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_PEERSUPP))
 152                         continue;
 153                 if (addr->a.sa.sa_family == AF_INET6 &&
 154                     (!(copy_flags & SCTP_ADDR6_ALLOWED) ||
 155                      !(copy_flags & SCTP_ADDR6_PEERSUPP)))
 156                         continue;
 157
 158                 laddr = addr->a;
 159                 /* also works for setting ipv6 address port */
 160                 laddr.v4.sin_port = htons(bp->port);
 161                 if (sctp_bind_addr_state(bp, &laddr) != -1)
 162                         continue;
 163
 164                 error = sctp_add_bind_addr(bp, &addr->a, sizeof(addr->a),
 165                                            SCTP_ADDR_SRC, GFP_ATOMIC);
 166                 if (error)
 167                         break;
 168         }
 169
 170         rcu_read_unlock();
 171         return error;
 172 }
 173
 174 /* Copy over any ip options */
 175 static void sctp_v4_copy_ip_options(struct sock *sk, struct sock *newsk)
 176 {
 177         struct inet_sock *newinet, *inet = inet_sk(sk);
 178         struct ip_options_rcu *inet_opt, *newopt = NULL;
 179
 180         newinet = inet_sk(newsk);
 181
 182         rcu_read_lock();
 183         inet_opt = rcu_dereference(inet->inet_opt);
 184         if (inet_opt) {
 185                 newopt = sock_kmalloc(newsk, sizeof(*inet_opt) +
 186                                       inet_opt->opt.optlen, GFP_ATOMIC);
 187                 if (newopt)
 188                         memcpy(newopt, inet_opt, sizeof(*inet_opt) +
 189                                inet_opt->opt.optlen);
 190                 else
 191                         pr_err("%s: Failed to copy ip options\n", __func__);
 192         }
 193         RCU_INIT_POINTER(newinet->inet_opt, newopt);
 194         rcu_read_unlock();
 195 }
 196
 197 /* Account for the IP options */
 198 static int sctp_v4_ip_options_len(struct sock *sk)
 199 {
 200         struct inet_sock *inet = inet_sk(sk);
 201         struct ip_options_rcu *inet_opt;
 202         int len = 0;
 203
 204         rcu_read_lock();
 205         inet_opt = rcu_dereference(inet->inet_opt);
 206         if (inet_opt)
 207                 len = inet_opt->opt.optlen;
 208
 209         rcu_read_unlock();
 210         return len;
 211 }
 212
 213 /* Initialize a sctp_addr from in incoming skb.  */
 214 static void sctp_v4_from_skb(union sctp_addr *addr, struct sk_buff *skb,
 215                              int is_saddr)
 216 {
 217         /* Always called on head skb, so this is safe */
 218         struct sctphdr *sh = sctp_hdr(skb);
 219         struct sockaddr_in *sa = &addr->v4;
 220
 221         addr->v4.sin_family = AF_INET;
 222
 223         if (is_saddr) {
 224                 sa->sin_port = sh->source;
 225                 sa->sin_addr.s_addr = ip_hdr(skb)->saddr;
 226         } else {
 227                 sa->sin_port = sh->dest;
 228                 sa->sin_addr.s_addr = ip_hdr(skb)->daddr;
 229         }
 230         memset(sa->sin_zero, 0, sizeof(sa->sin_zero));
 231 }
 232
 233 /* Initialize an sctp_addr from a socket. */
 234 static void sctp_v4_from_sk(union sctp_addr *addr, struct sock *sk)
 235 {
 236         addr->v4.sin_family = AF_INET;
 237         addr->v4.sin_port = 0;
 238         addr->v4.sin_addr.s_addr = inet_sk(sk)->inet_rcv_saddr;
 239         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 240 }
 241
 242 /* Initialize sk->sk_rcv_saddr from sctp_addr. */
 243 static void sctp_v4_to_sk_saddr(union sctp_addr *addr, struct sock *sk)
 244 {
 245         inet_sk(sk)->inet_rcv_saddr = addr->v4.sin_addr.s_addr;
 246 }
 247
 248 /* Initialize sk->sk_daddr from sctp_addr. */
 249 static void sctp_v4_to_sk_daddr(union sctp_addr *addr, struct sock *sk)
 250 {
 251         inet_sk(sk)->inet_daddr = addr->v4.sin_addr.s_addr;
 252 }
 253
 254 /* Initialize a sctp_addr from an address parameter. */
 255 static void sctp_v4_from_addr_param(union sctp_addr *addr,
 256                                     union sctp_addr_param *param,
 257                                     __be16 port, int iif)
 258 {
 259         addr->v4.sin_family = AF_INET;
 260         addr->v4.sin_port = port;
 261         addr->v4.sin_addr.s_addr = param->v4.addr.s_addr;
 262         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 263 }
 264
 265 /* Initialize an address parameter from a sctp_addr and return the length
 266  * of the address parameter.
 267  */
 268 static int sctp_v4_to_addr_param(const union sctp_addr *addr,
 269                                  union sctp_addr_param *param)
 270 {
 271         int length = sizeof(struct sctp_ipv4addr_param);
 272
 273         param->v4.param_hdr.type = SCTP_PARAM_IPV4_ADDRESS;
 274         param->v4.param_hdr.length = htons(length);
 275         param->v4.addr.s_addr = addr->v4.sin_addr.s_addr;
 276
 277         return length;
 278 }
 279
 280 /* Initialize a sctp_addr from a dst_entry. */
 281 static void sctp_v4_dst_saddr(union sctp_addr *saddr, struct flowi4 *fl4,
 282                               __be16 port)
 283 {
 284         saddr->v4.sin_family = AF_INET;
 285         saddr->v4.sin_port = port;
 286         saddr->v4.sin_addr.s_addr = fl4->saddr;
 287         memset(saddr->v4.sin_zero, 0, sizeof(saddr->v4.sin_zero));
 288 }
 289
 290 /* Compare two addresses exactly. */
 291 static int sctp_v4_cmp_addr(const union sctp_addr *addr1,
 292                             const union sctp_addr *addr2)
 293 {
 294         if (addr1->sa.sa_family != addr2->sa.sa_family)
 295                 return 0;
 296         if (addr1->v4.sin_port != addr2->v4.sin_port)
 297                 return 0;
 298         if (addr1->v4.sin_addr.s_addr != addr2->v4.sin_addr.s_addr)
 299                 return 0;
 300
 301         return 1;
 302 }
 303
 304 /* Initialize addr struct to INADDR_ANY. */
 305 static void sctp_v4_inaddr_any(union sctp_addr *addr, __be16 port)
 306 {
 307         addr->v4.sin_family = AF_INET;
 308         addr->v4.sin_addr.s_addr = htonl(INADDR_ANY);
 309         addr->v4.sin_port = port;
 310         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 311 }
 312
 313 /* Is this a wildcard address? */
 314 static int sctp_v4_is_any(const union sctp_addr *addr)
 315 {
 316         return htonl(INADDR_ANY) == addr->v4.sin_addr.s_addr;
 317 }
 318
 319 /* This function checks if the address is a valid address to be used for
 320  * SCTP binding.
