1 /* 2 * NETLINK Kernel-user communication protocol. 3 * 4 * Authors: Alan Cox <alan@redhat.com> 5 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 6 * 7 * This program is free software; you can redistribute it and/or 8 * modify it under the terms of the GNU General Public License 9 * as published by the Free Software Foundation; either version 10 * 2 of the License, or (at your option) any later version. 11 * 12 * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith 13 * added netlink_proto_exit 14 * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> 15 * use nlk_sk, as sk->protinfo is on a diet 8) 16 * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> 17 * - inc module use count of module that owns 18 * the kernel socket in case userspace opens 19 * socket of same protocol 20 * - remove all module support, since netlink is 21 * mandatory if CONFIG_NET=y these days 22 */ 23 24 #include <linux/config.h> 25 #include <linux/module.h> 26 27 #include <linux/kernel.h> 28 #include <linux/init.h> 29 #include <linux/signal.h> 30 #include <linux/sched.h> 31 #include <linux/errno.h> 32 #include <linux/string.h> 33 #include <linux/stat.h> 34 #include <linux/socket.h> 35 #include <linux/un.h> 36 #include <linux/fcntl.h> 37 #include <linux/termios.h> 38 #include <linux/sockios.h> 39 #include <linux/net.h> 40 #include <linux/fs.h> 41 #include <linux/slab.h> 42 #include <asm/uaccess.h> 43 #include <linux/skbuff.h> 44 #include <linux/netdevice.h> 45 #include <linux/rtnetlink.h> 46 #include <linux/proc_fs.h> 47 #include <linux/seq_file.h> 48 #include <linux/smp_lock.h> 49 #include <linux/notifier.h> 50 #include <linux/security.h> 51 #include <linux/jhash.h> 52 #include <linux/jiffies.h> 53 #include <linux/random.h> 54 #include <linux/bitops.h> 55 #include <linux/mm.h> 56 #include <linux/types.h> 57 #include <linux/audit.h> 58 59 #include <net/sock.h> 60 #include <net/scm.h> 61 62 #define Nprintk(a...) 63 #define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8) 64 65 struct netlink_sock { 66 /* struct sock has to be the first member of netlink_sock */ 67 struct sock sk; 68 u32 pid; 69 u32 dst_pid; 70 u32 dst_group; 71 u32 flags; 72 u32 subscriptions; 73 u32 ngroups; 74 unsigned long *groups; 75 unsigned long state; 76 wait_queue_head_t wait; 77 struct netlink_callback *cb; 78 spinlock_t cb_lock; 79 void (*data_ready)(struct sock *sk, int bytes); 80 struct module *module; 81 }; 82 83 #define NETLINK_KERNEL_SOCKET 0x1 84 #define NETLINK_RECV_PKTINFO 0x2 85 86 static inline struct netlink_sock *nlk_sk(struct sock *sk) 87 { 88 return (struct netlink_sock *)sk; 89 } 90 91 struct nl_pid_hash { 92 struct hlist_head *table; 93 unsigned long rehash_time; 94 95 unsigned int mask; 96 unsigned int shift; 97 98 unsigned int entries; 99 unsigned int max_shift; 100 101 u32 rnd; 102 }; 103 104 struct netlink_table { 105 struct nl_pid_hash hash; 106 struct hlist_head mc_list; 107 unsigned int nl_nonroot; 108 unsigned int groups; 109 struct module *module; 110 int registered; 111 }; 112 113 static struct netlink_table *nl_table; 114 115 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); 116 117 static int netlink_dump(struct sock *sk); 118 static void netlink_destroy_callback(struct netlink_callback *cb); 119 120 static DEFINE_RWLOCK(nl_table_lock); 121 static atomic_t nl_table_users = ATOMIC_INIT(0); 122 123 static struct notifier_block *netlink_chain; 124 125 static u32 netlink_group_mask(u32 group) 126 { 127 return group ? 1 << (group - 1) : 0; 128 } 129 130 static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid) 131 { 132 return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask]; 133 } 134 135 static void netlink_sock_destruct(struct sock *sk) 136 { 137 skb_queue_purge(&sk->sk_receive_queue); 138 139 if (!sock_flag(sk, SOCK_DEAD)) { 140 printk("Freeing alive netlink socket %p\n", sk); 141 return; 142 } 143 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc)); 144 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc)); 145 BUG_TRAP(!nlk_sk(sk)->cb); 146 BUG_TRAP(!nlk_sk(sk)->groups); 147 } 148 149 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP. 150 * Look, when several writers sleep and reader wakes them up, all but one 151 * immediately hit write lock and grab all the cpus. Exclusive sleep solves 152 * this, _but_ remember, it adds useless work on UP machines. 153 */ 154 155 static void netlink_table_grab(void) 156 { 157 write_lock_bh(&nl_table_lock); 158 159 if (atomic_read(&nl_table_users)) { 160 DECLARE_WAITQUEUE(wait, current); 161 162 add_wait_queue_exclusive(&nl_table_wait, &wait); 163 for(;;) { 164 set_current_state(TASK_UNINTERRUPTIBLE); 165 if (atomic_read(&nl_table_users) == 0) 166 break; 167 write_unlock_bh(&nl_table_lock); 168 schedule(); 169 write_lock_bh(&nl_table_lock); 170 } 171 172 __set_current_state(TASK_RUNNING); 173 remove_wait_queue(&nl_table_wait, &wait); 174 } 175 } 176 177 static __inline__ void netlink_table_ungrab(void) 178 { 179 write_unlock_bh(&nl_table_lock); 180 wake_up(&nl_table_wait); 181 } 182 183 static __inline__ void 184 netlink_lock_table(void) 185 { 186 /* read_lock() synchronizes us to netlink_table_grab */ 187 188 