1 // SPDX-License-Identifier: GPL-2.0-only 2 /* L2TP core. 3 * 4 * Copyright (c) 2008,2009,2010 Katalix Systems Ltd 5 * 6 * This file contains some code of the original L2TPv2 pppol2tp 7 * driver, which has the following copyright: 8 * 9 * Authors: Martijn van Oosterhout <kleptog@svana.org> 10 * James Chapman (jchapman@katalix.com) 11 * Contributors: 12 * Michal Ostrowski <mostrows@speakeasy.net> 13 * Arnaldo Carvalho de Melo <acme@xconectiva.com.br> 14 * David S. Miller (davem@redhat.com) 15 */ 16 17 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 18 19 #include <linux/module.h> 20 #include <linux/string.h> 21 #include <linux/list.h> 22 #include <linux/rculist.h> 23 #include <linux/uaccess.h> 24 25 #include <linux/kernel.h> 26 #include <linux/spinlock.h> 27 #include <linux/kthread.h> 28 #include <linux/sched.h> 29 #include <linux/slab.h> 30 #include <linux/errno.h> 31 #include <linux/jiffies.h> 32 33 #include <linux/netdevice.h> 34 #include <linux/net.h> 35 #include <linux/inetdevice.h> 36 #include <linux/skbuff.h> 37 #include <linux/init.h> 38 #include <linux/in.h> 39 #include <linux/ip.h> 40 #include <linux/udp.h> 41 #include <linux/l2tp.h> 42 #include <linux/sort.h> 43 #include <linux/file.h> 44 #include <linux/nsproxy.h> 45 #include <net/net_namespace.h> 46 #include <net/netns/generic.h> 47 #include <net/dst.h> 48 #include <net/ip.h> 49 #include <net/udp.h> 50 #include <net/udp_tunnel.h> 51 #include <net/inet_common.h> 52 #include <net/xfrm.h> 53 #include <net/protocol.h> 54 #include <net/inet6_connection_sock.h> 55 #include <net/inet_ecn.h> 56 #include <net/ip6_route.h> 57 #include <net/ip6_checksum.h> 58 59 #include <asm/byteorder.h> 60 #include <linux/atomic.h> 61 62 #include "l2tp_core.h" 63 64 #define CREATE_TRACE_POINTS 65 #include "trace.h" 66 67 #define L2TP_DRV_VERSION "V2.0" 68 69 /* L2TP header constants */ 70 #define L2TP_HDRFLAG_T 0x8000 71 #define L2TP_HDRFLAG_L 0x4000 72 #define L2TP_HDRFLAG_S 0x0800 73 #define L2TP_HDRFLAG_O 0x0200 74 #define L2TP_HDRFLAG_P 0x0100 75 76 #define L2TP_HDR_VER_MASK 0x000F 77 #define L2TP_HDR_VER_2 0x0002 78 #define L2TP_HDR_VER_3 0x0003 79 80 /* L2TPv3 default L2-specific sublayer */ 81 #define L2TP_SLFLAG_S 0x40000000 82 #define L2TP_SL_SEQ_MASK 0x00ffffff 83 84 #define L2TP_HDR_SIZE_MAX 14 85 86 /* Default trace flags */ 87 #define L2TP_DEFAULT_DEBUG_FLAGS 0 88 89 #define L2TP_DEPTH_NESTING 2 90 #if L2TP_DEPTH_NESTING == SINGLE_DEPTH_NESTING 91 #error "L2TP requires its own lockdep subclass" 92 #endif 93 94 /* Private data stored for received packets in the skb. 95 */ 96 struct l2tp_skb_cb { 97 u32 ns; 98 u16 has_seq; 99 u16 length; 100 unsigned long expires; 101 }; 102 103 #define L2TP_SKB_CB(skb) ((struct l2tp_skb_cb *)&(skb)->cb[sizeof(struct inet_skb_parm)]) 104 105 static struct workqueue_struct *l2tp_wq; 106 107 /* per-net private data for this module */ 108 static unsigned int l2tp_net_id; 109 struct l2tp_net { 110 /* Lock for write access to l2tp_tunnel_idr */ 111 spinlock_t l2tp_tunnel_idr_lock; 112 struct idr l2tp_tunnel_idr; 113 /* Lock for write access to l2tp_v[23]_session_idr/htable */ 114 spinlock_t l2tp_session_idr_lock; 115 struct idr l2tp_v2_session_idr; 116 struct idr l2tp_v3_session_idr; 117 struct hlist_head l2tp_v3_session_htable[16]; 118 }; 119 120 static inline u32 l2tp_v2_session_key(u16 tunnel_id, u16 session_id) 121 { 122 return ((u32)tunnel_id) << 16 | session_id; 123 } 124 125 static inline unsigned long l2tp_v3_session_hashkey(struct sock *sk, u32 session_id) 126 { 127 return ((unsigned long)sk) + session_id; 128 } 129 130 #if IS_ENABLED(CONFIG_IPV6) 131 static bool l2tp_sk_is_v6(struct sock *sk) 132 { 133 return sk->sk_family == PF_INET6 && 134 !ipv6_addr_v4mapped(&sk->sk_v6_daddr); 135 } 136 #endif 137 138 static inline struct l2tp_net *l2tp_pernet(const struct net *net) 139 { 140 return net_generic(net, l2tp_net_id); 141 } 142 143 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel) 144 { 145 trace_free_tunnel(tunnel); 146 sock_put(tunnel->sock); 147 /* the tunnel is freed in the socket destructor */ 148 } 149 150 static void l2tp_session_free(struct l2tp_session *session) 151 { 152 trace_free_session(session); 153 if (session->tunnel) 154 l2tp_tunnel_dec_refcount(session->tunnel); 155 kfree(session); 156 } 157 158 struct l2tp_tunnel *l2tp_sk_to_tunnel(struct sock *sk) 159 { 160 struct l2tp_tunnel *tunnel = sk->sk_user_data; 161 162 if (tunnel) 163 if (WARN_ON(tunnel->magic != L2TP_TUNNEL_MAGIC)) 164 return NULL; 165 166 return tunnel; 167 } 168 EXPORT_SYMBOL_GPL(l2tp_sk_to_tunnel); 169 170 void l2tp_tunnel_inc_refcount(struct l2tp_tunnel *tunnel) 171 { 172 refcount_inc(&tunnel->ref_count); 173 } 174 EXPORT_SYMBOL_GPL(l2tp_tunnel_inc_refcount); 175 176 void l2tp_tunnel_dec_refcount(struct l2tp_tunnel *tunnel) 177 { 178 if (refcount_dec_and_test(&tunnel->ref_count)) 179 l2tp_tunnel_free(tunnel); 180 } 181 EXPORT_SYMBOL_GPL(l2tp_tunnel_dec_refcount); 182 183 void l2tp_session_inc_refcount(struct l2tp_session *session) 184 { 185 refcount_inc(&session->ref_count); 186 } 187 EXPORT_SYMBOL_GPL(l2tp_session_inc_refcount); 188 189 void l2tp_session_dec_refcount(struct l2tp_session *session) 190 { 191 if (refcount_dec_and_test(&session->ref_count)) 192 l2tp_session_free(session); 193 } 194 EXPORT_SYMBOL_GPL(l2tp_session_dec_refcount); 195 196 /* Lookup a tunnel. A new reference is held on the returned tunnel. */ 197 struct l2tp_tunnel *l2tp_tunnel_get(const struct net *net, u32 tunnel_id) 198 { 199 const struct l2tp_net *pn = l2tp_pernet(net); 200 struct l2tp_tunnel *tunnel; 201 202 rcu_read_lock_bh(); 203 tunnel = idr_find(&pn->l2tp_tunnel_idr, tunnel_id); 204 if (tunnel && refcount_inc_not_zero(&tunnel->ref_count)) { 205 rcu_read_unlock_bh(); 206 return tunnel; 207 } 208 rcu_read_unlock_bh(); 209 210 return NULL; 211 } 212 EXPORT_SYMBOL_GPL(l2tp_tunnel_get); 213 214 struct l2tp_tunnel *l2tp_tunnel_get_nth(const struct net *net, int nth) 215 { 216 struct l2tp_net *pn = l2tp_pernet(net); 217 unsigned long tunnel_id, tmp; 218 struct l2tp_tunnel *tunnel; 219 int count = 0; 220 221 rcu_read_lock_bh(); 222 idr_for_each_entry_ul(&pn->l2tp_tunnel_idr, tunnel, tmp, tunnel_id) { 223 if (tunnel && ++count > nth && 224 refcount_inc_not_zero(&tunnel->ref_count)) { 225 rcu_read_unlock_bh(); 226 return tunnel; 227 } 228 } 229 rcu_read_unlock_bh(); 230 231 return NULL; 232 } 233 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_nth); 234 235 struct l2tp_session *l2tp_v3_session_get(const struct net *net, struct sock *sk, u32 session_id) 236 { 237 const struct l2tp_net *pn = l2tp_pernet(net); 238 struct l2tp_session *session; 239 240 rcu_read_lock_bh(); 241 session = idr_find(&pn->l2tp_v3_session_idr, session_id); 242 if (session && !hash_hashed(&session->hlist) && 243 refcount_inc_not_zero(&session->ref_count)) { 244 rcu_read_unlock_bh(); 245 return session; 246 } 247 248 /* If we get here and session is non-NULL, the session_id 249 * collides with one in another tunnel. If sk is non-NULL, 250 * find the session matching sk. 251 */ 252 if (session && sk) { 253 unsigned long key = l2tp_v3_session_hashkey(sk, session->session_id); 254 255 hash_for_each_possible_rcu(pn->l2tp_v3_session_htable, session, 256 hlist, key) { 257 if (session->tunnel->sock == sk && 258 refcount_inc_not_zero(&session->ref_count)) { 259 rcu_read_unlock_bh(); 260 return session; 261 } 262 } 263 } 264 rcu_read_unlock_bh(); 265 266 return NULL; 267 } 268 EXPORT_SYMBOL_GPL(l2tp_v3_session_get); 269 270 struct l2tp_session *l2tp_v2_session_get(const struct net *net, u16 tunnel_id, u16 session_id) 271 { 272 u32 session_key = l2tp_v2_session_key(tunnel_id, session_id); 273 const struct l2tp_net *pn = l2tp_pernet(net); 274 struct l2tp_session *session; 275 276 rcu_read_lock_bh(); 277 session = idr_find(&pn->l2tp_v2_session_idr, session_key); 278 if (session && refcount_inc_not_zero(&session->ref_count)) { 279 rcu_read_unlock_bh(); 280 return session; 281 } 282 rcu_read_unlock_bh(); 283 284 return NULL; 285 } 286 EXPORT_SYMBOL_GPL(l2tp_v2_session_get); 287 288 struct l2tp_session *l2tp_session_get(const struct net *net, struct sock *sk, int pver, 289 u32 tunnel_id, u32 session_id) 290 { 291 if (pver == L2TP_HDR_VER_2) 292 return l2tp_v2_session_get(net, tunnel_id, session_id); 293 else 294 return l2tp_v3_session_get(net, sk, session_id); 295 } 296 EXPORT_SYMBOL_GPL(l2tp_session_get); 297 298 struct l2tp_session *l2tp_session_get_nth(struct l2tp_tunnel *tunnel, int nth) 299 { 300 struct l2tp_session *session; 301 int count = 0; 302 303 rcu_read_lock_bh(); 304 list_for_each_entry_rcu(session, &tunnel->session_list, list) { 305 if (++count > nth) { 306 l2tp_session_inc_refcount(session); 307 rcu_read_unlock_bh(); 308 return session; 309 } 310 } 311 rcu_read_unlock_bh(); 312 313 return NULL; 314 } 315 EXPORT_SYMBOL_GPL(l2tp_session_get_nth); 316 317 /* Lookup a session by interface name. 318 * This is very inefficient but is only used by management interfaces. 319 */ 320 struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net, 321 const char *ifname) 322 { 323 struct l2tp_net *pn = l2tp_pernet(net); 324 unsigned long tunnel_id, tmp; 325 struct l2tp_session *session; 326 struct l2tp_tunnel *tunnel; 327 328 rcu_read_lock_bh(); 329 idr_for_each_entry_ul(&pn->l2tp_tunnel_idr, tunnel, tmp, tunnel_id) { 330 if (tunnel) { 331 list_for_each_entry_rcu(session, &tunnel->session_list, list) { 332 if (!strcmp(session->ifname, ifname)) { 333 l2tp_session_inc_refcount(session); 334 rcu_read_unlock_bh(); 335 336 return session; 337 } 338 } 339 } 340 } 341 rcu_read_unlock_bh(); 342 343 return NULL; 344 } 345 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname); 346 347 static void l2tp_session_coll_list_add(struct l2tp_session_coll_list *clist, 348 struct l2tp_session *session) 349 { 350 l2tp_session_inc_refcount(session); 351 WARN_ON_ONCE(session->coll_list); 352 session->coll_list = clist; 353 spin_lock(&clist->lock); 354 list_add(&session->clist, &clist->list); 355 spin_unlock(&clist->lock); 356 } 357 358 static int l2tp_session_collision_add(struct l2tp_net *pn, 359 struct l2tp_session *session1, 360 struct l2tp_session *session2) 361 { 362 struct l2tp_session_coll_list *clist; 363 364 lockdep_assert_held(&pn->l2tp_session_idr_lock); 365 366 if (!session2) 367 return -EEXIST; 368 369 /* If existing session is in IP-encap tunnel, refuse new session */ 370 if (session2->tunnel->encap == L2TP_ENCAPTYPE_IP) 371 return -EEXIST; 372 373 clist = session2->coll_list; 374 if (!clist) { 375 /* First collision. Allocate list to manage the collided sessions 376 * and add the existing session to the list. 377 */ 378 clist = kmalloc(sizeof(*clist), GFP_ATOMIC); 379 if (!clist) 380 return -ENOMEM; 381 382 spin_lock_init(&clist->lock); 383 INIT_LIST_HEAD(&clist->list); 384 refcount_set(&clist->ref_count, 1); 385 l2tp_session_coll_list_add(clist, session2); 386 } 387 388 /* If existing session isn't already in the session hlist, add it. */ 389 if (!hash_hashed(&session2->hlist)) 390 hash_add(pn->l2tp_v3_session_htable, &session2->hlist, 391 session2->hlist_key); 392 393 /* Add new session to the hlist and collision list */ 394 hash_add(pn->l2tp_v3_session_htable, &session1->hlist, 395 session1->hlist_key); 396 refcount_inc(&clist->ref_count); 397 l2tp_session_coll_list_add(clist, session1); 398 399 return 0; 400 } 401 402 static void l2tp_session_collision_del(struct l2tp_net *pn, 403 struct l2tp_session *session) 404 { 405 struct l2tp_session_coll_list *clist = session->coll_list; 406 unsigned long session_key = session->session_id; 407 struct l2tp_session *session2; 408 409 lockdep_assert_held(&pn->l2tp_session_idr_lock); 410 411 hash_del(&session->hlist); 412 413 if (clist) { 414 /* Remove session from its collision list. If there 415 * are other sessions with the same ID, replace this 416 * session's IDR entry with that session, otherwise 417 * remove the IDR entry. If this is the last session, 418 * the collision list data is freed. 419 */ 420 spin_lock(&clist->lock); 421 list_del_init(&session->clist); 422 session2 = list_first_entry_or_null(&clist->list, struct l2tp_session, clist); 423 if (session2) { 424 void *old = idr_replace(&pn->l2tp_v3_session_idr, session2, session_key); 425 426 WARN_ON_ONCE(IS_ERR_VALUE(old)); 427 } else { 428 void *removed = idr_remove(&pn->l2tp_v3_session_idr, session_key); 429 430 WARN_ON_ONCE(removed != session); 431 } 432 session->coll_list = NULL; 433 spin_unlock(&clist->lock); 434 if (refcount_dec_and_test(&clist->ref_count)) 435 kfree(clist); 436 l2tp_session_dec_refcount(session); 437 } 438 } 439 440 int l2tp_session_register(struct l2tp_session *session, 441 struct l2tp_tunnel *tunnel) 442 { 443 struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net); 444 struct l2tp_session *other_session = NULL; 445 u32 session_key; 446 int err; 447 448 spin_lock_bh(&tunnel->list_lock); 449 spin_lock_bh(&pn->l2tp_session_idr_lock); 450 451 if (!tunnel->acpt_newsess) { 452 err = -ENODEV; 453 goto out; 454 } 455 456 if (tunnel->version == L2TP_HDR_VER_3) { 457 session_key = session->session_id; 458 err = idr_alloc_u32(&pn->l2tp_v3_session_idr, NULL, 459 &session_key, session_key, GFP_ATOMIC); 460 /* IP encap expects session IDs to be globally unique, while 461 * UDP encap doesn't. This isn't per the RFC, which says that 462 * sessions are identified only by