net/l2tp/l2tp_core.c

   1 /*
   2  * L2TP core.
   3  *
   4  * Copyright (c) 2008,2009,2010 Katalix Systems Ltd
   5  *
   6  * This file contains some code of the original L2TPv2 pppol2tp
   7  * driver, which has the following copyright:
   8  *
   9  * Authors:     Martijn van Oosterhout <kleptog@svana.org>
  10  *              James Chapman (jchapman@katalix.com)
  11  * Contributors:
  12  *              Michal Ostrowski <mostrows@speakeasy.net>
  13  *              Arnaldo Carvalho de Melo <acme@xconectiva.com.br>
  14  *              David S. Miller (davem@redhat.com)
  15  *
  16  * This program is free software; you can redistribute it and/or modify
  17  * it under the terms of the GNU General Public License version 2 as
  18  * published by the Free Software Foundation.
  19  */
  20
  21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  22
  23 #include <linux/module.h>
  24 #include <linux/string.h>
  25 #include <linux/list.h>
  26 #include <linux/rculist.h>
  27 #include <linux/uaccess.h>
  28
  29 #include <linux/kernel.h>
  30 #include <linux/spinlock.h>
  31 #include <linux/kthread.h>
  32 #include <linux/sched.h>
  33 #include <linux/slab.h>
  34 #include <linux/errno.h>
  35 #include <linux/jiffies.h>
  36
  37 #include <linux/netdevice.h>
  38 #include <linux/net.h>
  39 #include <linux/inetdevice.h>
  40 #include <linux/skbuff.h>
  41 #include <linux/init.h>
  42 #include <linux/in.h>
  43 #include <linux/ip.h>
  44 #include <linux/udp.h>
  45 #include <linux/l2tp.h>
  46 #include <linux/hash.h>
  47 #include <linux/sort.h>
  48 #include <linux/file.h>
  49 #include <linux/nsproxy.h>
  50 #include <net/net_namespace.h>
  51 #include <net/netns/generic.h>
  52 #include <net/dst.h>
  53 #include <net/ip.h>
  54 #include <net/udp.h>
  55 #include <net/udp_tunnel.h>
  56 #include <net/inet_common.h>
  57 #include <net/xfrm.h>
  58 #include <net/protocol.h>
  59 #include <net/inet6_connection_sock.h>
  60 #include <net/inet_ecn.h>
  61 #include <net/ip6_route.h>
  62 #include <net/ip6_checksum.h>
  63
  64 #include <asm/byteorder.h>
  65 #include <linux/atomic.h>
  66
  67 #include "l2tp_core.h"
  68
  69 #define L2TP_DRV_VERSION        "V2.0"
  70
  71 /* L2TP header constants */
  72 #define L2TP_HDRFLAG_T     0x8000
  73 #define L2TP_HDRFLAG_L     0x4000
  74 #define L2TP_HDRFLAG_S     0x0800
  75 #define L2TP_HDRFLAG_O     0x0200
  76 #define L2TP_HDRFLAG_P     0x0100
  77
  78 #define L2TP_HDR_VER_MASK  0x000F
  79 #define L2TP_HDR_VER_2     0x0002
  80 #define L2TP_HDR_VER_3     0x0003
  81
  82 /* L2TPv3 default L2-specific sublayer */
  83 #define L2TP_SLFLAG_S      0x40000000
  84 #define L2TP_SL_SEQ_MASK   0x00ffffff
  85
  86 #define L2TP_HDR_SIZE_SEQ               10
  87 #define L2TP_HDR_SIZE_NOSEQ             6
  88
  89 /* Default trace flags */
  90 #define L2TP_DEFAULT_DEBUG_FLAGS        0
  91
  92 /* Private data stored for received packets in the skb.
  93  */
  94 struct l2tp_skb_cb {
  95         u32                     ns;
  96         u16                     has_seq;
  97         u16                     length;
  98         unsigned long           expires;
  99 };
 100
 101 #define L2TP_SKB_CB(skb)        ((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
 102
 103 static atomic_t l2tp_tunnel_count;
 104 static atomic_t l2tp_session_count;
 105 static struct workqueue_struct *l2tp_wq;
 106
 107 /* per-net private data for this module */
 108 static unsigned int l2tp_net_id;
 109 struct l2tp_net {
 110         struct list_head l2tp_tunnel_list;
 111         spinlock_t l2tp_tunnel_list_lock;
 112         struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
 113         spinlock_t l2tp_session_hlist_lock;
 114 };
 115
 116
 117 static inline struct l2tp_tunnel *l2tp_tunnel(struct sock *sk)
 118 {
 119         return sk->sk_user_data;
 120 }
 121
 122 static inline struct l2tp_net *l2tp_pernet(const struct net *net)
 123 {
 124         BUG_ON(!net);
 125
 126         return net_generic(net, l2tp_net_id);
 127 }
 128
 129 /* Session hash global list for L2TPv3.
 130  * The session_id SHOULD be random according to RFC3931, but several
 131  * L2TP implementations use incrementing session_ids.  So we do a real
 132  * hash on the session_id, rather than a simple bitmask.
 133  */
 134 static inline struct hlist_head *
 135 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
 136 {
 137         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
 138
 139 }
 140
 141 /* Lookup the tunnel socket, possibly involving the fs code if the socket is
 142  * owned by userspace.  A struct sock returned from this function must be
 143  * released using l2tp_tunnel_sock_put once you're done with it.
 144  */
 145 static struct sock *l2tp_tunnel_sock_lookup(struct l2tp_tunnel *tunnel)
 146 {
 147         int err = 0;
 148         struct socket *sock = NULL;
 149         struct sock *sk = NULL;
 150
 151         if (!tunnel)
 152                 goto out;
 153
 154         if (tunnel->fd >= 0) {
 155                 /* Socket is owned by userspace, who might be in the process
 156                  * of closing it.  Look the socket up using the fd to ensure
 157                  * consistency.
 158                  */
 159                 sock = sockfd_lookup(tunnel->fd, &err);
 160                 if (sock)
 161                         sk = sock->sk;
 162         } else {
 163                 /* Socket is owned by kernelspace */
 164                 sk = tunnel->sock;
 165                 sock_hold(sk);
 166         }
 167
 168 out:
 169         return sk;
 170 }
 171
 172 /* Drop a reference to a tunnel socket obtained via. l2tp_tunnel_sock_put */
 173 static void l2tp_tunnel_sock_put(struct sock *sk)
 174 {
 175         struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
 176         if (tunnel) {
 177                 if (tunnel->fd >= 0) {
 178                         /* Socket is owned by userspace */
 179                         sockfd_put(sk->sk_socket);
 180                 }
 181                 sock_put(sk);
 182         }
 183         sock_put(sk);
 184 }
 185
 186 /* Session hash list.
 187  * The session_id SHOULD be random according to RFC2661, but several
 188  * L2TP implementations (Cisco and Microsoft) use incrementing
 189  * session_ids.  So we do a real hash on the session_id, rather than a
 190  * simple bitmask.
 191  */
 192 static inline struct hlist_head *
 193 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
 194 {
 195         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
 196 }
 197
 198 /* Lookup a tunnel. A new reference is held on the returned tunnel. */
 199 struct l2tp_tunnel *l2tp_tunnel_get(const struct net *net, u32 tunnel_id)
 200 {
 201         const struct l2tp_net *pn = l2tp_pernet(net);
 202         struct l2tp_tunnel *tunnel;
 203
 204         rcu_read_lock_bh();
 205         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
 206                 if (tunnel->tunnel_id == tunnel_id) {
 207                         l2tp_tunnel_inc_refcount(tunnel);
 208                         rcu_read_unlock_bh();
 209
 210                         return tunnel;
 211                 }
 212         }
 213         rcu_read_unlock_bh();
 214
 215         return NULL;
 216 }
 217 EXPORT_SYMBOL_GPL(l2tp_tunnel_get);
 218
 219 /* Lookup a session. A new reference is held on the returned session.
 220  * Optionally calls session->ref() too if do_ref is true.
