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 struct workqueue_struct *l2tp_wq;
 104
 105 /* per-net private data for this module */
 106 static unsigned int l2tp_net_id;
 107 struct l2tp_net {
 108         struct list_head l2tp_tunnel_list;
 109         spinlock_t l2tp_tunnel_list_lock;
 110         struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
 111         spinlock_t l2tp_session_hlist_lock;
 112 };
 113
 114 #if IS_ENABLED(CONFIG_IPV6)
 115 static bool l2tp_sk_is_v6(struct sock *sk)
 116 {
 117         return sk->sk_family == PF_INET6 &&
 118                !ipv6_addr_v4mapped(&sk->sk_v6_daddr);
 119 }
 120 #endif
 121
 122 static inline struct l2tp_tunnel *l2tp_tunnel(struct sock *sk)
 123 {
 124         return sk->sk_user_data;
 125 }
 126
 127 static inline struct l2tp_net *l2tp_pernet(const struct net *net)
 128 {
 129         BUG_ON(!net);
 130
 131         return net_generic(net, l2tp_net_id);
 132 }
 133
 134 /* Session hash global list for L2TPv3.
 135  * The session_id SHOULD be random according to RFC3931, but several
 136  * L2TP implementations use incrementing session_ids.  So we do a real
 137  * hash on the session_id, rather than a simple bitmask.
 138  */
 139 static inline struct hlist_head *
 140 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
 141 {
 142         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
 143
 144 }
 145
 146 /* Session hash list.
 147  * The session_id SHOULD be random according to RFC2661, but several
 148  * L2TP implementations (Cisco and Microsoft) use incrementing
 149  * session_ids.  So we do a real hash on the session_id, rather than a
 150  * simple bitmask.
 151  */
 152 static inline struct hlist_head *
 153 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
 154 {
 155         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
 156 }
 157
 158 void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
 159 {
 160         sock_put(tunnel->sock);
 161         /* the tunnel is freed in the socket destructor */
 162 }
 163 EXPORT_SYMBOL(l2tp_tunnel_free);
 164
 165 /* Lookup a tunnel. A new reference is held on the returned tunnel. */
 166 struct l2tp_tunnel *l2tp_tunnel_get(const struct net *net, u32 tunnel_id)
 167 {
 168         const struct l2tp_net *pn = l2tp_pernet(net);
 169         struct l2tp_tunnel *tunnel;
 170
 171         rcu_read_lock_bh();
 172         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
 173                 if (tunnel->tunnel_id == tunnel_id) {
 174                         l2tp_tunnel_inc_refcount(tunnel);
 175                         rcu_read_unlock_bh();
 176
 177                         return tunnel;
 178                 }
 179         }
 180         rcu_read_unlock_bh();
 181
 182         return NULL;
 183 }
 184 EXPORT_SYMBOL_GPL(l2tp_tunnel_get);
 185
 186 struct l2tp_tunnel *l2tp_tunnel_get_nth(const struct net *net, int nth)
 187 {
 188         const struct l2tp_net *pn = l2tp_pernet(net);
 189         struct l2tp_tunnel *tunnel;
 190         int count = 0;
 191
 192         rcu_read_lock_bh();
 193         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
 194                 if (++count > nth) {
 195                         l2tp_tunnel_inc_refcount(tunnel);
 196                         rcu_read_unlock_bh();
 197                         return tunnel;
 198                 }
 199         }
 200         rcu_read_unlock_bh();
 201
 202         return NULL;
 203 }
 204 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_nth);
 205
 206 struct l2tp_session *l2tp_tunnel_get_session(struct l2tp_tunnel *tunnel,
 207                                              u32 session_id)
 208 {
 209         struct hlist_head *session_list;
 210         struct l2tp_session *session;
 211
 212         session_list = l2tp_session_id_hash(tunnel, session_id);
 213
 214         read_lock_bh(&tunnel->hlist_lock);
 215         hlist_for_each_entry(session, session_list, hlist)
 216                 if (session->session_id == session_id) {
 217                         l2tp_session_inc_refcount(session);
 218                         read_unlock_bh(&tunnel->hlist_lock);
 219
 220                         return session;
 221                 }
 222         read_unlock_bh(&tunnel->hlist_lock);
 223
 224         return NULL;
 225 }
 226 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_session);
 227
 228 struct l2tp_session *l2tp_session_get(const struct net *net, u32 session_id)
 229 {
 230         struct hlist_head *session_list;
 231         struct l2tp_session *session;
 232
 233         session_list = l2tp_session_id_hash_2(l2tp_pernet(net), session_id);
 234
 235         rcu_read_lock_bh();
 236         hlist_for_each_entry_rcu(session, session_list, global_hlist)
 237                 if (session->session_id == session_id) {
 238                         l2tp_session_inc_refcount(session);
 239                         rcu_read_unlock_bh();
 240
 241                         return session;
 242                 }
 243         rcu_read_unlock_bh();
 244
 245         return NULL;
 246 }
 247 EXPORT_SYMBOL_GPL(l2tp_session_get);
 248
 249 struct l2tp_session *l2tp_session_get_nth(struct l2tp_tunnel *tunnel, int nth)
 250 {
 251         int hash;
 252         struct l2tp_session *session;
 253         int count = 0;
 254
 255         read_lock_bh(&tunnel->hlist_lock);
 256         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
 257                 hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
 258                         if (++count > nth) {
 259                                 l2tp_session_inc_refcount(session);
 260                                 read_unlock_bh(&tunnel->hlist_lock);
 261                                 return session;
 262                         }
 263                 }
 264         }
 265
 266         read_unlock_bh(&tunnel->hlist_lock);
 267
 268         return NULL;
 269 }
 270 EXPORT_SYMBOL_GPL(l2tp_session_get_nth);
 271
 272 /* Lookup a session by interface name.
 273  * This is very inefficient but is only used by management interfaces.
 274  */
 275 struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net,
 276                                                 const char *ifname)
 277 {
 278         struct l2tp_net *pn = l2tp_pernet(net);
 279         int hash;
 280         struct l2tp_session *session;
 281
 282         rcu_read_lock_bh();
 283         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
 284                 hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
 285                         if (!strcmp(session->ifname, ifname)) {
 286                                 l2tp_session_inc_refcount(session);
 287                                 rcu_read_unlock_bh();
 288
 289                                 return session;
 290                         }
 291                 }
 292         }
 293
 294         rcu_read_unlock_bh();
 295
 296         return NULL;
 297 }
 298 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname);
 299
 300 int l2tp_session_register(struct l2tp_session *session,
 301                           struct l2tp_tunnel *tunnel)
 302 {
 303         struct l2tp_session *session_walk;
 304         struct hlist_head *g_head;
 305         struct hlist_head *head;
 306         struct l2tp_net *pn;
 307         int err;
 308
 309         head = l2tp_session_id_hash(tunnel, session->session_id);
 310
 311         write_lock_bh(&tunnel->hlist_lock);
 312         if (!tunnel->acpt_newsess) {
 313                 err = -ENODEV;
 314                 goto err_tlock;
 315         }
 316
 317         hlist_for_each_entry(session_walk, head, hlist)
 318                 if (session_walk->session_id == session->session_id) {
 319                         err = -EEXIST;
 320                         goto err_tlock;
 321                 }
 322
 323         if (tunnel->version == L2TP_HDR_VER_3) {
 324                 pn = l2tp_pernet(tunnel->l2tp_net);
 325                 g_head = l2tp_session_id_hash_2(pn, session->session_id);
 326
 327                 spin_lock_bh(&pn->l2tp_session_hlist_lock);
 328
 329                 hlist_for_each_entry(session_walk, g_head, global_hlist)
 330                         if (session_walk->session_id == session->session_id) {
 331                                 err = -EEXIST;
 332                                 goto err_tlock_pnlock;
 333                         }
 334
 335                 l2tp_tunnel_inc_refcount(tunnel);
 336                 hlist_add_head_rcu(&session->global_hlist, g_head);
 337
 338                 spin_unlock_bh(&pn->l2tp_session_hlist_lock);
 339         } else {
 340                 l2tp_tunnel_inc_refcount(tunnel);
 341         }
 342
 343         hlist_add_head(&session->hlist, head);
 344         write_unlock_bh(&tunnel->hlist_lock);
 345
 346         return 0;
 347
 348 err_tlock_pnlock:
 349         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
 350 err_tlock:
 351         write_unlock_bh(&tunnel->hlist_lock);
 352
 353         return err;
 354 }
 355 EXPORT_SYMBOL_GPL(l2tp_session_register);
 356
 357 /*****************************************************************************
 358  * Receive data handling
 359  *****************************************************************************/
 360
 361 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
 362  * number.