read_lock(&nl_table_lock); 189 atomic_inc(&nl_table_users); 190 read_unlock(&nl_table_lock); 191 } 192 193 static __inline__ void 194 netlink_unlock_table(void) 195 { 196 if (atomic_dec_and_test(&nl_table_users)) 197 wake_up(&nl_table_wait); 198 } 199 200 static __inline__ struct sock *netlink_lookup(int protocol, u32 pid) 201 { 202 struct nl_pid_hash *hash = &nl_table[protocol].hash; 203 struct hlist_head *head; 204 struct sock *sk; 205 struct hlist_node *node; 206 207 read_lock(&nl_table_lock); 208 head = nl_pid_hashfn(hash, pid); 209 sk_for_each(sk, node, head) { 210 if (nlk_sk(sk)->pid == pid) { 211 sock_hold(sk); 212 goto found; 213 } 214 } 215 sk = NULL; 216 found: 217 read_unlock(&nl_table_lock); 218 return sk; 219 } 220 221 static inline struct hlist_head *nl_pid_hash_alloc(size_t size) 222 { 223 if (size <= PAGE_SIZE) 224 return kmalloc(size, GFP_ATOMIC); 225 else 226 return (struct hlist_head *) 227 __get_free_pages(GFP_ATOMIC, get_order(size)); 228 } 229 230 static inline void nl_pid_hash_free(struct hlist_head *table, size_t size) 231 { 232 if (size <= PAGE_SIZE) 233 kfree(table); 234 else 235 free_pages((unsigned long)table, get_order(size)); 236 } 237 238 static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow) 239 { 240 unsigned int omask, mask, shift; 241 size_t osize, size; 242 struct hlist_head *otable, *table; 243 int i; 244 245 omask = mask = hash->mask; 246 osize = size = (mask + 1) * sizeof(*table); 247 shift = hash->shift; 248 249 if (grow) { 250 if (++shift > hash->max_shift) 251 return 0; 252 mask = mask * 2 + 1; 253 size *= 2; 254 } 255 256 table = nl_pid_hash_alloc(size); 257 if (!table) 258 return 0; 259 260 memset(table, 0, size); 261 otable = hash->table; 262 hash->table = table; 263 hash->mask = mask; 264 hash->shift = shift; 265 get_random_bytes(&hash->rnd, sizeof(hash->rnd)); 266 267 for (i = 0; i <= omask; i++) { 268 struct sock *sk; 269 struct hlist_node *node, *tmp; 270 271 sk_for_each_safe(sk, node, tmp, &otable[i]) 272 __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid)); 273 } 274 275 nl_pid_hash_free(otable, osize); 276 hash->rehash_time = jiffies + 10 * 60 * HZ; 277 return 1; 278 } 279 280 static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len) 281 { 282 int avg = hash->entries >> hash->shift; 283 284 if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1)) 285 return 1; 286 287 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { 288 nl_pid_hash_rehash(hash, 0); 289 return 1; 290 } 291 292 return 0; 293 } 294 295 static struct proto_ops netlink_ops; 296 297 static int netlink_insert(struct sock *sk, u32 pid) 298 { 299 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; 300 struct hlist_head *head; 301 int err = -EADDRINUSE; 302 struct sock *osk; 303 struct hlist_node *node; 304 int len; 305 306 netlink_table_grab(); 307 head = nl_pid_hashfn(hash, pid); 308 len = 0; 309 sk_for_each(osk, node, head) { 310 if (nlk_sk(osk)->pid == pid) 311 break; 312 len++; 313 } 314 if (node) 315 goto err; 316 317 err = -EBUSY; 318 if (nlk_sk(sk)->pid) 319 goto err; 320 321 err = -ENOMEM; 322 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) 323 goto err; 324 325 if (len && nl_pid_hash_dilute(hash, len)) 326 head = nl_pid_hashfn(hash, pid); 327 hash->entries++; 328 nlk_sk(sk)->pid = pid; 329 sk_add_node(sk, head); 330 err = 0; 331 332 err: 333 netlink_table_ungrab(); 334 return err; 335 } 336 337 static void netlink_remove(struct sock *sk) 338 { 339 netlink_table_grab(); 340 if (sk_del_node_init(sk)) 341 nl_table[sk->sk_protocol].hash.entries--; 342 if (nlk_sk(sk)->subscriptions) 343 __sk_del_bind_node(sk); 344 netlink_table_ungrab(); 345 } 346 347 static struct proto netlink_proto = { 348 .name = "NETLINK", 349 .owner = THIS_MODULE, 350 .obj_size = sizeof(struct netlink_sock), 351 }; 352 353 static int __netlink_create(struct socket *sock, int protocol) 354 { 355 struct sock *sk; 356 struct netlink_sock *nlk; 357 358 sock->ops = &netlink_ops; 359 360 sk = sk_alloc(PF_NETLINK, GFP_KERNEL, &netlink_proto, 1); 361 if (!sk) 362 return -ENOMEM; 363 364 sock_init_data(sock, sk); 365 366 nlk = nlk_sk(sk); 367 spin_lock_init(&nlk->cb_lock); 368 init_waitqueue_head(&nlk->wait); 369 370 sk->sk_destruct = netlink_sock_destruct; 371 sk->sk_protocol = protocol; 372 return 0; 373 } 374 375 static int netlink_create(struct socket *sock, int protocol) 376 { 377 struct module *module = NULL; 378 struct netlink_sock *nlk; 379 unsigned int groups; 380 int err = 0; 381 382 sock->state = SS_UNCONNECTED; 383 384 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) 385 return -ESOCKTNOSUPPORT; 386 387 if (protocol<0 || protocol >= MAX_LINKS) 388 return -EPROTONOSUPPORT; 389 390 netlink_lock_table(); 391 #ifdef CONFIG_KMOD 392 if (!nl_table[protocol].registered) { 393 netlink_unlock_table(); 394 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); 395 netlink_lock_table(); 396 } 397 #endif 398 if (nl_table[protocol].registered && 399 try_module_get(nl_table[protocol].module)) 400 module = nl_table[protocol].module; 401 groups = nl_table[protocol].groups; 402 netlink_unlock_table(); 403 404 if ((err = __netlink_create(sock, protocol) < 0)) 405 goto