1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /** -*- linux-c -*- *********************************************************** 3 * Linux PPP over Ethernet (PPPoX/PPPoE) Sockets 4 * 5 * PPPoX --- Generic PPP encapsulation socket family 6 * PPPoE --- PPP over Ethernet (RFC 2516) 7 * 8 * Version: 0.7.0 9 * 10 * 070228 : Fix to allow multiple sessions with same remote MAC and same 11 * session id by including the local device ifindex in the 12 * tuple identifying a session. This also ensures packets can't 13 * be injected into a session from interfaces other than the one 14 * specified by userspace. Florian Zumbiehl <florz@florz.de> 15 * (Oh, BTW, this one is YYMMDD, in case you were wondering ...) 16 * 220102 : Fix module use count on failure in pppoe_create, pppox_sk -acme 17 * 030700 : Fixed connect logic to allow for disconnect. 18 * 270700 : Fixed potential SMP problems; we must protect against 19 * simultaneous invocation of ppp_input 20 * and ppp_unregister_channel. 21 * 040800 : Respect reference count mechanisms on net-devices. 22 * 200800 : fix kfree(skb) in pppoe_rcv (acme) 23 * Module reference count is decremented in the right spot now, 24 * guards against sock_put not actually freeing the sk 25 * in pppoe_release. 26 * 051000 : Initialization cleanup. 27 * 111100 : Fix recvmsg. 28 * 050101 : Fix PADT processing. 29 * 140501 : Use pppoe_rcv_core to handle all backlog. (Alexey) 30 * 170701 : Do not lock_sock with rwlock held. (DaveM) 31 * Ignore discovery frames if user has socket 32 * locked. (DaveM) 33 * Ignore return value of dev_queue_xmit in __pppoe_xmit 34 * or else we may kfree an SKB twice. (DaveM) 35 * 190701 : When doing copies of skb's in __pppoe_xmit, always delete 36 * the original skb that was passed in on success, never on 37 * failure. Delete the copy of the skb on failure to avoid 38 * a memory leak. 39 * 081001 : Misc. cleanup (licence string, non-blocking, prevent 40 * reference of device on close). 41 * 121301 : New ppp channels interface; cannot unregister a channel 42 * from interrupts. Thus, we mark the socket as a ZOMBIE 43 * and do the unregistration later. 44 * 081002 : seq_file support for proc stuff -acme 45 * 111602 : Merge all 2.4 fixes into 2.5/2.6 tree. Label 2.5/2.6 46 * as version 0.7. Spacing cleanup. 47 * Author: Michal Ostrowski <mostrows@speakeasy.net> 48 * Contributors: 49 * Arnaldo Carvalho de Melo <acme@conectiva.com.br> 50 * David S. Miller (davem@redhat.com) 51 * 52 * License: 53 */ 54 55 #include <linux/string.h> 56 #include <linux/module.h> 57 #include <linux/kernel.h> 58 #include <linux/slab.h> 59 #include <linux/errno.h> 60 #include <linux/netdevice.h> 61 #include <linux/net.h> 62 #include <linux/inetdevice.h> 63 #include <linux/etherdevice.h> 64 #include <linux/skbuff.h> 65 #include <linux/init.h> 66 #include <linux/if_ether.h> 67 #include <linux/if_pppox.h> 68 #include <linux/ppp_channel.h> 69 #include <linux/ppp_defs.h> 70 #include <linux/ppp-ioctl.h> 71 #include <linux/notifier.h> 72 #include <linux/file.h> 73 #include <linux/proc_fs.h> 74 #include <linux/seq_file.h> 75 76 #include <linux/nsproxy.h> 77 #include <net/net_namespace.h> 78 #include <net/netns/generic.h> 79 #include <net/sock.h> 80 #include <net/gro.h> 81 82 #include <linux/uaccess.h> 83 84 #define PPPOE_HASH_BITS CONFIG_PPPOE_HASH_BITS 85 #define PPPOE_HASH_SIZE (1 << PPPOE_HASH_BITS) 86 #define PPPOE_HASH_MASK (PPPOE_HASH_SIZE - 1) 87 88 static const struct proto_ops pppoe_ops; 89 static const struct ppp_channel_ops pppoe_chan_ops; 90 91 /* per-net private data for this module */ 92 static unsigned int pppoe_net_id __read_mostly; 93 struct pppoe_net { 94 /* 95 * we could use _single_ hash table for all 96 * nets by injecting net id into the hash but 97 * it would increase hash chains and add 98 * a few additional math comparisons messy 99 * as well, moreover in case of SMP less locking 100 * controversy here 101 */ 102 struct pppox_sock __rcu *hash_table[PPPOE_HASH_SIZE]; 103 spinlock_t hash_lock; 104 }; 105 106 /* 107 * PPPoE could be in the following stages: 108 * 1) Discovery stage (to obtain remote MAC and Session ID) 109 * 2) Session stage (MAC and SID are known) 110 * 111 * Ethernet frames have a special tag for this but 112 * we use simpler approach based on session id 113 */ 114 static inline