* What interface did this skb arrive on? */
 556 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 557 {
 558         return inet_iif(skb);
 559 }
 560
 561 /* Was this packet marked by Explicit Congestion Notification? */
 562 static int sctp_v4_is_ce(const struct sk_buff *skb)
 563 {
 564         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 565 }
 566
 567 /* Create and initialize a new sk for the socket returned by accept(). */
 568 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 569                                              struct sctp_association *asoc,
 570                                              bool kern)
 571 {
 572         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 573                         sk->sk_prot, kern);
 574         struct inet_sock *newinet;
 575
 576         if (!newsk)
 577                 goto out;
 578
 579         sock_init_data(NULL, newsk);
 580
 581         sctp_copy_sock(newsk, sk, asoc);
 582         sock_reset_flag(newsk, SOCK_ZAPPED);
 583
 584         sctp_v4_copy_ip_options(sk, newsk);
 585
 586         newinet = inet_sk(newsk);
 587
 588         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 589
 590         sk_refcnt_debug_inc(newsk);
 591
 592         if (newsk->sk_prot->init(newsk)) {
 593                 sk_common_release(newsk);
 594                 newsk = NULL;
 595         }
 596
 597 out:
 598         return newsk;
 599 }
 600
 601 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
 602 {
 603         /* No address mapping for V4 sockets */
 604         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 605         return sizeof(struct sockaddr_in);
 606 }
 607
 608 /* Dump the v4 addr to the seq file. */
 609 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 610 {
 611         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 612 }
 613
 614 static void sctp_v4_ecn_capable(struct sock *sk)
 615 {
 616         INET_ECN_xmit(sk);
 617 }
 618
 619 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
 620 {
 621         struct net *net = from_timer(net, t, sctp.addr_wq_timer);
 622         struct sctp_sockaddr_entry *addrw, *temp;
 623         struct sctp_sock *sp;
 624
 625         spin_lock_bh(&net->sctp.addr_wq_lock);
 626
 627         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 628                 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
 629                          "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
 630                          addrw->state, addrw);
 631
 632 #if IS_ENABLED(CONFIG_IPV6)
 633                 /* Now we send an ASCONF for each association */
 634                 /* Note. we currently don't handle link local IPv6 addressees */
 635                 if (addrw->a.sa.sa_family == AF_INET6) {
 636                         struct in6_addr *in6;
 637
 638                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 639                             IPV6_ADDR_LINKLOCAL)
 640                                 goto free_next;
 641
 642                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 643                         if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
 644                             addrw->state == SCTP_ADDR_NEW) {
 645                                 unsigned long timeo_val;
 646
 647                                 pr_debug("%s: this is on DAD, trying %d sec "
 648                                          "later\n", __func__,
 649                                          SCTP_ADDRESS_TICK_DELAY);
 650
 651                                 timeo_val = jiffies;
 652                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 653                                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 654                                 break;
 655                         }
 656                 }
 657 #endif
 658                 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
 659                         struct sock *sk;
 660
 661                         sk = sctp_opt2sk(sp);
 662                         /* ignore bound-specific endpoints */
 663                         if (!sctp_is_ep_boundall(sk))
 664                                 continue;
 665                         bh_lock_sock(sk);
 666                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 667                                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
 668                         bh_unlock_sock(sk);
 669                 }
 670 #if IS_ENABLED(CONFIG_IPV6)
 671 free_next:
 672 #endif
 673                 list_del(&addrw->list);
 674                 kfree(addrw);
 675         }
 676         spin_unlock_bh(&net->sctp.addr_wq_lock);
 677 }
 678
 679 static void sctp_free_addr_wq(struct net *net)
 680 {
 681         struct sctp_sockaddr_entry *addrw;
 682         struct sctp_sockaddr_entry *temp;
 683
 684         spin_lock_bh(&net->sctp.addr_wq_lock);
 685         del_timer(&net->sctp.addr_wq_timer);
 686         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 687                 list_del(&addrw->list);
 688                 kfree(addrw);
 689         }
 690         spin_unlock_bh(&net->sctp.addr_wq_lock);
 691 }
 692
 693 /* lookup the entry for the same address in the addr_waitq
 694  * sctp_addr_wq MUST be locked
 695  */
 696 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
 697                                         struct sctp_sockaddr_entry *addr)
 698 {
 699         struct sctp_sockaddr_entry *addrw;
 700
 701         list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
 702                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 703                         continue;
 704                 if (addrw->a.sa.sa_family == AF_INET) {
 705                         if (addrw->a.v4.sin_addr.s_addr ==
 706                             addr->a.v4.sin_addr.s_addr)
 707                                 return addrw;
 708                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 709                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 710                             &addr->a.v6.sin6_addr))
 711                                 return addrw;
 712                 }
 713         }
 714         return NULL;
 715 }
 716
 717 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
 718 {
 719         struct sctp_sockaddr_entry *addrw;
 720         unsigned long timeo_val;
 721
 722         /* first, we check if an opposite message already exist in the queue.
