SAFE public projects git trees. - safe/jmp/linux-2.6/blob - drivers/net/pppol2tp.c

   1 /*****************************************************************************
   2  * Linux PPP over L2TP (PPPoX/PPPoL2TP) Sockets
   3  *
   4  * PPPoX    --- Generic PPP encapsulation socket family
   5  * PPPoL2TP --- PPP over L2TP (RFC 2661)
   6  *
   7  * Version:     1.0.0
   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  * License:
  17  *              This program is free software; you can redistribute it and/or
  18  *              modify it under the terms of the GNU General Public License
  19  *              as published by the Free Software Foundation; either version
  20  *              2 of the License, or (at your option) any later version.
  21  *
  22  */
  23
  24 /* This driver handles only L2TP data frames; control frames are handled by a
  25  * userspace application.
  26  *
  27  * To send data in an L2TP session, userspace opens a PPPoL2TP socket and
  28  * attaches it to a bound UDP socket with local tunnel_id / session_id and
  29  * peer tunnel_id / session_id set. Data can then be sent or received using
  30  * regular socket sendmsg() / recvmsg() calls. Kernel parameters of the socket
  31  * can be read or modified using ioctl() or [gs]etsockopt() calls.
  32  *
  33  * When a PPPoL2TP socket is connected with local and peer session_id values
  34  * zero, the socket is treated as a special tunnel management socket.
  35  *
  36  * Here's example userspace code to create a socket for sending/receiving data
  37  * over an L2TP session:-
  38  *
  39  *      struct sockaddr_pppol2tp sax;
  40  *      int fd;
  41  *      int session_fd;
  42  *
  43  *      fd = socket(AF_PPPOX, SOCK_DGRAM, PX_PROTO_OL2TP);
  44  *
  45  *      sax.sa_family = AF_PPPOX;
  46  *      sax.sa_protocol = PX_PROTO_OL2TP;
  47  *      sax.pppol2tp.fd = tunnel_fd;    // bound UDP socket
  48  *      sax.pppol2tp.addr.sin_addr.s_addr = addr->sin_addr.s_addr;
  49  *      sax.pppol2tp.addr.sin_port = addr->sin_port;
  50  *      sax.pppol2tp.addr.sin_family = AF_INET;
  51  *      sax.pppol2tp.s_tunnel  = tunnel_id;
  52  *      sax.pppol2tp.s_session = session_id;
  53  *      sax.pppol2tp.d_tunnel  = peer_tunnel_id;
  54  *      sax.pppol2tp.d_session = peer_session_id;
  55  *
  56  *      session_fd = connect(fd, (struct sockaddr *)&sax, sizeof(sax));
  57  *
  58  * A pppd plugin that allows PPP traffic to be carried over L2TP using
  59  * this driver is available from the OpenL2TP project at
  60  * http://openl2tp.sourceforge.net.
  61  */
  62
  63 #include <linux/module.h>
  64 #include <linux/version.h>
  65 #include <linux/string.h>
  66 #include <linux/list.h>
  67 #include <asm/uaccess.h>
  68
  69 #include <linux/kernel.h>
  70 #include <linux/spinlock.h>
  71 #include <linux/kthread.h>
  72 #include <linux/sched.h>
  73 #include <linux/slab.h>
  74 #include <linux/errno.h>
  75 #include <linux/jiffies.h>
  76
  77 #include <linux/netdevice.h>
  78 #include <linux/net.h>
  79 #include <linux/inetdevice.h>
  80 #include <linux/skbuff.h>
  81 #include <linux/init.h>
  82 #include <linux/ip.h>
  83 #include <linux/udp.h>
  84 #include <linux/if_pppox.h>
  85 #include <linux/if_pppol2tp.h>
  86 #include <net/sock.h>
  87 #include <linux/ppp_channel.h>
  88 #include <linux/ppp_defs.h>
  89 #include <linux/if_ppp.h>
  90 #include <linux/file.h>
  91 #include <linux/hash.h>
  92 #include <linux/sort.h>
  93 #include <linux/proc_fs.h>
  94 #include <net/net_namespace.h>
  95 #include <net/dst.h>
  96 #include <net/ip.h>
  97 #include <net/udp.h>
  98 #include <net/xfrm.h>
  99
 100 #include <asm/byteorder.h>
 101 #include <asm/atomic.h>
 102
 103
 104 #define PPPOL2TP_DRV_VERSION    "V1.0"
 105
 106 /* L2TP header constants */
 107 #define L2TP_HDRFLAG_T     0x8000
 108 #define L2TP_HDRFLAG_L     0x4000
 109 #define L2TP_HDRFLAG_S     0x0800
 110 #define L2TP_HDRFLAG_O     0x0200
 111 #define L2TP_HDRFLAG_P     0x0100
 112
 113 #define L2TP_HDR_VER_MASK  0x000F
 114 #define L2TP_HDR_VER       0x0002
 115
 116 /* Space for UDP, L2TP and PPP headers */
 117 #define PPPOL2TP_HEADER_OVERHEAD        40
 118
 119 /* Just some random numbers */
 120 #define L2TP_TUNNEL_MAGIC       0x42114DDA
 121 #define L2TP_SESSION_MAGIC      0x0C04EB7D
 122
 123 #define PPPOL2TP_HASH_BITS      4
 124 #define PPPOL2TP_HASH_SIZE      (1 << PPPOL2TP_HASH_BITS)
 125
 126 /* Default trace flags */
 127 #define PPPOL2TP_DEFAULT_DEBUG_FLAGS    0
 128
 129 #define PRINTK(_mask, _type, _lvl, _fmt, args...)                       \
 130         do {                                                            \
 131                 if ((_mask) & (_type))                                  \
 132                         printk(_lvl "PPPOL2TP: " _fmt, ##args);         \
 133         } while(0)
 134
 135 /* Number of bytes to build transmit L2TP headers.
 136  * Unfortunately the size is different depending on whether sequence numbers
 137  * are enabled.
 138  */
 139 #define PPPOL2TP_L2TP_HDR_SIZE_SEQ              10
 140 #define PPPOL2TP_L2TP_HDR_SIZE_NOSEQ            6
 141
 142 struct pppol2tp_tunnel;
 143
 144 /* Describes a session. It is the sk_user_data field in the PPPoL2TP
 145  * socket. Contains information to determine incoming packets and transmit
 146  * outgoing ones.
 147  */
 148 struct pppol2tp_session
 149 {
 150         int                     magic;          /* should be
 151                                                  * L2TP_SESSION_MAGIC */
 152         int                     owner;          /* pid that opened the socket */
 153
 154         struct sock             *sock;          /* Pointer to the session
 155                                                  * PPPoX socket */
 156         struct sock             *tunnel_sock;   /* Pointer to the tunnel UDP
 157                                                  * socket */
 158
 159         struct pppol2tp_addr    tunnel_addr;    /* Description of tunnel */
 160
 161         struct pppol2tp_tunnel  *tunnel;        /* back pointer to tunnel
 162                                                  * context */
 163
 164         char                    name[20];       /* "sess xxxxx/yyyyy", where
 165                                                  * x=tunnel_id, y=session_id */
 166         int                     mtu;
 167         int                     mru;
 168         int                     flags;          /* accessed by PPPIOCGFLAGS.
 169                                                  * Unused. */
 170         unsigned                recv_seq:1;     /* expect receive packets with
 171                                                  * sequence numbers? */
 172         unsigned                send_seq:1;     /* send packets with sequence
 173                                                  * numbers? */
 174         unsigned                lns_mode:1;     /* behave as LNS? LAC enables
 175                                                  * sequence numbers under
 176                                                  * control of LNS. */
 177         int                     debug;          /* bitmask of debug message
 178                                                  * categories */
 179         int                     reorder_timeout; /* configured reorder timeout
 180                                                   * (in jiffies) */
 181         u16                     nr;             /* session NR state (receive) */
 182         u16                     ns;             /* session NR state (send) */
 183         struct sk_buff_head     reorder_q;      /* receive reorder queue */
 184         struct pppol2tp_ioc_stats stats;
 185         struct hlist_node       hlist;          /* Hash list node */
 186 };
 187
 188 /* The sk_user_data field of the tunnel's UDP socket. It contains info to track
 189  * all the associated sessions so incoming packets can be sorted out
 190  */
 191 struct pppol2tp_tunnel
 192 {
 193         int                     magic;          /* Should be L2TP_TUNNEL_MAGIC */
 194         rwlock_t                hlist_lock;     /* protect session_hlist */
 195         struct hlist_head       session_hlist[PPPOL2TP_HASH_SIZE];
 196                                                 /* hashed list of sessions,
 197                                                  * hashed by id */
 198         int                     debug;          /* bitmask of debug message
 199                                                  * categories */
 200         char                    name[12];       /* "tunl xxxxx" */
 201         struct pppol2tp_ioc_stats stats;
 202
 203         void (*old_sk_destruct)(struct sock *);
 204
 205         struct sock             *sock;          /* Parent socket */
 206         struct list_head        list;           /* Keep a list of all open
 207                                                  * prepared sockets */
 208
 209         atomic_t                ref_count;
 210 };
 211
 212 /* Private data stored for received packets in the skb.
 213  */
 214 struct pppol2tp_skb_cb {
 215         u16                     ns;
 216         u16                     nr;
 217         u16                     has_seq;
 218         u16                     length;
 219         unsigned long           expires;
 220 };
 221
 222 #define PPPOL2TP_SKB_CB(skb)    ((struct pppol2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
 223
 224 static int pppol2tp_xmit(struct ppp_channel *chan, struct sk_buff *skb);
 225 static void pppol2tp_tunnel_free(struct pppol2tp_tunnel *tunnel);
 226
 227 static atomic_t pppol2tp_tunnel_count;
 228 static atomic_t pppol2tp_session_count;
 229 static struct ppp_channel_ops pppol2tp_chan_ops = { pppol2tp_xmit , NULL };
 230 static struct proto_ops pppol2tp_ops;
 231 static LIST_HEAD(pppol2tp_tunnel_list);
 232 static DEFINE_RWLOCK(pppol2tp_tunnel_list_lock);
 233
 234 /* Helpers to obtain tunnel/session contexts from sockets.
 235  */
 236 static inline struct pppol2tp_session *pppol2tp_sock_to_session(struct sock *sk)
 237 {
 238         struct pppol2tp_session *session;
 239
 240         if (sk == NULL)
 241                 return NULL;
 242
 243         session = (struct pppol2tp_session *)(sk->sk_user_data);
 244         if (session == NULL)
 245                 return NULL;
 246
 247         BUG_ON(session->magic != L2TP_SESSION_MAGIC);
 248
 249         return session;
 250 }
 251
 252 static inline struct pppol2tp_tunnel *pppol2tp_sock_to_tunnel(struct sock *sk)
 253 {
 254         struct pppol2tp_tunnel *tunnel;
 255
 256         if (sk == NULL)
 257                 return NULL;
 258
 259         tunnel = (struct pppol2tp_tunnel *)(sk->sk_user_data);
 260         if (tunnel == NULL)
 261                 return NULL;
 262
 263         BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
 264
 265         return tunnel;
 266 }
 267
 268 /* Tunnel reference counts. Incremented per session that is added to
 269  * the tunnel.
 270  */
 271 static inline void pppol2tp_tunnel_inc_refcount(struct pppol2tp_tunnel *tunnel)
 272 {
 273         atomic_inc(&tunnel->ref_count);
 274 }
 275
 276 static inline void pppol2tp_tunnel_dec_refcount(struct pppol2tp_tunnel *tunnel)
 277 {
 278         if (atomic_dec_and_test(&tunnel->ref_count))
 279                 pppol2tp_tunnel_free(tunnel);
 280 }
 281
 282 /* Session hash list.