bool stage_session(__be16 sid) 115 { 116 return sid != 0; 117 } 118 119 static inline struct pppoe_net *pppoe_pernet(struct net *net) 120 { 121 return net_generic(net, pppoe_net_id); 122 } 123 124 static inline int cmp_2_addr(struct pppoe_addr *a, struct pppoe_addr *b) 125 { 126 return a->sid == b->sid && ether_addr_equal(a->remote, b->remote); 127 } 128 129 static inline int cmp_addr(struct pppoe_addr *a, __be16 sid, char *addr) 130 { 131 return a->sid == sid && ether_addr_equal(a->remote, addr); 132 } 133 134 #if 8 % PPPOE_HASH_BITS 135 #error 8 must be a multiple of PPPOE_HASH_BITS 136 #endif 137 138 static u8 hash_item(__be16 sid, const u8 addr[ETH_ALEN]) 139 { 140 const u16 *addr16 = (const u16 *)addr; 141 unsigned int i; 142 u16 hash16; 143 u8 hash; 144 145 hash16 = addr16[0] ^ addr16[1] ^ addr16[2] ^ (__force u16)sid; 146 hash = (hash16 >> 8) ^ hash16; 147 for (i = 8; (i >>= 1) >= PPPOE_HASH_BITS;) 148 hash ^= hash >> i; 149 150 return hash & PPPOE_HASH_MASK; 151 } 152 153 /********************************************************************** 154 * 155 * Set/get/delete/rehash items (internal versions) 156 * 157 **********************************************************************/ 158 static struct pppox_sock *__get_item(struct pppoe_net *pn, __be16 sid, 159 unsigned char *addr, int ifindex) 160 { 161 int hash = hash_item(sid, addr); 162 struct pppox_sock *ret; 163 164 ret = rcu_dereference(pn->hash_table[hash]); 165 while (ret) { 166 if (cmp_addr(&ret->pppoe_pa, sid, addr) && 167 ret->pppoe_ifindex == ifindex) 168 return ret; 169 170 ret = rcu_dereference(ret->next); 171 } 172 173 return NULL; 174 } 175 176 static int __set_item(struct pppoe_net *pn, struct pppox_sock *po) 177 { 178 int hash = hash_item(po->pppoe_pa.sid, po->pppoe_pa.remote); 179 struct pppox_sock *ret, *first; 180 181 first = rcu_dereference_protected(pn->hash_table[hash], lockdep_is_held(&pn->hash_lock)); 182 ret = first; 183 while (ret) { 184 if (cmp_2_addr(&ret->pppoe_pa, &po->pppoe_pa) && 185 ret->pppoe_ifindex == po->pppoe_ifindex) 186 return -EALREADY; 187 188 ret = rcu_dereference_protected(ret->next, lockdep_is_held(&pn->hash_lock)); 189 } 190 191 RCU_INIT_POINTER(po->next, first); 192 rcu_assign_pointer(pn->hash_table[hash], po); 193 194 return 0; 195 } 196 197 static void __delete_item(struct pppoe_net *pn, const struct pppox_sock *po) 198 { 199 int hash = hash_item(po->pppoe_pa.sid, po->pppoe_pa.remote); 200 struct pppox_sock *ret, __rcu **src; 201 202 ret = rcu_dereference_protected(pn->hash_table[hash], lockdep_is_held(&pn->hash_lock)); 203 src = &pn->hash_table[hash]; 204 205 while (ret) { 206 if (ret == po) { 207 struct pppox_sock *next; 208 209 next = rcu_dereference_protected(ret->next, 210 lockdep_is_held(&pn->hash_lock)); 211 rcu_assign_pointer(*src, next); 212 break; 213 } 214 215 src = &ret->next; 216 ret = rcu_dereference_protected(ret->next, lockdep_is_held(&pn->hash_lock)); 217 } 218 } 219 220 /********************************************************************** 221 * 222 * Set/get/delete/rehash items 223 * 224 **********************************************************************/ 225 static inline struct pppox_sock *get_item(struct pppoe_net *pn, __be16 sid, 226 unsigned char *addr, int ifindex) 227 { 228 struct pppox_sock *po; 229 230 po = __get_item(pn, sid, addr, ifindex); 231 if (po && !refcount_inc_not_zero(&po->sk.sk_refcnt)) 232 po = NULL; 233 234 return po; 235 } 236 237 static void delete_item(struct pppoe_net *pn, const struct pppox_sock *po) 238 { 239 spin_lock(&pn->hash_lock); 240 __delete_item(pn, po); 241 spin_unlock(&pn->hash_lock); 242 } 243 244 /*************************************************************************** 245 * 246 * Handler for device events. 247 * Certain device events require that sockets be unconnected. 248 * 249 **************************************************************************/ 250 251 static void pppoe_flush_dev(struct net_device *dev) 252 { 253 struct pppoe_net *pn; 254 int i; 255 256 pn = pppoe_pernet(dev_net(dev)); 257 spin_lock(&pn->hash_lock); 258 for (i = 0; i < PPPOE_HASH_SIZE; i++) { 259 struct pppox_sock *po = rcu_dereference_protected(pn->hash_table[i], 260 lockdep_is_held(&pn->hash_lock)); 261 struct sock *sk; 262 263 while (po) { 264 while (po && po->pppoe_dev != dev) { 265 po = rcu_dereference_protected(po->next, 266 lockdep_is_held(&pn->hash_lock)); 267 } 268 269 if (!po) 270 break; 271 272 sk = &po->sk; 273 274 /* We always grab the socket lock, followed by the 275 * hash_lock, in that order. Since we should hold the 276 * sock lock while doing any unbinding, we need to 277 * release the lock we're holding. Hold a reference to 278 * the sock so it doesn't disappear as we're jumping 279 * between locks. 