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 #include <linux/module.h>
  22 #include <linux/string.h>
  23 #include <linux/list.h>
  24 #include <linux/uaccess.h>
  25
  26 #include <linux/kernel.h>
  27 #include <linux/spinlock.h>
  28 #include <linux/kthread.h>
  29 #include <linux/sched.h>
  30 #include <linux/slab.h>
  31 #include <linux/errno.h>
  32 #include <linux/jiffies.h>
  33
  34 #include <linux/netdevice.h>
  35 #include <linux/net.h>
  36 #include <linux/inetdevice.h>
  37 #include <linux/skbuff.h>
  38 #include <linux/init.h>
  39 #include <linux/in.h>
  40 #include <linux/ip.h>
  41 #include <linux/udp.h>
  42 #include <linux/l2tp.h>
  43 #include <linux/hash.h>
  44 #include <linux/sort.h>
  45 #include <linux/file.h>
  46 #include <linux/nsproxy.h>
  47 #include <net/net_namespace.h>
  48 #include <net/netns/generic.h>
  49 #include <net/dst.h>
  50 #include <net/ip.h>
  51 #include <net/udp.h>
  52 #include <net/inet_common.h>
  53 #include <net/xfrm.h>
  54 #include <net/protocol.h>
  55
  56 #include <asm/byteorder.h>
  57 #include <asm/atomic.h>
  58
  59 #include "l2tp_core.h"
  60
  61 #define L2TP_DRV_VERSION        "V2.0"
  62
  63 /* L2TP header constants */
  64 #define L2TP_HDRFLAG_T     0x8000
  65 #define L2TP_HDRFLAG_L     0x4000
  66 #define L2TP_HDRFLAG_S     0x0800
  67 #define L2TP_HDRFLAG_O     0x0200
  68 #define L2TP_HDRFLAG_P     0x0100
  69
  70 #define L2TP_HDR_VER_MASK  0x000F
  71 #define L2TP_HDR_VER_2     0x0002
  72 #define L2TP_HDR_VER_3     0x0003
  73
  74 /* L2TPv3 default L2-specific sublayer */
  75 #define L2TP_SLFLAG_S      0x40000000
  76 #define L2TP_SL_SEQ_MASK   0x00ffffff
  77
  78 #define L2TP_HDR_SIZE_SEQ               10
  79 #define L2TP_HDR_SIZE_NOSEQ             6
  80
  81 /* Default trace flags */
  82 #define L2TP_DEFAULT_DEBUG_FLAGS        0
  83
  84 #define PRINTK(_mask, _type, _lvl, _fmt, args...)                       \
  85         do {                                                            \
  86                 if ((_mask) & (_type))                                  \
  87                         printk(_lvl "L2TP: " _fmt, ##args);             \
  88         } while (0)
  89
  90 /* Private data stored for received packets in the skb.
  91  */
  92 struct l2tp_skb_cb {
  93         u32                     ns;
  94         u16                     has_seq;
  95         u16                     length;
  96         unsigned long           expires;
  97 };
  98
  99 #define L2TP_SKB_CB(skb)        ((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
 100
 101 static atomic_t l2tp_tunnel_count;
 102 static atomic_t l2tp_session_count;
 103
 104 /* per-net private data for this module */
 105 static unsigned int l2tp_net_id;
 106 struct l2tp_net {
 107         struct list_head l2tp_tunnel_list;
 108         rwlock_t l2tp_tunnel_list_lock;
 109         struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
 110         rwlock_t l2tp_session_hlist_lock;
 111 };
 112
 113 static inline struct l2tp_net *l2tp_pernet(struct net *net)
 114 {
 115         BUG_ON(!net);
 116
 117         return net_generic(net, l2tp_net_id);
 118 }
 119
 120 /* Session hash global list for L2TPv3.
 121  * The session_id SHOULD be random according to RFC3931, but several
 122  * L2TP implementations use incrementing session_ids.  So we do a real
 123  * hash on the session_id, rather than a simple bitmask.
 124  */
 125 static inline struct hlist_head *
 126 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
 127 {
 128         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
 129
 130 }
 131
 132 /* Lookup a session by id in the global session list
 133  */
 134 static struct l2tp_session *l2tp_session_find_2(struct net *net, u32 session_id)
 135 {
 136         struct l2tp_net *pn = l2tp_pernet(net);
 137         struct hlist_head *session_list =
 138                 l2tp_session_id_hash_2(pn, session_id);
 139         struct l2tp_session *session;
 140         struct hlist_node *walk;
 141
 142         read_lock_bh(&pn->l2tp_session_hlist_lock);
 143         hlist_for_each_entry(session, walk, session_list, global_hlist) {
 144                 if (session->session_id == session_id) {
 145                         read_unlock_bh(&pn->l2tp_session_hlist_lock);
 146                         return session;
 147                 }
 148         }
 149         read_unlock_bh(&pn->l2tp_session_hlist_lock);
 150
 151         return NULL;
 152 }
 153
 154 /* Session hash list.
 155  * The session_id SHOULD be random according to RFC2661, but several
 156  * L2TP implementations (Cisco and Microsoft) use incrementing
 157  * session_ids.  So we do a real hash on the session_id, rather than a
 158  * simple bitmask.
 159  */
 160 static inline struct hlist_head *
 161 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
 162 {
 163         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
 164 }
 165
 166 /* Lookup a session by id
 167  */
 168 struct l2tp_session *l2tp_session_find(struct net *net, struct l2tp_tunnel *tunnel, u32 session_id)
 169 {
 170         struct hlist_head *session_list;
 171         struct l2tp_session *session;
 172         struct hlist_node *walk;
 173
 174         /* In L2TPv3, session_ids are unique over all tunnels and we
 175          * sometimes need to look them up before we know the
 176          * tunnel.
 177          */
 178         if (tunnel == NULL)
 179                 return l2tp_session_find_2(net, session_id);
 180
 181         session_list = l2tp_session_id_hash(tunnel, session_id);
 182         read_lock_bh(&tunnel->hlist_lock);
 183         hlist_for_each_entry(session, walk, session_list, hlist) {
 184                 if (session->session_id == session_id) {
 185                         read_unlock_bh(&tunnel->hlist_lock);
 186                         return session;
 187                 }
 188         }
 189         read_unlock_bh(&tunnel->hlist_lock);
 190
 191         return NULL;
 192 }
 193 EXPORT_SYMBOL_GPL(l2tp_session_find);
 194
 195 struct l2tp_session *l2tp_session_find_nth(struct l2tp_tunnel *tunnel, int nth)
 196 {
 197         int hash;
 198         struct hlist_node *walk;
 199         struct l2tp_session *session;
 200         int count = 0;
 201
 202         read_lock_bh(&tunnel->hlist_lock);
 203         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
 204                 hlist_for_each_entry(session, walk, &tunnel->session_hlist[hash], hlist) {
 205                         if (++count > nth) {
 206                                 read_unlock_bh(&tunnel->hlist_lock);
 207                                 return session;
 208                         }
 209                 }
 210         }
 211
 212         read_unlock_bh(&tunnel->hlist_lock);
 213
 214         return NULL;
 215 }
 216 EXPORT_SYMBOL_GPL(l2tp_session_find_nth);
 217
 218 /* Lookup a session by interface name.
