Pileus Git - ~andy/linux/blob - 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/inet_common.h>
  56 #include <net/xfrm.h>
  57 #include <net/protocol.h>
  58 #include <net/inet6_connection_sock.h>
  59 #include <net/inet_ecn.h>
  60 #include <net/ip6_route.h>
  61 #include <net/ip6_checksum.h>
  62
  63 #include <asm/byteorder.h>
  64 #include <linux/atomic.h>
  65
  66 #include "l2tp_core.h"
  67
  68 #define L2TP_DRV_VERSION        "V2.0"
  69
  70 /* L2TP header constants */
  71 #define L2TP_HDRFLAG_T     0x8000
  72 #define L2TP_HDRFLAG_L     0x4000
  73 #define L2TP_HDRFLAG_S     0x0800
  74 #define L2TP_HDRFLAG_O     0x0200
  75 #define L2TP_HDRFLAG_P     0x0100
  76
  77 #define L2TP_HDR_VER_MASK  0x000F
  78 #define L2TP_HDR_VER_2     0x0002
  79 #define L2TP_HDR_VER_3     0x0003
  80
  81 /* L2TPv3 default L2-specific sublayer */
  82 #define L2TP_SLFLAG_S      0x40000000
  83 #define L2TP_SL_SEQ_MASK   0x00ffffff
  84
  85 #define L2TP_HDR_SIZE_SEQ               10
  86 #define L2TP_HDR_SIZE_NOSEQ             6
  87
  88 /* Default trace flags */
  89 #define L2TP_DEFAULT_DEBUG_FLAGS        0
  90
  91 /* Private data stored for received packets in the skb.
  92  */
  93 struct l2tp_skb_cb {
  94         u32                     ns;
  95         u16                     has_seq;
  96         u16                     length;
  97         unsigned long           expires;
  98 };
  99
 100 #define L2TP_SKB_CB(skb)        ((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
 101
 102 static atomic_t l2tp_tunnel_count;
 103 static atomic_t l2tp_session_count;
 104 static struct workqueue_struct *l2tp_wq;
 105
 106 /* per-net private data for this module */
 107 static unsigned int l2tp_net_id;
 108 struct l2tp_net {
 109         struct list_head l2tp_tunnel_list;
 110         spinlock_t l2tp_tunnel_list_lock;
 111         struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
 112         spinlock_t l2tp_session_hlist_lock;
 113 };
 114
 115 static void l2tp_session_set_header_len(struct l2tp_session *session, int version);
 116 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel);
 117
 118 static inline struct l2tp_tunnel *l2tp_tunnel(struct sock *sk)
 119 {
 120         return sk->sk_user_data;
 121 }
 122
 123 static inline struct l2tp_net *l2tp_pernet(struct net *net)
 124 {
 125         BUG_ON(!net);
 126
 127         return net_generic(net, l2tp_net_id);
 128 }
 129
 130 /* Tunnel reference counts. Incremented per session that is added to
 131  * the tunnel.
 132  */
 133 static inline void l2tp_tunnel_inc_refcount_1(struct l2tp_tunnel *tunnel)
 134 {
 135         atomic_inc(&tunnel->ref_count);
 136 }
 137
 138 static inline void l2tp_tunnel_dec_refcount_1(struct l2tp_tunnel *tunnel)
 139 {
 140         if (atomic_dec_and_test(&tunnel->ref_count))
 141                 l2tp_tunnel_free(tunnel);
 142 }
 143 #ifdef L2TP_REFCNT_DEBUG
 144 #define l2tp_tunnel_inc_refcount(_t)                                    \
 145 do {                                                                    \
 146         pr_debug("l2tp_tunnel_inc_refcount: %s:%d %s: cnt=%d\n",        \
 147                  __func__, __LINE__, (_t)->name,                        \
 148                  atomic_read(&_t->ref_count));                          \
 149         l2tp_tunnel_inc_refcount_1(_t);                                 \
 150 } while (0)
 151 #define l2tp_tunnel_dec_refcount(_t)
 152 do {                                                                    \
 153         pr_debug("l2tp_tunnel_dec_refcount: %s:%d %s: cnt=%d\n",        \
 154                  __func__, __LINE__, (_t)->name,                        \
 155                  atomic_read(&_t->ref_count));                          \
 156         l2tp_tunnel_dec_refcount_1(_t);                                 \
 157 } while (0)
 158 #else
 159 #define l2tp_tunnel_inc_refcount(t) l2tp_tunnel_inc_refcount_1(t)
 160 #define l2tp_tunnel_dec_refcount(t) l2tp_tunnel_dec_refcount_1(t)
 161 #endif
 162
 163 /* Session hash global list for L2TPv3.
 164  * The session_id SHOULD be random according to RFC3931, but several
 165  * L2TP implementations use incrementing session_ids.  So we do a real
 166  * hash on the session_id, rather than a simple bitmask.
 167  */
 168 static inline struct hlist_head *
 169 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
 170 {
 171         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
 172
 173 }
 174
 175 /* Lookup the tunnel socket, possibly involving the fs code if the socket is
 176  * owned by userspace.  A struct sock returned from this function must be
 177  * released using l2tp_tunnel_sock_put once you're done with it.
 178  */
 179 struct sock *l2tp_tunnel_sock_lookup(struct l2tp_tunnel *tunnel)
 180 {
 181         int err = 0;
 182         struct socket *sock = NULL;
 183         struct sock *sk = NULL;
 184
 185         if (!tunnel)
 186                 goto out;
 187
 188         if (tunnel->fd >= 0) {
 189                 /* Socket is owned by userspace, who might be in the process
 190                  * of closing it.  Look the socket up using the fd to ensure
 191                  * consistency.
 192                  */
 193                 sock = sockfd_lookup(tunnel->fd, &err);
 194                 if (sock)
 195                         sk = sock->sk;
 196         } else {
 197                 /* Socket is owned by kernelspace */
 198                 sk = tunnel->sock;
 199                 sock_hold(sk);
 200         }
 201
 202 out:
 203         return sk;
 204 }
 205 EXPORT_SYMBOL_GPL(l2tp_tunnel_sock_lookup);
 206
 207 /* Drop a reference to a tunnel socket obtained via. l2tp_tunnel_sock_put */
 208 void l2tp_tunnel_sock_put(struct sock *sk)
 209 {
 210         struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
 211         if (tunnel) {
 212                 if (tunnel->fd >= 0) {
 213                         /* Socket is owned by userspace */
 214                         sockfd_put(sk->sk_socket);
 215                 }
 216                 sock_put(sk);
 217         }
 218         sock_put(sk);
 219 }
 220 EXPORT_SYMBOL_GPL(l2tp_tunnel_sock_put);
 221
 222 /* Lookup a session by id in the global session list
 223  */
 224 static struct l2tp_session *l2tp_session_find_2(struct net *net, u32 session_id)
 225 {
 226         struct l2tp_net *pn = l2tp_pernet(net);
 227         struct hlist_head *session_list =
 228                 l2tp_session_id_hash_2(pn, session_id);
 229         struct l2tp_session *session;
 230
 231         rcu_read_lock_bh();
 232         hlist_for_each_entry_rcu(session, session_list, global_hlist) {
 233                 if (session->session_id == session_id) {
 234                         rcu_read_unlock_bh();
 235                         return session;
 236                 }
 237         }
 238         rcu_read_unlock_bh();
 239
 240         return NULL;
 241 }
 242
 243 /* Session hash list.
 244  * The session_id SHOULD be random according to RFC2661, but several
 245  * L2TP implementations (Cisco and Microsoft) use incrementing
 246  * session_ids.  So we do a real hash on the session_id, rather than a
 247  * simple bitmask.
 248  */
 249 static inline struct hlist_head *
 250 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
 251 {
 252         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
 253 }
 254
 255 /* Lookup a session by id
 256  */
 257 struct l2tp_session *l2tp_session_find(struct net *net, struct l2tp_tunnel *tunnel, u32 session_id)
 258 {
 259         struct hlist_head *session_list;
 260         struct l2tp_session *session;
 261
 262         /* In L2TPv3, session_ids are unique over all tunnels and we
 263          * sometimes need to look them up before we know the
 264          * tunnel.
 265          */
 266         if (tunnel == NULL)
 267                 return l2tp_session_find_2(net, session_id);
 268
 269         session_list = l2tp_session_id_hash(tunnel, session_id);
 270         read_lock_bh(&tunnel->hlist_lock);
 271         hlist_for_each_entry(session, session_list, hlist) {
 272                 if (session->session_id == session_id) {
 273                         read_unlock_bh(&tunnel->hlist_lock);
 274                         return session;
 275                 }
 276         }
 277         read_unlock_bh(&tunnel->hlist_lock);
 278
 279         return NULL;
 280 }
 281 EXPORT_SYMBOL_GPL(l2tp_session_find);
 282
 283 struct l2tp_session *l2tp_session_find_nth(struct l2tp_tunnel *tunnel, int nth)
 284 {
 285         int hash;
 286         struct l2tp_session *session;
 287         int count = 0;
 288
 289         read_lock_bh(&tunnel->hlist_lock);
 290         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
 291                 hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
 292                         if (++count > nth) {
 293                                 read_unlock_bh(&tunnel->hlist_lock);
 294                                 return session;
 295                         }
 296                 }
 297         }
 298
 299         read_unlock_bh(&tunnel->hlist_lock);
 300
 301         return NULL;
 302 }
 303 EXPORT_SYMBOL_GPL(l2tp_session_find_nth);
 304
 305 /* Lookup a session by interface name.