280 */ 281 282 sock_hold(sk); 283 spin_unlock(&pn->hash_lock); 284 lock_sock(sk); 285 286 if (po->pppoe_dev == dev && 287 sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND)) { 288 pppox_unbind_sock(sk); 289 sk->sk_state_change(sk); 290 po->pppoe_dev = NULL; 291 dev_put(dev); 292 } 293 294 release_sock(sk); 295 sock_put(sk); 296 297 /* Restart the process from the start of the current 298 * hash chain. We dropped locks so the world may have 299 * change from underneath us. 300 */ 301 302 BUG_ON(pppoe_pernet(dev_net(dev)) == NULL); 303 spin_lock(&pn->hash_lock); 304 po = rcu_dereference_protected(pn->hash_table[i], 305 lockdep_is_held(&pn->hash_lock)); 306 } 307 } 308 spin_unlock(&pn->hash_lock); 309 } 310 311 static int pppoe_device_event(struct notifier_block *this, 312 unsigned long event, void *ptr) 313 { 314 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 315 316 /* Only look at sockets that are using this specific device. */ 317 switch (event) { 318 case NETDEV_CHANGEADDR: 319 case NETDEV_CHANGEMTU: 320 /* A change in mtu or address is a bad thing, requiring 321 * LCP re-negotiation. 322 */ 323 324 case NETDEV_GOING_DOWN: 325 case NETDEV_DOWN: 326 /* Find every socket on this device and kill it. */ 327 pppoe_flush_dev(dev); 328 break; 329 330 default: 331 break; 332 } 333 334 return NOTIFY_DONE; 335 } 336 337 static struct notifier_block pppoe_notifier = { 338 .notifier_call = pppoe_device_event, 339 }; 340 341 /************************************************************************ 342 * 343 * Do the real work of receiving a PPPoE Session frame. 344 * 345 ***********************************************************************/ 346 static int pppoe_rcv_core(struct sock *sk, struct sk_buff *skb) 347 { 348 struct pppox_sock *po = pppox_sk(sk); 349 350 /* Backlog receive. Semantics of backlog rcv preclude any code from 351 * executing in lock_sock()/release_sock() bounds; meaning sk->sk_state 352 * can't change. 353 */ 354 355 if (sk->sk_state & PPPOX_BOUND) { 356 ppp_input(&po->chan, skb); 357 } else { 358 if (sock_queue_rcv_skb(sk, skb)) 359 goto abort_kfree; 360 } 361 362 return NET_RX_SUCCESS; 363 364 abort_kfree: 365 kfree_skb(skb); 366 return NET_RX_DROP; 367 } 368 369 /************************************************************************ 370 * 371 * Receive wrapper called in BH context. 372 * 373 ***********************************************************************/ 374 static int pppoe_rcv(struct sk_buff *skb, struct net_device *dev, 375 struct packet_type *pt, struct net_device *orig_dev) 376 { 377 struct pppoe_hdr *ph; 378 struct pppox_sock *po; 379 struct pppoe_net *pn; 380 int len; 381 382 if (skb->pkt_type == PACKET_OTHERHOST) 383 goto drop; 384 385 skb = skb_share_check(skb, GFP_ATOMIC); 386 if (!skb) 387 goto out; 388 389 if (skb_mac_header_len(skb) < ETH_HLEN) 390 goto drop; 391 392 if (!pskb_may_pull(skb, PPPOE_SES_HLEN)) 393 goto drop; 394 395 ph = pppoe_hdr(skb); 396 len = ntohs(ph->length); 397 398 skb_pull_rcsum(skb, sizeof(*ph)); 399 if (skb->len < len) 400 goto drop; 401 402 /* skb->data points to the PPP protocol header after skb_pull_rcsum. 403 * Drop PFC frames. 404 */ 405 if (ppp_skb_is_compressed_proto(skb)) 406 goto drop; 407 408 if (!skb_is_gso(skb) && pskb_trim_rcsum(skb, len)) 409 goto drop; 410 411 ph = pppoe_hdr(skb); 412 pn = pppoe_pernet(dev_net(dev)); 413 414 po = __get_item(pn, ph->sid, eth_hdr(skb)->h_source, dev->ifindex); 415 if (!po) 416 goto drop; 417 418 return __sk_receive_skb(&po->sk, skb, 0, 1, false); 419 420 drop: 421 kfree_skb(skb); 422 out: 423 return NET_RX_DROP; 424 } 425 426 static void pppoe_unbind_sock_work(struct work_struct *work) 427 { 428 struct pppox_sock *po = container_of(work, struct pppox_sock, 429 proto.pppoe.padt_work); 430 struct sock *sk = &po->sk; 431 432 lock_sock(sk); 433 if (po->pppoe_dev) { 434 dev_put(po->pppoe_dev); 435 po->pppoe_dev = NULL; 436 } 437 pppox_unbind_sock(sk); 438 release_sock(sk); 439 sock_put(sk); 440 } 441 442 /************************************************************************ 443 * 444 * Receive a PPPoE Discovery frame. 