 321  *
 322  * Output:
 323  * Return 0 - If the address is a non-unicast or an illegal address.
 324  * Return 1 - If the address is a unicast.
 325  */
 326 static int sctp_v4_addr_valid(union sctp_addr *addr,
 327                               struct sctp_sock *sp,
 328                               const struct sk_buff *skb)
 329 {
 330         /* IPv4 addresses not allowed */
 331         if (sp && ipv6_only_sock(sctp_opt2sk(sp)))
 332                 return 0;
 333
 334         /* Is this a non-unicast address or a unusable SCTP address? */
 335         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr))
 336                 return 0;
 337
 338         /* Is this a broadcast address? */
 339         if (skb && skb_rtable(skb)->rt_flags & RTCF_BROADCAST)
 340                 return 0;
 341
 342         return 1;
 343 }
 344
 345 /* Should this be available for binding?   */
 346 static int sctp_v4_available(union sctp_addr *addr, struct sctp_sock *sp)
 347 {
 348         struct net *net = sock_net(&sp->inet.sk);
 349         int ret = inet_addr_type(net, addr->v4.sin_addr.s_addr);
 350
 351
 352         if (addr->v4.sin_addr.s_addr != htonl(INADDR_ANY) &&
 353            ret != RTN_LOCAL &&
 354            !sp->inet.freebind &&
 355            !net->ipv4.sysctl_ip_nonlocal_bind)
 356                 return 0;
 357
 358         if (ipv6_only_sock(sctp_opt2sk(sp)))
 359                 return 0;
 360
 361         return 1;
 362 }
 363
 364 /* Checking the loopback, private and other address scopes as defined in
 365  * RFC 1918.   The IPv4 scoping is based on the draft for SCTP IPv4
 366  * scoping <draft-stewart-tsvwg-sctp-ipv4-00.txt>.
 367  *
 368  * Level 0 - unusable SCTP addresses
 369  * Level 1 - loopback address
 370  * Level 2 - link-local addresses
 371  * Level 3 - private addresses.
 372  * Level 4 - global addresses
 373  * For INIT and INIT-ACK address list, let L be the level of
 374  * of requested destination address, sender and receiver
 375  * SHOULD include all of its addresses with level greater
 376  * than or equal to L.
 377  *
 378  * IPv4 scoping can be controlled through sysctl option
 379  * net.sctp.addr_scope_policy
 380  */
 381 static enum sctp_scope sctp_v4_scope(union sctp_addr *addr)
 382 {
 383         enum sctp_scope retval;
 384
 385         /* Check for unusable SCTP addresses. */
 386         if (IS_IPV4_UNUSABLE_ADDRESS(addr->v4.sin_addr.s_addr)) {
 387                 retval =  SCTP_SCOPE_UNUSABLE;
 388         } else if (ipv4_is_loopback(addr->v4.sin_addr.s_addr)) {
 389                 retval = SCTP_SCOPE_LOOPBACK;
 390         } else if (ipv4_is_linklocal_169(addr->v4.sin_addr.s_addr)) {
 391                 retval = SCTP_SCOPE_LINK;
 392         } else if (ipv4_is_private_10(addr->v4.sin_addr.s_addr) ||
 393                    ipv4_is_private_172(addr->v4.sin_addr.s_addr) ||
 394                    ipv4_is_private_192(addr->v4.sin_addr.s_addr)) {
 395                 retval = SCTP_SCOPE_PRIVATE;
 396         } else {
 397                 retval = SCTP_SCOPE_GLOBAL;
 398         }
 399
 400         return retval;
 401 }
 402
 403 /* Returns a valid dst cache entry for the given source and destination ip
 404  * addresses. If an association is passed, trys to get a dst entry with a
 405  * source address that matches an address in the bind address list.
 406  */
 407 static void sctp_v4_get_dst(struct sctp_transport *t, union sctp_addr *saddr,
 408                                 struct flowi *fl, struct sock *sk)
 409 {
 410         struct sctp_association *asoc = t->asoc;
 411         struct rtable *rt;
 412         struct flowi4 *fl4 = &fl->u.ip4;
 413         struct sctp_bind_addr *bp;
 414         struct sctp_sockaddr_entry *laddr;
 415         struct dst_entry *dst = NULL;
 416         union sctp_addr *daddr = &t->ipaddr;
 417         union sctp_addr dst_saddr;
 418         __u8 tos = inet_sk(sk)->tos;
 419
 420         if (t->dscp & SCTP_DSCP_SET_MASK)
 421                 tos = t->dscp & SCTP_DSCP_VAL_MASK;
 422         memset(fl4, 0x0, sizeof(struct flowi4));
 423         fl4->daddr  = daddr->v4.sin_addr.s_addr;
 424         fl4->fl4_dport = daddr->v4.sin_port;
 425         fl4->flowi4_proto = IPPROTO_SCTP;
 426         if (asoc) {
 427                 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
 428                 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
 429                 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
 430         }
 431         if (saddr) {
 432                 fl4->saddr = saddr->v4.sin_addr.s_addr;
 433                 if (!fl4->fl4_sport)
 434                         fl4->fl4_sport = saddr->v4.sin_port;
 435         }
 436
 437         pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
 438                  &fl4->saddr);
 439
 440         rt = ip_route_output_key(sock_net(sk), fl4);
 441         if (!IS_ERR(rt))
 442                 dst = &rt->dst;
 443
 444         /* If there is no association or if a source address is passed, no
 445          * more validation is required.
 446          */
 447         if (!asoc || saddr)
 448                 goto out;
 449
 450         bp = &asoc->base.bind_addr;
 451
 452         if (dst) {
 453                 /* Walk through the bind address list and look for a bind
 454                  * address that matches the source address of the returned dst.