 723          * If we found such message, it is removed.
 724          * This operation is a bit stupid, but the DHCP client attaches the
 725          * new address after a couple of addition and deletion of that address
 726          */
 727
 728         spin_lock_bh(&net->sctp.addr_wq_lock);
 729         /* Offsets existing events in addr_wq */
 730         addrw = sctp_addr_wq_lookup(net, addr);
 731         if (addrw) {
 732                 if (addrw->state != cmd) {
 733                         pr_debug("%s: offsets existing entry for %d, addr:%pISc "
 734                                  "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
 735                                  &net->sctp.addr_waitq);
 736
 737                         list_del(&addrw->list);
 738                         kfree(addrw);
 739                 }
 740                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 741                 return;
 742         }
 743
 744         /* OK, we have to add the new address to the wait queue */
 745         addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 746         if (addrw == NULL) {
 747                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 748                 return;
 749         }
 750         addrw->state = cmd;
 751         list_add_tail(&addrw->list, &net->sctp.addr_waitq);
 752
 753         pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
 754                  __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
 755
 756         if (!timer_pending(&net->sctp.addr_wq_timer)) {
 757                 timeo_val = jiffies;
 758                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 759                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 760         }
 761         spin_unlock_bh(&net->sctp.addr_wq_lock);
 762 }
 763
 764 /* Event handler for inet address addition/deletion events.
 765  * The sctp_local_addr_list needs to be protocted by a spin lock since
 766  * multiple notifiers (say IPv4 and IPv6) may be running at the same
 767  * time and thus corrupt the list.
 768  * The reader side is protected with RCU.
 769  */
 770 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
 771                                void *ptr)
 772 {
 773         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
 774         struct sctp_sockaddr_entry *addr = NULL;
 775         struct sctp_sockaddr_entry *temp;
 776         struct net *net = dev_net(ifa->ifa_dev->dev);
 777         int found = 0;
 778
 779         switch (ev) {
 780         case NETDEV_UP:
 781                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
 782                 if (addr) {
 783                         addr->a.v4.sin_family = AF_INET;
 784                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 785                         addr->valid = 1;
 786                         spin_lock_bh(&net->sctp.local_addr_lock);
 787                         list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
 788                         sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
 789                         spin_unlock_bh(&net->sctp.local_addr_lock);
 790                 }
 791                 break;
 792         case NETDEV_DOWN:
 793                 spin_lock_bh(&net->sctp.local_addr_lock);
 794                 list_for_each_entry_safe(addr, temp,
 795                                         &net->sctp.local_addr_list, list) {
 796                         if (addr->a.sa.sa_family == AF_INET &&
 797                                         addr->a.v4.sin_addr.s_addr ==
 798                                         ifa->ifa_local) {
 799                                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
 800                                 found = 1;
 801                                 addr->valid = 0;
 802                                 list_del_rcu(&addr->list);
 803                                 break;
 804                         }
 805                 }
 806                 spin_unlock_bh(&net->sctp.local_addr_lock);
 807                 if (found)
 808                         kfree_rcu(addr, rcu);
 809                 break;
 810         }
 811
 812         return NOTIFY_DONE;
 813 }
 814
 815 /*
 816  * Initialize the control inode/socket with a control endpoint data
 817  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 818  */
 819 static int sctp_ctl_sock_init(struct net *net)
 820 {
 821         int err;
 822         sa_family_t family = PF_INET;
 823
 824         if (sctp_get_pf_specific(PF_INET6))
 825                 family = PF_INET6;
 826
 827         err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
 828                                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
 829
 830         /* If IPv6 socket could not be created, try the IPv4 socket */
 831         if (err < 0 && family == PF_INET6)
 832                 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
 833                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 834                                            net);
 835
 836         if (err < 0) {
 837                 pr_err("Failed to create the SCTP control socket\n");
 838                 return err;
 839         }
 840         return 0;
 841 }
 842
 843 /* Register address family specific functions. */
 844 int sctp_register_af(struct sctp_af *af)
 845 {
 846         switch (af->sa_family) {
 847         case AF_INET:
 848                 if (sctp_af_v4_specific)
 849                         return 0;
 850                 sctp_af_v4_specific = af;
 851                 break;
 852         case AF_INET6:
 853                 if (sctp_af_v6_specific)
 854                         return 0;
 855                 sctp_af_v6_specific = af;
 856                 break;
 857         default:
 858                 return 0;
 859         }
 860
 861         INIT_LIST_HEAD(&af->list);
 862         list_add_tail(&af->list, &sctp_address_families);
 863         return 1;
 864 }
 865
 866 /* Get the table of functions for manipulating a particular address
 867  * family.