the session ID, but is to 463 * support existing userspace which depends on it. 464 */ 465 if (err == -ENOSPC && tunnel->encap == L2TP_ENCAPTYPE_UDP) { 466 other_session = idr_find(&pn->l2tp_v3_session_idr, 467 session_key); 468 err = l2tp_session_collision_add(pn, session, 469 other_session); 470 } 471 } else { 472 session_key = l2tp_v2_session_key(tunnel->tunnel_id, 473 session->session_id); 474 err = idr_alloc_u32(&pn->l2tp_v2_session_idr, NULL, 475 &session_key, session_key, GFP_ATOMIC); 476 } 477 478 if (err) { 479 if (err == -ENOSPC) 480 err = -EEXIST; 481 goto out; 482 } 483 484 l2tp_tunnel_inc_refcount(tunnel); 485 list_add(&session->list, &tunnel->session_list); 486 487 if (tunnel->version == L2TP_HDR_VER_3) { 488 if (!other_session) 489 idr_replace(&pn->l2tp_v3_session_idr, session, session_key); 490 } else { 491 idr_replace(&pn->l2tp_v2_session_idr, session, session_key); 492 } 493 494 out: 495 spin_unlock_bh(&pn->l2tp_session_idr_lock); 496 spin_unlock_bh(&tunnel->list_lock); 497 498 if (!err) 499 trace_register_session(session); 500 501 return err; 502 } 503 EXPORT_SYMBOL_GPL(l2tp_session_register); 504 505 /***************************************************************************** 506 * Receive data handling 507 *****************************************************************************/ 508 509 /* Queue a skb in order. We come here only if the skb has an L2TP sequence 510 * number. 511 */ 512 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb) 513 { 514 struct sk_buff *skbp; 515 struct sk_buff *tmp; 516 u32 ns = L2TP_SKB_CB(skb)->ns; 517 518 spin_lock_bh(&session->reorder_q.lock); 519 skb_queue_walk_safe(&session->reorder_q, skbp, tmp) { 520 if (L2TP_SKB_CB(skbp)->ns > ns) { 521 __skb_queue_before(&session->reorder_q, skbp, skb); 522 atomic_long_inc(&session->stats.rx_oos_packets); 523 goto out; 524 } 525 } 526 527 __skb_queue_tail(&session->reorder_q, skb); 528 529 out: 530 spin_unlock_bh(&session->reorder_q.lock); 531 } 532 533 /* Dequeue a single skb. 534 */ 535 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb) 536 { 537 struct l2tp_tunnel *tunnel = session->tunnel; 538 int length = L2TP_SKB_CB(skb)->length; 539 540 /* We're about to requeue the skb, so return resources 541 * to its current owner (a socket receive buffer). 542 */ 543 skb_orphan(skb); 544 545 atomic_long_inc(&tunnel->stats.rx_packets); 546 atomic_long_add(length, &tunnel->stats.rx_bytes); 547 atomic_long_inc(&session->stats.rx_packets); 548 atomic_long_add(length, &session->stats.rx_bytes); 549 550 if (L2TP_SKB_CB(skb)->has_seq) { 551 /* Bump our Nr */ 552 session->nr++; 553 session->nr &= session->nr_max; 554 trace_session_seqnum_update(session); 555 } 556 557 /* call private receive handler */ 558 if (session->recv_skb) 559 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length); 560 else 561 kfree_skb(skb); 562 } 563 564 /* Dequeue skbs from the session's reorder_q, subject to packet order. 565 * Skbs that have been in the queue for too long are simply discarded. 566 */ 567 static void l2tp_recv_dequeue(struct l2tp_session *session) 568 { 569 struct sk_buff *skb; 570 struct sk_buff *tmp; 571 572 /* If the pkt at the head of the queue has the nr that we 573 * expect to send up next, dequeue it and any other 574 * in-sequence packets behind it. 575 */ 576 start: 577 spin_lock_bh(&session->reorder_q.lock); 578 skb_queue_walk_safe(&session->reorder_q, skb, tmp) { 579 struct l2tp_skb_cb *cb = L2TP_SKB_CB(skb); 580 581 /* If the packet has been pending on the queue for too long, discard it */ 582 if (time_after(jiffies, cb->expires)) { 583 atomic_long_inc(&session->stats.rx_seq_discards); 584 atomic_long_inc(&session->stats.rx_errors); 585 trace_session_pkt_expired(session, cb->ns); 586 session->reorder_skip = 1; 587 __skb_unlink(skb, &session->reorder_q); 588 kfree_skb(skb); 589 continue; 590 } 591 592 if (cb->has_seq) { 593 if (session->reorder_skip) { 594 session->reorder_skip = 0; 595 session->nr = cb->ns; 596 trace_session_seqnum_reset(session); 597 } 598 if (cb->ns != session->nr) 599 goto out; 600 } 601 __skb_unlink(skb, &session->reorder_q); 602 603 /* Process the skb. We release the queue lock while we 604 * do so to let other contexts process the queue. 605 */ 606 spin_unlock_bh(&session->reorder_q.lock); 607 l2tp_recv_dequeue_skb(session, skb); 608 goto start; 609 } 610 611 out: 612 spin_unlock_bh(&session->reorder_q.lock); 613 } 614 615 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr) 616 { 617 u32 nws; 618 619 if (nr >= session->nr) 620 nws = nr - session->nr; 621 else 622 nws = (session->nr_max + 1) - (session->nr - nr); 623 624 return nws < session->nr_window_size; 625 } 626 627 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if 628 * acceptable, else non-zero. 629 */ 630 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb) 631 { 632 struct l2tp_skb_cb *cb = L2TP_SKB_CB(skb); 633 634 if (!l2tp_seq_check_rx_window(session, cb->ns)) { 635 /* Packet sequence number is outside allowed window. 636 * Discard it. 637 */ 638 trace_session_pkt_outside_rx_window(session, cb->ns); 639 goto discard; 640 } 641 642 if (session->reorder_timeout != 0) { 643 /* Packet reordering enabled. Add skb to session's 644 * reorder queue, in order of ns. 645 */ 646 l2tp_recv_queue_skb(session, skb); 647 goto out; 648 } 649 650 /* Packet reordering disabled. Discard out-of-sequence packets, while 651 * tracking the number if in-sequence packets after the first OOS packet 652 * is seen. After nr_oos_count_max in-sequence packets, reset the 653 * sequence number to re-enable packet reception. 654 */ 655 if (cb->ns == session->nr) { 656 skb_queue_tail(&session->reorder_q, skb); 657 } else { 658 u32 nr_oos = cb->ns; 659 u32 nr_next = (session->nr_oos + 1) & session->nr_max; 660 661 if (nr_oos == nr_next) 662 session->nr_oos_count++; 663 else 664 session->nr_oos_count = 0; 665 666 session->nr_oos = nr_oos; 667 if (session->nr_oos_count > session->nr_oos_count_max) { 668 session->reorder_skip = 1; 669 } 670 if (!session->reorder_skip) { 671 atomic_long_inc(&session->stats.rx_seq_discards); 672 trace_session_pkt_oos(session, cb->ns); 673 goto discard; 674 } 675 skb_queue_tail(&session->reorder_q, skb); 676 } 677 678 out: 679 return 0; 680 681 discard: 682 return 1; 683 } 684 685 /* Do receive processing of L2TP data frames. We handle both L2TPv2 686 * and L2TPv3 data frames here. 687 * 688 * L2TPv2 Data Message Header 689 * 690 * 0 1 2 3 691 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 692 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 693 * |T|L|x|x|S|x|O|P|x|x|x|x| Ver | Length (opt) | 694 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 695 * | Tunnel ID | Session ID | 696 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 697 * | Ns (opt) | Nr (opt) | 698 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 699 * | Offset Size (opt) | Offset pad... (opt) 700 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 701 * 702 * Data frames are marked by T=0. All other fields are the same as 703 * those in L2TP control frames. 