445 * This is solely for detection of PADT frames 446 * 447 ***********************************************************************/ 448 static int pppoe_disc_rcv(struct sk_buff *skb, struct net_device *dev, 449 struct packet_type *pt, struct net_device *orig_dev) 450 451 { 452 struct pppoe_hdr *ph; 453 struct pppox_sock *po; 454 struct pppoe_net *pn; 455 456 skb = skb_share_check(skb, GFP_ATOMIC); 457 if (!skb) 458 goto out; 459 460 if (skb->pkt_type != PACKET_HOST) 461 goto abort; 462 463 if (skb_mac_header_len(skb) < ETH_HLEN) 464 goto abort; 465 466 if (!pskb_may_pull(skb, sizeof(struct pppoe_hdr))) 467 goto abort; 468 469 ph = pppoe_hdr(skb); 470 if (ph->code != PADT_CODE) 471 goto abort; 472 473 pn = pppoe_pernet(dev_net(dev)); 474 po = get_item(pn, ph->sid, eth_hdr(skb)->h_source, dev->ifindex); 475 if (po) 476 if (!schedule_work(&po->proto.pppoe.padt_work)) 477 sock_put(&po->sk); 478 479 abort: 480 kfree_skb(skb); 481 out: 482 return NET_RX_SUCCESS; /* Lies... :-) */ 483 } 484 485 static struct packet_type pppoes_ptype __read_mostly = { 486 .type = cpu_to_be16(ETH_P_PPP_SES), 487 .func = pppoe_rcv, 488 }; 489 490 static struct packet_type pppoed_ptype __read_mostly = { 491 .type = cpu_to_be16(ETH_P_PPP_DISC), 492 .func = pppoe_disc_rcv, 493 }; 494 495 static struct proto pppoe_sk_proto __read_mostly = { 496 .name = "PPPOE", 497 .owner = THIS_MODULE, 498 .obj_size = sizeof(struct pppox_sock), 499 }; 500 501 static void pppoe_destruct(struct sock *sk) 502 { 503 skb_queue_purge(&sk->sk_receive_queue); 504 } 505 506 /*********************************************************************** 507 * 508 * Initialize a new struct sock. 509 * 510 **********************************************************************/ 511 static int pppoe_create(struct net *net, struct socket *sock, int kern) 512 { 513 struct sock *sk; 514 515 sk = sk_alloc(net, PF_PPPOX, GFP_KERNEL, &pppoe_sk_proto, kern); 516 if (!sk) 517 return -ENOMEM; 518 519 sock_init_data(sock, sk); 520 sock_set_flag(sk, SOCK_RCU_FREE); 521 522 sock->state = SS_UNCONNECTED; 523 sock->ops = &pppoe_ops; 524 525 sk->sk_backlog_rcv = pppoe_rcv_core; 526 sk->sk_destruct = pppoe_destruct; 527 sk->sk_state = PPPOX_NONE; 528 sk->sk_type = SOCK_STREAM; 529 sk->sk_family = PF_PPPOX; 530 sk->sk_protocol = PX_PROTO_OE; 531 532 INIT_WORK(&pppox_sk(sk)->proto.pppoe.padt_work, 533 pppoe_unbind_sock_work); 534 535 return 0; 536 } 537 538 static int pppoe_release(struct socket *sock) 539 { 540 struct sock *sk = sock->sk; 541 struct pppox_sock *po; 542 struct pppoe_net *pn; 543 struct net *net = NULL; 544 545 if (!sk) 546 return 0; 547 548 lock_sock(sk); 549 if (sock_flag(sk, SOCK_DEAD)) { 550 release_sock(sk); 551 return -EBADF; 552 } 553 554 po = pppox_sk(sk); 555 556 if (po->pppoe_dev) { 557 dev_put(po->pppoe_dev); 558 po->pppoe_dev = NULL; 559 } 560 561 pppox_unbind_sock(sk); 562 563 /* Signal the death of the socket. */ 564 sk->sk_state = PPPOX_DEAD; 565 566 net = sock_net(sk); 567 pn = pppoe_pernet(net); 568 569 /* 570 * protect "po" from concurrent updates 571 * on pppoe_flush_dev 572 */ 573 delete_item(pn, po); 574 575 sock_orphan(sk); 576 sock->sk = NULL; 577 578 release_sock(sk); 579 sock_put(sk); 580 581 return 0; 582 } 583 584 static int pppoe_connect(struct socket *sock, struct sockaddr_unsized *uservaddr, 585 int sockaddr_len, int flags) 586 { 587 struct sock *sk = sock->sk; 588 struct sockaddr_pppox *sp = (struct sockaddr_pppox *)uservaddr; 589 struct pppox_sock *po = pppox_sk(sk); 590 struct net_device *dev = NULL; 591 struct pppoe_net *pn; 592 struct net *net = NULL; 593 int error; 594 595 lock_sock(sk); 596 597 error = -EINVAL; 598 599 if (sockaddr_len != sizeof(struct sockaddr_pppox)) 600 goto end; 601 602 if (sp->sa_protocol != PX_PROTO_OE) 603 goto end; 604 605 /* Check for already bound sockets */ 606 error = -EBUSY; 607 if ((sk->sk_state & PPPOX_CONNECTED) && 608 stage_session(sp->sa_addr.pppoe.sid)) 609 goto end; 610 611 /* Check for already disconnected sockets, on attempts to disconnect */ 612 error = -EALREADY; 613 if ((sk->sk_state & PPPOX_DEAD) && 614 !stage_session(sp->sa_addr.pppoe.sid)) 615 goto end; 616 617 error = 0; 618 619 /* Delete the old binding */ 620 if (stage_session(po->pppoe_pa.sid)) { 621 pppox_unbind_sock(sk); 622 pn = pppoe_pernet(sock_net(sk)); 623 delete_item(pn, po); 624 625 /* wait for concurrent readers before clearing the fields */ 626 synchronize_net(); 627 628 if (po->pppoe_dev) { 629 dev_put(po->pppoe_dev); 630 po->pppoe_dev = NULL; 631 } 632 633 po->pppoe_ifindex = 0; 634 memset(&po->pppoe_pa, 0, sizeof(po->pppoe_pa)); 635 memset(&po->chan, 0, sizeof(po->chan)); 636 po->next = NULL; 637 po->num = 0; 638 639 sk->sk_state = PPPOX_NONE; 640 } 641 642 /* Re-bind in session stage only */ 643 if (stage_session(sp->sa_addr.pppoe.sid)) { 644 error = -ENODEV; 645 net = sock_net(sk); 646 dev = dev_get_by_name(net, sp->sa_addr.pppoe.dev); 647 if (!dev) 648 goto err_put; 649 650 po->pppoe_dev = dev; 651 po->pppoe_ifindex = dev->ifindex; 652 pn = pppoe_pernet(net); 653 if (!(dev->flags & IFF_UP)) { 654 goto err_put; 655 } 656 657 memcpy(&po->pppoe_pa, 658 &sp->sa_addr.pppoe, 659 sizeof(struct pppoe_addr)); 660 661 spin_lock(&pn->hash_lock); 662 error = __set_item(pn, po); 663 spin_unlock(&pn->hash_lock); 664 if (error < 0) 665 goto err_put; 666 667 po->chan.hdrlen = (sizeof(struct pppoe_hdr) + 668 dev->hard_header_len); 669 670 po->chan.mtu = dev->mtu - sizeof(struct pppoe_hdr) - 2; 671 po->chan.private = sk; 672 po->chan.ops = &pppoe_chan_ops; 673 po->chan.direct_xmit = true; 674 675 error = ppp_register_net_channel(dev_net(dev), &po->chan); 676 if (error) { 677 delete_item(pn, po); 678 goto err_put; 679 } 680 681 sk->sk_state = PPPOX_CONNECTED; 682 } 683 684 po->num = sp->sa_addr.pppoe.sid; 685 686 end: 687 release_sock(sk); 688 return error; 689 err_put: 690 if (po->pppoe_dev) { 691 dev_put(po->pppoe_dev); 692 po->pppoe_dev = NULL; 693 } 694 goto end; 695 } 696 697 static int pppoe_getname(struct socket *sock, struct sockaddr *uaddr, 698 int peer) 699 { 700 int len = sizeof(struct sockaddr_pppox); 701 struct sockaddr_pppox sp; 702 703 sp.sa_family = AF_PPPOX; 704 sp.sa_protocol = PX_PROTO_OE; 705 memcpy(&sp.sa_addr.pppoe, &pppox_sk(sock->sk)->pppoe_pa, 706 sizeof(struct pppoe_addr)); 707 708 memcpy(uaddr, &sp, len); 709 710 return len; 711 } 712 713 static int pppoe_ioctl(struct socket *sock, unsigned int cmd, 714 unsigned long arg) 715 { 716 struct sock *sk = sock->sk; 717 struct pppox_sock *po = pppox_sk(sk); 718 int val; 719 int err; 720 721 switch (cmd) { 722 case PPPIOCGMRU: 723 err = -ENXIO; 724 if (!(sk->sk_state & PPPOX_CONNECTED)) 725 break; 726 727 err = -EFAULT; 728 if (put_user(po->pppoe_dev->mtu - 729 sizeof(struct pppoe_hdr) - 730 PPP_HDRLEN, 731 (int __user *)arg)) 732 break; 733 err = 0; 734 break; 735 736 case PPPIOCSMRU: 737 err = -ENXIO; 738 if (!(sk->sk_state & PPPOX_CONNECTED)) 739 break; 740 741 err = -EFAULT; 742 if (get_user(val, (int __user *)arg)) 743 break; 744 745 if (val < (po->pppoe_dev->mtu 746 - sizeof(struct pppoe_hdr) 747 - PPP_HDRLEN)) 748 err = 0; 749 else 750 err = -EINVAL; 751 break; 752 753 case PPPIOCSFLAGS: 754 err = -EFAULT; 755 if (get_user(val, (int __user *)arg)) 756 break; 757 err = 0; 758 break; 759 760 default: 761 err = -ENOTTY; 762 } 763 764 return err; 765 } 766 767 static int pppoe_sendmsg(struct socket *sock, struct msghdr *m, 768 size_t total_len) 769 { 770 struct sk_buff *skb; 771 struct sock *sk = sock->sk; 772 struct pppox_sock *po = pppox_sk(sk); 773 int error; 774 struct pppoe_hdr hdr; 775 struct pppoe_hdr *ph; 776 struct net_device *dev; 777 char *start; 778 int hlen; 779 780 lock_sock(sk); 781 if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED)) { 782 error = -ENOTCONN; 783 goto end; 784 } 785 786 hdr.ver = 1; 787 hdr.type = 1; 788 hdr.code = 0; 789 hdr.sid = po->num; 790 791 dev = po->pppoe_dev; 792 793 error = -EMSGSIZE; 794 if (total_len > (dev->mtu + dev->hard_header_len)) 795 goto end; 796 797 hlen = LL_RESERVED_SPACE(dev); 798 skb = sock_wmalloc(sk, hlen + sizeof(*ph) + total_len + 799 dev->needed_tailroom, 0, GFP_KERNEL); 800 if (!skb) { 801 error = -ENOMEM; 802 goto end; 803 } 804 805 /* Reserve space for headers. */ 806 skb_reserve(skb, hlen); 807 skb_reset_network_header(skb); 808 809 skb->dev = dev; 810 811 skb->priority = READ_ONCE(sk->sk_priority); 812 