 868  */
 869 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 870 {
 871         switch (family) {
 872         case AF_INET:
 873                 return sctp_af_v4_specific;
 874         case AF_INET6:
 875                 return sctp_af_v6_specific;
 876         default:
 877                 return NULL;
 878         }
 879 }
 880
 881 /* Common code to initialize a AF_INET msg_name. */
 882 static void sctp_inet_msgname(char *msgname, int *addr_len)
 883 {
 884         struct sockaddr_in *sin;
 885
 886         sin = (struct sockaddr_in *)msgname;
 887         *addr_len = sizeof(struct sockaddr_in);
 888         sin->sin_family = AF_INET;
 889         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 890 }
 891
 892 /* Copy the primary address of the peer primary address as the msg_name. */
 893 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 894                                     int *addr_len)
 895 {
 896         struct sockaddr_in *sin, *sinfrom;
 897
 898         if (msgname) {
 899                 struct sctp_association *asoc;
 900
 901                 asoc = event->asoc;
 902                 sctp_inet_msgname(msgname, addr_len);
 903                 sin = (struct sockaddr_in *)msgname;
 904                 sinfrom = &asoc->peer.primary_addr.v4;
 905                 sin->sin_port = htons(asoc->peer.port);
 906                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 907         }
 908 }
 909
 910 /* Initialize and copy out a msgname from an inbound skb. */
 911 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 912 {
 913         if (msgname) {
 914                 struct sctphdr *sh = sctp_hdr(skb);
 915                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 916
 917                 sctp_inet_msgname(msgname, len);
 918                 sin->sin_port = sh->source;
 919                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 920         }
 921 }
 922
 923 /* Do we support this AF? */
 924 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 925 {
 926         /* PF_INET only supports AF_INET addresses. */
 927         return AF_INET == family;
 928 }
 929
 930 /* Address matching with wildcards allowed. */
 931 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 932                               const union sctp_addr *addr2,
 933                               struct sctp_sock *opt)
 934 {
 935         /* PF_INET only supports AF_INET addresses. */
 936         if (addr1->sa.sa_family != addr2->sa.sa_family)
 937                 return 0;
 938         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 939             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 940                 return 1;
 941         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 942                 return 1;
 943
 944         return 0;
 945 }
 946
 947 /* Verify that provided sockaddr looks bindable.  Common verification has
 948  * already been taken care of.
 949  */
 950 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 951 {
 952         return sctp_v4_available(addr, opt);
 953 }
 954
 955 /* Verify that sockaddr looks sendable.  Common verification has already
 956  * been taken care of.
 957  */
 958 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 959 {
 960         return 1;
 961 }
 962
 963 /* Fill in Supported Address Type information for INIT and INIT-ACK
 964  * chunks.  Returns number of addresses supported.