 283  * The session_id SHOULD be random according to RFC2661, but several
 284  * L2TP implementations (Cisco and Microsoft) use incrementing
 285  * session_ids.  So we do a real hash on the session_id, rather than a
 286  * simple bitmask.
 287  */
 288 static inline struct hlist_head *
 289 pppol2tp_session_id_hash(struct pppol2tp_tunnel *tunnel, u16 session_id)
 290 {
 291         unsigned long hash_val = (unsigned long) session_id;
 292         return &tunnel->session_hlist[hash_long(hash_val, PPPOL2TP_HASH_BITS)];
 293 }
 294
 295 /* Lookup a session by id
 296  */
 297 static struct pppol2tp_session *
 298 pppol2tp_session_find(struct pppol2tp_tunnel *tunnel, u16 session_id)
 299 {
 300         struct hlist_head *session_list =
 301                 pppol2tp_session_id_hash(tunnel, session_id);
 302         struct pppol2tp_session *session;
 303         struct hlist_node *walk;
 304
 305         read_lock_bh(&tunnel->hlist_lock);
 306         hlist_for_each_entry(session, walk, session_list, hlist) {
 307                 if (session->tunnel_addr.s_session == session_id) {
 308                         read_unlock_bh(&tunnel->hlist_lock);
 309                         return session;
 310                 }
 311         }
 312         read_unlock_bh(&tunnel->hlist_lock);
 313
 314         return NULL;
 315 }
 316
 317 /* Lookup a tunnel by id
 318  */
 319 static struct pppol2tp_tunnel *pppol2tp_tunnel_find(u16 tunnel_id)
 320 {
 321         struct pppol2tp_tunnel *tunnel = NULL;
 322
 323         read_lock_bh(&pppol2tp_tunnel_list_lock);
 324         list_for_each_entry(tunnel, &pppol2tp_tunnel_list, list) {
 325                 if (tunnel->stats.tunnel_id == tunnel_id) {
 326                         read_unlock_bh(&pppol2tp_tunnel_list_lock);
 327                         return tunnel;
 328                 }
 329         }
 330         read_unlock_bh(&pppol2tp_tunnel_list_lock);
 331
 332         return NULL;
 333 }
 334
 335 /*****************************************************************************
 336  * Receive data handling
 337  *****************************************************************************/
 338
 339 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
 340  * number.
 341  */
 342 static void pppol2tp_recv_queue_skb(struct pppol2tp_session *session, struct sk_buff *skb)
 343 {
 344         struct sk_buff *skbp;
 345         struct sk_buff *tmp;
 346         u16 ns = PPPOL2TP_SKB_CB(skb)->ns;
 347
 348         spin_lock_bh(&session->reorder_q.lock);
 349         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
 350                 if (PPPOL2TP_SKB_CB(skbp)->ns > ns) {
 351                         __skb_insert(skb, skbp->prev, skbp, &session->reorder_q);
 352                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
 353                                "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
 354                                session->name, ns, PPPOL2TP_SKB_CB(skbp)->ns,
 355                                skb_queue_len(&session->reorder_q));
 356                         session->stats.rx_oos_packets++;
 357                         goto out;
 358                 }
 359         }
 360
 361         __skb_queue_tail(&session->reorder_q, skb);
 362
 363 out:
 364         spin_unlock_bh(&session->reorder_q.lock);
 365 }
 366
 367 /* Dequeue a single skb.
 368  */
 369 static void pppol2tp_recv_dequeue_skb(struct pppol2tp_session *session, struct sk_buff *skb)
 370 {
 371         struct pppol2tp_tunnel *tunnel = session->tunnel;
 372         int length = PPPOL2TP_SKB_CB(skb)->length;
 373         struct sock *session_sock = NULL;
 374
 375         /* We're about to requeue the skb, so return resources
 376          * to its current owner (a socket receive buffer).
 377          */
 378         skb_orphan(skb);
 379
 380         tunnel->stats.rx_packets++;
 381         tunnel->stats.rx_bytes += length;
 382         session->stats.rx_packets++;
 383         session->stats.rx_bytes += length;
 384
 385         if (PPPOL2TP_SKB_CB(skb)->has_seq) {
 386                 /* Bump our Nr */
 387                 session->nr++;
 388                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
 389                        "%s: updated nr to %hu\n", session->name, session->nr);
 390         }
 391
 392         /* If the socket is bound, send it in to PPP's input queue. Otherwise
 393          * queue it on the session socket.
 394          */
 395         session_sock = session->sock;
 396         if (session_sock->sk_state & PPPOX_BOUND) {
 397                 struct pppox_sock *po;
 398                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
 399                        "%s: recv %d byte data frame, passing to ppp\n",
 400                        session->name, length);
 401
 402                 /* We need to forget all info related to the L2TP packet
 403                  * gathered in the skb as we are going to reuse the same
 404                  * skb for the inner packet.
 405                  * Namely we need to:
 406                  * - reset xfrm (IPSec) information as it applies to
 407                  *   the outer L2TP packet and not to the inner one
 408                  * - release the dst to force a route lookup on the inner
 409                  *   IP packet since skb->dst currently points to the dst
 410                  *   of the UDP tunnel
 411                  * - reset netfilter information as it doesn't apply
 412                  *   to the inner packet either
 413                  */
 414                 secpath_reset(skb);
 415                 dst_release(skb->dst);
 416                 skb->dst = NULL;
 417                 nf_reset(skb);
 418
 419                 po = pppox_sk(session_sock);
 420                 ppp_input(&po->chan, skb);
 421         } else {
 422                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
 423                        "%s: socket not bound\n", session->name);
 424
 425                 /* Not bound. Nothing we can do, so discard. */
 426                 session->stats.rx_errors++;
 427                 kfree_skb(skb);
 428         }
 429
 430         sock_put(session->sock);
 431 }
 432
 433 /* Dequeue skbs from the session's reorder_q, subject to packet order.
 434  * Skbs that have been in the queue for too long are simply discarded.
 435  */
 436 static void pppol2tp_recv_dequeue(struct pppol2tp_session *session)
 437 {
 438         struct sk_buff *skb;
 439         struct sk_buff *tmp;
 440
 441         /* If the pkt at the head of the queue has the nr that we
 442          * expect to send up next, dequeue it and any other
 443          * in-sequence packets behind it.
 444          */
 445         spin_lock_bh(&session->reorder_q.lock);
 446         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
 447                 if (time_after(jiffies, PPPOL2TP_SKB_CB(skb)->expires)) {
 448                         session->stats.rx_seq_discards++;
 449                         session->stats.rx_errors++;
 450                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
 451                                "%s: oos pkt %hu len %d discarded (too old), "
 452                                "waiting for %hu, reorder_q_len=%d\n",
 453                                session->name, PPPOL2TP_SKB_CB(skb)->ns,
 454                                PPPOL2TP_SKB_CB(skb)->length, session->nr,
 455                                skb_queue_len(&session->reorder_q));
 456                         __skb_unlink(skb, &session->reorder_q);
 457                         kfree_skb(skb);
 458                         sock_put(session->sock);
 459                         continue;
 460                 }
 461
 462                 if (PPPOL2TP_SKB_CB(skb)->has_seq) {
 463                         if (PPPOL2TP_SKB_CB(skb)->ns != session->nr) {
 464                                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
 465                                        "%s: holding oos pkt %hu len %d, "
 466                                        "waiting for %hu, reorder_q_len=%d\n",
 467                                        session->name, PPPOL2TP_SKB_CB(skb)->ns,
 468                                        PPPOL2TP_SKB_CB(skb)->length, session->nr,
 469                                        skb_queue_len(&session->reorder_q));
 470                                 goto out;
 471                         }
 472                 }
 473                 __skb_unlink(skb, &session->reorder_q);
 474
 475                 /* Process the skb. We release the queue lock while we
 476                  * do so to let other contexts process the queue.
 477                  */
 478                 spin_unlock_bh(&session->reorder_q.lock);
 479                 pppol2tp_recv_dequeue_skb(session, skb);
 480                 spin_lock_bh(&session->reorder_q.lock);
 481         }
 482
 483 out:
 484         spin_unlock_bh(&session->reorder_q.lock);
 485 }
 486
 487 /* Internal receive frame. Do the real work of receiving an L2TP data frame
 488  * here. The skb is not on a list when we get here.
 489  * Returns 0 if the packet was a data packet and was successfully passed on.
 490  * Returns 1 if the packet was not a good data packet and could not be
 491  * forwarded.  All such packets are passed up to userspace to deal with.
 492  */
 493 static int pppol2tp_recv_core(struct sock *sock, struct sk_buff *skb)
 494 {
 495         struct pppol2tp_session *session = NULL;
 496         struct pppol2tp_tunnel *tunnel;
 497         unsigned char *ptr, *optr;
 498         u16 hdrflags;
 499         u16 tunnel_id, session_id;
 500         int length;
 501         int offset;
 502
 503         tunnel = pppol2tp_sock_to_tunnel(sock);
 504         if (tunnel == NULL)
 505                 goto no_tunnel;
 506
 507         /* UDP always verifies the packet length. */
 508         __skb_pull(skb, sizeof(struct udphdr));
 509
 510         /* Short packet? */
 511         if (!pskb_may_pull(skb, 12)) {
 512                 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
 513                        "%s: recv short packet (len=%d)\n", tunnel->name, skb->len);
 514                 goto error;
 515         }
 516
 517         /* Point to L2TP header */
 518         optr = ptr = skb->data;
 519
 520         /* Get L2TP header flags */
 521         hdrflags = ntohs(*(__be16*)ptr);
 522
 523         /* Trace packet contents, if enabled */
 524         if (tunnel->debug & PPPOL2TP_MSG_DATA) {
 525                 length = min(16u, skb->len);
 526                 if (!pskb_may_pull(skb, length))
 527                         goto error;
 528
 529                 printk(KERN_DEBUG "%s: recv: ", tunnel->name);
 530
 531                 offset = 0;
 532                 do {
 533                         printk(" %02X", ptr[offset]);
 534                 } while (++offset < length);
 535
 536                 printk("\n");
 537         }
 538
 539         /* Get length of L2TP packet */
 540         length = skb->len;
 541
 542         /* If type is control packet, it is handled by userspace. */
 543         if (hdrflags & L2TP_HDRFLAG_T) {
 544                 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
 545                        "%s: recv control packet, len=%d\n", tunnel->name, length);
 546                 goto error;
 547         }
 548
 549         /* Skip flags */
 550         ptr += 2;
 551
 552         /* If length is present, skip it */
 553         if (hdrflags & L2TP_HDRFLAG_L)
 554                 ptr += 2;
 555
 556         /* Extract tunnel and session ID */
 557         tunnel_id = ntohs(*(__be16 *) ptr);
 558         ptr += 2;
 559         session_id = ntohs(*(__be16 *) ptr);
 560         ptr += 2;
 561
 562         /* Find the session context */
 563         session = pppol2tp_session_find(tunnel, session_id);
 564         if (!session) {
 565                 /* Not found? Pass to userspace to deal with */
 566                 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
 567                        "%s: no socket found (%hu/%hu). Passing up.\n",
 568                        tunnel->name, tunnel_id, session_id);
 569                 goto error;
 570         }
 571         sock_hold(session->sock);
 572
 573         /* The ref count on the socket was increased by the above call since
 574          * we now hold a pointer to the session. Take care to do sock_put()
 575          * when exiting this function from now on...
 576          */
 577
 578         /* Handle the optional sequence numbers.  If we are the LAC,
 579          * enable/disable sequence numbers under the control of the LNS.  If
 580          * no sequence numbers present but we were expecting them, discard
 581          * frame.