skb->protocol = cpu_to_be16(ETH_P_PPP_SES); 813 814 ph = skb_put(skb, total_len + sizeof(struct pppoe_hdr)); 815 start = (char *)ph + sizeof(*ph); 816 817 error = memcpy_from_msg(start, m, total_len); 818 if (error < 0) { 819 kfree_skb(skb); 820 goto end; 821 } 822 823 error = total_len; 824 dev_hard_header(skb, dev, ETH_P_PPP_SES, 825 po->pppoe_pa.remote, NULL, total_len); 826 827 ph = pppoe_hdr(skb); 828 memcpy(ph, &hdr, sizeof(struct pppoe_hdr)); 829 830 ph->length = htons(total_len); 831 832 dev_queue_xmit(skb); 833 834 end: 835 release_sock(sk); 836 return error; 837 } 838 839 /************************************************************************ 840 * 841 * xmit function called by generic PPP driver 842 * sends PPP frame over PPPoE socket 843 * 844 ***********************************************************************/ 845 static int pppoe_xmit(struct ppp_channel *chan, struct sk_buff *skb) 846 { 847 struct sock *sk = chan->private; 848 struct pppox_sock *po = pppox_sk(sk); 849 struct net_device *dev = po->pppoe_dev; 850 struct pppoe_hdr *ph; 851 int data_len = skb->len; 852 853 /* The higher-level PPP code (ppp_unregister_channel()) ensures the PPP 854 * xmit operations conclude prior to an unregistration call. Thus 855 * sk->sk_state cannot change, so we don't need to do lock_sock(). 856 * But, we also can't do a lock_sock since that introduces a potential 857 * deadlock as we'd reverse the lock ordering used when calling 858 * ppp_unregister_channel(). 859 */ 860 861 if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED)) 862 goto abort; 863 864 if (!dev) 865 goto abort; 866 867 /* Copy the data if there is no space for the header or if it's 868 * read-only. 869 */ 870 if (skb_cow_head(skb, LL_RESERVED_SPACE(dev) + sizeof(*ph))) 871 goto abort; 872 873 __skb_push(skb, sizeof(*ph)); 874 skb_reset_network_header(skb); 875 876 ph = pppoe_hdr(skb); 877 ph->ver = 1; 878 ph->type = 1; 879 ph->code = 0; 880 ph->sid = po->num; 881 ph->length = htons(data_len); 882 883 skb->protocol = cpu_to_be16(ETH_P_PPP_SES); 884 skb->dev = dev; 885 886 dev_hard_header(skb, dev, ETH_P_PPP_SES, 887 po->pppoe_pa.remote, NULL, data_len); 888 889 dev_queue_xmit(skb); 890 return 1; 891 892 abort: 893 kfree_skb(skb); 894 return 1; 895 } 896 897 static int pppoe_fill_forward_path(struct net_device_path_ctx *ctx, 898 struct net_device_path *path, 899 const struct ppp_channel *chan) 900 { 901 struct sock *sk = chan->private; 902 struct pppox_sock *po = pppox_sk(sk); 903 struct net_device *dev = po->pppoe_dev; 904 905 if (sock_flag(sk, SOCK_DEAD) || 906 !(sk->sk_state & PPPOX_CONNECTED) || !dev) 907 return -1; 908 909 path->type = DEV_PATH_PPPOE; 910 path->encap.proto = htons(ETH_P_PPP_SES); 911 path->encap.id = be16_to_cpu(po->num); 912 memcpy(path->encap.h_dest, po->pppoe_pa.remote, ETH_ALEN); 913 memcpy(ctx->daddr, po->pppoe_pa.remote, ETH_ALEN); 914 path->dev = ctx->dev; 915 ctx->dev = dev; 916 917 return 0; 918 } 919 920 static const struct ppp_channel_ops pppoe_chan_ops = { 921 .start_xmit = pppoe_xmit, 922 .fill_forward_path = pppoe_fill_forward_path, 923 }; 924 925 static int pppoe_recvmsg(struct socket *sock, struct msghdr *m, 926 size_t total_len, int flags) 927 { 928 struct sock *sk = sock->sk; 929 struct sk_buff *skb; 930 int error = 0; 931 932 if (sk->sk_state & PPPOX_BOUND) 933 return -EIO; 934 935 skb = skb_recv_datagram(sk, flags, &error); 936 if (!skb) 937 return error; 938 939 total_len = min_t(size_t, total_len, skb->len); 940 error = skb_copy_datagram_msg(skb, 0, m, total_len); 941 if (error == 0) { 942 consume_skb(skb); 943 return total_len; 944 } 945 946 kfree_skb(skb); 947 return error; 948 } 949 950 #ifdef CONFIG_PROC_FS 951 static int pppoe_seq_show(struct seq_file *seq, void *v) 952 { 953 struct pppox_sock *po; 954 char *dev_name; 955 956 if (v == SEQ_START_TOKEN) { 957 seq_puts(seq, "Id Address Device\n"); 958 goto out; 959 } 960 961 po = v; 962 dev_name = po->pppoe_pa.dev; 963 964 seq_printf(seq, "%08X %pM %8s\n", 965 po->pppoe_pa.sid, po->pppoe_pa.remote, dev_name); 966 out: 967 return 0; 968 } 969 970 static inline struct pppox_sock *pppoe_get_idx(struct pppoe_net *pn, loff_t pos) 971 { 972 struct pppox_sock *po; 973 int i; 974 975 for (i = 0; i < PPPOE_HASH_SIZE; i++) { 976 po = rcu_dereference(pn->hash_table[i]); 