 455                  */
 456                 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
 457                 rcu_read_lock();
 458                 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 459                         if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
 460                             (laddr->state != SCTP_ADDR_SRC &&
 461                             !asoc->src_out_of_asoc_ok))
 462                                 continue;
 463                         if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
 464                                 goto out_unlock;
 465                 }
 466                 rcu_read_unlock();
 467
 468                 /* None of the bound addresses match the source address of the
 469                  * dst. So release it.
 470                  */
 471                 dst_release(dst);
 472                 dst = NULL;
 473         }
 474
 475         /* Walk through the bind address list and try to get a dst that
 476          * matches a bind address as the source address.
 477          */
 478         rcu_read_lock();
 479         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 480                 struct net_device *odev;
 481
 482                 if (!laddr->valid)
 483                         continue;
 484                 if (laddr->state != SCTP_ADDR_SRC ||
 485                     AF_INET != laddr->a.sa.sa_family)
 486                         continue;
 487
 488                 fl4->fl4_sport = laddr->a.v4.sin_port;
 489                 flowi4_update_output(fl4,
 490                                      asoc->base.sk->sk_bound_dev_if,
 491                                      RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
 492                                      daddr->v4.sin_addr.s_addr,
 493                                      laddr->a.v4.sin_addr.s_addr);
 494
 495                 rt = ip_route_output_key(sock_net(sk), fl4);
 496                 if (IS_ERR(rt))
 497                         continue;
 498
 499                 /* Ensure the src address belongs to the output
 500                  * interface.
 501                  */
 502                 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
 503                                      false);
 504                 if (!odev || odev->ifindex != fl4->flowi4_oif) {
 505                         if (!dst)
 506                                 dst = &rt->dst;
 507                         else
 508                                 dst_release(&rt->dst);
 509                         continue;
 510                 }
 511
 512                 dst_release(dst);
 513                 dst = &rt->dst;
 514                 break;
 515         }
 516
 517 out_unlock:
 518         rcu_read_unlock();
 519 out:
 520         t->dst = dst;
 521         if (dst)
 522                 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
 523                          &fl4->daddr, &fl4->saddr);
 524         else
 525                 pr_debug("no route\n");
 526 }
 527
 528 /* For v4, the source address is cached in the route entry(dst). So no need
 529  * to cache it separately and hence this is an empty routine.
 530  */
 531 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 532                               struct sctp_transport *t,
 533                               struct flowi *fl)
 534 {
 535         union sctp_addr *saddr = &t->saddr;
 536         struct rtable *rt = (struct rtable *)t->dst;
 537
 538         if (rt) {
 539                 saddr->v4.sin_family = AF_INET;
 540                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 541         }
 542 }
 543
 544 /* What interface did this skb arrive on? */
 545 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 546 {
 547         return inet_iif(skb);
 548 }
 549
 550 /* Was this packet marked by Explicit Congestion Notification? */
 551 static int sctp_v4_is_ce(const struct sk_buff *skb)
 552 {
 553         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 554 }
 555
 556 /* Create and initialize a new sk for the socket returned by accept(). */
 557 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 558                                              struct sctp_association *asoc,
 559                                              bool kern)
 560 {
 561         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 562                         sk->sk_prot, kern);
 563         struct inet_sock *newinet;
 564
 565         if (!newsk)
 566                 goto out;
 567
 568         sock_init_data(NULL, newsk);
 569
 570         sctp_copy_sock(newsk, sk, asoc);
 571         sock_reset_flag(newsk, SOCK_ZAPPED);
 572
 573         sctp_v4_copy_ip_options(sk, newsk);
 574
 575         newinet = inet_sk(newsk);
 576
 577         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 578
 579         sk_refcnt_debug_inc(newsk);
 580
 581         if (newsk->sk_prot->init(newsk)) {
 582                 sk_common_release(newsk);
 583                 newsk = NULL;
 584         }
 585
 586 out:
 587         return newsk;
 588 }
 589
 590 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
 591 {
 592         /* No address mapping for V4 sockets */
 593         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 594         return sizeof(struct sockaddr_in);
 595 }
 596
 597 /* Dump the v4 addr to the seq file. */
 598 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 599 {
 600         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 601 }
 602
 603 static void sctp_v4_ecn_capable(struct sock *sk)
 604 {
 605         INET_ECN_xmit(sk);
 606 }
 607
 608 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
 609 {
 610         struct net *net = from_timer(net, t, sctp.addr_wq_timer);
 611         struct sctp_sockaddr_entry *addrw, *temp;
 612         struct sctp_sock *sp;
 613
 614         spin_lock_bh(&net->sctp.addr_wq_lock);
 615
 616         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 617                 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
 618                          "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
 619                          addrw->state, addrw);
 620
 621 #if IS_ENABLED(CONFIG_IPV6)
 622                 /* Now we send an ASCONF for each association */
 623                 /* Note. we currently don't handle link local IPv6 addressees */
 624                 if (addrw->a.sa.sa_family == AF_INET6) {
 625                         struct in6_addr *in6;
 626
 627                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 628                             IPV6_ADDR_LINKLOCAL)
 629                                 goto free_next;
 630
 631                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 632                         if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
 633                             addrw->state == SCTP_ADDR_NEW) {
 634                                 unsigned long timeo_val;
 635
 636                                 pr_debug("%s: this is on DAD, trying %d sec "
 637                                          "later\n", __func__,
 638                                          SCTP_ADDRESS_TICK_DELAY);
 639
 640                                 timeo_val = jiffies;
 641                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 642                                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 643                                 break;
 644                         }
 645                 }
 646 #endif
 647                 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
 648                         struct sock *sk;
 649
 650                         sk = sctp_opt2sk(sp);
 651                         /* ignore bound-specific endpoints */
 652                         if (!sctp_is_ep_boundall(sk))
 653                                 continue;
 654                         bh_lock_sock(sk);
 655                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 656                                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
 657                         bh_unlock_sock(sk);
 658                 }
 659 #if IS_ENABLED(CONFIG_IPV6)
 660 free_next:
 661 #endif
 662                 list_del(&addrw->list);
 663                 kfree(addrw);
 664         }
 665         spin_unlock_bh(&net->sctp.addr_wq_lock);
 666 }
 667
 668 static void sctp_free_addr_wq(struct net *net)
 669 {
 670         struct sctp_sockaddr_entry *addrw;
 671         struct sctp_sockaddr_entry *temp;
 672
 673         spin_lock_bh(&net->sctp.addr_wq_lock);
 674         del_timer(&net->sctp.addr_wq_timer);
 675         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 676                 list_del(&addrw->list);
 677                 kfree(addrw);
 678         }
 679         spin_unlock_bh(&net->sctp.addr_wq_lock);
 680 }
 681
 682 /* lookup the entry for the same address in the addr_waitq
 683  * sctp_addr_wq MUST be locked
 684  */
 685 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
 686                                         struct sctp_sockaddr_entry *addr)
 687 {
 688         struct sctp_sockaddr_entry *addrw;
 689
 690         list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
 691                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 692                         continue;
 693                 if (addrw->a.sa.sa_family == AF_INET) {
 694                         if (addrw->a.v4.sin_addr.s_addr ==
 695                             addr->a.v4.sin_addr.s_addr)
 696                                 return addrw;
 697                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 698                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 699                             &addr->a.v6.sin6_addr))
 700                                 return addrw;
 701                 }
 702         }
 703         return NULL;
 704 }
 705
 706 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
 707 {
 708         struct sctp_sockaddr_entry *addrw;
 709         unsigned long timeo_val;
 710
 711         /* first, we check if an opposite message already exist in the queue.
