1 /* 2 * NET An implementation of the SOCKET network access protocol. 3 * 4 * Version: @(#)socket.c 1.1.93 18/02/95 5 * 6 * Authors: Orest Zborowski, <[email protected]> 7 * Ross Biro 8 * Fred N. van Kempen, <[email protected]> 9 * 10 * Fixes: 11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in 12 * shutdown() 13 * Alan Cox : verify_area() fixes 14 * Alan Cox : Removed DDI 15 * Jonathan Kamens : SOCK_DGRAM reconnect bug 16 * Alan Cox : Moved a load of checks to the very 17 * top level. 18 * Alan Cox : Move address structures to/from user 19 * mode above the protocol layers. 20 * Rob Janssen : Allow 0 length sends. 21 * Alan Cox : Asynchronous I/O support (cribbed from the 22 * tty drivers). 23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style) 24 * Jeff Uphoff : Made max number of sockets command-line 25 * configurable. 26 * Matti Aarnio : Made the number of sockets dynamic, 27 * to be allocated when needed, and mr. 28 * Uphoff's max is used as max to be 29 * allowed to allocate. 30 * Linus : Argh. removed all the socket allocation 31 * altogether: it's in the inode now. 32 * Alan Cox : Made sock_alloc()/sock_release() public 33 * for NetROM and future kernel nfsd type 34 * stuff. 35 * Alan Cox : sendmsg/recvmsg basics. 36 * Tom Dyas : Export net symbols. 37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n". 38 * Alan Cox : Added thread locking to sys_* calls 39 * for sockets. May have errors at the 40 * moment. 41 * Kevin Buhr : Fixed the dumb errors in the above. 42 * Andi Kleen : Some small cleanups, optimizations, 43 * and fixed a copy_from_user() bug. 44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0) 45 * Tigran Aivazian : Made listen(2) backlog sanity checks 46 * protocol-independent 47 * 48 * 49 * This program is free software; you can redistribute it and/or 50 * modify it under the terms of the GNU General Public License 51 * as published by the Free Software Foundation; either version 52 * 2 of the License, or (at your option) any later version. 53 * 54 * 55 * This module is effectively the top level interface to the BSD socket 56 * paradigm. 57 * 58 * Based upon Swansea University Computer Society NET3.039 59 */ 60 61 #include <linux/mm.h> 62 #include <linux/socket.h> 63 #include <linux/file.h> 64 #include <linux/net.h> 65 #include <linux/interrupt.h> 66 #include <linux/thread_info.h> 67 #include <linux/rcupdate.h> 68 #include <linux/netdevice.h> 69 #include <linux/proc_fs.h> 70 #include <linux/seq_file.h> 71 #include <linux/mutex.h> 72 #include <linux/wanrouter.h> 73 #include <linux/if_bridge.h> 74 #include <linux/if_frad.h> 75 #include <linux/if_vlan.h> 76 #include <linux/init.h> 77 #include <linux/poll.h> 78 #include <linux/cache.h> 79 #include <linux/module.h> 80 #include <linux/highmem.h> 81 #include <linux/mount.h> 82 #include <linux/security.h> 83 #include <linux/syscalls.h> 84 #include <linux/compat.h> 85 #include <linux/kmod.h> 86 #include <linux/audit.h> 87 #include <linux/wireless.h> 88 #include <linux/nsproxy.h> 89 #include <linux/magic.h> 90 #include <linux/slab.h> 91 92 #include <asm/uaccess.h> 93 #include <asm/unistd.h> 94 95 #include <net/compat.h> 96 #include <net/wext.h> 97 #include <net/cls_cgroup.h> 98 99 #include <net/sock.h> 100 #include <linux/netfilter.h> 101 102 #include <linux/if_tun.h> 103 #include <linux/ipv6_route.h> 104 #include <linux/route.h> 105 #include <linux/sockios.h> 106 #include <linux/atalk.h> 107 108 static int sock_no_open(struct inode *irrelevant, struct file *dontcare); 109 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, 110 unsigned long nr_segs, loff_t pos); 111 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, 112 unsigned long nr_segs, loff_t pos); 113 static int sock_mmap(struct file *file, struct vm_area_struct *vma); 114 115 static int sock_close(struct inode *inode, struct file *file); 116 static unsigned int sock_poll(struct file *file, 117 struct poll_table_struct *wait); 118 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 119 #ifdef CONFIG_COMPAT 120 static long compat_sock_ioctl(struct file *file, 121 unsigned int cmd, unsigned long arg); 122 #endif 123 static int sock_fasync(int fd, struct file *filp, int on); 124 static ssize_t sock_sendpage(struct file *file, struct page *page, 125 int offset, size_t size, loff_t *ppos, int more); 126 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 127 struct pipe_inode_info *pipe, size_t len, 128 unsigned int flags); 129 130 /* 131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear 132 * in the operation structures but are done directly via the socketcall() multiplexor. 133 */ 134 135 static const struct file_operations socket_file_ops = { 136 .owner = THIS_MODULE, 137 .llseek = no_llseek, 138 .aio_read = sock_aio_read, 139 .aio_write = sock_aio_write, 140 .poll = sock_poll, 141 .unlocked_ioctl = sock_ioctl, 142 #ifdef CONFIG_COMPAT 143 .compat_ioctl = compat_sock_ioctl, 144 #endif 145 .mmap = sock_mmap, 146 .open = sock_no_open, /* special open code to disallow open via /proc */ 147 .release = sock_close, 148 .fasync = sock_fasync, 149 .sendpage = sock_sendpage, 150 .splice_write = generic_splice_sendpage, 151 .splice_read = sock_splice_read, 152 }; 153 154 /* 155 * The protocol list. Each protocol is registered in here. 156 */ 157 158 static DEFINE_SPINLOCK(net_family_lock); 159 static const struct net_proto_family *net_families[NPROTO] __read_mostly; 160 161 /* 162 * Statistics counters of the socket lists 163 */ 164 165 static DEFINE_PER_CPU(int, sockets_in_use); 166 167 /* 168 * Support routines. 169 * Move socket addresses back and forth across the kernel/user 170 * divide and look after the messy bits. 171 */ 172 173 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain - 174 16 for IP, 16 for IPX, 175 24 for IPv6, 176 about 80 for AX.25 177 must be at least one bigger than 178 the AF_UNIX size (see net/unix/af_unix.c 179 :unix_mkname()). 180 */ 181 182 /** 183 * move_addr_to_kernel - copy a socket address into kernel space 184 * @uaddr: Address in user space 185 * @kaddr: Address in kernel space 186 * @ulen: Length in user space 187 * 188 * The address is copied into kernel space. If the provided address is 189 * too long an error code of -EINVAL is returned. If the copy gives 190 * invalid addresses -EFAULT is returned. On a success 0 is returned. 191 */ 192 193 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr) 194 { 195 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) 196 return -EINVAL; 197 if (ulen == 0) 198 return 0; 199 if (copy_from_user(kaddr, uaddr, ulen)) 200 return -EFAULT; 201 return audit_sockaddr(ulen, kaddr); 202 } 203 204 /** 205 * move_addr_to_user - copy an address to user space 206 * @kaddr: kernel space address 207 * @klen: length of address in kernel 208 * @uaddr: user space address 209 * @ulen: pointer to user length field 210 * 211 * The value pointed to by ulen on entry is the buffer length available. 212 * This is overwritten with the buffer space used. -EINVAL is returned 213 * if an overlong buffer is specified or a negative buffer size. -EFAULT 214 * is returned if either the buffer or the length field are not 215 * accessible. 216 * After copying the data up to the limit the user specifies, the true 217 * length of the data is written over the length limit the user 218 * specified. Zero is returned for a success. 219 */ 220 221 int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr, 222 int __user *ulen) 223 { 224 int err; 225 int len; 226 227 err = get_user(len, ulen); 228 if (err) 229 return err; 230 if (len > klen) 231 len = klen; 232 if (len < 0 || len > sizeof(struct sockaddr_storage)) 233 return -EINVAL; 234 if (len) { 235 if (audit_sockaddr(klen, kaddr)) 236 return -ENOMEM; 237 if (copy_to_user(uaddr, kaddr, len)) 238 return -EFAULT; 239 } 240 /* 241 * "fromlen shall refer to the value before truncation.." 