 582          */
 583         if (hdrflags & L2TP_HDRFLAG_S) {
 584                 u16 ns, nr;
 585                 ns = ntohs(*(__be16 *) ptr);
 586                 ptr += 2;
 587                 nr = ntohs(*(__be16 *) ptr);
 588                 ptr += 2;
 589
 590                 /* Received a packet with sequence numbers. If we're the LNS,
 591                  * check if we sre sending sequence numbers and if not,
 592                  * configure it so.
 593                  */
 594                 if ((!session->lns_mode) && (!session->send_seq)) {
 595                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_INFO,
 596                                "%s: requested to enable seq numbers by LNS\n",
 597                                session->name);
 598                         session->send_seq = -1;
 599                 }
 600
 601                 /* Store L2TP info in the skb */
 602                 PPPOL2TP_SKB_CB(skb)->ns = ns;
 603                 PPPOL2TP_SKB_CB(skb)->nr = nr;
 604                 PPPOL2TP_SKB_CB(skb)->has_seq = 1;
 605
 606                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
 607                        "%s: recv data ns=%hu, nr=%hu, session nr=%hu\n",
 608                        session->name, ns, nr, session->nr);
 609         } else {
 610                 /* No sequence numbers.
 611                  * If user has configured mandatory sequence numbers, discard.
 612                  */
 613                 if (session->recv_seq) {
 614                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_WARNING,
 615                                "%s: recv data has no seq numbers when required. "
 616                                "Discarding\n", session->name);
 617                         session->stats.rx_seq_discards++;
 618                         goto discard;
 619                 }
 620
 621                 /* If we're the LAC and we're sending sequence numbers, the
 622                  * LNS has requested that we no longer send sequence numbers.
 623                  * If we're the LNS and we're sending sequence numbers, the
 624                  * LAC is broken. Discard the frame.
 625                  */
 626                 if ((!session->lns_mode) && (session->send_seq)) {
 627                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_INFO,
 628                                "%s: requested to disable seq numbers by LNS\n",
 629                                session->name);
 630                         session->send_seq = 0;
 631                 } else if (session->send_seq) {
 632                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_WARNING,
 633                                "%s: recv data has no seq numbers when required. "
 634                                "Discarding\n", session->name);
 635                         session->stats.rx_seq_discards++;
 636                         goto discard;
 637                 }
 638
 639                 /* Store L2TP info in the skb */
 640                 PPPOL2TP_SKB_CB(skb)->has_seq = 0;
 641         }
 642
 643         /* If offset bit set, skip it. */
 644         if (hdrflags & L2TP_HDRFLAG_O) {
 645                 offset = ntohs(*(__be16 *)ptr);
 646                 ptr += 2 + offset;
 647         }
 648
 649         offset = ptr - optr;
 650         if (!pskb_may_pull(skb, offset))
 651                 goto discard;
 652
 653         __skb_pull(skb, offset);
 654
 655         /* Skip PPP header, if present.  In testing, Microsoft L2TP clients
 656          * don't send the PPP header (PPP header compression enabled), but
 657          * other clients can include the header. So we cope with both cases
 658          * here. The PPP header is always FF03 when using L2TP.
 659          *
 660          * Note that skb->data[] isn't dereferenced from a u16 ptr here since
 661          * the field may be unaligned.
 662          */
 663         if (!pskb_may_pull(skb, 2))
 664                 goto discard;
 665
 666         if ((skb->data[0] == 0xff) && (skb->data[1] == 0x03))
 667                 skb_pull(skb, 2);
 668
 669         /* Prepare skb for adding to the session's reorder_q.  Hold
 670          * packets for max reorder_timeout or 1 second if not
 671          * reordering.
 672          */
 673         PPPOL2TP_SKB_CB(skb)->length = length;
 674         PPPOL2TP_SKB_CB(skb)->expires = jiffies +
 675                 (session->reorder_timeout ? session->reorder_timeout : HZ);
 676
 677         /* Add packet to the session's receive queue. Reordering is done here, if
 678          * enabled. Saved L2TP protocol info is stored in skb->sb[].
 679          */
 680         if (PPPOL2TP_SKB_CB(skb)->has_seq) {
 681                 if (session->reorder_timeout != 0) {
 682                         /* Packet reordering enabled. Add skb to session's
 683                          * reorder queue, in order of ns.
 684                          */
 685                         pppol2tp_recv_queue_skb(session, skb);
 686                 } else {
 687                         /* Packet reordering disabled. Discard out-of-sequence
 688                          * packets
 689                          */
 690                         if (PPPOL2TP_SKB_CB(skb)->ns != session->nr) {
 691                                 session->stats.rx_seq_discards++;
 692                                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
 693                                        "%s: oos pkt %hu len %d discarded, "
 694                                        "waiting for %hu, reorder_q_len=%d\n",
 695                                        session->name, PPPOL2TP_SKB_CB(skb)->ns,
 696                                        PPPOL2TP_SKB_CB(skb)->length, session->nr,
 697                                        skb_queue_len(&session->reorder_q));
 698                                 goto discard;
 699                         }
 700                         skb_queue_tail(&session->reorder_q, skb);
 701                 }
 702         } else {
 703                 /* No sequence numbers. Add the skb to the tail of the
 704                  * reorder queue. This ensures that it will be
 705                  * delivered after all previous sequenced skbs.
 706                  */
 707                 skb_queue_tail(&session->reorder_q, skb);
 708         }
 709
 710         /* Try to dequeue as many skbs from reorder_q as we can. */
 711         pppol2tp_recv_dequeue(session);
 712
 713         return 0;
 714
 715 discard:
 716         session->stats.rx_errors++;
 717         kfree_skb(skb);
 718         sock_put(session->sock);
 719
 720         return 0;
 721
 722 error:
 723         /* Put UDP header back */
 724         __skb_push(skb, sizeof(struct udphdr));
 725
 726 no_tunnel:
 727         return 1;
 728 }
 729
 730 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
 731  * Return codes:
 732  * 0 : success.
 733  * <0: error
 734  * >0: skb should be passed up to userspace as UDP.
 735  */
 736 static int pppol2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
 737 {
 738         struct pppol2tp_tunnel *tunnel;
 739
 740         tunnel = pppol2tp_sock_to_tunnel(sk);
 741         if (tunnel == NULL)
 742                 goto pass_up;
 743
 744         PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
 745                "%s: received %d bytes\n", tunnel->name, skb->len);
 746
 747         if (pppol2tp_recv_core(sk, skb))
 748                 goto pass_up;
 749
 750         return 0;
 751
 752 pass_up:
 753         return 1;
 754 }
 755
 756 /* Receive message. This is the recvmsg for the PPPoL2TP socket.
 757  */
 758 static int pppol2tp_recvmsg(struct kiocb *iocb, struct socket *sock,
 759                             struct msghdr *msg, size_t len,
 760                             int flags)
 761 {
 762         int err;
 763         struct sk_buff *skb;
 764         struct sock *sk = sock->sk;
 765
 766         err = -EIO;
 767         if (sk->sk_state & PPPOX_BOUND)
 768                 goto end;
 769
 770         msg->msg_namelen = 0;
 771
 772         err = 0;
 773         skb = skb_recv_datagram(sk, flags & ~MSG_DONTWAIT,
 774                                 flags & MSG_DONTWAIT, &err);
 775         if (skb) {
 776                 err = memcpy_toiovec(msg->msg_iov, (unsigned char *) skb->data,
 777                                      skb->len);
 778                 if (err < 0)
 779                         goto do_skb_free;
 780                 err = skb->len;
 781         }
 782 do_skb_free:
 783         kfree_skb(skb);
 784 end:
 785         return err;
 786 }
 787
 788 /************************************************************************
 789  * Transmit handling
 790  ***********************************************************************/
 791
 792 /* Tell how big L2TP headers are for a particular session. This
 793  * depends on whether sequence numbers are being used.
 794  */
 795 static inline int pppol2tp_l2tp_header_len(struct pppol2tp_session *session)
 796 {
 797         if (session->send_seq)
 798                 return PPPOL2TP_L2TP_HDR_SIZE_SEQ;
 799
 800         return PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
 801 }
 802
 803 /* Build an L2TP header for the session into the buffer provided.
 804  */
 805 static void pppol2tp_build_l2tp_header(struct pppol2tp_session *session,
 806                                        void *buf)
 807 {
 808         __be16 *bufp = buf;
 809         u16 flags = L2TP_HDR_VER;
 810
 811         if (session->send_seq)
 812                 flags |= L2TP_HDRFLAG_S;
 813
 814         /* Setup L2TP header.
 815          * FIXME: Can this ever be unaligned? Is direct dereferencing of
 816          * 16-bit header fields safe here for all architectures?
 817          */
 818         *bufp++ = htons(flags);
 819         *bufp++ = htons(session->tunnel_addr.d_tunnel);
 820         *bufp++ = htons(session->tunnel_addr.d_session);
 821         if (session->send_seq) {
 822                 *bufp++ = htons(session->ns);
 823                 *bufp++ = 0;
 824                 session->ns++;
 825                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
 826                        "%s: updated ns to %hu\n", session->name, session->ns);
 827         }
 828 }
 829
 830 /* This is the sendmsg for the PPPoL2TP pppol2tp_session socket.  We come here
 831  * when a user application does a sendmsg() on the session socket. L2TP and
 832  * PPP headers must be inserted into the user's data.
 833  */
 834 static int pppol2tp_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
 835                             size_t total_len)
 836 {
 837         static const unsigned char ppph[2] = { 0xff, 0x03 };
 838         struct sock *sk = sock->sk;
 839         struct inet_sock *inet;
 840         __wsum csum = 0;
 841         struct sk_buff *skb;
 842         int error;
 843         int hdr_len;
 844         struct pppol2tp_session *session;
 845         struct pppol2tp_tunnel *tunnel;
 846         struct udphdr *uh;
 847         unsigned int len;
 848
 849         error = -ENOTCONN;
 850         if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED))
 851                 goto error;
 852
 853         /* Get session and tunnel contexts */
 854         error = -EBADF;
 855         session = pppol2tp_sock_to_session(sk);
 856         if (session == NULL)
 857                 goto error;
 858
 859         tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
 860         if (tunnel == NULL)
 861                 goto error;
 862
 863         /* What header length is configured for this session? */
 864         hdr_len = pppol2tp_l2tp_header_len(session);
 865
 866         /* Allocate a socket buffer */
 867         error = -ENOMEM;
 868         skb = sock_wmalloc(sk, NET_SKB_PAD + sizeof(struct iphdr) +
 869                            sizeof(struct udphdr) + hdr_len +
 870                            sizeof(ppph) + total_len,
 871                            0, GFP_KERNEL);
 872         if (!skb)
 873                 goto error;
 874
 875         /* Reserve space for headers. */
 876         skb_reserve(skb, NET_SKB_PAD);
 877         skb_reset_network_header(skb);
 878         skb_reserve(skb, sizeof(struct iphdr));
 879         skb_reset_transport_header(skb);
 880
 881         /* Build UDP header */
 882         inet = inet_sk(session->tunnel_sock);
 883         uh = (struct udphdr *) skb->data;
 884         uh->source = inet->sport;
 885         uh->dest = inet->dport;
 886         uh->len = htons(hdr_len + sizeof(ppph) + total_len);
 887         uh->check = 0;
 888         skb_put(skb, sizeof(struct udphdr));
 889
 890         /* Build L2TP header */
 891         pppol2tp_build_l2tp_header(session, skb->data);
 892         skb_put(skb, hdr_len);
 893
 894         /* Add PPP header */
 895         skb->data[0] = ppph[0];
 896         skb->data[1] = ppph[1];
 897         skb_put(skb, 2);
 898
 899         /* Copy user data into skb */
 900         error = memcpy_fromiovec(skb->data, m->msg_iov, total_len);
 901         if (error < 0) {
 902                 kfree_skb(skb);
 903                 goto error;
 904         }
 905         skb_put(skb, total_len);
 906
 907         /* Calculate UDP checksum if configured to do so */
 908         if (session->tunnel_sock->sk_no_check != UDP_CSUM_NOXMIT)
 909                 csum = udp_csum_outgoing(sk, skb);
 910
 911         /* Debug */
 912         if (session->send_seq)
 913                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
 914                        "%s: send %Zd bytes, ns=%hu\n", session->name,
 915                        total_len, session->ns - 1);
 916         else
 917                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
 918                        "%s: send %Zd bytes\n", session->name, total_len);
 919
 920         if (session->debug & PPPOL2TP_MSG_DATA) {
 921                 int i;
 922                 unsigned char *datap = skb->data;
 923
 924                 printk(KERN_DEBUG "%s: xmit:", session->name);
 925                 for (i = 0; i < total_len; i++) {
 926                         printk(" %02X", *datap++);
 927                         if (i == 15) {
 928                                 printk(" ...");
 929                                 break;
 930                         }
 931                 }
 932                 printk("\n");
 933         }
 934
 935         /* Queue the packet to IP for output */
 936         len = skb->len;
 937         error = ip_queue_xmit(skb, 1);
 938
 939         /* Update stats */
 940         if (error >= 0) {
 941                 tunnel->stats.tx_packets++;
 942                 tunnel->stats.tx_bytes += len;
 943                 session->stats.tx_packets++;
 944                 session->stats.tx_bytes += len;
 945         } else {
 946                 tunnel->stats.tx_errors++;
 947                 session->stats.tx_errors++;
 948         }
 949
 950 error:
 951         return error;
 952 }
 953
 954 /* Transmit function called by generic PPP driver.  Sends PPP frame
 955  * over PPPoL2TP socket.