 219  * This is very inefficient but is only used by management interfaces.
 220  */
 221 struct l2tp_session *l2tp_session_find_by_ifname(struct net *net, char *ifname)
 222 {
 223         struct l2tp_net *pn = l2tp_pernet(net);
 224         int hash;
 225         struct hlist_node *walk;
 226         struct l2tp_session *session;
 227
 228         read_lock_bh(&pn->l2tp_session_hlist_lock);
 229         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
 230                 hlist_for_each_entry(session, walk, &pn->l2tp_session_hlist[hash], global_hlist) {
 231                         if (!strcmp(session->ifname, ifname)) {
 232                                 read_unlock_bh(&pn->l2tp_session_hlist_lock);
 233                                 return session;
 234                         }
 235                 }
 236         }
 237
 238         read_unlock_bh(&pn->l2tp_session_hlist_lock);
 239
 240         return NULL;
 241 }
 242 EXPORT_SYMBOL_GPL(l2tp_session_find_by_ifname);
 243
 244 /* Lookup a tunnel by id
 245  */
 246 struct l2tp_tunnel *l2tp_tunnel_find(struct net *net, u32 tunnel_id)
 247 {
 248         struct l2tp_tunnel *tunnel;
 249         struct l2tp_net *pn = l2tp_pernet(net);
 250
 251         read_lock_bh(&pn->l2tp_tunnel_list_lock);
 252         list_for_each_entry(tunnel, &pn->l2tp_tunnel_list, list) {
 253                 if (tunnel->tunnel_id == tunnel_id) {
 254                         read_unlock_bh(&pn->l2tp_tunnel_list_lock);
 255                         return tunnel;
 256                 }
 257         }
 258         read_unlock_bh(&pn->l2tp_tunnel_list_lock);
 259
 260         return NULL;
 261 }
 262 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
 263
 264 struct l2tp_tunnel *l2tp_tunnel_find_nth(struct net *net, int nth)
 265 {
 266         struct l2tp_net *pn = l2tp_pernet(net);
 267         struct l2tp_tunnel *tunnel;
 268         int count = 0;
 269
 270         read_lock_bh(&pn->l2tp_tunnel_list_lock);
 271         list_for_each_entry(tunnel, &pn->l2tp_tunnel_list, list) {
 272                 if (++count > nth) {
 273                         read_unlock_bh(&pn->l2tp_tunnel_list_lock);
 274                         return tunnel;
 275                 }
 276         }
 277
 278         read_unlock_bh(&pn->l2tp_tunnel_list_lock);
 279
 280         return NULL;
 281 }
 282 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
 283
 284 /*****************************************************************************
 285  * Receive data handling
 286  *****************************************************************************/
 287
 288 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
 289  * number.
 290  */
 291 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
 292 {
 293         struct sk_buff *skbp;
 294         struct sk_buff *tmp;
 295         u32 ns = L2TP_SKB_CB(skb)->ns;
 296
 297         spin_lock_bh(&session->reorder_q.lock);
 298         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
 299                 if (L2TP_SKB_CB(skbp)->ns > ns) {
 300                         __skb_queue_before(&session->reorder_q, skbp, skb);
 301                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 302                                "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
 303                                session->name, ns, L2TP_SKB_CB(skbp)->ns,
 304                                skb_queue_len(&session->reorder_q));
 305                         session->stats.rx_oos_packets++;
 306                         goto out;
 307                 }
 308         }
 309
 310         __skb_queue_tail(&session->reorder_q, skb);
 311
 312 out:
 313         spin_unlock_bh(&session->reorder_q.lock);
 314 }
 315
 316 /* Dequeue a single skb.
 317  */
 318 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
 319 {
 320         struct l2tp_tunnel *tunnel = session->tunnel;
 321         int length = L2TP_SKB_CB(skb)->length;
 322
 323         /* We're about to requeue the skb, so return resources
 324          * to its current owner (a socket receive buffer).
 325          */
 326         skb_orphan(skb);
 327
 328         tunnel->stats.rx_packets++;
 329         tunnel->stats.rx_bytes += length;
 330         session->stats.rx_packets++;
 331         session->stats.rx_bytes += length;
 332
 333         if (L2TP_SKB_CB(skb)->has_seq) {
 334                 /* Bump our Nr */
 335                 session->nr++;
 336                 if (tunnel->version == L2TP_HDR_VER_2)
 337                         session->nr &= 0xffff;
 338                 else
 339                         session->nr &= 0xffffff;
 340
 341                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 342                        "%s: updated nr to %hu\n", session->name, session->nr);
 343         }
 344
 345         /* call private receive handler */
 346         if (session->recv_skb != NULL)
 347                 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
 348         else
 349                 kfree_skb(skb);
 350
 351         if (session->deref)
 352                 (*session->deref)(session);
 353 }
 354
 355 /* Dequeue skbs from the session's reorder_q, subject to packet order.
 356  * Skbs that have been in the queue for too long are simply discarded.
 357  */
 358 static void l2tp_recv_dequeue(struct l2tp_session *session)
 359 {
 360         struct sk_buff *skb;
 361         struct sk_buff *tmp;
 362
 363         /* If the pkt at the head of the queue has the nr that we
 364          * expect to send up next, dequeue it and any other
 365          * in-sequence packets behind it.
 366          */
 367         spin_lock_bh(&session->reorder_q.lock);
 368         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
 369                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
 370                         session->stats.rx_seq_discards++;
 371                         session->stats.rx_errors++;
 372                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 373                                "%s: oos pkt %u len %d discarded (too old), "
 374                                "waiting for %u, reorder_q_len=%d\n",
 375                                session->name, L2TP_SKB_CB(skb)->ns,
 376                                L2TP_SKB_CB(skb)->length, session->nr,
 377                                skb_queue_len(&session->reorder_q));
 378                         __skb_unlink(skb, &session->reorder_q);
 379                         kfree_skb(skb);
 380                         if (session->deref)
 381                                 (*session->deref)(session);
 382                         continue;
 383                 }
 384
 385                 if (L2TP_SKB_CB(skb)->has_seq) {
 386                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
 387                                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 388                                        "%s: holding oos pkt %u len %d, "
 389                                        "waiting for %u, reorder_q_len=%d\n",
 390                                        session->name, L2TP_SKB_CB(skb)->ns,
 391                                        L2TP_SKB_CB(skb)->length, session->nr,
 392                                        skb_queue_len(&session->reorder_q));
 393                                 goto out;
 394                         }
 395                 }
 396                 __skb_unlink(skb, &session->reorder_q);
 397
 398                 /* Process the skb. We release the queue lock while we
 399                  * do so to let other contexts process the queue.