 363  */
 364 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
 365 {
 366         struct sk_buff *skbp;
 367         struct sk_buff *tmp;
 368         u32 ns = L2TP_SKB_CB(skb)->ns;
 369
 370         spin_lock_bh(&session->reorder_q.lock);
 371         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
 372                 if (L2TP_SKB_CB(skbp)->ns > ns) {
 373                         __skb_queue_before(&session->reorder_q, skbp, skb);
 374                         l2tp_dbg(session, L2TP_MSG_SEQ,
 375                                  "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
 376                                  session->name, ns, L2TP_SKB_CB(skbp)->ns,
 377                                  skb_queue_len(&session->reorder_q));
 378                         atomic_long_inc(&session->stats.rx_oos_packets);
 379                         goto out;
 380                 }
 381         }
 382
 383         __skb_queue_tail(&session->reorder_q, skb);
 384
 385 out:
 386         spin_unlock_bh(&session->reorder_q.lock);
 387 }
 388
 389 /* Dequeue a single skb.
 390  */
 391 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
 392 {
 393         struct l2tp_tunnel *tunnel = session->tunnel;
 394         int length = L2TP_SKB_CB(skb)->length;
 395
 396         /* We're about to requeue the skb, so return resources
 397          * to its current owner (a socket receive buffer).
 398          */
 399         skb_orphan(skb);
 400
 401         atomic_long_inc(&tunnel->stats.rx_packets);
 402         atomic_long_add(length, &tunnel->stats.rx_bytes);
 403         atomic_long_inc(&session->stats.rx_packets);
 404         atomic_long_add(length, &session->stats.rx_bytes);
 405
 406         if (L2TP_SKB_CB(skb)->has_seq) {
 407                 /* Bump our Nr */
 408                 session->nr++;
 409                 session->nr &= session->nr_max;
 410
 411                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
 412                          session->name, session->nr);
 413         }
 414
 415         /* call private receive handler */
 416         if (session->recv_skb != NULL)
 417                 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
 418         else
 419                 kfree_skb(skb);
 420 }
 421
 422 /* Dequeue skbs from the session's reorder_q, subject to packet order.
 423  * Skbs that have been in the queue for too long are simply discarded.
 424  */
 425 static void l2tp_recv_dequeue(struct l2tp_session *session)
 426 {
 427         struct sk_buff *skb;
 428         struct sk_buff *tmp;
 429
 430         /* If the pkt at the head of the queue has the nr that we
 431          * expect to send up next, dequeue it and any other
 432          * in-sequence packets behind it.
 433          */
 434 start:
 435         spin_lock_bh(&session->reorder_q.lock);
 436         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
 437                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
 438                         atomic_long_inc(&session->stats.rx_seq_discards);
 439                         atomic_long_inc(&session->stats.rx_errors);
 440                         l2tp_dbg(session, L2TP_MSG_SEQ,
 441                                  "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
 442                                  session->name, L2TP_SKB_CB(skb)->ns,
 443                                  L2TP_SKB_CB(skb)->length, session->nr,
 444                                  skb_queue_len(&session->reorder_q));
 445                         session->reorder_skip = 1;
 446                         __skb_unlink(skb, &session->reorder_q);
 447                         kfree_skb(skb);
 448                         continue;
 449                 }
 450
 451                 if (L2TP_SKB_CB(skb)->has_seq) {
 452                         if (session->reorder_skip) {
 453                                 l2tp_dbg(session, L2TP_MSG_SEQ,
 454                                          "%s: advancing nr to next pkt: %u -> %u",
 455                                          session->name, session->nr,
 456                                          L2TP_SKB_CB(skb)->ns);
 457                                 session->reorder_skip = 0;
 458                                 session->nr = L2TP_SKB_CB(skb)->ns;
 459                         }
 460                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
 461                                 l2tp_dbg(session, L2TP_MSG_SEQ,
 462                                          "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
 463                                          session->name, L2TP_SKB_CB(skb)->ns,
 464                                          L2TP_SKB_CB(skb)->length, session->nr,
 465                                          skb_queue_len(&session->reorder_q));
 466                                 goto out;
 467                         }
 468                 }
 469                 __skb_unlink(skb, &session->reorder_q);
 470
 471                 /* Process the skb. We release the queue lock while we
 472                  * do so to let other contexts process the queue.
 473                  */
 474                 spin_unlock_bh(&session->reorder_q.lock);
 475                 l2tp_recv_dequeue_skb(session, skb);
 476                 goto start;
 477         }
 478
 479 out:
 480         spin_unlock_bh(&session->reorder_q.lock);
 481 }
 482
 483 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
 484 {
 485         u32 nws;
 486
 487         if (nr >= session->nr)
 488                 nws = nr - session->nr;
 489         else
 490                 nws = (session->nr_max + 1) - (session->nr - nr);
 491
 492         return nws < session->nr_window_size;
 493 }
 494
 495 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
 496  * acceptable, else non-zero.
 497  */
 498 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
 499 {
 500         if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
 501                 /* Packet sequence number is outside allowed window.
 502                  * Discard it.
 503                  */
 504                 l2tp_dbg(session, L2TP_MSG_SEQ,
 505                          "%s: pkt %u len %d discarded, outside window, nr=%u\n",
 506                          session->name, L2TP_SKB_CB(skb)->ns,
 507                          L2TP_SKB_CB(skb)->length, session->nr);
 508                 goto discard;
 509         }
 510
 511         if (session->reorder_timeout != 0) {
 512                 /* Packet reordering enabled. Add skb to session's
 513                  * reorder queue, in order of ns.
 514                  */
 515                 l2tp_recv_queue_skb(session, skb);
 516                 goto out;
 517         }
 518
 519         /* Packet reordering disabled. Discard out-of-sequence packets, while
 520          * tracking the number if in-sequence packets after the first OOS packet
 521          * is seen. After nr_oos_count_max in-sequence packets, reset the
 522          * sequence number to re-enable packet reception.
