1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Generic socket support routines. Memory allocators, socket lock/release 7 * handler for protocols to use and generic option handler. 8 * 9 * 10 * Version: $Id: sock.c,v 1.117 2002/02/01 22:01:03 davem Exp $ 11 * 12 * Authors: Ross Biro 13 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 14 * Florian La Roche, <flla@stud.uni-sb.de> 15 * Alan Cox, <A.Cox@swansea.ac.uk> 16 * 17 * Fixes: 18 * Alan Cox : Numerous verify_area() problems 19 * Alan Cox : Connecting on a connecting socket 20 * now returns an error for tcp. 21 * Alan Cox : sock->protocol is set correctly. 22 * and is not sometimes left as 0. 23 * Alan Cox : connect handles icmp errors on a 24 * connect properly. Unfortunately there 25 * is a restart syscall nasty there. I 26 * can't match BSD without hacking the C 27 * library. Ideas urgently sought! 28 * Alan Cox : Disallow bind() to addresses that are 29 * not ours - especially broadcast ones!! 30 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost) 31 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets, 32 * instead they leave that for the DESTROY timer. 33 * Alan Cox : Clean up error flag in accept 34 * Alan Cox : TCP ack handling is buggy, the DESTROY timer 35 * was buggy. Put a remove_sock() in the handler 36 * for memory when we hit 0. Also altered the timer 37 * code. The ACK stuff can wait and needs major 38 * TCP layer surgery. 39 * Alan Cox : Fixed TCP ack bug, removed remove sock 40 * and fixed timer/inet_bh race. 41 * Alan Cox : Added zapped flag for TCP 42 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code 43 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb 44 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources 45 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing. 46 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so... 47 * Rick Sladkey : Relaxed UDP rules for matching packets. 48 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support 49 * Pauline Middelink : identd support 50 * Alan Cox : Fixed connect() taking signals I think. 51 * Alan Cox : SO_LINGER supported 52 * Alan Cox : Error reporting fixes 53 * Anonymous : inet_create tidied up (sk->reuse setting) 54 * Alan Cox : inet sockets don't set sk->type! 55 * Alan Cox : Split socket option code 56 * Alan Cox : Callbacks 57 * Alan Cox : Nagle flag for Charles & Johannes stuff 58 * Alex : Removed restriction on inet fioctl 59 * Alan Cox : Splitting INET from NET core 60 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt() 61 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code 62 * Alan Cox : Split IP from generic code 63 * Alan Cox : New kfree_skbmem() 64 * Alan Cox : Make SO_DEBUG superuser only. 65 * Alan Cox : Allow anyone to clear SO_DEBUG 66 * (compatibility fix) 67 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput. 68 * Alan Cox : Allocator for a socket is settable. 69 * Alan Cox : SO_ERROR includes soft errors. 70 * Alan Cox : Allow NULL arguments on some SO_ opts 71 * Alan Cox : Generic socket allocation to make hooks 72 * easier (suggested by Craig Metz). 73 * Michael Pall : SO_ERROR returns positive errno again 74 * Steve Whitehouse: Added default destructor to free 75 * protocol private data. 76 * Steve Whitehouse: Added various other default routines 77 * common to several socket families. 78 * Chris Evans : Call suser() check last on F_SETOWN 79 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER. 80 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s() 81 * Andi Kleen : Fix write_space callback 82 * Chris Evans : Security fixes - signedness again 83 * Arnaldo C. Melo : cleanups, use skb_queue_purge 84 * 85 * To Fix: 86 * 87 * 88 * This program is free software; you can redistribute it and/or 89 * modify it under the terms of the GNU General Public License 90 * as published by the Free Software Foundation; either version 91 * 2 of the License, or (at your option) any later version. 92 */ 93 94 #include <linux/capability.h> 95 #include <linux/errno.h> 96 #include <linux/types.h> 97 #include <linux/socket.h> 98 #include <linux/in.h> 99 #include <linux/kernel.h> 100 #include <linux/module.h> 101 #include <linux/proc_fs.h> 102 #include <linux/seq_file.h> 103 #include <linux/sched.h> 104 #include <linux/timer.h> 105 #include <linux/string.h> 106 #include <linux/sockios.h> 107 #include <linux/net.h> 108 #include <linux/mm.h> 109 #include <linux/slab.h> 110 #include <linux/interrupt.h> 111 #include <linux/poll.h> 112 #include <linux/tcp.h> 113 #include <linux/init.h> 114 115 #include <asm/uaccess.h> 116 #include <asm/system.h> 117 118 #include <linux/netdevice.h> 119 #include <net/protocol.h> 120 #include <linux/skbuff.h> 121 #include <net/request_sock.h> 122 #include <net/sock.h> 123 #include <net/xfrm.h> 124 #include <linux/ipsec.h> 125 126 #include <linux/filter.h> 127 128 #ifdef CONFIG_INET 129 #include <net/tcp.h> 130 #endif 131 132 /* 133 * Each address family might have different locking rules, so we have 134 * one slock key per address family: 135 */ 136 static struct lock_class_key af_family_keys[AF_MAX]; 137 static struct lock_class_key af_family_slock_keys[AF_MAX]; 138 139 #ifdef CONFIG_DEBUG_LOCK_ALLOC 140 /* 141 * Make lock validator output more readable. (we pre-construct these 142 * strings build-time, so that runtime initialization of socket 143 * locks is fast): 144 */ 145 static const char *af_family_key_strings[AF_MAX+1] = { 146 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" , 147 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK", 148 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" , 149 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" , 150 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" , 151 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" , 152 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" , 153 "sk_lock-21" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" , 154 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" , 155 "sk_lock-27" , "sk_lock-28" , "sk_lock-29" , 156 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-AF_MAX" 157 }; 158 static const char *af_family_slock_key_strings[AF_MAX+1] = { 159 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" , 160 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK", 161 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" , 162 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" , 163 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" , 164 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" , 165 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" , 166 "slock-21" , "slock-AF_SNA" , "slock-AF_IRDA" , 167 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" , 168 "slock-27" , "slock-28" , "slock-29" , 169 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_MAX" 170 }; 171 #endif 172 173 /* 174 * sk_callback_lock locking rules are per-address-family, 175 * so split the lock classes by using a per-AF key: 176 */ 177 static struct lock_class_key af_callback_keys[AF_MAX]; 178 179 /* Take into consideration the size of the struct sk_buff overhead in the 180 * determination of these values, since that is non-constant across 181 * platforms. This makes socket queueing behavior and performance 182 * not depend upon such differences. 