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 flowi _fl;
 413         struct flowi4 *fl4 = &_fl.u.ip4;
 414         struct sctp_bind_addr *bp;
 415         struct sctp_sockaddr_entry *laddr;
 416         struct dst_entry *dst = NULL;
 417         union sctp_addr *daddr = &t->ipaddr;
 418         union sctp_addr dst_saddr;
 419         __u8 tos = inet_sk(sk)->tos;
 420
 421         if (t->dscp & SCTP_DSCP_SET_MASK)
 422                 tos = t->dscp & SCTP_DSCP_VAL_MASK;
 423         memset(&_fl, 0x0, sizeof(_fl));
 424         fl4->daddr  = daddr->v4.sin_addr.s_addr;
 425         fl4->fl4_dport = daddr->v4.sin_port;
 426         fl4->flowi4_proto = IPPROTO_SCTP;
 427         if (asoc) {
 428                 fl4->flowi4_tos = RT_CONN_FLAGS_TOS(asoc->base.sk, tos);
 429                 fl4->flowi4_oif = asoc->base.sk->sk_bound_dev_if;
 430                 fl4->fl4_sport = htons(asoc->base.bind_addr.port);
 431         }
 432         if (saddr) {
 433                 fl4->saddr = saddr->v4.sin_addr.s_addr;
 434                 if (!fl4->fl4_sport)
 435                         fl4->fl4_sport = saddr->v4.sin_port;
 436         }
 437
 438         pr_debug("%s: dst:%pI4, src:%pI4 - ", __func__, &fl4->daddr,
 439                  &fl4->saddr);
 440
 441         rt = ip_route_output_key(sock_net(sk), fl4);
 442         if (!IS_ERR(rt)) {
 443                 dst = &rt->dst;
 444                 t->dst = dst;
 445                 memcpy(fl, &_fl, sizeof(_fl));
 446         }
 447
 448         /* If there is no association or if a source address is passed, no
 449          * more validation is required.
 450          */
 451         if (!asoc || saddr)
 452                 goto out;
 453
 454         bp = &asoc->base.bind_addr;
 455
 456         if (dst) {
 457                 /* Walk through the bind address list and look for a bind
 458                  * address that matches the source address of the returned dst.
 459                  */
 460                 sctp_v4_dst_saddr(&dst_saddr, fl4, htons(bp->port));
 461                 rcu_read_lock();
 462                 list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 463                         if (!laddr->valid || (laddr->state == SCTP_ADDR_DEL) ||
 464                             (laddr->state != SCTP_ADDR_SRC &&
 465                             !asoc->src_out_of_asoc_ok))
 466                                 continue;
 467                         if (sctp_v4_cmp_addr(&dst_saddr, &laddr->a))
 468                                 goto out_unlock;
 469                 }
 470                 rcu_read_unlock();
 471
 472                 /* None of the bound addresses match the source address of the
 473                  * dst. So release it.
 474                  */
 475                 dst_release(dst);
 476                 dst = NULL;
 477         }
 478
 479         /* Walk through the bind address list and try to get a dst that
 480          * matches a bind address as the source address.
 481          */
 482         rcu_read_lock();
 483         list_for_each_entry_rcu(laddr, &bp->address_list, list) {
 484                 struct net_device *odev;
 485
 486                 if (!laddr->valid)
 487                         continue;
 488                 if (laddr->state != SCTP_ADDR_SRC ||
 489                     AF_INET != laddr->a.sa.sa_family)
 490                         continue;
 491
 492                 fl4->fl4_sport = laddr->a.v4.sin_port;
 493                 flowi4_update_output(fl4,
 494                                      asoc->base.sk->sk_bound_dev_if,
 495                                      RT_CONN_FLAGS_TOS(asoc->base.sk, tos),
 496                                      daddr->v4.sin_addr.s_addr,
 497                                      laddr->a.v4.sin_addr.s_addr);
 498
 499                 rt = ip_route_output_key(sock_net(sk), fl4);
 500                 if (IS_ERR(rt))
 501                         continue;
 502
 503                 /* Ensure the src address belongs to the output
 504                  * interface.
 505                  */
 506                 odev = __ip_dev_find(sock_net(sk), laddr->a.v4.sin_addr.s_addr,
 507                                      false);
 508                 if (!odev || odev->ifindex != fl4->flowi4_oif) {
 509                         if (!dst) {
 510                                 dst = &rt->dst;
 511                                 t->dst = dst;
 512                                 memcpy(fl, &_fl, sizeof(_fl));
 513                         } else {
 514                                 dst_release(&rt->dst);
 515                         }
 516                         continue;
 517                 }
 518
 519                 dst_release(dst);
 520                 dst = &rt->dst;
 521                 t->dst = dst;
 522                 memcpy(fl, &_fl, sizeof(_fl));
 523                 break;
 524         }
 525
 526 out_unlock:
 527         rcu_read_unlock();
 528 out:
 529         if (dst) {
 530                 pr_debug("rt_dst:%pI4, rt_src:%pI4\n",
 531                          &fl->u.ip4.daddr, &fl->u.ip4.saddr);
 532         } else {
 533                 t->dst = NULL;
 534                 pr_debug("no route\n");
 535         }
 536 }
 537
 538 /* For v4, the source address is cached in the route entry(dst). So no need
 539  * to cache it separately and hence this is an empty routine.