 712          * If we found such message, it is removed.
 713          * This operation is a bit stupid, but the DHCP client attaches the
 714          * new address after a couple of addition and deletion of that address
 715          */
 716
 717         spin_lock_bh(&net->sctp.addr_wq_lock);
 718         /* Offsets existing events in addr_wq */
 719         addrw = sctp_addr_wq_lookup(net, addr);
 720         if (addrw) {
 721                 if (addrw->state != cmd) {
 722                         pr_debug("%s: offsets existing entry for %d, addr:%pISc "
 723                                  "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
 724                                  &net->sctp.addr_waitq);
 725
 726                         list_del(&addrw->list);
 727                         kfree(addrw);
 728                 }
 729                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 730                 return;
 731         }
 732
 733         /* OK, we have to add the new address to the wait queue */
 734         addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 735         if (addrw == NULL) {
 736                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 737                 return;
 738         }
 739         addrw->state = cmd;
 740         list_add_tail(&addrw->list, &net->sctp.addr_waitq);
 741
 742         pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
 743                  __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
 744
 745         if (!timer_pending(&net->sctp.addr_wq_timer)) {
 746                 timeo_val = jiffies;
 747                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 748                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 749         }
 750         spin_unlock_bh(&net->sctp.addr_wq_lock);
 751 }
 752
 753 /* Event handler for inet address addition/deletion events.
 754  * The sctp_local_addr_list needs to be protocted by a spin lock since
 755  * multiple notifiers (say IPv4 and IPv6) may be running at the same
 756  * time and thus corrupt the list.
 757  * The reader side is protected with RCU.
 758  */
 759 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
 760                                void *ptr)
 761 {
 762         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
 763         struct sctp_sockaddr_entry *addr = NULL;
 764         struct sctp_sockaddr_entry *temp;
 765         struct net *net = dev_net(ifa->ifa_dev->dev);
 766         int found = 0;
 767
 768         switch (ev) {
 769         case NETDEV_UP:
 770                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
 771                 if (addr) {
 772                         addr->a.v4.sin_family = AF_INET;
 773                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 774                         addr->valid = 1;
 775                         spin_lock_bh(&net->sctp.local_addr_lock);
 776                         list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
 777                         sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
 778                         spin_unlock_bh(&net->sctp.local_addr_lock);
 779                 }
 780                 break;
 781         case NETDEV_DOWN:
 782                 spin_lock_bh(&net->sctp.local_addr_lock);
 783                 list_for_each_entry_safe(addr, temp,
 784                                         &net->sctp.local_addr_list, list) {
 785                         if (addr->a.sa.sa_family == AF_INET &&
 786                                         addr->a.v4.sin_addr.s_addr ==
 787                                         ifa->ifa_local) {
 788                                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
 789                                 found = 1;
 790                                 addr->valid = 0;
 791                                 list_del_rcu(&addr->list);
 792                                 break;
 793                         }
 794                 }
 795                 spin_unlock_bh(&net->sctp.local_addr_lock);
 796                 if (found)
 797                         kfree_rcu(addr, rcu);
 798                 break;
 799         }
 800
 801         return NOTIFY_DONE;
 802 }
 803
 804 /*
 805  * Initialize the control inode/socket with a control endpoint data
 806  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 807  */
 808 static int sctp_ctl_sock_init(struct net *net)
 809 {
 810         int err;
 811         sa_family_t family = PF_INET;
 812
 813         if (sctp_get_pf_specific(PF_INET6))
 814                 family = PF_INET6;
 815
 816         err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
 817                                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
 818
 819         /* If IPv6 socket could not be created, try the IPv4 socket */
 820         if (err < 0 && family == PF_INET6)
 821                 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
 822                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 823                                            net);
 824
 825         if (err < 0) {
 826                 pr_err("Failed to create the SCTP control socket\n");
 827                 return err;
 828         }
 829         return 0;
 830 }
 831
 832 /* Register address family specific functions. */
 833 int sctp_register_af(struct sctp_af *af)
 834 {
 835         switch (af->sa_family) {
 836         case AF_INET:
 837                 if (sctp_af_v4_specific)
 838                         return 0;
 839                 sctp_af_v4_specific = af;
 840                 break;
 841         case AF_INET6:
 842                 if (sctp_af_v6_specific)
 843                         return 0;
 844                 sctp_af_v6_specific = af;
 845                 break;
 846         default:
 847                 return 0;
 848         }
 849
 850         INIT_LIST_HEAD(&af->list);
 851         list_add_tail(&af->list, &sctp_address_families);
 852         return 1;
 853 }
 854
 855 /* Get the table of functions for manipulating a particular address
 856  * family.