183 */ 184 #define _SK_MEM_PACKETS 256 185 #define _SK_MEM_OVERHEAD (sizeof(struct sk_buff) + 256) 186 #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) 187 #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) 188 189 /* Run time adjustable parameters. */ 190 __u32 sysctl_wmem_max = SK_WMEM_MAX; 191 __u32 sysctl_rmem_max = SK_RMEM_MAX; 192 __u32 sysctl_wmem_default = SK_WMEM_MAX; 193 __u32 sysctl_rmem_default = SK_RMEM_MAX; 194 195 /* Maximal space eaten by iovec or ancilliary data plus some space */ 196 int sysctl_optmem_max = sizeof(unsigned long)*(2*UIO_MAXIOV + 512); 197 198 static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen) 199 { 200 struct timeval tv; 201 202 if (optlen < sizeof(tv)) 203 return -EINVAL; 204 if (copy_from_user(&tv, optval, sizeof(tv))) 205 return -EFAULT; 206 207 *timeo_p = MAX_SCHEDULE_TIMEOUT; 208 if (tv.tv_sec == 0 && tv.tv_usec == 0) 209 return 0; 210 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1)) 211 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ); 212 return 0; 213 } 214 215 static void sock_warn_obsolete_bsdism(const char *name) 216 { 217 static int warned; 218 static char warncomm[TASK_COMM_LEN]; 219 if (strcmp(warncomm, current->comm) && warned < 5) { 220 strcpy(warncomm, current->comm); 221 printk(KERN_WARNING "process `%s' is using obsolete " 222 "%s SO_BSDCOMPAT\n", warncomm, name); 223 warned++; 224 } 225 } 226 227 static void sock_disable_timestamp(struct sock *sk) 228 { 229 if (sock_flag(sk, SOCK_TIMESTAMP)) { 230 sock_reset_flag(sk, SOCK_TIMESTAMP); 231 net_disable_timestamp(); 232 } 233 } 234 235 236 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 237 { 238 int err = 0; 239 int skb_len; 240 241 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces 242 number of warnings when compiling with -W --ANK 243 */ 244 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >= 245 (unsigned)sk->sk_rcvbuf) { 246 err = -ENOMEM; 247 goto out; 248 } 249 250 /* It would be deadlock, if sock_queue_rcv_skb is used 251 with socket lock! We assume that users of this 252 function are lock free. 253 */ 254 err = sk_filter(sk, skb, 1); 255 if (err) 256 goto out; 257 258 skb->dev = NULL; 259 skb_set_owner_r(skb, sk); 260 261 /* Cache the SKB length before we tack it onto the receive 262 * queue. Once it is added it no longer belongs to us and 263 * may be freed by other threads of control pulling packets 264 * from the queue. 265 */ 266 skb_len = skb->len; 267 268 skb_queue_tail(&sk->sk_receive_queue, skb); 269 270 if (!sock_flag(sk, SOCK_DEAD)) 271 sk->sk_data_ready(sk, skb_len); 272 out: 273 return err; 274 } 275 EXPORT_SYMBOL(sock_queue_rcv_skb); 276 277 int sk_receive_skb(struct sock *sk, struct sk_buff *skb) 278 { 279 int rc = NET_RX_SUCCESS; 280 281 if (sk_filter(sk, skb, 0)) 282 goto discard_and_relse; 283 284 skb->dev = NULL; 285 286 bh_lock_sock(sk); 287 if (!sock_owned_by_user(sk)) { 288 /* 289 * trylock + unlock semantics: 290 */ 291 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_); 292 293 rc = sk->sk_backlog_rcv(sk, skb); 294 295 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_); 296 } else 297 sk_add_backlog(sk, skb); 298 bh_unlock_sock(sk); 299 out: 300 sock_put(sk); 301 return rc; 302 discard_and_relse: 303 kfree_skb(skb); 304 goto out; 305 } 306 EXPORT_SYMBOL(sk_receive_skb); 307 308 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie) 309 { 310 struct dst_entry *dst = sk->sk_dst_cache; 311 312 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) { 313 sk->sk_dst_cache = NULL; 314 dst_release(dst); 315 return NULL; 316 } 317 318 return dst; 319 } 320 EXPORT_SYMBOL(__sk_dst_check); 321 322 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie) 323 { 324 struct dst_entry *dst = sk_dst_get(sk); 325 326 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) { 327 sk_dst_reset(sk); 328 dst_release(dst); 329 return NULL; 330 } 331 332 return dst; 333 } 334 EXPORT_SYMBOL(sk_dst_check); 335 336 /* 337 * This is meant for all protocols to use and covers goings on 338 * at the socket level. Everything here is generic. 339 */ 340 341 int sock_setsockopt(struct socket *sock, int level, int optname, 342 char __user *optval, int optlen) 343 { 344 struct sock *sk=sock->sk; 345 struct sk_filter *filter; 346 int val; 347 int valbool; 348 struct linger ling; 349 int ret = 0; 350 351 /* 352 * Options without arguments 353 */ 354 355 #ifdef SO_DONTLINGER /* Compatibility item... */ 356 if (optname == SO_DONTLINGER) { 357 lock_sock(sk); 358 sock_reset_flag(sk, SOCK_LINGER); 359 release_sock(sk); 360 return 0; 361 } 362 #endif 363 364 if(optlen<sizeof(int)) 365 return(-EINVAL); 366 367 if (get_user(val, (int __user *)optval)) 368 return -EFAULT; 369 370 valbool = val?1:0; 371 372 lock_sock(sk); 373 374 switch(optname) 375 { 376 case SO_DEBUG: 377 if(val && !capable(CAP_NET_ADMIN)) 378 { 379 ret = -EACCES; 380 } 381 else if (valbool) 382 sock_set_flag(sk, SOCK_DBG); 383 else 384 sock_reset_flag(sk, SOCK_DBG); 385 break; 386 case SO_REUSEADDR: 387 sk->sk_reuse = valbool; 388 break; 389 case SO_TYPE: 390 case SO_ERROR: 391 ret = -ENOPROTOOPT; 392 break; 393 case SO_DONTROUTE: 394 if (valbool) 395 sock_set_flag(sk, SOCK_LOCALROUTE); 396 else 397 sock_reset_flag(sk, SOCK_LOCALROUTE); 398 break; 399 case SO_BROADCAST: 400 sock_valbool_flag(sk, SOCK_BROADCAST, valbool); 401 break; 402 case SO_SNDBUF: 403 /* Don't error on this BSD doesn't and if you think 404 about it this is right. Otherwise apps have to 405 play 'guess the biggest size' games. RCVBUF/SNDBUF 406 are treated in BSD as hints */ 407 408 if (val > sysctl_wmem_max) 409 val = sysctl_wmem_max; 410 set_sndbuf: 411 sk->sk_userlocks |= SOCK_SNDBUF_LOCK; 412 if ((val * 2) < SOCK_MIN_SNDBUF) 413 sk->sk_sndbuf = SOCK_MIN_SNDBUF; 414 else 415 sk->sk_sndbuf = val * 2; 416 417 /* 418 * Wake up sending tasks if we 419 * upped the value. 420 */ 421 sk->sk_write_space(sk); 422 break; 423 424 case SO_SNDBUFFORCE: 425 if (!capable(CAP_NET_ADMIN)) { 426 ret = -EPERM; 427 break; 428 } 429 goto set_sndbuf; 430 431 case SO_RCVBUF: 432 /* Don't error on this BSD doesn't and if you think 433 about it this is right. Otherwise apps have to 434 play 'guess the biggest size' games. RCVBUF/SNDBUF 435 are treated in BSD as hints */ 436 437 if (val > sysctl_rmem_max) 438 val = sysctl_rmem_max; 439 set_rcvbuf: 440 sk->sk_userlocks |= SOCK_RCVBUF_LOCK; 441 /* 442 * We double it on the way in to account for 443 * "struct sk_buff" etc. overhead. Applications 444 * assume that the SO_RCVBUF setting they make will 445 * allow that much actual data to be received on that 446 * socket. 