 306  * This is very inefficient but is only used by management interfaces.
 307  */
 308 struct l2tp_session *l2tp_session_find_by_ifname(struct net *net, char *ifname)
 309 {
 310         struct l2tp_net *pn = l2tp_pernet(net);
 311         int hash;
 312         struct l2tp_session *session;
 313
 314         rcu_read_lock_bh();
 315         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
 316                 hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
 317                         if (!strcmp(session->ifname, ifname)) {
 318                                 rcu_read_unlock_bh();
 319                                 return session;
 320                         }
 321                 }
 322         }
 323
 324         rcu_read_unlock_bh();
 325
 326         return NULL;
 327 }
 328 EXPORT_SYMBOL_GPL(l2tp_session_find_by_ifname);
 329
 330 /* Lookup a tunnel by id
 331  */
 332 struct l2tp_tunnel *l2tp_tunnel_find(struct net *net, u32 tunnel_id)
 333 {
 334         struct l2tp_tunnel *tunnel;
 335         struct l2tp_net *pn = l2tp_pernet(net);
 336
 337         rcu_read_lock_bh();
 338         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
 339                 if (tunnel->tunnel_id == tunnel_id) {
 340                         rcu_read_unlock_bh();
 341                         return tunnel;
 342                 }
 343         }
 344         rcu_read_unlock_bh();
 345
 346         return NULL;
 347 }
 348 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
 349
 350 struct l2tp_tunnel *l2tp_tunnel_find_nth(struct net *net, int nth)
 351 {
 352         struct l2tp_net *pn = l2tp_pernet(net);
 353         struct l2tp_tunnel *tunnel;
 354         int count = 0;
 355
 356         rcu_read_lock_bh();
 357         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
 358                 if (++count > nth) {
 359                         rcu_read_unlock_bh();
 360                         return tunnel;
 361                 }
 362         }
 363
 364         rcu_read_unlock_bh();
 365
 366         return NULL;
 367 }
 368 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
 369
 370 /*****************************************************************************
 371  * Receive data handling
 372  *****************************************************************************/
 373
 374 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
 375  * number.
 376  */
 377 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
 378 {
 379         struct sk_buff *skbp;
 380         struct sk_buff *tmp;
 381         u32 ns = L2TP_SKB_CB(skb)->ns;
 382
 383         spin_lock_bh(&session->reorder_q.lock);
 384         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
 385                 if (L2TP_SKB_CB(skbp)->ns > ns) {
 386                         __skb_queue_before(&session->reorder_q, skbp, skb);
 387                         l2tp_dbg(session, L2TP_MSG_SEQ,
 388                                  "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
 389                                  session->name, ns, L2TP_SKB_CB(skbp)->ns,
 390                                  skb_queue_len(&session->reorder_q));
 391                         atomic_long_inc(&session->stats.rx_oos_packets);
 392                         goto out;
 393                 }
 394         }
 395
 396         __skb_queue_tail(&session->reorder_q, skb);
 397
 398 out:
 399         spin_unlock_bh(&session->reorder_q.lock);
 400 }
 401
 402 /* Dequeue a single skb.
 403  */
 404 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
 405 {
 406         struct l2tp_tunnel *tunnel = session->tunnel;
 407         int length = L2TP_SKB_CB(skb)->length;
 408
 409         /* We're about to requeue the skb, so return resources
 410          * to its current owner (a socket receive buffer).
 411          */
 412         skb_orphan(skb);
 413
 414         atomic_long_inc(&tunnel->stats.rx_packets);
 415         atomic_long_add(length, &tunnel->stats.rx_bytes);
 416         atomic_long_inc(&session->stats.rx_packets);
 417         atomic_long_add(length, &session->stats.rx_bytes);
 418
 419         if (L2TP_SKB_CB(skb)->has_seq) {
 420                 /* Bump our Nr */
 421                 session->nr++;
 422                 session->nr &= session->nr_max;
 423
 424                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
 425                          session->name, session->nr);
 426         }
 427
 428         /* call private receive handler */
 429         if (session->recv_skb != NULL)
 430                 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
 431         else
 432                 kfree_skb(skb);
 433
 434         if (session->deref)
 435                 (*session->deref)(session);
 436 }
 437
 438 /* Dequeue skbs from the session's reorder_q, subject to packet order.
 439  * Skbs that have been in the queue for too long are simply discarded.
 440  */
 441 static void l2tp_recv_dequeue(struct l2tp_session *session)
 442 {
 443         struct sk_buff *skb;
 444         struct sk_buff *tmp;
 445
 446         /* If the pkt at the head of the queue has the nr that we
 447          * expect to send up next, dequeue it and any other
 448          * in-sequence packets behind it.
 449          */
 450 start:
 451         spin_lock_bh(&session->reorder_q.lock);
 452         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
 453                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
 454                         atomic_long_inc(&session->stats.rx_seq_discards);
 455                         atomic_long_inc(&session->stats.rx_errors);
 456                         l2tp_dbg(session, L2TP_MSG_SEQ,
 457                                  "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
 458                                  session->name, L2TP_SKB_CB(skb)->ns,
 459                                  L2TP_SKB_CB(skb)->length, session->nr,
 460                                  skb_queue_len(&session->reorder_q));
 461                         session->reorder_skip = 1;
 462                         __skb_unlink(skb, &session->reorder_q);
 463                         kfree_skb(skb);
 464                         if (session->deref)
 465                                 (*session->deref)(session);
 466                         continue;
 467                 }
 468
 469                 if (L2TP_SKB_CB(skb)->has_seq) {
 470                         if (session->reorder_skip) {
 471                                 l2tp_dbg(session, L2TP_MSG_SEQ,
 472                                          "%s: advancing nr to next pkt: %u -> %u",
 473                                          session->name, session->nr,
 474                                          L2TP_SKB_CB(skb)->ns);
 475                                 session->reorder_skip = 0;
 476                                 session->nr = L2TP_SKB_CB(skb)->ns;
 477                         }
 478                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
 479                                 l2tp_dbg(session, L2TP_MSG_SEQ,
 480                                          "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
 481                                          session->name, L2TP_SKB_CB(skb)->ns,
 482                                          L2TP_SKB_CB(skb)->length, session->nr,
 483                                          skb_queue_len(&session->reorder_q));
 484                                 goto out;
 485                         }
 486                 }
 487                 __skb_unlink(skb, &session->reorder_q);
 488
 489                 /* Process the skb. We release the queue lock while we
 490                  * do so to let other contexts process the queue.
 491                  */
 492                 spin_unlock_bh(&session->reorder_q.lock);
 493                 l2tp_recv_dequeue_skb(session, skb);
 494                 goto start;
 495         }
 496
 497 out:
 498         spin_unlock_bh(&session->reorder_q.lock);
 499 }
 500
 501 static inline int l2tp_verify_udp_checksum(struct sock *sk,
 502                                            struct sk_buff *skb)
 503 {
 504         struct udphdr *uh = udp_hdr(skb);
 505         u16 ulen = ntohs(uh->len);
 506         __wsum psum;
 507
 508         if (sk->sk_no_check || skb_csum_unnecessary(skb))
 509                 return 0;
 510
 511 #if IS_ENABLED(CONFIG_IPV6)
 512         if (sk->sk_family == PF_INET6 && !l2tp_tunnel(sk)->v4mapped) {
 513                 if (!uh->check) {
 514                         LIMIT_NETDEBUG(KERN_INFO "L2TP: IPv6: checksum is 0\n");
 515                         return 1;
 516                 }
 517                 if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
 518                     !csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
 519                                      &ipv6_hdr(skb)->daddr, ulen,
 520                                      IPPROTO_UDP, skb->csum)) {
 521                         skb->ip_summed = CHECKSUM_UNNECESSARY;
 522                         return 0;
 523                 }
 524                 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
 525                                                          &ipv6_hdr(skb)->daddr,
 526                                                          skb->len, IPPROTO_UDP,
 527                                                          0));
 528         } else
 529 #endif
 530         {
 531                 struct inet_sock *inet;
 532                 if (!uh->check)
 533                         return 0;
 534                 inet = inet_sk(sk);
 535                 psum = csum_tcpudp_nofold(inet->inet_saddr, inet->inet_daddr,
 536                                           ulen, IPPROTO_UDP, 0);
 537
 538                 if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
 539                     !csum_fold(csum_add(psum, skb->csum)))
 540                         return 0;
 541                 skb->csum = psum;
 542         }
 543
 544         return __skb_checksum_complete(skb);
 545 }
 546
 547 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
 548 {
 549         u32 nws;
 550
 551         if (nr >= session->nr)
 552                 nws = nr - session->nr;
 553         else
 554                 nws = (session->nr_max + 1) - (session->nr - nr);
 555
 556         return nws < session->nr_window_size;
 557 }
 558
 559 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
 560  * acceptable, else non-zero.