 400                  */
 401                 spin_unlock_bh(&session->reorder_q.lock);
 402                 l2tp_recv_dequeue_skb(session, skb);
 403                 spin_lock_bh(&session->reorder_q.lock);
 404         }
 405
 406 out:
 407         spin_unlock_bh(&session->reorder_q.lock);
 408 }
 409
 410 static inline int l2tp_verify_udp_checksum(struct sock *sk,
 411                                            struct sk_buff *skb)
 412 {
 413         struct udphdr *uh = udp_hdr(skb);
 414         u16 ulen = ntohs(uh->len);
 415         struct inet_sock *inet;
 416         __wsum psum;
 417
 418         if (sk->sk_no_check || skb_csum_unnecessary(skb) || !uh->check)
 419                 return 0;
 420
 421         inet = inet_sk(sk);
 422         psum = csum_tcpudp_nofold(inet->inet_saddr, inet->inet_daddr, ulen,
 423                                   IPPROTO_UDP, 0);
 424
 425         if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
 426             !csum_fold(csum_add(psum, skb->csum)))
 427                 return 0;
 428
 429         skb->csum = psum;
 430
 431         return __skb_checksum_complete(skb);
 432 }
 433
 434 /* Do receive processing of L2TP data frames. We handle both L2TPv2
 435  * and L2TPv3 data frames here.
 436  *
 437  * L2TPv2 Data Message Header
 438  *
 439  *  0                   1                   2                   3
 440  *  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
 441  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 442  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
 443  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 444  * |           Tunnel ID           |           Session ID          |
 445  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 446  * |             Ns (opt)          |             Nr (opt)          |
 447  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 448  * |      Offset Size (opt)        |    Offset pad... (opt)
 449  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 450  *
 451  * Data frames are marked by T=0. All other fields are the same as
 452  * those in L2TP control frames.
 453  *
 454  * L2TPv3 Data Message Header
 455  *
 456  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 457  * |                      L2TP Session Header                      |
 458  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 459  * |                      L2-Specific Sublayer                     |
 460  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 461  * |                        Tunnel Payload                      ...
 462  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 463  *
 464  * L2TPv3 Session Header Over IP
 465  *
 466  *  0                   1                   2                   3
 467  *  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
 468  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 469  * |                           Session ID                          |
 470  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 471  * |               Cookie (optional, maximum 64 bits)...
 472  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 473  *                                                                 |
 474  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 475  *
 476  * L2TPv3 L2-Specific Sublayer Format
 477  *
 478  *  0                   1                   2                   3
 479  *  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
 480  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 481  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
 482  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 483  *
 484  * Cookie value, sublayer format and offset (pad) are negotiated with
 485  * the peer when the session is set up. Unlike L2TPv2, we do not need
 486  * to parse the packet header to determine if optional fields are
 487  * present.
 488  *
 489  * Caller must already have parsed the frame and determined that it is
 490  * a data (not control) frame before coming here. Fields up to the
 491  * session-id have already been parsed and ptr points to the data
 492  * after the session-id.
 493  */
 494 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
 495                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
 496                       int length, int (*payload_hook)(struct sk_buff *skb))
 497 {
 498         struct l2tp_tunnel *tunnel = session->tunnel;
 499         int offset;
 500         u32 ns, nr;
 501
 502         /* The ref count is increased since we now hold a pointer to
 503          * the session. Take care to decrement the refcnt when exiting
 504          * this function from now on...
 505          */
 506         l2tp_session_inc_refcount(session);
 507         if (session->ref)
 508                 (*session->ref)(session);
 509
 510         /* Parse and check optional cookie */
 511         if (session->peer_cookie_len > 0) {
 512                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
 513                         PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
 514                                "%s: cookie mismatch (%u/%u). Discarding.\n",
 515                                tunnel->name, tunnel->tunnel_id, session->session_id);
 516                         session->stats.rx_cookie_discards++;
 517                         goto discard;
 518                 }
 519                 ptr += session->peer_cookie_len;
 520         }
 521
 522         /* Handle the optional sequence numbers. Sequence numbers are
 523          * in different places for L2TPv2 and L2TPv3.
 524          *
 525          * If we are the LAC, enable/disable sequence numbers under
 526          * the control of the LNS.  If no sequence numbers present but
 527          * we were expecting them, discard frame.
 528          */
 529         ns = nr = 0;
 530         L2TP_SKB_CB(skb)->has_seq = 0;
 531         if (tunnel->version == L2TP_HDR_VER_2) {
 532                 if (hdrflags & L2TP_HDRFLAG_S) {
 533                         ns = ntohs(*(__be16 *) ptr);
 534                         ptr += 2;
 535                         nr = ntohs(*(__be16 *) ptr);
 536                         ptr += 2;
 537
 538                         /* Store L2TP info in the skb */
 539                         L2TP_SKB_CB(skb)->ns = ns;
 540                         L2TP_SKB_CB(skb)->has_seq = 1;
 541
 542                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 543                                "%s: recv data ns=%u, nr=%u, session nr=%u\n",
 544                                session->name, ns, nr, session->nr);
 545                 }
 546         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
 547                 u32 l2h = ntohl(*(__be32 *) ptr);
 548
 549                 if (l2h & 0x40000000) {
 550                         ns = l2h & 0x00ffffff;
 551
 552                         /* Store L2TP info in the skb */
 553                         L2TP_SKB_CB(skb)->ns = ns;
 554                         L2TP_SKB_CB(skb)->has_seq = 1;
 555
 556                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 557                                "%s: recv data ns=%u, session nr=%u\n",
 558                                session->name, ns, session->nr);
 559                 }
 560         }
 561
 562         /* Advance past L2-specific header, if present */
 563         ptr += session->l2specific_len;
 564
 565         if (L2TP_SKB_CB(skb)->has_seq) {
 566                 /* Received a packet with sequence numbers. If we're the LNS,
 567                  * check if we sre sending sequence numbers and if not,
 568                  * configure it so.
 569                  */
 570                 if ((!session->lns_mode) && (!session->send_seq)) {
 571                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_INFO,
 572                                "%s: requested to enable seq numbers by LNS\n",
 573                                session->name);
 574                         session->send_seq = -1;
 575                         l2tp_session_set_header_len(session, tunnel->version);
 576                 }
 577         } else {
 578                 /* No sequence numbers.
 579                  * If user has configured mandatory sequence numbers, discard.
 580                  */
 581                 if (session->recv_seq) {
 582                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_WARNING,
 583                                "%s: recv data has no seq numbers when required. "
 584                                "Discarding\n", session->name);
 585                         session->stats.rx_seq_discards++;
 586                         goto discard;
 587                 }
 588
 589                 /* If we're the LAC and we're sending sequence numbers, the
 590                  * LNS has requested that we no longer send sequence numbers.