 965  */
 966 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 967                                      __be16 *types)
 968 {
 969         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 970         return 1;
 971 }
 972
 973 /* Wrapper routine that calls the ip transmit routine. */
 974 static inline int sctp_v4_xmit(struct sk_buff *skb,
 975                                struct sctp_transport *transport)
 976 {
 977         struct inet_sock *inet = inet_sk(skb->sk);
 978         __u8 dscp = inet->tos;
 979
 980         pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
 981                  skb->len, &transport->fl.u.ip4.saddr,
 982                  &transport->fl.u.ip4.daddr);
 983
 984         if (transport->dscp & SCTP_DSCP_SET_MASK)
 985                 dscp = transport->dscp & SCTP_DSCP_VAL_MASK;
 986
 987         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
 988                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 989
 990         SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
 991
 992         return __ip_queue_xmit(&inet->sk, skb, &transport->fl, dscp);
 993 }
 994
 995 static struct sctp_af sctp_af_inet;
 996
 997 static struct sctp_pf sctp_pf_inet = {
 998         .event_msgname = sctp_inet_event_msgname,
 999         .skb_msgname   = sctp_inet_skb_msgname,
1000         .af_supported  = sctp_inet_af_supported,
1001         .cmp_addr      = sctp_inet_cmp_addr,
1002         .bind_verify   = sctp_inet_bind_verify,
1003         .send_verify   = sctp_inet_send_verify,
1004         .supported_addrs = sctp_inet_supported_addrs,
1005         .create_accept_sk = sctp_v4_create_accept_sk,
1006         .addr_to_user  = sctp_v4_addr_to_user,
1007         .to_sk_saddr   = sctp_v4_to_sk_saddr,
1008         .to_sk_daddr   = sctp_v4_to_sk_daddr,
1009         .copy_ip_options = sctp_v4_copy_ip_options,
1010         .af            = &sctp_af_inet
1011 };
1012
1013 /* Notifier for inetaddr addition/deletion events.  */
1014 static struct notifier_block sctp_inetaddr_notifier = {
1015         .notifier_call = sctp_inetaddr_event,
1016 };
1017
1018 /* Socket operations.  */
1019 static const struct proto_ops inet_seqpacket_ops = {
1020         .family            = PF_INET,
1021         .owner             = THIS_MODULE,
1022         .release           = inet_release,      /* Needs to be wrapped... */
1023         .bind              = inet_bind,
1024         .connect           = sctp_inet_connect,
1025         .socketpair        = sock_no_socketpair,
1026         .accept            = inet_accept,
1027         .getname           = inet_getname,      /* Semantics are different.  */
1028         .poll              = sctp_poll,
1029         .ioctl             = inet_ioctl,
1030         .gettstamp         = sock_gettstamp,
1031         .listen            = sctp_inet_listen,
1032         .shutdown          = inet_shutdown,     /* Looks harmless.  */
1033         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1034         .getsockopt        = sock_common_getsockopt,
1035         .sendmsg           = inet_sendmsg,
1036         .recvmsg           = inet_recvmsg,
1037         .mmap              = sock_no_mmap,
1038         .sendpage          = sock_no_sendpage,
1039 };
1040
1041 /* Registration with AF_INET family.  */
1042 static struct inet_protosw sctp_seqpacket_protosw = {
1043         .type       = SOCK_SEQPACKET,
1044         .protocol   = IPPROTO_SCTP,
1045         .prot       = &sctp_prot,
1046         .ops        = &inet_seqpacket_ops,
1047         .flags      = SCTP_PROTOSW_FLAG
1048 };
1049 static struct inet_protosw sctp_stream_protosw = {
1050         .type       = SOCK_STREAM,
1051         .protocol   = IPPROTO_SCTP,
1052         .prot       = &sctp_prot,
1053         .ops        = &inet_seqpacket_ops,
1054         .flags      = SCTP_PROTOSW_FLAG
1055 };
1056
1057 /* Register with IP layer.  */
1058 static const struct net_protocol sctp_protocol = {
1059         .handler     = sctp_rcv,
1060         .err_handler = sctp_v4_err,
1061         .no_policy   = 1,
1062         .netns_ok    = 1,
1063         .icmp_strict_tag_validation = 1,
1064 };
1065
1066 /* IPv4 address related functions.  */
1067 static struct sctp_af sctp_af_inet = {
1068         .sa_family         = AF_INET,
1069         .sctp_xmit         = sctp_v4_xmit,
1070         .setsockopt        = ip_setsockopt,
1071         .getsockopt        = ip_getsockopt,
1072         .get_dst           = sctp_v4_get_dst,
1073         .get_saddr         = sctp_v4_get_saddr,
1074         .copy_addrlist     = sctp_v4_copy_addrlist,
1075         .from_skb          = sctp_v4_from_skb,
1076         .from_sk           = sctp_v4_from_sk,
1077         .from_addr_param   = sctp_v4_from_addr_param,
1078         .to_addr_param     = sctp_v4_to_addr_param,
1079         .cmp_addr          = sctp_v4_cmp_addr,
1080         .addr_valid        = sctp_v4_addr_valid,
1081         .inaddr_any        = sctp_v4_inaddr_any,