 561  */
 562 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
 563 {
 564         if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
 565                 /* Packet sequence number is outside allowed window.
 566                  * Discard it.
 567                  */
 568                 l2tp_dbg(session, L2TP_MSG_SEQ,
 569                          "%s: pkt %u len %d discarded, outside window, nr=%u\n",
 570                          session->name, L2TP_SKB_CB(skb)->ns,
 571                          L2TP_SKB_CB(skb)->length, session->nr);
 572                 goto discard;
 573         }
 574
 575         if (session->reorder_timeout != 0) {
 576                 /* Packet reordering enabled. Add skb to session's
 577                  * reorder queue, in order of ns.
 578                  */
 579                 l2tp_recv_queue_skb(session, skb);
 580                 goto out;
 581         }
 582
 583         /* Packet reordering disabled. Discard out-of-sequence packets, while
 584          * tracking the number if in-sequence packets after the first OOS packet
 585          * is seen. After nr_oos_count_max in-sequence packets, reset the
 586          * sequence number to re-enable packet reception.
 587          */
 588         if (L2TP_SKB_CB(skb)->ns == session->nr) {
 589                 skb_queue_tail(&session->reorder_q, skb);
 590         } else {
 591                 u32 nr_oos = L2TP_SKB_CB(skb)->ns;
 592                 u32 nr_next = (session->nr_oos + 1) & session->nr_max;
 593
 594                 if (nr_oos == nr_next)
 595                         session->nr_oos_count++;
 596                 else
 597                         session->nr_oos_count = 0;
 598
 599                 session->nr_oos = nr_oos;
 600                 if (session->nr_oos_count > session->nr_oos_count_max) {
 601                         session->reorder_skip = 1;
 602                         l2tp_dbg(session, L2TP_MSG_SEQ,
 603                                  "%s: %d oos packets received. Resetting sequence numbers\n",
 604                                  session->name, session->nr_oos_count);
 605                 }
 606                 if (!session->reorder_skip) {
 607                         atomic_long_inc(&session->stats.rx_seq_discards);
 608                         l2tp_dbg(session, L2TP_MSG_SEQ,
 609                                  "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
 610                                  session->name, L2TP_SKB_CB(skb)->ns,
 611                                  L2TP_SKB_CB(skb)->length, session->nr,
 612                                  skb_queue_len(&session->reorder_q));
 613                         goto discard;
 614                 }
 615                 skb_queue_tail(&session->reorder_q, skb);
 616         }
 617
 618 out:
 619         return 0;
 620
 621 discard:
 622         return 1;
 623 }
 624
 625 /* Do receive processing of L2TP data frames. We handle both L2TPv2
 626  * and L2TPv3 data frames here.
 627  *
 628  * L2TPv2 Data Message Header
 629  *
 630  *  0                   1                   2                   3
 631  *  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
 632  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 633  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
 634  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 635  * |           Tunnel ID           |           Session ID          |
 636  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 637  * |             Ns (opt)          |             Nr (opt)          |
 638  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 639  * |      Offset Size (opt)        |    Offset pad... (opt)
 640  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 641  *
 642  * Data frames are marked by T=0. All other fields are the same as
 643  * those in L2TP control frames.
 644  *
 645  * L2TPv3 Data Message Header
 646  *
 647  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 648  * |                      L2TP Session Header                      |
 649  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 650  * |                      L2-Specific Sublayer                     |
 651  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 652  * |                        Tunnel Payload                      ...
 653  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 654  *
 655  * L2TPv3 Session Header Over IP
 656  *
 657  *  0                   1                   2                   3
 658  *  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
 659  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 660  * |                           Session ID                          |
 661  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 662  * |               Cookie (optional, maximum 64 bits)...
 663  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 664  *                                                                 |
 665  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 666  *
 667  * L2TPv3 L2-Specific Sublayer Format
 668  *
 669  *  0                   1                   2                   3
 670  *  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
 671  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 672  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
 673  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 674  *
 675  * Cookie value, sublayer format and offset (pad) are negotiated with
 676  * the peer when the session is set up. Unlike L2TPv2, we do not need
 677  * to parse the packet header to determine if optional fields are
 678  * present.
 679  *
 680  * Caller must already have parsed the frame and determined that it is
 681  * a data (not control) frame before coming here. Fields up to the
 682  * session-id have already been parsed and ptr points to the data
 683  * after the session-id.
 684  */
 685 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
 686                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
 687                       int length, int (*payload_hook)(struct sk_buff *skb))
 688 {
 689         struct l2tp_tunnel *tunnel = session->tunnel;
 690         int offset;
 691         u32 ns, nr;
 692
 693         /* The ref count is increased since we now hold a pointer to
 694          * the session. Take care to decrement the refcnt when exiting
 695          * this function from now on...
 696          */
 697         l2tp_session_inc_refcount(session);
 698         if (session->ref)
 699                 (*session->ref)(session);
 700
 701         /* Parse and check optional cookie */
 702         if (session->peer_cookie_len > 0) {
 703                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
 704                         l2tp_info(tunnel, L2TP_MSG_DATA,
 705                                   "%s: cookie mismatch (%u/%u). Discarding.\n",
 706                                   tunnel->name, tunnel->tunnel_id,
 707                                   session->session_id);
 708                         atomic_long_inc(&session->stats.rx_cookie_discards);
 709                         goto discard;
 710                 }
 711                 ptr += session->peer_cookie_len;
 712         }
 713
 714         /* Handle the optional sequence numbers. Sequence numbers are
 715          * in different places for L2TPv2 and L2TPv3.
 716          *
 717          * If we are the LAC, enable/disable sequence numbers under
 718          * the control of the LNS.  If no sequence numbers present but
 719          * we were expecting them, discard frame.
 720          */
 721         ns = nr = 0;
 722         L2TP_SKB_CB(skb)->has_seq = 0;
 723         if (tunnel->version == L2TP_HDR_VER_2) {
 724                 if (hdrflags & L2TP_HDRFLAG_S) {
 725                         ns = ntohs(*(__be16 *) ptr);
 726                         ptr += 2;
 727                         nr = ntohs(*(__be16 *) ptr);
 728                         ptr += 2;
 729
 730                         /* Store L2TP info in the skb */
 731                         L2TP_SKB_CB(skb)->ns = ns;
 732                         L2TP_SKB_CB(skb)->has_seq = 1;
 733
 734                         l2tp_dbg(session, L2TP_MSG_SEQ,
 735                                  "%s: recv data ns=%u, nr=%u, session nr=%u\n",
 736                                  session->name, ns, nr, session->nr);
 737                 }
 738         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
 739                 u32 l2h = ntohl(*(__be32 *) ptr);
 740
 741                 if (l2h & 0x40000000) {
 742                         ns = l2h & 0x00ffffff;
 743
 744                         /* Store L2TP info in the skb */
 745                         L2TP_SKB_CB(skb)->ns = ns;
 746                         L2TP_SKB_CB(skb)->has_seq = 1;
 747
 748                         l2tp_dbg(session, L2TP_MSG_SEQ,
 749                                  "%s: recv data ns=%u, session nr=%u\n",
 750                                  session->name, ns, session->nr);
 751                 }
 752         }
 753
 754         /* Advance past L2-specific header, if present */
 755         ptr += session->l2specific_len;
 756
 757         if (L2TP_SKB_CB(skb)->has_seq) {
 758                 /* Received a packet with sequence numbers. If we're the LNS,
 759                  * check if we sre sending sequence numbers and if not,
 760                  * configure it so.
 761                  */
 762                 if ((!session->lns_mode) && (!session->send_seq)) {
 763                         l2tp_info(session, L2TP_MSG_SEQ,
 764                                   "%s: requested to enable seq numbers by LNS\n",
 765                                   session->name);
 766                         session->send_seq = -1;
 767                         l2tp_session_set_header_len(session, tunnel->version);
 768                 }
 769         } else {
 770                 /* No sequence numbers.
 771                  * If user has configured mandatory sequence numbers, discard.
 772                  */
 773                 if (session->recv_seq) {
 774                         l2tp_warn(session, L2TP_MSG_SEQ,
 775                                   "%s: recv data has no seq numbers when required. Discarding.\n",
 776                                   session->name);
 777                         atomic_long_inc(&session->stats.rx_seq_discards);
 778                         goto discard;
 779                 }
 780
 781                 /* If we're the LAC and we're sending sequence numbers, the
 782                  * LNS has requested that we no longer send sequence numbers.
 783                  * If we're the LNS and we're sending sequence numbers, the
 784                  * LAC is broken. Discard the frame.