 591                  * If we're the LNS and we're sending sequence numbers, the
 592                  * LAC is broken. Discard the frame.
 593                  */
 594                 if ((!session->lns_mode) && (session->send_seq)) {
 595                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_INFO,
 596                                "%s: requested to disable seq numbers by LNS\n",
 597                                session->name);
 598                         session->send_seq = 0;
 599                         l2tp_session_set_header_len(session, tunnel->version);
 600                 } else if (session->send_seq) {
 601                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_WARNING,
 602                                "%s: recv data has no seq numbers when required. "
 603                                "Discarding\n", session->name);
 604                         session->stats.rx_seq_discards++;
 605                         goto discard;
 606                 }
 607         }
 608
 609         /* Session data offset is handled differently for L2TPv2 and
 610          * L2TPv3. For L2TPv2, there is an optional 16-bit value in
 611          * the header. For L2TPv3, the offset is negotiated using AVPs
 612          * in the session setup control protocol.
 613          */
 614         if (tunnel->version == L2TP_HDR_VER_2) {
 615                 /* If offset bit set, skip it. */
 616                 if (hdrflags & L2TP_HDRFLAG_O) {
 617                         offset = ntohs(*(__be16 *)ptr);
 618                         ptr += 2 + offset;
 619                 }
 620         } else
 621                 ptr += session->offset;
 622
 623         offset = ptr - optr;
 624         if (!pskb_may_pull(skb, offset))
 625                 goto discard;
 626
 627         __skb_pull(skb, offset);
 628
 629         /* If caller wants to process the payload before we queue the
 630          * packet, do so now.
 631          */
 632         if (payload_hook)
 633                 if ((*payload_hook)(skb))
 634                         goto discard;
 635
 636         /* Prepare skb for adding to the session's reorder_q.  Hold
 637          * packets for max reorder_timeout or 1 second if not
 638          * reordering.
 639          */
 640         L2TP_SKB_CB(skb)->length = length;
 641         L2TP_SKB_CB(skb)->expires = jiffies +
 642                 (session->reorder_timeout ? session->reorder_timeout : HZ);
 643
 644         /* Add packet to the session's receive queue. Reordering is done here, if
 645          * enabled. Saved L2TP protocol info is stored in skb->sb[].
 646          */
 647         if (L2TP_SKB_CB(skb)->has_seq) {
 648                 if (session->reorder_timeout != 0) {
 649                         /* Packet reordering enabled. Add skb to session's
 650                          * reorder queue, in order of ns.
 651                          */
 652                         l2tp_recv_queue_skb(session, skb);
 653                 } else {
 654                         /* Packet reordering disabled. Discard out-of-sequence
 655                          * packets
 656                          */
 657                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
 658                                 session->stats.rx_seq_discards++;
 659                                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 660                                        "%s: oos pkt %u len %d discarded, "
 661                                        "waiting for %u, reorder_q_len=%d\n",
 662                                        session->name, L2TP_SKB_CB(skb)->ns,
 663                                        L2TP_SKB_CB(skb)->length, session->nr,
 664                                        skb_queue_len(&session->reorder_q));
 665                                 goto discard;
 666                         }
 667                         skb_queue_tail(&session->reorder_q, skb);
 668                 }
 669         } else {
 670                 /* No sequence numbers. Add the skb to the tail of the
 671                  * reorder queue. This ensures that it will be
 672                  * delivered after all previous sequenced skbs.
 673                  */
 674                 skb_queue_tail(&session->reorder_q, skb);
 675         }
 676
 677         /* Try to dequeue as many skbs from reorder_q as we can. */
 678         l2tp_recv_dequeue(session);
 679
 680         l2tp_session_dec_refcount(session);
 681
 682         return;
 683
 684 discard:
 685         session->stats.rx_errors++;
 686         kfree_skb(skb);
 687
 688         if (session->deref)
 689                 (*session->deref)(session);
 690
 691         l2tp_session_dec_refcount(session);
 692 }
 693 EXPORT_SYMBOL(l2tp_recv_common);
 694
 695 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
 696  * here. The skb is not on a list when we get here.
 697  * Returns 0 if the packet was a data packet and was successfully passed on.
 698  * Returns 1 if the packet was not a good data packet and could not be
 699  * forwarded.  All such packets are passed up to userspace to deal with.
 700  */
 701 int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
 702                        int (*payload_hook)(struct sk_buff *skb))
 703 {
 704         struct l2tp_session *session = NULL;
 705         unsigned char *ptr, *optr;
 706         u16 hdrflags;
 707         u32 tunnel_id, session_id;
 708         int offset;
 709         u16 version;
 710         int length;
 711
 712         if (tunnel->sock && l2tp_verify_udp_checksum(tunnel->sock, skb))
 713                 goto discard_bad_csum;
 714
 715         /* UDP always verifies the packet length. */
 716         __skb_pull(skb, sizeof(struct udphdr));
 717
 718         /* Short packet? */
 719         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
 720                 PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
 721                        "%s: recv short packet (len=%d)\n", tunnel->name, skb->len);
 722                 goto error;
 723         }
 724
 725         /* Point to L2TP header */
 726         optr = ptr = skb->data;
 727
 728         /* Trace packet contents, if enabled */
 729         if (tunnel->debug & L2TP_MSG_DATA) {
 730                 length = min(32u, skb->len);
 731                 if (!pskb_may_pull(skb, length))
 732                         goto error;
 733
 734                 printk(KERN_DEBUG "%s: recv: ", tunnel->name);
 735
 736                 offset = 0;
 737                 do {
 738                         printk(" %02X", ptr[offset]);
 739                 } while (++offset < length);
 740
 741                 printk("\n");
 742         }
 743
 744         /* Get L2TP header flags */
 745         hdrflags = ntohs(*(__be16 *) ptr);
 746
 747         /* Check protocol version */
 748         version = hdrflags & L2TP_HDR_VER_MASK;
 749         if (version != tunnel->version) {
 750                 PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
 751                        "%s: recv protocol version mismatch: got %d expected %d\n",
 752                        tunnel->name, version, tunnel->version);
 753                 goto error;
 754         }
 755
 756         /* Get length of L2TP packet */
 757         length = skb->len;
 758
 759         /* If type is control packet, it is handled by userspace. */
 760         if (hdrflags & L2TP_HDRFLAG_T) {
 761                 PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_DEBUG,
 762                        "%s: recv control packet, len=%d\n", tunnel->name, length);
 763                 goto error;
 764         }
 765
 766         /* Skip flags */
 767         ptr += 2;
 768
 769         if (tunnel->version == L2TP_HDR_VER_2) {
 770                 /* If length is present, skip it */
 771                 if (hdrflags & L2TP_HDRFLAG_L)
 772                         ptr += 2;
 773
 774                 /* Extract tunnel and session ID */
 775                 tunnel_id = ntohs(*(__be16 *) ptr);
 776                 ptr += 2;
 777                 session_id = ntohs(*(__be16 *) ptr);
 778                 ptr += 2;
 779         } else {
 780                 ptr += 2;       /* skip reserved bits */
 781                 tunnel_id = tunnel->tunnel_id;
 782                 session_id = ntohl(*(__be32 *) ptr);
 783                 ptr += 4;
 784         }
 785
 786         /* Find the session context */
 787         session = l2tp_session_find(tunnel->l2tp_net, tunnel, session_id);
 788         if (!session || !session->recv_skb) {
 789                 /* Not found? Pass to userspace to deal with */
 790                 PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
 791                        "%s: no session found (%u/%u). Passing up.\n",
 792                        tunnel->name, tunnel_id, session_id);
 793                 goto error;
 794         }
 795
 796         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
 797
 798         return 0;
 799
 800 discard_bad_csum:
 801         LIMIT_NETDEBUG("%s: UDP: bad checksum\n", tunnel->name);
 802         UDP_INC_STATS_USER(tunnel->l2tp_net, UDP_MIB_INERRORS, 0);
 803         tunnel->stats.rx_errors++;
 804         kfree_skb(skb);
 805
 806         return 0;
 807
 808 error:
 809         /* Put UDP header back */
 810         __skb_push(skb, sizeof(struct udphdr));
 811
 812         return 1;
 813 }
 814 EXPORT_SYMBOL_GPL(l2tp_udp_recv_core);
 815
 816 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
 817  * Return codes:
 818  * 0 : success.