 221  */
 222 struct l2tp_session *l2tp_session_get(const struct net *net,
 223                                       struct l2tp_tunnel *tunnel,
 224                                       u32 session_id, bool do_ref)
 225 {
 226         struct hlist_head *session_list;
 227         struct l2tp_session *session;
 228
 229         if (!tunnel) {
 230                 struct l2tp_net *pn = l2tp_pernet(net);
 231
 232                 session_list = l2tp_session_id_hash_2(pn, session_id);
 233
 234                 rcu_read_lock_bh();
 235                 hlist_for_each_entry_rcu(session, session_list, global_hlist) {
 236                         if (session->session_id == session_id) {
 237                                 l2tp_session_inc_refcount(session);
 238                                 if (do_ref && session->ref)
 239                                         session->ref(session);
 240                                 rcu_read_unlock_bh();
 241
 242                                 return session;
 243                         }
 244                 }
 245                 rcu_read_unlock_bh();
 246
 247                 return NULL;
 248         }
 249
 250         session_list = l2tp_session_id_hash(tunnel, session_id);
 251         read_lock_bh(&tunnel->hlist_lock);
 252         hlist_for_each_entry(session, session_list, hlist) {
 253                 if (session->session_id == session_id) {
 254                         l2tp_session_inc_refcount(session);
 255                         if (do_ref && session->ref)
 256                                 session->ref(session);
 257                         read_unlock_bh(&tunnel->hlist_lock);
 258
 259                         return session;
 260                 }
 261         }
 262         read_unlock_bh(&tunnel->hlist_lock);
 263
 264         return NULL;
 265 }
 266 EXPORT_SYMBOL_GPL(l2tp_session_get);
 267
 268 struct l2tp_session *l2tp_session_get_nth(struct l2tp_tunnel *tunnel, int nth,
 269                                           bool do_ref)
 270 {
 271         int hash;
 272         struct l2tp_session *session;
 273         int count = 0;
 274
 275         read_lock_bh(&tunnel->hlist_lock);
 276         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
 277                 hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
 278                         if (++count > nth) {
 279                                 l2tp_session_inc_refcount(session);
 280                                 if (do_ref && session->ref)
 281                                         session->ref(session);
 282                                 read_unlock_bh(&tunnel->hlist_lock);
 283                                 return session;
 284                         }
 285                 }
 286         }
 287
 288         read_unlock_bh(&tunnel->hlist_lock);
 289
 290         return NULL;
 291 }
 292 EXPORT_SYMBOL_GPL(l2tp_session_get_nth);
 293
 294 /* Lookup a session by interface name.
 295  * This is very inefficient but is only used by management interfaces.
 296  */
 297 struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net,
 298                                                 const char *ifname,
 299                                                 bool do_ref)
 300 {
 301         struct l2tp_net *pn = l2tp_pernet(net);
 302         int hash;
 303         struct l2tp_session *session;
 304
 305         rcu_read_lock_bh();
 306         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
 307                 hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
 308                         if (!strcmp(session->ifname, ifname)) {
 309                                 l2tp_session_inc_refcount(session);
 310                                 if (do_ref && session->ref)
 311                                         session->ref(session);
 312                                 rcu_read_unlock_bh();
 313
 314                                 return session;
 315                         }
 316                 }
 317         }
 318
 319         rcu_read_unlock_bh();
 320
 321         return NULL;
 322 }
 323 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname);
 324
 325 static int l2tp_session_add_to_tunnel(struct l2tp_tunnel *tunnel,
 326                                       struct l2tp_session *session)
 327 {
 328         struct l2tp_session *session_walk;
 329         struct hlist_head *g_head;
 330         struct hlist_head *head;
 331         struct l2tp_net *pn;
 332         int err;
 333
 334         head = l2tp_session_id_hash(tunnel, session->session_id);
 335
 336         write_lock_bh(&tunnel->hlist_lock);
 337         if (!tunnel->acpt_newsess) {
 338                 err = -ENODEV;
 339                 goto err_tlock;
 340         }
 341
 342         hlist_for_each_entry(session_walk, head, hlist)
 343                 if (session_walk->session_id == session->session_id) {
 344                         err = -EEXIST;
 345                         goto err_tlock;
 346                 }
 347
 348         if (tunnel->version == L2TP_HDR_VER_3) {
 349                 pn = l2tp_pernet(tunnel->l2tp_net);
 350                 g_head = l2tp_session_id_hash_2(l2tp_pernet(tunnel->l2tp_net),
 351                                                 session->session_id);
 352
 353                 spin_lock_bh(&pn->l2tp_session_hlist_lock);
 354
 355                 hlist_for_each_entry(session_walk, g_head, global_hlist)
 356                         if (session_walk->session_id == session->session_id) {
 357                                 err = -EEXIST;
 358                                 goto err_tlock_pnlock;
 359                         }
 360
 361                 l2tp_tunnel_inc_refcount(tunnel);
 362                 sock_hold(tunnel->sock);
 363                 hlist_add_head_rcu(&session->global_hlist, g_head);
 364
 365                 spin_unlock_bh(&pn->l2tp_session_hlist_lock);
 366         } else {
 367                 l2tp_tunnel_inc_refcount(tunnel);
 368                 sock_hold(tunnel->sock);
 369         }
 370
 371         hlist_add_head(&session->hlist, head);
 372         write_unlock_bh(&tunnel->hlist_lock);
 373
 374         return 0;
 375
 376 err_tlock_pnlock:
 377         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
 378 err_tlock:
 379         write_unlock_bh(&tunnel->hlist_lock);
 380
 381         return err;
 382 }
 383
 384 /* Lookup a tunnel by id
 385  */
 386 struct l2tp_tunnel *l2tp_tunnel_find(const struct net *net, u32 tunnel_id)
 387 {
 388         struct l2tp_tunnel *tunnel;
 389         struct l2tp_net *pn = l2tp_pernet(net);
 390
 391         rcu_read_lock_bh();
 392         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
 393                 if (tunnel->tunnel_id == tunnel_id) {
 394                         rcu_read_unlock_bh();
 395                         return tunnel;
 396                 }
 397         }
 398         rcu_read_unlock_bh();
 399
 400         return NULL;
 401 }
 402 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
 403
 404 struct l2tp_tunnel *l2tp_tunnel_find_nth(const struct net *net, int nth)
 405 {
 406         struct l2tp_net *pn = l2tp_pernet(net);
 407         struct l2tp_tunnel *tunnel;
 408         int count = 0;
 409
 410         rcu_read_lock_bh();
 411         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
 412                 if (++count > nth) {
 413                         rcu_read_unlock_bh();
 414                         return tunnel;
 415                 }
 416         }
 417
 418         rcu_read_unlock_bh();
 419
 420         return NULL;
 421 }
 422 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
 423
 424 /*****************************************************************************
 425  * Receive data handling
 426  *****************************************************************************/
 427
 428 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
 429  * number.
 430  */
 431 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
 432 {
 433         struct sk_buff *skbp;
 434         struct sk_buff *tmp;
 435         u32 ns = L2TP_SKB_CB(skb)->ns;
 436
 437         spin_lock_bh(&session->reorder_q.lock);
 438         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
 439                 if (L2TP_SKB_CB(skbp)->ns > ns) {
 440                         __skb_queue_before(&session->reorder_q, skbp, skb);
 441                         l2tp_dbg(session, L2TP_MSG_SEQ,
 442                                  "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
 443                                  session->name, ns, L2TP_SKB_CB(skbp)->ns,
 444                                  skb_queue_len(&session->reorder_q));
 445                         atomic_long_inc(&session->stats.rx_oos_packets);
 446                         goto out;
 447                 }
 448         }
 449
 450         __skb_queue_tail(&session->reorder_q, skb);
 451
 452 out:
 453         spin_unlock_bh(&session->reorder_q.lock);
 454 }
 455
 456 /* Dequeue a single skb.
 457  */
 458 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
 459 {
 460         struct l2tp_tunnel *tunnel = session->tunnel;
 461         int length = L2TP_SKB_CB(skb)->length;
 462
 463         /* We're about to requeue the skb, so return resources
 464          * to its current owner (a socket receive buffer).
 465          */
 466         skb_orphan(skb);
 467
 468         atomic_long_inc(&tunnel->stats.rx_packets);
 469         atomic_long_add(length, &tunnel->stats.rx_bytes);
 470         atomic_long_inc(&session->stats.rx_packets);
 471         atomic_long_add(length, &session->stats.rx_bytes);
 472
 473         if (L2TP_SKB_CB(skb)->has_seq) {
 474                 /* Bump our Nr */
 475                 session->nr++;
 476                 session->nr &= session->nr_max;
 477
 478                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
 479                          session->name, session->nr);
 480         }
 481
 482         /* call private receive handler */
 483         if (session->recv_skb != NULL)
 484                 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
 485         else
 486                 kfree_skb(skb);
 487
 488         if (session->deref)
 489                 (*session->deref)(session);
 490 }
 491
 492 /* Dequeue skbs from the session's reorder_q, subject to packet order.
 493  * Skbs that have been in the queue for too long are simply discarded.