 523          */
 524         if (L2TP_SKB_CB(skb)->ns == session->nr) {
 525                 skb_queue_tail(&session->reorder_q, skb);
 526         } else {
 527                 u32 nr_oos = L2TP_SKB_CB(skb)->ns;
 528                 u32 nr_next = (session->nr_oos + 1) & session->nr_max;
 529
 530                 if (nr_oos == nr_next)
 531                         session->nr_oos_count++;
 532                 else
 533                         session->nr_oos_count = 0;
 534
 535                 session->nr_oos = nr_oos;
 536                 if (session->nr_oos_count > session->nr_oos_count_max) {
 537                         session->reorder_skip = 1;
 538                         l2tp_dbg(session, L2TP_MSG_SEQ,
 539                                  "%s: %d oos packets received. Resetting sequence numbers\n",
 540                                  session->name, session->nr_oos_count);
 541                 }
 542                 if (!session->reorder_skip) {
 543                         atomic_long_inc(&session->stats.rx_seq_discards);
 544                         l2tp_dbg(session, L2TP_MSG_SEQ,
 545                                  "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
 546                                  session->name, L2TP_SKB_CB(skb)->ns,
 547                                  L2TP_SKB_CB(skb)->length, session->nr,
 548                                  skb_queue_len(&session->reorder_q));
 549                         goto discard;
 550                 }
 551                 skb_queue_tail(&session->reorder_q, skb);
 552         }
 553
 554 out:
 555         return 0;
 556
 557 discard:
 558         return 1;
 559 }
 560
 561 /* Do receive processing of L2TP data frames. We handle both L2TPv2
 562  * and L2TPv3 data frames here.
 563  *
 564  * L2TPv2 Data Message Header
 565  *
 566  *  0                   1                   2                   3
 567  *  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
 568  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 569  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
 570  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 571  * |           Tunnel ID           |           Session ID          |
 572  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 573  * |             Ns (opt)          |             Nr (opt)          |
 574  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 575  * |      Offset Size (opt)        |    Offset pad... (opt)
 576  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 577  *
 578  * Data frames are marked by T=0. All other fields are the same as
 579  * those in L2TP control frames.
 580  *
 581  * L2TPv3 Data Message Header
 582  *
 583  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 584  * |                      L2TP Session Header                      |
 585  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 586  * |                      L2-Specific Sublayer                     |
 587  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 588  * |                        Tunnel Payload                      ...
 589  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 590  *
 591  * L2TPv3 Session Header Over IP
 592  *
 593  *  0                   1                   2                   3
 594  *  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
 595  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 596  * |                           Session ID                          |
 597  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 598  * |               Cookie (optional, maximum 64 bits)...
 599  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 600  *                                                                 |
 601  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 602  *
 603  * L2TPv3 L2-Specific Sublayer Format
 604  *
 605  *  0                   1                   2                   3
 606  *  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
 607  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 608  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
 609  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 610  *
 611  * Cookie value and sublayer format are negotiated with the peer when
 612  * the session is set up. Unlike L2TPv2, we do not need to parse the
 613  * packet header to determine if optional fields are present.
 614  *
 615  * Caller must already have parsed the frame and determined that it is
 616  * a data (not control) frame before coming here. Fields up to the
 617  * session-id have already been parsed and ptr points to the data
 618  * after the session-id.
 619  */
 620 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
 621                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
 622                       int length)
 623 {
 624         struct l2tp_tunnel *tunnel = session->tunnel;
 625         int offset;
 626         u32 ns, nr;
 627
 628         /* Parse and check optional cookie */
 629         if (session->peer_cookie_len > 0) {
 630                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
 631                         l2tp_info(tunnel, L2TP_MSG_DATA,
 632                                   "%s: cookie mismatch (%u/%u). Discarding.\n",
 633                                   tunnel->name, tunnel->tunnel_id,
 634                                   session->session_id);
 635                         atomic_long_inc(&session->stats.rx_cookie_discards);
 636                         goto discard;
 637                 }
 638                 ptr += session->peer_cookie_len;
 639         }
 640
 641         /* Handle the optional sequence numbers. Sequence numbers are
 642          * in different places for L2TPv2 and L2TPv3.
 643          *
 644          * If we are the LAC, enable/disable sequence numbers under
 645          * the control of the LNS.  If no sequence numbers present but
 646          * we were expecting them, discard frame.
 647          */
 648         ns = nr = 0;
 649         L2TP_SKB_CB(skb)->has_seq = 0;
 650         if (tunnel->version == L2TP_HDR_VER_2) {
 651                 if (hdrflags & L2TP_HDRFLAG_S) {
 652                         ns = ntohs(*(__be16 *) ptr);
 653                         ptr += 2;
 654                         nr = ntohs(*(__be16 *) ptr);
 655                         ptr += 2;
 656
 657                         /* Store L2TP info in the skb */
 658                         L2TP_SKB_CB(skb)->ns = ns;
 659                         L2TP_SKB_CB(skb)->has_seq = 1;
 660
 661                         l2tp_dbg(session, L2TP_MSG_SEQ,
 662                                  "%s: recv data ns=%u, nr=%u, session nr=%u\n",
 663                                  session->name, ns, nr, session->nr);
 664                 }
 665         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
 666                 u32 l2h = ntohl(*(__be32 *) ptr);
 667
 668                 if (l2h & 0x40000000) {
 669                         ns = l2h & 0x00ffffff;
 670
 671                         /* Store L2TP info in the skb */
 672                         L2TP_SKB_CB(skb)->ns = ns;
 673                         L2TP_SKB_CB(skb)->has_seq = 1;
 674
 675                         l2tp_dbg(session, L2TP_MSG_SEQ,
 676                                  "%s: recv data ns=%u, session nr=%u\n",
 677                                  session->name, ns, session->nr);
 678                 }
 679                 ptr += 4;
 680         }
 681
 682         if (L2TP_SKB_CB(skb)->has_seq) {
 683                 /* Received a packet with sequence numbers. If we're the LNS,
 684                  * check if we sre sending sequence numbers and if not,
 685                  * configure it so.
 686                  */
 687                 if ((!session->lns_mode) && (!session->send_seq)) {
 688                         l2tp_info(session, L2TP_MSG_SEQ,
 689                                   "%s: requested to enable seq numbers by LNS\n",
 690                                   session->name);
 691                         session->send_seq = 1;
 692                         l2tp_session_set_header_len(session, tunnel->version);
 693                 }
 694         } else {
 695                 /* No sequence numbers.
 696                  * If user has configured mandatory sequence numbers, discard.
 697                  */
 698                 if (session->recv_seq) {
 699                         l2tp_warn(session, L2TP_MSG_SEQ,
 700                                   "%s: recv data has no seq numbers when required. Discarding.\n",
 701                                   session->name);
 702                         atomic_long_inc(&session->stats.rx_seq_discards);
 703                         goto discard;
 704                 }
 705
 706                 /* If we're the LAC and we're sending sequence numbers, the
 707                  * LNS has requested that we no longer send sequence numbers.
 708                  * If we're the LNS and we're sending sequence numbers, the
 709                  * LAC is broken. Discard the frame.