 819  * <0: error
 820  * >0: skb should be passed up to userspace as UDP.
 821  */
 822 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
 823 {
 824         struct l2tp_tunnel *tunnel;
 825
 826         tunnel = l2tp_sock_to_tunnel(sk);
 827         if (tunnel == NULL)
 828                 goto pass_up;
 829
 830         PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_DEBUG,
 831                "%s: received %d bytes\n", tunnel->name, skb->len);
 832
 833         if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
 834                 goto pass_up_put;
 835
 836         sock_put(sk);
 837         return 0;
 838
 839 pass_up_put:
 840         sock_put(sk);
 841 pass_up:
 842         return 1;
 843 }
 844 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
 845
 846 /************************************************************************
 847  * Transmit handling
 848  ***********************************************************************/
 849
 850 /* Build an L2TP header for the session into the buffer provided.
 851  */
 852 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
 853 {
 854         struct l2tp_tunnel *tunnel = session->tunnel;
 855         __be16 *bufp = buf;
 856         __be16 *optr = buf;
 857         u16 flags = L2TP_HDR_VER_2;
 858         u32 tunnel_id = tunnel->peer_tunnel_id;
 859         u32 session_id = session->peer_session_id;
 860
 861         if (session->send_seq)
 862                 flags |= L2TP_HDRFLAG_S;
 863
 864         /* Setup L2TP header. */
 865         *bufp++ = htons(flags);
 866         *bufp++ = htons(tunnel_id);
 867         *bufp++ = htons(session_id);
 868         if (session->send_seq) {
 869                 *bufp++ = htons(session->ns);
 870                 *bufp++ = 0;
 871                 session->ns++;
 872                 session->ns &= 0xffff;
 873                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 874                        "%s: updated ns to %u\n", session->name, session->ns);
 875         }
 876
 877         return bufp - optr;
 878 }
 879
 880 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
 881 {
 882         struct l2tp_tunnel *tunnel = session->tunnel;
 883         char *bufp = buf;
 884         char *optr = bufp;
 885
 886         /* Setup L2TP header. The header differs slightly for UDP and
 887          * IP encapsulations. For UDP, there is 4 bytes of flags.
 888          */
 889         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
 890                 u16 flags = L2TP_HDR_VER_3;
 891                 *((__be16 *) bufp) = htons(flags);
 892                 bufp += 2;
 893                 *((__be16 *) bufp) = 0;
 894                 bufp += 2;
 895         }
 896
 897         *((__be32 *) bufp) = htonl(session->peer_session_id);
 898         bufp += 4;
 899         if (session->cookie_len) {
 900                 memcpy(bufp, &session->cookie[0], session->cookie_len);
 901                 bufp += session->cookie_len;
 902         }
 903         if (session->l2specific_len) {
 904                 if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
 905                         u32 l2h = 0;
 906                         if (session->send_seq) {
 907                                 l2h = 0x40000000 | session->ns;
 908                                 session->ns++;
 909                                 session->ns &= 0xffffff;
 910                                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
 911                                        "%s: updated ns to %u\n", session->name, session->ns);
 912                         }
 913
 914                         *((__be32 *) bufp) = htonl(l2h);
 915                 }
 916                 bufp += session->l2specific_len;
 917         }
 918         if (session->offset)
 919                 bufp += session->offset;
 920
 921         return bufp - optr;
 922 }
 923
 924 int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb, size_t data_len)
 925 {
 926         struct l2tp_tunnel *tunnel = session->tunnel;
 927         unsigned int len = skb->len;
 928         int error;
 929
 930         /* Debug */
 931         if (session->send_seq)
 932                 PRINTK(session->debug, L2TP_MSG_DATA, KERN_DEBUG,
 933                        "%s: send %Zd bytes, ns=%u\n", session->name,
 934                        data_len, session->ns - 1);
 935         else
 936                 PRINTK(session->debug, L2TP_MSG_DATA, KERN_DEBUG,
 937                        "%s: send %Zd bytes\n", session->name, data_len);
 938
 939         if (session->debug & L2TP_MSG_DATA) {
 940                 int i;
 941                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
 942                 unsigned char *datap = skb->data + uhlen;
 943
 944                 printk(KERN_DEBUG "%s: xmit:", session->name);
 945                 for (i = 0; i < (len - uhlen); i++) {
 946                         printk(" %02X", *datap++);
 947                         if (i == 31) {
 948                                 printk(" ...");
 949                                 break;
 950                         }
 951                 }
 952                 printk("\n");
 953         }
 954
 955         /* Queue the packet to IP for output */
 956         error = ip_queue_xmit(skb, 1);
 957
 958         /* Update stats */
 959         if (error >= 0) {
 960                 tunnel->stats.tx_packets++;
 961                 tunnel->stats.tx_bytes += len;
 962                 session->stats.tx_packets++;
 963                 session->stats.tx_bytes += len;
 964         } else {
 965                 tunnel->stats.tx_errors++;
 966                 session->stats.tx_errors++;
 967         }
 968
 969         return 0;
 970 }
 971 EXPORT_SYMBOL_GPL(l2tp_xmit_core);
 972
 973 /* Automatically called when the skb is freed.
 974  */
 975 static void l2tp_sock_wfree(struct sk_buff *skb)
 976 {
 977         sock_put(skb->sk);
 978 }
 979
 980 /* For data skbs that we transmit, we associate with the tunnel socket
 981  * but don't do accounting.