 494  */
 495 static void l2tp_recv_dequeue(struct l2tp_session *session)
 496 {
 497         struct sk_buff *skb;
 498         struct sk_buff *tmp;
 499
 500         /* If the pkt at the head of the queue has the nr that we
 501          * expect to send up next, dequeue it and any other
 502          * in-sequence packets behind it.
 503          */
 504 start:
 505         spin_lock_bh(&session->reorder_q.lock);
 506         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
 507                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
 508                         atomic_long_inc(&session->stats.rx_seq_discards);
 509                         atomic_long_inc(&session->stats.rx_errors);
 510                         l2tp_dbg(session, L2TP_MSG_SEQ,
 511                                  "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
 512                                  session->name, L2TP_SKB_CB(skb)->ns,
 513                                  L2TP_SKB_CB(skb)->length, session->nr,
 514                                  skb_queue_len(&session->reorder_q));
 515                         session->reorder_skip = 1;
 516                         __skb_unlink(skb, &session->reorder_q);
 517                         kfree_skb(skb);
 518                         if (session->deref)
 519                                 (*session->deref)(session);
 520                         continue;
 521                 }
 522
 523                 if (L2TP_SKB_CB(skb)->has_seq) {
 524                         if (session->reorder_skip) {
 525                                 l2tp_dbg(session, L2TP_MSG_SEQ,
 526                                          "%s: advancing nr to next pkt: %u -> %u",
 527                                          session->name, session->nr,
 528                                          L2TP_SKB_CB(skb)->ns);
 529                                 session->reorder_skip = 0;
 530                                 session->nr = L2TP_SKB_CB(skb)->ns;
 531                         }
 532                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
 533                                 l2tp_dbg(session, L2TP_MSG_SEQ,
 534                                          "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
 535                                          session->name, L2TP_SKB_CB(skb)->ns,
 536                                          L2TP_SKB_CB(skb)->length, session->nr,
 537                                          skb_queue_len(&session->reorder_q));
 538                                 goto out;
 539                         }
 540                 }
 541                 __skb_unlink(skb, &session->reorder_q);
 542
 543                 /* Process the skb. We release the queue lock while we
 544                  * do so to let other contexts process the queue.
 545                  */
 546                 spin_unlock_bh(&session->reorder_q.lock);
 547                 l2tp_recv_dequeue_skb(session, skb);
 548                 goto start;
 549         }
 550
 551 out:
 552         spin_unlock_bh(&session->reorder_q.lock);
 553 }
 554
 555 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
 556 {
 557         u32 nws;
 558
 559         if (nr >= session->nr)
 560                 nws = nr - session->nr;
 561         else
 562                 nws = (session->nr_max + 1) - (session->nr - nr);
 563
 564         return nws < session->nr_window_size;
 565 }
 566
 567 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
 568  * acceptable, else non-zero.
 569  */
 570 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
 571 {
 572         if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
 573                 /* Packet sequence number is outside allowed window.
 574                  * Discard it.
 575                  */
 576                 l2tp_dbg(session, L2TP_MSG_SEQ,
 577                          "%s: pkt %u len %d discarded, outside window, nr=%u\n",
 578                          session->name, L2TP_SKB_CB(skb)->ns,
 579                          L2TP_SKB_CB(skb)->length, session->nr);
 580                 goto discard;
 581         }
 582
 583         if (session->reorder_timeout != 0) {
 584                 /* Packet reordering enabled. Add skb to session's
 585                  * reorder queue, in order of ns.
 586                  */
 587                 l2tp_recv_queue_skb(session, skb);
 588                 goto out;
 589         }
 590
 591         /* Packet reordering disabled. Discard out-of-sequence packets, while
 592          * tracking the number if in-sequence packets after the first OOS packet
 593          * is seen. After nr_oos_count_max in-sequence packets, reset the
 594          * sequence number to re-enable packet reception.
 595          */
 596         if (L2TP_SKB_CB(skb)->ns == session->nr) {
 597                 skb_queue_tail(&session->reorder_q, skb);
 598         } else {
 599                 u32 nr_oos = L2TP_SKB_CB(skb)->ns;
 600                 u32 nr_next = (session->nr_oos + 1) & session->nr_max;
 601
 602                 if (nr_oos == nr_next)
 603                         session->nr_oos_count++;
 604                 else
 605                         session->nr_oos_count = 0;
 606
 607                 session->nr_oos = nr_oos;
 608                 if (session->nr_oos_count > session->nr_oos_count_max) {
 609                         session->reorder_skip = 1;
 610                         l2tp_dbg(session, L2TP_MSG_SEQ,
 611                                  "%s: %d oos packets received. Resetting sequence numbers\n",
 612                                  session->name, session->nr_oos_count);
 613                 }
 614                 if (!session->reorder_skip) {
 615                         atomic_long_inc(&session->stats.rx_seq_discards);
 616                         l2tp_dbg(session, L2TP_MSG_SEQ,
 617                                  "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
 618                                  session->name, L2TP_SKB_CB(skb)->ns,
 619                                  L2TP_SKB_CB(skb)->length, session->nr,
 620                                  skb_queue_len(&session->reorder_q));
 621                         goto discard;
 622                 }
 623                 skb_queue_tail(&session->reorder_q, skb);
 624         }
 625
 626 out:
 627         return 0;
 628
 629 discard:
 630         return 1;
 631 }
 632
 633 /* Do receive processing of L2TP data frames. We handle both L2TPv2
 634  * and L2TPv3 data frames here.
 635  *
 636  * L2TPv2 Data Message Header
 637  *
 638  *  0                   1                   2                   3
 639  *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
 640  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 641  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
 642  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 643  * |           Tunnel ID           |           Session ID          |
 644  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 645  * |             Ns (opt)          |             Nr (opt)          |
 646  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 647  * |      Offset Size (opt)        |    Offset pad... (opt)
 648  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 649  *
 650  * Data frames are marked by T=0. All other fields are the same as
 651  * those in L2TP control frames.
 652  *
 653  * L2TPv3 Data Message Header
 654  *
 655  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 656  * |                      L2TP Session Header                      |
 657  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 658  * |                      L2-Specific Sublayer                     |
 659  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 660  * |                        Tunnel Payload                      ...
 661  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 662  *
 663  * L2TPv3 Session Header Over IP
 664  *
 665  *  0                   1                   2                   3
 666  *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
 667  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 668  * |                           Session ID                          |
 669  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 670  * |               Cookie (optional, maximum 64 bits)...
 671  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 672  *                                                                 |
 673  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 674  *
 675  * L2TPv3 L2-Specific Sublayer Format
 676  *
 677  *  0                   1                   2                   3
 678  *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
 679  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 680  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
 681  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 682  *
 683  * Cookie value, sublayer format and offset (pad) are negotiated with
 684  * the peer when the session is set up. Unlike L2TPv2, we do not need
 685  * to parse the packet header to determine if optional fields are
 686  * present.
 687  *
 688  * Caller must already have parsed the frame and determined that it is
 689  * a data (not control) frame before coming here. Fields up to the
 690  * session-id have already been parsed and ptr points to the data
 691  * after the session-id.
 692  *
 693  * session->ref() must have been called prior to l2tp_recv_common().
 694  * session->deref() will be called automatically after skb is processed.
 695  */
 696 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
 697                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
 698                       int length, int (*payload_hook)(struct sk_buff *skb))
 699 {
 700         struct l2tp_tunnel *tunnel = session->tunnel;
 701         int offset;
 702         u32 ns, nr;
 703
 704         /* Parse and check optional cookie */
 705         if (session->peer_cookie_len > 0) {
 706                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
 707                         l2tp_info(tunnel, L2TP_MSG_DATA,
 708                                   "%s: cookie mismatch (%u/%u). Discarding.\n",
 709                                   tunnel->name, tunnel->tunnel_id,
 710                                   session->session_id);
 711                         atomic_long_inc(&session->stats.rx_cookie_discards);
 712                         goto discard;
 713                 }
 714                 ptr += session->peer_cookie_len;
 715         }
 716
 717         /* Handle the optional sequence numbers. Sequence numbers are
 718          * in different places for L2TPv2 and L2TPv3.
 719          *
 720          * If we are the LAC, enable/disable sequence numbers under
 721          * the control of the LNS.  If no sequence numbers present but
 722          * we were expecting them, discard frame.