 540  */
 541 static void sctp_v4_get_saddr(struct sctp_sock *sk,
 542                               struct sctp_transport *t,
 543                               struct flowi *fl)
 544 {
 545         union sctp_addr *saddr = &t->saddr;
 546         struct rtable *rt = (struct rtable *)t->dst;
 547
 548         if (rt) {
 549                 saddr->v4.sin_family = AF_INET;
 550                 saddr->v4.sin_addr.s_addr = fl->u.ip4.saddr;
 551         }
 552 }
 553
 554 /* What interface did this skb arrive on? */
 555 static int sctp_v4_skb_iif(const struct sk_buff *skb)
 556 {
 557         return inet_iif(skb);
 558 }
 559
 560 /* Was this packet marked by Explicit Congestion Notification? */
 561 static int sctp_v4_is_ce(const struct sk_buff *skb)
 562 {
 563         return INET_ECN_is_ce(ip_hdr(skb)->tos);
 564 }
 565
 566 /* Create and initialize a new sk for the socket returned by accept(). */
 567 static struct sock *sctp_v4_create_accept_sk(struct sock *sk,
 568                                              struct sctp_association *asoc,
 569                                              bool kern)
 570 {
 571         struct sock *newsk = sk_alloc(sock_net(sk), PF_INET, GFP_KERNEL,
 572                         sk->sk_prot, kern);
 573         struct inet_sock *newinet;
 574
 575         if (!newsk)
 576                 goto out;
 577
 578         sock_init_data(NULL, newsk);
 579
 580         sctp_copy_sock(newsk, sk, asoc);
 581         sock_reset_flag(newsk, SOCK_ZAPPED);
 582
 583         sctp_v4_copy_ip_options(sk, newsk);
 584
 585         newinet = inet_sk(newsk);
 586
 587         newinet->inet_daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
 588
 589         sk_refcnt_debug_inc(newsk);
 590
 591         if (newsk->sk_prot->init(newsk)) {
 592                 sk_common_release(newsk);
 593                 newsk = NULL;
 594         }
 595
 596 out:
 597         return newsk;
 598 }
 599
 600 static int sctp_v4_addr_to_user(struct sctp_sock *sp, union sctp_addr *addr)
 601 {
 602         /* No address mapping for V4 sockets */
 603         memset(addr->v4.sin_zero, 0, sizeof(addr->v4.sin_zero));
 604         return sizeof(struct sockaddr_in);
 605 }
 606
 607 /* Dump the v4 addr to the seq file. */
 608 static void sctp_v4_seq_dump_addr(struct seq_file *seq, union sctp_addr *addr)
 609 {
 610         seq_printf(seq, "%pI4 ", &addr->v4.sin_addr);
 611 }
 612
 613 static void sctp_v4_ecn_capable(struct sock *sk)
 614 {
 615         INET_ECN_xmit(sk);
 616 }
 617
 618 static void sctp_addr_wq_timeout_handler(struct timer_list *t)
 619 {
 620         struct net *net = from_timer(net, t, sctp.addr_wq_timer);
 621         struct sctp_sockaddr_entry *addrw, *temp;
 622         struct sctp_sock *sp;
 623
 624         spin_lock_bh(&net->sctp.addr_wq_lock);
 625
 626         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 627                 pr_debug("%s: the first ent in wq:%p is addr:%pISc for cmd:%d at "
 628                          "entry:%p\n", __func__, &net->sctp.addr_waitq, &addrw->a.sa,
 629                          addrw->state, addrw);
 630
 631 #if IS_ENABLED(CONFIG_IPV6)
 632                 /* Now we send an ASCONF for each association */
 633                 /* Note. we currently don't handle link local IPv6 addressees */
 634                 if (addrw->a.sa.sa_family == AF_INET6) {
 635                         struct in6_addr *in6;
 636
 637                         if (ipv6_addr_type(&addrw->a.v6.sin6_addr) &
 638                             IPV6_ADDR_LINKLOCAL)
 639                                 goto free_next;
 640
 641                         in6 = (struct in6_addr *)&addrw->a.v6.sin6_addr;
 642                         if (ipv6_chk_addr(net, in6, NULL, 0) == 0 &&
 643                             addrw->state == SCTP_ADDR_NEW) {
 644                                 unsigned long timeo_val;
 645
 646                                 pr_debug("%s: this is on DAD, trying %d sec "
 647                                          "later\n", __func__,
 648                                          SCTP_ADDRESS_TICK_DELAY);
 649
 650                                 timeo_val = jiffies;
 651                                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 652                                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 653                                 break;
 654                         }
 655                 }
 656 #endif
 657                 list_for_each_entry(sp, &net->sctp.auto_asconf_splist, auto_asconf_list) {
 658                         struct sock *sk;
 659
 660                         sk = sctp_opt2sk(sp);
 661                         /* ignore bound-specific endpoints */
 662                         if (!sctp_is_ep_boundall(sk))
 663                                 continue;
 664                         bh_lock_sock(sk);
 665                         if (sctp_asconf_mgmt(sp, addrw) < 0)
 666                                 pr_debug("%s: sctp_asconf_mgmt failed\n", __func__);
 667                         bh_unlock_sock(sk);
 668                 }
 669 #if IS_ENABLED(CONFIG_IPV6)
 670 free_next:
 671 #endif
 672                 list_del(&addrw->list);
 673                 kfree(addrw);
 674         }
 675         spin_unlock_bh(&net->sctp.addr_wq_lock);
 676 }
 677
 678 static void sctp_free_addr_wq(struct net *net)
 679 {
 680         struct sctp_sockaddr_entry *addrw;
 681         struct sctp_sockaddr_entry *temp;
 682
 683         spin_lock_bh(&net->sctp.addr_wq_lock);
 684         del_timer(&net->sctp.addr_wq_timer);
 685         list_for_each_entry_safe(addrw, temp, &net->sctp.addr_waitq, list) {
 686                 list_del(&addrw->list);
 687                 kfree(addrw);
 688         }
 689         spin_unlock_bh(&net->sctp.addr_wq_lock);
 690 }
 691
 692 /* lookup the entry for the same address in the addr_waitq
 693  * sctp_addr_wq MUST be locked
 694  */
 695 static struct sctp_sockaddr_entry *sctp_addr_wq_lookup(struct net *net,
 696                                         struct sctp_sockaddr_entry *addr)
 697 {
 698         struct sctp_sockaddr_entry *addrw;
 699
 700         list_for_each_entry(addrw, &net->sctp.addr_waitq, list) {
 701                 if (addrw->a.sa.sa_family != addr->a.sa.sa_family)
 702                         continue;
 703                 if (addrw->a.sa.sa_family == AF_INET) {
 704                         if (addrw->a.v4.sin_addr.s_addr ==
 705                             addr->a.v4.sin_addr.s_addr)
 706                                 return addrw;
 707                 } else if (addrw->a.sa.sa_family == AF_INET6) {
 708                         if (ipv6_addr_equal(&addrw->a.v6.sin6_addr,
 709                             &addr->a.v6.sin6_addr))
 710                                 return addrw;
 711                 }
 712         }
 713         return NULL;
 714 }
 715
 716 void sctp_addr_wq_mgmt(struct net *net, struct sctp_sockaddr_entry *addr, int cmd)
 717 {
 718         struct sctp_sockaddr_entry *addrw;
 719         unsigned long timeo_val;
 720
 721         /* first, we check if an opposite message already exist in the queue.