704 * 705 * L2TPv3 Data Message Header 706 * 707 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 708 * | L2TP Session Header | 709 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 710 * | L2-Specific Sublayer | 711 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 712 * | Tunnel Payload ... 713 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 714 * 715 * L2TPv3 Session Header Over IP 716 * 717 * 0 1 2 3 718 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 719 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 720 * | Session ID | 721 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 722 * | Cookie (optional, maximum 64 bits)... 723 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 724 * | 725 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 726 * 727 * L2TPv3 L2-Specific Sublayer Format 728 * 729 * 0 1 2 3 730 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 731 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 732 * |x|S|x|x|x|x|x|x| Sequence Number | 733 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 734 * 735 * Cookie value and sublayer format are negotiated with the peer when 736 * the session is set up. Unlike L2TPv2, we do not need to parse the 737 * packet header to determine if optional fields are present. 738 * 739 * Caller must already have parsed the frame and determined that it is 740 * a data (not control) frame before coming here. Fields up to the 741 * session-id have already been parsed and ptr points to the data 742 * after the session-id. 743 */ 744 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb, 745 unsigned char *ptr, unsigned char *optr, u16 hdrflags, 746 int length) 747 { 748 struct l2tp_tunnel *tunnel = session->tunnel; 749 int offset; 750 751 /* Parse and check optional cookie */ 752 if (session->peer_cookie_len > 0) { 753 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) { 754 pr_debug_ratelimited("%s: cookie mismatch (%u/%u). Discarding.\n", 755 tunnel->name, tunnel->tunnel_id, 756 session->session_id); 757 atomic_long_inc(&session->stats.rx_cookie_discards); 758 goto discard; 759 } 760 ptr += session->peer_cookie_len; 761 } 762 763 /* Handle the optional sequence numbers. Sequence numbers are 764 * in different places for L2TPv2 and L2TPv3. 765 * 766 * If we are the LAC, enable/disable sequence numbers under 767 * the control of the LNS. If no sequence numbers present but 768 * we were expecting them, discard frame. 769 */ 770 L2TP_SKB_CB(skb)->has_seq = 0; 771 if (tunnel->version == L2TP_HDR_VER_2) { 772 if (hdrflags & L2TP_HDRFLAG_S) { 773 /* Store L2TP info in the skb */ 774 L2TP_SKB_CB(skb)->ns = ntohs(*(__be16 *)ptr); 775 L2TP_SKB_CB(skb)->has_seq = 1; 776 ptr += 2; 777 /* Skip past nr in the header */ 778 ptr += 2; 779 780 } 781 } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) { 782 u32 l2h = ntohl(*(__be32 *)ptr); 783 784 if (l2h & 0x40000000) { 785 /* Store L2TP info in the skb */ 786 L2TP_SKB_CB(skb)->ns = l2h & 0x00ffffff; 787 L2TP_SKB_CB(skb)->has_seq = 1; 788 } 789 ptr += 4; 790 } 791 792 if (L2TP_SKB_CB(skb)->has_seq) { 793 /* Received a packet with sequence numbers. If we're the LAC, 794 * check if we sre sending sequence numbers and if not, 795 * configure it so. 796 */ 797 if (!session->lns_mode && !session->send_seq) { 798 trace_session_seqnum_lns_enable(session); 799 session->send_seq = 1; 800 l2tp_session_set_header_len(session, tunnel->version); 801 } 802 } else { 803 /* No sequence numbers. 804 * If user has configured mandatory sequence numbers, discard. 805 */ 806 if (session->recv_seq) { 807 pr_debug_ratelimited("%s: recv data has no seq numbers when required. Discarding.\n", 808 session->name); 809 atomic_long_inc(&session->stats.rx_seq_discards); 810 goto discard; 811 } 812 813 /* If we're the LAC and we're sending sequence numbers, the 814 * LNS has requested that we no longer send sequence numbers. 815 * If we're the LNS and we're sending sequence numbers, the 816 * LAC is broken. Discard the frame. 817 */ 818 if (!session->lns_mode && session->send_seq) { 819 trace_session_seqnum_lns_disable(session); 820 session->send_seq = 0; 821 l2tp_session_set_header_len(session, tunnel->version); 822 } else if (session->send_seq) { 823 pr_debug_ratelimited("%s: recv data has no seq numbers when required. Discarding.\n", 824 session->name); 825 atomic_long_inc(&session->stats.rx_seq_discards); 826 goto discard; 827 } 828 } 829 830 /* Session data offset is defined only for L2TPv2 and is 831 * indicated by an optional 16-bit value in the header. 832 */ 833 if (tunnel->version == L2TP_HDR_VER_2) { 834 /* If offset bit set, skip it. */ 835 if (hdrflags & L2TP_HDRFLAG_O) { 836 offset = ntohs(*(__be16 *)ptr); 837 ptr += 2 + offset; 838 } 839 } 840 841 offset = ptr - optr; 842 if (!pskb_may_pull(skb, offset)) 843 goto discard; 844 845 __skb_pull(skb, offset); 846 847 /* Prepare skb for adding to the session's reorder_q. Hold 848 * packets for max reorder_timeout or 1 second if not 849 * reordering. 850 */ 851 L2TP_SKB_CB(skb)->length = length; 852 L2TP_SKB_CB(skb)->expires = jiffies + 853 (session->reorder_timeout ? session->reorder_timeout : HZ); 854 855 /* Add packet to the session's receive queue. Reordering is done here, if 856 * enabled. Saved L2TP protocol info is stored in skb->sb[]. 