1082         .is_any            = sctp_v4_is_any,
1083         .available         = sctp_v4_available,
1084         .scope             = sctp_v4_scope,
1085         .skb_iif           = sctp_v4_skb_iif,
1086         .is_ce             = sctp_v4_is_ce,
1087         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1088         .ecn_capable       = sctp_v4_ecn_capable,
1089         .net_header_len    = sizeof(struct iphdr),
1090         .sockaddr_len      = sizeof(struct sockaddr_in),
1091         .ip_options_len    = sctp_v4_ip_options_len,
1092 };
1093
1094 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1095 {
1096         switch (family) {
1097         case PF_INET:
1098                 return sctp_pf_inet_specific;
1099         case PF_INET6:
1100                 return sctp_pf_inet6_specific;
1101         default:
1102                 return NULL;
1103         }
1104 }
1105
1106 /* Register the PF specific function table.  */
1107 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1108 {
1109         switch (family) {
1110         case PF_INET:
1111                 if (sctp_pf_inet_specific)
1112                         return 0;
1113                 sctp_pf_inet_specific = pf;
1114                 break;
1115         case PF_INET6:
1116                 if (sctp_pf_inet6_specific)
1117                         return 0;
1118                 sctp_pf_inet6_specific = pf;
1119                 break;
1120         default:
1121                 return 0;
1122         }
1123         return 1;
1124 }
1125
1126 static inline int init_sctp_mibs(struct net *net)
1127 {
1128         net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1129         if (!net->sctp.sctp_statistics)
1130                 return -ENOMEM;
1131         return 0;
1132 }
1133
1134 static inline void cleanup_sctp_mibs(struct net *net)
1135 {
1136         free_percpu(net->sctp.sctp_statistics);
1137 }
1138
1139 static void sctp_v4_pf_init(void)
1140 {
1141         /* Initialize the SCTP specific PF functions. */
1142         sctp_register_pf(&sctp_pf_inet, PF_INET);
1143         sctp_register_af(&sctp_af_inet);
1144 }
1145
1146 static void sctp_v4_pf_exit(void)
1147 {
1148         list_del(&sctp_af_inet.list);
1149 }
1150
1151 static int sctp_v4_protosw_init(void)
1152 {
1153         int rc;
1154
1155         rc = proto_register(&sctp_prot, 1);
1156         if (rc)
1157                 return rc;
1158
1159         /* Register SCTP(UDP and TCP style) with socket layer.  */
1160         inet_register_protosw(&sctp_seqpacket_protosw);
1161         inet_register_protosw(&sctp_stream_protosw);
1162
1163         return 0;
1164 }
1165
1166 static void sctp_v4_protosw_exit(void)
1167 {
1168         inet_unregister_protosw(&sctp_stream_protosw);
1169         inet_unregister_protosw(&sctp_seqpacket_protosw);
1170         proto_unregister(&sctp_prot);
1171 }
1172
1173 static int sctp_v4_add_protocol(void)
1174 {
1175         /* Register notifier for inet address additions/deletions. */
1176         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1177
1178         /* Register SCTP with inet layer.  */
1179         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1180                 return -EAGAIN;
1181
1182         return 0;
1183 }
1184
1185 static void sctp_v4_del_protocol(void)
1186 {
1187         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1188         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1189 }
1190
1191 static int __net_init sctp_defaults_init(struct net *net)
1192 {
1193         int status;
1194
1195         /*
1196          * 14. Suggested SCTP Protocol Parameter Values
1197          */
1198         /* The following protocol parameters are RECOMMENDED:  */
1199         /* RTO.Initial              - 3  seconds */
1200         net->sctp.rto_initial                   = SCTP_RTO_INITIAL;
1201         /* RTO.Min                  - 1  second */
1202         net->sctp.rto_min                       = SCTP_RTO_MIN;
1203         /* RTO.Max                 -  60 seconds */
1204         net->sctp.rto_max                       = SCTP_RTO_MAX;
1205         /* RTO.Alpha                - 1/8 */
1206         net->sctp.rto_alpha                     = SCTP_RTO_ALPHA;
1207         /* RTO.Beta                 - 1/4 */
1208         net->sctp.rto_beta                      = SCTP_RTO_BETA;
1209
1210         /* Valid.Cookie.Life        - 60  seconds */
1211         net->sctp.valid_cookie_life             = SCTP_DEFAULT_COOKIE_LIFE;
1212
1213         /* Whether Cookie Preservative is enabled(1) or not(0) */
1214         net->sctp.cookie_preserve_enable        = 1;
1215
1216         /* Default sctp sockets to use md5 as their hmac alg */
1217 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1218         net->sctp.sctp_hmac_alg                 = "md5";
1219 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1220         net->sctp.sctp_hmac_alg                 = "sha1";
1221 #else
1222         net->sctp.sctp_hmac_alg                 = NULL;