 722          * If we found such message, it is removed.
 723          * This operation is a bit stupid, but the DHCP client attaches the
 724          * new address after a couple of addition and deletion of that address
 725          */
 726
 727         spin_lock_bh(&net->sctp.addr_wq_lock);
 728         /* Offsets existing events in addr_wq */
 729         addrw = sctp_addr_wq_lookup(net, addr);
 730         if (addrw) {
 731                 if (addrw->state != cmd) {
 732                         pr_debug("%s: offsets existing entry for %d, addr:%pISc "
 733                                  "in wq:%p\n", __func__, addrw->state, &addrw->a.sa,
 734                                  &net->sctp.addr_waitq);
 735
 736                         list_del(&addrw->list);
 737                         kfree(addrw);
 738                 }
 739                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 740                 return;
 741         }
 742
 743         /* OK, we have to add the new address to the wait queue */
 744         addrw = kmemdup(addr, sizeof(struct sctp_sockaddr_entry), GFP_ATOMIC);
 745         if (addrw == NULL) {
 746                 spin_unlock_bh(&net->sctp.addr_wq_lock);
 747                 return;
 748         }
 749         addrw->state = cmd;
 750         list_add_tail(&addrw->list, &net->sctp.addr_waitq);
 751
 752         pr_debug("%s: add new entry for cmd:%d, addr:%pISc in wq:%p\n",
 753                  __func__, addrw->state, &addrw->a.sa, &net->sctp.addr_waitq);
 754
 755         if (!timer_pending(&net->sctp.addr_wq_timer)) {
 756                 timeo_val = jiffies;
 757                 timeo_val += msecs_to_jiffies(SCTP_ADDRESS_TICK_DELAY);
 758                 mod_timer(&net->sctp.addr_wq_timer, timeo_val);
 759         }
 760         spin_unlock_bh(&net->sctp.addr_wq_lock);
 761 }
 762
 763 /* Event handler for inet address addition/deletion events.
 764  * The sctp_local_addr_list needs to be protocted by a spin lock since
 765  * multiple notifiers (say IPv4 and IPv6) may be running at the same
 766  * time and thus corrupt the list.
 767  * The reader side is protected with RCU.
 768  */
 769 static int sctp_inetaddr_event(struct notifier_block *this, unsigned long ev,
 770                                void *ptr)
 771 {
 772         struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
 773         struct sctp_sockaddr_entry *addr = NULL;
 774         struct sctp_sockaddr_entry *temp;
 775         struct net *net = dev_net(ifa->ifa_dev->dev);
 776         int found = 0;
 777
 778         switch (ev) {
 779         case NETDEV_UP:
 780                 addr = kzalloc(sizeof(*addr), GFP_ATOMIC);
 781                 if (addr) {
 782                         addr->a.v4.sin_family = AF_INET;
 783                         addr->a.v4.sin_addr.s_addr = ifa->ifa_local;
 784                         addr->valid = 1;
 785                         spin_lock_bh(&net->sctp.local_addr_lock);
 786                         list_add_tail_rcu(&addr->list, &net->sctp.local_addr_list);
 787                         sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_NEW);
 788                         spin_unlock_bh(&net->sctp.local_addr_lock);
 789                 }
 790                 break;
 791         case NETDEV_DOWN:
 792                 spin_lock_bh(&net->sctp.local_addr_lock);
 793                 list_for_each_entry_safe(addr, temp,
 794                                         &net->sctp.local_addr_list, list) {
 795                         if (addr->a.sa.sa_family == AF_INET &&
 796                                         addr->a.v4.sin_addr.s_addr ==
 797                                         ifa->ifa_local) {
 798                                 sctp_addr_wq_mgmt(net, addr, SCTP_ADDR_DEL);
 799                                 found = 1;
 800                                 addr->valid = 0;
 801                                 list_del_rcu(&addr->list);
 802                                 break;
 803                         }
 804                 }
 805                 spin_unlock_bh(&net->sctp.local_addr_lock);
 806                 if (found)
 807                         kfree_rcu(addr, rcu);
 808                 break;
 809         }
 810
 811         return NOTIFY_DONE;
 812 }
 813
 814 /*
 815  * Initialize the control inode/socket with a control endpoint data
 816  * structure.  This endpoint is reserved exclusively for the OOTB processing.
 817  */
 818 static int sctp_ctl_sock_init(struct net *net)
 819 {
 820         int err;
 821         sa_family_t family = PF_INET;
 822
 823         if (sctp_get_pf_specific(PF_INET6))
 824                 family = PF_INET6;
 825
 826         err = inet_ctl_sock_create(&net->sctp.ctl_sock, family,
 827                                    SOCK_SEQPACKET, IPPROTO_SCTP, net);
 828
 829         /* If IPv6 socket could not be created, try the IPv4 socket */
 830         if (err < 0 && family == PF_INET6)
 831                 err = inet_ctl_sock_create(&net->sctp.ctl_sock, AF_INET,
 832                                            SOCK_SEQPACKET, IPPROTO_SCTP,
 833                                            net);
 834
 835         if (err < 0) {
 836                 pr_err("Failed to create the SCTP control socket\n");
 837                 return err;
 838         }
 839         return 0;
 840 }
 841
 842 /* Register address family specific functions. */
 843 int sctp_register_af(struct sctp_af *af)
 844 {
 845         switch (af->sa_family) {
 846         case AF_INET:
 847                 if (sctp_af_v4_specific)
 848                         return 0;
 849                 sctp_af_v4_specific = af;
 850                 break;
 851         case AF_INET6:
 852                 if (sctp_af_v6_specific)
 853                         return 0;
 854                 sctp_af_v6_specific = af;
 855                 break;
 856         default:
 857                 return 0;
 858         }
 859
 860         INIT_LIST_HEAD(&af->list);
 861         list_add_tail(&af->list, &sctp_address_families);
 862         return 1;
 863 }
 864
 865 /* Get the table of functions for manipulating a particular address
 866  * family.
