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