1223 #endif
1224
1225         /* Max.Burst                - 4 */
1226         net->sctp.max_burst                     = SCTP_DEFAULT_MAX_BURST;
1227
1228         /* Disable of Primary Path Switchover by default */
1229         net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1230
1231         /* Enable pf state by default */
1232         net->sctp.pf_enable = 1;
1233
1234         /* Ignore pf exposure feature by default */
1235         net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1236
1237         /* Association.Max.Retrans  - 10 attempts
1238          * Path.Max.Retrans         - 5  attempts (per destination address)
1239          * Max.Init.Retransmits     - 8  attempts
1240          */
1241         net->sctp.max_retrans_association       = 10;
1242         net->sctp.max_retrans_path              = 5;
1243         net->sctp.max_retrans_init              = 8;
1244
1245         /* Sendbuffer growth        - do per-socket accounting */
1246         net->sctp.sndbuf_policy                 = 0;
1247
1248         /* Rcvbuffer growth         - do per-socket accounting */
1249         net->sctp.rcvbuf_policy                 = 0;
1250
1251         /* HB.interval              - 30 seconds */
1252         net->sctp.hb_interval                   = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1253
1254         /* delayed SACK timeout */
1255         net->sctp.sack_timeout                  = SCTP_DEFAULT_TIMEOUT_SACK;
1256
1257         /* Disable ADDIP by default. */
1258         net->sctp.addip_enable = 0;
1259         net->sctp.addip_noauth = 0;
1260         net->sctp.default_auto_asconf = 0;
1261
1262         /* Enable PR-SCTP by default. */
1263         net->sctp.prsctp_enable = 1;
1264
1265         /* Disable RECONF by default. */
1266         net->sctp.reconf_enable = 0;
1267
1268         /* Disable AUTH by default. */
1269         net->sctp.auth_enable = 0;
1270
1271         /* Enable ECN by default. */
1272         net->sctp.ecn_enable = 1;
1273
1274         /* Set SCOPE policy to enabled */
1275         net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1276
1277         /* Set the default rwnd update threshold */
1278         net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1279
1280         /* Initialize maximum autoclose timeout. */
1281         net->sctp.max_autoclose         = INT_MAX / HZ;
1282
1283         status = sctp_sysctl_net_register(net);
1284         if (status)
1285                 goto err_sysctl_register;
1286
1287         /* Allocate and initialise sctp mibs.  */
1288         status = init_sctp_mibs(net);
1289         if (status)
1290                 goto err_init_mibs;
1291
1292 #ifdef CONFIG_PROC_FS
1293         /* Initialize proc fs directory.  */
1294         status = sctp_proc_init(net);
1295         if (status)
1296                 goto err_init_proc;
1297 #endif
1298
1299         sctp_dbg_objcnt_init(net);
1300
1301         /* Initialize the local address list. */
1302         INIT_LIST_HEAD(&net->sctp.local_addr_list);
1303         spin_lock_init(&net->sctp.local_addr_lock);
1304         sctp_get_local_addr_list(net);
1305
1306         /* Initialize the address event list */
1307         INIT_LIST_HEAD(&net->sctp.addr_waitq);
1308         INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1309         spin_lock_init(&net->sctp.addr_wq_lock);
1310         net->sctp.addr_wq_timer.expires = 0;
1311         timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1312
1313         return 0;
1314
1315 #ifdef CONFIG_PROC_FS
1316 err_init_proc:
1317         cleanup_sctp_mibs(net);
1318 #endif
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 #ifdef CONFIG_PROC_FS
1332         remove_proc_subtree("sctp", net->proc_net);
1333         net->sctp.proc_net_sctp = NULL;
1334 #endif
1335         cleanup_sctp_mibs(net);
1336         sctp_sysctl_net_unregister(net);
1337 }
1338
1339 static struct pernet_operations sctp_defaults_ops = {
1340         .init = sctp_defaults_init,
1341         .exit = sctp_defaults_exit,
1342 };
1343
1344 static int __net_init sctp_ctrlsock_init(struct net *net)
1345 {
1346         int status;
1347
1348         /* Initialize the control inode/socket for handling OOTB packets.  */
1349         status = sctp_ctl_sock_init(net);
1350         if (status)
1351                 pr_err("Failed to initialize the SCTP control sock\n");
1352
1353         return status;
1354 }
1355
1356 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1357 {
1358         /* Free the control endpoint.  */
1359         inet_ctl_sock_destroy(net->sctp.ctl_sock);
1360 }
1361
1362 static struct pernet_operations sctp_ctrlsock_ops = {
1363         .init = sctp_ctrlsock_init,
1364         .exit = sctp_ctrlsock_exit,
1365 };
1366
1367 /* Initialize the universe into something sensible.  */
1368 static __init int sctp_init(void)
1369 {
1370         unsigned long nr_pages = totalram_pages();
1371         unsigned long limit;
1372         unsigned long goal;