977 while (po) { 978 if (!pos--) 979 goto out; 980 po = rcu_dereference(po->next); 981 } 982 } 983 984 out: 985 return po; 986 } 987 988 static void *pppoe_seq_start(struct seq_file *seq, loff_t *pos) 989 __acquires(RCU) 990 { 991 struct pppoe_net *pn = pppoe_pernet(seq_file_net(seq)); 992 loff_t l = *pos; 993 994 rcu_read_lock(); 995 return l ? pppoe_get_idx(pn, --l) : SEQ_START_TOKEN; 996 } 997 998 static void *pppoe_seq_next(struct seq_file *seq, void *v, loff_t *pos) 999 { 1000 struct pppoe_net *pn = pppoe_pernet(seq_file_net(seq)); 1001 struct pppox_sock *po, *next; 1002 1003 ++*pos; 1004 if (v == SEQ_START_TOKEN) { 1005 po = pppoe_get_idx(pn, 0); 1006 goto out; 1007 } 1008 po = v; 1009 next = rcu_dereference(po->next); 1010 if (next) 1011 po = next; 1012 else { 1013 int hash = hash_item(po->pppoe_pa.sid, po->pppoe_pa.remote); 1014 1015 po = NULL; 1016 while (++hash < PPPOE_HASH_SIZE) { 1017 po = rcu_dereference(pn->hash_table[hash]); 1018 if (po) 1019 break; 1020 } 1021 } 1022 1023 out: 1024 return po; 1025 } 1026 1027 static void pppoe_seq_stop(struct seq_file *seq, void *v) 1028 __releases(RCU) 1029 { 1030 rcu_read_unlock(); 1031 } 1032 1033 static const struct seq_operations pppoe_seq_ops = { 1034 .start = pppoe_seq_start, 1035 .next = pppoe_seq_next, 1036 .stop = pppoe_seq_stop, 1037 .show = pppoe_seq_show, 1038 }; 1039 #endif /* CONFIG_PROC_FS */ 1040 1041 static const struct proto_ops pppoe_ops = { 1042 .family = AF_PPPOX, 1043 .owner = THIS_MODULE, 1044 .release = pppoe_release, 1045 .bind = sock_no_bind, 1046 .connect = pppoe_connect, 1047 .socketpair = sock_no_socketpair, 1048 .accept = sock_no_accept, 1049 .getname = pppoe_getname, 1050 .poll = datagram_poll, 1051 .listen = sock_no_listen, 1052 .shutdown = sock_no_shutdown, 1053 .sendmsg = pppoe_sendmsg, 1054 .recvmsg = pppoe_recvmsg, 1055 .mmap = sock_no_mmap, 1056 .ioctl = pppox_ioctl, 1057 #ifdef CONFIG_COMPAT 1058 .compat_ioctl = pppox_compat_ioctl, 1059 #endif 1060 }; 1061 1062 static const struct pppox_proto pppoe_proto = { 1063 .create = pppoe_create, 1064 .ioctl = pppoe_ioctl, 1065 .owner = THIS_MODULE, 1066 }; 1067 1068 static __net_init int pppoe_init_net(struct net *net) 1069 { 1070 struct pppoe_net *pn = pppoe_pernet(net); 1071 struct proc_dir_entry *pde; 1072 1073 spin_lock_init(&pn->hash_lock); 1074 1075 pde = proc_create_net("pppoe", 0444, net->proc_net, 1076 &pppoe_seq_ops, sizeof(struct seq_net_private)); 1077 #ifdef CONFIG_PROC_FS 1078 if (!pde) 1079 return -ENOMEM; 1080 #endif 1081 1082 return 0; 1083 } 1084 1085 static __net_exit void pppoe_exit_net(struct net *net) 1086 { 1087 remove_proc_entry("pppoe", net->proc_net); 1088 } 1089 1090 static struct pernet_operations pppoe_net_ops = { 1091 .init = pppoe_init_net, 1092 .exit = pppoe_exit_net, 1093 .id = &pppoe_net_id, 1094 .size = sizeof(struct pppoe_net), 1095 }; 1096 1097 static u16 1098 compare_pppoe_header(const struct pppoe_hdr *phdr, 1099 const struct pppoe_hdr *phdr2) 1100 { 1101 __be16 proto = *(const __be16 *)(phdr + 1); 1102 __be16 proto2 = *(const __be16 *)(phdr2 + 1); 1103 1104 return (__force u16)((phdr->sid ^ phdr2->sid) | (proto ^ proto2)); 1105 } 1106 1107 static __be16 pppoe_hdr_proto(const struct pppoe_hdr *phdr) 1108 { 1109 __be16 proto = *(const __be16 *)(phdr + 1); 1110 1111 switch (proto) { 1112 case cpu_to_be16(PPP_IP): 1113 return cpu_to_be16(ETH_P_IP); 1114 #if IS_ENABLED(CONFIG_IPV6) 1115 case cpu_to_be16(PPP_IPV6): 1116 return cpu_to_be16(ETH_P_IPV6); 1117 #endif 1118 default: 1119 return 0; 1120 } 1121 } 1122 1123 static struct sk_buff *pppoe_gro_receive(struct list_head *head, 1124 struct sk_buff *skb) 1125 { 1126 const struct packet_offload *ptype; 1127 unsigned int hlen, off_pppoe; 1128 const struct pppoe_hdr *phdr; 1129 struct sk_buff *pp = NULL; 1130 struct sk_buff *p; 1131 int flush = 1; 1132 __be16 type; 1133 1134 off_pppoe = skb_gro_offset(skb); 1135 hlen = off_pppoe + PPPOE_SES_HLEN; 1136 phdr = skb_gro_header(skb, hlen, off_pppoe); 1137 if (unlikely(!phdr)) 1138 goto out; 1139 1140 /* filter for session packets (type:1, ver:1, code:0) */ 1141 if (*(const __be16 *)phdr != cpu_to_be16(0x1100)) 1142 goto out; 1143 1144 /* ignore packets with padding or invalid length */ 1145 if (skb_gro_len(skb) != be16_to_cpu(phdr->length) + sizeof(*phdr)) 1146 goto out; 1147 