447 * 448 * Applications are unaware that "struct sk_buff" and 449 * other overheads allocate from the receive buffer 450 * during socket buffer allocation. 451 * 452 * And after considering the possible alternatives, 453 * returning the value we actually used in getsockopt 454 * is the most desirable behavior. 455 */ 456 if ((val * 2) < SOCK_MIN_RCVBUF) 457 sk->sk_rcvbuf = SOCK_MIN_RCVBUF; 458 else 459 sk->sk_rcvbuf = val * 2; 460 break; 461 462 case SO_RCVBUFFORCE: 463 if (!capable(CAP_NET_ADMIN)) { 464 ret = -EPERM; 465 break; 466 } 467 goto set_rcvbuf; 468 469 case SO_KEEPALIVE: 470 #ifdef CONFIG_INET 471 if (sk->sk_protocol == IPPROTO_TCP) 472 tcp_set_keepalive(sk, valbool); 473 #endif 474 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool); 475 break; 476 477 case SO_OOBINLINE: 478 sock_valbool_flag(sk, SOCK_URGINLINE, valbool); 479 break; 480 481 case SO_NO_CHECK: 482 sk->sk_no_check = valbool; 483 break; 484 485 case SO_PRIORITY: 486 if ((val >= 0 && val <= 6) || capable(CAP_NET_ADMIN)) 487 sk->sk_priority = val; 488 else 489 ret = -EPERM; 490 break; 491 492 case SO_LINGER: 493 if(optlen<sizeof(ling)) { 494 ret = -EINVAL; /* 1003.1g */ 495 break; 496 } 497 if (copy_from_user(&ling,optval,sizeof(ling))) { 498 ret = -EFAULT; 499 break; 500 } 501 if (!ling.l_onoff) 502 sock_reset_flag(sk, SOCK_LINGER); 503 else { 504 #if (BITS_PER_LONG == 32) 505 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ) 506 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT; 507 else 508 #endif 509 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ; 510 sock_set_flag(sk, SOCK_LINGER); 511 } 512 break; 513 514 case SO_BSDCOMPAT: 515 sock_warn_obsolete_bsdism("setsockopt"); 516 break; 517 518 case SO_PASSCRED: 519 if (valbool) 520 set_bit(SOCK_PASSCRED, &sock->flags); 521 else 522 clear_bit(SOCK_PASSCRED, &sock->flags); 523 break; 524 525 case SO_TIMESTAMP: 526 if (valbool) { 527 sock_set_flag(sk, SOCK_RCVTSTAMP); 528 sock_enable_timestamp(sk); 529 } else 530 sock_reset_flag(sk, SOCK_RCVTSTAMP); 531 break; 532 533 case SO_RCVLOWAT: 534 if (val < 0) 535 val = INT_MAX; 536 sk->sk_rcvlowat = val ? : 1; 537 break; 538 539 case SO_RCVTIMEO: 540 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen); 541 break; 542 543 case SO_SNDTIMEO: 544 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen); 545 break; 546 547 #ifdef CONFIG_NETDEVICES 548 case SO_BINDTODEVICE: 549 { 550 char devname[IFNAMSIZ]; 551 552 /* Sorry... */ 553 if (!capable(CAP_NET_RAW)) { 554 ret = -EPERM; 555 break; 556 } 557 558 /* Bind this socket to a particular device like "eth0", 559 * as specified in the passed interface name. If the 560 * name is "" or the option length is zero the socket 561 * is not bound. 562 */ 563 564 if (!valbool) { 565 sk->sk_bound_dev_if = 0; 566 } else { 567 if (optlen > IFNAMSIZ - 1) 568 optlen = IFNAMSIZ - 1; 569 memset(devname, 0, sizeof(devname)); 570 if (copy_from_user(devname, optval, optlen)) { 571 ret = -EFAULT; 572 break; 573 } 574 575 /* Remove any cached route for this socket. */ 576 sk_dst_reset(sk); 577 578 if (devname[0] == '\0') { 579 sk->sk_bound_dev_if = 0; 580 } else { 581 struct net_device *dev = dev_get_by_name(devname); 582 if (!dev) { 583 ret = -ENODEV; 584 break; 585 } 586 sk->sk_bound_dev_if = dev->ifindex; 587 dev_put(dev); 588 } 589 } 590 break; 591 } 592 #endif 593 594 595 case SO_ATTACH_FILTER: 596 ret = -EINVAL; 597 if (optlen == sizeof(struct sock_fprog)) { 598 struct sock_fprog fprog; 599 600 ret = -EFAULT; 601 if (copy_from_user(&fprog, optval, sizeof(fprog))) 602 break; 603 604 ret = sk_attach_filter(&fprog, sk); 605 } 606 break; 607 608 case SO_DETACH_FILTER: 609 spin_lock_bh(&sk->sk_lock.slock); 610 filter = sk->sk_filter; 611 if (filter) { 612 sk->sk_filter = NULL; 613 spin_unlock_bh(&sk->sk_lock.slock); 614 sk_filter_release(sk, filter); 615 break; 616 } 617 spin_unlock_bh(&sk->sk_lock.slock); 618 ret = -ENONET; 619 break; 620 621 case SO_PASSSEC: 622 if (valbool) 623 set_bit(SOCK_PASSSEC, &sock->flags); 624 else 625 clear_bit(SOCK_PASSSEC, &sock->flags); 626 break; 627 628 /* We implement the SO_SNDLOWAT etc to 629 not be settable (1003.1g 5.3) */ 630 default: 631 ret = -ENOPROTOOPT; 632 break; 633 } 634 release_sock(sk); 635 return ret; 636 } 637 638 639 int sock_getsockopt(struct socket *sock, int level, int optname, 640 char __user *optval, int __user *optlen) 641 { 642 struct sock *sk = sock->sk; 643 644 union 645 { 646 int val; 647 struct linger ling; 648 struct timeval tm; 649 } v; 650 651 unsigned int lv = sizeof(int); 652 int len; 653 654 if(get_user(len,optlen)) 655 return -EFAULT; 656 if(len < 0) 657 return -EINVAL; 658 659 switch(optname) 660 { 661 case SO_DEBUG: 662 v.val = sock_flag(sk, SOCK_DBG); 663 break; 664 665 case SO_DONTROUTE: 666 v.val = sock_flag(sk, SOCK_LOCALROUTE); 667 break; 668 669 case SO_BROADCAST: 670 v.val = !!sock_flag(sk, SOCK_BROADCAST); 671 break; 672 673 case SO_SNDBUF: 674 v.val = sk->sk_sndbuf; 675 break; 676 677 case SO_RCVBUF: 678 v.val = sk->sk_rcvbuf; 679 break; 680 681 case SO_REUSEADDR: 682 v.val = sk->sk_reuse; 683 break; 684 685 case SO_KEEPALIVE: 686 v.val = !!sock_flag(sk, SOCK_KEEPOPEN); 687 break; 688 689 case SO_TYPE: 690 v.val = sk->sk_type; 691 break; 692 693 case SO_ERROR: 694 v.val = -sock_error(sk); 695 if(v.val==0) 696 v.val = xchg(&sk->sk_err_soft, 0); 697 break; 698 699 case SO_OOBINLINE: 700 v.val = !!sock_flag(sk, SOCK_URGINLINE); 701 break; 702 703 case SO_NO_CHECK: 704 v.val = sk->sk_no_check; 705 break; 706 707 case SO_PRIORITY: 708 v.val = sk->sk_priority; 709 break; 710 711 case SO_LINGER: 712 lv = sizeof(v.ling); 713 v.ling.l_onoff = !!sock_flag(sk, SOCK_LINGER); 714 v.ling.l_linger = sk->sk_lingertime / HZ; 715 break; 716 717 case SO_BSDCOMPAT: 718 sock_warn_obsolete_bsdism("getsockopt"); 719 break; 720 721 case SO_TIMESTAMP: 722 v.val = sock_flag(sk, SOCK_RCVTSTAMP); 723 break; 724 725 case SO_RCVTIMEO: 726 lv=sizeof(struct timeval); 727 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) { 728 v.tm.tv_sec = 0; 729 v.tm.tv_usec = 0; 730 } else { 731 v.tm.tv_sec = sk->sk_rcvtimeo / HZ; 732 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ; 733 } 734 break; 735 736 case SO_SNDTIMEO: 737 lv=sizeof(struct timeval); 738 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) { 739 v.tm.tv_sec = 0; 740 v.tm.tv_usec = 0; 741 } else { 742 v.tm.tv_sec = sk->sk_sndtimeo / HZ; 743 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ; 744 } 745 break; 746 747 case SO_RCVLOWAT: 748 v.val = sk->sk_rcvlowat; 749 break; 750 751 case SO_SNDLOWAT: 752 v.val=1; 753 break; 754 755 case SO_PASSCRED: 756 v.val = test_bit(SOCK_PASSCRED, &sock->flags) ? 