1373         int max_entry_order;
1374         int num_entries;
1375         int max_share;
1376         int status;
1377         int order;
1378         int i;
1379
1380         sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1381
1382         /* Allocate bind_bucket and chunk caches. */
1383         status = -ENOBUFS;
1384         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1385                                                sizeof(struct sctp_bind_bucket),
1386                                                0, SLAB_HWCACHE_ALIGN,
1387                                                NULL);
1388         if (!sctp_bucket_cachep)
1389                 goto out;
1390
1391         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1392                                                sizeof(struct sctp_chunk),
1393                                                0, SLAB_HWCACHE_ALIGN,
1394                                                NULL);
1395         if (!sctp_chunk_cachep)
1396                 goto err_chunk_cachep;
1397
1398         status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1399         if (status)
1400                 goto err_percpu_counter_init;
1401
1402         /* Implementation specific variables. */
1403
1404         /* Initialize default stream count setup information. */
1405         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1406         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1407
1408         /* Initialize handle used for association ids. */
1409         idr_init(&sctp_assocs_id);
1410
1411         limit = nr_free_buffer_pages() / 8;
1412         limit = max(limit, 128UL);
1413         sysctl_sctp_mem[0] = limit / 4 * 3;
1414         sysctl_sctp_mem[1] = limit;
1415         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1416
1417         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1418         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1419         max_share = min(4UL*1024*1024, limit);
1420
1421         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1422         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1423         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1424
1425         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1426         sysctl_sctp_wmem[1] = 16*1024;
1427         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1428
1429         /* Size and allocate the association hash table.
1430          * The methodology is similar to that of the tcp hash tables.
1431          * Though not identical.  Start by getting a goal size
1432          */
1433         if (nr_pages >= (128 * 1024))
1434                 goal = nr_pages >> (22 - PAGE_SHIFT);
1435         else
1436                 goal = nr_pages >> (24 - PAGE_SHIFT);
1437
1438         /* Then compute the page order for said goal */
1439         order = get_order(goal);
1440
1441         /* Now compute the required page order for the maximum sized table we
1442          * want to create
1443          */
1444         max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1445                                     sizeof(struct sctp_bind_hashbucket));
1446
1447         /* Limit the page order by that maximum hash table size */
1448         order = min(order, max_entry_order);
1449
1450         /* Allocate and initialize the endpoint hash table.  */
1451         sctp_ep_hashsize = 64;
1452         sctp_ep_hashtable =
1453                 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1454         if (!sctp_ep_hashtable) {
1455                 pr_err("Failed endpoint_hash alloc\n");
1456                 status = -ENOMEM;
1457                 goto err_ehash_alloc;
1458         }
1459         for (i = 0; i < sctp_ep_hashsize; i++) {
1460                 rwlock_init(&sctp_ep_hashtable[i].lock);
1461                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1462         }
1463
1464         /* Allocate and initialize the SCTP port hash table.
1465          * Note that order is initalized to start at the max sized
1466          * table we want to support.  If we can't get that many pages
1467          * reduce the order and try again
1468          */
1469         do {
1470                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1471                         __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1472         } while (!sctp_port_hashtable && --order > 0);
1473
1474         if (!sctp_port_hashtable) {
1475                 pr_err("Failed bind hash alloc\n");
1476                 status = -ENOMEM;
1477                 goto err_bhash_alloc;
1478         }
1479
1480         /* Now compute the number of entries that will fit in the
1481          * port hash space we allocated
1482          */
1483         num_entries = (1UL << order) * PAGE_SIZE /
1484                       sizeof(struct sctp_bind_hashbucket);
1485
1486         /* And finish by rounding it down to the nearest power of two.