 710                  */
 711                 if ((!session->lns_mode) && (session->send_seq)) {
 712                         l2tp_info(session, L2TP_MSG_SEQ,
 713                                   "%s: requested to disable seq numbers by LNS\n",
 714                                   session->name);
 715                         session->send_seq = 0;
 716                         l2tp_session_set_header_len(session, tunnel->version);
 717                 } else if (session->send_seq) {
 718                         l2tp_warn(session, L2TP_MSG_SEQ,
 719                                   "%s: recv data has no seq numbers when required. Discarding.\n",
 720                                   session->name);
 721                         atomic_long_inc(&session->stats.rx_seq_discards);
 722                         goto discard;
 723                 }
 724         }
 725
 726         /* Session data offset is defined only for L2TPv2 and is
 727          * indicated by an optional 16-bit value in the header.
 728          */
 729         if (tunnel->version == L2TP_HDR_VER_2) {
 730                 /* If offset bit set, skip it. */
 731                 if (hdrflags & L2TP_HDRFLAG_O) {
 732                         offset = ntohs(*(__be16 *)ptr);
 733                         ptr += 2 + offset;
 734                 }
 735         }
 736
 737         offset = ptr - optr;
 738         if (!pskb_may_pull(skb, offset))
 739                 goto discard;
 740
 741         __skb_pull(skb, offset);
 742
 743         /* Prepare skb for adding to the session's reorder_q.  Hold
 744          * packets for max reorder_timeout or 1 second if not
 745          * reordering.
 746          */
 747         L2TP_SKB_CB(skb)->length = length;
 748         L2TP_SKB_CB(skb)->expires = jiffies +
 749                 (session->reorder_timeout ? session->reorder_timeout : HZ);
 750
 751         /* Add packet to the session's receive queue. Reordering is done here, if
 752          * enabled. Saved L2TP protocol info is stored in skb->sb[].
 753          */
 754         if (L2TP_SKB_CB(skb)->has_seq) {
 755                 if (l2tp_recv_data_seq(session, skb))
 756                         goto discard;
 757         } else {
 758                 /* No sequence numbers. Add the skb to the tail of the
 759                  * reorder queue. This ensures that it will be
 760                  * delivered after all previous sequenced skbs.
 761                  */
 762                 skb_queue_tail(&session->reorder_q, skb);
 763         }
 764
 765         /* Try to dequeue as many skbs from reorder_q as we can. */
 766         l2tp_recv_dequeue(session);
 767
 768         return;
 769
 770 discard:
 771         atomic_long_inc(&session->stats.rx_errors);
 772         kfree_skb(skb);
 773 }
 774 EXPORT_SYMBOL(l2tp_recv_common);
 775
 776 /* Drop skbs from the session's reorder_q
 777  */
 778 static int l2tp_session_queue_purge(struct l2tp_session *session)
 779 {
 780         struct sk_buff *skb = NULL;
 781         BUG_ON(!session);
 782         BUG_ON(session->magic != L2TP_SESSION_MAGIC);
 783         while ((skb = skb_dequeue(&session->reorder_q))) {
 784                 atomic_long_inc(&session->stats.rx_errors);
 785                 kfree_skb(skb);
 786         }
 787         return 0;
 788 }
 789
 790 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
 791  * here. The skb is not on a list when we get here.
 792  * Returns 0 if the packet was a data packet and was successfully passed on.
 793  * Returns 1 if the packet was not a good data packet and could not be
 794  * forwarded.  All such packets are passed up to userspace to deal with.
 795  */
 796 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb)
 797 {
 798         struct l2tp_session *session = NULL;
 799         unsigned char *ptr, *optr;
 800         u16 hdrflags;
 801         u32 tunnel_id, session_id;
 802         u16 version;
 803         int length;
 804
 805         /* UDP has verifed checksum */
 806
 807         /* UDP always verifies the packet length. */
 808         __skb_pull(skb, sizeof(struct udphdr));
 809
 810         /* Short packet? */
 811         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
 812                 l2tp_info(tunnel, L2TP_MSG_DATA,
 813                           "%s: recv short packet (len=%d)\n",
 814                           tunnel->name, skb->len);
 815                 goto error;
 816         }
 817
 818         /* Trace packet contents, if enabled */
 819         if (tunnel->debug & L2TP_MSG_DATA) {
 820                 length = min(32u, skb->len);
 821                 if (!pskb_may_pull(skb, length))
 822                         goto error;
 823
 824                 pr_debug("%s: recv\n", tunnel->name);
 825                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
 826         }
 827
 828         /* Point to L2TP header */
 829         optr = ptr = skb->data;
 830
 831         /* Get L2TP header flags */
 832         hdrflags = ntohs(*(__be16 *) ptr);
 833
 834         /* Check protocol version */
 835         version = hdrflags & L2TP_HDR_VER_MASK;
 836         if (version != tunnel->version) {
 837                 l2tp_info(tunnel, L2TP_MSG_DATA,
 838                           "%s: recv protocol version mismatch: got %d expected %d\n",
 839                           tunnel->name, version, tunnel->version);
 840                 goto error;
 841         }
 842
 843         /* Get length of L2TP packet */
 844         length = skb->len;
 845
 846         /* If type is control packet, it is handled by userspace. */
 847         if (hdrflags & L2TP_HDRFLAG_T) {
 848                 l2tp_dbg(tunnel, L2TP_MSG_DATA,
 849                          "%s: recv control packet, len=%d\n",
 850                          tunnel->name, length);
 851                 goto error;
 852         }
 853
 854         /* Skip flags */
 855         ptr += 2;
 856
 857         if (tunnel->version == L2TP_HDR_VER_2) {
 858                 /* If length is present, skip it */
 859                 if (hdrflags & L2TP_HDRFLAG_L)
 860                         ptr += 2;
 861
 862                 /* Extract tunnel and session ID */
 863                 tunnel_id = ntohs(*(__be16 *) ptr);
 864                 ptr += 2;
 865                 session_id = ntohs(*(__be16 *) ptr);
 866                 ptr += 2;
 867         } else {
 868                 ptr += 2;       /* skip reserved bits */
 869                 tunnel_id = tunnel->tunnel_id;
 870                 session_id = ntohl(*(__be32 *) ptr);
 871                 ptr += 4;
 872         }
 873
 874         /* Find the session context */
 875         session = l2tp_tunnel_get_session(tunnel, session_id);
 876         if (!session || !session->recv_skb) {
 877                 if (session)
 878                         l2tp_session_dec_refcount(session);
 879
 880                 /* Not found? Pass to userspace to deal with */
 881                 l2tp_info(tunnel, L2TP_MSG_DATA,
 882                           "%s: no session found (%u/%u). Passing up.\n",
 883                           tunnel->name, tunnel_id, session_id);
 884                 goto error;
 885         }
 886
 887         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length);
 888         l2tp_session_dec_refcount(session);
 889
 890         return 0;
 891
 892 error:
 893         /* Put UDP header back */
 894         __skb_push(skb, sizeof(struct udphdr));
 895
 896         return 1;
 897 }
 898
 899 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
 900  * Return codes:
 901  * 0 : success.
 902  * <0: error
 903  * >0: skb should be passed up to userspace as UDP.
 904  */
 905 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
 906 {
 907         struct l2tp_tunnel *tunnel;
 908
 909         tunnel = l2tp_tunnel(sk);
 910         if (tunnel == NULL)
 911                 goto pass_up;
 912
 913         l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
 914                  tunnel->name, skb->len);
 915
 916         if (l2tp_udp_recv_core(tunnel, skb))
 917                 goto pass_up;
 918
 919         return 0;
 920
 921 pass_up:
 922         return 1;
 923 }
 924 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
 925
 926 /************************************************************************
 927  * Transmit handling
 928  ***********************************************************************/
 929
 930 /* Build an L2TP header for the session into the buffer provided.