out_module; 406 407 nlk = nlk_sk(sock->sk); 408 nlk->module = module; 409 out: 410 return err; 411 412 out_module: 413 module_put(module); 414 goto out; 415 } 416 417 static int netlink_release(struct socket *sock) 418 { 419 struct sock *sk = sock->sk; 420 struct netlink_sock *nlk; 421 422 if (!sk) 423 return 0; 424 425 netlink_remove(sk); 426 nlk = nlk_sk(sk); 427 428 spin_lock(&nlk->cb_lock); 429 if (nlk->cb) { 430 nlk->cb->done(nlk->cb); 431 netlink_destroy_callback(nlk->cb); 432 nlk->cb = NULL; 433 } 434 spin_unlock(&nlk->cb_lock); 435 436 /* OK. Socket is unlinked, and, therefore, 437 no new packets will arrive */ 438 439 sock_orphan(sk); 440 sock->sk = NULL; 441 wake_up_interruptible_all(&nlk->wait); 442 443 skb_queue_purge(&sk->sk_write_queue); 444 445 if (nlk->pid && !nlk->subscriptions) { 446 struct netlink_notify n = { 447 .protocol = sk->sk_protocol, 448 .pid = nlk->pid, 449 }; 450 notifier_call_chain(&netlink_chain, NETLINK_URELEASE, &n); 451 } 452 453 if (nlk->module) 454 module_put(nlk->module); 455 456 if (nlk->flags & NETLINK_KERNEL_SOCKET) { 457 netlink_table_grab(); 458 nl_table[sk->sk_protocol].module = NULL; 459 nl_table[sk->sk_protocol].registered = 0; 460 netlink_table_ungrab(); 461 } 462 463 kfree(nlk->groups); 464 nlk->groups = NULL; 465 466 sock_put(sk); 467 return 0; 468 } 469 470 static int netlink_autobind(struct socket *sock) 471 { 472 struct sock *sk = sock->sk; 473 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; 474 struct hlist_head *head; 475 struct sock *osk; 476 struct hlist_node *node; 477 s32 pid = current->pid; 478 int err; 479 static s32 rover = -4097; 480 481 retry: 482 cond_resched(); 483 netlink_table_grab(); 484 head = nl_pid_hashfn(hash, pid); 485 sk_for_each(osk, node, head) { 486 if (nlk_sk(osk)->pid == pid) { 487 /* Bind collision, search negative pid values. */ 488 pid = rover--; 489 if (rover > -4097) 490 rover = -4097; 491 netlink_table_ungrab(); 492 goto retry; 493 } 494 } 495 netlink_table_ungrab(); 496 497 err = netlink_insert(sk, pid); 498 if (err == -EADDRINUSE) 499 goto retry; 500 501 /* If 2 threads race to autobind, that is fine. */ 502 if (err == -EBUSY) 503 err = 0; 504 505 return err; 506 } 507 508 static inline int netlink_capable(struct socket *sock, unsigned int flag) 509 { 510 return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) || 511 capable(CAP_NET_ADMIN); 512 } 513 514 static void 515 netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) 516 { 517 struct netlink_sock *nlk = nlk_sk(sk); 518 519 if (nlk->subscriptions && !subscriptions) 520 __sk_del_bind_node(sk); 521 else if (!nlk->subscriptions && subscriptions) 522 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); 523 nlk->subscriptions = subscriptions; 524 } 525 526 static int netlink_alloc_groups(struct sock *sk) 527 { 528 struct netlink_sock *nlk = nlk_sk(sk); 529 unsigned int groups; 530 int err = 0; 531 532 netlink_lock_table(); 533 groups = nl_table[sk->sk_protocol].groups; 534 if (!nl_table[sk->sk_protocol].registered) 535 err = -ENOENT; 536 netlink_unlock_table(); 537 538 if (err) 539 return err; 540 541 nlk->groups = kmalloc(NLGRPSZ(groups), GFP_KERNEL); 542 if (nlk->groups == NULL) 543 return -ENOMEM; 544 memset(nlk->groups, 0, NLGRPSZ(groups)); 545 nlk->ngroups = groups; 546 return 0; 547 } 548 549 static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len) 550 { 551 struct sock *sk = sock->sk; 552 struct netlink_sock *nlk = nlk_sk(sk); 553 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; 554 int err; 555 556 if (nladdr->nl_family != AF_NETLINK) 557 return -EINVAL; 558 559 /* Only superuser is allowed to listen multicasts */ 560 if (nladdr->nl_groups) { 561 if (!netlink_capable(sock, NL_NONROOT_RECV)) 562 return -EPERM; 563 if (nlk->groups == NULL) { 564 err = netlink_alloc_groups(sk); 565 if (err) 566 return err; 567 } 568 } 569 570 if (nlk->pid) { 571 if (nladdr->nl_pid != nlk->pid) 572 return -EINVAL; 573 } else { 574 err = nladdr->nl_pid ? 575 netlink_insert(sk, nladdr->nl_pid) : 576 netlink_autobind(sock); 577 if (err) 578 return err; 579 } 580 581 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) 582 return 0; 583 584 netlink_table_grab(); 585 netlink_update_subscriptions(sk, nlk->subscriptions + 586 hweight32(nladdr->nl_groups) - 587 hweight32(nlk->groups[0])); 588 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; 589 netlink_table_ungrab(); 590 591 return 0; 592 } 593 594 static int netlink_connect(struct socket *sock, struct sockaddr *addr, 595 int alen, int flags) 596 { 597 int err = 0; 598 struct sock *sk = sock->sk; 599 struct netlink_sock *nlk = nlk_sk(sk); 600 struct sockaddr_nl *nladdr=(struct sockaddr_nl*)addr; 601 602 if (addr->sa_family == AF_UNSPEC) { 603 sk->sk_state = NETLINK_UNCONNECTED; 604 nlk->dst_pid = 0; 605 nlk->dst_group = 0; 606 return 0; 607 } 608 if (addr->sa_family != AF_NETLINK) 609 return -EINVAL; 610 611 /* Only superuser is allowed to send multicasts */ 612 if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND)) 613 return -EPERM; 614 615 if (!nlk->pid) 616 err = netlink_autobind(sock); 617 618 if (err == 0) { 619 sk->sk_state = NETLINK_CONNECTED; 620 nlk->dst_pid = nladdr->nl_pid; 621 nlk->dst_group = ffs(nladdr->nl_groups); 622 } 623 624 return err; 625 } 626 627 static int netlink_getname(struct socket *sock, struct sockaddr *addr, int *addr_len, int peer) 628 { 629 struct sock *sk = sock->sk; 630 struct netlink_sock *nlk = nlk_sk(sk); 631 struct sockaddr_nl *nladdr=(struct sockaddr_nl *)addr; 632 633 nladdr->nl_family = AF_NETLINK; 634 nladdr->nl_pad = 0; 635 *addr_len = sizeof(*nladdr); 636 637 if (peer) { 638 nladdr->nl_pid = nlk->dst_pid; 639 nladdr->nl_groups = netlink_group_mask(nlk->dst_group); 640 } else { 641 nladdr->nl_pid = nlk->pid; 642 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; 643 } 644 return 0; 645 } 646 647 static void netlink_overrun(struct sock *sk) 648 { 649 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) { 650 sk->sk_err = ENOBUFS; 651 sk->sk_error_report(sk); 652 } 653 } 654 655 static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid) 656 { 657 int protocol = ssk->sk_protocol; 658 struct sock *sock; 659 struct netlink_sock *nlk; 660 661 sock = netlink_lookup(protocol, pid); 662 if (!sock) 663 return ERR_PTR(-ECONNREFUSED); 664 665 /* Don't bother queuing skb if kernel socket has no input function */ 666 nlk = nlk_sk(sock); 667 if ((nlk->pid == 0 && !nlk->data_ready) || 668 (sock->sk_state == NETLINK_CONNECTED && 669 nlk->dst_pid != nlk_sk(ssk)->pid)) { 670 sock_put(sock); 671 return ERR_PTR(-ECONNREFUSED); 672 } 673 return sock; 674 } 675 676 struct sock *netlink_getsockbyfilp(struct file *filp) 677 { 678 struct inode *inode = filp->f_dentry->d_inode; 679 struct sock *sock; 680 681 if (!S_ISSOCK(inode->i_mode)) 682 return ERR_PTR(-ENOTSOCK); 683 684 sock = SOCKET_I(inode)->sk; 685 if (sock->sk_family != AF_NETLINK) 686 return ERR_PTR(-EINVAL); 687 688 sock_hold(sock); 689 return sock; 690 } 691 692 /* 693 * Attach a skb to a netlink socket. 694 * The caller must hold a reference to the destination socket. On error, the 695 * reference is dropped. The skb is not send to the destination, just all 696 * all error checks are performed and memory in the queue is reserved. 697 * Return values: 698 * < 0: error. skb freed, reference to sock dropped. 699 * 0: continue 700 * 1: repeat lookup - reference dropped while waiting for socket memory. 701 */ 702 int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock, long timeo) 703 { 704 struct netlink_sock *nlk; 705 706 nlk = nlk_sk(sk); 707 708 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || 709 test_bit(0, &nlk->state)) { 710 DECLARE_WAITQUEUE(wait, current); 711 if (!timeo) { 712 if (!nlk->pid) 713 netlink_overrun(sk); 714 sock_put(sk); 715 kfree_skb(skb); 716 return -EAGAIN; 717 } 718 719 __set_current_state(TASK_INTERRUPTIBLE); 720 add_wait_queue(&nlk->wait, &wait); 721 722 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || 723 test_bit(0, &nlk->state)) && 724 !sock_flag(sk, SOCK_DEAD)) 725 timeo = schedule_timeout(timeo); 726 727 __set_current_state(TASK_RUNNING); 728 remove_wait_queue(&nlk->wait, &wait); 729 sock_put(sk); 730 731 if (signal_pending(current)) { 732 kfree_skb(skb); 733 return sock_intr_errno(timeo); 734 } 735 return 1; 736 } 737 skb_set_owner_r(skb, sk); 738 return 0; 739 } 740 741 int netlink_sendskb(struct sock *sk, struct sk_buff *skb, int protocol) 742 { 743 int len = skb->len; 744 745 skb_queue_tail(&sk->sk_receive_queue, skb); 746 sk->sk_data_ready(sk, len); 747 sock_put(sk); 748 return len; 749 } 750 751 void netlink_detachskb(struct sock *sk, struct sk_buff *skb) 752 { 753 kfree_skb(skb); 754 sock_put(sk); 755 } 756 757 static inline struct sk_buff *netlink_trim(struct sk_buff *skb, 758 gfp_t allocation) 759 { 760 int delta; 761 762 skb_orphan(skb); 763 764 delta = skb->end - skb->tail; 765 if (delta * 2 < skb->truesize) 766 return skb; 767 768 if (skb_shared(skb)) { 769 struct sk_buff *nskb = skb_clone(skb, allocation); 770 if (!nskb) 771 return skb; 772 kfree_skb(skb); 773 skb = nskb; 774 } 775 776 if (!pskb_expand_head(skb, 0, -delta, allocation)) 777 skb->truesize -= delta; 778 779 return skb; 780 } 781 782 int netlink_unicast(struct sock *ssk, struct sk_buff *skb, u32 pid, int nonblock) 783 { 784 struct sock *sk; 785 int err; 786 long timeo; 787 788 skb = netlink_trim(skb, gfp_any()); 789 790 timeo = sock_sndtimeo(ssk, nonblock); 791 retry: 792 sk = netlink_getsockbypid(ssk, pid); 793 if (IS_ERR(sk)) { 794 kfree_skb(skb); 795 return PTR_ERR(sk); 796 } 797 err = netlink_attachskb(sk, skb, nonblock, timeo); 798 if (err == 1) 799 goto retry; 800 if (err) 801 return err; 802 803 return netlink_sendskb(sk, skb, ssk->sk_protocol); 804 } 805 806 static __inline__ int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) 807 { 808 struct netlink_sock *nlk = nlk_sk(sk); 809 810 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && 811 !test_bit(0, &nlk->state)) { 812 skb_set_owner_r(skb, sk); 813 skb_queue_tail(&sk->sk_receive_queue, skb); 814 sk->sk_data_ready(sk, skb->len); 815 return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf; 816 } 817 return -1; 818 } 819 820 struct netlink_broadcast_data { 821 struct sock *exclude_sk; 822 u32 pid; 823 u32 group; 824 int failure; 825 int congested; 826 int delivered; 827 gfp_t allocation; 828 struct sk_buff *skb, *skb2; 829 }; 830 831 static inline int do_one_broadcast(struct sock *sk, 832 struct netlink_broadcast_data *p) 833 { 834 struct netlink_sock *nlk = nlk_sk(sk); 835 int val; 836 837 if (p->exclude_sk == sk) 838 goto out; 839 840 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups || 841 !test_bit(p->group - 1, nlk->groups)) 842 goto out; 843 844 if (p->failure) { 845 netlink_overrun(sk); 846 goto out; 847 } 848 849 sock_hold(sk); 850 if (p->skb2 == NULL) { 851 if (skb_shared(p->skb)) { 852 p->skb2 = skb_clone(p->skb, p->allocation); 853 } else { 854 p->skb2 = skb_get(p->skb); 855 /* 856 * skb ownership may have been set when 857 * delivered to a previous socket. 