 982  */
 983 static inline void l2tp_skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
 984 {
 985         sock_hold(sk);
 986         skb->sk = sk;
 987         skb->destructor = l2tp_sock_wfree;
 988 }
 989
 990 /* If caller requires the skb to have a ppp header, the header must be
 991  * inserted in the skb data before calling this function.
 992  */
 993 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
 994 {
 995         int data_len = skb->len;
 996         struct l2tp_tunnel *tunnel = session->tunnel;
 997         struct sock *sk = tunnel->sock;
 998         struct udphdr *uh;
 999         struct inet_sock *inet;
1000         __wsum csum;
1001         int old_headroom;
1002         int new_headroom;
1003         int headroom;
1004         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1005         int udp_len;
1006
1007         /* Check that there's enough headroom in the skb to insert IP,
1008          * UDP and L2TP headers. If not enough, expand it to
1009          * make room. Adjust truesize.
1010          */
1011         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1012                 uhlen + hdr_len;
1013         old_headroom = skb_headroom(skb);
1014         if (skb_cow_head(skb, headroom))
1015                 goto abort;
1016
1017         new_headroom = skb_headroom(skb);
1018         skb_orphan(skb);
1019         skb->truesize += new_headroom - old_headroom;
1020
1021         /* Setup L2TP header */
1022         session->build_header(session, __skb_push(skb, hdr_len));
1023
1024         /* Reset skb netfilter state */
1025         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1026         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1027                               IPSKB_REROUTED);
1028         nf_reset(skb);
1029
1030         /* Get routing info from the tunnel socket */
1031         skb_dst_drop(skb);
1032         skb_dst_set(skb, dst_clone(__sk_dst_get(sk)));
1033
1034         switch (tunnel->encap) {
1035         case L2TP_ENCAPTYPE_UDP:
1036                 /* Setup UDP header */
1037                 inet = inet_sk(sk);
1038                 __skb_push(skb, sizeof(*uh));
1039                 skb_reset_transport_header(skb);
1040                 uh = udp_hdr(skb);
1041                 uh->source = inet->inet_sport;
1042                 uh->dest = inet->inet_dport;
1043                 udp_len = uhlen + hdr_len + data_len;
1044                 uh->len = htons(udp_len);
1045                 uh->check = 0;
1046
1047                 /* Calculate UDP checksum if configured to do so */
1048                 if (sk->sk_no_check == UDP_CSUM_NOXMIT)
1049                         skb->ip_summed = CHECKSUM_NONE;
1050                 else if ((skb_dst(skb) && skb_dst(skb)->dev) &&
1051                          (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM))) {
1052                         skb->ip_summed = CHECKSUM_COMPLETE;
1053                         csum = skb_checksum(skb, 0, udp_len, 0);
1054                         uh->check = csum_tcpudp_magic(inet->inet_saddr,
1055                                                       inet->inet_daddr,
1056                                                       udp_len, IPPROTO_UDP, csum);
1057                         if (uh->check == 0)
1058                                 uh->check = CSUM_MANGLED_0;
1059                 } else {
1060                         skb->ip_summed = CHECKSUM_PARTIAL;
1061                         skb->csum_start = skb_transport_header(skb) - skb->head;
1062                         skb->csum_offset = offsetof(struct udphdr, check);
1063                         uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
1064                                                        inet->inet_daddr,
1065                                                        udp_len, IPPROTO_UDP, 0);
1066                 }
1067                 break;
1068
1069         case L2TP_ENCAPTYPE_IP:
1070                 break;
1071         }
1072
1073         l2tp_skb_set_owner_w(skb, sk);
1074
1075         l2tp_xmit_core(session, skb, data_len);
1076
1077 abort:
1078         return 0;
1079 }
1080 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1081
1082 /*****************************************************************************
1083  * Tinnel and session create/destroy.
1084  *****************************************************************************/
1085
1086 /* Tunnel socket destruct hook.
1087  * The tunnel context is deleted only when all session sockets have been
1088  * closed.
1089  */
1090 void l2tp_tunnel_destruct(struct sock *sk)
1091 {
1092         struct l2tp_tunnel *tunnel;
1093
1094         tunnel = sk->sk_user_data;
1095         if (tunnel == NULL)
1096                 goto end;
1097
1098         PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1099                "%s: closing...\n", tunnel->name);
1100
1101         /* Close all sessions */
1102         l2tp_tunnel_closeall(tunnel);
1103
1104         switch (tunnel->encap) {
1105         case L2TP_ENCAPTYPE_UDP:
1106                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1107                 (udp_sk(sk))->encap_type = 0;
1108                 (udp_sk(sk))->encap_rcv = NULL;
1109                 break;
1110         case L2TP_ENCAPTYPE_IP:
1111                 break;
1112         }
1113
1114         /* Remove hooks into tunnel socket */
1115         tunnel->sock = NULL;
1116         sk->sk_destruct = tunnel->old_sk_destruct;
1117         sk->sk_user_data = NULL;
1118
1119         /* Call the original destructor */
1120         if (sk->sk_destruct)
1121                 (*sk->sk_destruct)(sk);
1122
1123         /* We're finished with the socket */
1124         l2tp_tunnel_dec_refcount(tunnel);
1125
1126 end:
1127         return;
1128 }
1129 EXPORT_SYMBOL(l2tp_tunnel_destruct);
1130
1131 /* When the tunnel is closed, all the attached sessions need to go too.
1132  */
1133 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1134 {
1135         int hash;
1136         struct hlist_node *walk;
1137         struct hlist_node *tmp;
1138         struct l2tp_session *session;
1139
1140         BUG_ON(tunnel == NULL);
1141
1142         PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1143                "%s: closing all sessions...\n", tunnel->name);
1144
1145         write_lock_bh(&tunnel->hlist_lock);
1146         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1147 again:
1148                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1149                         session = hlist_entry(walk, struct l2tp_session, hlist);
1150
1151                         PRINTK(session->debug, L2TP_MSG_CONTROL, KERN_INFO,
1152                                "%s: closing session\n", session->name);
1153
1154                         hlist_del_init(&session->hlist);
1155
1156                         /* Since we should hold the sock lock while
1157                          * doing any unbinding, we need to release the
1158                          * lock we're holding before taking that lock.
1159                          * Hold a reference to the sock so it doesn't
1160                          * disappear as we're jumping between locks.
1161                          */
1162                         if (session->ref != NULL)
1163                                 (*session->ref)(session);
1164
1165                         write_unlock_bh(&tunnel->hlist_lock);
1166
1167                         if (tunnel->version != L2TP_HDR_VER_2) {
1168                                 struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1169
1170                                 write_lock_bh(&pn->l2tp_session_hlist_lock);
1171                                 hlist_del_init(&session->global_hlist);
1172                                 write_unlock_bh(&pn->l2tp_session_hlist_lock);
1173                         }
1174
1175                         if (session->session_close != NULL)
1176                                 (*session->session_close)(session);
1177
1178                         if (session->deref != NULL)
1179                                 (*session->deref)(session);
1180
1181                         write_lock_bh(&tunnel->hlist_lock);
1182
1183                         /* Now restart from the beginning of this hash
1184                          * chain.  We always remove a session from the
1185                          * list so we are guaranteed to make forward
1186                          * progress.