 785                  */
 786                 if ((!session->lns_mode) && (session->send_seq)) {
 787                         l2tp_info(session, L2TP_MSG_SEQ,
 788                                   "%s: requested to disable seq numbers by LNS\n",
 789                                   session->name);
 790                         session->send_seq = 0;
 791                         l2tp_session_set_header_len(session, tunnel->version);
 792                 } else if (session->send_seq) {
 793                         l2tp_warn(session, L2TP_MSG_SEQ,
 794                                   "%s: recv data has no seq numbers when required. Discarding.\n",
 795                                   session->name);
 796                         atomic_long_inc(&session->stats.rx_seq_discards);
 797                         goto discard;
 798                 }
 799         }
 800
 801         /* Session data offset is handled differently for L2TPv2 and
 802          * L2TPv3. For L2TPv2, there is an optional 16-bit value in
 803          * the header. For L2TPv3, the offset is negotiated using AVPs
 804          * in the session setup control protocol.
 805          */
 806         if (tunnel->version == L2TP_HDR_VER_2) {
 807                 /* If offset bit set, skip it. */
 808                 if (hdrflags & L2TP_HDRFLAG_O) {
 809                         offset = ntohs(*(__be16 *)ptr);
 810                         ptr += 2 + offset;
 811                 }
 812         } else
 813                 ptr += session->offset;
 814
 815         offset = ptr - optr;
 816         if (!pskb_may_pull(skb, offset))
 817                 goto discard;
 818
 819         __skb_pull(skb, offset);
 820
 821         /* If caller wants to process the payload before we queue the
 822          * packet, do so now.
 823          */
 824         if (payload_hook)
 825                 if ((*payload_hook)(skb))
 826                         goto discard;
 827
 828         /* Prepare skb for adding to the session's reorder_q.  Hold
 829          * packets for max reorder_timeout or 1 second if not
 830          * reordering.
 831          */
 832         L2TP_SKB_CB(skb)->length = length;
 833         L2TP_SKB_CB(skb)->expires = jiffies +
 834                 (session->reorder_timeout ? session->reorder_timeout : HZ);
 835
 836         /* Add packet to the session's receive queue. Reordering is done here, if
 837          * enabled. Saved L2TP protocol info is stored in skb->sb[].
 838          */
 839         if (L2TP_SKB_CB(skb)->has_seq) {
 840                 if (l2tp_recv_data_seq(session, skb))
 841                         goto discard;
 842         } else {
 843                 /* No sequence numbers. Add the skb to the tail of the
 844                  * reorder queue. This ensures that it will be
 845                  * delivered after all previous sequenced skbs.
 846                  */
 847                 skb_queue_tail(&session->reorder_q, skb);
 848         }
 849
 850         /* Try to dequeue as many skbs from reorder_q as we can. */
 851         l2tp_recv_dequeue(session);
 852
 853         l2tp_session_dec_refcount(session);
 854
 855         return;
 856
 857 discard:
 858         atomic_long_inc(&session->stats.rx_errors);
 859         kfree_skb(skb);
 860
 861         if (session->deref)
 862                 (*session->deref)(session);
 863
 864         l2tp_session_dec_refcount(session);
 865 }
 866 EXPORT_SYMBOL(l2tp_recv_common);
 867
 868 /* Drop skbs from the session's reorder_q
 869  */
 870 int l2tp_session_queue_purge(struct l2tp_session *session)
 871 {
 872         struct sk_buff *skb = NULL;
 873         BUG_ON(!session);
 874         BUG_ON(session->magic != L2TP_SESSION_MAGIC);
 875         while ((skb = skb_dequeue(&session->reorder_q))) {
 876                 atomic_long_inc(&session->stats.rx_errors);
 877                 kfree_skb(skb);
 878                 if (session->deref)
 879                         (*session->deref)(session);
 880         }
 881         return 0;
 882 }
 883 EXPORT_SYMBOL_GPL(l2tp_session_queue_purge);
 884
 885 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
 886  * here. The skb is not on a list when we get here.
 887  * Returns 0 if the packet was a data packet and was successfully passed on.
 888  * Returns 1 if the packet was not a good data packet and could not be
 889  * forwarded.  All such packets are passed up to userspace to deal with.
 890  */
 891 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
 892                               int (*payload_hook)(struct sk_buff *skb))
 893 {
 894         struct l2tp_session *session = NULL;
 895         unsigned char *ptr, *optr;
 896         u16 hdrflags;
 897         u32 tunnel_id, session_id;
 898         u16 version;
 899         int length;
 900
 901         if (tunnel->sock && l2tp_verify_udp_checksum(tunnel->sock, skb))
 902                 goto discard_bad_csum;
 903
 904         /* UDP always verifies the packet length. */
 905         __skb_pull(skb, sizeof(struct udphdr));
 906
 907         /* Short packet? */
 908         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
 909                 l2tp_info(tunnel, L2TP_MSG_DATA,
 910                           "%s: recv short packet (len=%d)\n",
 911                           tunnel->name, skb->len);
 912                 goto error;
 913         }
 914
 915         /* Trace packet contents, if enabled */
 916         if (tunnel->debug & L2TP_MSG_DATA) {
 917                 length = min(32u, skb->len);
 918                 if (!pskb_may_pull(skb, length))
 919                         goto error;
 920
 921                 pr_debug("%s: recv\n", tunnel->name);
 922                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
 923         }
 924
 925         /* Point to L2TP header */
 926         optr = ptr = skb->data;
 927
 928         /* Get L2TP header flags */
 929         hdrflags = ntohs(*(__be16 *) ptr);
 930
 931         /* Check protocol version */
 932         version = hdrflags & L2TP_HDR_VER_MASK;
 933         if (version != tunnel->version) {
 934                 l2tp_info(tunnel, L2TP_MSG_DATA,
 935                           "%s: recv protocol version mismatch: got %d expected %d\n",
 936                           tunnel->name, version, tunnel->version);
 937                 goto error;
 938         }
 939
 940         /* Get length of L2TP packet */
 941         length = skb->len;
 942
 943         /* If type is control packet, it is handled by userspace. */
 944         if (hdrflags & L2TP_HDRFLAG_T) {
 945                 l2tp_dbg(tunnel, L2TP_MSG_DATA,
 946                          "%s: recv control packet, len=%d\n",
 947                          tunnel->name, length);
 948                 goto error;
 949         }
 950
 951         /* Skip flags */
 952         ptr += 2;
 953
 954         if (tunnel->version == L2TP_HDR_VER_2) {
 955                 /* If length is present, skip it */
 956                 if (hdrflags & L2TP_HDRFLAG_L)
 957                         ptr += 2;
 958
 959                 /* Extract tunnel and session ID */
 960                 tunnel_id = ntohs(*(__be16 *) ptr);
 961                 ptr += 2;
 962                 session_id = ntohs(*(__be16 *) ptr);
 963                 ptr += 2;
 964         } else {
 965                 ptr += 2;       /* skip reserved bits */
 966                 tunnel_id = tunnel->tunnel_id;
 967                 session_id = ntohl(*(__be32 *) ptr);
 968                 ptr += 4;
 969         }
 970
 971         /* Find the session context */
 972         session = l2tp_session_find(tunnel->l2tp_net, tunnel, session_id);
 973         if (!session || !session->recv_skb) {
 974                 /* Not found? Pass to userspace to deal with */
 975                 l2tp_info(tunnel, L2TP_MSG_DATA,
 976                           "%s: no session found (%u/%u). Passing up.\n",
 977                           tunnel->name, tunnel_id, session_id);
 978                 goto error;
 979         }
 980
 981         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
 982
 983         return 0;
 984
 985 discard_bad_csum:
 986         LIMIT_NETDEBUG("%s: UDP: bad checksum\n", tunnel->name);
 987         UDP_INC_STATS_USER(tunnel->l2tp_net, UDP_MIB_INERRORS, 0);
 988         atomic_long_inc(&tunnel->stats.rx_errors);
 989         kfree_skb(skb);
 990
 991         return 0;
 992
 993 error:
 994         /* Put UDP header back */
 995         __skb_push(skb, sizeof(struct udphdr));
 996
 997         return 1;
 998 }
 999
1000 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
1001  * Return codes:
1002  * 0 : success.
1003  * <0: error
1004  * >0: skb should be passed up to userspace as UDP.
1005  */
1006 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1007 {
1008         struct l2tp_tunnel *tunnel;
1009
1010         tunnel = l2tp_sock_to_tunnel(sk);
1011         if (tunnel == NULL)
1012                 goto pass_up;
1013
1014         l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
1015                  tunnel->name, skb->len);
1016
1017         if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
1018                 goto pass_up_put;
1019
1020         sock_put(sk);
1021         return 0;
1022
1023 pass_up_put:
1024         sock_put(sk);
1025 pass_up:
1026         return 1;
1027 }
1028 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
1029
1030 /************************************************************************
1031  * Transmit handling
1032  ***********************************************************************/
1033
1034 /* Build an L2TP header for the session into the buffer provided.