 857  */
 858 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 859 {
 860         switch (family) {
 861         case AF_INET:
 862                 return sctp_af_v4_specific;
 863         case AF_INET6:
 864                 return sctp_af_v6_specific;
 865         default:
 866                 return NULL;
 867         }
 868 }
 869
 870 /* Common code to initialize a AF_INET msg_name. */
 871 static void sctp_inet_msgname(char *msgname, int *addr_len)
 872 {
 873         struct sockaddr_in *sin;
 874
 875         sin = (struct sockaddr_in *)msgname;
 876         *addr_len = sizeof(struct sockaddr_in);
 877         sin->sin_family = AF_INET;
 878         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 879 }
 880
 881 /* Copy the primary address of the peer primary address as the msg_name. */
 882 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 883                                     int *addr_len)
 884 {
 885         struct sockaddr_in *sin, *sinfrom;
 886
 887         if (msgname) {
 888                 struct sctp_association *asoc;
 889
 890                 asoc = event->asoc;
 891                 sctp_inet_msgname(msgname, addr_len);
 892                 sin = (struct sockaddr_in *)msgname;
 893                 sinfrom = &asoc->peer.primary_addr.v4;
 894                 sin->sin_port = htons(asoc->peer.port);
 895                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 896         }
 897 }
 898
 899 /* Initialize and copy out a msgname from an inbound skb. */
 900 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 901 {
 902         if (msgname) {
 903                 struct sctphdr *sh = sctp_hdr(skb);
 904                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 905
 906                 sctp_inet_msgname(msgname, len);
 907                 sin->sin_port = sh->source;
 908                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 909         }
 910 }
 911
 912 /* Do we support this AF? */
 913 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 914 {
 915         /* PF_INET only supports AF_INET addresses. */
 916         return AF_INET == family;
 917 }
 918
 919 /* Address matching with wildcards allowed. */
 920 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 921                               const union sctp_addr *addr2,
 922                               struct sctp_sock *opt)
 923 {
 924         /* PF_INET only supports AF_INET addresses. */
 925         if (addr1->sa.sa_family != addr2->sa.sa_family)
 926                 return 0;
 927         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 928             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 929                 return 1;
 930         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 931                 return 1;
 932
 933         return 0;
 934 }
 935
 936 /* Verify that provided sockaddr looks bindable.  Common verification has
 937  * already been taken care of.
 938  */
 939 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 940 {
 941         return sctp_v4_available(addr, opt);
 942 }
 943
 944 /* Verify that sockaddr looks sendable.  Common verification has already
 945  * been taken care of.
 946  */
 947 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 948 {
 949         return 1;
 950 }
 951
 952 /* Fill in Supported Address Type information for INIT and INIT-ACK
 953  * chunks.  Returns number of addresses supported.
 954  */
 955 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 956                                      __be16 *types)
 957 {
 958         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 959         return 1;
 960 }
 961
 962 /* Wrapper routine that calls the ip transmit routine. */
 963 static inline int sctp_v4_xmit(struct sk_buff *skb,
 964                                struct sctp_transport *transport)
 965 {
 966         struct inet_sock *inet = inet_sk(skb->sk);
 967         __u8 dscp = inet->tos;
 968
 969         pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
 970                  skb->len, &transport->fl.u.ip4.saddr,
 971                  &transport->fl.u.ip4.daddr);
 972
 973         if (transport->dscp & SCTP_DSCP_SET_MASK)
 974                 dscp = transport->dscp & SCTP_DSCP_VAL_MASK;
 975
 976         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
 977                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 978
 979         SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
 980
 981         return __ip_queue_xmit(&inet->sk, skb, &transport->fl, dscp);
 982 }
 983
 984 static struct sctp_af sctp_af_inet;
 985
 986 static struct sctp_pf sctp_pf_inet = {
 987         .event_msgname = sctp_inet_event_msgname,
 988         .skb_msgname   = sctp_inet_skb_msgname,
 989         .af_supported  = sctp_inet_af_supported,
 990         .cmp_addr      = sctp_inet_cmp_addr,
 991         .bind_verify   = sctp_inet_bind_verify,
 992         .send_verify   = sctp_inet_send_verify,
 993         .supported_addrs = sctp_inet_supported_addrs,
 994         .create_accept_sk = sctp_v4_create_accept_sk,
 995         .addr_to_user  = sctp_v4_addr_to_user,
 996         .to_sk_saddr   = sctp_v4_to_sk_saddr,
 997         .to_sk_daddr   = sctp_v4_to_sk_daddr,
 998         .copy_ip_options = sctp_v4_copy_ip_options,
 999         .af            = &sctp_af_inet
1000 };
1001
1002 /* Notifier for inetaddr addition/deletion events.  */
1003 static struct notifier_block sctp_inetaddr_notifier = {
1004         .notifier_call = sctp_inetaddr_event,
1005 };
1006
1007 /* Socket operations.  */
1008 static const struct proto_ops inet_seqpacket_ops = {
1009         .family            = PF_INET,
1010         .owner             = THIS_MODULE,
1011         .release           = inet_release,      /* Needs to be wrapped... */
1012         .bind              = inet_bind,
1013         .connect           = sctp_inet_connect,
1014         .socketpair        = sock_no_socketpair,
1015         .accept            = inet_accept,
1016         .getname           = inet_getname,      /* Semantics are different.  */
1017         .poll              = sctp_poll,
1018         .ioctl             = inet_ioctl,
1019         .gettstamp         = sock_gettstamp,
1020         .listen            = sctp_inet_listen,
1021         .shutdown          = inet_shutdown,     /* Looks harmless.  */
1022         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1023         .getsockopt        = sock_common_getsockopt,
1024         .sendmsg           = inet_sendmsg,
1025         .recvmsg           = inet_recvmsg,
1026         .mmap              = sock_no_mmap,
1027         .sendpage          = sock_no_sendpage,
1028 #ifdef CONFIG_COMPAT
1029         .compat_setsockopt = compat_sock_common_setsockopt,
1030         .compat_getsockopt = compat_sock_common_getsockopt,
1031 #endif
1032 };
1033
1034 /* Registration with AF_INET family.  */
1035 static struct inet_protosw sctp_seqpacket_protosw = {
1036         .type       = SOCK_SEQPACKET,
1037         .protocol   = IPPROTO_SCTP,
1038         .prot       = &sctp_prot,
1039         .ops        = &inet_seqpacket_ops,
1040         .flags      = SCTP_PROTOSW_FLAG