 723          */
 724         ns = nr = 0;
 725         L2TP_SKB_CB(skb)->has_seq = 0;
 726         if (tunnel->version == L2TP_HDR_VER_2) {
 727                 if (hdrflags & L2TP_HDRFLAG_S) {
 728                         ns = ntohs(*(__be16 *) ptr);
 729                         ptr += 2;
 730                         nr = ntohs(*(__be16 *) ptr);
 731                         ptr += 2;
 732
 733                         /* Store L2TP info in the skb */
 734                         L2TP_SKB_CB(skb)->ns = ns;
 735                         L2TP_SKB_CB(skb)->has_seq = 1;
 736
 737                         l2tp_dbg(session, L2TP_MSG_SEQ,
 738                                  "%s: recv data ns=%u, nr=%u, session nr=%u\n",
 739                                  session->name, ns, nr, session->nr);
 740                 }
 741         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
 742                 u32 l2h = ntohl(*(__be32 *) ptr);
 743
 744                 if (l2h & 0x40000000) {
 745                         ns = l2h & 0x00ffffff;
 746
 747                         /* Store L2TP info in the skb */
 748                         L2TP_SKB_CB(skb)->ns = ns;
 749                         L2TP_SKB_CB(skb)->has_seq = 1;
 750
 751                         l2tp_dbg(session, L2TP_MSG_SEQ,
 752                                  "%s: recv data ns=%u, session nr=%u\n",
 753                                  session->name, ns, session->nr);
 754                 }
 755         }
 756
 757         /* Advance past L2-specific header, if present */
 758         ptr += session->l2specific_len;
 759
 760         if (L2TP_SKB_CB(skb)->has_seq) {
 761                 /* Received a packet with sequence numbers. If we're the LNS,
 762                  * check if we sre sending sequence numbers and if not,
 763                  * configure it so.
 764                  */
 765                 if ((!session->lns_mode) && (!session->send_seq)) {
 766                         l2tp_info(session, L2TP_MSG_SEQ,
 767                                   "%s: requested to enable seq numbers by LNS\n",
 768                                   session->name);
 769                         session->send_seq = 1;
 770                         l2tp_session_set_header_len(session, tunnel->version);
 771                 }
 772         } else {
 773                 /* No sequence numbers.
 774                  * If user has configured mandatory sequence numbers, discard.
 775                  */
 776                 if (session->recv_seq) {
 777                         l2tp_warn(session, L2TP_MSG_SEQ,
 778                                   "%s: recv data has no seq numbers when required. Discarding.\n",
 779                                   session->name);
 780                         atomic_long_inc(&session->stats.rx_seq_discards);
 781                         goto discard;
 782                 }
 783
 784                 /* If we're the LAC and we're sending sequence numbers, the
 785                  * LNS has requested that we no longer send sequence numbers.
 786                  * If we're the LNS and we're sending sequence numbers, the
 787                  * LAC is broken. Discard the frame.
 788                  */
 789                 if ((!session->lns_mode) && (session->send_seq)) {
 790                         l2tp_info(session, L2TP_MSG_SEQ,
 791                                   "%s: requested to disable seq numbers by LNS\n",
 792                                   session->name);
 793                         session->send_seq = 0;
 794                         l2tp_session_set_header_len(session, tunnel->version);
 795                 } else if (session->send_seq) {
 796                         l2tp_warn(session, L2TP_MSG_SEQ,
 797                                   "%s: recv data has no seq numbers when required. Discarding.\n",
 798                                   session->name);
 799                         atomic_long_inc(&session->stats.rx_seq_discards);
 800                         goto discard;
 801                 }
 802         }
 803
 804         /* Session data offset is handled differently for L2TPv2 and
 805          * L2TPv3. For L2TPv2, there is an optional 16-bit value in
 806          * the header. For L2TPv3, the offset is negotiated using AVPs
 807          * in the session setup control protocol.
 808          */
 809         if (tunnel->version == L2TP_HDR_VER_2) {
 810                 /* If offset bit set, skip it. */
 811                 if (hdrflags & L2TP_HDRFLAG_O) {
 812                         offset = ntohs(*(__be16 *)ptr);
 813                         ptr += 2 + offset;
 814                 }
 815         } else
 816                 ptr += session->offset;
 817
 818         offset = ptr - optr;
 819         if (!pskb_may_pull(skb, offset))
 820                 goto discard;
 821
 822         __skb_pull(skb, offset);
 823
 824         /* If caller wants to process the payload before we queue the
 825          * packet, do so now.
 826          */
 827         if (payload_hook)
 828                 if ((*payload_hook)(skb))
 829                         goto discard;
 830
 831         /* Prepare skb for adding to the session's reorder_q.  Hold
 832          * packets for max reorder_timeout or 1 second if not
 833          * reordering.
 834          */
 835         L2TP_SKB_CB(skb)->length = length;
 836         L2TP_SKB_CB(skb)->expires = jiffies +
 837                 (session->reorder_timeout ? session->reorder_timeout : HZ);
 838
 839         /* Add packet to the session's receive queue. Reordering is done here, if
 840          * enabled. Saved L2TP protocol info is stored in skb->sb[].
 841          */
 842         if (L2TP_SKB_CB(skb)->has_seq) {
 843                 if (l2tp_recv_data_seq(session, skb))
 844                         goto discard;
 845         } else {
 846                 /* No sequence numbers. Add the skb to the tail of the
 847                  * reorder queue. This ensures that it will be
 848                  * delivered after all previous sequenced skbs.
 849                  */
 850                 skb_queue_tail(&session->reorder_q, skb);
 851         }
 852
 853         /* Try to dequeue as many skbs from reorder_q as we can. */
 854         l2tp_recv_dequeue(session);
 855
 856         return;
 857
 858 discard:
 859         atomic_long_inc(&session->stats.rx_errors);
 860         kfree_skb(skb);
 861
 862         if (session->deref)
 863                 (*session->deref)(session);
 864 }
 865 EXPORT_SYMBOL(l2tp_recv_common);
 866
 867 /* Drop skbs from the session's reorder_q
 868  */
 869 int l2tp_session_queue_purge(struct l2tp_session *session)
 870 {
 871         struct sk_buff *skb = NULL;
 872         BUG_ON(!session);
 873         BUG_ON(session->magic != L2TP_SESSION_MAGIC);
 874         while ((skb = skb_dequeue(&session->reorder_q))) {
 875                 atomic_long_inc(&session->stats.rx_errors);
 876                 kfree_skb(skb);
 877                 if (session->deref)
 878                         (*session->deref)(session);
 879         }
 880         return 0;
 881 }
 882 EXPORT_SYMBOL_GPL(l2tp_session_queue_purge);
 883
 884 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
 885  * here. The skb is not on a list when we get here.
 886  * Returns 0 if the packet was a data packet and was successfully passed on.
 887  * Returns 1 if the packet was not a good data packet and could not be
 888  * forwarded.  All such packets are passed up to userspace to deal with.
 889  */
 890 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
 891                               int (*payload_hook)(struct sk_buff *skb))
 892 {
 893         struct l2tp_session *session = NULL;
 894         unsigned char *ptr, *optr;
 895         u16 hdrflags;
 896         u32 tunnel_id, session_id;
 897         u16 version;
 898         int length;
 899
 900         /* UDP has verifed checksum */
 901
 902         /* UDP always verifies the packet length. */
 903         __skb_pull(skb, sizeof(struct udphdr));
 904
 905         /* Short packet? */
 906         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
 907                 l2tp_info(tunnel, L2TP_MSG_DATA,
 908                           "%s: recv short packet (len=%d)\n",
 909                           tunnel->name, skb->len);
 910                 goto error;
 911         }
 912
 913         /* Trace packet contents, if enabled */
 914         if (tunnel->debug & L2TP_MSG_DATA) {
 915                 length = min(32u, skb->len);
 916                 if (!pskb_may_pull(skb, length))
 917                         goto error;
 918
 919                 pr_debug("%s: recv\n", tunnel->name);
 920                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
 921         }
 922
 923         /* Point to L2TP header */
 924         optr = ptr = skb->data;
 925
 926         /* Get L2TP header flags */
 927         hdrflags = ntohs(*(__be16 *) ptr);
 928
 929         /* Check protocol version */
 930         version = hdrflags & L2TP_HDR_VER_MASK;
 931         if (version != tunnel->version) {
 932                 l2tp_info(tunnel, L2TP_MSG_DATA,
 933                           "%s: recv protocol version mismatch: got %d expected %d\n",
 934                           tunnel->name, version, tunnel->version);
 935                 goto error;
 936         }
 937
 938         /* Get length of L2TP packet */
 939         length = skb->len;
 940
 941         /* If type is control packet, it is handled by userspace. */
 942         if (hdrflags & L2TP_HDRFLAG_T) {
 943                 l2tp_dbg(tunnel, L2TP_MSG_DATA,
 944                          "%s: recv control packet, len=%d\n",
 945                          tunnel->name, length);
 946                 goto error;
 947         }
 948
 949         /* Skip flags */
 950         ptr += 2;
 951
 952         if (tunnel->version == L2TP_HDR_VER_2) {
 953                 /* If length is present, skip it */
 954                 if (hdrflags & L2TP_HDRFLAG_L)
 955                         ptr += 2;
 956
 957                 /* Extract tunnel and session ID */
 958                 tunnel_id = ntohs(*(__be16 *) ptr);
 959                 ptr += 2;
 960                 session_id = ntohs(*(__be16 *) ptr);
 961                 ptr += 2;
 962         } else {
 963                 ptr += 2;       /* skip reserved bits */
 964                 tunnel_id = tunnel->tunnel_id;
 965                 session_id = ntohl(*(__be32 *) ptr);
 966                 ptr += 4;
 967         }
 968
 969         /* Find the session context */
 970         session = l2tp_session_get(tunnel->l2tp_net, tunnel, session_id, true);
 971         if (!session || !session->recv_skb) {
 972                 if (session) {
 973                         if (session->deref)
 974                                 session->deref(session);
 975                         l2tp_session_dec_refcount(session);
 976                 }
 977
 978                 /* Not found? Pass to userspace to deal with */
 979                 l2tp_info(tunnel, L2TP_MSG_DATA,
 980                           "%s: no session found (%u/%u). Passing up.\n",
 981                           tunnel->name, tunnel_id, session_id);
 982                 goto error;
 983         }
 984
 985         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
 986         l2tp_session_dec_refcount(session);
 987
 988         return 0;
 989
 990 error:
 991         /* Put UDP header back */
 992         __skb_push(skb, sizeof(struct udphdr));
 993
 994         return 1;
 995 }
 996
 997 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
 998  * Return codes:
 999  * 0 : success.