1035  */
1036 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
1037 {
1038         struct l2tp_tunnel *tunnel = session->tunnel;
1039         __be16 *bufp = buf;
1040         __be16 *optr = buf;
1041         u16 flags = L2TP_HDR_VER_2;
1042         u32 tunnel_id = tunnel->peer_tunnel_id;
1043         u32 session_id = session->peer_session_id;
1044
1045         if (session->send_seq)
1046                 flags |= L2TP_HDRFLAG_S;
1047
1048         /* Setup L2TP header. */
1049         *bufp++ = htons(flags);
1050         *bufp++ = htons(tunnel_id);
1051         *bufp++ = htons(session_id);
1052         if (session->send_seq) {
1053                 *bufp++ = htons(session->ns);
1054                 *bufp++ = 0;
1055                 session->ns++;
1056                 session->ns &= 0xffff;
1057                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
1058                          session->name, session->ns);
1059         }
1060
1061         return bufp - optr;
1062 }
1063
1064 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
1065 {
1066         struct l2tp_tunnel *tunnel = session->tunnel;
1067         char *bufp = buf;
1068         char *optr = bufp;
1069
1070         /* Setup L2TP header. The header differs slightly for UDP and
1071          * IP encapsulations. For UDP, there is 4 bytes of flags.
1072          */
1073         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1074                 u16 flags = L2TP_HDR_VER_3;
1075                 *((__be16 *) bufp) = htons(flags);
1076                 bufp += 2;
1077                 *((__be16 *) bufp) = 0;
1078                 bufp += 2;
1079         }
1080
1081         *((__be32 *) bufp) = htonl(session->peer_session_id);
1082         bufp += 4;
1083         if (session->cookie_len) {
1084                 memcpy(bufp, &session->cookie[0], session->cookie_len);
1085                 bufp += session->cookie_len;
1086         }
1087         if (session->l2specific_len) {
1088                 if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
1089                         u32 l2h = 0;
1090                         if (session->send_seq) {
1091                                 l2h = 0x40000000 | session->ns;
1092                                 session->ns++;
1093                                 session->ns &= 0xffffff;
1094                                 l2tp_dbg(session, L2TP_MSG_SEQ,
1095                                          "%s: updated ns to %u\n",
1096                                          session->name, session->ns);
1097                         }
1098
1099                         *((__be32 *) bufp) = htonl(l2h);
1100                 }
1101                 bufp += session->l2specific_len;
1102         }
1103         if (session->offset)
1104                 bufp += session->offset;
1105
1106         return bufp - optr;
1107 }
1108
1109 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1110                           struct flowi *fl, size_t data_len)
1111 {
1112         struct l2tp_tunnel *tunnel = session->tunnel;
1113         unsigned int len = skb->len;
1114         int error;
1115
1116         /* Debug */
1117         if (session->send_seq)
1118                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %Zd bytes, ns=%u\n",
1119                          session->name, data_len, session->ns - 1);
1120         else
1121                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %Zd bytes\n",
1122                          session->name, data_len);
1123
1124         if (session->debug & L2TP_MSG_DATA) {
1125                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1126                 unsigned char *datap = skb->data + uhlen;
1127
1128                 pr_debug("%s: xmit\n", session->name);
1129                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1130                                      datap, min_t(size_t, 32, len - uhlen));
1131         }
1132
1133         /* Queue the packet to IP for output */
1134         skb->local_df = 1;
1135 #if IS_ENABLED(CONFIG_IPV6)
1136         if (skb->sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1137                 error = inet6_csk_xmit(skb, NULL);
1138         else
1139 #endif
1140                 error = ip_queue_xmit(skb, fl);
1141
1142         /* Update stats */
1143         if (error >= 0) {
1144                 atomic_long_inc(&tunnel->stats.tx_packets);
1145                 atomic_long_add(len, &tunnel->stats.tx_bytes);
1146                 atomic_long_inc(&session->stats.tx_packets);
1147                 atomic_long_add(len, &session->stats.tx_bytes);
1148         } else {
1149                 atomic_long_inc(&tunnel->stats.tx_errors);
1150                 atomic_long_inc(&session->stats.tx_errors);
1151         }
1152
1153         return 0;
1154 }
1155
1156 /* Automatically called when the skb is freed.
1157  */
1158 static void l2tp_sock_wfree(struct sk_buff *skb)
1159 {
1160         sock_put(skb->sk);
1161 }
1162
1163 /* For data skbs that we transmit, we associate with the tunnel socket
1164  * but don't do accounting.
1165  */
1166 static inline void l2tp_skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1167 {
1168         sock_hold(sk);
1169         skb->sk = sk;
1170         skb->destructor = l2tp_sock_wfree;
1171 }
1172
1173 #if IS_ENABLED(CONFIG_IPV6)
1174 static void l2tp_xmit_ipv6_csum(struct sock *sk, struct sk_buff *skb,
1175                                 int udp_len)
1176 {
1177         struct ipv6_pinfo *np = inet6_sk(sk);
1178         struct udphdr *uh = udp_hdr(skb);
1179
1180         if (!skb_dst(skb) || !skb_dst(skb)->dev ||
1181             !(skb_dst(skb)->dev->features & NETIF_F_IPV6_CSUM)) {
1182                 __wsum csum = skb_checksum(skb, 0, udp_len, 0);
1183                 skb->ip_summed = CHECKSUM_UNNECESSARY;
1184                 uh->check = csum_ipv6_magic(&np->saddr, &sk->sk_v6_daddr, udp_len,
1185                                             IPPROTO_UDP, csum);
1186                 if (uh->check == 0)
1187                         uh->check = CSUM_MANGLED_0;
1188         } else {
1189                 skb->ip_summed = CHECKSUM_PARTIAL;
1190                 skb->csum_start = skb_transport_header(skb) - skb->head;
1191                 skb->csum_offset = offsetof(struct udphdr, check);
1192                 uh->check = ~csum_ipv6_magic(&np->saddr, &sk->sk_v6_daddr,
1193                                              udp_len, IPPROTO_UDP, 0);
1194         }
1195 }
1196 #endif
1197
1198 /* If caller requires the skb to have a ppp header, the header must be
1199  * inserted in the skb data before calling this function.
1200  */
1201 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1202 {
1203         int data_len = skb->len;
1204         struct l2tp_tunnel *tunnel = session->tunnel;
1205         struct sock *sk = tunnel->sock;
1206         struct flowi *fl;
1207         struct udphdr *uh;
1208         struct inet_sock *inet;
1209         __wsum csum;
1210         int headroom;
1211         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1212         int udp_len;
1213         int ret = NET_XMIT_SUCCESS;
1214
1215         /* Check that there's enough headroom in the skb to insert IP,
1216          * UDP and L2TP headers. If not enough, expand it to
1217          * make room. Adjust truesize.
1218          */
1219         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1220                 uhlen + hdr_len;
1221         if (skb_cow_head(skb, headroom)) {
1222                 kfree_skb(skb);
1223                 return NET_XMIT_DROP;
1224         }
1225
1226         skb_orphan(skb);
1227         /* Setup L2TP header */
1228         session->build_header(session, __skb_push(skb, hdr_len));
1229
1230         /* Reset skb netfilter state */
1231         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1232         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1233                               IPSKB_REROUTED);
1234         nf_reset(skb);
1235
1236         bh_lock_sock(sk);
1237         if (sock_owned_by_user(sk)) {
1238                 kfree_skb(skb);
1239                 ret = NET_XMIT_DROP;
1240                 goto out_unlock;
1241         }
1242
1243         /* Get routing info from the tunnel socket */
1244         skb_dst_drop(skb);
1245         skb_dst_set(skb, dst_clone(__sk_dst_check(sk, 0)));
1246
1247         inet = inet_sk(sk);
1248         fl = &inet->cork.fl;
1249         switch (tunnel->encap) {
1250         case L2TP_ENCAPTYPE_UDP:
1251                 /* Setup UDP header */
1252                 __skb_push(skb, sizeof(*uh));
1253                 skb_reset_transport_header(skb);
1254                 uh = udp_hdr(skb);
1255                 uh->source = inet->inet_sport;
1256                 uh->dest = inet->inet_dport;
1257                 udp_len = uhlen + hdr_len + data_len;
1258                 uh->len = htons(udp_len);
1259                 uh->check = 0;
1260
1261                 /* Calculate UDP checksum if configured to do so */
1262 #if IS_ENABLED(CONFIG_IPV6)
1263                 if (sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1264                         l2tp_xmit_ipv6_csum(sk, skb, udp_len);
1265                 else
1266 #endif
1267                 if (sk->sk_no_check == UDP_CSUM_NOXMIT)
1268                         skb->ip_summed = CHECKSUM_NONE;
1269                 else if ((skb_dst(skb) && skb_dst(skb)->dev) &&
1270                          (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM))) {
1271                         skb->ip_summed = CHECKSUM_COMPLETE;
1272                         csum = skb_checksum(skb, 0, udp_len, 0);
1273                         uh->check = csum_tcpudp_magic(inet->inet_saddr,
1274                                                       inet->inet_daddr,
1275                                                       udp_len, IPPROTO_UDP, csum);
1276                         if (uh->check == 0)
1277                                 uh->check = CSUM_MANGLED_0;
1278                 } else {
1279                         skb->ip_summed = CHECKSUM_PARTIAL;
1280                         skb->csum_start = skb_transport_header(skb) - skb->head;
1281                         skb->csum_offset = offsetof(struct udphdr, check);
1282                         uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
1283                                                        inet->inet_daddr,
1284                                                        udp_len, IPPROTO_UDP, 0);
1285                 }
1286                 break;
1287
1288         case L2TP_ENCAPTYPE_IP:
1289                 break;
1290         }
1291
1292         l2tp_skb_set_owner_w(skb, sk);
1293
1294         l2tp_xmit_core(session, skb, fl, data_len);
1295 out_unlock:
1296         bh_unlock_sock(sk);
1297
1298         return ret;
1299 }
1300 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1301
1302 /*****************************************************************************
1303  * Tinnel and session create/destroy.