1041 };
1042 static struct inet_protosw sctp_stream_protosw = {
1043         .type       = SOCK_STREAM,
1044         .protocol   = IPPROTO_SCTP,
1045         .prot       = &sctp_prot,
1046         .ops        = &inet_seqpacket_ops,
1047         .flags      = SCTP_PROTOSW_FLAG
1048 };
1049
1050 /* Register with IP layer.  */
1051 static const struct net_protocol sctp_protocol = {
1052         .handler     = sctp_rcv,
1053         .err_handler = sctp_v4_err,
1054         .no_policy   = 1,
1055         .netns_ok    = 1,
1056         .icmp_strict_tag_validation = 1,
1057 };
1058
1059 /* IPv4 address related functions.  */
1060 static struct sctp_af sctp_af_inet = {
1061         .sa_family         = AF_INET,
1062         .sctp_xmit         = sctp_v4_xmit,
1063         .setsockopt        = ip_setsockopt,
1064         .getsockopt        = ip_getsockopt,
1065         .get_dst           = sctp_v4_get_dst,
1066         .get_saddr         = sctp_v4_get_saddr,
1067         .copy_addrlist     = sctp_v4_copy_addrlist,
1068         .from_skb          = sctp_v4_from_skb,
1069         .from_sk           = sctp_v4_from_sk,
1070         .from_addr_param   = sctp_v4_from_addr_param,
1071         .to_addr_param     = sctp_v4_to_addr_param,
1072         .cmp_addr          = sctp_v4_cmp_addr,
1073         .addr_valid        = sctp_v4_addr_valid,
1074         .inaddr_any        = sctp_v4_inaddr_any,
1075         .is_any            = sctp_v4_is_any,
1076         .available         = sctp_v4_available,
1077         .scope             = sctp_v4_scope,
1078         .skb_iif           = sctp_v4_skb_iif,
1079         .is_ce             = sctp_v4_is_ce,
1080         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1081         .ecn_capable       = sctp_v4_ecn_capable,
1082         .net_header_len    = sizeof(struct iphdr),
1083         .sockaddr_len      = sizeof(struct sockaddr_in),
1084         .ip_options_len    = sctp_v4_ip_options_len,
1085 #ifdef CONFIG_COMPAT
1086         .compat_setsockopt = compat_ip_setsockopt,
1087         .compat_getsockopt = compat_ip_getsockopt,
1088 #endif
1089 };
1090
1091 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1092 {
1093         switch (family) {
1094         case PF_INET:
1095                 return sctp_pf_inet_specific;
1096         case PF_INET6:
1097                 return sctp_pf_inet6_specific;
1098         default:
1099                 return NULL;
1100         }
1101 }
1102
1103 /* Register the PF specific function table.  */
1104 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1105 {
1106         switch (family) {
1107         case PF_INET:
1108                 if (sctp_pf_inet_specific)
1109                         return 0;
1110                 sctp_pf_inet_specific = pf;
1111                 break;
1112         case PF_INET6:
1113                 if (sctp_pf_inet6_specific)
1114                         return 0;
1115                 sctp_pf_inet6_specific = pf;
1116                 break;
1117         default:
1118                 return 0;
1119         }
1120         return 1;
1121 }
1122
1123 static inline int init_sctp_mibs(struct net *net)
1124 {
1125         net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1126         if (!net->sctp.sctp_statistics)
1127                 return -ENOMEM;
1128         return 0;
1129 }
1130
1131 static inline void cleanup_sctp_mibs(struct net *net)
1132 {
1133         free_percpu(net->sctp.sctp_statistics);
1134 }
1135
1136 static void sctp_v4_pf_init(void)
1137 {
1138         /* Initialize the SCTP specific PF functions. */
1139         sctp_register_pf(&sctp_pf_inet, PF_INET);
1140         sctp_register_af(&sctp_af_inet);
1141 }
1142
1143 static void sctp_v4_pf_exit(void)
1144 {
1145         list_del(&sctp_af_inet.list);
1146 }
1147
1148 static int sctp_v4_protosw_init(void)
1149 {
1150         int rc;
1151
1152         rc = proto_register(&sctp_prot, 1);
1153         if (rc)
1154                 return rc;
1155
1156         /* Register SCTP(UDP and TCP style) with socket layer.  */
1157         inet_register_protosw(&sctp_seqpacket_protosw);
1158         inet_register_protosw(&sctp_stream_protosw);
1159
1160         return 0;
1161 }
1162
1163 static void sctp_v4_protosw_exit(void)
1164 {
1165         inet_unregister_protosw(&sctp_stream_protosw);
1166         inet_unregister_protosw(&sctp_seqpacket_protosw);
1167         proto_unregister(&sctp_prot);
1168 }
1169
1170 static int sctp_v4_add_protocol(void)
1171 {
1172         /* Register notifier for inet address additions/deletions. */
1173         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1174
1175         /* Register SCTP with inet layer.  */
1176         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1177                 return -EAGAIN;
1178
1179         return 0;
1180 }
1181
1182 static void sctp_v4_del_protocol(void)
1183 {
1184         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1185         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1186 }
1187
1188 static int __net_init sctp_defaults_init(struct net *net)
1189 {
1190         int status;
1191
1192         /*
1193          * 14. Suggested SCTP Protocol Parameter Values
1194          */
1195         /* The following protocol parameters are RECOMMENDED:  */
1196         /* RTO.Initial              - 3  seconds */
1197         net->sctp.rto_initial                   = SCTP_RTO_INITIAL;
1198         /* RTO.Min                  - 1  second */
1199         net->sctp.rto_min                       = SCTP_RTO_MIN;
1200         /* RTO.Max                 -  60 seconds */
1201         net->sctp.rto_max                       = SCTP_RTO_MAX;
1202         /* RTO.Alpha                - 1/8 */
1203         net->sctp.rto_alpha                     = SCTP_RTO_ALPHA;
1204         /* RTO.Beta                 - 1/4 */
1205         net->sctp.rto_beta                      = SCTP_RTO_BETA;
1206
1207         /* Valid.Cookie.Life        - 60  seconds */
1208         net->sctp.valid_cookie_life             = SCTP_DEFAULT_COOKIE_LIFE;
1209
1210         /* Whether Cookie Preservative is enabled(1) or not(0) */
1211         net->sctp.cookie_preserve_enable        = 1;
1212
1213         /* Default sctp sockets to use md5 as their hmac alg */
1214 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1215         net->sctp.sctp_hmac_alg                 = "md5";
1216 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1217         net->sctp.sctp_hmac_alg                 = "sha1";
1218 #else
1219         net->sctp.sctp_hmac_alg                 = NULL;
1220 #endif
1221
1222         /* Max.Burst                - 4 */
1223         net->sctp.max_burst                     = SCTP_DEFAULT_MAX_BURST;
1224
1225         /* Disable of Primary Path Switchover by default */
1226         net->sctp.ps_retrans = SCTP_PS_RETRANS_MAX;
1227
1228         /* Enable pf state by default */
1229         net->sctp.pf_enable = 1;
1230
1231         /* Ignore pf exposure feature by default */
1232         net->sctp.pf_expose = SCTP_PF_EXPOSE_UNSET;
1233
1234         /* Association.Max.Retrans  - 10 attempts
1235          * Path.Max.Retrans         - 5  attempts (per destination address)