1000  * <0: error
1001  * >0: skb should be passed up to userspace as UDP.
1002  */
1003 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1004 {
1005         struct l2tp_tunnel *tunnel;
1006
1007         tunnel = l2tp_sock_to_tunnel(sk);
1008         if (tunnel == NULL)
1009                 goto pass_up;
1010
1011         l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
1012                  tunnel->name, skb->len);
1013
1014         if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
1015                 goto pass_up_put;
1016
1017         sock_put(sk);
1018         return 0;
1019
1020 pass_up_put:
1021         sock_put(sk);
1022 pass_up:
1023         return 1;
1024 }
1025 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
1026
1027 /************************************************************************
1028  * Transmit handling
1029  ***********************************************************************/
1030
1031 /* Build an L2TP header for the session into the buffer provided.
1032  */
1033 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
1034 {
1035         struct l2tp_tunnel *tunnel = session->tunnel;
1036         __be16 *bufp = buf;
1037         __be16 *optr = buf;
1038         u16 flags = L2TP_HDR_VER_2;
1039         u32 tunnel_id = tunnel->peer_tunnel_id;
1040         u32 session_id = session->peer_session_id;
1041
1042         if (session->send_seq)
1043                 flags |= L2TP_HDRFLAG_S;
1044
1045         /* Setup L2TP header. */
1046         *bufp++ = htons(flags);
1047         *bufp++ = htons(tunnel_id);
1048         *bufp++ = htons(session_id);
1049         if (session->send_seq) {
1050                 *bufp++ = htons(session->ns);
1051                 *bufp++ = 0;
1052                 session->ns++;
1053                 session->ns &= 0xffff;
1054                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
1055                          session->name, session->ns);
1056         }
1057
1058         return bufp - optr;
1059 }
1060
1061 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
1062 {
1063         struct l2tp_tunnel *tunnel = session->tunnel;
1064         char *bufp = buf;
1065         char *optr = bufp;
1066
1067         /* Setup L2TP header. The header differs slightly for UDP and
1068          * IP encapsulations. For UDP, there is 4 bytes of flags.
1069          */
1070         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1071                 u16 flags = L2TP_HDR_VER_3;
1072                 *((__be16 *) bufp) = htons(flags);
1073                 bufp += 2;
1074                 *((__be16 *) bufp) = 0;
1075                 bufp += 2;
1076         }
1077
1078         *((__be32 *) bufp) = htonl(session->peer_session_id);
1079         bufp += 4;
1080         if (session->cookie_len) {
1081                 memcpy(bufp, &session->cookie[0], session->cookie_len);
1082                 bufp += session->cookie_len;
1083         }
1084         if (session->l2specific_len) {
1085                 if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
1086                         u32 l2h = 0;
1087                         if (session->send_seq) {
1088                                 l2h = 0x40000000 | session->ns;
1089                                 session->ns++;
1090                                 session->ns &= 0xffffff;
1091                                 l2tp_dbg(session, L2TP_MSG_SEQ,
1092                                          "%s: updated ns to %u\n",
1093                                          session->name, session->ns);
1094                         }
1095
1096                         *((__be32 *) bufp) = htonl(l2h);
1097                 }
1098                 bufp += session->l2specific_len;
1099         }
1100         if (session->offset)
1101                 bufp += session->offset;
1102
1103         return bufp - optr;
1104 }
1105
1106 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1107                           struct flowi *fl, size_t data_len)
1108 {
1109         struct l2tp_tunnel *tunnel = session->tunnel;
1110         unsigned int len = skb->len;
1111         int error;
1112
1113         /* Debug */
1114         if (session->send_seq)
1115                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes, ns=%u\n",
1116                          session->name, data_len, session->ns - 1);
1117         else
1118                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes\n",
1119                          session->name, data_len);
1120
1121         if (session->debug & L2TP_MSG_DATA) {
1122                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1123                 unsigned char *datap = skb->data + uhlen;
1124
1125                 pr_debug("%s: xmit\n", session->name);
1126                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1127                                      datap, min_t(size_t, 32, len - uhlen));
1128         }
1129
1130         /* Queue the packet to IP for output */
1131         skb->ignore_df = 1;
1132 #if IS_ENABLED(CONFIG_IPV6)
1133         if (tunnel->sock->sk_family == PF_INET6 && !tunnel->v4mapped)
1134                 error = inet6_csk_xmit(tunnel->sock, skb, NULL);
1135         else
1136 #endif
1137                 error = ip_queue_xmit(tunnel->sock, skb, fl);
1138
1139         /* Update stats */
1140         if (error >= 0) {
1141                 atomic_long_inc(&tunnel->stats.tx_packets);
1142                 atomic_long_add(len, &tunnel->stats.tx_bytes);
1143                 atomic_long_inc(&session->stats.tx_packets);
1144                 atomic_long_add(len, &session->stats.tx_bytes);
1145         } else {
1146                 atomic_long_inc(&tunnel->stats.tx_errors);
1147                 atomic_long_inc(&session->stats.tx_errors);
1148         }
1149
1150         return 0;
1151 }
1152
1153 /* If caller requires the skb to have a ppp header, the header must be
1154  * inserted in the skb data before calling this function.
1155  */
1156 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1157 {
1158         int data_len = skb->len;
1159         struct l2tp_tunnel *tunnel = session->tunnel;
1160         struct sock *sk = tunnel->sock;
1161         struct flowi *fl;
1162         struct udphdr *uh;
1163         struct inet_sock *inet;
1164         int headroom;
1165         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1166         int udp_len;
1167         int ret = NET_XMIT_SUCCESS;
1168
1169         /* Check that there's enough headroom in the skb to insert IP,
1170          * UDP and L2TP headers. If not enough, expand it to
1171          * make room. Adjust truesize.