 931  */
 932 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
 933 {
 934         struct l2tp_tunnel *tunnel = session->tunnel;
 935         __be16 *bufp = buf;
 936         __be16 *optr = buf;
 937         u16 flags = L2TP_HDR_VER_2;
 938         u32 tunnel_id = tunnel->peer_tunnel_id;
 939         u32 session_id = session->peer_session_id;
 940
 941         if (session->send_seq)
 942                 flags |= L2TP_HDRFLAG_S;
 943
 944         /* Setup L2TP header. */
 945         *bufp++ = htons(flags);
 946         *bufp++ = htons(tunnel_id);
 947         *bufp++ = htons(session_id);
 948         if (session->send_seq) {
 949                 *bufp++ = htons(session->ns);
 950                 *bufp++ = 0;
 951                 session->ns++;
 952                 session->ns &= 0xffff;
 953                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
 954                          session->name, session->ns);
 955         }
 956
 957         return bufp - optr;
 958 }
 959
 960 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
 961 {
 962         struct l2tp_tunnel *tunnel = session->tunnel;
 963         char *bufp = buf;
 964         char *optr = bufp;
 965
 966         /* Setup L2TP header. The header differs slightly for UDP and
 967          * IP encapsulations. For UDP, there is 4 bytes of flags.
 968          */
 969         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
 970                 u16 flags = L2TP_HDR_VER_3;
 971                 *((__be16 *) bufp) = htons(flags);
 972                 bufp += 2;
 973                 *((__be16 *) bufp) = 0;
 974                 bufp += 2;
 975         }
 976
 977         *((__be32 *) bufp) = htonl(session->peer_session_id);
 978         bufp += 4;
 979         if (session->cookie_len) {
 980                 memcpy(bufp, &session->cookie[0], session->cookie_len);
 981                 bufp += session->cookie_len;
 982         }
 983         if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
 984                 u32 l2h = 0;
 985
 986                 if (session->send_seq) {
 987                         l2h = 0x40000000 | session->ns;
 988                         session->ns++;
 989                         session->ns &= 0xffffff;
 990                         l2tp_dbg(session, L2TP_MSG_SEQ,
 991                                  "%s: updated ns to %u\n",
 992                                  session->name, session->ns);
 993                 }
 994
 995                 *((__be32 *)bufp) = htonl(l2h);
 996                 bufp += 4;
 997         }
 998
 999         return bufp - optr;
1000 }
1001
1002 static void l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1003                            struct flowi *fl, size_t data_len)
1004 {
1005         struct l2tp_tunnel *tunnel = session->tunnel;
1006         unsigned int len = skb->len;
1007         int error;
1008
1009         /* Debug */
1010         if (session->send_seq)
1011                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes, ns=%u\n",
1012                          session->name, data_len, session->ns - 1);
1013         else
1014                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes\n",
1015                          session->name, data_len);
1016
1017         if (session->debug & L2TP_MSG_DATA) {
1018                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1019                 unsigned char *datap = skb->data + uhlen;
1020
1021                 pr_debug("%s: xmit\n", session->name);
1022                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1023                                      datap, min_t(size_t, 32, len - uhlen));
1024         }
1025
1026         /* Queue the packet to IP for output */
1027         skb->ignore_df = 1;
1028 #if IS_ENABLED(CONFIG_IPV6)
1029         if (l2tp_sk_is_v6(tunnel->sock))
1030                 error = inet6_csk_xmit(tunnel->sock, skb, NULL);
1031         else
1032 #endif
1033                 error = ip_queue_xmit(tunnel->sock, skb, fl);
1034
1035         /* Update stats */
1036         if (error >= 0) {
1037                 atomic_long_inc(&tunnel->stats.tx_packets);
1038                 atomic_long_add(len, &tunnel->stats.tx_bytes);
1039                 atomic_long_inc(&session->stats.tx_packets);
1040                 atomic_long_add(len, &session->stats.tx_bytes);
1041         } else {
1042                 atomic_long_inc(&tunnel->stats.tx_errors);
1043                 atomic_long_inc(&session->stats.tx_errors);
1044         }
1045 }
1046
1047 /* If caller requires the skb to have a ppp header, the header must be
1048  * inserted in the skb data before calling this function.
1049  */
1050 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1051 {
1052         int data_len = skb->len;
1053         struct l2tp_tunnel *tunnel = session->tunnel;
1054         struct sock *sk = tunnel->sock;
1055         struct flowi *fl;
1056         struct udphdr *uh;
1057         struct inet_sock *inet;
1058         int headroom;
1059         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1060         int udp_len;
1061         int ret = NET_XMIT_SUCCESS;
1062
1063         /* Check that there's enough headroom in the skb to insert IP,
1064          * UDP and L2TP headers. If not enough, expand it to
1065          * make room. Adjust truesize.
1066          */
1067         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1068                 uhlen + hdr_len;
1069         if (skb_cow_head(skb, headroom)) {
1070                 kfree_skb(skb);
1071                 return NET_XMIT_DROP;
1072         }
1073
1074         /* Setup L2TP header */
1075         session->build_header(session, __skb_push(skb, hdr_len));
1076
1077         /* Reset skb netfilter state */
1078         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1079         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1080                               IPSKB_REROUTED);
1081         nf_reset(skb);
1082
1083         bh_lock_sock(sk);
1084         if (sock_owned_by_user(sk)) {
1085                 kfree_skb(skb);
1086                 ret = NET_XMIT_DROP;
1087                 goto out_unlock;
1088         }
1089
1090         /* The user-space may change the connection status for the user-space
1091          * provided socket at run time: we must check it under the socket lock
1092          */
1093         if (tunnel->fd >= 0 && sk->sk_state != TCP_ESTABLISHED) {
1094                 kfree_skb(skb);
1095                 ret = NET_XMIT_DROP;
1096                 goto out_unlock;
1097         }
1098
1099         /* Get routing info from the tunnel socket */
1100         skb_dst_drop(skb);
1101         skb_dst_set(skb, sk_dst_check(sk, 0));
1102
1103         inet = inet_sk(sk);
1104         fl = &inet->cork.fl;
1105         switch (tunnel->encap) {
1106         case L2TP_ENCAPTYPE_UDP:
1107                 /* Setup UDP header */
1108                 __skb_push(skb, sizeof(*uh));
1109                 skb_reset_transport_header(skb);
1110                 uh = udp_hdr(skb);
1111                 uh->source = inet->inet_sport;
1112                 uh->dest = inet->inet_dport;
1113                 udp_len = uhlen + hdr_len + data_len;
1114                 uh->len = htons(udp_len);
1115
1116                 /* Calculate UDP checksum if configured to do so */
1117 #if IS_ENABLED(CONFIG_IPV6)
1118                 if (l2tp_sk_is_v6(sk))
1119                         udp6_set_csum(udp_get_no_check6_tx(sk),
1120                                       skb, &inet6_sk(sk)->saddr,
1121                                       &sk->sk_v6_daddr, udp_len);
1122                 else
1123 #endif
1124                 udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr,
1125                              inet->inet_daddr, udp_len);
1126                 break;
1127
1128         case L2TP_ENCAPTYPE_IP:
1129                 break;
1130         }
1131
1132         l2tp_xmit_core(session, skb, fl, data_len);
1133 out_unlock:
1134         bh_unlock_sock(sk);
1135
1136         return ret;
1137 }
1138 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1139
1140 /*****************************************************************************
1141  * Tinnel and session create/destroy.