 867  */
 868 struct sctp_af *sctp_get_af_specific(sa_family_t family)
 869 {
 870         switch (family) {
 871         case AF_INET:
 872                 return sctp_af_v4_specific;
 873         case AF_INET6:
 874                 return sctp_af_v6_specific;
 875         default:
 876                 return NULL;
 877         }
 878 }
 879
 880 /* Common code to initialize a AF_INET msg_name. */
 881 static void sctp_inet_msgname(char *msgname, int *addr_len)
 882 {
 883         struct sockaddr_in *sin;
 884
 885         sin = (struct sockaddr_in *)msgname;
 886         *addr_len = sizeof(struct sockaddr_in);
 887         sin->sin_family = AF_INET;
 888         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 889 }
 890
 891 /* Copy the primary address of the peer primary address as the msg_name. */
 892 static void sctp_inet_event_msgname(struct sctp_ulpevent *event, char *msgname,
 893                                     int *addr_len)
 894 {
 895         struct sockaddr_in *sin, *sinfrom;
 896
 897         if (msgname) {
 898                 struct sctp_association *asoc;
 899
 900                 asoc = event->asoc;
 901                 sctp_inet_msgname(msgname, addr_len);
 902                 sin = (struct sockaddr_in *)msgname;
 903                 sinfrom = &asoc->peer.primary_addr.v4;
 904                 sin->sin_port = htons(asoc->peer.port);
 905                 sin->sin_addr.s_addr = sinfrom->sin_addr.s_addr;
 906         }
 907 }
 908
 909 /* Initialize and copy out a msgname from an inbound skb. */
 910 static void sctp_inet_skb_msgname(struct sk_buff *skb, char *msgname, int *len)
 911 {
 912         if (msgname) {
 913                 struct sctphdr *sh = sctp_hdr(skb);
 914                 struct sockaddr_in *sin = (struct sockaddr_in *)msgname;
 915
 916                 sctp_inet_msgname(msgname, len);
 917                 sin->sin_port = sh->source;
 918                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 919         }
 920 }
 921
 922 /* Do we support this AF? */
 923 static int sctp_inet_af_supported(sa_family_t family, struct sctp_sock *sp)
 924 {
 925         /* PF_INET only supports AF_INET addresses. */
 926         return AF_INET == family;
 927 }
 928
 929 /* Address matching with wildcards allowed. */
 930 static int sctp_inet_cmp_addr(const union sctp_addr *addr1,
 931                               const union sctp_addr *addr2,
 932                               struct sctp_sock *opt)
 933 {
 934         /* PF_INET only supports AF_INET addresses. */
 935         if (addr1->sa.sa_family != addr2->sa.sa_family)
 936                 return 0;
 937         if (htonl(INADDR_ANY) == addr1->v4.sin_addr.s_addr ||
 938             htonl(INADDR_ANY) == addr2->v4.sin_addr.s_addr)
 939                 return 1;
 940         if (addr1->v4.sin_addr.s_addr == addr2->v4.sin_addr.s_addr)
 941                 return 1;
 942
 943         return 0;
 944 }
 945
 946 /* Verify that provided sockaddr looks bindable.  Common verification has
 947  * already been taken care of.
 948  */
 949 static int sctp_inet_bind_verify(struct sctp_sock *opt, union sctp_addr *addr)
 950 {
 951         return sctp_v4_available(addr, opt);
 952 }
 953
 954 /* Verify that sockaddr looks sendable.  Common verification has already
 955  * been taken care of.
 956  */
 957 static int sctp_inet_send_verify(struct sctp_sock *opt, union sctp_addr *addr)
 958 {
 959         return 1;
 960 }
 961
 962 /* Fill in Supported Address Type information for INIT and INIT-ACK
 963  * chunks.  Returns number of addresses supported.
 964  */
 965 static int sctp_inet_supported_addrs(const struct sctp_sock *opt,
 966                                      __be16 *types)
 967 {
 968         types[0] = SCTP_PARAM_IPV4_ADDRESS;
 969         return 1;
 970 }
 971
 972 /* Wrapper routine that calls the ip transmit routine. */
 973 static inline int sctp_v4_xmit(struct sk_buff *skb,
 974                                struct sctp_transport *transport)
 975 {
 976         struct inet_sock *inet = inet_sk(skb->sk);
 977         __u8 dscp = inet->tos;
 978
 979         pr_debug("%s: skb:%p, len:%d, src:%pI4, dst:%pI4\n", __func__, skb,
 980                  skb->len, &transport->fl.u.ip4.saddr,
 981                  &transport->fl.u.ip4.daddr);
 982
 983         if (transport->dscp & SCTP_DSCP_SET_MASK)
 984                 dscp = transport->dscp & SCTP_DSCP_VAL_MASK;
 985
 986         inet->pmtudisc = transport->param_flags & SPP_PMTUD_ENABLE ?