1148 type = pppoe_hdr_proto(phdr); 1149 ptype = gro_find_receive_by_type(type); 1150 if (!ptype) 1151 goto out; 1152 1153 flush = 0; 1154 1155 list_for_each_entry(p, head, list) { 1156 const struct pppoe_hdr *phdr2; 1157 1158 if (!NAPI_GRO_CB(p)->same_flow) 1159 continue; 1160 1161 phdr2 = (const struct pppoe_hdr *)(p->data + off_pppoe); 1162 if (compare_pppoe_header(phdr, phdr2)) 1163 NAPI_GRO_CB(p)->same_flow = 0; 1164 } 1165 1166 skb_gro_pull(skb, PPPOE_SES_HLEN); 1167 skb_gro_postpull_rcsum(skb, phdr, PPPOE_SES_HLEN); 1168 1169 pp = indirect_call_gro_receive_inet(ptype->callbacks.gro_receive, 1170 ipv6_gro_receive, inet_gro_receive, 1171 head, skb); 1172 1173 out: 1174 skb_gro_flush_final(skb, pp, flush); 1175 1176 return pp; 1177 } 1178 1179 static int pppoe_gro_complete(struct sk_buff *skb, int nhoff) 1180 { 1181 struct pppoe_hdr *phdr = (struct pppoe_hdr *)(skb->data + nhoff); 1182 __be16 type = pppoe_hdr_proto(phdr); 1183 struct packet_offload *ptype; 1184 unsigned int len; 1185 1186 ptype = gro_find_complete_by_type(type); 1187 if (!ptype) 1188 return -ENOENT; 1189 1190 len = skb->len - (nhoff + sizeof(*phdr)); 1191 len = min(len, 0xFFFFU); 1192 phdr->length = cpu_to_be16(len); 1193 1194 return INDIRECT_CALL_INET(ptype->callbacks.gro_complete, 1195 ipv6_gro_complete, inet_gro_complete, 1196 skb, nhoff + PPPOE_SES_HLEN); 1197 } 1198 1199 static struct sk_buff *pppoe_gso_segment(struct sk_buff *skb, 1200 netdev_features_t features) 1201 { 1202 struct sk_buff *segs = ERR_PTR(-EINVAL); 1203 struct packet_offload *ptype; 1204 struct pppoe_hdr *phdr; 1205 __be16 orig_type, type; 1206 int len, nhoff; 1207 1208 skb_reset_network_header(skb); 1209 nhoff = skb_network_header(skb) - skb_mac_header(skb); 1210 1211 if (unlikely(!pskb_may_pull(skb, PPPOE_SES_HLEN))) 1212 goto out; 1213 1214 phdr = (struct pppoe_hdr *)skb_network_header(skb); 1215 type = pppoe_hdr_proto(phdr); 1216 ptype = gro_find_complete_by_type(type); 1217 if (!ptype) 1218 goto out; 1219 1220 orig_type = skb->protocol; 1221 __skb_pull(skb, PPPOE_SES_HLEN); 1222 features &= ~NETIF_F_GSO_SOFTWARE; 1223 segs = ptype->callbacks.gso_segment(skb, features); 1224 if (IS_ERR_OR_NULL(segs)) 1225 goto out; 1226 1227 skb = segs; 1228 do { 1229 phdr = (struct pppoe_hdr *)(skb_mac_header(skb) + nhoff); 1230 len = skb->len - (nhoff + sizeof(*phdr)); 1231 phdr->length = cpu_to_be16(len); 1232 skb->network_header = (u8 *)phdr - skb->head; 1233 skb->protocol = orig_type; 1234 skb_reset_mac_len(skb); 1235 } while ((skb = skb->next)); 1236 1237 out: 1238 return segs; 1239 } 1240 1241 static struct packet_offload pppoe_packet_offload __read_mostly = { 1242 .type = cpu_to_be16(ETH_P_PPP_SES), 1243 .priority = 20, 1244 .callbacks = { 1245 .gro_receive = pppoe_gro_receive, 1246 .gro_complete = pppoe_gro_complete, 1247 .gso_segment = pppoe_gso_segment, 1248 }, 1249 }; 1250 1251 static int __init pppoe_init(void) 1252 { 1253 int err; 1254 1255 err = register_pernet_device(&pppoe_net_ops); 1256 if (err) 1257 goto out; 1258 1259 err = proto_register(&pppoe_sk_proto, 0); 1260 if (err) 1261 goto out_unregister_net_ops; 1262 1263 err = register_pppox_proto(PX_PROTO_OE, &pppoe_proto); 1264 if (err) 1265 goto out_unregister_pppoe_proto; 1266 1267 if (IS_ENABLED(CONFIG_INET)) 1268 dev_add_offload(&pppoe_packet_offload); 1269 dev_add_pack(&pppoes_ptype); 1270 dev_add_pack(&pppoed_ptype); 1271 register_netdevice_notifier(&pppoe_notifier); 1272 1273 return 0; 1274 1275 out_unregister_pppoe_proto: 1276 proto_unregister(&pppoe_sk_proto); 1277 out_unregister_net_ops: 1278 unregister_pernet_device(&pppoe_net_ops); 1279 out: 1280 return err; 1281 } 1282 1283 static void __exit pppoe_exit(void) 1284 { 1285 unregister_netdevice_notifier(&pppoe_notifier); 1286 dev_remove_pack(&pppoed_ptype); 1287 dev_remove_pack(&pppoes_ptype); 1288 if (IS_ENABLED(CONFIG_INET)) 1289 dev_remove_offload(&pppoe_packet_offload); 1290 unregister_pppox_proto(PX_PROTO_OE); 1291 proto_unregister(&pppoe_sk_proto); 1292 unregister_pernet_device(&pppoe_net_ops); 1293 } 1294 1295 module_init(pppoe_init); 1296 module_exit(pppoe_exit); 1297 1298 MODULE_AUTHOR("Michal Ostrowski <mostrows@speakeasy.net>"); 1299 MODULE_DESCRIPTION("PPP over Ethernet driver"); 1300 MODULE_LICENSE("GPL"); 1301 MODULE_ALIAS_NET_PF_PROTO(PF_PPPOX, PX_PROTO_OE); 1302