857 */ 858 if (L2TP_SKB_CB(skb)->has_seq) { 859 if (l2tp_recv_data_seq(session, skb)) 860 goto discard; 861 } else { 862 /* No sequence numbers. Add the skb to the tail of the 863 * reorder queue. This ensures that it will be 864 * delivered after all previous sequenced skbs. 865 */ 866 skb_queue_tail(&session->reorder_q, skb); 867 } 868 869 /* Try to dequeue as many skbs from reorder_q as we can. */ 870 l2tp_recv_dequeue(session); 871 872 return; 873 874 discard: 875 atomic_long_inc(&session->stats.rx_errors); 876 kfree_skb(skb); 877 } 878 EXPORT_SYMBOL_GPL(l2tp_recv_common); 879 880 /* Drop skbs from the session's reorder_q 881 */ 882 static void l2tp_session_queue_purge(struct l2tp_session *session) 883 { 884 struct sk_buff *skb = NULL; 885 886 while ((skb = skb_dequeue(&session->reorder_q))) { 887 atomic_long_inc(&session->stats.rx_errors); 888 kfree_skb(skb); 889 } 890 } 891 892 /* UDP encapsulation receive handler. See net/ipv4/udp.c for details. */ 893 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb) 894 { 895 struct l2tp_session *session = NULL; 896 struct l2tp_tunnel *tunnel = NULL; 897 struct net *net = sock_net(sk); 898 unsigned char *ptr, *optr; 899 u16 hdrflags; 900 u16 version; 901 int length; 902 903 /* UDP has verified checksum */ 904 905 /* UDP always verifies the packet length. */ 906 __skb_pull(skb, sizeof(struct udphdr)); 907 908 /* Short packet? */ 909 if (!pskb_may_pull(skb, L2TP_HDR_SIZE_MAX)) 910 goto pass; 911 912 /* Point to L2TP header */ 913 optr = skb->data; 914 ptr = skb->data; 915 916 /* Get L2TP header flags */ 917 hdrflags = ntohs(*(__be16 *)ptr); 918 919 /* Get protocol version */ 920 version = hdrflags & L2TP_HDR_VER_MASK; 921 922 /* Get length of L2TP packet */ 923 length = skb->len; 924 925 /* If type is control packet, it is handled by userspace. */ 926 if (hdrflags & L2TP_HDRFLAG_T) 927 goto pass; 928 929 /* Skip flags */ 930 ptr += 2; 931 932 if (version == L2TP_HDR_VER_2) { 933 u16 tunnel_id, session_id; 934 935 /* If length is present, skip it */ 936 if (hdrflags & L2TP_HDRFLAG_L) 937 ptr += 2; 938 939 /* Extract tunnel and session ID */ 940 tunnel_id = ntohs(*(__be16 *)ptr); 941 ptr += 2; 942 session_id = ntohs(*(__be16 *)ptr); 943 ptr += 2; 944 945 session = l2tp_v2_session_get(net, tunnel_id, session_id); 946 } else { 947 u32 session_id; 948 949 ptr += 2; /* skip reserved bits */ 950 session_id = ntohl(*(__be32 *)ptr); 951 ptr += 4; 952 953 session = l2tp_v3_session_get(net, sk, session_id); 954 } 955 956 if (!session || !session->recv_skb) { 957 if (session) 958 l2tp_session_dec_refcount(session); 959 960 /* Not found? Pass to userspace to deal with */ 961 goto pass; 962 } 963 964 tunnel = session->tunnel; 965 966 /* Check protocol version */ 967 if (version != tunnel->version) 968 goto invalid; 969 970 if (version == L2TP_HDR_VER_3 && 971 l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr)) { 972 l2tp_session_dec_refcount(session); 973 goto invalid; 974 } 975 976 l2tp_recv_common(session, skb, ptr, optr, hdrflags, length); 977 l2tp_session_dec_refcount(session); 978 979 return 0; 980 981 invalid: 982 atomic_long_inc(&tunnel->stats.rx_invalid); 983 984 pass: 985 /* Put UDP header back */ 986 __skb_push(skb, sizeof(struct udphdr)); 987 988 return 1; 989 } 990 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv); 991 992 /* UDP encapsulation receive error handler. See net/ipv4/udp.c for details. */ 993 static void l2tp_udp_encap_err_recv(struct sock *sk, struct sk_buff *skb, int err, 994 __be16 port, u32 info, u8 *payload) 995 { 996 sk->sk_err = err; 997 sk_error_report(sk); 998 999 if (ip_hdr(skb)->version == IPVERSION) { 1000 if (inet_test_bit(RECVERR, sk)) 1001 return ip_icmp_error(sk, skb, err, port, info, payload); 1002 #if IS_ENABLED(CONFIG_IPV6) 1003 } else { 1004 if (inet6_test_bit(RECVERR6, sk)) 1005 return ipv6_icmp_error(sk, skb, err, port, info, payload); 1006 #endif 1007 } 1008 } 1009 1010 /************************************************************************ 1011 * Transmit handling 1012 ***********************************************************************/ 1013 1014 /* Build an L2TP header for the session into the buffer provided. 1015 */ 1016 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf) 1017 { 1018 struct l2tp_tunnel *tunnel = session->tunnel; 1019 __be16 *bufp = buf; 1020 __be16 *optr = buf; 1021 u16 flags = L2TP_HDR_VER_2; 1022 u32 tunnel_id = tunnel->peer_tunnel_id; 1023 u32 session_id = session->peer_session_id; 1024 1025 if (session->send_seq) 1026 flags |= L2TP_HDRFLAG_S; 1027 1028 /* Setup L2TP header. */ 1029 *bufp++ = htons(flags); 1030 *bufp++ = htons(tunnel_id); 1031 *bufp++ = htons(session_id); 1032 if (session->send_seq) { 1033 *bufp++ = htons(session->ns); 1034 *bufp++ = 0; 1035 session->ns++; 1036 session->ns &= 0xffff; 1037 trace_session_seqnum_update(session); 1038 } 1039 1040 return bufp - optr; 1041 } 1042 1043 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf) 1044 { 1045 struct l2tp_tunnel *tunnel = session->tunnel; 1046 char *bufp = buf; 1047 char *optr = bufp; 1048 1049 /* Setup L2TP header. The header differs slightly for UDP and 1050 * IP encapsulations. For UDP, there is 4 bytes of flags. 1051 */ 1052 if (tunnel->encap == L2TP_ENCAPTYPE_UDP) { 1053 u16 flags = L2TP_HDR_VER_3; 1054 *((__be16 *)bufp) = htons(flags); 1055 bufp += 2; 1056 *((__be16 *)bufp) = 0; 1057 bufp += 2; 1058 } 1059 1060 *((__be32 *)bufp) = htonl(session->peer_session_id); 1061 bufp += 4; 1062 if (session->cookie_len) { 1063 memcpy(bufp, &session->cookie[0], session->cookie_len); 1064 bufp += session->cookie_len; 1065 } 1066 if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) { 1067 u32 l2h = 0; 1068 1069 if (session->send_seq) { 1070 l2h = 0x40000000 | session->ns; 1071 session->ns++; 1072 session->ns &= 0xffffff; 1073 trace_session_seqnum_update(session); 1074 } 1075 1076 *((__be32 *)bufp) = htonl(l2h); 1077 bufp += 4; 1078 } 1079 1080 return bufp - optr; 1081 } 1082 1083 /* Queue the packet to IP for output: tunnel socket lock must be held */ 1084 static int l2tp_xmit_queue(struct l2tp_tunnel *tunnel, struct sk_buff *skb, struct flowi *fl) 1085 { 1086 int err; 1087 1088 skb->ignore_df = 1; 1089 skb_dst_drop(skb); 1090 #if IS_ENABLED(CONFIG_IPV6) 1091 if (l2tp_sk_is_v6(tunnel->sock)) 1092 err = inet6_csk_xmit(tunnel->sock, skb, NULL); 1093 else 1094 #endif 1095 err = ip_queue_xmit(tunnel->sock, skb, fl); 1096 1097 return err >= 0 ? NET_XMIT_SUCCESS : NET_XMIT_DROP; 1098 } 1099 1100 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb, unsigned int *len) 1101 { 1102 struct l2tp_tunnel *tunnel = session->tunnel; 1103 unsigned int data_len = skb->len; 1104 struct sock *sk = tunnel->sock; 1105 int headroom, uhlen, udp_len; 1106 int ret = NET_XMIT_SUCCESS; 1107 struct inet_sock *inet; 1108 struct udphdr *uh; 1109 1110 /* Check that there's enough headroom in the skb to insert IP, 1111 * UDP and L2TP headers. If not enough, expand it to 1112 * make room. Adjust truesize. 1113 */ 1114 uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(*uh) : 0; 1115 headroom = NET_SKB_PAD + sizeof(struct iphdr) + uhlen + session->hdr_len; 1116 if (skb_cow_head(skb, headroom)) { 1117 kfree_skb(skb); 1118 return NET_XMIT_DROP; 1119 } 1120 1121 /* Setup L2TP header */ 1122 if (tunnel->version == L2TP_HDR_VER_2) 1123 l2tp_build_l2tpv2_header(session, __skb_push(skb, session->hdr_len)); 1124 else 1125 l2tp_build_l2tpv3_header(session, __skb_push(skb, session->hdr_len)); 1126 1127 /* Reset skb netfilter state */ 1128 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1129 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED | IPSKB_REROUTED); 1130 nf_reset_ct(skb); 1131 1132 /* L2TP uses its own lockdep subclass to avoid lockdep splats caused by 1133 * nested socket calls on the same lockdep socket class. This can 1134 * happen when data from a user socket is routed over l2tp, which uses 1135 * another userspace socket. 