 987                          IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
 988
 989         SCTP_INC_STATS(sock_net(&inet->sk), SCTP_MIB_OUTSCTPPACKS);
 990
 991         return __ip_queue_xmit(&inet->sk, skb, &transport->fl, dscp);
 992 }
 993
 994 static struct sctp_af sctp_af_inet;
 995
 996 static struct sctp_pf sctp_pf_inet = {
 997         .event_msgname = sctp_inet_event_msgname,
 998         .skb_msgname   = sctp_inet_skb_msgname,
 999         .af_supported  = sctp_inet_af_supported,
1000         .cmp_addr      = sctp_inet_cmp_addr,
1001         .bind_verify   = sctp_inet_bind_verify,
1002         .send_verify   = sctp_inet_send_verify,
1003         .supported_addrs = sctp_inet_supported_addrs,
1004         .create_accept_sk = sctp_v4_create_accept_sk,
1005         .addr_to_user  = sctp_v4_addr_to_user,
1006         .to_sk_saddr   = sctp_v4_to_sk_saddr,
1007         .to_sk_daddr   = sctp_v4_to_sk_daddr,
1008         .copy_ip_options = sctp_v4_copy_ip_options,
1009         .af            = &sctp_af_inet
1010 };
1011
1012 /* Notifier for inetaddr addition/deletion events.  */
1013 static struct notifier_block sctp_inetaddr_notifier = {
1014         .notifier_call = sctp_inetaddr_event,
1015 };
1016
1017 /* Socket operations.  */
1018 static const struct proto_ops inet_seqpacket_ops = {
1019         .family            = PF_INET,
1020         .owner             = THIS_MODULE,
1021         .release           = inet_release,      /* Needs to be wrapped... */
1022         .bind              = inet_bind,
1023         .connect           = sctp_inet_connect,
1024         .socketpair        = sock_no_socketpair,
1025         .accept            = inet_accept,
1026         .getname           = inet_getname,      /* Semantics are different.  */
1027         .poll              = sctp_poll,
1028         .ioctl             = inet_ioctl,
1029         .gettstamp         = sock_gettstamp,
1030         .listen            = sctp_inet_listen,
1031         .shutdown          = inet_shutdown,     /* Looks harmless.  */
1032         .setsockopt        = sock_common_setsockopt, /* IP_SOL IP_OPTION is a problem */
1033         .getsockopt        = sock_common_getsockopt,
1034         .sendmsg           = inet_sendmsg,
1035         .recvmsg           = inet_recvmsg,
1036         .mmap              = sock_no_mmap,
1037         .sendpage          = sock_no_sendpage,
1038 #ifdef CONFIG_COMPAT
1039         .compat_setsockopt = compat_sock_common_setsockopt,
1040         .compat_getsockopt = compat_sock_common_getsockopt,
1041 #endif
1042 };
1043
1044 /* Registration with AF_INET family.  */
1045 static struct inet_protosw sctp_seqpacket_protosw = {
1046         .type       = SOCK_SEQPACKET,
1047         .protocol   = IPPROTO_SCTP,
1048         .prot       = &sctp_prot,
1049         .ops        = &inet_seqpacket_ops,
1050         .flags      = SCTP_PROTOSW_FLAG
1051 };
1052 static struct inet_protosw sctp_stream_protosw = {
1053         .type       = SOCK_STREAM,
1054         .protocol   = IPPROTO_SCTP,
1055         .prot       = &sctp_prot,
1056         .ops        = &inet_seqpacket_ops,
1057         .flags      = SCTP_PROTOSW_FLAG
1058 };
1059
1060 /* Register with IP layer.  */
1061 static const struct net_protocol sctp_protocol = {
1062         .handler     = sctp_rcv,
1063         .err_handler = sctp_v4_err,
1064         .no_policy   = 1,
1065         .netns_ok    = 1,
1066         .icmp_strict_tag_validation = 1,
1067 };
1068
1069 /* IPv4 address related functions.  */
1070 static struct sctp_af sctp_af_inet = {
1071         .sa_family         = AF_INET,
1072         .sctp_xmit         = sctp_v4_xmit,
1073         .setsockopt        = ip_setsockopt,
1074         .getsockopt        = ip_getsockopt,
1075         .get_dst           = sctp_v4_get_dst,
1076         .get_saddr         = sctp_v4_get_saddr,
1077         .copy_addrlist     = sctp_v4_copy_addrlist,
1078         .from_skb          = sctp_v4_from_skb,
1079         .from_sk           = sctp_v4_from_sk,
1080         .from_addr_param   = sctp_v4_from_addr_param,
1081         .to_addr_param     = sctp_v4_to_addr_param,
1082         .cmp_addr          = sctp_v4_cmp_addr,
1083         .addr_valid        = sctp_v4_addr_valid,
1084         .inaddr_any        = sctp_v4_inaddr_any,
1085         .is_any            = sctp_v4_is_any,
1086         .available         = sctp_v4_available,
1087         .scope             = sctp_v4_scope,
1088         .skb_iif           = sctp_v4_skb_iif,
1089         .is_ce             = sctp_v4_is_ce,
1090         .seq_dump_addr     = sctp_v4_seq_dump_addr,
1091         .ecn_capable       = sctp_v4_ecn_capable,
1092         .net_header_len    = sizeof(struct iphdr),
1093         .sockaddr_len      = sizeof(struct sockaddr_in),
1094         .ip_options_len    = sctp_v4_ip_options_len,
1095 #ifdef CONFIG_COMPAT
1096         .compat_setsockopt = compat_ip_setsockopt,
1097         .compat_getsockopt = compat_ip_getsockopt,
1098 #endif
1099 };
1100
1101 struct sctp_pf *sctp_get_pf_specific(sa_family_t family)
1102 {
1103         switch (family) {
1104         case PF_INET:
1105                 return sctp_pf_inet_specific;
1106         case PF_INET6:
1107                 return sctp_pf_inet6_specific;
1108         default:
1109                 return NULL;
1110         }
1111 }
1112
1113 /* Register the PF specific function table.  */
1114 int sctp_register_pf(struct sctp_pf *pf, sa_family_t family)
1115 {
1116         switch (family) {
1117         case PF_INET:
1118                 if (sctp_pf_inet_specific)
1119                         return 0;
1120                 sctp_pf_inet_specific = pf;
1121                 break;
1122         case PF_INET6:
1123                 if (sctp_pf_inet6_specific)
1124                         return 0;
1125                 sctp_pf_inet6_specific = pf;
1126                 break;
1127         default:
1128                 return 0;
1129         }
1130         return 1;
1131 }
1132
1133 static inline int init_sctp_mibs(struct net *net)
1134 {
1135         net->sctp.sctp_statistics = alloc_percpu(struct sctp_mib);
1136         if (!net->sctp.sctp_statistics)
1137                 return -ENOMEM;