 956  *
 957  * This is almost the same as pppol2tp_sendmsg(), but rather than
 958  * being called with a msghdr from userspace, it is called with a skb
 959  * from the kernel.
 960  *
 961  * The supplied skb from ppp doesn't have enough headroom for the
 962  * insertion of L2TP, UDP and IP headers so we need to allocate more
 963  * headroom in the skb. This will create a cloned skb. But we must be
 964  * careful in the error case because the caller will expect to free
 965  * the skb it supplied, not our cloned skb. So we take care to always
 966  * leave the original skb unfreed if we return an error.
 967  */
 968 static int pppol2tp_xmit(struct ppp_channel *chan, struct sk_buff *skb)
 969 {
 970         static const u8 ppph[2] = { 0xff, 0x03 };
 971         struct sock *sk = (struct sock *) chan->private;
 972         struct sock *sk_tun;
 973         int hdr_len;
 974         struct pppol2tp_session *session;
 975         struct pppol2tp_tunnel *tunnel;
 976         int rc;
 977         int headroom;
 978         int data_len = skb->len;
 979         struct inet_sock *inet;
 980         __wsum csum = 0;
 981         struct udphdr *uh;
 982         unsigned int len;
 983         int old_headroom;
 984         int new_headroom;
 985
 986         if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED))
 987                 goto abort;
 988
 989         /* Get session and tunnel contexts from the socket */
 990         session = pppol2tp_sock_to_session(sk);
 991         if (session == NULL)
 992                 goto abort;
 993
 994         sk_tun = session->tunnel_sock;
 995         if (sk_tun == NULL)
 996                 goto abort;
 997         tunnel = pppol2tp_sock_to_tunnel(sk_tun);
 998         if (tunnel == NULL)
 999                 goto abort;
1000
1001         /* What header length is configured for this session? */
1002         hdr_len = pppol2tp_l2tp_header_len(session);
1003
1004         /* Check that there's enough headroom in the skb to insert IP,
1005          * UDP and L2TP and PPP headers. If not enough, expand it to
1006          * make room. Adjust truesize.
1007          */
1008         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1009                 sizeof(struct udphdr) + hdr_len + sizeof(ppph);
1010         old_headroom = skb_headroom(skb);
1011         if (skb_cow_head(skb, headroom))
1012                 goto abort;
1013
1014         new_headroom = skb_headroom(skb);
1015         skb_orphan(skb);
1016         skb->truesize += new_headroom - old_headroom;
1017
1018         /* Setup PPP header */
1019         __skb_push(skb, sizeof(ppph));
1020         skb->data[0] = ppph[0];
1021         skb->data[1] = ppph[1];
1022
1023         /* Setup L2TP header */
1024         pppol2tp_build_l2tp_header(session, __skb_push(skb, hdr_len));
1025
1026         /* Setup UDP header */
1027         inet = inet_sk(sk_tun);
1028         __skb_push(skb, sizeof(*uh));
1029         skb_reset_transport_header(skb);
1030         uh = udp_hdr(skb);
1031         uh->source = inet->sport;
1032         uh->dest = inet->dport;
1033         uh->len = htons(sizeof(struct udphdr) + hdr_len + sizeof(ppph) + data_len);
1034         uh->check = 0;
1035
1036         /* *BROKEN* Calculate UDP checksum if configured to do so */
1037         if (sk_tun->sk_no_check != UDP_CSUM_NOXMIT)
1038                 csum = udp_csum_outgoing(sk_tun, skb);
1039
1040         /* Debug */
1041         if (session->send_seq)
1042                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
1043                        "%s: send %d bytes, ns=%hu\n", session->name,
1044                        data_len, session->ns - 1);
1045         else
1046                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
1047                        "%s: send %d bytes\n", session->name, data_len);
1048
1049         if (session->debug & PPPOL2TP_MSG_DATA) {
1050                 int i;
1051                 unsigned char *datap = skb->data;
1052
1053                 printk(KERN_DEBUG "%s: xmit:", session->name);
1054                 for (i = 0; i < data_len; i++) {
1055                         printk(" %02X", *datap++);
1056                         if (i == 31) {
1057                                 printk(" ...");
1058                                 break;
1059                         }
1060                 }
1061                 printk("\n");
1062         }
1063
1064         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1065         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1066                               IPSKB_REROUTED);
1067         nf_reset(skb);
1068
1069         /* Get routing info from the tunnel socket */
1070         dst_release(skb->dst);
1071         skb->dst = dst_clone(__sk_dst_get(sk_tun));
1072         skb->sk = sk_tun;
1073
1074         /* Queue the packet to IP for output */
1075         len = skb->len;
1076         rc = ip_queue_xmit(skb, 1);
1077
1078         /* Update stats */
1079         if (rc >= 0) {
1080                 tunnel->stats.tx_packets++;
1081                 tunnel->stats.tx_bytes += len;
1082                 session->stats.tx_packets++;
1083                 session->stats.tx_bytes += len;
1084         } else {
1085                 tunnel->stats.tx_errors++;
1086                 session->stats.tx_errors++;
1087         }
1088
1089         return 1;
1090
1091 abort:
1092         /* Free the original skb */
1093         kfree_skb(skb);
1094         return 1;
1095 }
1096
1097 /*****************************************************************************
1098  * Session (and tunnel control) socket create/destroy.
1099  *****************************************************************************/
1100
1101 /* When the tunnel UDP socket is closed, all the attached sockets need to go
1102  * too.
1103  */
1104 static void pppol2tp_tunnel_closeall(struct pppol2tp_tunnel *tunnel)
1105 {
1106         int hash;
1107         struct hlist_node *walk;
1108         struct hlist_node *tmp;
1109         struct pppol2tp_session *session;
1110         struct sock *sk;
1111
1112         if (tunnel == NULL)
1113                 BUG();
1114
1115         PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1116                "%s: closing all sessions...\n", tunnel->name);
1117
1118         write_lock_bh(&tunnel->hlist_lock);
1119         for (hash = 0; hash < PPPOL2TP_HASH_SIZE; hash++) {
1120 again:
1121                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1122                         struct sk_buff *skb;
1123
1124                         session = hlist_entry(walk, struct pppol2tp_session, hlist);
1125
1126                         sk = session->sock;
1127
1128                         PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1129                                "%s: closing session\n", session->name);
1130
1131                         hlist_del_init(&session->hlist);
1132
1133                         /* Since we should hold the sock lock while
1134                          * doing any unbinding, we need to release the
1135                          * lock we're holding before taking that lock.
1136                          * Hold a reference to the sock so it doesn't
1137                          * disappear as we're jumping between locks.
1138                          */
1139                         sock_hold(sk);
1140                         write_unlock_bh(&tunnel->hlist_lock);
1141                         lock_sock(sk);
1142
1143                         if (sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND)) {
1144                                 pppox_unbind_sock(sk);
1145                                 sk->sk_state = PPPOX_DEAD;
1146                                 sk->sk_state_change(sk);
1147                         }
1148
1149                         /* Purge any queued data */
1150                         skb_queue_purge(&sk->sk_receive_queue);
1151                         skb_queue_purge(&sk->sk_write_queue);
1152                         while ((skb = skb_dequeue(&session->reorder_q))) {
1153                                 kfree_skb(skb);
1154                                 sock_put(sk);
1155                         }
1156
1157                         release_sock(sk);
1158                         sock_put(sk);
1159
1160                         /* Now restart from the beginning of this hash
1161                          * chain.  We always remove a session from the
1162                          * list so we are guaranteed to make forward
1163                          * progress.
1164                          */
1165                         write_lock_bh(&tunnel->hlist_lock);
1166                         goto again;
1167                 }
1168         }
1169         write_unlock_bh(&tunnel->hlist_lock);
1170 }
1171
1172 /* Really kill the tunnel.
1173  * Come here only when all sessions have been cleared from the tunnel.
1174  */
1175 static void pppol2tp_tunnel_free(struct pppol2tp_tunnel *tunnel)
1176 {
1177         /* Remove from socket list */
1178         write_lock_bh(&pppol2tp_tunnel_list_lock);
1179         list_del_init(&tunnel->list);
1180         write_unlock_bh(&pppol2tp_tunnel_list_lock);
1181
1182         atomic_dec(&pppol2tp_tunnel_count);
1183         kfree(tunnel);
1184 }
1185
1186 /* Tunnel UDP socket destruct hook.
1187  * The tunnel context is deleted only when all session sockets have been
1188  * closed.
1189  */
1190 static void pppol2tp_tunnel_destruct(struct sock *sk)
1191 {
1192         struct pppol2tp_tunnel *tunnel;
1193
1194         tunnel = pppol2tp_sock_to_tunnel(sk);
1195         if (tunnel == NULL)
1196                 goto end;
1197
1198         PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1199                "%s: closing...\n", tunnel->name);
1200
1201         /* Close all sessions */
1202         pppol2tp_tunnel_closeall(tunnel);
1203
1204         /* No longer an encapsulation socket. See net/ipv4/udp.c */
1205         (udp_sk(sk))->encap_type = 0;
1206         (udp_sk(sk))->encap_rcv = NULL;
1207
1208         /* Remove hooks into tunnel socket */
1209         tunnel->sock = NULL;
1210         sk->sk_destruct = tunnel->old_sk_destruct;
1211         sk->sk_user_data = NULL;
1212
1213         /* Call original (UDP) socket descructor */
1214         if (sk->sk_destruct != NULL)
1215                 (*sk->sk_destruct)(sk);
1216
1217         pppol2tp_tunnel_dec_refcount(tunnel);
1218
1219 end:
1220         return;
1221 }
1222
1223 /* Really kill the session socket. (Called from sock_put() if
1224  * refcnt == 0.)