858 */ 859 skb_orphan(p->skb2); 860 } 861 } 862 if (p->skb2 == NULL) { 863 netlink_overrun(sk); 864 /* Clone failed. Notify ALL listeners. */ 865 p->failure = 1; 866 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { 867 netlink_overrun(sk); 868 } else { 869 p->congested |= val; 870 p->delivered = 1; 871 p->skb2 = NULL; 872 } 873 sock_put(sk); 874 875 out: 876 return 0; 877 } 878 879 int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid, 880 u32 group, gfp_t allocation) 881 { 882 struct netlink_broadcast_data info; 883 struct hlist_node *node; 884 struct sock *sk; 885 886 skb = netlink_trim(skb, allocation); 887 888 info.exclude_sk = ssk; 889 info.pid = pid; 890 info.group = group; 891 info.failure = 0; 892 info.congested = 0; 893 info.delivered = 0; 894 info.allocation = allocation; 895 info.skb = skb; 896 info.skb2 = NULL; 897 898 /* While we sleep in clone, do not allow to change socket list */ 899 900 netlink_lock_table(); 901 902 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) 903 do_one_broadcast(sk, &info); 904 905 kfree_skb(skb); 906 907 netlink_unlock_table(); 908 909 if (info.skb2) 910 kfree_skb(info.skb2); 911 912 if (info.delivered) { 913 if (info.congested && (allocation & __GFP_WAIT)) 914 yield(); 915 return 0; 916 } 917 if (info.failure) 918 return -ENOBUFS; 919 return -ESRCH; 920 } 921 922 struct netlink_set_err_data { 923 struct sock *exclude_sk; 924 u32 pid; 925 u32 group; 926 int code; 927 }; 928 929 static inline int do_one_set_err(struct sock *sk, 930 struct netlink_set_err_data *p) 931 { 932 struct netlink_sock *nlk = nlk_sk(sk); 933 934 if (sk == p->exclude_sk) 935 goto out; 936 937 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups || 938 !test_bit(p->group - 1, nlk->groups)) 939 goto out; 940 941 sk->sk_err = p->code; 942 sk->sk_error_report(sk); 943 out: 944 return 0; 945 } 946 947 void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code) 948 { 949 struct netlink_set_err_data info; 950 struct hlist_node *node; 951 struct sock *sk; 952 953 info.exclude_sk = ssk; 954 info.pid = pid; 955 info.group = group; 956 info.code = code; 957 958 read_lock(&nl_table_lock); 959 960 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) 961 do_one_set_err(sk, &info); 962 963 read_unlock(&nl_table_lock); 964 } 965 966 static int netlink_setsockopt(struct socket *sock, int level, int optname, 967 char __user *optval, int optlen) 968 { 969 struct sock *sk = sock->sk; 970 struct netlink_sock *nlk = nlk_sk(sk); 971 int val = 0, err; 972 973 if (level != SOL_NETLINK) 974 return -ENOPROTOOPT; 975 976 if (optlen >= sizeof(int) && 977 get_user(val, (int __user *)optval)) 978 return -EFAULT; 979 980 switch (optname) { 981 case NETLINK_PKTINFO: 982 if (val) 983 nlk->flags |= NETLINK_RECV_PKTINFO; 984 else 985 nlk->flags &= ~NETLINK_RECV_PKTINFO; 986 err = 0; 987 break; 988 case NETLINK_ADD_MEMBERSHIP: 989 case NETLINK_DROP_MEMBERSHIP: { 990 unsigned int subscriptions; 991 int old, new = optname == NETLINK_ADD_MEMBERSHIP ? 1 : 0; 992 993 if (!netlink_capable(sock, NL_NONROOT_RECV)) 994 return -EPERM; 995 if (nlk->groups == NULL) { 996 err = netlink_alloc_groups(sk); 997 if (err) 998 return err; 999 } 1000 if (!val || val - 1 >= nlk->ngroups) 1001 return -EINVAL; 1002 netlink_table_grab(); 1003 old = test_bit(val - 1, nlk->groups); 1004 subscriptions = nlk->subscriptions - old + new; 1005 if (new) 1006 __set_bit(val - 1, nlk->groups); 1007 else 1008 __clear_bit(val - 1, nlk->groups); 1009 netlink_update_subscriptions(sk, subscriptions); 1010 netlink_table_ungrab(); 1011 err = 0; 1012 break; 1013 } 1014 default: 1015 err = -ENOPROTOOPT; 1016 } 1017 return err; 1018 } 1019 1020 static int netlink_getsockopt(struct socket *sock, int level, int optname, 1021 char __user *optval, int __user *optlen) 1022 { 1023 struct sock *sk = sock->sk; 1024 struct netlink_sock *nlk = nlk_sk(sk); 1025 int len, val, err; 1026 1027 if (level != SOL_NETLINK) 1028 return -ENOPROTOOPT; 1029 1030 if (get_user(len, optlen)) 1031 return -EFAULT; 1032 if (len < 0) 1033 return -EINVAL; 1034 1035 switch (optname) { 1036 case NETLINK_PKTINFO: 1037 if (len < sizeof(int)) 1038 return -EINVAL; 1039 len = sizeof(int); 1040 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; 1041 put_user(len, optlen); 1042 put_user(val, optval); 1043 err = 0; 1044 break; 1045 default: 1046 err = -ENOPROTOOPT; 1047 } 1048 return err; 1049 } 1050 1051 static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) 