1142  *****************************************************************************/
1143
1144 /* Tunnel socket destruct hook.
1145  * The tunnel context is deleted only when all session sockets have been
1146  * closed.
1147  */
1148 static void l2tp_tunnel_destruct(struct sock *sk)
1149 {
1150         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1151
1152         if (tunnel == NULL)
1153                 goto end;
1154
1155         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1156
1157         /* Disable udp encapsulation */
1158         switch (tunnel->encap) {
1159         case L2TP_ENCAPTYPE_UDP:
1160                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1161                 (udp_sk(sk))->encap_type = 0;
1162                 (udp_sk(sk))->encap_rcv = NULL;
1163                 (udp_sk(sk))->encap_destroy = NULL;
1164                 break;
1165         case L2TP_ENCAPTYPE_IP:
1166                 break;
1167         }
1168
1169         /* Remove hooks into tunnel socket */
1170         sk->sk_destruct = tunnel->old_sk_destruct;
1171         sk->sk_user_data = NULL;
1172
1173         /* Call the original destructor */
1174         if (sk->sk_destruct)
1175                 (*sk->sk_destruct)(sk);
1176
1177         kfree_rcu(tunnel, rcu);
1178 end:
1179         return;
1180 }
1181
1182 /* When the tunnel is closed, all the attached sessions need to go too.
1183  */
1184 static void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1185 {
1186         int hash;
1187         struct hlist_node *walk;
1188         struct hlist_node *tmp;
1189         struct l2tp_session *session;
1190
1191         BUG_ON(tunnel == NULL);
1192
1193         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1194                   tunnel->name);
1195
1196         write_lock_bh(&tunnel->hlist_lock);
1197         tunnel->acpt_newsess = false;
1198         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1199 again:
1200                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1201                         session = hlist_entry(walk, struct l2tp_session, hlist);
1202
1203                         l2tp_info(session, L2TP_MSG_CONTROL,
1204                                   "%s: closing session\n", session->name);
1205
1206                         hlist_del_init(&session->hlist);
1207
1208                         if (test_and_set_bit(0, &session->dead))
1209                                 goto again;
1210
1211                         write_unlock_bh(&tunnel->hlist_lock);
1212
1213                         __l2tp_session_unhash(session);
1214                         l2tp_session_queue_purge(session);
1215
1216                         if (session->session_close != NULL)
1217                                 (*session->session_close)(session);
1218
1219                         l2tp_session_dec_refcount(session);
1220
1221                         write_lock_bh(&tunnel->hlist_lock);
1222
1223                         /* Now restart from the beginning of this hash
1224                          * chain.  We always remove a session from the
1225                          * list so we are guaranteed to make forward
1226                          * progress.
1227                          */
1228                         goto again;
1229                 }
1230         }
1231         write_unlock_bh(&tunnel->hlist_lock);
1232 }
1233
1234 /* Tunnel socket destroy hook for UDP encapsulation */
1235 static void l2tp_udp_encap_destroy(struct sock *sk)
1236 {
1237         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1238
1239         if (tunnel)
1240                 l2tp_tunnel_delete(tunnel);
1241 }
1242
1243 /* Workqueue tunnel deletion function */
1244 static void l2tp_tunnel_del_work(struct work_struct *work)
1245 {
1246         struct l2tp_tunnel *tunnel = container_of(work, struct l2tp_tunnel,
1247                                                   del_work);
1248         struct sock *sk = tunnel->sock;
1249         struct socket *sock = sk->sk_socket;
1250         struct l2tp_net *pn;
1251
1252         l2tp_tunnel_closeall(tunnel);
1253
1254         /* If the tunnel socket was created within the kernel, use
1255          * the sk API to release it here.
1256          */
1257         if (tunnel->fd < 0) {
1258                 if (sock) {
1259                         kernel_sock_shutdown(sock, SHUT_RDWR);
1260                         sock_release(sock);
1261                 }
1262         }
1263
1264         /* Remove the tunnel struct from the tunnel list */
1265         pn = l2tp_pernet(tunnel->l2tp_net);
1266         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1267         list_del_rcu(&tunnel->list);
1268         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1269
1270         /* drop initial ref */
1271         l2tp_tunnel_dec_refcount(tunnel);
1272
1273         /* drop workqueue ref */
1274         l2tp_tunnel_dec_refcount(tunnel);
1275 }
1276
1277 /* Create a socket for the tunnel, if one isn't set up by
1278  * userspace. This is used for static tunnels where there is no
1279  * managing L2TP daemon.
1280  *
1281  * Since we don't want these sockets to keep a namespace alive by
1282  * themselves, we drop the socket's namespace refcount after creation.
1283  * These sockets are freed when the namespace exits using the pernet
1284  * exit hook.
1285  */
1286 static int l2tp_tunnel_sock_create(struct net *net,
1287                                 u32 tunnel_id,
1288                                 u32 peer_tunnel_id,
1289                                 struct l2tp_tunnel_cfg *cfg,
1290                                 struct socket **sockp)
1291 {
1292         int err = -EINVAL;
1293         struct socket *sock = NULL;
1294         struct udp_port_cfg udp_conf;
1295
1296         switch (cfg->encap) {
1297         case L2TP_ENCAPTYPE_UDP:
1298                 memset(&udp_conf, 0, sizeof(udp_conf));
1299
1300 #if IS_ENABLED(CONFIG_IPV6)
1301                 if (cfg->local_ip6 && cfg->peer_ip6) {
1302                         udp_conf.family = AF_INET6;
1303                         memcpy(&udp_conf.local_ip6, cfg->local_ip6,
1304                                sizeof(udp_conf.local_ip6));
1305                         memcpy(&udp_conf.peer_ip6, cfg->peer_ip6,
1306                                sizeof(udp_conf.peer_ip6));
1307                         udp_conf.use_udp6_tx_checksums =
1308                           ! cfg->udp6_zero_tx_checksums;
1309                         udp_conf.use_udp6_rx_checksums =
1310                           ! cfg->udp6_zero_rx_checksums;
1311                 } else
1312 #endif
1313                 {
1314                         udp_conf.family = AF_INET;
1315                         udp_conf.local_ip = cfg->local_ip;
1316                         udp_conf.peer_ip = cfg->peer_ip;
1317                         udp_conf.use_udp_checksums = cfg->use_udp_checksums;
1318                 }
1319
1320                 udp_conf.local_udp_port = htons(cfg->local_udp_port);
1321                 udp_conf.peer_udp_port = htons(cfg->peer_udp_port);
1322
1323                 err = udp_sock_create(net, &udp_conf, &sock);
1324                 if (err < 0)
1325                         goto out;
1326
1327                 break;
1328
1329         case L2TP_ENCAPTYPE_IP:
1330 #if IS_ENABLED(CONFIG_IPV6)
1331                 if (cfg->local_ip6 && cfg->peer_ip6) {
1332                         struct sockaddr_l2tpip6 ip6_addr = {0};
1333
1334                         err = sock_create_kern(net, AF_INET6, SOCK_DGRAM,
1335                                           IPPROTO_L2TP, &sock);
1336                         if (err < 0)
1337                                 goto out;
1338
1339                         ip6_addr.l2tp_family = AF_INET6;
1340                         memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1341                                sizeof(ip6_addr.l2tp_addr));
1342                         ip6_addr.l2tp_conn_id = tunnel_id;