1236          * Max.Init.Retransmits     - 8  attempts
1237          */
1238         net->sctp.max_retrans_association       = 10;
1239         net->sctp.max_retrans_path              = 5;
1240         net->sctp.max_retrans_init              = 8;
1241
1242         /* Sendbuffer growth        - do per-socket accounting */
1243         net->sctp.sndbuf_policy                 = 0;
1244
1245         /* Rcvbuffer growth         - do per-socket accounting */
1246         net->sctp.rcvbuf_policy                 = 0;
1247
1248         /* HB.interval              - 30 seconds */
1249         net->sctp.hb_interval                   = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1250
1251         /* delayed SACK timeout */
1252         net->sctp.sack_timeout                  = SCTP_DEFAULT_TIMEOUT_SACK;
1253
1254         /* Disable ADDIP by default. */
1255         net->sctp.addip_enable = 0;
1256         net->sctp.addip_noauth = 0;
1257         net->sctp.default_auto_asconf = 0;
1258
1259         /* Enable PR-SCTP by default. */
1260         net->sctp.prsctp_enable = 1;
1261
1262         /* Disable RECONF by default. */
1263         net->sctp.reconf_enable = 0;
1264
1265         /* Disable AUTH by default. */
1266         net->sctp.auth_enable = 0;
1267
1268         /* Enable ECN by default. */
1269         net->sctp.ecn_enable = 1;
1270
1271         /* Set SCOPE policy to enabled */
1272         net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1273
1274         /* Set the default rwnd update threshold */
1275         net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1276
1277         /* Initialize maximum autoclose timeout. */
1278         net->sctp.max_autoclose         = INT_MAX / HZ;
1279
1280         status = sctp_sysctl_net_register(net);
1281         if (status)
1282                 goto err_sysctl_register;
1283
1284         /* Allocate and initialise sctp mibs.  */
1285         status = init_sctp_mibs(net);
1286         if (status)
1287                 goto err_init_mibs;
1288
1289 #ifdef CONFIG_PROC_FS
1290         /* Initialize proc fs directory.  */
1291         status = sctp_proc_init(net);
1292         if (status)
1293                 goto err_init_proc;
1294 #endif
1295
1296         sctp_dbg_objcnt_init(net);
1297
1298         /* Initialize the local address list. */
1299         INIT_LIST_HEAD(&net->sctp.local_addr_list);
1300         spin_lock_init(&net->sctp.local_addr_lock);
1301         sctp_get_local_addr_list(net);
1302
1303         /* Initialize the address event list */
1304         INIT_LIST_HEAD(&net->sctp.addr_waitq);
1305         INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1306         spin_lock_init(&net->sctp.addr_wq_lock);
1307         net->sctp.addr_wq_timer.expires = 0;
1308         timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1309
1310         return 0;
1311
1312 #ifdef CONFIG_PROC_FS
1313 err_init_proc:
1314         cleanup_sctp_mibs(net);
1315 #endif
1316 err_init_mibs:
1317         sctp_sysctl_net_unregister(net);
1318 err_sysctl_register:
1319         return status;
1320 }
1321
1322 static void __net_exit sctp_defaults_exit(struct net *net)
1323 {
1324         /* Free the local address list */
1325         sctp_free_addr_wq(net);
1326         sctp_free_local_addr_list(net);
1327
1328 #ifdef CONFIG_PROC_FS
1329         remove_proc_subtree("sctp", net->proc_net);
1330         net->sctp.proc_net_sctp = NULL;
1331 #endif
1332         cleanup_sctp_mibs(net);
1333         sctp_sysctl_net_unregister(net);
1334 }
1335
1336 static struct pernet_operations sctp_defaults_ops = {
1337         .init = sctp_defaults_init,
1338         .exit = sctp_defaults_exit,
1339 };
1340
1341 static int __net_init sctp_ctrlsock_init(struct net *net)
1342 {
1343         int status;
1344
1345         /* Initialize the control inode/socket for handling OOTB packets.  */
1346         status = sctp_ctl_sock_init(net);
1347         if (status)
1348                 pr_err("Failed to initialize the SCTP control sock\n");
1349
1350         return status;
1351 }
1352
1353 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1354 {
1355         /* Free the control endpoint.  */
1356         inet_ctl_sock_destroy(net->sctp.ctl_sock);
1357 }
1358
1359 static struct pernet_operations sctp_ctrlsock_ops = {
1360         .init = sctp_ctrlsock_init,
1361         .exit = sctp_ctrlsock_exit,
1362 };
1363
1364 /* Initialize the universe into something sensible.  */
1365 static __init int sctp_init(void)
1366 {
1367         int i;
1368         int status = -EINVAL;
1369         unsigned long goal;
1370         unsigned long limit;
1371         unsigned long nr_pages = totalram_pages();
1372         int max_share;
1373         int order;
1374         int num_entries;
1375         int max_entry_order;
1376
1377         sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1378
1379         /* Allocate bind_bucket and chunk caches. */
1380         status = -ENOBUFS;
1381         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1382                                                sizeof(struct sctp_bind_bucket),
1383                                                0, SLAB_HWCACHE_ALIGN,
1384                                                NULL);
1385         if (!sctp_bucket_cachep)
1386                 goto out;
1387
1388         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1389                                                sizeof(struct sctp_chunk),
1390                                                0, SLAB_HWCACHE_ALIGN,
1391                                                NULL);
1392         if (!sctp_chunk_cachep)
1393                 goto err_chunk_cachep;
1394
1395         status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1396         if (status)
1397                 goto err_percpu_counter_init;
1398
1399         /* Implementation specific variables. */
1400
1401         /* Initialize default stream count setup information. */
1402         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1403         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1404
1405         /* Initialize handle used for association ids. */
1406         idr_init(&sctp_assocs_id);
1407
1408         limit = nr_free_buffer_pages() / 8;
1409         limit = max(limit, 128UL);
1410         sysctl_sctp_mem[0] = limit / 4 * 3;
1411         sysctl_sctp_mem[1] = limit;
1412         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1413
1414         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1415         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1416         max_share = min(4UL*1024*1024, limit);
1417
1418         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1419         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1420         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1421
1422         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1423         sysctl_sctp_wmem[1] = 16*1024;
1424         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1425
1426         /* Size and allocate the association hash table.
1427          * The methodology is similar to that of the tcp hash tables.