1187                          */
1188                         goto again;
1189                 }
1190         }
1191         write_unlock_bh(&tunnel->hlist_lock);
1192 }
1193 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1194
1195 /* Really kill the tunnel.
1196  * Come here only when all sessions have been cleared from the tunnel.
1197  */
1198 void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
1199 {
1200         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1201
1202         BUG_ON(atomic_read(&tunnel->ref_count) != 0);
1203         BUG_ON(tunnel->sock != NULL);
1204
1205         PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1206                "%s: free...\n", tunnel->name);
1207
1208         /* Remove from tunnel list */
1209         write_lock_bh(&pn->l2tp_tunnel_list_lock);
1210         list_del_init(&tunnel->list);
1211         write_unlock_bh(&pn->l2tp_tunnel_list_lock);
1212
1213         atomic_dec(&l2tp_tunnel_count);
1214         kfree(tunnel);
1215 }
1216 EXPORT_SYMBOL_GPL(l2tp_tunnel_free);
1217
1218 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)
1219 {
1220         struct l2tp_tunnel *tunnel = NULL;
1221         int err;
1222         struct socket *sock = NULL;
1223         struct sock *sk = NULL;
1224         struct l2tp_net *pn;
1225         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1226
1227         /* Get the tunnel socket from the fd, which was opened by
1228          * the userspace L2TP daemon.
1229          */
1230         err = -EBADF;
1231         sock = sockfd_lookup(fd, &err);
1232         if (!sock) {
1233                 printk(KERN_ERR "tunl %hu: sockfd_lookup(fd=%d) returned %d\n",
1234                        tunnel_id, fd, err);
1235                 goto err;
1236         }
1237
1238         sk = sock->sk;
1239
1240         if (cfg != NULL)
1241                 encap = cfg->encap;
1242
1243         /* Quick sanity checks */
1244         switch (encap) {
1245         case L2TP_ENCAPTYPE_UDP:
1246                 err = -EPROTONOSUPPORT;
1247                 if (sk->sk_protocol != IPPROTO_UDP) {
1248                         printk(KERN_ERR "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1249                                tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1250                         goto err;
1251                 }
1252                 break;
1253         case L2TP_ENCAPTYPE_IP:
1254                 err = -EPROTONOSUPPORT;
1255                 if (sk->sk_protocol != IPPROTO_L2TP) {
1256                         printk(KERN_ERR "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1257                                tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1258                         goto err;
1259                 }
1260                 break;
1261         }
1262
1263         /* Check if this socket has already been prepped */
1264         tunnel = (struct l2tp_tunnel *)sk->sk_user_data;
1265         if (tunnel != NULL) {
1266                 /* This socket has already been prepped */
1267                 err = -EBUSY;
1268                 goto err;
1269         }
1270
1271         tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1272         if (tunnel == NULL) {
1273                 err = -ENOMEM;
1274                 goto err;
1275         }
1276
1277         tunnel->version = version;
1278         tunnel->tunnel_id = tunnel_id;
1279         tunnel->peer_tunnel_id = peer_tunnel_id;
1280         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1281
1282         tunnel->magic = L2TP_TUNNEL_MAGIC;
1283         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1284         rwlock_init(&tunnel->hlist_lock);
1285
1286         /* The net we belong to */
1287         tunnel->l2tp_net = net;
1288         pn = l2tp_pernet(net);
1289
1290         if (cfg != NULL)
1291                 tunnel->debug = cfg->debug;
1292
1293         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1294         tunnel->encap = encap;
1295         if (encap == L2TP_ENCAPTYPE_UDP) {
1296                 /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1297                 udp_sk(sk)->encap_type = UDP_ENCAP_L2TPINUDP;
1298                 udp_sk(sk)->encap_rcv = l2tp_udp_encap_recv;
1299         }
1300
1301         sk->sk_user_data = tunnel;
1302
1303         /* Hook on the tunnel socket destructor so that we can cleanup
1304          * if the tunnel socket goes away.
1305          */
1306         tunnel->old_sk_destruct = sk->sk_destruct;
1307         sk->sk_destruct = &l2tp_tunnel_destruct;
1308         tunnel->sock = sk;
1309         sk->sk_allocation = GFP_ATOMIC;
1310
1311         /* Add tunnel to our list */
1312         INIT_LIST_HEAD(&tunnel->list);
1313         write_lock_bh(&pn->l2tp_tunnel_list_lock);
1314         list_add(&tunnel->list, &pn->l2tp_tunnel_list);
1315         write_unlock_bh(&pn->l2tp_tunnel_list_lock);
1316         atomic_inc(&l2tp_tunnel_count);
1317
1318         /* Bump the reference count. The tunnel context is deleted
1319          * only when this drops to zero.
1320          */
1321         l2tp_tunnel_inc_refcount(tunnel);
1322
1323         err = 0;
1324 err:
1325         if (tunnelp)
1326                 *tunnelp = tunnel;
1327
1328         if (sock)
1329                 sockfd_put(sock);
1330
1331         return err;
1332 }
1333 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1334
1335 /* This function is used by the netlink TUNNEL_DELETE command.
1336  */
1337 int l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1338 {
1339         int err = 0;
1340
1341         /* Force the tunnel socket to close. This will eventually
1342          * cause the tunnel to be deleted via the normal socket close
1343          * mechanisms when userspace closes the tunnel socket.
1344          */
1345         if ((tunnel->sock != NULL) && (tunnel->sock->sk_socket != NULL))
1346                 err = inet_shutdown(tunnel->sock->sk_socket, 2);
1347
1348         return err;
1349 }
1350 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1351
1352 /* Really kill the session.
1353  */
1354 void l2tp_session_free(struct l2tp_session *session)
1355 {
1356         struct l2tp_tunnel *tunnel;
1357
1358         BUG_ON(atomic_read(&session->ref_count) != 0);
1359
1360         tunnel = session->tunnel;
1361         if (tunnel != NULL) {
1362                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1363
1364                 /* Delete the session from the hash */
1365                 write_lock_bh(&tunnel->hlist_lock);
1366                 hlist_del_init(&session->hlist);
1367                 write_unlock_bh(&tunnel->hlist_lock);
1368
1369                 /* Unlink from the global hash if not L2TPv2 */
1370                 if (tunnel->version != L2TP_HDR_VER_2) {
1371                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1372
1373                         write_lock_bh(&pn->l2tp_session_hlist_lock);
1374                         hlist_del_init(&session->global_hlist);
1375                         write_unlock_bh(&pn->l2tp_session_hlist_lock);
1376                 }
1377
1378                 if (session->session_id != 0)
1379                         atomic_dec(&l2tp_session_count);
1380
1381                 sock_put(tunnel->sock);
1382
1383                 /* This will delete the tunnel context if this
1384                  * is the last session on the tunnel.