1225  */
1226 static void pppol2tp_session_destruct(struct sock *sk)
1227 {
1228         struct pppol2tp_session *session = NULL;
1229
1230         if (sk->sk_user_data != NULL) {
1231                 struct pppol2tp_tunnel *tunnel;
1232
1233                 session = pppol2tp_sock_to_session(sk);
1234                 if (session == NULL)
1235                         goto out;
1236
1237                 /* Don't use pppol2tp_sock_to_tunnel() here to
1238                  * get the tunnel context because the tunnel
1239                  * socket might have already been closed (its
1240                  * sk->sk_user_data will be NULL) so use the
1241                  * session's private tunnel ptr instead.
1242                  */
1243                 tunnel = session->tunnel;
1244                 if (tunnel != NULL) {
1245                         BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1246
1247                         /* If session_id is zero, this is a null
1248                          * session context, which was created for a
1249                          * socket that is being used only to manage
1250                          * tunnels.
1251                          */
1252                         if (session->tunnel_addr.s_session != 0) {
1253                                 /* Delete the session socket from the
1254                                  * hash
1255                                  */
1256                                 write_lock_bh(&tunnel->hlist_lock);
1257                                 hlist_del_init(&session->hlist);
1258                                 write_unlock_bh(&tunnel->hlist_lock);
1259
1260                                 atomic_dec(&pppol2tp_session_count);
1261                         }
1262
1263                         /* This will delete the tunnel context if this
1264                          * is the last session on the tunnel.
1265                          */
1266                         session->tunnel = NULL;
1267                         session->tunnel_sock = NULL;
1268                         pppol2tp_tunnel_dec_refcount(tunnel);
1269                 }
1270         }
1271
1272         kfree(session);
1273 out:
1274         return;
1275 }
1276
1277 /* Called when the PPPoX socket (session) is closed.
1278  */
1279 static int pppol2tp_release(struct socket *sock)
1280 {
1281         struct sock *sk = sock->sk;
1282         struct pppol2tp_session *session;
1283         int error;
1284
1285         if (!sk)
1286                 return 0;
1287
1288         error = -EBADF;
1289         lock_sock(sk);
1290         if (sock_flag(sk, SOCK_DEAD) != 0)
1291                 goto error;
1292
1293         pppox_unbind_sock(sk);
1294
1295         /* Signal the death of the socket. */
1296         sk->sk_state = PPPOX_DEAD;
1297         sock_orphan(sk);
1298         sock->sk = NULL;
1299
1300         session = pppol2tp_sock_to_session(sk);
1301
1302         /* Purge any queued data */
1303         skb_queue_purge(&sk->sk_receive_queue);
1304         skb_queue_purge(&sk->sk_write_queue);
1305         if (session != NULL) {
1306                 struct sk_buff *skb;
1307                 while ((skb = skb_dequeue(&session->reorder_q))) {
1308                         kfree_skb(skb);
1309                         sock_put(sk);
1310                 }
1311         }
1312
1313         release_sock(sk);
1314
1315         /* This will delete the session context via
1316          * pppol2tp_session_destruct() if the socket's refcnt drops to
1317          * zero.
1318          */
1319         sock_put(sk);
1320
1321         return 0;
1322
1323 error:
1324         release_sock(sk);
1325         return error;
1326 }
1327
1328 /* Internal function to prepare a tunnel (UDP) socket to have PPPoX
1329  * sockets attached to it.
1330  */
1331 static struct sock *pppol2tp_prepare_tunnel_socket(int fd, u16 tunnel_id,
1332                                                    int *error)
1333 {
1334         int err;
1335         struct socket *sock = NULL;
1336         struct sock *sk;
1337         struct pppol2tp_tunnel *tunnel;
1338         struct sock *ret = NULL;
1339
1340         /* Get the tunnel UDP socket from the fd, which was opened by
1341          * the userspace L2TP daemon.
1342          */
1343         err = -EBADF;
1344         sock = sockfd_lookup(fd, &err);
1345         if (!sock) {
1346                 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1347                        "tunl %hu: sockfd_lookup(fd=%d) returned %d\n",
1348                        tunnel_id, fd, err);
1349                 goto err;
1350         }
1351
1352         sk = sock->sk;
1353
1354         /* Quick sanity checks */
1355         err = -EPROTONOSUPPORT;
1356         if (sk->sk_protocol != IPPROTO_UDP) {
1357                 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1358                        "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1359                        tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1360                 goto err;
1361         }
1362         err = -EAFNOSUPPORT;
1363         if (sock->ops->family != AF_INET) {
1364                 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1365                        "tunl %hu: fd %d wrong family, got %d, expected %d\n",
1366                        tunnel_id, fd, sock->ops->family, AF_INET);
1367                 goto err;
1368         }
1369
1370         err = -ENOTCONN;
1371
1372         /* Check if this socket has already been prepped */
1373         tunnel = (struct pppol2tp_tunnel *)sk->sk_user_data;
1374         if (tunnel != NULL) {
1375                 /* User-data field already set */
1376                 err = -EBUSY;
1377                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1378
1379                 /* This socket has already been prepped */
1380                 ret = tunnel->sock;
1381                 goto out;
1382         }
1383
1384         /* This socket is available and needs prepping. Create a new tunnel
1385          * context and init it.
1386          */
1387         sk->sk_user_data = tunnel = kzalloc(sizeof(struct pppol2tp_tunnel), GFP_KERNEL);
1388         if (sk->sk_user_data == NULL) {
1389                 err = -ENOMEM;
1390                 goto err;
1391         }
1392
1393         tunnel->magic = L2TP_TUNNEL_MAGIC;
1394         sprintf(&tunnel->name[0], "tunl %hu", tunnel_id);
1395
1396         tunnel->stats.tunnel_id = tunnel_id;
1397         tunnel->debug = PPPOL2TP_DEFAULT_DEBUG_FLAGS;
1398
1399         /* Hook on the tunnel socket destructor so that we can cleanup
1400          * if the tunnel socket goes away.
1401          */
1402         tunnel->old_sk_destruct = sk->sk_destruct;
1403         sk->sk_destruct = &pppol2tp_tunnel_destruct;
1404
1405         tunnel->sock = sk;
1406         sk->sk_allocation = GFP_ATOMIC;
1407
1408         /* Misc init */
1409         rwlock_init(&tunnel->hlist_lock);
1410
1411         /* Add tunnel to our list */
1412         INIT_LIST_HEAD(&tunnel->list);
1413         write_lock_bh(&pppol2tp_tunnel_list_lock);
1414         list_add(&tunnel->list, &pppol2tp_tunnel_list);
1415         write_unlock_bh(&pppol2tp_tunnel_list_lock);
1416         atomic_inc(&pppol2tp_tunnel_count);
1417
1418         /* Bump the reference count. The tunnel context is deleted
1419          * only when this drops to zero.
1420          */
1421         pppol2tp_tunnel_inc_refcount(tunnel);
1422
1423         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1424         (udp_sk(sk))->encap_type = UDP_ENCAP_L2TPINUDP;
1425         (udp_sk(sk))->encap_rcv = pppol2tp_udp_encap_recv;
1426
1427         ret = tunnel->sock;
1428
1429         *error = 0;
1430 out:
1431         if (sock)
1432                 sockfd_put(sock);
1433
1434         return ret;
1435
1436 err:
1437         *error = err;
1438         goto out;
1439 }
1440
1441 static struct proto pppol2tp_sk_proto = {
1442         .name     = "PPPOL2TP",
1443         .owner    = THIS_MODULE,
1444         .obj_size = sizeof(struct pppox_sock),
1445 };
1446
1447 /* socket() handler. Initialize a new struct sock.
1448  */
1449 static int pppol2tp_create(struct net *net, struct socket *sock)
1450 {
1451         int error = -ENOMEM;
1452         struct sock *sk;
1453
1454         sk = sk_alloc(net, PF_PPPOX, GFP_KERNEL, &pppol2tp_sk_proto);
1455         if (!sk)
1456                 goto out;
1457
1458         sock_init_data(sock, sk);
1459
1460         sock->state  = SS_UNCONNECTED;
1461         sock->ops    = &pppol2tp_ops;
1462
1463         sk->sk_backlog_rcv = pppol2tp_recv_core;
1464         sk->sk_protocol    = PX_PROTO_OL2TP;
1465         sk->sk_family      = PF_PPPOX;
1466         sk->sk_state       = PPPOX_NONE;
1467         sk->sk_type        = SOCK_STREAM;
1468         sk->sk_destruct    = pppol2tp_session_destruct;
1469
1470         error = 0;
1471
1472 out:
1473         return error;
1474 }
1475
1476 /* connect() handler. Attach a PPPoX socket to a tunnel UDP socket
1477  */
1478 static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
1479                             int sockaddr_len, int flags)
1480 {
1481         struct sock *sk = sock->sk;
1482         struct sockaddr_pppol2tp *sp = (struct sockaddr_pppol2tp *) uservaddr;
1483         struct pppox_sock *po = pppox_sk(sk);
1484         struct sock *tunnel_sock = NULL;
1485         struct pppol2tp_session *session = NULL;
1486         struct pppol2tp_tunnel *tunnel;
1487         struct dst_entry *dst;
1488         int error = 0;
1489
1490         lock_sock(sk);
1491
1492         error = -EINVAL;
1493         if (sp->sa_protocol != PX_PROTO_OL2TP)
1494                 goto end;
1495
1496         /* Check for already bound sockets */
1497         error = -EBUSY;
1498         if (sk->sk_state & PPPOX_CONNECTED)
1499                 goto end;
1500
1501         /* We don't supporting rebinding anyway */
1502         error = -EALREADY;
1503         if (sk->sk_user_data)
1504                 goto end; /* socket is already attached */
1505
1506         /* Don't bind if s_tunnel is 0 */
1507         error = -EINVAL;
1508         if (sp->pppol2tp.s_tunnel == 0)
1509                 goto end;
1510
1511         /* Special case: prepare tunnel socket if s_session and
1512          * d_session is 0. Otherwise look up tunnel using supplied
1513          * tunnel id.
1514          */
1515         if ((sp->pppol2tp.s_session == 0) && (sp->pppol2tp.d_session == 0)) {
1516                 tunnel_sock = pppol2tp_prepare_tunnel_socket(sp->pppol2tp.fd,
1517                                                              sp->pppol2tp.s_tunnel,
1518                                                              &error);
1519                 if (tunnel_sock == NULL)
1520                         goto end;
1521
1522                 tunnel = tunnel_sock->sk_user_data;
1523         } else {
1524                 tunnel = pppol2tp_tunnel_find(sp->pppol2tp.s_tunnel);
1525
1526                 /* Error if we can't find the tunnel */
1527                 error = -ENOENT;
1528                 if (tunnel == NULL)
1529                         goto end;
1530
1531                 tunnel_sock = tunnel->sock;
1532         }
1533
1534         /* Check that this session doesn't already exist */
1535         error = -EEXIST;
1536         session = pppol2tp_session_find(tunnel, sp->pppol2tp.s_session);
1537         if (session != NULL)
1538                 goto end;
1539
1540         /* Allocate and initialize a new session context. */
1541         session = kzalloc(sizeof(struct pppol2tp_session), GFP_KERNEL);
1542         if (session == NULL) {
1543                 error = -ENOMEM;
1544                 goto end;
1545         }
1546
1547         skb_queue_head_init(&session->reorder_q);
1548
1549         session->magic       = L2TP_SESSION_MAGIC;
1550         session->owner       = current->pid;
1551         session->sock        = sk;
1552         session->tunnel      = tunnel;
1553         session->tunnel_sock = tunnel_sock;
1554         session->tunnel_addr = sp->pppol2tp;
1555         sprintf(&session->name[0], "sess %hu/%hu",
1556                 session->tunnel_addr.s_tunnel,
1557                 session->tunnel_addr.s_session);
1558
1559         session->stats.tunnel_id  = session->tunnel_addr.s_tunnel;
1560         session->stats.session_id = session->tunnel_addr.s_session;
1561
1562         INIT_HLIST_NODE(&session->hlist);
1563
1564         /* Inherit debug options from tunnel */
1565         session->debug = tunnel->debug;
1566
1567         /* Default MTU must allow space for UDP/L2TP/PPP
1568          * headers.