1138         return 0;
1139 }
1140
1141 static inline void cleanup_sctp_mibs(struct net *net)
1142 {
1143         free_percpu(net->sctp.sctp_statistics);
1144 }
1145
1146 static void sctp_v4_pf_init(void)
1147 {
1148         /* Initialize the SCTP specific PF functions. */
1149         sctp_register_pf(&sctp_pf_inet, PF_INET);
1150         sctp_register_af(&sctp_af_inet);
1151 }
1152
1153 static void sctp_v4_pf_exit(void)
1154 {
1155         list_del(&sctp_af_inet.list);
1156 }
1157
1158 static int sctp_v4_protosw_init(void)
1159 {
1160         int rc;
1161
1162         rc = proto_register(&sctp_prot, 1);
1163         if (rc)
1164                 return rc;
1165
1166         /* Register SCTP(UDP and TCP style) with socket layer.  */
1167         inet_register_protosw(&sctp_seqpacket_protosw);
1168         inet_register_protosw(&sctp_stream_protosw);
1169
1170         return 0;
1171 }
1172
1173 static void sctp_v4_protosw_exit(void)
1174 {
1175         inet_unregister_protosw(&sctp_stream_protosw);
1176         inet_unregister_protosw(&sctp_seqpacket_protosw);
1177         proto_unregister(&sctp_prot);
1178 }
1179
1180 static int sctp_v4_add_protocol(void)
1181 {
1182         /* Register notifier for inet address additions/deletions. */
1183         register_inetaddr_notifier(&sctp_inetaddr_notifier);
1184
1185         /* Register SCTP with inet layer.  */
1186         if (inet_add_protocol(&sctp_protocol, IPPROTO_SCTP) < 0)
1187                 return -EAGAIN;
1188
1189         return 0;
1190 }
1191
1192 static void sctp_v4_del_protocol(void)
1193 {
1194         inet_del_protocol(&sctp_protocol, IPPROTO_SCTP);
1195         unregister_inetaddr_notifier(&sctp_inetaddr_notifier);
1196 }
1197
1198 static int __net_init sctp_defaults_init(struct net *net)
1199 {
1200         int status;
1201
1202         /*
1203          * 14. Suggested SCTP Protocol Parameter Values
1204          */
1205         /* The following protocol parameters are RECOMMENDED:  */
1206         /* RTO.Initial              - 3  seconds */
1207         net->sctp.rto_initial                   = SCTP_RTO_INITIAL;
1208         /* RTO.Min                  - 1  second */
1209         net->sctp.rto_min                       = SCTP_RTO_MIN;
1210         /* RTO.Max                 -  60 seconds */
1211         net->sctp.rto_max                       = SCTP_RTO_MAX;
1212         /* RTO.Alpha                - 1/8 */
1213         net->sctp.rto_alpha                     = SCTP_RTO_ALPHA;
1214         /* RTO.Beta                 - 1/4 */
1215         net->sctp.rto_beta                      = SCTP_RTO_BETA;
1216
1217         /* Valid.Cookie.Life        - 60  seconds */
1218         net->sctp.valid_cookie_life             = SCTP_DEFAULT_COOKIE_LIFE;
1219
1220         /* Whether Cookie Preservative is enabled(1) or not(0) */
1221         net->sctp.cookie_preserve_enable        = 1;
1222
1223         /* Default sctp sockets to use md5 as their hmac alg */
1224 #if defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_MD5)
1225         net->sctp.sctp_hmac_alg                 = "md5";
1226 #elif defined (CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1)
1227         net->sctp.sctp_hmac_alg                 = "sha1";
1228 #else
1229         net->sctp.sctp_hmac_alg                 = NULL;
1230 #endif
1231
1232         /* Max.Burst                - 4 */
1233         net->sctp.max_burst                     = SCTP_DEFAULT_MAX_BURST;
1234
1235         /* Enable pf state by default */
1236         net->sctp.pf_enable = 1;
1237
1238         /* Association.Max.Retrans  - 10 attempts
1239          * Path.Max.Retrans         - 5  attempts (per destination address)
1240          * Max.Init.Retransmits     - 8  attempts
1241          */
1242         net->sctp.max_retrans_association       = 10;
1243         net->sctp.max_retrans_path              = 5;
1244         net->sctp.max_retrans_init              = 8;
1245
1246         /* Sendbuffer growth        - do per-socket accounting */
1247         net->sctp.sndbuf_policy                 = 0;
1248
1249         /* Rcvbuffer growth         - do per-socket accounting */
1250         net->sctp.rcvbuf_policy                 = 0;
1251
1252         /* HB.interval              - 30 seconds */
1253         net->sctp.hb_interval                   = SCTP_DEFAULT_TIMEOUT_HEARTBEAT;
1254
1255         /* delayed SACK timeout */
1256         net->sctp.sack_timeout                  = SCTP_DEFAULT_TIMEOUT_SACK;
1257
1258         /* Disable ADDIP by default. */
1259         net->sctp.addip_enable = 0;
1260         net->sctp.addip_noauth = 0;
1261         net->sctp.default_auto_asconf = 0;
1262
1263         /* Enable PR-SCTP by default. */
1264         net->sctp.prsctp_enable = 1;
1265
1266         /* Disable RECONF by default. */
1267         net->sctp.reconf_enable = 0;
1268
1269         /* Disable AUTH by default. */
1270         net->sctp.auth_enable = 0;
1271
1272         /* Enable ECN by default. */
1273         net->sctp.ecn_enable = 1;
1274
1275         /* Set SCOPE policy to enabled */
1276         net->sctp.scope_policy = SCTP_SCOPE_POLICY_ENABLE;
1277
1278         /* Set the default rwnd update threshold */
1279         net->sctp.rwnd_upd_shift = SCTP_DEFAULT_RWND_SHIFT;
1280
1281         /* Initialize maximum autoclose timeout. */
1282         net->sctp.max_autoclose         = INT_MAX / HZ;
1283
1284         status = sctp_sysctl_net_register(net);
1285         if (status)
1286                 goto err_sysctl_register;
1287
1288         /* Allocate and initialise sctp mibs.  */
1289         status = init_sctp_mibs(net);
1290         if (status)
1291                 goto err_init_mibs;
1292
1293 #ifdef CONFIG_PROC_FS
1294         /* Initialize proc fs directory.  */
1295         status = sctp_proc_init(net);
1296         if (status)
1297                 goto err_init_proc;
1298 #endif
1299
1300         sctp_dbg_objcnt_init(net);
1301
1302         /* Initialize the local address list. */
1303         INIT_LIST_HEAD(&net->sctp.local_addr_list);
1304         spin_lock_init(&net->sctp.local_addr_lock);
1305         sctp_get_local_addr_list(net);
1306