1052 { 1053 struct nl_pktinfo info; 1054 1055 info.group = NETLINK_CB(skb).dst_group; 1056 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); 1057 } 1058 1059 static inline void netlink_rcv_wake(struct sock *sk) 1060 { 1061 struct netlink_sock *nlk = nlk_sk(sk); 1062 1063 if (skb_queue_empty(&sk->sk_receive_queue)) 1064 clear_bit(0, &nlk->state); 1065 if (!test_bit(0, &nlk->state)) 1066 wake_up_interruptible(&nlk->wait); 1067 } 1068 1069 static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, 1070 struct msghdr *msg, size_t len) 1071 { 1072 struct sock_iocb *siocb = kiocb_to_siocb(kiocb); 1073 struct sock *sk = sock->sk; 1074 struct netlink_sock *nlk = nlk_sk(sk); 1075 struct sockaddr_nl *addr=msg->msg_name; 1076 u32 dst_pid; 1077 u32 dst_group; 1078 struct sk_buff *skb; 1079 int err; 1080 struct scm_cookie scm; 1081 1082 if (msg->msg_flags&MSG_OOB) 1083 return -EOPNOTSUPP; 1084 1085 if (NULL == siocb->scm) 1086 siocb->scm = &scm; 1087 err = scm_send(sock, msg, siocb->scm); 1088 if (err < 0) 1089 return err; 1090 1091 if (msg->msg_namelen) { 1092 if (addr->nl_family != AF_NETLINK) 1093 return -EINVAL; 1094 dst_pid = addr->nl_pid; 1095 dst_group = ffs(addr->nl_groups); 1096 if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND)) 1097 return -EPERM; 1098 } else { 1099 dst_pid = nlk->dst_pid; 1100 dst_group = nlk->dst_group; 1101 } 1102 1103 if (!nlk->pid) { 1104 err = netlink_autobind(sock); 1105 if (err) 1106 goto out; 1107 } 1108 1109 err = -EMSGSIZE; 1110 if (len > sk->sk_sndbuf - 32) 1111 goto out; 1112 err = -ENOBUFS; 1113 skb = alloc_skb(len, GFP_KERNEL); 1114 if (skb==NULL) 1115 goto out; 1116 1117 NETLINK_CB(skb).pid = nlk->pid; 1118 NETLINK_CB(skb).dst_pid = dst_pid; 1119 NETLINK_CB(skb).dst_group = dst_group; 1120 NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context); 1121 memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred)); 1122 1123 /* What can I do? Netlink is asynchronous, so that 1124 we will have to save current capabilities to 1125 check them, when this message will be delivered 1126 to corresponding kernel module. --ANK (980802) 1127 */ 1128 1129 err = -EFAULT; 1130 if (memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len)) { 1131 kfree_skb(skb); 1132 goto out; 1133 } 1134 1135 err = security_netlink_send(sk, skb); 1136 if (err) { 1137 kfree_skb(skb); 1138 goto out; 1139 } 1140 1141 if (dst_group) { 1142 atomic_inc(&skb->users); 1143 netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL); 1144 } 1145 err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT); 1146 1147 out: 1148 return err; 1149 } 1150 1151 static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, 1152 struct msghdr *msg, size_t len, 1153 int flags) 1154 { 1155 struct sock_iocb *siocb = kiocb_to_siocb(kiocb); 1156 struct scm_cookie scm; 1157 struct sock *sk = sock->sk; 1158 struct netlink_sock *nlk = nlk_sk(sk); 1159 int noblock = flags&MSG_DONTWAIT; 1160 size_t copied; 1161 struct sk_buff *skb; 1162 int err; 1163 1164 if (flags&MSG_OOB) 1165 return -EOPNOTSUPP; 1166 1167 copied = 0; 1168 1169 skb = skb_recv_datagram(sk,flags,noblock,&err); 1170 if (skb==NULL) 1171 goto out; 1172 1173 msg->msg_namelen = 0; 1174 1175 copied = skb->len; 1176 if (len < copied) { 1177 msg->msg_flags |= MSG_TRUNC; 1178 copied = len; 1179 } 1180 1181 skb->h.raw = skb->data; 1182 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 1183 1184 if (msg->msg_name) { 1185 struct sockaddr_nl *addr = (struct sockaddr_nl*)msg->msg_name; 1186 addr->nl_family = AF_NETLINK; 1187 addr->nl_pad = 0; 1188 addr->nl_pid = NETLINK_CB(skb).pid; 1189 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); 1190 msg->msg_namelen = sizeof(*addr); 1191 } 1192 1193 if (NULL == siocb->scm) { 1194 memset(&scm, 0, sizeof(scm)); 1195 siocb->scm = &scm; 1196 } 1197 siocb->scm->creds = *NETLINK_CREDS(skb); 1198 skb_free_datagram(sk, skb); 1199 1200 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) 1201 netlink_dump(sk); 1202 1203 scm_recv(sock, msg, siocb->scm, flags); 1204 if (nlk->flags & NETLINK_RECV_PKTINFO) 1205 netlink_cmsg_recv_pktinfo(msg, skb); 1206 1207 out: 1208 netlink_rcv_wake(sk); 1209 return err ? : copied; 1210 } 1211 1212 static void netlink_data_ready(struct sock *sk, int len) 1213 { 1214 struct netlink_sock *nlk = nlk_sk(sk); 1215 1216 if (nlk->data_ready) 1217 nlk->data_ready(sk, len); 1218 netlink_rcv_wake(sk); 1219 } 1220 1221 /* 1222 * We export these functions to other modules. They provide a 1223 * complete set of kernel non-blocking support for message 1224 * queueing. 1225 */ 1226 1227 struct sock * 1228 netlink_kernel_create(int unit, unsigned int groups, 1229 void (*input)(struct sock *sk, int len), 1230 struct module *module) 1231 { 1232 struct socket *sock; 1233 struct sock *sk; 1234 struct netlink_sock *nlk; 1235 1236 if (!nl_table) 1237 return NULL; 1238 1239 if (unit<0 || unit>=MAX_LINKS) 1240 return NULL; 1241 1242 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) 1243 return NULL; 1244 1245 if (__netlink_create(sock, unit) < 0) 1246 goto out_sock_release; 1247 1248 sk = sock->sk; 1249 sk->sk_data_ready = netlink_data_ready; 1250 if (input) 1251 nlk_sk(sk)->data_ready = input; 1252 1253 if (netlink_insert(sk, 0)) 1254 goto out_sock_release; 1255 1256 nlk = nlk_sk(sk); 1257 nlk->flags |= NETLINK_KERNEL_SOCKET; 1258 1259 netlink_table_grab(); 1260 nl_table[unit].groups = groups < 32 ? 