1136 */ 1137 spin_lock_nested(&sk->sk_lock.slock, L2TP_DEPTH_NESTING); 1138 1139 if (sock_owned_by_user(sk)) { 1140 kfree_skb(skb); 1141 ret = NET_XMIT_DROP; 1142 goto out_unlock; 1143 } 1144 1145 /* The user-space may change the connection status for the user-space 1146 * provided socket at run time: we must check it under the socket lock 1147 */ 1148 if (tunnel->fd >= 0 && sk->sk_state != TCP_ESTABLISHED) { 1149 kfree_skb(skb); 1150 ret = NET_XMIT_DROP; 1151 goto out_unlock; 1152 } 1153 1154 /* Report transmitted length before we add encap header, which keeps 1155 * statistics consistent for both UDP and IP encap tx/rx paths. 1156 */ 1157 *len = skb->len; 1158 1159 inet = inet_sk(sk); 1160 switch (tunnel->encap) { 1161 case L2TP_ENCAPTYPE_UDP: 1162 /* Setup UDP header */ 1163 __skb_push(skb, sizeof(*uh)); 1164 skb_reset_transport_header(skb); 1165 uh = udp_hdr(skb); 1166 uh->source = inet->inet_sport; 1167 uh->dest = inet->inet_dport; 1168 udp_len = uhlen + session->hdr_len + data_len; 1169 uh->len = htons(udp_len); 1170 1171 /* Calculate UDP checksum if configured to do so */ 1172 #if IS_ENABLED(CONFIG_IPV6) 1173 if (l2tp_sk_is_v6(sk)) 1174 udp6_set_csum(udp_get_no_check6_tx(sk), 1175 skb, &inet6_sk(sk)->saddr, 1176 &sk->sk_v6_daddr, udp_len); 1177 else 1178 #endif 1179 udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr, 1180 inet->inet_daddr, udp_len); 1181 break; 1182 1183 case L2TP_ENCAPTYPE_IP: 1184 break; 1185 } 1186 1187 ret = l2tp_xmit_queue(tunnel, skb, &inet->cork.fl); 1188 1189 out_unlock: 1190 spin_unlock(&sk->sk_lock.slock); 1191 1192 return ret; 1193 } 1194 1195 /* If caller requires the skb to have a ppp header, the header must be 1196 * inserted in the skb data before calling this function. 1197 */ 1198 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb) 1199 { 1200 unsigned int len = 0; 1201 int ret; 1202 1203 ret = l2tp_xmit_core(session, skb, &len); 1204 if (ret == NET_XMIT_SUCCESS) { 1205 atomic_long_inc(&session->tunnel->stats.tx_packets); 1206 atomic_long_add(len, &session->tunnel->stats.tx_bytes); 1207 atomic_long_inc(&session->stats.tx_packets); 1208 atomic_long_add(len, &session->stats.tx_bytes); 1209 } else { 1210 atomic_long_inc(&session->tunnel->stats.tx_errors); 1211 atomic_long_inc(&session->stats.tx_errors); 1212 } 1213 return ret; 1214 } 1215 EXPORT_SYMBOL_GPL(l2tp_xmit_skb); 1216 1217 /***************************************************************************** 1218 * Tinnel and session create/destroy. 1219 *****************************************************************************/ 1220 1221 /* Tunnel socket destruct hook. 1222 * The tunnel context is deleted only when all session sockets have been 1223 * closed. 1224 */ 1225 static void l2tp_tunnel_destruct(struct sock *sk) 1226 { 1227 struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk); 1228 1229 if (!tunnel) 1230 goto end; 1231 1232 /* Disable udp encapsulation */ 1233 switch (tunnel->encap) { 1234 case L2TP_ENCAPTYPE_UDP: 1235 /* No longer an encapsulation socket. See net/ipv4/udp.c */ 1236 WRITE_ONCE(udp_sk(sk)->encap_type, 0); 1237 udp_sk(sk)->encap_rcv = NULL; 1238 udp_sk(sk)->encap_destroy = NULL; 1239 break; 1240 case L2TP_ENCAPTYPE_IP: 1241 break; 1242 } 1243 1244 /* Remove hooks into tunnel socket */ 1245 write_lock_bh(&sk->sk_callback_lock); 1246 sk->sk_destruct = tunnel->old_sk_destruct; 1247 sk->sk_user_data = NULL; 1248 write_unlock_bh(&sk->sk_callback_lock); 1249 1250 /* Call the original destructor */ 1251 if (sk->sk_destruct) 1252 (*sk->sk_destruct)(sk); 1253 1254 kfree_rcu(tunnel, rcu); 1255 end: 1256 return; 1257 } 1258 1259 /* Remove an l2tp session from l2tp_core's lists. */ 1260 static void l2tp_session_unhash(struct l2tp_session *session) 1261 { 1262 struct l2tp_tunnel *tunnel = session->tunnel; 1263 1264 if (tunnel) { 1265 struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net); 1266 struct l2tp_session *removed = session; 1267 1268 spin_lock_bh(&tunnel->list_lock); 1269 spin_lock_bh(&pn->l2tp_session_idr_lock); 1270 1271 /* Remove from the per-tunnel list */ 1272 list_del_init(&session->list); 1273 1274 /* Remove from per-net IDR */ 1275 if (tunnel->version == L2TP_HDR_VER_3) { 1276 if (hash_hashed(&session->hlist)) 1277 l2tp_session_collision_del(pn, session); 1278 else 1279 removed = idr_remove(&pn->l2tp_v3_session_idr, 1280 session->session_id); 1281 } else { 1282 u32 session_key = l2tp_v2_session_key(tunnel->tunnel_id, 1283 session->session_id); 1284 removed = idr_remove(&pn->l2tp_v2_session_idr, 1285 session_key); 1286 } 1287 WARN_ON_ONCE(removed && removed != session); 1288 1289 spin_unlock_bh(&pn->l2tp_session_idr_lock); 1290 spin_unlock_bh(&tunnel->list_lock); 1291 1292 synchronize_rcu(); 1293 } 1294 } 1295 1296 /* When the tunnel is closed, all the attached sessions need to go too. 1297 */ 1298 static void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel) 1299 { 1300 struct l2tp_session *session; 1301 1302 spin_lock_bh(&tunnel->list_lock); 1303 tunnel->acpt_newsess = false; 1304 for (;;) { 1305 session = list_first_entry_or_null(&tunnel->session_list, 1306 struct l2tp_session, list); 1307 if (!session) 1308 break; 1309 l2tp_session_inc_refcount(session); 1310 list_del_init(&session->list); 1311 spin_unlock_bh(&tunnel->list_lock); 1312 l2tp_session_delete(session); 1313 spin_lock_bh(&tunnel->list_lock); 1314 l2tp_session_dec_refcount(session); 1315 } 1316 spin_unlock_bh(&tunnel->list_lock); 1317 } 1318 1319 /* Tunnel socket destroy hook for UDP encapsulation */ 1320 static void l2tp_udp_encap_destroy(struct sock *sk) 1321 { 1322 struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk); 1323 1324 if (tunnel) 1325 l2tp_tunnel_delete(tunnel); 1326 } 1327 1328 static void l2tp_tunnel_remove(struct net *net, struct l2tp_tunnel *tunnel) 1329 { 1330 struct l2tp_net *pn = l2tp_pernet(net); 1331 1332 spin_lock_bh(&pn->l2tp_tunnel_idr_lock); 1333 idr_remove(&pn->l2tp_tunnel_idr, tunnel->tunnel_id); 1334 spin_unlock_bh(&pn->l2tp_tunnel_idr_lock); 1335 } 1336 1337 /* Workqueue tunnel deletion function */ 1338 static void l2tp_tunnel_del_work(struct work_struct *work) 1339 { 1340 struct l2tp_tunnel *tunnel = container_of(work, struct l2tp_tunnel, 1341 del_work); 1342 struct sock *sk = tunnel->sock; 1343 struct socket *sock = sk->sk_socket; 1344 1345 l2tp_tunnel_closeall(tunnel); 1346 1347 /* If the tunnel socket was created within the kernel, use 1348 * the sk API to release it here. 1349 */ 1350 if (tunnel->fd < 0) { 1351 if (sock) { 1352 kernel_sock_shutdown(sock, SHUT_RDWR); 1353 sock_release(sock); 1354 } 1355 } 1356 1357 l2tp_tunnel_remove(tunnel->l2tp_net, tunnel); 1358 /* drop initial ref */ 1359 l2tp_tunnel_dec_refcount(tunnel); 1360 1361 /* drop workqueue ref */ 1362 l2tp_tunnel_dec_refcount(tunnel); 1363 } 1364 1365 /* Create a socket for the tunnel, if one isn't set up by 1366 * userspace. This is used for static tunnels where there is no 1367 * managing L2TP daemon. 