1304  *****************************************************************************/
1305
1306 /* Tunnel socket destruct hook.
1307  * The tunnel context is deleted only when all session sockets have been
1308  * closed.
1309  */
1310 static void l2tp_tunnel_destruct(struct sock *sk)
1311 {
1312         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1313         struct l2tp_net *pn;
1314
1315         if (tunnel == NULL)
1316                 goto end;
1317
1318         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1319
1320
1321         /* Disable udp encapsulation */
1322         switch (tunnel->encap) {
1323         case L2TP_ENCAPTYPE_UDP:
1324                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1325                 (udp_sk(sk))->encap_type = 0;
1326                 (udp_sk(sk))->encap_rcv = NULL;
1327                 (udp_sk(sk))->encap_destroy = NULL;
1328                 break;
1329         case L2TP_ENCAPTYPE_IP:
1330                 break;
1331         }
1332
1333         /* Remove hooks into tunnel socket */
1334         sk->sk_destruct = tunnel->old_sk_destruct;
1335         sk->sk_user_data = NULL;
1336         tunnel->sock = NULL;
1337
1338         /* Remove the tunnel struct from the tunnel list */
1339         pn = l2tp_pernet(tunnel->l2tp_net);
1340         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1341         list_del_rcu(&tunnel->list);
1342         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1343         atomic_dec(&l2tp_tunnel_count);
1344
1345         l2tp_tunnel_closeall(tunnel);
1346         l2tp_tunnel_dec_refcount(tunnel);
1347
1348         /* Call the original destructor */
1349         if (sk->sk_destruct)
1350                 (*sk->sk_destruct)(sk);
1351 end:
1352         return;
1353 }
1354
1355 /* When the tunnel is closed, all the attached sessions need to go too.
1356  */
1357 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1358 {
1359         int hash;
1360         struct hlist_node *walk;
1361         struct hlist_node *tmp;
1362         struct l2tp_session *session;
1363
1364         BUG_ON(tunnel == NULL);
1365
1366         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1367                   tunnel->name);
1368
1369         write_lock_bh(&tunnel->hlist_lock);
1370         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1371 again:
1372                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1373                         session = hlist_entry(walk, struct l2tp_session, hlist);
1374
1375                         l2tp_info(session, L2TP_MSG_CONTROL,
1376                                   "%s: closing session\n", session->name);
1377
1378                         hlist_del_init(&session->hlist);
1379
1380                         if (session->ref != NULL)
1381                                 (*session->ref)(session);
1382
1383                         write_unlock_bh(&tunnel->hlist_lock);
1384
1385                         __l2tp_session_unhash(session);
1386                         l2tp_session_queue_purge(session);
1387
1388                         if (session->session_close != NULL)
1389                                 (*session->session_close)(session);
1390
1391                         if (session->deref != NULL)
1392                                 (*session->deref)(session);
1393
1394                         l2tp_session_dec_refcount(session);
1395
1396                         write_lock_bh(&tunnel->hlist_lock);
1397
1398                         /* Now restart from the beginning of this hash
1399                          * chain.  We always remove a session from the
1400                          * list so we are guaranteed to make forward
1401                          * progress.
1402                          */
1403                         goto again;
1404                 }
1405         }
1406         write_unlock_bh(&tunnel->hlist_lock);
1407 }
1408 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1409
1410 /* Tunnel socket destroy hook for UDP encapsulation */
1411 static void l2tp_udp_encap_destroy(struct sock *sk)
1412 {
1413         struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
1414         if (tunnel) {
1415                 l2tp_tunnel_closeall(tunnel);
1416                 sock_put(sk);
1417         }
1418 }
1419
1420 /* Really kill the tunnel.
1421  * Come here only when all sessions have been cleared from the tunnel.
1422  */
1423 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
1424 {
1425         BUG_ON(atomic_read(&tunnel->ref_count) != 0);
1426         BUG_ON(tunnel->sock != NULL);
1427         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: free...\n", tunnel->name);
1428         kfree_rcu(tunnel, rcu);
1429 }
1430
1431 /* Workqueue tunnel deletion function */
1432 static void l2tp_tunnel_del_work(struct work_struct *work)
1433 {
1434         struct l2tp_tunnel *tunnel = NULL;
1435         struct socket *sock = NULL;
1436         struct sock *sk = NULL;
1437
1438         tunnel = container_of(work, struct l2tp_tunnel, del_work);
1439         sk = l2tp_tunnel_sock_lookup(tunnel);
1440         if (!sk)
1441                 return;
1442
1443         sock = sk->sk_socket;
1444
1445         /* If the tunnel socket was created by userspace, then go through the
1446          * inet layer to shut the socket down, and let userspace close it.
1447          * Otherwise, if we created the socket directly within the kernel, use
1448          * the sk API to release it here.
1449          * In either case the tunnel resources are freed in the socket
1450          * destructor when the tunnel socket goes away.
1451          */
1452         if (tunnel->fd >= 0) {
1453                 if (sock)
1454                         inet_shutdown(sock, 2);
1455         } else {
1456                 if (sock)
1457                         kernel_sock_shutdown(sock, SHUT_RDWR);
1458                 sk_release_kernel(sk);
1459         }
1460
1461         l2tp_tunnel_sock_put(sk);
1462 }
1463
1464 /* Create a socket for the tunnel, if one isn't set up by
1465  * userspace. This is used for static tunnels where there is no
1466  * managing L2TP daemon.
1467  *
1468  * Since we don't want these sockets to keep a namespace alive by
1469  * themselves, we drop the socket's namespace refcount after creation.
1470  * These sockets are freed when the namespace exits using the pernet
1471  * exit hook.