1569          */
1570         session->mtu = session->mru = 1500 - PPPOL2TP_HEADER_OVERHEAD;
1571
1572         /* If PMTU discovery was enabled, use the MTU that was discovered */
1573         dst = sk_dst_get(sk);
1574         if (dst != NULL) {
1575                 u32 pmtu = dst_mtu(__sk_dst_get(sk));
1576                 if (pmtu != 0)
1577                         session->mtu = session->mru = pmtu -
1578                                 PPPOL2TP_HEADER_OVERHEAD;
1579                 dst_release(dst);
1580         }
1581
1582         /* Special case: if source & dest session_id == 0x0000, this socket is
1583          * being created to manage the tunnel. Don't add the session to the
1584          * session hash list, just set up the internal context for use by
1585          * ioctl() and sockopt() handlers.
1586          */
1587         if ((session->tunnel_addr.s_session == 0) &&
1588             (session->tunnel_addr.d_session == 0)) {
1589                 error = 0;
1590                 sk->sk_user_data = session;
1591                 goto out_no_ppp;
1592         }
1593
1594         /* Get tunnel context from the tunnel socket */
1595         tunnel = pppol2tp_sock_to_tunnel(tunnel_sock);
1596         if (tunnel == NULL) {
1597                 error = -EBADF;
1598                 goto end;
1599         }
1600
1601         /* Right now, because we don't have a way to push the incoming skb's
1602          * straight through the UDP layer, the only header we need to worry
1603          * about is the L2TP header. This size is different depending on
1604          * whether sequence numbers are enabled for the data channel.
1605          */
1606         po->chan.hdrlen = PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
1607
1608         po->chan.private = sk;
1609         po->chan.ops     = &pppol2tp_chan_ops;
1610         po->chan.mtu     = session->mtu;
1611
1612         error = ppp_register_channel(&po->chan);
1613         if (error)
1614                 goto end;
1615
1616         /* This is how we get the session context from the socket. */
1617         sk->sk_user_data = session;
1618
1619         /* Add session to the tunnel's hash list */
1620         write_lock_bh(&tunnel->hlist_lock);
1621         hlist_add_head(&session->hlist,
1622                        pppol2tp_session_id_hash(tunnel,
1623                                                 session->tunnel_addr.s_session));
1624         write_unlock_bh(&tunnel->hlist_lock);
1625
1626         atomic_inc(&pppol2tp_session_count);
1627
1628 out_no_ppp:
1629         pppol2tp_tunnel_inc_refcount(tunnel);
1630         sk->sk_state = PPPOX_CONNECTED;
1631         PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1632                "%s: created\n", session->name);
1633
1634 end:
1635         release_sock(sk);
1636
1637         if (error != 0) {
1638                 if (session)
1639                         PRINTK(session->debug,
1640                                 PPPOL2TP_MSG_CONTROL, KERN_WARNING,
1641                                 "%s: connect failed: %d\n",
1642                                 session->name, error);
1643                 else
1644                         PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_WARNING,
1645                                 "connect failed: %d\n", error);
1646         }
1647
1648         return error;
1649 }
1650
1651 /* getname() support.
1652  */
1653 static int pppol2tp_getname(struct socket *sock, struct sockaddr *uaddr,
1654                             int *usockaddr_len, int peer)
1655 {
1656         int len = sizeof(struct sockaddr_pppol2tp);
1657         struct sockaddr_pppol2tp sp;
1658         int error = 0;
1659         struct pppol2tp_session *session;
1660
1661         error = -ENOTCONN;
1662         if (sock->sk->sk_state != PPPOX_CONNECTED)
1663                 goto end;
1664
1665         session = pppol2tp_sock_to_session(sock->sk);
1666         if (session == NULL) {
1667                 error = -EBADF;
1668                 goto end;
1669         }
1670
1671         sp.sa_family    = AF_PPPOX;
1672         sp.sa_protocol  = PX_PROTO_OL2TP;
1673         memcpy(&sp.pppol2tp, &session->tunnel_addr,
1674                sizeof(struct pppol2tp_addr));
1675
1676         memcpy(uaddr, &sp, len);
1677
1678         *usockaddr_len = len;
1679
1680         error = 0;
1681
1682 end:
1683         return error;
1684 }
1685
1686 /****************************************************************************
1687  * ioctl() handlers.
1688  *
1689  * The PPPoX socket is created for L2TP sessions: tunnels have their own UDP
1690  * sockets. However, in order to control kernel tunnel features, we allow
1691  * userspace to create a special "tunnel" PPPoX socket which is used for
1692  * control only.  Tunnel PPPoX sockets have session_id == 0 and simply allow
1693  * the user application to issue L2TP setsockopt(), getsockopt() and ioctl()
1694  * calls.
1695  ****************************************************************************/
1696
1697 /* Session ioctl helper.
1698  */
1699 static int pppol2tp_session_ioctl(struct pppol2tp_session *session,
1700                                   unsigned int cmd, unsigned long arg)
1701 {
1702         struct ifreq ifr;
1703         int err = 0;
1704         struct sock *sk = session->sock;
1705         int val = (int) arg;
1706
1707         PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_DEBUG,
1708                "%s: pppol2tp_session_ioctl(cmd=%#x, arg=%#lx)\n",
1709                session->name, cmd, arg);
1710
1711         sock_hold(sk);
1712
1713         switch (cmd) {
1714         case SIOCGIFMTU:
1715                 err = -ENXIO;
1716                 if (!(sk->sk_state & PPPOX_CONNECTED))
1717                         break;
1718
1719                 err = -EFAULT;
1720                 if (copy_from_user(&ifr, (void __user *) arg, sizeof(struct ifreq)))
1721                         break;
1722                 ifr.ifr_mtu = session->mtu;
1723                 if (copy_to_user((void __user *) arg, &ifr, sizeof(struct ifreq)))
1724                         break;
1725
1726                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1727                        "%s: get mtu=%d\n", session->name, session->mtu);
1728                 err = 0;
1729                 break;
1730
1731         case SIOCSIFMTU:
1732                 err = -ENXIO;
1733                 if (!(sk->sk_state & PPPOX_CONNECTED))
1734                         break;
1735
1736                 err = -EFAULT;
1737                 if (copy_from_user(&ifr, (void __user *) arg, sizeof(struct ifreq)))
1738                         break;
1739
1740                 session->mtu = ifr.ifr_mtu;
1741
1742                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1743                        "%s: set mtu=%d\n", session->name, session->mtu);
1744                 err = 0;
1745                 break;
1746
1747         case PPPIOCGMRU:
1748                 err = -ENXIO;
1749                 if (!(sk->sk_state & PPPOX_CONNECTED))
1750                         break;
1751
1752                 err = -EFAULT;
1753                 if (put_user(session->mru, (int __user *) arg))
1754                         break;
1755
1756                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1757                        "%s: get mru=%d\n", session->name, session->mru);
1758                 err = 0;
1759                 break;
1760
1761         case PPPIOCSMRU:
1762                 err = -ENXIO;
1763                 if (!(sk->sk_state & PPPOX_CONNECTED))
1764                         break;
1765
1766                 err = -EFAULT;
1767                 if (get_user(val,(int __user *) arg))
1768                         break;
1769
1770                 session->mru = val;
1771                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1772                        "%s: set mru=%d\n", session->name, session->mru);
1773                 err = 0;
1774                 break;
1775
1776         case PPPIOCGFLAGS:
1777                 err = -EFAULT;
1778                 if (put_user(session->flags, (int __user *) arg))
1779                         break;
1780
1781                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1782                        "%s: get flags=%d\n", session->name, session->flags);
1783                 err = 0;
1784                 break;
1785
1786         case PPPIOCSFLAGS:
1787                 err = -EFAULT;
1788                 if (get_user(val, (int __user *) arg))
1789                         break;
1790                 session->flags = val;
1791                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1792                        "%s: set flags=%d\n", session->name, session->flags);
1793                 err = 0;
1794                 break;
1795
1796         case PPPIOCGL2TPSTATS:
1797                 err = -ENXIO;
1798                 if (!(sk->sk_state & PPPOX_CONNECTED))
1799                         break;
1800
1801                 if (copy_to_user((void __user *) arg, &session->stats,
1802                                  sizeof(session->stats)))
1803                         break;
1804                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1805                        "%s: get L2TP stats\n", session->name);
1806                 err = 0;
1807                 break;
1808
1809         default:
1810                 err = -ENOSYS;
1811                 break;
1812         }
1813
1814         sock_put(sk);
1815
1816         return err;
1817 }
1818
1819 /* Tunnel ioctl helper.
1820  *
1821  * Note the special handling for PPPIOCGL2TPSTATS below. If the ioctl data
1822  * specifies a session_id, the session ioctl handler is called. This allows an
1823  * application to retrieve session stats via a tunnel socket.
1824  */
1825 static int pppol2tp_tunnel_ioctl(struct pppol2tp_tunnel *tunnel,
1826                                  unsigned int cmd, unsigned long arg)
1827 {
1828         int err = 0;
1829         struct sock *sk = tunnel->sock;
1830         struct pppol2tp_ioc_stats stats_req;
1831
1832         PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_DEBUG,
1833                "%s: pppol2tp_tunnel_ioctl(cmd=%#x, arg=%#lx)\n", tunnel->name,
1834                cmd, arg);
1835
1836         sock_hold(sk);
1837
1838         switch (cmd) {
1839         case PPPIOCGL2TPSTATS:
1840                 err = -ENXIO;
1841                 if (!(sk->sk_state & PPPOX_CONNECTED))
1842                         break;
1843
1844                 if (copy_from_user(&stats_req, (void __user *) arg,
1845                                    sizeof(stats_req))) {
1846                         err = -EFAULT;
1847                         break;
1848                 }
1849                 if (stats_req.session_id != 0) {
1850                         /* resend to session ioctl handler */
1851                         struct pppol2tp_session *session =
1852                                 pppol2tp_session_find(tunnel, stats_req.session_id);
1853                         if (session != NULL)
1854                                 err = pppol2tp_session_ioctl(session, cmd, arg);
1855                         else
1856                                 err = -EBADR;
1857                         break;
1858                 }
1859 #ifdef CONFIG_XFRM
1860                 tunnel->stats.using_ipsec = (sk->sk_policy[0] || sk->sk_policy[1]) ? 1 : 0;
1861 #endif
1862                 if (copy_to_user((void __user *) arg, &tunnel->stats,
1863                                  sizeof(tunnel->stats))) {
1864                         err = -EFAULT;
1865                         break;
1866                 }
1867                 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1868                        "%s: get L2TP stats\n", tunnel->name);
1869                 err = 0;
1870                 break;
1871
1872         default:
1873                 err = -ENOSYS;
1874                 break;
1875         }
1876
1877         sock_put(sk);
1878
1879         return err;
1880 }
1881
1882 /* Main ioctl() handler.
1883  * Dispatch to tunnel or session helpers depending on the socket.