1 : 0; 757 break; 758 759 case SO_PEERCRED: 760 if (len > sizeof(sk->sk_peercred)) 761 len = sizeof(sk->sk_peercred); 762 if (copy_to_user(optval, &sk->sk_peercred, len)) 763 return -EFAULT; 764 goto lenout; 765 766 case SO_PEERNAME: 767 { 768 char address[128]; 769 770 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2)) 771 return -ENOTCONN; 772 if (lv < len) 773 return -EINVAL; 774 if (copy_to_user(optval, address, len)) 775 return -EFAULT; 776 goto lenout; 777 } 778 779 /* Dubious BSD thing... Probably nobody even uses it, but 780 * the UNIX standard wants it for whatever reason... -DaveM 781 */ 782 case SO_ACCEPTCONN: 783 v.val = sk->sk_state == TCP_LISTEN; 784 break; 785 786 case SO_PASSSEC: 787 v.val = test_bit(SOCK_PASSSEC, &sock->flags) ? 1 : 0; 788 break; 789 790 case SO_PEERSEC: 791 return security_socket_getpeersec_stream(sock, optval, optlen, len); 792 793 default: 794 return(-ENOPROTOOPT); 795 } 796 if (len > lv) 797 len = lv; 798 if (copy_to_user(optval, &v, len)) 799 return -EFAULT; 800 lenout: 801 if (put_user(len, optlen)) 802 return -EFAULT; 803 return 0; 804 } 805 806 /* 807 * Initialize an sk_lock. 808 * 809 * (We also register the sk_lock with the lock validator.) 810 */ 811 static void inline sock_lock_init(struct sock *sk) 812 { 813 spin_lock_init(&sk->sk_lock.slock); 814 sk->sk_lock.owner = NULL; 815 init_waitqueue_head(&sk->sk_lock.wq); 816 /* 817 * Make sure we are not reinitializing a held lock: 818 */ 819 debug_check_no_locks_freed((void *)&sk->sk_lock, sizeof(sk->sk_lock)); 820 821 /* 822 * Mark both the sk_lock and the sk_lock.slock as a 823 * per-address-family lock class: 824 */ 825 lockdep_set_class_and_name(&sk->sk_lock.slock, 826 af_family_slock_keys + sk->sk_family, 827 af_family_slock_key_strings[sk->sk_family]); 828 lockdep_init_map(&sk->sk_lock.dep_map, 829 af_family_key_strings[sk->sk_family], 830 af_family_keys + sk->sk_family); 831 } 832 833 /** 834 * sk_alloc - All socket objects are allocated here 835 * @family: protocol family 836 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc) 837 * @prot: struct proto associated with this new sock instance 838 * @zero_it: if we should zero the newly allocated sock 839 */ 840 struct sock *sk_alloc(int family, gfp_t priority, 841 struct proto *prot, int zero_it) 842 { 843 struct sock *sk = NULL; 844 kmem_cache_t *slab = prot->slab; 845 846 if (slab != NULL) 847 sk = kmem_cache_alloc(slab, priority); 848 else 849 sk = kmalloc(prot->obj_size, priority); 850 851 if (sk) { 852 if (zero_it) { 853 memset(sk, 0, prot->obj_size); 854 sk->sk_family = family; 855 /* 856 * See comment in struct sock definition to understand 857 * why we need sk_prot_creator -acme 858 */ 859 sk->sk_prot = sk->sk_prot_creator = prot; 860 sock_lock_init(sk); 861 } 862 863 if (security_sk_alloc(sk, family, priority)) 864 goto out_free; 865 866 if (!try_module_get(prot->owner)) 867 goto out_free; 868 } 869 return sk; 870 871 out_free: 872 if (slab != NULL) 873 kmem_cache_free(slab, sk); 874 else 875 kfree(sk); 876 return NULL; 877 } 878 879 void sk_free(struct sock *sk) 880 { 881 struct sk_filter *filter; 882 struct module *owner = sk->sk_prot_creator->owner; 883 884 if (sk->sk_destruct) 885 sk->sk_destruct(sk); 886 887 filter = sk->sk_filter; 888 if (filter) { 889 sk_filter_release(sk, filter); 890 sk->sk_filter = NULL; 891 } 892 893 sock_disable_timestamp(sk); 894 895 if (atomic_read(&sk->sk_omem_alloc)) 896 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n", 897 __FUNCTION__, atomic_read(&sk->sk_omem_alloc)); 898 899 security_sk_free(sk); 900 if (sk->sk_prot_creator->slab != NULL) 901 kmem_cache_free(sk->sk_prot_creator->slab, sk); 902 else 903 kfree(sk); 904 module_put(owner); 905 } 906 907 struct sock *sk_clone(const struct sock *sk, const gfp_t priority) 908 { 909 struct sock *newsk = sk_alloc(sk->sk_family, priority, sk->sk_prot, 0); 910 911 if (newsk != NULL) { 912 struct sk_filter *filter; 913 914 memcpy(newsk, sk, sk->sk_prot->obj_size); 915 916 /* SANITY */ 917 sk_node_init(&newsk->sk_node); 918 sock_lock_init(newsk); 919 bh_lock_sock(newsk); 920 921 atomic_set(&newsk->sk_rmem_alloc, 0); 922 atomic_set(&newsk->sk_wmem_alloc, 0); 923 atomic_set(&newsk->sk_omem_alloc, 0); 924 skb_queue_head_init(&newsk->sk_receive_queue); 925 skb_queue_head_init(&newsk->sk_write_queue); 926 #ifdef CONFIG_NET_DMA 927 skb_queue_head_init(&newsk->sk_async_wait_queue); 928 #endif 929 930 rwlock_init(&newsk->sk_dst_lock); 931 rwlock_init(&newsk->sk_callback_lock); 932 lockdep_set_class(&newsk->sk_callback_lock, 933 af_callback_keys + newsk->sk_family); 934 935 newsk->sk_dst_cache = NULL; 936 newsk->sk_wmem_queued = 0; 937 newsk->sk_forward_alloc = 0; 938 newsk->sk_send_head = NULL; 939 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL; 940 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK; 941 942 sock_reset_flag(newsk, SOCK_DONE); 943 skb_queue_head_init(&newsk->sk_error_queue); 944 945 filter = newsk->sk_filter; 946 if (filter != NULL) 947 sk_filter_charge(newsk, filter); 948 949 if (unlikely(xfrm_sk_clone_policy(newsk))) { 950 /* It is still raw copy of parent, so invalidate 951 * destructor and make plain sk_free() */ 952 newsk->sk_destruct = NULL; 953 sk_free(newsk); 954 newsk = NULL; 955 goto out; 956 } 957 958 newsk->sk_err = 0; 959 newsk->sk_priority = 0; 960 atomic_set(&newsk->sk_refcnt, 2); 961 962 /* 963 * Increment the counter in the same struct proto as the master 964 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that 965 * is the same as sk->sk_prot->socks, as this field was copied 966 * with memcpy). 967 * 968 * This _changes_ the previous behaviour, where 969 * tcp_create_openreq_child always was incrementing the 970 * equivalent to tcp_prot->socks (inet_sock_nr), so this have 971 * to be taken into account in all callers. -acme 972 */ 973 sk_refcnt_debug_inc(newsk); 974 newsk->sk_socket = NULL; 975 newsk->sk_sleep = NULL; 976 977 if (newsk->sk_prot->sockets_allocated) 978 atomic_inc(newsk->sk_prot->sockets_allocated); 979 } 980 out: 981 return newsk; 982 } 983 984 EXPORT_SYMBOL_GPL(sk_clone); 985 986 void __init sk_init(void) 987 { 988 if (num_physpages <= 4096) { 989 sysctl_wmem_max = 32767; 990 sysctl_rmem_max = 32767; 991 sysctl_wmem_default = 32767; 992 sysctl_rmem_default = 32767; 993 } else if (num_physpages >= 131072) { 994 sysctl_wmem_max = 131071; 995 sysctl_rmem_max = 131071; 996 } 997 } 998 999 /* 1000 * Simple resource managers for sockets. 