1487          * This wastes some memory of course, but it's needed because
1488          * the hash function operates based on the assumption that
1489          * the number of entries is a power of two.
1490          */
1491         sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1492
1493         for (i = 0; i < sctp_port_hashsize; i++) {
1494                 spin_lock_init(&sctp_port_hashtable[i].lock);
1495                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1496         }
1497
1498         status = sctp_transport_hashtable_init();
1499         if (status)
1500                 goto err_thash_alloc;
1501
1502         pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1503                 num_entries);
1504
1505         sctp_sysctl_register();
1506
1507         INIT_LIST_HEAD(&sctp_address_families);
1508         sctp_v4_pf_init();
1509         sctp_v6_pf_init();
1510         sctp_sched_ops_init();
1511
1512         status = register_pernet_subsys(&sctp_defaults_ops);
1513         if (status)
1514                 goto err_register_defaults;
1515
1516         status = sctp_v4_protosw_init();
1517         if (status)
1518                 goto err_protosw_init;
1519
1520         status = sctp_v6_protosw_init();
1521         if (status)
1522                 goto err_v6_protosw_init;
1523
1524         status = register_pernet_subsys(&sctp_ctrlsock_ops);
1525         if (status)
1526                 goto err_register_ctrlsock;
1527
1528         status = sctp_v4_add_protocol();
1529         if (status)
1530                 goto err_add_protocol;
1531
1532         /* Register SCTP with inet6 layer.  */
1533         status = sctp_v6_add_protocol();
1534         if (status)
1535                 goto err_v6_add_protocol;
1536
1537         if (sctp_offload_init() < 0)
1538                 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1539
1540 out:
1541         return status;
1542 err_v6_add_protocol:
1543         sctp_v4_del_protocol();
1544 err_add_protocol:
1545         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1546 err_register_ctrlsock:
1547         sctp_v6_protosw_exit();
1548 err_v6_protosw_init:
1549         sctp_v4_protosw_exit();
1550 err_protosw_init:
1551         unregister_pernet_subsys(&sctp_defaults_ops);
1552 err_register_defaults:
1553         sctp_v4_pf_exit();
1554         sctp_v6_pf_exit();
1555         sctp_sysctl_unregister();
1556         free_pages((unsigned long)sctp_port_hashtable,
1557                    get_order(sctp_port_hashsize *
1558                              sizeof(struct sctp_bind_hashbucket)));
1559 err_bhash_alloc:
1560         sctp_transport_hashtable_destroy();
1561 err_thash_alloc:
1562         kfree(sctp_ep_hashtable);
1563 err_ehash_alloc:
1564         percpu_counter_destroy(&sctp_sockets_allocated);
1565 err_percpu_counter_init:
1566         kmem_cache_destroy(sctp_chunk_cachep);
1567 err_chunk_cachep:
1568         kmem_cache_destroy(sctp_bucket_cachep);
1569         goto out;
1570 }
1571
1572 /* Exit handler for the SCTP protocol.  */
1573 static __exit void sctp_exit(void)
1574 {
1575         /* BUG.  This should probably do something useful like clean
1576          * up all the remaining associations and all that memory.
1577          */
1578
1579         /* Unregister with inet6/inet layers. */
1580         sctp_v6_del_protocol();
1581         sctp_v4_del_protocol();
1582
1583         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1584
1585         /* Free protosw registrations */
1586         sctp_v6_protosw_exit();
1587         sctp_v4_protosw_exit();
1588
1589         unregister_pernet_subsys(&sctp_defaults_ops);
1590
1591         /* Unregister with socket layer. */
1592         sctp_v6_pf_exit();
1593         sctp_v4_pf_exit();
1594
1595         sctp_sysctl_unregister();
1596
1597         free_pages((unsigned long)sctp_port_hashtable,
1598                    get_order(sctp_port_hashsize *
1599                              sizeof(struct sctp_bind_hashbucket)));
1600         kfree(sctp_ep_hashtable);
1601         sctp_transport_hashtable_destroy();
1602
1603         percpu_counter_destroy(&sctp_sockets_allocated);
1604
1605         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1606
1607         kmem_cache_destroy(sctp_chunk_cachep);
1608         kmem_cache_destroy(sctp_bucket_cachep);
1609 }
1610
1611 module_init(sctp_init);
1612 module_exit(sctp_exit);
1613
1614 /*
1615  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1616  */
1617 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1618 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1619 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1620 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1621 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1622 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1623 MODULE_LICENSE("GPL");