1343                         err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1344                                           sizeof(ip6_addr));
1345                         if (err < 0)
1346                                 goto out;
1347
1348                         ip6_addr.l2tp_family = AF_INET6;
1349                         memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1350                                sizeof(ip6_addr.l2tp_addr));
1351                         ip6_addr.l2tp_conn_id = peer_tunnel_id;
1352                         err = kernel_connect(sock,
1353                                              (struct sockaddr *) &ip6_addr,
1354                                              sizeof(ip6_addr), 0);
1355                         if (err < 0)
1356                                 goto out;
1357                 } else
1358 #endif
1359                 {
1360                         struct sockaddr_l2tpip ip_addr = {0};
1361
1362                         err = sock_create_kern(net, AF_INET, SOCK_DGRAM,
1363                                           IPPROTO_L2TP, &sock);
1364                         if (err < 0)
1365                                 goto out;
1366
1367                         ip_addr.l2tp_family = AF_INET;
1368                         ip_addr.l2tp_addr = cfg->local_ip;
1369                         ip_addr.l2tp_conn_id = tunnel_id;
1370                         err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1371                                           sizeof(ip_addr));
1372                         if (err < 0)
1373                                 goto out;
1374
1375                         ip_addr.l2tp_family = AF_INET;
1376                         ip_addr.l2tp_addr = cfg->peer_ip;
1377                         ip_addr.l2tp_conn_id = peer_tunnel_id;
1378                         err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1379                                              sizeof(ip_addr), 0);
1380                         if (err < 0)
1381                                 goto out;
1382                 }
1383                 break;
1384
1385         default:
1386                 goto out;
1387         }
1388
1389 out:
1390         *sockp = sock;
1391         if ((err < 0) && sock) {
1392                 kernel_sock_shutdown(sock, SHUT_RDWR);
1393                 sock_release(sock);
1394                 *sockp = NULL;
1395         }
1396
1397         return err;
1398 }
1399
1400 static struct lock_class_key l2tp_socket_class;
1401
1402 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)
1403 {
1404         struct l2tp_tunnel *tunnel = NULL;
1405         int err;
1406         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1407
1408         if (cfg != NULL)
1409                 encap = cfg->encap;
1410
1411         tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1412         if (tunnel == NULL) {
1413                 err = -ENOMEM;
1414                 goto err;
1415         }
1416
1417         tunnel->version = version;
1418         tunnel->tunnel_id = tunnel_id;
1419         tunnel->peer_tunnel_id = peer_tunnel_id;
1420         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1421
1422         tunnel->magic = L2TP_TUNNEL_MAGIC;
1423         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1424         rwlock_init(&tunnel->hlist_lock);
1425         tunnel->acpt_newsess = true;
1426
1427         if (cfg != NULL)
1428                 tunnel->debug = cfg->debug;
1429
1430         tunnel->encap = encap;
1431
1432         refcount_set(&tunnel->ref_count, 1);
1433         tunnel->fd = fd;
1434
1435         /* Init delete workqueue struct */
1436         INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1437
1438         INIT_LIST_HEAD(&tunnel->list);
1439
1440         err = 0;
1441 err:
1442         if (tunnelp)
1443                 *tunnelp = tunnel;
1444
1445         return err;
1446 }
1447 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1448
1449 static int l2tp_validate_socket(const struct sock *sk, const struct net *net,
1450                                 enum l2tp_encap_type encap)
1451 {
1452         if (!net_eq(sock_net(sk), net))
1453                 return -EINVAL;
1454
1455         if (sk->sk_type != SOCK_DGRAM)
1456                 return -EPROTONOSUPPORT;
1457
1458         if ((encap == L2TP_ENCAPTYPE_UDP && sk->sk_protocol != IPPROTO_UDP) ||
1459             (encap == L2TP_ENCAPTYPE_IP && sk->sk_protocol != IPPROTO_L2TP))
1460                 return -EPROTONOSUPPORT;
1461
1462         if (sk->sk_user_data)
1463                 return -EBUSY;
1464
1465         return 0;
1466 }
1467
1468 int l2tp_tunnel_register(struct l2tp_tunnel *tunnel, struct net *net,
1469                          struct l2tp_tunnel_cfg *cfg)
1470 {
1471         struct l2tp_tunnel *tunnel_walk;
1472         struct l2tp_net *pn;
1473         struct socket *sock;
1474         struct sock *sk;
1475         int ret;
1476
1477         if (tunnel->fd < 0) {
1478                 ret = l2tp_tunnel_sock_create(net, tunnel->tunnel_id,
1479                                               tunnel->peer_tunnel_id, cfg,
1480                                               &sock);
1481                 if (ret < 0)
1482                         goto err;
1483         } else {
1484                 sock = sockfd_lookup(tunnel->fd, &ret);
1485                 if (!sock)
1486                         goto err;
1487
1488                 ret = l2tp_validate_socket(sock->sk, net, tunnel->encap);
1489                 if (ret < 0)
1490                         goto err_sock;
1491         }
1492
1493         sk = sock->sk;
1494
1495         sock_hold(sk);
1496         tunnel->sock = sk;
1497         tunnel->l2tp_net = net;
1498
1499         pn = l2tp_pernet(net);
1500
1501         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1502         list_for_each_entry(tunnel_walk, &pn->l2tp_tunnel_list, list) {
1503                 if (tunnel_walk->tunnel_id == tunnel->tunnel_id) {
1504                         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1505
1506                         ret = -EEXIST;
1507                         goto err_sock;
1508                 }
1509         }
1510         list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1511         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1512
1513         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1514                 struct udp_tunnel_sock_cfg udp_cfg = {
1515                         .sk_user_data = tunnel,
1516                         .encap_type = UDP_ENCAP_L2TPINUDP,
1517                         .encap_rcv = l2tp_udp_encap_recv,
1518                         .encap_destroy = l2tp_udp_encap_destroy,
1519                 };
1520
1521                 setup_udp_tunnel_sock(net, sock, &udp_cfg);
1522         } else {
1523                 sk->sk_user_data = tunnel;
1524         }
1525
1526         tunnel->old_sk_destruct = sk->sk_destruct;
1527         sk->sk_destruct = &l2tp_tunnel_destruct;
1528         lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class,
1529                                    "l2tp_sock");
1530         sk->sk_allocation = GFP_ATOMIC;
1531
1532         if (tunnel->fd >= 0)
1533                 sockfd_put(sock);
1534
1535         return 0;
1536
1537 err_sock:
1538         if (tunnel->fd < 0)
1539                 sock_release(sock);
1540         else
1541                 sockfd_put(sock);
1542 err:
1543         return ret;
1544 }
1545 EXPORT_SYMBOL_GPL(l2tp_tunnel_register);
1546
1547 /* This function is used by the netlink TUNNEL_DELETE command.
1548  */
1549 void l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1550 {
1551         if (!test_and_set_bit(0, &tunnel->dead)) {
1552                 l2tp_tunnel_inc_refcount(tunnel);
1553                 queue_work(l2tp_wq, &tunnel->del_work);
1554         }
1555 }
1556 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1557
1558 /* Really kill the session.