1001 */ 1002 1003 1004 /* 1005 * Write buffer destructor automatically called from kfree_skb. 1006 */ 1007 void sock_wfree(struct sk_buff *skb) 1008 { 1009 struct sock *sk = skb->sk; 1010 1011 /* In case it might be waiting for more memory. */ 1012 atomic_sub(skb->truesize, &sk->sk_wmem_alloc); 1013 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) 1014 sk->sk_write_space(sk); 1015 sock_put(sk); 1016 } 1017 1018 /* 1019 * Read buffer destructor automatically called from kfree_skb. 1020 */ 1021 void sock_rfree(struct sk_buff *skb) 1022 { 1023 struct sock *sk = skb->sk; 1024 1025 atomic_sub(skb->truesize, &sk->sk_rmem_alloc); 1026 } 1027 1028 1029 int sock_i_uid(struct sock *sk) 1030 { 1031 int uid; 1032 1033 read_lock(&sk->sk_callback_lock); 1034 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0; 1035 read_unlock(&sk->sk_callback_lock); 1036 return uid; 1037 } 1038 1039 unsigned long sock_i_ino(struct sock *sk) 1040 { 1041 unsigned long ino; 1042 1043 read_lock(&sk->sk_callback_lock); 1044 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0; 1045 read_unlock(&sk->sk_callback_lock); 1046 return ino; 1047 } 1048 1049 /* 1050 * Allocate a skb from the socket's send buffer. 1051 */ 1052 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force, 1053 gfp_t priority) 1054 { 1055 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) { 1056 struct sk_buff * skb = alloc_skb(size, priority); 1057 if (skb) { 1058 skb_set_owner_w(skb, sk); 1059 return skb; 1060 } 1061 } 1062 return NULL; 1063 } 1064 1065 /* 1066 * Allocate a skb from the socket's receive buffer. 1067 */ 1068 struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force, 1069 gfp_t priority) 1070 { 1071 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) { 1072 struct sk_buff *skb = alloc_skb(size, priority); 1073 if (skb) { 1074 skb_set_owner_r(skb, sk); 1075 return skb; 1076 } 1077 } 1078 return NULL; 1079 } 1080 1081 /* 1082 * Allocate a memory block from the socket's option memory buffer. 1083 */ 1084 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority) 1085 { 1086 if ((unsigned)size <= sysctl_optmem_max && 1087 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) { 1088 void *mem; 1089 /* First do the add, to avoid the race if kmalloc 1090 * might sleep. 1091 */ 1092 atomic_add(size, &sk->sk_omem_alloc); 1093 mem = kmalloc(size, priority); 1094 if (mem) 1095 return mem; 1096 atomic_sub(size, &sk->sk_omem_alloc); 1097 } 1098 return NULL; 1099 } 1100 1101 /* 1102 * Free an option memory block. 1103 */ 1104 void sock_kfree_s(struct sock *sk, void *mem, int size) 1105 { 1106 kfree(mem); 1107 atomic_sub(size, &sk->sk_omem_alloc); 1108 } 1109 1110 /* It is almost wait_for_tcp_memory minus release_sock/lock_sock. 1111 I think, these locks should be removed for datagram sockets. 1112 */ 1113 static long sock_wait_for_wmem(struct sock * sk, long timeo) 1114 { 1115 DEFINE_WAIT(wait); 1116 1117 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); 1118 for (;;) { 1119 if (!timeo) 1120 break; 1121 if (signal_pending(current)) 1122 break; 1123 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); 1124 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE); 1125 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) 1126 break; 1127 if (sk->sk_shutdown & SEND_SHUTDOWN) 1128 break; 1129 if (sk->sk_err) 1130 break; 1131 timeo = schedule_timeout(timeo); 1132 } 1133 finish_wait(sk->sk_sleep, &wait); 1134 return timeo; 1135 } 1136 1137 1138 /* 1139 * Generic send/receive buffer handlers 1140 */ 1141 1142 static struct sk_buff *sock_alloc_send_pskb(struct sock *sk, 1143 unsigned long header_len, 1144 unsigned long data_len, 1145 int noblock, int *errcode) 1146 { 1147 struct sk_buff *skb; 1148 gfp_t gfp_mask; 1149 long timeo; 1150 int err; 1151 1152 gfp_mask = sk->sk_allocation; 1153 if (gfp_mask & __GFP_WAIT) 1154 gfp_mask |= __GFP_REPEAT; 1155 1156 timeo = sock_sndtimeo(sk, noblock); 1157 while (1) { 1158 err = sock_error(sk); 1159 if (err != 0) 1160 goto failure; 1161 1162 err = -EPIPE; 1163 if (sk->sk_shutdown & SEND_SHUTDOWN) 1164 goto failure; 1165 1166 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) { 1167 skb = alloc_skb(header_len, sk->sk_allocation); 1168 if (skb) { 1169 int npages; 1170 int i; 1171 1172 /* No pages, we're done... */ 1173 if (!data_len) 1174 break; 1175 1176 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT; 1177 skb->truesize += data_len; 1178 skb_shinfo(skb)->nr_frags = npages; 1179 for (i = 0; i < npages; i++) { 1180 struct page *page; 1181 skb_frag_t *frag; 1182 1183 page = alloc_pages(sk->sk_allocation, 0); 1184 if (!page) { 1185 err = -ENOBUFS; 1186 skb_shinfo(skb)->nr_frags = i; 1187 kfree_skb(skb); 1188 goto failure; 1189 } 1190 1191 frag = &skb_shinfo(skb)->frags[i]; 1192 frag->page = page; 1193 frag->page_offset = 0; 1194 frag->size = (data_len >= PAGE_SIZE ? 1195 PAGE_SIZE : 1196 data_len); 1197 data_len -= PAGE_SIZE; 1198 } 1199 1200 /* Full success... */ 1201 break; 1202 } 1203 err = -ENOBUFS; 1204 goto failure; 1205 } 1206 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); 1207 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); 1208 err = -EAGAIN; 1209 if (!timeo) 1210 goto failure; 1211 if (signal_pending(current)) 1212 goto interrupted; 1213 timeo = sock_wait_for_wmem(sk, timeo); 1214 } 1215 1216 skb_set_owner_w(skb, sk); 1217 return skb; 1218 1219 interrupted: 1220 err = sock_intr_errno(timeo); 1221 failure: 1222 *errcode = err; 1223 return NULL; 1224 } 1225 1226 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size, 1227 int noblock, int *errcode) 1228 { 1229 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode); 1230 } 1231 1232 static void __lock_sock(struct sock *sk) 1233 { 1234 DEFINE_WAIT(wait); 1235 1236 for(;;) { 1237 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait, 1238 TASK_UNINTERRUPTIBLE); 1239 spin_unlock_bh(&sk->sk_lock.slock); 1240 schedule(); 1241 spin_lock_bh(&sk->sk_lock.slock); 1242 if(!sock_owned_by_user(sk)) 1243 break; 1244 } 1245 finish_wait(&sk->sk_lock.wq, &wait); 1246 } 1247 1248 static void __release_sock(struct sock *sk) 1249 { 1250 struct sk_buff *skb = sk->sk_backlog.head; 1251 1252 do { 1253 sk->sk_backlog.head = sk->sk_backlog.tail = NULL; 1254 bh_unlock_sock(sk); 1255 1256 do { 1257 struct sk_buff *next = skb->next; 1258 1259 skb->next = NULL; 1260 sk->sk_backlog_rcv(sk, skb); 1261 1262 /* 1263 * We are in process context here with softirqs 1264 * disabled, use cond_resched_softirq() to preempt. 1265 * This is safe to do because we've taken the backlog 1266 * queue private: 1267 */ 1268 cond_resched_softirq(); 1269 1270 skb = next; 1271 } while (skb != NULL); 1272 1273 bh_lock_sock(sk); 1274 } while((skb = sk->sk_backlog.head) != NULL); 1275 } 1276 1277 /** 1278 * sk_wait_data - wait for data to arrive at sk_receive_queue 1279 * @sk: sock to wait on 1280 * @timeo: for how long 1281 * 1282 * Now socket state including sk->sk_err is changed only under lock, 1283 * hence we may omit checks after joining wait queue. 