1307         /* Initialize the address event list */
1308         INIT_LIST_HEAD(&net->sctp.addr_waitq);
1309         INIT_LIST_HEAD(&net->sctp.auto_asconf_splist);
1310         spin_lock_init(&net->sctp.addr_wq_lock);
1311         net->sctp.addr_wq_timer.expires = 0;
1312         timer_setup(&net->sctp.addr_wq_timer, sctp_addr_wq_timeout_handler, 0);
1313
1314         return 0;
1315
1316 #ifdef CONFIG_PROC_FS
1317 err_init_proc:
1318         cleanup_sctp_mibs(net);
1319 #endif
1320 err_init_mibs:
1321         sctp_sysctl_net_unregister(net);
1322 err_sysctl_register:
1323         return status;
1324 }
1325
1326 static void __net_exit sctp_defaults_exit(struct net *net)
1327 {
1328         /* Free the local address list */
1329         sctp_free_addr_wq(net);
1330         sctp_free_local_addr_list(net);
1331
1332 #ifdef CONFIG_PROC_FS
1333         remove_proc_subtree("sctp", net->proc_net);
1334         net->sctp.proc_net_sctp = NULL;
1335 #endif
1336         cleanup_sctp_mibs(net);
1337         sctp_sysctl_net_unregister(net);
1338 }
1339
1340 static struct pernet_operations sctp_defaults_ops = {
1341         .init = sctp_defaults_init,
1342         .exit = sctp_defaults_exit,
1343 };
1344
1345 static int __net_init sctp_ctrlsock_init(struct net *net)
1346 {
1347         int status;
1348
1349         /* Initialize the control inode/socket for handling OOTB packets.  */
1350         status = sctp_ctl_sock_init(net);
1351         if (status)
1352                 pr_err("Failed to initialize the SCTP control sock\n");
1353
1354         return status;
1355 }
1356
1357 static void __net_exit sctp_ctrlsock_exit(struct net *net)
1358 {
1359         /* Free the control endpoint.  */
1360         inet_ctl_sock_destroy(net->sctp.ctl_sock);
1361 }
1362
1363 static struct pernet_operations sctp_ctrlsock_ops = {
1364         .init = sctp_ctrlsock_init,
1365         .exit = sctp_ctrlsock_exit,
1366 };
1367
1368 /* Initialize the universe into something sensible.  */
1369 static __init int sctp_init(void)
1370 {
1371         int i;
1372         int status = -EINVAL;
1373         unsigned long goal;
1374         unsigned long limit;
1375         unsigned long nr_pages = totalram_pages();
1376         int max_share;
1377         int order;
1378         int num_entries;
1379         int max_entry_order;
1380
1381         sock_skb_cb_check_size(sizeof(struct sctp_ulpevent));
1382
1383         /* Allocate bind_bucket and chunk caches. */
1384         status = -ENOBUFS;
1385         sctp_bucket_cachep = kmem_cache_create("sctp_bind_bucket",
1386                                                sizeof(struct sctp_bind_bucket),
1387                                                0, SLAB_HWCACHE_ALIGN,
1388                                                NULL);
1389         if (!sctp_bucket_cachep)
1390                 goto out;
1391
1392         sctp_chunk_cachep = kmem_cache_create("sctp_chunk",
1393                                                sizeof(struct sctp_chunk),
1394                                                0, SLAB_HWCACHE_ALIGN,
1395                                                NULL);
1396         if (!sctp_chunk_cachep)
1397                 goto err_chunk_cachep;
1398
1399         status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
1400         if (status)
1401                 goto err_percpu_counter_init;
1402
1403         /* Implementation specific variables. */
1404
1405         /* Initialize default stream count setup information. */
1406         sctp_max_instreams              = SCTP_DEFAULT_INSTREAMS;
1407         sctp_max_outstreams             = SCTP_DEFAULT_OUTSTREAMS;
1408
1409         /* Initialize handle used for association ids. */
1410         idr_init(&sctp_assocs_id);
1411
1412         limit = nr_free_buffer_pages() / 8;
1413         limit = max(limit, 128UL);
1414         sysctl_sctp_mem[0] = limit / 4 * 3;
1415         sysctl_sctp_mem[1] = limit;
1416         sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
1417
1418         /* Set per-socket limits to no more than 1/128 the pressure threshold*/
1419         limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
1420         max_share = min(4UL*1024*1024, limit);
1421
1422         sysctl_sctp_rmem[0] = SK_MEM_QUANTUM; /* give each asoc 1 page min */
1423         sysctl_sctp_rmem[1] = 1500 * SKB_TRUESIZE(1);
1424         sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
1425
1426         sysctl_sctp_wmem[0] = SK_MEM_QUANTUM;
1427         sysctl_sctp_wmem[1] = 16*1024;
1428         sysctl_sctp_wmem[2] = max(64*1024, max_share);
1429
1430         /* Size and allocate the association hash table.
1431          * The methodology is similar to that of the tcp hash tables.
1432          * Though not identical.  Start by getting a goal size
1433          */
1434         if (nr_pages >= (128 * 1024))
1435                 goal = nr_pages >> (22 - PAGE_SHIFT);
1436         else
1437                 goal = nr_pages >> (24 - PAGE_SHIFT);
1438
1439         /* Then compute the page order for said goal */
1440         order = get_order(goal);
1441
1442         /* Now compute the required page order for the maximum sized table we
1443          * want to create
1444          */
1445         max_entry_order = get_order(MAX_SCTP_PORT_HASH_ENTRIES *
1446                                     sizeof(struct sctp_bind_hashbucket));
1447
1448         /* Limit the page order by that maximum hash table size */
1449         order = min(order, max_entry_order);
1450
1451         /* Allocate and initialize the endpoint hash table.  */
1452         sctp_ep_hashsize = 64;
1453         sctp_ep_hashtable =
1454                 kmalloc_array(64, sizeof(struct sctp_hashbucket), GFP_KERNEL);
1455         if (!sctp_ep_hashtable) {
1456                 pr_err("Failed endpoint_hash alloc\n");
1457                 status = -ENOMEM;
1458                 goto err_ehash_alloc;
1459         }
1460         for (i = 0; i < sctp_ep_hashsize; i++) {
1461                 rwlock_init(&sctp_ep_hashtable[i].lock);
1462                 INIT_HLIST_HEAD(&sctp_ep_hashtable[i].chain);
1463         }
1464
1465         /* Allocate and initialize the SCTP port hash table.