1172          */
1173         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1174                 uhlen + hdr_len;
1175         if (skb_cow_head(skb, headroom)) {
1176                 kfree_skb(skb);
1177                 return NET_XMIT_DROP;
1178         }
1179
1180         /* Setup L2TP header */
1181         session->build_header(session, __skb_push(skb, hdr_len));
1182
1183         /* Reset skb netfilter state */
1184         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1185         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1186                               IPSKB_REROUTED);
1187         nf_reset(skb);
1188
1189         bh_lock_sock(sk);
1190         if (sock_owned_by_user(sk)) {
1191                 kfree_skb(skb);
1192                 ret = NET_XMIT_DROP;
1193                 goto out_unlock;
1194         }
1195
1196         /* Get routing info from the tunnel socket */
1197         skb_dst_drop(skb);
1198         skb_dst_set(skb, dst_clone(__sk_dst_check(sk, 0)));
1199
1200         inet = inet_sk(sk);
1201         fl = &inet->cork.fl;
1202         switch (tunnel->encap) {
1203         case L2TP_ENCAPTYPE_UDP:
1204                 /* Setup UDP header */
1205                 __skb_push(skb, sizeof(*uh));
1206                 skb_reset_transport_header(skb);
1207                 uh = udp_hdr(skb);
1208                 uh->source = inet->inet_sport;
1209                 uh->dest = inet->inet_dport;
1210                 udp_len = uhlen + hdr_len + data_len;
1211                 uh->len = htons(udp_len);
1212
1213                 /* Calculate UDP checksum if configured to do so */
1214 #if IS_ENABLED(CONFIG_IPV6)
1215                 if (sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1216                         udp6_set_csum(udp_get_no_check6_tx(sk),
1217                                       skb, &inet6_sk(sk)->saddr,
1218                                       &sk->sk_v6_daddr, udp_len);
1219                 else
1220 #endif
1221                 udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr,
1222                              inet->inet_daddr, udp_len);
1223                 break;
1224
1225         case L2TP_ENCAPTYPE_IP:
1226                 break;
1227         }
1228
1229         l2tp_xmit_core(session, skb, fl, data_len);
1230 out_unlock:
1231         bh_unlock_sock(sk);
1232
1233         return ret;
1234 }
1235 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1236
1237 /*****************************************************************************
1238  * Tinnel and session create/destroy.
1239  *****************************************************************************/
1240
1241 /* Tunnel socket destruct hook.
1242  * The tunnel context is deleted only when all session sockets have been
1243  * closed.
1244  */
1245 static void l2tp_tunnel_destruct(struct sock *sk)
1246 {
1247         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1248         struct l2tp_net *pn;
1249
1250         if (tunnel == NULL)
1251                 goto end;
1252
1253         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1254
1255
1256         /* Disable udp encapsulation */
1257         switch (tunnel->encap) {
1258         case L2TP_ENCAPTYPE_UDP:
1259                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1260                 (udp_sk(sk))->encap_type = 0;
1261                 (udp_sk(sk))->encap_rcv = NULL;
1262                 (udp_sk(sk))->encap_destroy = NULL;
1263                 break;
1264         case L2TP_ENCAPTYPE_IP:
1265                 break;
1266         }
1267
1268         /* Remove hooks into tunnel socket */
1269         sk->sk_destruct = tunnel->old_sk_destruct;
1270         sk->sk_user_data = NULL;
1271
1272         /* Remove the tunnel struct from the tunnel list */
1273         pn = l2tp_pernet(tunnel->l2tp_net);
1274         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1275         list_del_rcu(&tunnel->list);
1276         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1277         atomic_dec(&l2tp_tunnel_count);
1278
1279         l2tp_tunnel_closeall(tunnel);
1280
1281         tunnel->sock = NULL;
1282         l2tp_tunnel_dec_refcount(tunnel);
1283
1284         /* Call the original destructor */
1285         if (sk->sk_destruct)
1286                 (*sk->sk_destruct)(sk);
1287 end:
1288         return;
1289 }
1290
1291 /* When the tunnel is closed, all the attached sessions need to go too.
1292  */
1293 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1294 {
1295         int hash;
1296         struct hlist_node *walk;
1297         struct hlist_node *tmp;
1298         struct l2tp_session *session;
1299
1300         BUG_ON(tunnel == NULL);
1301
1302         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1303                   tunnel->name);
1304
1305         write_lock_bh(&tunnel->hlist_lock);
1306         tunnel->acpt_newsess = false;
1307         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1308 again:
1309                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1310                         session = hlist_entry(walk, struct l2tp_session, hlist);
1311
1312                         l2tp_info(session, L2TP_MSG_CONTROL,
1313                                   "%s: closing session\n", session->name);
1314
1315                         hlist_del_init(&session->hlist);
1316
1317                         if (session->ref != NULL)
1318                                 (*session->ref)(session);
1319
1320                         write_unlock_bh(&tunnel->hlist_lock);
1321
1322                         __l2tp_session_unhash(session);
1323                         l2tp_session_queue_purge(session);
1324
1325                         if (session->session_close != NULL)
1326                                 (*session->session_close)(session);
1327
1328                         if (session->deref != NULL)
1329                                 (*session->deref)(session);
1330
1331                         l2tp_session_dec_refcount(session);
1332
1333                         write_lock_bh(&tunnel->hlist_lock);
1334
1335                         /* Now restart from the beginning of this hash
1336                          * chain.  We always remove a session from the
1337                          * list so we are guaranteed to make forward
1338                          * progress.
1339                          */
1340                         goto again;
1341                 }
1342         }
1343         write_unlock_bh(&tunnel->hlist_lock);
1344 }
1345 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1346
1347 /* Tunnel socket destroy hook for UDP encapsulation */
1348 static void l2tp_udp_encap_destroy(struct sock *sk)
1349 {
1350         struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
1351         if (tunnel) {
1352                 l2tp_tunnel_closeall(tunnel);
1353                 sock_put(sk);
1354         }
1355 }
1356
1357 /* Workqueue tunnel deletion function */
1358 static void l2tp_tunnel_del_work(struct work_struct *work)
1359 {
1360         struct l2tp_tunnel *tunnel = NULL;
1361         struct socket *sock = NULL;
1362         struct sock *sk = NULL;
1363
1364         tunnel = container_of(work, struct l2tp_tunnel, del_work);
1365
1366         l2tp_tunnel_closeall(tunnel);
1367
1368         sk = l2tp_tunnel_sock_lookup(tunnel);
1369         if (!sk)
1370                 goto out;
1371
1372         sock = sk->sk_socket;
1373
1374         /* If the tunnel socket was created by userspace, then go through the
1375          * inet layer to shut the socket down, and let userspace close it.
1376          * Otherwise, if we created the socket directly within the kernel, use
1377          * the sk API to release it here.
1378          * In either case the tunnel resources are freed in the socket
1379          * destructor when the tunnel socket goes away.
1380          */
1381         if (tunnel->fd >= 0) {
1382                 if (sock)
1383                         inet_shutdown(sock, 2);
1384         } else {
1385                 if (sock) {
1386                         kernel_sock_shutdown(sock, SHUT_RDWR);
1387                         sock_release(sock);
1388                 }
1389         }
1390
1391         l2tp_tunnel_sock_put(sk);
1392 out:
1393         l2tp_tunnel_dec_refcount(tunnel);
1394 }
1395
1396 /* Create a socket for the tunnel, if one isn't set up by
1397  * userspace. This is used for static tunnels where there is no
1398  * managing L2TP daemon.
1399  *
1400  * Since we don't want these sockets to keep a namespace alive by
1401  * themselves, we drop the socket's namespace refcount after creation.
1402  * These sockets are freed when the namespace exits using the pernet
1403  * exit hook.