1284 * We check receive queue before schedule() only as optimization; 1285 * it is very likely that release_sock() added new data. 1286 */ 1287 int sk_wait_data(struct sock *sk, long *timeo) 1288 { 1289 int rc; 1290 DEFINE_WAIT(wait); 1291 1292 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE); 1293 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); 1294 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue)); 1295 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); 1296 finish_wait(sk->sk_sleep, &wait); 1297 return rc; 1298 } 1299 1300 EXPORT_SYMBOL(sk_wait_data); 1301 1302 /* 1303 * Set of default routines for initialising struct proto_ops when 1304 * the protocol does not support a particular function. In certain 1305 * cases where it makes no sense for a protocol to have a "do nothing" 1306 * function, some default processing is provided. 1307 */ 1308 1309 int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len) 1310 { 1311 return -EOPNOTSUPP; 1312 } 1313 1314 int sock_no_connect(struct socket *sock, struct sockaddr *saddr, 1315 int len, int flags) 1316 { 1317 return -EOPNOTSUPP; 1318 } 1319 1320 int sock_no_socketpair(struct socket *sock1, struct socket *sock2) 1321 { 1322 return -EOPNOTSUPP; 1323 } 1324 1325 int sock_no_accept(struct socket *sock, struct socket *newsock, int flags) 1326 { 1327 return -EOPNOTSUPP; 1328 } 1329 1330 int sock_no_getname(struct socket *sock, struct sockaddr *saddr, 1331 int *len, int peer) 1332 { 1333 return -EOPNOTSUPP; 1334 } 1335 1336 unsigned int sock_no_poll(struct file * file, struct socket *sock, poll_table *pt) 1337 { 1338 return 0; 1339 } 1340 1341 int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 1342 { 1343 return -EOPNOTSUPP; 1344 } 1345 1346 int sock_no_listen(struct socket *sock, int backlog) 1347 { 1348 return -EOPNOTSUPP; 1349 } 1350 1351 int sock_no_shutdown(struct socket *sock, int how) 1352 { 1353 return -EOPNOTSUPP; 1354 } 1355 1356 int sock_no_setsockopt(struct socket *sock, int level, int optname, 1357 char __user *optval, int optlen) 1358 { 1359 return -EOPNOTSUPP; 1360 } 1361 1362 int sock_no_getsockopt(struct socket *sock, int level, int optname, 1363 char __user *optval, int __user *optlen) 1364 { 1365 return -EOPNOTSUPP; 1366 } 1367 1368 int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m, 1369 size_t len) 1370 { 1371 return -EOPNOTSUPP; 1372 } 1373 1374 int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m, 1375 size_t len, int flags) 1376 { 1377 return -EOPNOTSUPP; 1378 } 1379 1380 int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma) 1381 { 1382 /* Mirror missing mmap method error code */ 1383 return -ENODEV; 1384 } 1385 1386 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags) 1387 { 1388 ssize_t res; 1389 struct msghdr msg = {.msg_flags = flags}; 1390 struct kvec iov; 1391 char *kaddr = kmap(page); 1392 iov.iov_base = kaddr + offset; 1393 iov.iov_len = size; 1394 res = kernel_sendmsg(sock, &msg, &iov, 1, size); 1395 kunmap(page); 1396 return res; 1397 } 1398 1399 /* 1400 * Default Socket Callbacks 1401 */ 1402 1403 static void sock_def_wakeup(struct sock *sk) 1404 { 1405 read_lock(&sk->sk_callback_lock); 1406 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) 1407 wake_up_interruptible_all(sk->sk_sleep); 1408 read_unlock(&sk->sk_callback_lock); 1409 } 1410 1411 static void sock_def_error_report(struct sock *sk) 1412 { 1413 read_lock(&sk->sk_callback_lock); 1414 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) 1415 wake_up_interruptible(sk->sk_sleep); 1416 sk_wake_async(sk,0,POLL_ERR); 1417 read_unlock(&sk->sk_callback_lock); 1418 } 1419 1420 static void sock_def_readable(struct sock *sk, int len) 1421 { 1422 read_lock(&sk->sk_callback_lock); 1423 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) 1424 wake_up_interruptible(sk->sk_sleep); 1425 sk_wake_async(sk,1,POLL_IN); 1426 read_unlock(&sk->sk_callback_lock); 1427 } 1428 1429 static void sock_def_write_space(struct sock *sk) 1430 { 1431 read_lock(&sk->sk_callback_lock); 1432 1433 /* Do not wake up a writer until he can make "significant" 1434 * progress. --DaveM 1435 */ 1436 if((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) { 1437 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) 1438 wake_up_interruptible(sk->sk_sleep); 1439 1440 /* Should agree with poll, otherwise some programs break */ 1441 if (sock_writeable(sk)) 1442 sk_wake_async(sk, 2, POLL_OUT); 1443 } 1444 1445 read_unlock(&sk->sk_callback_lock); 1446 } 1447 1448 static void sock_def_destruct(struct sock *sk) 1449 { 1450 kfree(sk->sk_protinfo); 1451 } 1452 1453 void sk_send_sigurg(struct sock *sk) 1454 { 1455 if (sk->sk_socket && sk->sk_socket->file) 1456 if (send_sigurg(&sk->sk_socket->file->f_owner)) 1457 sk_wake_async(sk, 3, POLL_PRI); 1458 } 1459 1460 void sk_reset_timer(struct sock *sk, struct timer_list* timer, 1461 unsigned long expires) 1462 { 1463 if (!mod_timer(timer, expires)) 1464 sock_hold(sk); 1465 } 1466 1467 EXPORT_SYMBOL(sk_reset_timer); 1468 1469 void sk_stop_timer(struct sock *sk, struct timer_list* timer) 1470 { 1471 if (timer_pending(timer) && del_timer(timer)) 1472 __sock_put(sk); 1473 } 1474 1475 EXPORT_SYMBOL(sk_stop_timer); 1476 1477 void sock_init_data(struct socket *sock, struct sock *sk) 1478 { 1479 skb_queue_head_init(&sk->sk_receive_queue); 1480 skb_queue_head_init(&sk->sk_write_queue); 1481 skb_queue_head_init(&sk->sk_error_queue); 1482 #ifdef CONFIG_NET_DMA 1483 skb_queue_head_init(&sk->sk_async_wait_queue); 1484 #endif 1485 1486 sk->sk_send_head = NULL; 1487 1488 init_timer(&sk->sk_timer); 1489 1490 sk->sk_allocation = GFP_KERNEL; 1491 sk->sk_rcvbuf = sysctl_rmem_default; 1492 sk->sk_sndbuf = sysctl_wmem_default; 1493 sk->sk_state = TCP_CLOSE; 1494 sk->sk_socket = sock; 1495 1496 sock_set_flag(sk, SOCK_ZAPPED); 1497 1498 if(sock) 1499 { 1500 sk->sk_type = sock->type; 1501 sk->sk_sleep = &sock->wait; 1502 sock->sk = sk; 1503 } else 1504 sk->sk_sleep = NULL; 1505 1506 rwlock_init(&sk->sk_dst_lock); 1507 rwlock_init(&sk->sk_callback_lock); 1508 lockdep_set_class(&sk->sk_callback_lock, 1509 af_callback_keys + sk->sk_family); 1510 1511 sk->sk_state_change = sock_def_wakeup; 1512 sk->sk_data_ready = sock_def_readable; 1513 sk->sk_write_space = sock_def_write_space; 1514 sk->sk_error_report = sock_def_error_report; 1515 sk->sk_destruct = sock_def_destruct; 1516 1517 sk->sk_sndmsg_page = NULL; 1518 sk->sk_sndmsg_off = 0; 1519 1520 sk->sk_peercred.pid = 0; 1521 sk->sk_peercred.uid = -1; 1522 sk->sk_peercred.gid = -1; 1523 sk->sk_write_pending = 0; 1524 sk->sk_rcvlowat = 1; 1525 