1472  */
1473 static int l2tp_tunnel_sock_create(struct net *net,
1474                                 u32 tunnel_id,
1475                                 u32 peer_tunnel_id,
1476                                 struct l2tp_tunnel_cfg *cfg,
1477                                 struct socket **sockp)
1478 {
1479         int err = -EINVAL;
1480         struct socket *sock = NULL;
1481         struct sockaddr_in udp_addr = {0};
1482         struct sockaddr_l2tpip ip_addr = {0};
1483 #if IS_ENABLED(CONFIG_IPV6)
1484         struct sockaddr_in6 udp6_addr = {0};
1485         struct sockaddr_l2tpip6 ip6_addr = {0};
1486 #endif
1487
1488         switch (cfg->encap) {
1489         case L2TP_ENCAPTYPE_UDP:
1490 #if IS_ENABLED(CONFIG_IPV6)
1491                 if (cfg->local_ip6 && cfg->peer_ip6) {
1492                         err = sock_create_kern(AF_INET6, SOCK_DGRAM, 0, &sock);
1493                         if (err < 0)
1494                                 goto out;
1495
1496                         sk_change_net(sock->sk, net);
1497
1498                         udp6_addr.sin6_family = AF_INET6;
1499                         memcpy(&udp6_addr.sin6_addr, cfg->local_ip6,
1500                                sizeof(udp6_addr.sin6_addr));
1501                         udp6_addr.sin6_port = htons(cfg->local_udp_port);
1502                         err = kernel_bind(sock, (struct sockaddr *) &udp6_addr,
1503                                           sizeof(udp6_addr));
1504                         if (err < 0)
1505                                 goto out;
1506
1507                         udp6_addr.sin6_family = AF_INET6;
1508                         memcpy(&udp6_addr.sin6_addr, cfg->peer_ip6,
1509                                sizeof(udp6_addr.sin6_addr));
1510                         udp6_addr.sin6_port = htons(cfg->peer_udp_port);
1511                         err = kernel_connect(sock,
1512                                              (struct sockaddr *) &udp6_addr,
1513                                              sizeof(udp6_addr), 0);
1514                         if (err < 0)
1515                                 goto out;
1516                 } else
1517 #endif
1518                 {
1519                         err = sock_create_kern(AF_INET, SOCK_DGRAM, 0, &sock);
1520                         if (err < 0)
1521                                 goto out;
1522
1523                         sk_change_net(sock->sk, net);
1524
1525                         udp_addr.sin_family = AF_INET;
1526                         udp_addr.sin_addr = cfg->local_ip;
1527                         udp_addr.sin_port = htons(cfg->local_udp_port);
1528                         err = kernel_bind(sock, (struct sockaddr *) &udp_addr,
1529                                           sizeof(udp_addr));
1530                         if (err < 0)
1531                                 goto out;
1532
1533                         udp_addr.sin_family = AF_INET;
1534                         udp_addr.sin_addr = cfg->peer_ip;
1535                         udp_addr.sin_port = htons(cfg->peer_udp_port);
1536                         err = kernel_connect(sock,
1537                                              (struct sockaddr *) &udp_addr,
1538                                              sizeof(udp_addr), 0);
1539                         if (err < 0)
1540                                 goto out;
1541                 }
1542
1543                 if (!cfg->use_udp_checksums)
1544                         sock->sk->sk_no_check = UDP_CSUM_NOXMIT;
1545
1546                 break;
1547
1548         case L2TP_ENCAPTYPE_IP:
1549 #if IS_ENABLED(CONFIG_IPV6)
1550                 if (cfg->local_ip6 && cfg->peer_ip6) {
1551                         err = sock_create_kern(AF_INET6, SOCK_DGRAM,
1552                                           IPPROTO_L2TP, &sock);
1553                         if (err < 0)
1554                                 goto out;
1555
1556                         sk_change_net(sock->sk, net);
1557
1558                         ip6_addr.l2tp_family = AF_INET6;
1559                         memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1560                                sizeof(ip6_addr.l2tp_addr));
1561                         ip6_addr.l2tp_conn_id = tunnel_id;
1562                         err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1563                                           sizeof(ip6_addr));
1564                         if (err < 0)
1565                                 goto out;
1566
1567                         ip6_addr.l2tp_family = AF_INET6;
1568                         memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1569                                sizeof(ip6_addr.l2tp_addr));
1570                         ip6_addr.l2tp_conn_id = peer_tunnel_id;
1571                         err = kernel_connect(sock,
1572                                              (struct sockaddr *) &ip6_addr,
1573                                              sizeof(ip6_addr), 0);
1574                         if (err < 0)
1575                                 goto out;
1576                 } else
1577 #endif
1578                 {
1579                         err = sock_create_kern(AF_INET, SOCK_DGRAM,
1580                                           IPPROTO_L2TP, &sock);
1581                         if (err < 0)
1582                                 goto out;
1583
1584                         sk_change_net(sock->sk, net);
1585
1586                         ip_addr.l2tp_family = AF_INET;
1587                         ip_addr.l2tp_addr = cfg->local_ip;
1588                         ip_addr.l2tp_conn_id = tunnel_id;
1589                         err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1590                                           sizeof(ip_addr));
1591                         if (err < 0)
1592                                 goto out;
1593
1594                         ip_addr.l2tp_family = AF_INET;
1595                         ip_addr.l2tp_addr = cfg->peer_ip;
1596                         ip_addr.l2tp_conn_id = peer_tunnel_id;
1597                         err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1598                                              sizeof(ip_addr), 0);
1599                         if (err < 0)
1600                                 goto out;
1601                 }
1602                 break;
1603
1604         default:
1605                 goto out;
1606         }
1607
1608 out:
1609         *sockp = sock;
1610         if ((err < 0) && sock) {
1611                 kernel_sock_shutdown(sock, SHUT_RDWR);
1612                 sk_release_kernel(sock->sk);
1613                 *sockp = NULL;
1614         }
1615
1616         return err;
1617 }
1618
1619 static struct lock_class_key l2tp_socket_class;
1620
1621 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)
1622 {
1623         struct l2tp_tunnel *tunnel = NULL;
1624         int err;
1625         struct socket *sock = NULL;
1626         struct sock *sk = NULL;
1627         struct l2tp_net *pn;
1628         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1629
1630         /* Get the tunnel socket from the fd, which was opened by
1631          * the userspace L2TP daemon. If not specified, create a
1632          * kernel socket.
1633          */
1634         if (fd < 0) {
1635                 err = l2tp_tunnel_sock_create(net, tunnel_id, peer_tunnel_id,
1636                                 cfg, &sock);
1637                 if (err < 0)
1638                         goto err;
1639         } else {
1640                 sock = sockfd_lookup(fd, &err);
1641                 if (!sock) {
1642                         pr_err("tunl %u: sockfd_lookup(fd=%d) returned %d\n",
1643                                tunnel_id, fd, err);
1644                         err = -EBADF;
1645                         goto err;
1646                 }
1647
1648                 /* Reject namespace mismatches */
1649                 if (!net_eq(sock_net(sock->sk), net)) {
1650                         pr_err("tunl %u: netns mismatch\n", tunnel_id);
1651                         err = -EINVAL;
1652                         goto err;
1653                 }
1654         }
1655
1656         sk = sock->sk;
1657
1658         if (cfg != NULL)
1659                 encap = cfg->encap;
1660
1661         /* Quick sanity checks */
1662         switch (encap) {
1663         case L2TP_ENCAPTYPE_UDP:
1664                 err = -EPROTONOSUPPORT;
1665                 if (sk->sk_protocol != IPPROTO_UDP) {
1666                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1667                                tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1668                         goto err;
1669                 }
1670                 break;
1671         case L2TP_ENCAPTYPE_IP:
1672                 err = -EPROTONOSUPPORT;
1673                 if (sk->sk_protocol != IPPROTO_L2TP) {
1674                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1675                                tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1676                         goto err;
1677                 }
1678                 break;
1679         }
1680
1681         /* Check if this socket has already been prepped */
1682         tunnel = l2tp_tunnel(sk);
1683         if (tunnel != NULL) {
1684                 /* This socket has already been prepped */
1685                 err = -EBUSY;
1686                 goto err;
1687         }
1688
1689         tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1690         if (tunnel == NULL) {
1691                 err = -ENOMEM;
1692                 goto err;
1693         }
1694
1695         tunnel->version = version;
1696         tunnel->tunnel_id = tunnel_id;
1697         tunnel->peer_tunnel_id = peer_tunnel_id;
1698         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1699
1700         tunnel->magic = L2TP_TUNNEL_MAGIC;
1701         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1702         rwlock_init(&tunnel->hlist_lock);
1703
1704         /* The net we belong to */
1705         tunnel->l2tp_net = net;
1706         pn = l2tp_pernet(net);
1707
1708         if (cfg != NULL)
1709                 tunnel->debug = cfg->debug;
1710
1711 #if IS_ENABLED(CONFIG_IPV6)
1712         if (sk->sk_family == PF_INET6) {
1713                 struct ipv6_pinfo *np = inet6_sk(sk);
1714
1715                 if (ipv6_addr_v4mapped(&np->saddr) &&
1716                     ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
1717                         struct inet_sock *inet = inet_sk(sk);
1718
1719                         tunnel->v4mapped = true;
1720                         inet->inet_saddr = np->saddr.s6_addr32[3];
1721                         inet->inet_rcv_saddr = sk->sk_v6_rcv_saddr.s6_addr32[3];
1722                         inet->inet_daddr = sk->sk_v6_daddr.s6_addr32[3];
1723                 } else {
1724                         tunnel->v4mapped = false;
1725                 }
1726         }
1727 #endif
1728
1729         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1730         tunnel->encap = encap;
1731         if (encap == L2TP_ENCAPTYPE_UDP) {
1732                 /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1733                 udp_sk(sk)->encap_type = UDP_ENCAP_L2TPINUDP;
1734                 udp_sk(sk)->encap_rcv = l2tp_udp_encap_recv;
1735                 udp_sk(sk)->encap_destroy = l2tp_udp_encap_destroy;
1736 #if IS_ENABLED(CONFIG_IPV6)
1737                 if (sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1738                         udpv6_encap_enable();
1739                 else
1740 #endif
1741                 udp_encap_enable();
1742         }
1743
1744         sk->sk_user_data = tunnel;
1745
1746         /* Hook on the tunnel socket destructor so that we can cleanup
1747          * if the tunnel socket goes away.
1748          */
1749         tunnel->old_sk_destruct = sk->sk_destruct;
1750         sk->sk_destruct = &l2tp_tunnel_destruct;
1751         tunnel->sock = sk;
1752         tunnel->fd = fd;
1753         lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class, "l2tp_sock");
1754
1755         sk->sk_allocation = GFP_ATOMIC;
1756
1757         /* Init delete workqueue struct */
1758         INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1759
1760         /* Add tunnel to our list */
1761         INIT_LIST_HEAD(&tunnel->list);
1762         atomic_inc(&l2tp_tunnel_count);
1763
1764         /* Bump the reference count. The tunnel context is deleted
1765          * only when this drops to zero. Must be done before list insertion
1766          */
1767         l2tp_tunnel_inc_refcount(tunnel);
1768         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1769         list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1770         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1771
1772         err = 0;
1773 err:
1774         if (tunnelp)
1775                 *tunnelp = tunnel;
1776
1777         /* If tunnel's socket was created by the kernel, it doesn't
1778          *  have a file.