1404  */
1405 static int l2tp_tunnel_sock_create(struct net *net,
1406                                 u32 tunnel_id,
1407                                 u32 peer_tunnel_id,
1408                                 struct l2tp_tunnel_cfg *cfg,
1409                                 struct socket **sockp)
1410 {
1411         int err = -EINVAL;
1412         struct socket *sock = NULL;
1413         struct udp_port_cfg udp_conf;
1414
1415         switch (cfg->encap) {
1416         case L2TP_ENCAPTYPE_UDP:
1417                 memset(&udp_conf, 0, sizeof(udp_conf));
1418
1419 #if IS_ENABLED(CONFIG_IPV6)
1420                 if (cfg->local_ip6 && cfg->peer_ip6) {
1421                         udp_conf.family = AF_INET6;
1422                         memcpy(&udp_conf.local_ip6, cfg->local_ip6,
1423                                sizeof(udp_conf.local_ip6));
1424                         memcpy(&udp_conf.peer_ip6, cfg->peer_ip6,
1425                                sizeof(udp_conf.peer_ip6));
1426                         udp_conf.use_udp6_tx_checksums =
1427                           ! cfg->udp6_zero_tx_checksums;
1428                         udp_conf.use_udp6_rx_checksums =
1429                           ! cfg->udp6_zero_rx_checksums;
1430                 } else
1431 #endif
1432                 {
1433                         udp_conf.family = AF_INET;
1434                         udp_conf.local_ip = cfg->local_ip;
1435                         udp_conf.peer_ip = cfg->peer_ip;
1436                         udp_conf.use_udp_checksums = cfg->use_udp_checksums;
1437                 }
1438
1439                 udp_conf.local_udp_port = htons(cfg->local_udp_port);
1440                 udp_conf.peer_udp_port = htons(cfg->peer_udp_port);
1441
1442                 err = udp_sock_create(net, &udp_conf, &sock);
1443                 if (err < 0)
1444                         goto out;
1445
1446                 break;
1447
1448         case L2TP_ENCAPTYPE_IP:
1449 #if IS_ENABLED(CONFIG_IPV6)
1450                 if (cfg->local_ip6 && cfg->peer_ip6) {
1451                         struct sockaddr_l2tpip6 ip6_addr = {0};
1452
1453                         err = sock_create_kern(net, AF_INET6, SOCK_DGRAM,
1454                                           IPPROTO_L2TP, &sock);
1455                         if (err < 0)
1456                                 goto out;
1457
1458                         ip6_addr.l2tp_family = AF_INET6;
1459                         memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1460                                sizeof(ip6_addr.l2tp_addr));
1461                         ip6_addr.l2tp_conn_id = tunnel_id;
1462                         err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1463                                           sizeof(ip6_addr));
1464                         if (err < 0)
1465                                 goto out;
1466
1467                         ip6_addr.l2tp_family = AF_INET6;
1468                         memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1469                                sizeof(ip6_addr.l2tp_addr));
1470                         ip6_addr.l2tp_conn_id = peer_tunnel_id;
1471                         err = kernel_connect(sock,
1472                                              (struct sockaddr *) &ip6_addr,
1473                                              sizeof(ip6_addr), 0);
1474                         if (err < 0)
1475                                 goto out;
1476                 } else
1477 #endif
1478                 {
1479                         struct sockaddr_l2tpip ip_addr = {0};
1480
1481                         err = sock_create_kern(net, AF_INET, SOCK_DGRAM,
1482                                           IPPROTO_L2TP, &sock);
1483                         if (err < 0)
1484                                 goto out;
1485
1486                         ip_addr.l2tp_family = AF_INET;
1487                         ip_addr.l2tp_addr = cfg->local_ip;
1488                         ip_addr.l2tp_conn_id = tunnel_id;
1489                         err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1490                                           sizeof(ip_addr));
1491                         if (err < 0)
1492                                 goto out;
1493
1494                         ip_addr.l2tp_family = AF_INET;
1495                         ip_addr.l2tp_addr = cfg->peer_ip;
1496                         ip_addr.l2tp_conn_id = peer_tunnel_id;
1497                         err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1498                                              sizeof(ip_addr), 0);
1499                         if (err < 0)
1500                                 goto out;
1501                 }
1502                 break;
1503
1504         default:
1505                 goto out;
1506         }
1507
1508 out:
1509         *sockp = sock;
1510         if ((err < 0) && sock) {
1511                 kernel_sock_shutdown(sock, SHUT_RDWR);
1512                 sock_release(sock);
1513                 *sockp = NULL;
1514         }
1515
1516         return err;
1517 }
1518
1519 static struct lock_class_key l2tp_socket_class;
1520
1521 int l2tp_tunnel_create(struct net *net, int fd, int version, u32 tunnel_id, u32 peer_tunnel_id, struct l2tp_tunnel_cfg *cfg, struct l2tp_tunnel **tunnelp)
1522 {
1523         struct l2tp_tunnel *tunnel = NULL;
1524         int err;
1525         struct socket *sock = NULL;
1526         struct sock *sk = NULL;
1527         struct l2tp_net *pn;
1528         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1529
1530         /* Get the tunnel socket from the fd, which was opened by
1531          * the userspace L2TP daemon. If not specified, create a
1532          * kernel socket.
1533          */
1534         if (fd < 0) {
1535                 err = l2tp_tunnel_sock_create(net, tunnel_id, peer_tunnel_id,
1536                                 cfg, &sock);
1537                 if (err < 0)
1538                         goto err;
1539         } else {
1540                 sock = sockfd_lookup(fd, &err);
1541                 if (!sock) {
1542                         pr_err("tunl %u: sockfd_lookup(fd=%d) returned %d\n",
1543                                tunnel_id, fd, err);
1544                         err = -EBADF;
1545                         goto err;
1546                 }
1547
1548                 /* Reject namespace mismatches */
1549                 if (!net_eq(sock_net(sock->sk), net)) {
1550                         pr_err("tunl %u: netns mismatch\n", tunnel_id);
1551                         err = -EINVAL;
1552                         goto err;
1553                 }
1554         }
1555
1556         sk = sock->sk;
1557
1558         if (cfg != NULL)
1559                 encap = cfg->encap;
1560
1561         /* Quick sanity checks */
1562         switch (encap) {
1563         case L2TP_ENCAPTYPE_UDP:
1564                 err = -EPROTONOSUPPORT;
1565                 if (sk->sk_protocol != IPPROTO_UDP) {
1566                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1567                                tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1568                         goto err;
1569                 }
1570                 break;
1571         case L2TP_ENCAPTYPE_IP:
1572                 err = -EPROTONOSUPPORT;
1573                 if (sk->sk_protocol != IPPROTO_L2TP) {
1574                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1575                                tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1576                         goto err;
1577                 }
1578                 break;
1579         }
1580
1581         /* Check if this socket has already been prepped */
1582         tunnel = l2tp_tunnel(sk);
1583         if (tunnel != NULL) {
1584                 /* This socket has already been prepped */
1585                 err = -EBUSY;
1586                 goto err;
1587         }
1588
1589         tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1590         if (tunnel == NULL) {
1591                 err = -ENOMEM;
1592                 goto err;
1593         }
1594
1595         tunnel->version = version;
1596         tunnel->tunnel_id = tunnel_id;
1597         tunnel->peer_tunnel_id = peer_tunnel_id;
1598         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1599
1600         tunnel->magic = L2TP_TUNNEL_MAGIC;
1601         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1602         rwlock_init(&tunnel->hlist_lock);
1603         tunnel->acpt_newsess = true;
1604
1605         /* The net we belong to */
1606         tunnel->l2tp_net = net;
1607         pn = l2tp_pernet(net);
1608
1609         if (cfg != NULL)
1610                 tunnel->debug = cfg->debug;
1611
1612 #if IS_ENABLED(CONFIG_IPV6)
1613         if (sk->sk_family == PF_INET6) {
1614                 struct ipv6_pinfo *np = inet6_sk(sk);
1615
1616                 if (ipv6_addr_v4mapped(&np->saddr) &&
1617                     ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
1618                         struct inet_sock *inet = inet_sk(sk);
1619
1620                         tunnel->v4mapped = true;
1621                         inet->inet_saddr = np->saddr.s6_addr32[3];
1622                         inet->inet_rcv_saddr = sk->sk_v6_rcv_saddr.s6_addr32[3];
1623                         inet->inet_daddr = sk->sk_v6_daddr.s6_addr32[3];
1624                 } else {
1625                         tunnel->v4mapped = false;
1626                 }
1627         }
1628 #endif
1629
1630         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1631         tunnel->encap = encap;
1632         if (encap == L2TP_ENCAPTYPE_UDP) {
1633                 struct udp_tunnel_sock_cfg udp_cfg = { };
1634
1635                 udp_cfg.sk_user_data = tunnel;
1636                 udp_cfg.encap_type = UDP_ENCAP_L2TPINUDP;
1637                 udp_cfg.encap_rcv = l2tp_udp_encap_recv;
1638                 udp_cfg.encap_destroy = l2tp_udp_encap_destroy;
1639
1640                 setup_udp_tunnel_sock(net, sock, &udp_cfg);
1641         } else {
1642                 sk->sk_user_data = tunnel;
1643         }
1644
1645         /* Hook on the tunnel socket destructor so that we can cleanup
1646          * if the tunnel socket goes away.
1647          */
1648         tunnel->old_sk_destruct = sk->sk_destruct;
1649         sk->sk_destruct = &l2tp_tunnel_destruct;
1650         tunnel->sock = sk;
1651         tunnel->fd = fd;
1652         lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class, "l2tp_sock");
1653
1654         sk->sk_allocation = GFP_ATOMIC;
1655
1656         /* Init delete workqueue struct */
1657         INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1658
1659         /* Add tunnel to our list */
1660         INIT_LIST_HEAD(&tunnel->list);
1661         atomic_inc(&l2tp_tunnel_count);
1662
1663         /* Bump the reference count. The tunnel context is deleted
1664          * only when this drops to zero. Must be done before list insertion
1665          */
1666         refcount_set(&tunnel->ref_count, 1);
1667         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1668         list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1669         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1670
1671         err = 0;
1672 err:
1673         if (tunnelp)
1674                 *tunnelp = tunnel;
1675
1676         /* If tunnel's socket was created by the kernel, it doesn't
1677          *  have a file.