1368 * 1369 * Since we don't want these sockets to keep a namespace alive by 1370 * themselves, we drop the socket's namespace refcount after creation. 1371 * These sockets are freed when the namespace exits using the pernet 1372 * exit hook. 1373 */ 1374 static int l2tp_tunnel_sock_create(struct net *net, 1375 u32 tunnel_id, 1376 u32 peer_tunnel_id, 1377 struct l2tp_tunnel_cfg *cfg, 1378 struct socket **sockp) 1379 { 1380 int err = -EINVAL; 1381 struct socket *sock = NULL; 1382 struct udp_port_cfg udp_conf; 1383 1384 switch (cfg->encap) { 1385 case L2TP_ENCAPTYPE_UDP: 1386 memset(&udp_conf, 0, sizeof(udp_conf)); 1387 1388 #if IS_ENABLED(CONFIG_IPV6) 1389 if (cfg->local_ip6 && cfg->peer_ip6) { 1390 udp_conf.family = AF_INET6; 1391 memcpy(&udp_conf.local_ip6, cfg->local_ip6, 1392 sizeof(udp_conf.local_ip6)); 1393 memcpy(&udp_conf.peer_ip6, cfg->peer_ip6, 1394 sizeof(udp_conf.peer_ip6)); 1395 udp_conf.use_udp6_tx_checksums = 1396 !cfg->udp6_zero_tx_checksums; 1397 udp_conf.use_udp6_rx_checksums = 1398 !cfg->udp6_zero_rx_checksums; 1399 } else 1400 #endif 1401 { 1402 udp_conf.family = AF_INET; 1403 udp_conf.local_ip = cfg->local_ip; 1404 udp_conf.peer_ip = cfg->peer_ip; 1405 udp_conf.use_udp_checksums = cfg->use_udp_checksums; 1406 } 1407 1408 udp_conf.local_udp_port = htons(cfg->local_udp_port); 1409 udp_conf.peer_udp_port = htons(cfg->peer_udp_port); 1410 1411 err = udp_sock_create(net, &udp_conf, &sock); 1412 if (err < 0) 1413 goto out; 1414 1415 break; 1416 1417 case L2TP_ENCAPTYPE_IP: 1418 #if IS_ENABLED(CONFIG_IPV6) 1419 if (cfg->local_ip6 && cfg->peer_ip6) { 1420 struct sockaddr_l2tpip6 ip6_addr = {0}; 1421 1422 err = sock_create_kern(net, AF_INET6, SOCK_DGRAM, 1423 IPPROTO_L2TP, &sock); 1424 if (err < 0) 1425 goto out; 1426 1427 ip6_addr.l2tp_family = AF_INET6; 1428 memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6, 1429 sizeof(ip6_addr.l2tp_addr)); 1430 ip6_addr.l2tp_conn_id = tunnel_id; 1431 err = kernel_bind(sock, (struct sockaddr *)&ip6_addr, 1432 sizeof(ip6_addr)); 1433 if (err < 0) 1434 goto out; 1435 1436 ip6_addr.l2tp_family = AF_INET6; 1437 memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6, 1438 sizeof(ip6_addr.l2tp_addr)); 1439 ip6_addr.l2tp_conn_id = peer_tunnel_id; 1440 err = kernel_connect(sock, 1441 (struct sockaddr *)&ip6_addr, 1442 sizeof(ip6_addr), 0); 1443 if (err < 0) 1444 goto out; 1445 } else 1446 #endif 1447 { 1448 struct sockaddr_l2tpip ip_addr = {0}; 1449 1450 err = sock_create_kern(net, AF_INET, SOCK_DGRAM, 1451 IPPROTO_L2TP, &sock); 1452 if (err < 0) 1453 goto out; 1454 1455 ip_addr.l2tp_family = AF_INET; 1456 ip_addr.l2tp_addr = cfg->local_ip; 1457 ip_addr.l2tp_conn_id = tunnel_id; 1458 err = kernel_bind(sock, (struct sockaddr *)&ip_addr, 1459 sizeof(ip_addr)); 1460 if (err < 0) 1461 goto out; 1462 1463 ip_addr.l2tp_family = AF_INET; 1464 ip_addr.l2tp_addr = cfg->peer_ip; 1465 ip_addr.l2tp_conn_id = peer_tunnel_id; 1466 err = kernel_connect(sock, (struct sockaddr *)&ip_addr, 1467 sizeof(ip_addr), 0); 1468 if (err < 0) 1469 goto out; 1470 } 1471 break; 1472 1473 default: 1474 goto out; 1475 } 1476 1477 out: 1478 *sockp = sock; 1479 if (err < 0 && sock) { 1480 kernel_sock_shutdown(sock, SHUT_RDWR); 1481 sock_release(sock); 1482 *sockp = NULL; 1483 } 1484 1485 return err; 1486 } 1487 1488 int l2tp_tunnel_create(int fd, int version, u32 tunnel_id, u32 peer_tunnel_id, 1489 struct l2tp_tunnel_cfg *cfg, struct l2tp_tunnel **tunnelp) 1490 { 1491 struct l2tp_tunnel *tunnel = NULL; 1492 int err; 1493 enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP; 1494 1495 if (cfg) 1496 encap = cfg->encap; 1497 1498 tunnel = kzalloc(sizeof(*tunnel), GFP_KERNEL); 1499 if (!tunnel) { 1500 err = -ENOMEM; 1501 goto err; 1502 } 1503 1504 tunnel->version = version; 1505 tunnel->tunnel_id = tunnel_id; 1506 tunnel->peer_tunnel_id = peer_tunnel_id; 1507 1508 tunnel->magic = L2TP_TUNNEL_MAGIC; 1509 sprintf(&tunnel->name[0], "tunl %u", tunnel_id); 1510 spin_lock_init(&tunnel->list_lock); 1511 tunnel->acpt_newsess = true; 1512 INIT_LIST_HEAD(&tunnel->session_list); 1513 1514 tunnel->encap = encap; 1515 1516 refcount_set(&tunnel->ref_count, 1); 1517 tunnel->fd = fd; 1518 1519 /* Init delete workqueue struct */ 1520 INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work); 1521 1522 err = 0; 1523 err: 1524 if (tunnelp) 1525 *tunnelp = tunnel; 1526 1527 return err; 1528 } 1529 EXPORT_SYMBOL_GPL(l2tp_tunnel_create); 1530 1531 static int l2tp_validate_socket(const struct sock *sk, const struct net *net, 1532 enum l2tp_encap_type encap) 1533 { 1534 if (!net_eq(sock_net(sk), net)) 1535 return -EINVAL; 1536 1537 if (sk->sk_type != SOCK_DGRAM) 1538 return -EPROTONOSUPPORT; 1539 1540 if (sk->sk_family != PF_INET && sk->sk_family != PF_INET6) 1541 return -EPROTONOSUPPORT; 1542 1543 if ((encap == L2TP_ENCAPTYPE_UDP && sk->sk_protocol != IPPROTO_UDP) || 1544 (encap == L2TP_ENCAPTYPE_IP && sk->sk_protocol != IPPROTO_L2TP)) 1545 return -EPROTONOSUPPORT; 1546 1547 if (sk->sk_user_data) 1548 return -EBUSY; 1549 1550 return 0; 1551 } 1552 1553 int l2tp_tunnel_register(struct l2tp_tunnel *tunnel, struct net *net, 1554 struct l2tp_tunnel_cfg *cfg) 1555 { 1556 struct l2tp_net *pn = l2tp_pernet(net); 1557 u32 tunnel_id = tunnel->tunnel_id; 1558 struct socket *sock; 1559 struct sock *sk; 1560 int ret; 1561 1562 spin_lock_bh(&pn->l2tp_tunnel_idr_lock); 1563 ret = idr_alloc_u32(&pn->l2tp_tunnel_idr, NULL, &tunnel_id, tunnel_id, 1564 GFP_ATOMIC); 1565 spin_unlock_bh(&pn->l2tp_tunnel_idr_lock); 1566 if (ret) 1567 return ret == -ENOSPC ? -EEXIST : ret; 1568 1569 if (tunnel->fd < 0) { 1570 ret = l2tp_tunnel_sock_create(net, tunnel->tunnel_id, 1571 tunnel->peer_tunnel_id, cfg, 1572 &sock); 1573 if (ret < 0) 1574 goto err; 1575 } else { 1576 sock = sockfd_lookup(tunnel->fd, &ret); 1577 if (!sock) 1578 goto err; 1579 } 1580 1581 sk = sock->sk; 1582 lock_sock(sk); 1583 write_lock_bh(&sk->sk_callback_lock); 1584 ret = l2tp_validate_socket(sk, net, tunnel->encap); 1585 if (ret < 0) 1586 goto err_inval_sock; 1587 rcu_assign_sk_user_data(sk, tunnel); 1588 write_unlock_bh(&sk->sk_callback_lock); 1589 1590 if (tunnel->encap == L2TP_ENCAPTYPE_UDP) { 1591 struct udp_tunnel_sock_cfg udp_cfg = { 1592 .sk_user_data = tunnel, 1593 .encap_type = UDP_ENCAP_L2TPINUDP, 1594 .encap_rcv = l2tp_udp_encap_recv, 1595 .encap_err_rcv = l2tp_udp_encap_err_recv, 1596 .encap_destroy = l2tp_udp_encap_destroy, 1597 }; 1598 1599 setup_udp_tunnel_sock(net, sock, &udp_cfg); 1600 } 1601 1602 tunnel->old_sk_destruct = sk->sk_destruct; 1603 sk->sk_destruct = &l2tp_tunnel_destruct; 1604 sk->sk_allocation = GFP_ATOMIC; 1605 release_sock(sk); 1606 1607 sock_hold(sk); 1608 tunnel->sock = sk; 1609 tunnel->l2tp_net = net; 1610 1611 spin_lock_bh(&pn->l2tp_tunnel_idr_lock); 1612 idr_replace(&pn->l2tp_tunnel_idr, tunnel, tunnel->tunnel_id); 1613 spin_unlock_bh(&pn->l2tp_tunnel_idr_lock); 1614 1615 trace_register_tunnel(tunnel); 1616 1617 if (tunnel->fd >= 0) 1618 sockfd_put(sock); 1619 1620 return 0; 1621 1622 err_inval_sock: 1623 write_unlock_bh(&sk->sk_callback_lock); 1624 release_sock(sk); 1625 1626 if (tunnel->fd < 0) 1627 sock_release(sock); 1628 else 1629 sockfd_put(sock); 1630 err: 1631 l2tp_tunnel_remove(net, tunnel); 1632 return ret; 1633 } 1634 EXPORT_SYMBOL_GPL(l2tp_tunnel_register); 1635 1636 /* This function is used by the netlink TUNNEL_DELETE command. 1637 */ 1638 void l2tp_tunnel_delete(struct l2tp_tunnel *tunnel) 1639 { 1640 if (!test_and_set_bit(0, &tunnel->dead)) { 1641 trace_delete_tunnel(tunnel); 1642 l2tp_tunnel_inc_refcount(tunnel); 1643 queue_work(l2tp_wq, &tunnel->del_work); 1644 } 1645 } 1646 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete); 1647 1648 void l2tp_session_delete(struct l2tp_session *session) 1649 { 1650 if (test_and_set_bit(0, &session->dead)) 1651 return; 1652 1653 trace_delete_session(session); 1654 l2tp_session_unhash(session); 1655 l2tp_session_queue_purge(session); 1656 if (session->session_close) 1657 (*session->session_close)(session); 1658 1659 l2tp_session_dec_refcount(session); 1660 } 1661 EXPORT_SYMBOL_GPL(l2tp_session_delete); 1662 1663 /* We come here whenever a session's send_seq, cookie_len or 1664 * l2specific_type parameters are set. 1665 */ 1666 void l2tp_session_set_header_len(struct l2tp_session *session, int version) 1667 { 1668 if (version == L2TP_HDR_VER_2) { 1669 session->hdr_len = 6; 1670 if (session->send_seq) 1671 session->hdr_len += 4; 1672 } else { 1673 session->hdr_len = 4 + session->cookie_len; 1674 session->hdr_len += l2tp_get_l2specific_len(session); 1675 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP) 1676 session->hdr_len += 4; 1677 } 1678 } 1679 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len); 1680 1681 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id, 1682 u32 peer_session_id, struct l2tp_session_cfg *cfg) 1683 { 1684 struct l2tp_session *session; 1685 1686 session = kzalloc(sizeof(*session) + priv_size, GFP_KERNEL); 1687 if (session) { 1688 session->magic = L2TP_SESSION_MAGIC; 1689 session->tunnel = tunnel; 1690 1691 session->session_id = session_id; 1692 session->peer_session_id = peer_session_id; 1693 session->nr = 0; 1694 if (tunnel->version == L2TP_HDR_VER_2) 1695 session->nr_max = 0xffff; 1696 else 1697 session->nr_max = 0xffffff; 1698 session->nr_window_size = session->nr_max / 2; 1699 session->nr_oos_count_max = 4; 1700 1701 /* Use NR of first received packet */ 1702 session->reorder_skip = 1; 1703 1704 sprintf(&session->name[0], "sess %u/%u", 1705 tunnel->tunnel_id, session->session_id); 1706 1707 skb_queue_head_init(&session->reorder_q); 1708 1709 session->hlist_key = l2tp_v3_session_hashkey(tunnel->sock, session->session_id); 1710 INIT_HLIST_NODE(&session->hlist); 1711 INIT_LIST_HEAD(&session->clist); 1712 INIT_LIST_HEAD(&session->list); 1713 1714 if (cfg) { 1715 session->pwtype = cfg->pw_type; 1716 session->send_seq = cfg->send_seq; 1717 session->recv_seq = cfg->recv_seq; 1718 session->lns_mode = cfg->lns_mode; 1719 session->reorder_timeout = cfg->reorder_timeout; 1720 session->l2specific_type = cfg->l2specific_type; 1721 session->cookie_len = cfg->cookie_len; 1722 memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len); 1723 session->peer_cookie_len = cfg->peer_cookie_len; 1724 memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len); 1725 } 1726 1727 l2tp_session_set_header_len(session, tunnel->version); 1728 1729 refcount_set(&session->ref_count, 1); 1730 1731 return session; 1732 } 1733 1734 return ERR_PTR(-ENOMEM); 1735 } 1736 EXPORT_SYMBOL_GPL(l2tp_session_create); 1737 1738 /***************************************************************************** 1739 * Init and cleanup 1740 *****************************************************************************/ 1741 1742 static __net_init int l2tp_init_net(struct net *net) 1743 { 1744 struct l2tp_net *pn = net_generic(net, l2tp_net_id); 1745 1746 idr_init(&pn->l2tp_tunnel_idr); 1747 spin_lock_init(&pn->l2tp_tunnel_idr_lock); 1748 1749 idr_init(&pn->l2tp_v2_session_idr); 1750 idr_init(&pn->l2tp_v3_session_idr); 1751 spin_lock_init(&pn->l2tp_session_idr_lock); 1752 1753 return 0; 1754 } 1755 1756 static __net_exit void l2tp_exit_net(struct net *net) 1757 { 1758 struct l2tp_net *pn = l2tp_pernet(net); 1759 struct l2tp_tunnel *tunnel = NULL; 1760 unsigned long tunnel_id, tmp; 1761 1762 rcu_read_lock_bh(); 1763 idr_for_each_entry_ul(&pn->l2tp_tunnel_idr, tunnel, tmp, tunnel_id) { 1764 if (tunnel) 1765 l2tp_tunnel_delete(tunnel); 1766 } 1767 rcu_read_unlock_bh(); 1768 1769 if (l2tp_wq) 1770 flush_workqueue(l2tp_wq); 1771 rcu_barrier(); 1772 1773 idr_destroy(&pn->l2tp_v2_session_idr); 1774 idr_destroy(&pn->l2tp_v3_session_idr); 1775 idr_destroy(&pn->l2tp_tunnel_idr); 1776 } 1777 1778 static struct pernet_operations l2tp_net_ops = { 1779 .init = l2tp_init_net, 1780 .exit = l2tp_exit_net, 1781 .id = &l2tp_net_id, 1782 .size = sizeof(struct l2tp_net), 1783 }; 1784 1785 static int __init l2tp_init(void) 1786 { 1787 int rc = 0; 1788 1789 rc = register_pernet_device(&l2tp_net_ops); 1790 if (rc) 1791 goto out; 1792 1793 l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0); 1794 if (!l2tp_wq) { 1795 pr_err("alloc_workqueue failed\n"); 1796 unregister_pernet_device(&l2tp_net_ops); 1797 rc = -ENOMEM; 1798 goto out; 1799 } 1800 1801 pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION); 1802 1803 out: 1804 return rc; 1805 } 1806 1807 static void __exit l2tp_exit(void) 1808 { 1809 unregister_pernet_device(&l2tp_net_ops); 1810 if (l2tp_wq) { 1811 destroy_workqueue(l2tp_wq); 1812 l2tp_wq = NULL; 1813 } 1814 } 1815 1816 module_init(l2tp_init); 1817 module_exit(l2tp_exit); 1818 1819 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>"); 1820 MODULE_DESCRIPTION("L2TP core"); 1821 MODULE_LICENSE("GPL"); 1822 MODULE_VERSION(L2TP_DRV_VERSION); 1823