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT; 1526 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT; 1527 1528 sk->sk_stamp.tv_sec = -1L; 1529 sk->sk_stamp.tv_usec = -1L; 1530 1531 atomic_set(&sk->sk_refcnt, 1); 1532 } 1533 1534 void fastcall lock_sock(struct sock *sk) 1535 { 1536 might_sleep(); 1537 spin_lock_bh(&sk->sk_lock.slock); 1538 if (sk->sk_lock.owner) 1539 __lock_sock(sk); 1540 sk->sk_lock.owner = (void *)1; 1541 spin_unlock(&sk->sk_lock.slock); 1542 /* 1543 * The sk_lock has mutex_lock() semantics here: 1544 */ 1545 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_); 1546 local_bh_enable(); 1547 } 1548 1549 EXPORT_SYMBOL(lock_sock); 1550 1551 void fastcall release_sock(struct sock *sk) 1552 { 1553 /* 1554 * The sk_lock has mutex_unlock() semantics: 1555 */ 1556 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_); 1557 1558 spin_lock_bh(&sk->sk_lock.slock); 1559 if (sk->sk_backlog.tail) 1560 __release_sock(sk); 1561 sk->sk_lock.owner = NULL; 1562 if (waitqueue_active(&sk->sk_lock.wq)) 1563 wake_up(&sk->sk_lock.wq); 1564 spin_unlock_bh(&sk->sk_lock.slock); 1565 } 1566 EXPORT_SYMBOL(release_sock); 1567 1568 int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp) 1569 { 1570 if (!sock_flag(sk, SOCK_TIMESTAMP)) 1571 sock_enable_timestamp(sk); 1572 if (sk->sk_stamp.tv_sec == -1) 1573 return -ENOENT; 1574 if (sk->sk_stamp.tv_sec == 0) 1575 do_gettimeofday(&sk->sk_stamp); 1576 return copy_to_user(userstamp, &sk->sk_stamp, sizeof(struct timeval)) ? 1577 -EFAULT : 0; 1578 } 1579 EXPORT_SYMBOL(sock_get_timestamp); 1580 1581 void sock_enable_timestamp(struct sock *sk) 1582 { 1583 if (!sock_flag(sk, SOCK_TIMESTAMP)) { 1584 sock_set_flag(sk, SOCK_TIMESTAMP); 1585 net_enable_timestamp(); 1586 } 1587 } 1588 EXPORT_SYMBOL(sock_enable_timestamp); 1589 1590 /* 1591 * Get a socket option on an socket. 1592 * 1593 * FIX: POSIX 1003.1g is very ambiguous here. It states that 1594 * asynchronous errors should be reported by getsockopt. We assume 1595 * this means if you specify SO_ERROR (otherwise whats the point of it). 1596 */ 1597 int sock_common_getsockopt(struct socket *sock, int level, int optname, 1598 char __user *optval, int __user *optlen) 1599 { 1600 struct sock *sk = sock->sk; 1601 1602 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen); 1603 } 1604 1605 EXPORT_SYMBOL(sock_common_getsockopt); 1606 1607 #ifdef CONFIG_COMPAT 1608 int compat_sock_common_getsockopt(struct socket *sock, int level, int optname, 1609 char __user *optval, int __user *optlen) 1610 { 1611 struct sock *sk = sock->sk; 1612 1613 if (sk->sk_prot->compat_setsockopt != NULL) 1614 return sk->sk_prot->compat_getsockopt(sk, level, optname, 1615 optval, optlen); 1616 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen); 1617 } 1618 EXPORT_SYMBOL(compat_sock_common_getsockopt); 1619 #endif 1620 1621 int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, 1622 struct msghdr *msg, size_t size, int flags) 1623 { 1624 struct sock *sk = sock->sk; 1625 int addr_len = 0; 1626 int err; 1627 1628 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT, 1629 flags & ~MSG_DONTWAIT, &addr_len); 1630 if (err >= 0) 1631 msg->msg_namelen = addr_len; 1632 return err; 1633 } 1634 1635 EXPORT_SYMBOL(sock_common_recvmsg); 1636 1637 /* 1638 * Set socket options on an inet socket. 1639 */ 1640 int sock_common_setsockopt(struct socket *sock, int level, int optname, 1641 char __user *optval, int optlen) 1642 { 1643 struct sock *sk = sock->sk; 1644 1645 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen); 1646 } 1647 1648 EXPORT_SYMBOL(sock_common_setsockopt); 1649 1650 #ifdef CONFIG_COMPAT 1651 int compat_sock_common_setsockopt(struct socket *sock, int level, int optname, 1652 char __user *optval, int optlen) 1653 { 1654 struct sock *sk = sock->sk; 1655 1656 if (sk->sk_prot->compat_setsockopt != NULL) 1657 return sk->sk_prot->compat_setsockopt(sk, level, optname, 1658 optval, optlen); 1659 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen); 1660 } 1661 EXPORT_SYMBOL(compat_sock_common_setsockopt); 1662 #endif 1663 1664 void sk_common_release(struct sock *sk) 1665 { 1666 if (sk->sk_prot->destroy) 1667 sk->sk_prot->destroy(sk); 1668 1669 /* 1670 * Observation: when sock_common_release is called, processes have 1671 * no access to socket. But net still has. 1672 * Step one, detach it from networking: 1673 * 1674 * A. Remove from hash tables. 1675 */ 1676 1677 sk->sk_prot->unhash(sk); 1678 1679 /* 1680 * In this point socket cannot receive new packets, but it is possible 1681 * that some packets are in flight because some CPU runs receiver and 1682 * did hash table lookup before we unhashed socket. They will achieve 1683 * receive queue and will be purged by socket destructor. 1684 * 1685 * Also we still have packets pending on receive queue and probably, 1686 * our own packets waiting in device queues. sock_destroy will drain 1687 * receive queue, but transmitted packets will delay socket destruction 1688 * until the last reference will be released. 1689 */ 1690 1691 sock_orphan(sk); 1692 1693 xfrm_sk_free_policy(sk); 1694 1695 sk_refcnt_debug_release(sk); 1696 sock_put(sk); 1697 } 1698 1699 EXPORT_SYMBOL(sk_common_release); 1700 1701 static DEFINE_RWLOCK(proto_list_lock); 1702 static LIST_HEAD(proto_list); 1703 1704 int proto_register(struct proto *prot, int alloc_slab) 1705 { 1706 char *request_sock_slab_name = NULL; 1707 char *timewait_sock_slab_name; 1708 int rc = -ENOBUFS; 1709 1710 if (alloc_slab) { 1711 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0, 1712 SLAB_HWCACHE_ALIGN, NULL, NULL); 1713 1714 if (prot->slab == NULL) { 1715 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n", 1716 prot->name); 1717 goto out; 1718 } 1719 1720 if (prot->rsk_prot != NULL) { 1721 static const char mask[] = "request_sock_%s"; 1722 1723 request_sock_slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL); 1724 if (request_sock_slab_name == NULL) 1725 goto out_free_sock_slab; 1726 1727 sprintf(request_sock_slab_name, mask, prot->name); 1728 prot->rsk_prot->slab = kmem_cache_create(request_sock_slab_name, 1729 prot->rsk_prot->obj_size, 0, 1730 SLAB_HWCACHE_ALIGN, NULL, NULL); 1731 1732 if (prot->rsk_prot->slab == NULL) { 1733 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n", 1734 prot->name); 1735 goto out_free_request_sock_slab_name; 1736 } 1737 } 1738 1739 if (prot->twsk_prot != NULL) { 1740 static const char mask[] = "tw_sock_%s"; 1741 1742 timewait_sock_slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL); 1743 1744 if (timewait_sock_slab_name == NULL) 1745 goto out_free_request_sock_slab; 1746 1747 sprintf(timewait_sock_slab_name, mask, prot->name); 1748 prot->twsk_prot->twsk_slab = 1749 kmem_cache_create(timewait_sock_slab_name, 1750 prot->twsk_prot->twsk_obj_size, 1751 