1466          * Note that order is initalized to start at the max sized
1467          * table we want to support.  If we can't get that many pages
1468          * reduce the order and try again
1469          */
1470         do {
1471                 sctp_port_hashtable = (struct sctp_bind_hashbucket *)
1472                         __get_free_pages(GFP_KERNEL | __GFP_NOWARN, order);
1473         } while (!sctp_port_hashtable && --order > 0);
1474
1475         if (!sctp_port_hashtable) {
1476                 pr_err("Failed bind hash alloc\n");
1477                 status = -ENOMEM;
1478                 goto err_bhash_alloc;
1479         }
1480
1481         /* Now compute the number of entries that will fit in the
1482          * port hash space we allocated
1483          */
1484         num_entries = (1UL << order) * PAGE_SIZE /
1485                       sizeof(struct sctp_bind_hashbucket);
1486
1487         /* And finish by rounding it down to the nearest power of two
1488          * this wastes some memory of course, but its needed because
1489          * the hash function operates based on the assumption that
1490          * that the number of entries is a power of two
1491          */
1492         sctp_port_hashsize = rounddown_pow_of_two(num_entries);
1493
1494         for (i = 0; i < sctp_port_hashsize; i++) {
1495                 spin_lock_init(&sctp_port_hashtable[i].lock);
1496                 INIT_HLIST_HEAD(&sctp_port_hashtable[i].chain);
1497         }
1498
1499         status = sctp_transport_hashtable_init();
1500         if (status)
1501                 goto err_thash_alloc;
1502
1503         pr_info("Hash tables configured (bind %d/%d)\n", sctp_port_hashsize,
1504                 num_entries);
1505
1506         sctp_sysctl_register();
1507
1508         INIT_LIST_HEAD(&sctp_address_families);
1509         sctp_v4_pf_init();
1510         sctp_v6_pf_init();
1511         sctp_sched_ops_init();
1512
1513         status = register_pernet_subsys(&sctp_defaults_ops);
1514         if (status)
1515                 goto err_register_defaults;
1516
1517         status = sctp_v4_protosw_init();
1518         if (status)
1519                 goto err_protosw_init;
1520
1521         status = sctp_v6_protosw_init();
1522         if (status)
1523                 goto err_v6_protosw_init;
1524
1525         status = register_pernet_subsys(&sctp_ctrlsock_ops);
1526         if (status)
1527                 goto err_register_ctrlsock;
1528
1529         status = sctp_v4_add_protocol();
1530         if (status)
1531                 goto err_add_protocol;
1532
1533         /* Register SCTP with inet6 layer.  */
1534         status = sctp_v6_add_protocol();
1535         if (status)
1536                 goto err_v6_add_protocol;
1537
1538         if (sctp_offload_init() < 0)
1539                 pr_crit("%s: Cannot add SCTP protocol offload\n", __func__);
1540
1541 out:
1542         return status;
1543 err_v6_add_protocol:
1544         sctp_v4_del_protocol();
1545 err_add_protocol:
1546         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1547 err_register_ctrlsock:
1548         sctp_v6_protosw_exit();
1549 err_v6_protosw_init:
1550         sctp_v4_protosw_exit();
1551 err_protosw_init:
1552         unregister_pernet_subsys(&sctp_defaults_ops);
1553 err_register_defaults:
1554         sctp_v4_pf_exit();
1555         sctp_v6_pf_exit();
1556         sctp_sysctl_unregister();
1557         free_pages((unsigned long)sctp_port_hashtable,
1558                    get_order(sctp_port_hashsize *
1559                              sizeof(struct sctp_bind_hashbucket)));
1560 err_bhash_alloc:
1561         sctp_transport_hashtable_destroy();
1562 err_thash_alloc:
1563         kfree(sctp_ep_hashtable);
1564 err_ehash_alloc:
1565         percpu_counter_destroy(&sctp_sockets_allocated);
1566 err_percpu_counter_init:
1567         kmem_cache_destroy(sctp_chunk_cachep);
1568 err_chunk_cachep:
1569         kmem_cache_destroy(sctp_bucket_cachep);
1570         goto out;
1571 }
1572
1573 /* Exit handler for the SCTP protocol.  */
1574 static __exit void sctp_exit(void)
1575 {
1576         /* BUG.  This should probably do something useful like clean
1577          * up all the remaining associations and all that memory.
1578          */
1579
1580         /* Unregister with inet6/inet layers. */
1581         sctp_v6_del_protocol();
1582         sctp_v4_del_protocol();
1583
1584         unregister_pernet_subsys(&sctp_ctrlsock_ops);
1585
1586         /* Free protosw registrations */
1587         sctp_v6_protosw_exit();
1588         sctp_v4_protosw_exit();
1589
1590         unregister_pernet_subsys(&sctp_defaults_ops);
1591
1592         /* Unregister with socket layer. */
1593         sctp_v6_pf_exit();
1594         sctp_v4_pf_exit();
1595
1596         sctp_sysctl_unregister();
1597
1598         free_pages((unsigned long)sctp_port_hashtable,
1599                    get_order(sctp_port_hashsize *
1600                              sizeof(struct sctp_bind_hashbucket)));
1601         kfree(sctp_ep_hashtable);
1602         sctp_transport_hashtable_destroy();
1603
1604         percpu_counter_destroy(&sctp_sockets_allocated);
1605
1606         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1607
1608         kmem_cache_destroy(sctp_chunk_cachep);
1609         kmem_cache_destroy(sctp_bucket_cachep);
1610 }
1611
1612 module_init(sctp_init);
1613 module_exit(sctp_exit);
1614
1615 /*
1616  * __stringify doesn't likes enums, so use IPPROTO_SCTP value (132) directly.
1617  */
1618 MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-132");
1619 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-132");
1620 MODULE_AUTHOR("Linux Kernel SCTP developers <linux-sctp@vger.kernel.org>");
1621 MODULE_DESCRIPTION("Support for the SCTP protocol (RFC2960)");
1622 module_param_named(no_checksums, sctp_checksum_disable, bool, 0644);
1623 MODULE_PARM_DESC(no_checksums, "Disable checksums computing and verification");
1624 MODULE_LICENSE("GPL");