1428          * Though not identical.  Start by getting a goal size
1429          */
1430         if (nr_pages >= (128 * 1024))
1431                 goal = nr_pages >> (22 - PAGE_SHIFT);
1432         else
1433                 goal = nr_pages >> (24 - PAGE_SHIFT);
1434
1435         /* Then compute the page order for said goal */
1436         order = get_order(goal);
1437
1438         /* Now compute the required page order for the maximum sized table we
1439          * want to create
1440          */
1441         max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1442                                     sizeof(struct sctp_bind_hashbucket));
1443
1444         /* Limit the page order by that maximum hash table size */
1445         order = min(order, max_entry_order);
1446
1447         /* Allocate and initialize the endpoint hash table.  */
1448         sctp_ep_hashsize = 64;
1449         sctp_ep_hashtable =
1450                 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1451         if (!sctp_ep_hashtable) {
1452                 pr_err("Failed endpoint_hash alloc\n");
1453                 status = -ENOMEM;
1454                 goto err_ehash_alloc;
1455         }
1456         for (i = 0; i < sctp_ep_hashsize; i++) {
1457                 rwlock_init(&sctp_ep_hashtable[i].lock);
1458                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1459         }
1460
1461         /* Allocate and initialize the SCTP port hash table.
1462          * Note that order is initalized to start at the max sized
1463          * table we want to support.  If we can't get that many pages
1464          * reduce the order and try again
1465          */
1466         do {
1467                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1468                         __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1469         } while (!sctp_port_hashtable && --order > 0);
1470
1471         if (!sctp_port_hashtable) {
1472                 pr_err("Failed bind hash alloc\n");
1473                 status = -ENOMEM;
1474                 goto err_bhash_alloc;
1475         }
1476
1477         /* Now compute the number of entries that will fit in the
1478          * port hash space we allocated
1479          */
1480         num_entries = (1UL << order) * PAGE_SIZE /
1481                       sizeof(struct sctp_bind_hashbucket);
1482
1483         /* And finish by rounding it down to the nearest power of two
1484          * this wastes some memory of course, but its needed because
1485          * the hash function operates based on the assumption that
1486          * that the number of entries is a power of two
1487          */
1488         sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1489
1490         for (i = 0; i < sctp_port_hashsize; i++) {
1491                 spin_lock_init(&sctp_port_hashtable[i].lock);
1492                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1493         }
1494
1495         status = sctp_transport_hashtable_init();
1496         if (status)
1497                 goto err_thash_alloc;
1498
1499         pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1500                 num_entries);
1501
1502         sctp_sysctl_register();
1503
1504         INIT_LIST_HEAD(&sctp_address_families);
1505         sctp_v4_pf_init();
1506         sctp_v6_pf_init();
1507         sctp_sched_ops_init();
1508
1509         status = register_pernet_subsys(&sctp_defaults_ops);
1510         if (status)
1511                 goto err_register_defaults;
1512
1513         status = sctp_v4_protosw_init();
1514         if (status)
1515                 goto err_protosw_init;
1516
1517         status = sctp_v6_protosw_init();
1518         if (status)
1519                 goto err_v6_protosw_init;
1520
1521         status = register_pernet_subsys(&sctp_ctrlsock_ops);
1522         if (status)
1523                 goto err_register_ctrlsock;
1524
1525         status = sctp_v4_add_protocol();
1526         if (status)
1527                 goto err_add_protocol;
1528
1529         /* Register SCTP with inet6 layer.  */
1530         status = sctp_v6_add_protocol();
1531         if (status)
1532                 goto err_v6_add_protocol;
1533
1534         if (sctp_offload_init() < 0)
1535                 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1536
1537 out:
1538         return status;
1539 err_v6_add_protocol:
1540         sctp_v4_del_protocol();
1541 err_add_protocol:
1542         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1543 err_register_ctrlsock:
1544         sctp_v6_protosw_exit();
1545 err_v6_protosw_init:
1546         sctp_v4_protosw_exit();
1547 err_protosw_init:
1548         unregister_pernet_subsys(&sctp_defaults_ops);
1549 err_register_defaults:
1550         sctp_v4_pf_exit();
1551         sctp_v6_pf_exit();
1552         sctp_sysctl_unregister();
1553         free_pages((unsigned long)sctp_port_hashtable,
1554                    get_order(sctp_port_hashsize *
1555                              sizeof(struct sctp_bind_hashbucket)));
1556 err_bhash_alloc:
1557         sctp_transport_hashtable_destroy();
1558 err_thash_alloc:
1559         kfree(sctp_ep_hashtable);
1560 err_ehash_alloc:
1561         percpu_counter_destroy(&sctp_sockets_allocated);
1562 err_percpu_counter_init:
1563         kmem_cache_destroy(sctp_chunk_cachep);
1564 err_chunk_cachep:
1565         kmem_cache_destroy(sctp_bucket_cachep);
1566         goto out;
1567 }
1568
1569 /* Exit handler for the SCTP protocol.  */
1570 static __exit void sctp_exit(void)
1571 {
1572         /* BUG.  This should probably do something useful like clean
1573          * up all the remaining associations and all that memory.
1574          */
1575
1576         /* Unregister with inet6/inet layers. */
1577         sctp_v6_del_protocol();
1578         sctp_v4_del_protocol();
1579
1580         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1581
1582         /* Free protosw registrations */
1583         sctp_v6_protosw_exit();
1584         sctp_v4_protosw_exit();
1585
1586         unregister_pernet_subsys(&sctp_defaults_ops);
1587
1588         /* Unregister with socket layer. */
1589         sctp_v6_pf_exit();
1590         sctp_v4_pf_exit();
1591
1592         sctp_sysctl_unregister();
1593
1594         free_pages((unsigned long)sctp_port_hashtable,
1595                    get_order(sctp_port_hashsize *
1596                              sizeof(struct sctp_bind_hashbucket)));
1597         kfree(sctp_ep_hashtable);
1598         sctp_transport_hashtable_destroy();
1599
1600         percpu_counter_destroy(&sctp_sockets_allocated);
1601
1602         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1603
1604         kmem_cache_destroy(sctp_chunk_cachep);
1605         kmem_cache_destroy(sctp_bucket_cachep);
1606 }
1607
1608 module_init(sctp_init);
1609 module_exit(sctp_exit);
1610
1611 /*
1612  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1613  */
1614 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1615 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1616 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1617 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1618 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1619 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1620 MODULE_LICENSE("GPL");