1779          */
1780         if (sock && sock->file)
1781                 sockfd_put(sock);
1782
1783         return err;
1784 }
1785 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1786
1787 /* This function is used by the netlink TUNNEL_DELETE command.
1788  */
1789 int l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1790 {
1791         l2tp_tunnel_closeall(tunnel);
1792         return (false == queue_work(l2tp_wq, &tunnel->del_work));
1793 }
1794 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1795
1796 /* Really kill the session.
1797  */
1798 void l2tp_session_free(struct l2tp_session *session)
1799 {
1800         struct l2tp_tunnel *tunnel = session->tunnel;
1801
1802         BUG_ON(atomic_read(&session->ref_count) != 0);
1803
1804         if (tunnel) {
1805                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1806                 if (session->session_id != 0)
1807                         atomic_dec(&l2tp_session_count);
1808                 sock_put(tunnel->sock);
1809                 session->tunnel = NULL;
1810                 l2tp_tunnel_dec_refcount(tunnel);
1811         }
1812
1813         kfree(session);
1814
1815         return;
1816 }
1817 EXPORT_SYMBOL_GPL(l2tp_session_free);
1818
1819 /* Remove an l2tp session from l2tp_core's hash lists.
1820  * Provides a tidyup interface for pseudowire code which can't just route all
1821  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1822  * callback.
1823  */
1824 void __l2tp_session_unhash(struct l2tp_session *session)
1825 {
1826         struct l2tp_tunnel *tunnel = session->tunnel;
1827
1828         /* Remove the session from core hashes */
1829         if (tunnel) {
1830                 /* Remove from the per-tunnel hash */
1831                 write_lock_bh(&tunnel->hlist_lock);
1832                 hlist_del_init(&session->hlist);
1833                 write_unlock_bh(&tunnel->hlist_lock);
1834
1835                 /* For L2TPv3 we have a per-net hash: remove from there, too */
1836                 if (tunnel->version != L2TP_HDR_VER_2) {
1837                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1838                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1839                         hlist_del_init_rcu(&session->global_hlist);
1840                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1841                         synchronize_rcu();
1842                 }
1843         }
1844 }
1845 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1846
1847 /* This function is used by the netlink SESSION_DELETE command and by
1848    pseudowire modules.
1849  */
1850 int l2tp_session_delete(struct l2tp_session *session)
1851 {
1852         if (session->ref)
1853                 (*session->ref)(session);
1854         __l2tp_session_unhash(session);
1855         l2tp_session_queue_purge(session);
1856         if (session->session_close != NULL)
1857                 (*session->session_close)(session);
1858         if (session->deref)
1859                 (*session->deref)(session);
1860         l2tp_session_dec_refcount(session);
1861         return 0;
1862 }
1863 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1864
1865 /* We come here whenever a session's send_seq, cookie_len or
1866  * l2specific_len parameters are set.
1867  */
1868 static void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1869 {
1870         if (version == L2TP_HDR_VER_2) {
1871                 session->hdr_len = 6;
1872                 if (session->send_seq)
1873                         session->hdr_len += 4;
1874         } else {
1875                 session->hdr_len = 4 + session->cookie_len + session->l2specific_len + session->offset;
1876                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1877                         session->hdr_len += 4;
1878         }
1879
1880 }
1881
1882 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)
1883 {
1884         struct l2tp_session *session;
1885
1886         session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1887         if (session != NULL) {
1888                 session->magic = L2TP_SESSION_MAGIC;
1889                 session->tunnel = tunnel;
1890
1891                 session->session_id = session_id;
1892                 session->peer_session_id = peer_session_id;
1893                 session->nr = 0;
1894                 if (tunnel->version == L2TP_HDR_VER_2)
1895                         session->nr_max = 0xffff;
1896                 else
1897                         session->nr_max = 0xffffff;
1898                 session->nr_window_size = session->nr_max / 2;
1899                 session->nr_oos_count_max = 4;
1900
1901                 /* Use NR of first received packet */
1902                 session->reorder_skip = 1;
1903
1904                 sprintf(&session->name[0], "sess %u/%u",
1905                         tunnel->tunnel_id, session->session_id);
1906
1907                 skb_queue_head_init(&session->reorder_q);
1908
1909                 INIT_HLIST_NODE(&session->hlist);
1910                 INIT_HLIST_NODE(&session->global_hlist);
1911
1912                 /* Inherit debug options from tunnel */
1913                 session->debug = tunnel->debug;
1914
1915                 if (cfg) {
1916                         session->pwtype = cfg->pw_type;
1917                         session->debug = cfg->debug;
1918                         session->mtu = cfg->mtu;
1919                         session->mru = cfg->mru;
1920                         session->send_seq = cfg->send_seq;
1921                         session->recv_seq = cfg->recv_seq;
1922                         session->lns_mode = cfg->lns_mode;
1923                         session->reorder_timeout = cfg->reorder_timeout;
1924                         session->offset = cfg->offset;
1925                         session->l2specific_type = cfg->l2specific_type;
1926                         session->l2specific_len = cfg->l2specific_len;
1927                         session->cookie_len = cfg->cookie_len;
1928                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1929                         session->peer_cookie_len = cfg->peer_cookie_len;
1930                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1931                 }
1932
1933                 if (tunnel->version == L2TP_HDR_VER_2)
1934                         session->build_header = l2tp_build_l2tpv2_header;
1935                 else
1936                         session->build_header = l2tp_build_l2tpv3_header;
1937
1938                 l2tp_session_set_header_len(session, tunnel->version);
1939
1940                 /* Bump the reference count. The session context is deleted
1941                  * only when this drops to zero.
1942                  */
1943                 l2tp_session_inc_refcount(session);
1944                 l2tp_tunnel_inc_refcount(tunnel);
1945
1946                 /* Ensure tunnel socket isn't deleted */
1947                 sock_hold(tunnel->sock);
1948
1949                 /* Add session to the tunnel's hash list */
1950                 write_lock_bh(&tunnel->hlist_lock);
1951                 hlist_add_head(&session->hlist,
1952                                l2tp_session_id_hash(tunnel, session_id));
1953                 write_unlock_bh(&tunnel->hlist_lock);
1954
1955                 /* And to the global session list if L2TPv3 */
1956                 if (tunnel->version != L2TP_HDR_VER_2) {
1957                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1958
1959                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1960                         hlist_add_head_rcu(&session->global_hlist,
1961                                            l2tp_session_id_hash_2(pn, session_id));
1962                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1963                 }
1964
1965                 /* Ignore management session in session count value */
1966                 if (session->session_id != 0)
1967                         atomic_inc(&l2tp_session_count);
1968         }
1969
1970         return session;
1971 }
1972 EXPORT_SYMBOL_GPL(l2tp_session_create);
1973
1974 /*****************************************************************************
1975  * Init and cleanup
1976  *****************************************************************************/
1977
1978 static __net_init int l2tp_init_net(struct net *net)
1979 {
1980         struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1981         int hash;
1982
1983         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1984         spin_lock_init(&pn->l2tp_tunnel_list_lock);
1985
1986         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1987                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1988
1989         spin_lock_init(&pn->l2tp_session_hlist_lock);
1990
1991         return 0;
1992 }
1993
1994 static __net_exit void l2tp_exit_net(struct net *net)
1995 {
1996         struct l2tp_net *pn = l2tp_pernet(net);
1997         struct l2tp_tunnel *tunnel = NULL;
1998
1999         rcu_read_lock_bh();
2000         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
2001                 (void)l2tp_tunnel_delete(tunnel);
2002         }
2003         rcu_read_unlock_bh();
2004 }
2005
2006 static struct pernet_operations l2tp_net_ops = {
2007         .init = l2tp_init_net,
2008         .exit = l2tp_exit_net,
2009         .id   = &l2tp_net_id,
2010         .size = sizeof(struct l2tp_net),
2011 };
2012
2013 static int __init l2tp_init(void)
2014 {
2015         int rc = 0;
2016
2017         rc = register_pernet_device(&l2tp_net_ops);
2018         if (rc)
2019                 goto out;
2020
2021         l2tp_wq = alloc_workqueue("l2tp", WQ_NON_REENTRANT | WQ_UNBOUND, 0);
2022         if (!l2tp_wq) {
2023                 pr_err("alloc_workqueue failed\n");
2024                 rc = -ENOMEM;
2025                 goto out;
2026         }
2027
2028         pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
2029
2030 out:
2031         return rc;
2032 }
2033
2034 static void __exit l2tp_exit(void)
2035 {
2036         unregister_pernet_device(&l2tp_net_ops);
2037         if (l2tp_wq) {
2038                 destroy_workqueue(l2tp_wq);
2039                 l2tp_wq = NULL;
2040         }
2041 }
2042
2043 module_init(l2tp_init);
2044 module_exit(l2tp_exit);
2045
2046 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
2047 MODULE_DESCRIPTION("L2TP core");
2048 MODULE_LICENSE("GPL");
2049 MODULE_VERSION(L2TP_DRV_VERSION);
2050