1678          */
1679         if (sock && sock->file)
1680                 sockfd_put(sock);
1681
1682         return err;
1683 }
1684 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1685
1686 /* This function is used by the netlink TUNNEL_DELETE command.
1687  */
1688 int l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1689 {
1690         l2tp_tunnel_inc_refcount(tunnel);
1691         if (false == queue_work(l2tp_wq, &tunnel->del_work)) {
1692                 l2tp_tunnel_dec_refcount(tunnel);
1693                 return 1;
1694         }
1695         return 0;
1696 }
1697 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1698
1699 /* Really kill the session.
1700  */
1701 void l2tp_session_free(struct l2tp_session *session)
1702 {
1703         struct l2tp_tunnel *tunnel = session->tunnel;
1704
1705         BUG_ON(refcount_read(&session->ref_count) != 0);
1706
1707         if (tunnel) {
1708                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1709                 if (session->session_id != 0)
1710                         atomic_dec(&l2tp_session_count);
1711                 sock_put(tunnel->sock);
1712                 session->tunnel = NULL;
1713                 l2tp_tunnel_dec_refcount(tunnel);
1714         }
1715
1716         kfree(session);
1717 }
1718 EXPORT_SYMBOL_GPL(l2tp_session_free);
1719
1720 /* Remove an l2tp session from l2tp_core's hash lists.
1721  * Provides a tidyup interface for pseudowire code which can't just route all
1722  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1723  * callback.
1724  */
1725 void __l2tp_session_unhash(struct l2tp_session *session)
1726 {
1727         struct l2tp_tunnel *tunnel = session->tunnel;
1728
1729         /* Remove the session from core hashes */
1730         if (tunnel) {
1731                 /* Remove from the per-tunnel hash */
1732                 write_lock_bh(&tunnel->hlist_lock);
1733                 hlist_del_init(&session->hlist);
1734                 write_unlock_bh(&tunnel->hlist_lock);
1735
1736                 /* For L2TPv3 we have a per-net hash: remove from there, too */
1737                 if (tunnel->version != L2TP_HDR_VER_2) {
1738                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1739                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1740                         hlist_del_init_rcu(&session->global_hlist);
1741                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1742                         synchronize_rcu();
1743                 }
1744         }
1745 }
1746 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1747
1748 /* This function is used by the netlink SESSION_DELETE command and by
1749    pseudowire modules.
1750  */
1751 int l2tp_session_delete(struct l2tp_session *session)
1752 {
1753         if (session->ref)
1754                 (*session->ref)(session);
1755         __l2tp_session_unhash(session);
1756         l2tp_session_queue_purge(session);
1757         if (session->session_close != NULL)
1758                 (*session->session_close)(session);
1759         if (session->deref)
1760                 (*session->deref)(session);
1761         l2tp_session_dec_refcount(session);
1762         return 0;
1763 }
1764 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1765
1766 /* We come here whenever a session's send_seq, cookie_len or
1767  * l2specific_len parameters are set.
1768  */
1769 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1770 {
1771         if (version == L2TP_HDR_VER_2) {
1772                 session->hdr_len = 6;
1773                 if (session->send_seq)
1774                         session->hdr_len += 4;
1775         } else {
1776                 session->hdr_len = 4 + session->cookie_len + session->l2specific_len + session->offset;
1777                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1778                         session->hdr_len += 4;
1779         }
1780
1781 }
1782 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1783
1784 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id, u32 peer_session_id, struct l2tp_session_cfg *cfg)
1785 {
1786         struct l2tp_session *session;
1787         int err;
1788
1789         session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1790         if (session != NULL) {
1791                 session->magic = L2TP_SESSION_MAGIC;
1792                 session->tunnel = tunnel;
1793
1794                 session->session_id = session_id;
1795                 session->peer_session_id = peer_session_id;
1796                 session->nr = 0;
1797                 if (tunnel->version == L2TP_HDR_VER_2)
1798                         session->nr_max = 0xffff;
1799                 else
1800                         session->nr_max = 0xffffff;
1801                 session->nr_window_size = session->nr_max / 2;
1802                 session->nr_oos_count_max = 4;
1803
1804                 /* Use NR of first received packet */
1805                 session->reorder_skip = 1;
1806
1807                 sprintf(&session->name[0], "sess %u/%u",
1808                         tunnel->tunnel_id, session->session_id);
1809
1810                 skb_queue_head_init(&session->reorder_q);
1811
1812                 INIT_HLIST_NODE(&session->hlist);
1813                 INIT_HLIST_NODE(&session->global_hlist);
1814
1815                 /* Inherit debug options from tunnel */
1816                 session->debug = tunnel->debug;
1817
1818                 if (cfg) {
1819                         session->pwtype = cfg->pw_type;
1820                         session->debug = cfg->debug;
1821                         session->mtu = cfg->mtu;
1822                         session->mru = cfg->mru;
1823                         session->send_seq = cfg->send_seq;
1824                         session->recv_seq = cfg->recv_seq;
1825                         session->lns_mode = cfg->lns_mode;
1826                         session->reorder_timeout = cfg->reorder_timeout;
1827                         session->offset = cfg->offset;
1828                         session->l2specific_type = cfg->l2specific_type;
1829                         session->l2specific_len = cfg->l2specific_len;
1830                         session->cookie_len = cfg->cookie_len;
1831                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1832                         session->peer_cookie_len = cfg->peer_cookie_len;
1833                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1834                 }
1835
1836                 if (tunnel->version == L2TP_HDR_VER_2)
1837                         session->build_header = l2tp_build_l2tpv2_header;
1838                 else
1839                         session->build_header = l2tp_build_l2tpv3_header;
1840
1841                 l2tp_session_set_header_len(session, tunnel->version);
1842
1843                 refcount_set(&session->ref_count, 1);
1844
1845                 err = l2tp_session_add_to_tunnel(tunnel, session);
1846                 if (err) {
1847                         kfree(session);
1848
1849                         return ERR_PTR(err);
1850                 }
1851
1852                 /* Ignore management session in session count value */
1853                 if (session->session_id != 0)
1854                         atomic_inc(&l2tp_session_count);
1855
1856                 return session;
1857         }
1858
1859         return ERR_PTR(-ENOMEM);
1860 }
1861 EXPORT_SYMBOL_GPL(l2tp_session_create);
1862
1863 /*****************************************************************************
1864  * Init and cleanup
1865  *****************************************************************************/
1866
1867 static __net_init int l2tp_init_net(struct net *net)
1868 {
1869         struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1870         int hash;
1871
1872         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1873         spin_lock_init(&pn->l2tp_tunnel_list_lock);
1874
1875         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1876                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1877
1878         spin_lock_init(&pn->l2tp_session_hlist_lock);
1879
1880         return 0;
1881 }
1882
1883 static __net_exit void l2tp_exit_net(struct net *net)
1884 {
1885         struct l2tp_net *pn = l2tp_pernet(net);
1886         struct l2tp_tunnel *tunnel = NULL;
1887
1888         rcu_read_lock_bh();
1889         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1890                 (void)l2tp_tunnel_delete(tunnel);
1891         }
1892         rcu_read_unlock_bh();
1893
1894         flush_workqueue(l2tp_wq);
1895         rcu_barrier();
1896 }
1897
1898 static struct pernet_operations l2tp_net_ops = {
1899         .init = l2tp_init_net,
1900         .exit = l2tp_exit_net,
1901         .id   = &l2tp_net_id,
1902         .size = sizeof(struct l2tp_net),
1903 };
1904
1905 static int __init l2tp_init(void)
1906 {
1907         int rc = 0;
1908
1909         rc = register_pernet_device(&l2tp_net_ops);
1910         if (rc)
1911                 goto out;
1912
1913         l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1914         if (!l2tp_wq) {
1915                 pr_err("alloc_workqueue failed\n");
1916                 unregister_pernet_device(&l2tp_net_ops);
1917                 rc = -ENOMEM;
1918                 goto out;
1919         }
1920
1921         pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
1922
1923 out:
1924         return rc;
1925 }
1926
1927 static void __exit l2tp_exit(void)
1928 {
1929         unregister_pernet_device(&l2tp_net_ops);
1930         if (l2tp_wq) {
1931                 destroy_workqueue(l2tp_wq);
1932                 l2tp_wq = NULL;
1933         }
1934 }
1935
1936 module_init(l2tp_init);
1937 module_exit(l2tp_exit);
1938
1939 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1940 MODULE_DESCRIPTION("L2TP core");
1941 MODULE_LICENSE("GPL");
1942 MODULE_VERSION(L2TP_DRV_VERSION);
1943