1884  */
1885 static int pppol2tp_ioctl(struct socket *sock, unsigned int cmd,
1886                           unsigned long arg)
1887 {
1888         struct sock *sk = sock->sk;
1889         struct pppol2tp_session *session;
1890         struct pppol2tp_tunnel *tunnel;
1891         int err;
1892
1893         if (!sk)
1894                 return 0;
1895
1896         err = -EBADF;
1897         if (sock_flag(sk, SOCK_DEAD) != 0)
1898                 goto end;
1899
1900         err = -ENOTCONN;
1901         if ((sk->sk_user_data == NULL) ||
1902             (!(sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND))))
1903                 goto end;
1904
1905         /* Get session context from the socket */
1906         err = -EBADF;
1907         session = pppol2tp_sock_to_session(sk);
1908         if (session == NULL)
1909                 goto end;
1910
1911         /* Special case: if session's session_id is zero, treat ioctl as a
1912          * tunnel ioctl
1913          */
1914         if ((session->tunnel_addr.s_session == 0) &&
1915             (session->tunnel_addr.d_session == 0)) {
1916                 err = -EBADF;
1917                 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
1918                 if (tunnel == NULL)
1919                         goto end;
1920
1921                 err = pppol2tp_tunnel_ioctl(tunnel, cmd, arg);
1922                 goto end;
1923         }
1924
1925         err = pppol2tp_session_ioctl(session, cmd, arg);
1926
1927 end:
1928         return err;
1929 }
1930
1931 /*****************************************************************************
1932  * setsockopt() / getsockopt() support.
1933  *
1934  * The PPPoX socket is created for L2TP sessions: tunnels have their own UDP
1935  * sockets. In order to control kernel tunnel features, we allow userspace to
1936  * create a special "tunnel" PPPoX socket which is used for control only.
1937  * Tunnel PPPoX sockets have session_id == 0 and simply allow the user
1938  * application to issue L2TP setsockopt(), getsockopt() and ioctl() calls.
1939  *****************************************************************************/
1940
1941 /* Tunnel setsockopt() helper.
1942  */
1943 static int pppol2tp_tunnel_setsockopt(struct sock *sk,
1944                                       struct pppol2tp_tunnel *tunnel,
1945                                       int optname, int val)
1946 {
1947         int err = 0;
1948
1949         switch (optname) {
1950         case PPPOL2TP_SO_DEBUG:
1951                 tunnel->debug = val;
1952                 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1953                        "%s: set debug=%x\n", tunnel->name, tunnel->debug);
1954                 break;
1955
1956         default:
1957                 err = -ENOPROTOOPT;
1958                 break;
1959         }
1960
1961         return err;
1962 }
1963
1964 /* Session setsockopt helper.
1965  */
1966 static int pppol2tp_session_setsockopt(struct sock *sk,
1967                                        struct pppol2tp_session *session,
1968                                        int optname, int val)
1969 {
1970         int err = 0;
1971
1972         switch (optname) {
1973         case PPPOL2TP_SO_RECVSEQ:
1974                 if ((val != 0) && (val != 1)) {
1975                         err = -EINVAL;
1976                         break;
1977                 }
1978                 session->recv_seq = val ? -1 : 0;
1979                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1980                        "%s: set recv_seq=%d\n", session->name,
1981                        session->recv_seq);
1982                 break;
1983
1984         case PPPOL2TP_SO_SENDSEQ:
1985                 if ((val != 0) && (val != 1)) {
1986                         err = -EINVAL;
1987                         break;
1988                 }
1989                 session->send_seq = val ? -1 : 0;
1990                 {
1991                         struct sock *ssk      = session->sock;
1992                         struct pppox_sock *po = pppox_sk(ssk);
1993                         po->chan.hdrlen = val ? PPPOL2TP_L2TP_HDR_SIZE_SEQ :
1994                                 PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
1995                 }
1996                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1997                        "%s: set send_seq=%d\n", session->name, session->send_seq);
1998                 break;
1999
2000         case PPPOL2TP_SO_LNSMODE:
2001                 if ((val != 0) && (val != 1)) {
2002                         err = -EINVAL;
2003                         break;
2004                 }
2005                 session->lns_mode = val ? -1 : 0;
2006                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2007                        "%s: set lns_mode=%d\n", session->name,
2008                        session->lns_mode);
2009                 break;
2010
2011         case PPPOL2TP_SO_DEBUG:
2012                 session->debug = val;
2013                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2014                        "%s: set debug=%x\n", session->name, session->debug);
2015                 break;
2016
2017         case PPPOL2TP_SO_REORDERTO:
2018                 session->reorder_timeout = msecs_to_jiffies(val);
2019                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2020                        "%s: set reorder_timeout=%d\n", session->name,
2021                        session->reorder_timeout);
2022                 break;
2023
2024         default:
2025                 err = -ENOPROTOOPT;
2026                 break;
2027         }
2028
2029         return err;
2030 }
2031
2032 /* Main setsockopt() entry point.
2033  * Does API checks, then calls either the tunnel or session setsockopt
2034  * handler, according to whether the PPPoL2TP socket is a for a regular
2035  * session or the special tunnel type.
2036  */
2037 static int pppol2tp_setsockopt(struct socket *sock, int level, int optname,
2038                                char __user *optval, int optlen)
2039 {
2040         struct sock *sk = sock->sk;
2041         struct pppol2tp_session *session = sk->sk_user_data;
2042         struct pppol2tp_tunnel *tunnel;
2043         int val;
2044         int err;
2045
2046         if (level != SOL_PPPOL2TP)
2047                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
2048
2049         if (optlen < sizeof(int))
2050                 return -EINVAL;
2051
2052         if (get_user(val, (int __user *)optval))
2053                 return -EFAULT;
2054
2055         err = -ENOTCONN;
2056         if (sk->sk_user_data == NULL)
2057                 goto end;
2058
2059         /* Get session context from the socket */
2060         err = -EBADF;
2061         session = pppol2tp_sock_to_session(sk);
2062         if (session == NULL)
2063                 goto end;
2064
2065         /* Special case: if session_id == 0x0000, treat as operation on tunnel
2066          */
2067         if ((session->tunnel_addr.s_session == 0) &&
2068             (session->tunnel_addr.d_session == 0)) {
2069                 err = -EBADF;
2070                 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
2071                 if (tunnel == NULL)
2072                         goto end;
2073
2074                 err = pppol2tp_tunnel_setsockopt(sk, tunnel, optname, val);
2075         } else
2076                 err = pppol2tp_session_setsockopt(sk, session, optname, val);
2077
2078         err = 0;
2079
2080 end:
2081         return err;
2082 }
2083
2084 /* Tunnel getsockopt helper. Called with sock locked.
2085  */
2086 static int pppol2tp_tunnel_getsockopt(struct sock *sk,
2087                                       struct pppol2tp_tunnel *tunnel,
2088                                       int optname, int *val)
2089 {
2090         int err = 0;
2091
2092         switch (optname) {
2093         case PPPOL2TP_SO_DEBUG:
2094                 *val = tunnel->debug;
2095                 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2096                        "%s: get debug=%x\n", tunnel->name, tunnel->debug);
2097                 break;
2098
2099         default:
2100                 err = -ENOPROTOOPT;
2101                 break;
2102         }
2103
2104         return err;
2105 }
2106
2107 /* Session getsockopt helper. Called with sock locked.
2108  */
2109 static int pppol2tp_session_getsockopt(struct sock *sk,
2110                                        struct pppol2tp_session *session,
2111                                        int optname, int *val)
2112 {
2113         int err = 0;
2114
2115         switch (optname) {
2116         case PPPOL2TP_SO_RECVSEQ:
2117                 *val = session->recv_seq;
2118                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2119                        "%s: get recv_seq=%d\n", session->name, *val);
2120                 break;
2121
2122         case PPPOL2TP_SO_SENDSEQ:
2123                 *val = session->send_seq;
2124                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2125                        "%s: get send_seq=%d\n", session->name, *val);
2126                 break;
2127
2128         case PPPOL2TP_SO_LNSMODE:
2129                 *val = session->lns_mode;
2130                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2131                        "%s: get lns_mode=%d\n", session->name, *val);
2132                 break;
2133
2134         case PPPOL2TP_SO_DEBUG:
2135                 *val = session->debug;
2136                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2137                        "%s: get debug=%d\n", session->name, *val);
2138                 break;
2139
2140         case PPPOL2TP_SO_REORDERTO:
2141                 *val = (int) jiffies_to_msecs(session->reorder_timeout);
2142                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2143                        "%s: get reorder_timeout=%d\n", session->name, *val);
2144                 break;
2145
2146         default:
2147                 err = -ENOPROTOOPT;
2148         }
2149
2150         return err;
2151 }
2152
2153 /* Main getsockopt() entry point.
2154  * Does API checks, then calls either the tunnel or session getsockopt
2155  * handler, according to whether the PPPoX socket is a for a regular session
2156  * or the special tunnel type.