1385                  */
1386                 session->tunnel = NULL;
1387                 l2tp_tunnel_dec_refcount(tunnel);
1388         }
1389
1390         kfree(session);
1391
1392         return;
1393 }
1394 EXPORT_SYMBOL_GPL(l2tp_session_free);
1395
1396 /* This function is used by the netlink SESSION_DELETE command and by
1397    pseudowire modules.
1398  */
1399 int l2tp_session_delete(struct l2tp_session *session)
1400 {
1401         if (session->session_close != NULL)
1402                 (*session->session_close)(session);
1403
1404         l2tp_session_dec_refcount(session);
1405
1406         return 0;
1407 }
1408 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1409
1410
1411 /* We come here whenever a session's send_seq, cookie_len or
1412  * l2specific_len parameters are set.
1413  */
1414 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1415 {
1416         if (version == L2TP_HDR_VER_2) {
1417                 session->hdr_len = 6;
1418                 if (session->send_seq)
1419                         session->hdr_len += 4;
1420         } else {
1421                 session->hdr_len = 4 + session->cookie_len + session->l2specific_len + session->offset;
1422                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1423                         session->hdr_len += 4;
1424         }
1425
1426 }
1427 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1428
1429 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)
1430 {
1431         struct l2tp_session *session;
1432
1433         session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1434         if (session != NULL) {
1435                 session->magic = L2TP_SESSION_MAGIC;
1436                 session->tunnel = tunnel;
1437
1438                 session->session_id = session_id;
1439                 session->peer_session_id = peer_session_id;
1440                 session->nr = 1;
1441
1442                 sprintf(&session->name[0], "sess %u/%u",
1443                         tunnel->tunnel_id, session->session_id);
1444
1445                 skb_queue_head_init(&session->reorder_q);
1446
1447                 INIT_HLIST_NODE(&session->hlist);
1448                 INIT_HLIST_NODE(&session->global_hlist);
1449
1450                 /* Inherit debug options from tunnel */
1451                 session->debug = tunnel->debug;
1452
1453                 if (cfg) {
1454                         session->pwtype = cfg->pw_type;
1455                         session->debug = cfg->debug;
1456                         session->mtu = cfg->mtu;
1457                         session->mru = cfg->mru;
1458                         session->send_seq = cfg->send_seq;
1459                         session->recv_seq = cfg->recv_seq;
1460                         session->lns_mode = cfg->lns_mode;
1461                         session->reorder_timeout = cfg->reorder_timeout;
1462                         session->offset = cfg->offset;
1463                         session->l2specific_type = cfg->l2specific_type;
1464                         session->l2specific_len = cfg->l2specific_len;
1465                         session->cookie_len = cfg->cookie_len;
1466                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1467                         session->peer_cookie_len = cfg->peer_cookie_len;
1468                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1469                 }
1470
1471                 if (tunnel->version == L2TP_HDR_VER_2)
1472                         session->build_header = l2tp_build_l2tpv2_header;
1473                 else
1474                         session->build_header = l2tp_build_l2tpv3_header;
1475
1476                 l2tp_session_set_header_len(session, tunnel->version);
1477
1478                 /* Bump the reference count. The session context is deleted
1479                  * only when this drops to zero.
1480                  */
1481                 l2tp_session_inc_refcount(session);
1482                 l2tp_tunnel_inc_refcount(tunnel);
1483
1484                 /* Ensure tunnel socket isn't deleted */
1485                 sock_hold(tunnel->sock);
1486
1487                 /* Add session to the tunnel's hash list */
1488                 write_lock_bh(&tunnel->hlist_lock);
1489                 hlist_add_head(&session->hlist,
1490                                l2tp_session_id_hash(tunnel, session_id));
1491                 write_unlock_bh(&tunnel->hlist_lock);
1492
1493                 /* And to the global session list if L2TPv3 */
1494                 if (tunnel->version != L2TP_HDR_VER_2) {
1495                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1496
1497                         write_lock_bh(&pn->l2tp_session_hlist_lock);
1498                         hlist_add_head(&session->global_hlist,
1499                                        l2tp_session_id_hash_2(pn, session_id));
1500                         write_unlock_bh(&pn->l2tp_session_hlist_lock);
1501                 }
1502
1503                 /* Ignore management session in session count value */
1504                 if (session->session_id != 0)
1505                         atomic_inc(&l2tp_session_count);
1506         }
1507
1508         return session;
1509 }
1510 EXPORT_SYMBOL_GPL(l2tp_session_create);
1511
1512 /*****************************************************************************
1513  * Init and cleanup
1514  *****************************************************************************/
1515
1516 static __net_init int l2tp_init_net(struct net *net)
1517 {
1518         struct l2tp_net *pn;
1519         int err;
1520         int hash;
1521
1522         pn = kzalloc(sizeof(*pn), GFP_KERNEL);
1523         if (!pn)
1524                 return -ENOMEM;
1525
1526         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1527         rwlock_init(&pn->l2tp_tunnel_list_lock);
1528
1529         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1530                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1531
1532         rwlock_init(&pn->l2tp_session_hlist_lock);
1533
1534         err = net_assign_generic(net, l2tp_net_id, pn);
1535         if (err)
1536                 goto out;
1537
1538         return 0;
1539
1540 out:
1541         kfree(pn);
1542         return err;
1543 }
1544
1545 static __net_exit void l2tp_exit_net(struct net *net)
1546 {
1547         struct l2tp_net *pn;
1548
1549         pn = net_generic(net, l2tp_net_id);
1550         /*
1551          * if someone has cached our net then
1552          * further net_generic call will return NULL
1553          */
1554         net_assign_generic(net, l2tp_net_id, NULL);
1555         kfree(pn);
1556 }
1557
1558 static struct pernet_operations l2tp_net_ops = {
1559         .init = l2tp_init_net,
1560         .exit = l2tp_exit_net,
1561         .id   = &l2tp_net_id,
1562         .size = sizeof(struct l2tp_net),
1563 };
1564
1565 static int __init l2tp_init(void)
1566 {
1567         int rc = 0;
1568
1569         rc = register_pernet_device(&l2tp_net_ops);
1570         if (rc)
1571                 goto out;
1572
1573         printk(KERN_INFO "L2TP core driver, %s\n", L2TP_DRV_VERSION);
1574
1575 out:
1576         return rc;
1577 }
1578
1579 static void __exit l2tp_exit(void)
1580 {
1581         unregister_pernet_device(&l2tp_net_ops);
1582 }
1583
1584 module_init(l2tp_init);
1585 module_exit(l2tp_exit);
1586
1587 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1588 MODULE_DESCRIPTION("L2TP core");
1589 MODULE_LICENSE("GPL");
1590 MODULE_VERSION(L2TP_DRV_VERSION);
1591