1559  */
1560 void l2tp_session_free(struct l2tp_session *session)
1561 {
1562         struct l2tp_tunnel *tunnel = session->tunnel;
1563
1564         BUG_ON(refcount_read(&session->ref_count) != 0);
1565
1566         if (tunnel) {
1567                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1568                 l2tp_tunnel_dec_refcount(tunnel);
1569         }
1570
1571         kfree(session);
1572 }
1573 EXPORT_SYMBOL_GPL(l2tp_session_free);
1574
1575 /* Remove an l2tp session from l2tp_core's hash lists.
1576  * Provides a tidyup interface for pseudowire code which can't just route all
1577  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1578  * callback.
1579  */
1580 void __l2tp_session_unhash(struct l2tp_session *session)
1581 {
1582         struct l2tp_tunnel *tunnel = session->tunnel;
1583
1584         /* Remove the session from core hashes */
1585         if (tunnel) {
1586                 /* Remove from the per-tunnel hash */
1587                 write_lock_bh(&tunnel->hlist_lock);
1588                 hlist_del_init(&session->hlist);
1589                 write_unlock_bh(&tunnel->hlist_lock);
1590
1591                 /* For L2TPv3 we have a per-net hash: remove from there, too */
1592                 if (tunnel->version != L2TP_HDR_VER_2) {
1593                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1594                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1595                         hlist_del_init_rcu(&session->global_hlist);
1596                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1597                         synchronize_rcu();
1598                 }
1599         }
1600 }
1601 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1602
1603 /* This function is used by the netlink SESSION_DELETE command and by
1604    pseudowire modules.
1605  */
1606 int l2tp_session_delete(struct l2tp_session *session)
1607 {
1608         if (test_and_set_bit(0, &session->dead))
1609                 return 0;
1610
1611         __l2tp_session_unhash(session);
1612         l2tp_session_queue_purge(session);
1613         if (session->session_close != NULL)
1614                 (*session->session_close)(session);
1615
1616         l2tp_session_dec_refcount(session);
1617
1618         return 0;
1619 }
1620 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1621
1622 /* We come here whenever a session's send_seq, cookie_len or
1623  * l2specific_type parameters are set.
1624  */
1625 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1626 {
1627         if (version == L2TP_HDR_VER_2) {
1628                 session->hdr_len = 6;
1629                 if (session->send_seq)
1630                         session->hdr_len += 4;
1631         } else {
1632                 session->hdr_len = 4 + session->cookie_len;
1633                 session->hdr_len += l2tp_get_l2specific_len(session);
1634                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1635                         session->hdr_len += 4;
1636         }
1637
1638 }
1639 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1640
1641 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)
1642 {
1643         struct l2tp_session *session;
1644
1645         session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1646         if (session != NULL) {
1647                 session->magic = L2TP_SESSION_MAGIC;
1648                 session->tunnel = tunnel;
1649
1650                 session->session_id = session_id;
1651                 session->peer_session_id = peer_session_id;
1652                 session->nr = 0;
1653                 if (tunnel->version == L2TP_HDR_VER_2)
1654                         session->nr_max = 0xffff;
1655                 else
1656                         session->nr_max = 0xffffff;
1657                 session->nr_window_size = session->nr_max / 2;
1658                 session->nr_oos_count_max = 4;
1659
1660                 /* Use NR of first received packet */
1661                 session->reorder_skip = 1;
1662
1663                 sprintf(&session->name[0], "sess %u/%u",
1664                         tunnel->tunnel_id, session->session_id);
1665
1666                 skb_queue_head_init(&session->reorder_q);
1667
1668                 INIT_HLIST_NODE(&session->hlist);
1669                 INIT_HLIST_NODE(&session->global_hlist);
1670
1671                 /* Inherit debug options from tunnel */
1672                 session->debug = tunnel->debug;
1673
1674                 if (cfg) {
1675                         session->pwtype = cfg->pw_type;
1676                         session->debug = cfg->debug;
1677                         session->send_seq = cfg->send_seq;
1678                         session->recv_seq = cfg->recv_seq;
1679                         session->lns_mode = cfg->lns_mode;
1680                         session->reorder_timeout = cfg->reorder_timeout;
1681                         session->l2specific_type = cfg->l2specific_type;
1682                         session->cookie_len = cfg->cookie_len;
1683                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1684                         session->peer_cookie_len = cfg->peer_cookie_len;
1685                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1686                 }
1687
1688                 if (tunnel->version == L2TP_HDR_VER_2)
1689                         session->build_header = l2tp_build_l2tpv2_header;
1690                 else
1691                         session->build_header = l2tp_build_l2tpv3_header;
1692
1693                 l2tp_session_set_header_len(session, tunnel->version);
1694
1695                 refcount_set(&session->ref_count, 1);
1696
1697                 return session;
1698         }
1699
1700         return ERR_PTR(-ENOMEM);
1701 }
1702 EXPORT_SYMBOL_GPL(l2tp_session_create);
1703
1704 /*****************************************************************************
1705  * Init and cleanup
1706  *****************************************************************************/
1707
1708 static __net_init int l2tp_init_net(struct net *net)
1709 {
1710         struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1711         int hash;
1712
1713         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1714         spin_lock_init(&pn->l2tp_tunnel_list_lock);
1715
1716         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1717                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1718
1719         spin_lock_init(&pn->l2tp_session_hlist_lock);
1720
1721         return 0;
1722 }
1723
1724 static __net_exit void l2tp_exit_net(struct net *net)
1725 {
1726         struct l2tp_net *pn = l2tp_pernet(net);
1727         struct l2tp_tunnel *tunnel = NULL;
1728         int hash;
1729
1730         rcu_read_lock_bh();
1731         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1732                 l2tp_tunnel_delete(tunnel);
1733         }
1734         rcu_read_unlock_bh();
1735
1736         flush_workqueue(l2tp_wq);
1737         rcu_barrier();
1738
1739         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1740                 WARN_ON_ONCE(!hlist_empty(&pn->l2tp_session_hlist[hash]));
1741 }
1742
1743 static struct pernet_operations l2tp_net_ops = {
1744         .init = l2tp_init_net,
1745         .exit = l2tp_exit_net,
1746         .id   = &l2tp_net_id,
1747         .size = sizeof(struct l2tp_net),
1748 };
1749
1750 static int __init l2tp_init(void)
1751 {
1752         int rc = 0;
1753
1754         rc = register_pernet_device(&l2tp_net_ops);
1755         if (rc)
1756                 goto out;
1757
1758         l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1759         if (!l2tp_wq) {
1760                 pr_err("alloc_workqueue failed\n");
1761                 unregister_pernet_device(&l2tp_net_ops);
1762                 rc = -ENOMEM;
1763                 goto out;
1764         }
1765
1766         pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
1767
1768 out:
1769         return rc;
1770 }
1771
1772 static void __exit l2tp_exit(void)
1773 {
1774         unregister_pernet_device(&l2tp_net_ops);
1775         if (l2tp_wq) {
1776                 destroy_workqueue(l2tp_wq);
1777                 l2tp_wq = NULL;
1778         }
1779 }
1780
1781 module_init(l2tp_init);
1782 module_exit(l2tp_exit);
1783
1784 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1785 MODULE_DESCRIPTION("L2TP core");
1786 MODULE_LICENSE("GPL");
1787 MODULE_VERSION(L2TP_DRV_VERSION);