2157  */
2158 static int pppol2tp_getsockopt(struct socket *sock, int level,
2159                                int optname, char __user *optval, int __user *optlen)
2160 {
2161         struct sock *sk = sock->sk;
2162         struct pppol2tp_session *session = sk->sk_user_data;
2163         struct pppol2tp_tunnel *tunnel;
2164         int val, len;
2165         int err;
2166
2167         if (level != SOL_PPPOL2TP)
2168                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
2169
2170         if (get_user(len, (int __user *) optlen))
2171                 return -EFAULT;
2172
2173         len = min_t(unsigned int, len, sizeof(int));
2174
2175         if (len < 0)
2176                 return -EINVAL;
2177
2178         err = -ENOTCONN;
2179         if (sk->sk_user_data == NULL)
2180                 goto end;
2181
2182         /* Get the session context */
2183         err = -EBADF;
2184         session = pppol2tp_sock_to_session(sk);
2185         if (session == NULL)
2186                 goto end;
2187
2188         /* Special case: if session_id == 0x0000, treat as operation on tunnel */
2189         if ((session->tunnel_addr.s_session == 0) &&
2190             (session->tunnel_addr.d_session == 0)) {
2191                 err = -EBADF;
2192                 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
2193                 if (tunnel == NULL)
2194                         goto end;
2195
2196                 err = pppol2tp_tunnel_getsockopt(sk, tunnel, optname, &val);
2197         } else
2198                 err = pppol2tp_session_getsockopt(sk, session, optname, &val);
2199
2200         err = -EFAULT;
2201         if (put_user(len, (int __user *) optlen))
2202                 goto end;
2203
2204         if (copy_to_user((void __user *) optval, &val, len))
2205                 goto end;
2206
2207         err = 0;
2208 end:
2209         return err;
2210 }
2211
2212 /*****************************************************************************
2213  * /proc filesystem for debug
2214  *****************************************************************************/
2215
2216 #ifdef CONFIG_PROC_FS
2217
2218 #include <linux/seq_file.h>
2219
2220 struct pppol2tp_seq_data {
2221         struct pppol2tp_tunnel *tunnel; /* current tunnel */
2222         struct pppol2tp_session *session; /* NULL means get first session in tunnel */
2223 };
2224
2225 static struct pppol2tp_session *next_session(struct pppol2tp_tunnel *tunnel, struct pppol2tp_session *curr)
2226 {
2227         struct pppol2tp_session *session = NULL;
2228         struct hlist_node *walk;
2229         int found = 0;
2230         int next = 0;
2231         int i;
2232
2233         read_lock_bh(&tunnel->hlist_lock);
2234         for (i = 0; i < PPPOL2TP_HASH_SIZE; i++) {
2235                 hlist_for_each_entry(session, walk, &tunnel->session_hlist[i], hlist) {
2236                         if (curr == NULL) {
2237                                 found = 1;
2238                                 goto out;
2239                         }
2240                         if (session == curr) {
2241                                 next = 1;
2242                                 continue;
2243                         }
2244                         if (next) {
2245                                 found = 1;
2246                                 goto out;
2247                         }
2248                 }
2249         }
2250 out:
2251         read_unlock_bh(&tunnel->hlist_lock);
2252         if (!found)
2253                 session = NULL;
2254
2255         return session;
2256 }
2257
2258 static struct pppol2tp_tunnel *next_tunnel(struct pppol2tp_tunnel *curr)
2259 {
2260         struct pppol2tp_tunnel *tunnel = NULL;
2261
2262         read_lock_bh(&pppol2tp_tunnel_list_lock);
2263         if (list_is_last(&curr->list, &pppol2tp_tunnel_list)) {
2264                 goto out;
2265         }
2266         tunnel = list_entry(curr->list.next, struct pppol2tp_tunnel, list);
2267 out:
2268         read_unlock_bh(&pppol2tp_tunnel_list_lock);
2269
2270         return tunnel;
2271 }
2272
2273 static void *pppol2tp_seq_start(struct seq_file *m, loff_t *offs)
2274 {
2275         struct pppol2tp_seq_data *pd = SEQ_START_TOKEN;
2276         loff_t pos = *offs;
2277
2278         if (!pos)
2279                 goto out;
2280
2281         BUG_ON(m->private == NULL);
2282         pd = m->private;
2283
2284         if (pd->tunnel == NULL) {
2285                 if (!list_empty(&pppol2tp_tunnel_list))
2286                         pd->tunnel = list_entry(pppol2tp_tunnel_list.next, struct pppol2tp_tunnel, list);
2287         } else {
2288                 pd->session = next_session(pd->tunnel, pd->session);
2289                 if (pd->session == NULL) {
2290                         pd->tunnel = next_tunnel(pd->tunnel);
2291                 }
2292         }
2293
2294         /* NULL tunnel and session indicates end of list */
2295         if ((pd->tunnel == NULL) && (pd->session == NULL))
2296                 pd = NULL;
2297
2298 out:
2299         return pd;
2300 }
2301
2302 static void *pppol2tp_seq_next(struct seq_file *m, void *v, loff_t *pos)
2303 {
2304         (*pos)++;
2305         return NULL;
2306 }
2307
2308 static void pppol2tp_seq_stop(struct seq_file *p, void *v)
2309 {
2310         /* nothing to do */
2311 }
2312
2313 static void pppol2tp_seq_tunnel_show(struct seq_file *m, void *v)
2314 {
2315         struct pppol2tp_tunnel *tunnel = v;
2316
2317         seq_printf(m, "\nTUNNEL '%s', %c %d\n",
2318                    tunnel->name,
2319                    (tunnel == tunnel->sock->sk_user_data) ? 'Y':'N',
2320                    atomic_read(&tunnel->ref_count) - 1);
2321         seq_printf(m, " %08x %llu/%llu/%llu %llu/%llu/%llu\n",
2322                    tunnel->debug,
2323                    (unsigned long long)tunnel->stats.tx_packets,
2324                    (unsigned long long)tunnel->stats.tx_bytes,
2325                    (unsigned long long)tunnel->stats.tx_errors,
2326                    (unsigned long long)tunnel->stats.rx_packets,
2327                    (unsigned long long)tunnel->stats.rx_bytes,
2328                    (unsigned long long)tunnel->stats.rx_errors);
2329 }
2330
2331 static void pppol2tp_seq_session_show(struct seq_file *m, void *v)
2332 {
2333         struct pppol2tp_session *session = v;
2334
2335         seq_printf(m, "  SESSION '%s' %08X/%d %04X/%04X -> "
2336                    "%04X/%04X %d %c\n",
2337                    session->name,
2338                    ntohl(session->tunnel_addr.addr.sin_addr.s_addr),
2339                    ntohs(session->tunnel_addr.addr.sin_port),
2340                    session->tunnel_addr.s_tunnel,
2341                    session->tunnel_addr.s_session,
2342                    session->tunnel_addr.d_tunnel,
2343                    session->tunnel_addr.d_session,
2344                    session->sock->sk_state,
2345                    (session == session->sock->sk_user_data) ?
2346                    'Y' : 'N');
2347         seq_printf(m, "   %d/%d/%c/%c/%s %08x %u\n",
2348                    session->mtu, session->mru,
2349                    session->recv_seq ? 'R' : '-',
2350                    session->send_seq ? 'S' : '-',
2351                    session->lns_mode ? "LNS" : "LAC",
2352                    session->debug,
2353                    jiffies_to_msecs(session->reorder_timeout));
2354         seq_printf(m, "   %hu/%hu %llu/%llu/%llu %llu/%llu/%llu\n",
2355                    session->nr, session->ns,
2356                    (unsigned long long)session->stats.tx_packets,
2357                    (unsigned long long)session->stats.tx_bytes,
2358                    (unsigned long long)session->stats.tx_errors,
2359                    (unsigned long long)session->stats.rx_packets,
2360                    (unsigned long long)session->stats.rx_bytes,
2361                    (unsigned long long)session->stats.rx_errors);
2362 }
2363
2364 static int pppol2tp_seq_show(struct seq_file *m, void *v)
2365 {
2366         struct pppol2tp_seq_data *pd = v;
2367
2368         /* display header on line 1 */
2369         if (v == SEQ_START_TOKEN) {
2370                 seq_puts(m, "PPPoL2TP driver info, " PPPOL2TP_DRV_VERSION "\n");
2371                 seq_puts(m, "TUNNEL name, user-data-ok session-count\n");
2372                 seq_puts(m, " debug tx-pkts/bytes/errs rx-pkts/bytes/errs\n");
2373                 seq_puts(m, "  SESSION name, addr/port src-tid/sid "
2374                          "dest-tid/sid state user-data-ok\n");
2375                 seq_puts(m, "   mtu/mru/rcvseq/sendseq/lns debug reorderto\n");
2376                 seq_puts(m, "   nr/ns tx-pkts/bytes/errs rx-pkts/bytes/errs\n");
2377                 goto out;
2378         }
2379
2380         /* Show the tunnel or session context.
2381          */
2382         if (pd->session == NULL)
2383                 pppol2tp_seq_tunnel_show(m, pd->tunnel);
2384         else
2385                 pppol2tp_seq_session_show(m, pd->session);
2386
2387 out:
2388         return 0;
2389 }
2390
2391 static struct seq_operations pppol2tp_seq_ops = {
2392         .start          = pppol2tp_seq_start,
2393         .next           = pppol2tp_seq_next,
2394         .stop           = pppol2tp_seq_stop,
2395         .show           = pppol2tp_seq_show,
2396 };
2397
2398 /* Called when our /proc file is opened. We allocate data for use when
2399  * iterating our tunnel / session contexts and store it in the private
2400  * data of the seq_file.
2401  */
2402 static int pppol2tp_proc_open(struct inode *inode, struct file *file)
2403 {
2404         struct seq_file *m;
2405         struct pppol2tp_seq_data *pd;
2406         int ret = 0;
2407
2408         ret = seq_open(file, &pppol2tp_seq_ops);
2409         if (ret < 0)
2410                 goto out;
2411
2412         m = file->private_data;
2413
2414         /* Allocate and fill our proc_data for access later */
2415         ret = -ENOMEM;
2416         m->private = kzalloc(sizeof(struct pppol2tp_seq_data), GFP_KERNEL);
2417         if (m->private == NULL)
2418                 goto out;
2419
2420         pd = m->private;
2421         ret = 0;
2422
2423 out:
2424         return ret;
2425 }
2426
2427 /* Called when /proc file access completes.
2428  */
2429 static int pppol2tp_proc_release(struct inode *inode, struct file *file)
2430 {
2431         struct seq_file *m = (struct seq_file *)file->private_data;
2432
2433         kfree(m->private);
2434         m->private = NULL;
2435
2436         return seq_release(inode, file);
2437 }
2438
2439 static struct file_operations pppol2tp_proc_fops = {
2440         .owner          = THIS_MODULE,
2441         .open           = pppol2tp_proc_open,
2442         .read           = seq_read,
2443         .llseek         = seq_lseek,
2444         .release        = pppol2tp_proc_release,
2445 };
2446
2447 static struct proc_dir_entry *pppol2tp_proc;
2448
2449 #endif /* CONFIG_PROC_FS */
2450
2451 /*****************************************************************************
2452  * Init and cleanup
2453  *****************************************************************************/
2454
2455 static struct proto_ops pppol2tp_ops = {
2456         .family         = AF_PPPOX,
2457         .owner          = THIS_MODULE,
2458         .release        = pppol2tp_release,
2459         .bind           = sock_no_bind,
2460         .connect        = pppol2tp_connect,
2461         .socketpair     = sock_no_socketpair,
2462         .accept         = sock_no_accept,
2463         .getname        = pppol2tp_getname,
2464         .poll           = datagram_poll,
2465         .listen         = sock_no_listen,
2466         .shutdown       = sock_no_shutdown,
2467         .setsockopt     = pppol2tp_setsockopt,
2468         .getsockopt     = pppol2tp_getsockopt,
2469         .sendmsg        = pppol2tp_sendmsg,
2470         .recvmsg        = pppol2tp_recvmsg,
2471         .mmap           = sock_no_mmap,
2472         .ioctl          = pppox_ioctl,
2473 };
2474
2475 static struct pppox_proto pppol2tp_proto = {
2476         .create         = pppol2tp_create,
2477         .ioctl          = pppol2tp_ioctl
2478 };
2479
2480 static int __init pppol2tp_init(void)
2481 {
2482         int err;
2483
2484         err = proto_register(&pppol2tp_sk_proto, 0);
2485         if (err)
2486                 goto out;
2487         err = register_pppox_proto(PX_PROTO_OL2TP, &pppol2tp_proto);
2488         if (err)
2489                 goto out_unregister_pppol2tp_proto;
2490
2491 #ifdef CONFIG_PROC_FS
2492         pppol2tp_proc = proc_net_fops_create(&init_net, "pppol2tp", 0,
2493                                              &pppol2tp_proc_fops);
2494         if (!pppol2tp_proc) {
2495                 err = -ENOMEM;
2496                 goto out_unregister_pppox_proto;
2497         }
2498 #endif /* CONFIG_PROC_FS */
2499         printk(KERN_INFO "PPPoL2TP kernel driver, %s\n",
2500                PPPOL2TP_DRV_VERSION);
2501
2502 out:
2503         return err;
2504 #ifdef CONFIG_PROC_FS
2505 out_unregister_pppox_proto:
2506         unregister_pppox_proto(PX_PROTO_OL2TP);
2507 #endif
2508 out_unregister_pppol2tp_proto:
2509         proto_unregister(&pppol2tp_sk_proto);
2510         goto out;
2511 }
2512
2513 static void __exit pppol2tp_exit(void)
2514 {
2515         unregister_pppox_proto(PX_PROTO_OL2TP);
2516
2517 #ifdef CONFIG_PROC_FS
2518         remove_proc_entry("pppol2tp", init_net.proc_net);
2519 #endif
2520         proto_unregister(&pppol2tp_sk_proto);
2521 }
2522
2523 module_init(pppol2tp_init);
2524 module_exit(pppol2tp_exit);
2525
2526 MODULE_AUTHOR("Martijn van Oosterhout <kleptog@svana.org>, "
2527               "James Chapman <jchapman@katalix.com>");
2528 MODULE_DESCRIPTION("PPP over L2TP over UDP");
2529 MODULE_LICENSE("GPL");
2530 MODULE_VERSION(PPPOL2TP_DRV_VERSION);