0, SLAB_HWCACHE_ALIGN, 1752 NULL, NULL); 1753 if (prot->twsk_prot->twsk_slab == NULL) 1754 goto out_free_timewait_sock_slab_name; 1755 } 1756 } 1757 1758 write_lock(&proto_list_lock); 1759 list_add(&prot->node, &proto_list); 1760 write_unlock(&proto_list_lock); 1761 rc = 0; 1762 out: 1763 return rc; 1764 out_free_timewait_sock_slab_name: 1765 kfree(timewait_sock_slab_name); 1766 out_free_request_sock_slab: 1767 if (prot->rsk_prot && prot->rsk_prot->slab) { 1768 kmem_cache_destroy(prot->rsk_prot->slab); 1769 prot->rsk_prot->slab = NULL; 1770 } 1771 out_free_request_sock_slab_name: 1772 kfree(request_sock_slab_name); 1773 out_free_sock_slab: 1774 kmem_cache_destroy(prot->slab); 1775 prot->slab = NULL; 1776 goto out; 1777 } 1778 1779 EXPORT_SYMBOL(proto_register); 1780 1781 void proto_unregister(struct proto *prot) 1782 { 1783 write_lock(&proto_list_lock); 1784 list_del(&prot->node); 1785 write_unlock(&proto_list_lock); 1786 1787 if (prot->slab != NULL) { 1788 kmem_cache_destroy(prot->slab); 1789 prot->slab = NULL; 1790 } 1791 1792 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) { 1793 const char *name = kmem_cache_name(prot->rsk_prot->slab); 1794 1795 kmem_cache_destroy(prot->rsk_prot->slab); 1796 kfree(name); 1797 prot->rsk_prot->slab = NULL; 1798 } 1799 1800 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) { 1801 const char *name = kmem_cache_name(prot->twsk_prot->twsk_slab); 1802 1803 kmem_cache_destroy(prot->twsk_prot->twsk_slab); 1804 kfree(name); 1805 prot->twsk_prot->twsk_slab = NULL; 1806 } 1807 } 1808 1809 EXPORT_SYMBOL(proto_unregister); 1810 1811 #ifdef CONFIG_PROC_FS 1812 static inline struct proto *__proto_head(void) 1813 { 1814 return list_entry(proto_list.next, struct proto, node); 1815 } 1816 1817 static inline struct proto *proto_head(void) 1818 { 1819 return list_empty(&proto_list) ? NULL : __proto_head(); 1820 } 1821 1822 static inline struct proto *proto_next(struct proto *proto) 1823 { 1824 return proto->node.next == &proto_list ? NULL : 1825 list_entry(proto->node.next, struct proto, node); 1826 } 1827 1828 static inline struct proto *proto_get_idx(loff_t pos) 1829 { 1830 struct proto *proto; 1831 loff_t i = 0; 1832 1833 list_for_each_entry(proto, &proto_list, node) 1834 if (i++ == pos) 1835 goto out; 1836 1837 proto = NULL; 1838 out: 1839 return proto; 1840 } 1841 1842 static void *proto_seq_start(struct seq_file *seq, loff_t *pos) 1843 { 1844 read_lock(&proto_list_lock); 1845 return *pos ? proto_get_idx(*pos - 1) : SEQ_START_TOKEN; 1846 } 1847 1848 static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1849 { 1850 ++*pos; 1851 return v == SEQ_START_TOKEN ? proto_head() : proto_next(v); 1852 } 1853 1854 static void proto_seq_stop(struct seq_file *seq, void *v) 1855 { 1856 read_unlock(&proto_list_lock); 1857 } 1858 1859 static char proto_method_implemented(const void *method) 1860 { 1861 return method == NULL ? 'n' : 'y'; 1862 } 1863 1864 static void proto_seq_printf(struct seq_file *seq, struct proto *proto) 1865 { 1866 seq_printf(seq, "%-9s %4u %6d %6d %-3s %6u %-3s %-10s " 1867 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n", 1868 proto->name, 1869 proto->obj_size, 1870 proto->sockets_allocated != NULL ? atomic_read(proto->sockets_allocated) : -1, 1871 proto->memory_allocated != NULL ? atomic_read(proto->memory_allocated) : -1, 1872 proto->memory_pressure != NULL ? *proto->memory_pressure ? "yes" : "no" : "NI", 1873 proto->max_header, 1874 proto->slab == NULL ? "no" : "yes", 1875 module_name(proto->owner), 1876 proto_method_implemented(proto->close), 1877 proto_method_implemented(proto->connect), 1878 proto_method_implemented(proto->disconnect), 1879 proto_method_implemented(proto->accept), 1880 proto_method_implemented(proto->ioctl), 1881 proto_method_implemented(proto->init), 1882 proto_method_implemented(proto->destroy), 1883 proto_method_implemented(proto->shutdown), 1884 proto_method_implemented(proto->setsockopt), 1885 proto_method_implemented(proto->getsockopt), 1886 proto_method_implemented(proto->sendmsg), 1887 proto_method_implemented(proto->recvmsg), 1888 proto_method_implemented(proto->sendpage), 1889 proto_method_implemented(proto->bind), 1890 proto_method_implemented(proto->backlog_rcv), 1891 proto_method_implemented(proto->hash), 1892 proto_method_implemented(proto->unhash), 1893 proto_method_implemented(proto->get_port), 1894 proto_method_implemented(proto->enter_memory_pressure)); 1895 } 1896 1897 static int proto_seq_show(struct seq_file *seq, void *v) 1898 { 1899 if (v == SEQ_START_TOKEN) 1900 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s", 1901 "protocol", 1902 "size", 1903 "sockets", 1904 "memory", 1905 "press", 1906 "maxhdr", 1907 "slab", 1908 "module", 1909 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n"); 1910 else 1911 proto_seq_printf(seq, v); 1912 return 0; 1913 } 1914 1915 static struct seq_operations proto_seq_ops = { 1916 .start = proto_seq_start, 1917 .next = proto_seq_next, 1918 .stop = proto_seq_stop, 1919 .show = proto_seq_show, 1920 }; 1921 1922 static int proto_seq_open(struct inode *inode, struct file *file) 1923 { 1924 return seq_open(file, &proto_seq_ops); 1925 } 1926 1927 static struct file_operations proto_seq_fops = { 1928 .owner = THIS_MODULE, 1929 .open = proto_seq_open, 1930 .read = seq_read, 1931 .llseek = seq_lseek, 1932 .release = seq_release, 1933 }; 1934 1935 static int __init proto_init(void) 1936 { 1937 /* register /proc/net/protocols */ 1938 return proc_net_fops_create("protocols", S_IRUGO, &proto_seq_fops) == NULL ? -ENOBUFS : 0; 1939 } 1940 1941 subsys_initcall(proto_init); 1942 1943 #endif /* PROC_FS */ 1944 1945 EXPORT_SYMBOL(sk_alloc); 1946 EXPORT_SYMBOL(sk_free); 1947 EXPORT_SYMBOL(sk_send_sigurg); 1948 EXPORT_SYMBOL(sock_alloc_send_skb); 1949 EXPORT_SYMBOL(sock_init_data); 1950 EXPORT_SYMBOL(sock_kfree_s); 1951 EXPORT_SYMBOL(sock_kmalloc); 1952 EXPORT_SYMBOL(sock_no_accept); 1953 EXPORT_SYMBOL(sock_no_bind); 1954 EXPORT_SYMBOL(sock_no_connect); 1955 EXPORT_SYMBOL(sock_no_getname); 1956 EXPORT_SYMBOL(sock_no_getsockopt); 1957 EXPORT_SYMBOL(sock_no_ioctl); 1958 EXPORT_SYMBOL(sock_no_listen); 1959 EXPORT_SYMBOL(sock_no_mmap); 1960 EXPORT_SYMBOL(sock_no_poll); 1961 EXPORT_SYMBOL(sock_no_recvmsg); 1962 EXPORT_SYMBOL(sock_no_sendmsg); 1963 EXPORT_SYMBOL(sock_no_sendpage); 1964 EXPORT_SYMBOL(sock_no_setsockopt); 1965 EXPORT_SYMBOL(sock_no_shutdown); 1966 EXPORT_SYMBOL(sock_no_socketpair); 1967 EXPORT_SYMBOL(sock_rfree); 1968 EXPORT_SYMBOL(sock_setsockopt); 1969 EXPORT_SYMBOL(sock_wfree); 1970 EXPORT_SYMBOL(sock_wmalloc); 1971 EXPORT_SYMBOL(sock_i_uid); 1972 EXPORT_SYMBOL(sock_i_ino); 1973 EXPORT_SYMBOL(sysctl_optmem_max); 1974 #ifdef CONFIG_SYSCTL 1975 EXPORT_SYMBOL(sysctl_rmem_max); 1976 EXPORT_SYMBOL(sysctl_wmem_max); 1977 #endif 1978