32 : groups; 1261 nl_table[unit].module = module; 1262 nl_table[unit].registered = 1; 1263 netlink_table_ungrab(); 1264 1265 return sk; 1266 1267 out_sock_release: 1268 sock_release(sock); 1269 return NULL; 1270 } 1271 1272 void netlink_set_nonroot(int protocol, unsigned int flags) 1273 { 1274 if ((unsigned int)protocol < MAX_LINKS) 1275 nl_table[protocol].nl_nonroot = flags; 1276 } 1277 1278 static void netlink_destroy_callback(struct netlink_callback *cb) 1279 { 1280 if (cb->skb) 1281 kfree_skb(cb->skb); 1282 kfree(cb); 1283 } 1284 1285 /* 1286 * It looks a bit ugly. 1287 * It would be better to create kernel thread. 1288 */ 1289 1290 static int netlink_dump(struct sock *sk) 1291 { 1292 struct netlink_sock *nlk = nlk_sk(sk); 1293 struct netlink_callback *cb; 1294 struct sk_buff *skb; 1295 struct nlmsghdr *nlh; 1296 int len; 1297 1298 skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL); 1299 if (!skb) 1300 return -ENOBUFS; 1301 1302 spin_lock(&nlk->cb_lock); 1303 1304 cb = nlk->cb; 1305 if (cb == NULL) { 1306 spin_unlock(&nlk->cb_lock); 1307 kfree_skb(skb); 1308 return -EINVAL; 1309 } 1310 1311 len = cb->dump(skb, cb); 1312 1313 if (len > 0) { 1314 spin_unlock(&nlk->cb_lock); 1315 skb_queue_tail(&sk->sk_receive_queue, skb); 1316 sk->sk_data_ready(sk, len); 1317 return 0; 1318 } 1319 1320 nlh = NLMSG_NEW_ANSWER(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); 1321 memcpy(NLMSG_DATA(nlh), &len, sizeof(len)); 1322 skb_queue_tail(&sk->sk_receive_queue, skb); 1323 sk->sk_data_ready(sk, skb->len); 1324 1325 cb->done(cb); 1326 nlk->cb = NULL; 1327 spin_unlock(&nlk->cb_lock); 1328 1329 netlink_destroy_callback(cb); 1330 return 0; 1331 1332 nlmsg_failure: 1333 return -ENOBUFS; 1334 } 1335 1336 int netlink_dump_start(struct sock *ssk, struct sk_buff *skb, 1337 struct nlmsghdr *nlh, 1338 int (*dump)(struct sk_buff *skb, struct netlink_callback*), 1339 int (*done)(struct netlink_callback*)) 1340 { 1341 struct netlink_callback *cb; 1342 struct sock *sk; 1343 struct netlink_sock *nlk; 1344 1345 cb = kmalloc(sizeof(*cb), GFP_KERNEL); 1346 if (cb == NULL) 1347 return -ENOBUFS; 1348 1349 memset(cb, 0, sizeof(*cb)); 1350 cb->dump = dump; 1351 cb->done = done; 1352 cb->nlh = nlh; 1353 atomic_inc(&skb->users); 1354 cb->skb = skb; 1355 1356 sk = netlink_lookup(ssk->sk_protocol, NETLINK_CB(skb).pid); 1357 if (sk == NULL) { 1358 netlink_destroy_callback(cb); 1359 return -ECONNREFUSED; 1360 } 1361 nlk = nlk_sk(sk); 1362 /* A dump is in progress... */ 1363 spin_lock(&nlk->cb_lock); 1364 if (nlk->cb) { 1365 spin_unlock(&nlk->cb_lock); 1366 netlink_destroy_callback(cb); 1367 sock_put(sk); 1368 return -EBUSY; 1369 } 1370 nlk->cb = cb; 1371 spin_unlock(&nlk->cb_lock); 1372 1373 netlink_dump(sk); 1374 sock_put(sk); 1375 return 0; 1376 } 1377 1378 void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) 1379 { 1380 struct sk_buff *skb; 1381 struct nlmsghdr *rep; 1382 struct nlmsgerr *errmsg; 1383 int size; 1384 1385 if (err == 0) 1386 size = NLMSG_SPACE(sizeof(struct nlmsgerr)); 1387 else 1388 size = NLMSG_SPACE(4 + NLMSG_ALIGN(nlh->nlmsg_len)); 1389 1390 skb = alloc_skb(size, GFP_KERNEL); 1391 if (!skb) { 1392 struct sock *sk; 1393 1394 sk = netlink_lookup(in_skb->sk->sk_protocol, 1395 NETLINK_CB(in_skb).pid); 1396 if (sk) { 1397 sk->sk_err = ENOBUFS; 1398 sk->sk_error_report(sk); 1399 sock_put(sk); 1400 } 1401 return; 1402 } 1403 1404 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, 1405 NLMSG_ERROR, sizeof(struct nlmsgerr), 0); 1406 errmsg = NLMSG_DATA(rep); 1407 errmsg->error = err; 1408 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(struct nlmsghdr)); 1409 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT); 1410 } 1411 1412 1413 #ifdef CONFIG_PROC_FS 1414 struct nl_seq_iter { 1415 int link; 1416 int hash_idx; 1417 }; 1418 1419 static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) 1420 { 1421 struct nl_seq_iter *iter = seq->private; 1422 int i, j; 1423 struct sock *s; 1424 struct hlist_node *node; 1425 loff_t off = 0; 1426 1427 for (i=0; i<MAX_LINKS; i++) { 1428 struct nl_pid_hash *hash = &nl_table[i].hash; 1429 1430 for (j = 0; j <= hash->mask; j++) { 1431 sk_for_each(s, node, &hash->table[j]) { 1432 if (off == pos) { 1433 iter->link = i; 1434 iter->hash_idx = j; 1435 return s; 1436 } 1437 ++off; 1438 } 1439 } 1440 } 1441 return NULL; 1442 } 1443 1444 static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) 1445 { 1446 read_lock(&nl_table_lock); 1447 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; 1448 } 1449 1450 static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1451 { 1452 struct sock *s; 1453 struct nl_seq_iter *iter; 1454 int i, j; 1455 1456 ++*pos; 1457 1458 if (v == SEQ_START_TOKEN) 1459 return netlink_seq_socket_idx(seq, 0); 1460 1461 s = sk_next(v); 1462 if (s) 1463 return s; 1464 1465 iter = seq->private; 1466 i = iter->link; 1467 j = iter->hash_idx + 1; 1468 1469 do { 1470 struct nl_pid_hash *hash = &nl_table[i].hash; 1471 1472 for (; j <= hash->mask; j++) { 1473 s = sk_head(&hash->table[j]); 1474 if (s) { 1475 iter->link = i; 1476 iter->hash_idx = j; 1477 return s; 1478 } 1479 } 1480 1481 j = 0; 1482 } while (++i < MAX_LINKS); 1483 1484 return NULL; 1485 } 1486 1487 static void netlink_seq_stop(struct seq_file *seq, void *v) 1488 { 1489 read_unlock(&nl_table_lock); 1490 } 1491 1492 1493 static int netlink_seq_show(struct seq_file *seq, void *v) 1494 { 1495 if (v == SEQ_START_TOKEN) 1496 seq_puts(seq, 1497 "sk Eth Pid Groups " 1498 "Rmem Wmem Dump Locks\n"); 1499 else { 1500 struct sock *s = v; 1501 struct netlink_sock *nlk = nlk_sk(s); 1502 1503 seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n", 1504 s, 1505 s->sk_protocol, 1506 nlk->pid, 1507 nlk->groups ? (u32)nlk->groups[0] : 0, 1508 atomic_read(&s->sk_rmem_alloc), 1509 atomic_read(&s->sk_wmem_alloc), 1510 nlk->cb, 1511 atomic_read(&s->sk_refcnt) 1512 ); 1513 1514 } 1515 return 0; 1516 } 1517 1518 static struct seq_operations netlink_seq_ops = { 1519 .start = netlink_seq_start, 1520 .next = netlink_seq_next, 1521 .stop = netlink_seq_stop, 1522 .show = netlink_seq_show, 1523 }; 1524 1525 1526 static int netlink_seq_open(struct inode *inode, struct file *file) 1527 { 1528 struct seq_file *seq; 1529 struct nl_seq_iter *iter; 1530 int err; 1531 1532 iter = kmalloc(sizeof(*iter), GFP_KERNEL); 1533 if (!iter) 1534 return -ENOMEM; 1535 1536 err = seq_open(file, &netlink_seq_ops); 1537 if (err) { 1538 kfree(iter); 1539 return err; 1540 } 1541 1542 memset(iter, 0, sizeof(*iter)); 1543 seq = file->private_data; 1544 seq->private = iter; 1545 return 0; 1546 } 1547 1548 static struct file_operations netlink_seq_fops = { 1549 .owner = THIS_MODULE, 1550 .open = netlink_seq_open, 1551 .read = seq_read, 1552 .llseek = seq_lseek, 1553 .release = seq_release_private, 1554 }; 1555 1556 #endif 1557 1558 int netlink_register_notifier(struct notifier_block *nb) 1559 { 1560 return notifier_chain_register(&netlink_chain, nb); 1561 } 1562 1563 int netlink_unregister_notifier(struct notifier_block *nb) 1564 { 1565 return notifier_chain_unregister(&netlink_chain, nb); 1566 } 1567 1568 static struct proto_ops netlink_ops = { 1569 .family = PF_NETLINK, 1570 .owner = THIS_MODULE, 1571 .release = netlink_release, 1572 .bind = netlink_bind, 1573 .connect = netlink_connect, 1574 .socketpair = sock_no_socketpair, 1575 .accept = sock_no_accept, 1576 .getname = netlink_getname, 1577 .poll = datagram_poll, 1578 .ioctl = sock_no_ioctl, 1579 .listen = sock_no_listen, 1580 .shutdown = sock_no_shutdown, 1581 .setsockopt = netlink_setsockopt, 1582 .getsockopt = netlink_getsockopt, 1583 .sendmsg = netlink_sendmsg, 1584 .recvmsg = netlink_recvmsg, 1585 .mmap = sock_no_mmap, 1586 .sendpage = sock_no_sendpage, 1587 }; 1588 1589 static struct net_proto_family netlink_family_ops = { 1590 .family = PF_NETLINK, 1591 .create = netlink_create, 1592 .owner = THIS_MODULE, /* for consistency 8) */ 1593 }; 1594 1595 extern void netlink_skb_parms_too_large(void); 1596 1597 static int __init netlink_proto_init(void) 1598 { 1599 struct sk_buff *dummy_skb; 1600 int i; 1601 unsigned long max; 1602 unsigned int order; 1603 int err = proto_register(&netlink_proto, 0); 1604 1605 if (err != 0) 1606 goto out; 1607 1608 if (sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb)) 1609 netlink_skb_parms_too_large(); 1610 1611 nl_table = kmalloc(sizeof(*nl_table) * MAX_LINKS, GFP_KERNEL); 1612 if (!nl_table) { 1613 enomem: 1614 printk(KERN_CRIT "netlink_init: Cannot allocate nl_table\n"); 1615 return -ENOMEM; 1616 } 1617 1618 memset(nl_table, 0, sizeof(*nl_table) * MAX_LINKS); 1619 1620 if (num_physpages >= (128 * 1024)) 1621 max = num_physpages >> (21 - PAGE_SHIFT); 1622 else 1623 max = num_physpages >> (23 - PAGE_SHIFT); 1624 1625 order = get_bitmask_order(max) - 1 + PAGE_SHIFT; 1626 max = (1UL << order) / sizeof(struct hlist_head); 1627 order = get_bitmask_order(max > UINT_MAX ? UINT_MAX : max) - 1; 1628 1629 for (i = 0; i < MAX_LINKS; i++) { 1630 struct nl_pid_hash *hash = &nl_table[i].hash; 1631 1632 hash->table = nl_pid_hash_alloc(1 * sizeof(*hash->table)); 1633 if (!hash->table) { 1634 while (i-- > 0) 1635 nl_pid_hash_free(nl_table[i].hash.table, 1636 1 * sizeof(*hash->table)); 1637 kfree(nl_table); 1638 goto enomem; 1639 } 1640 memset(hash->table, 0, 1 * sizeof(*hash->table)); 1641 hash->max_shift = order; 1642 hash->shift = 0; 1643 hash->mask = 0; 1644 hash->rehash_time = jiffies; 1645 } 1646 1647 sock_register(&netlink_family_ops); 1648 #ifdef CONFIG_PROC_FS 1649 proc_net_fops_create("netlink", 0, &netlink_seq_fops); 1650 #endif 1651 /* The netlink device handler may be needed early. */ 1652 rtnetlink_init(); 1653 out: 1654 return err; 1655 } 1656 1657 core_initcall(netlink_proto_init); 1658 1659 EXPORT_SYMBOL(netlink_ack); 1660 EXPORT_SYMBOL(netlink_broadcast); 1661 EXPORT_SYMBOL(netlink_dump_start); 1662 EXPORT_SYMBOL(netlink_kernel_create); 1663 EXPORT_SYMBOL(netlink_register_notifier); 1664 EXPORT_SYMBOL(netlink_set_err); 1665 EXPORT_SYMBOL(netlink_set_nonroot); 1666 EXPORT_SYMBOL(netlink_unicast); 1667 EXPORT_SYMBOL(netlink_unregister_notifier); 1668 1669