242 * 1003.1g 243 */ 244 return __put_user(klen, ulen); 245 } 246 247 static struct kmem_cache *sock_inode_cachep __read_mostly; 248 249 static struct inode *sock_alloc_inode(struct super_block *sb) 250 { 251 struct socket_alloc *ei; 252 253 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); 254 if (!ei) 255 return NULL; 256 ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL); 257 if (!ei->socket.wq) { 258 kmem_cache_free(sock_inode_cachep, ei); 259 return NULL; 260 } 261 init_waitqueue_head(&ei->socket.wq->wait); 262 ei->socket.wq->fasync_list = NULL; 263 264 ei->socket.state = SS_UNCONNECTED; 265 ei->socket.flags = 0; 266 ei->socket.ops = NULL; 267 ei->socket.sk = NULL; 268 ei->socket.file = NULL; 269 270 return &ei->vfs_inode; 271 } 272 273 274 static void wq_free_rcu(struct rcu_head *head) 275 { 276 struct socket_wq *wq = container_of(head, struct socket_wq, rcu); 277 278 kfree(wq); 279 } 280 281 static void sock_destroy_inode(struct inode *inode) 282 { 283 struct socket_alloc *ei; 284 285 ei = container_of(inode, struct socket_alloc, vfs_inode); 286 call_rcu(&ei->socket.wq->rcu, wq_free_rcu); 287 kmem_cache_free(sock_inode_cachep, ei); 288 } 289 290 static void init_once(void *foo) 291 { 292 struct socket_alloc *ei = (struct socket_alloc *)foo; 293 294 inode_init_once(&ei->vfs_inode); 295 } 296 297 static int init_inodecache(void) 298 { 299 sock_inode_cachep = kmem_cache_create("sock_inode_cache", 300 sizeof(struct socket_alloc), 301 0, 302 (SLAB_HWCACHE_ALIGN | 303 SLAB_RECLAIM_ACCOUNT | 304 SLAB_MEM_SPREAD), 305 init_once); 306 if (sock_inode_cachep == NULL) 307 return -ENOMEM; 308 return 0; 309 } 310 311 static const struct super_operations sockfs_ops = { 312 .alloc_inode = sock_alloc_inode, 313 .destroy_inode = sock_destroy_inode, 314 .statfs = simple_statfs, 315 }; 316 317 static int sockfs_get_sb(struct file_system_type *fs_type, 318 int flags, const char *dev_name, void *data, 319 struct vfsmount *mnt) 320 { 321 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC, 322 mnt); 323 } 324 325 static struct vfsmount *sock_mnt __read_mostly; 326 327 static struct file_system_type sock_fs_type = { 328 .name = "sockfs", 329 .get_sb = sockfs_get_sb, 330 .kill_sb = kill_anon_super, 331 }; 332 333 /* 334 * sockfs_dname() is called from d_path(). 335 */ 336 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) 337 { 338 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", 339 dentry->d_inode->i_ino); 340 } 341 342 static const struct dentry_operations sockfs_dentry_operations = { 343 .d_dname = sockfs_dname, 344 }; 345 346 /* 347 * Obtains the first available file descriptor and sets it up for use. 348 * 349 * These functions create file structures and maps them to fd space 350 * of the current process. On success it returns file descriptor 351 * and file struct implicitly stored in sock->file. 352 * Note that another thread may close file descriptor before we return 353 * from this function. We use the fact that now we do not refer 354 * to socket after mapping. If one day we will need it, this 355 * function will increment ref. count on file by 1. 356 * 357 * In any case returned fd MAY BE not valid! 358 * This race condition is unavoidable 359 * with shared fd spaces, we cannot solve it inside kernel, 360 * but we take care of internal coherence yet. 361 */ 362 363 static int sock_alloc_file(struct socket *sock, struct file **f, int flags) 364 { 365 struct qstr name = { .name = "" }; 366 struct path path; 367 struct file *file; 368 int fd; 369 370 fd = get_unused_fd_flags(flags); 371 if (unlikely(fd < 0)) 372 return fd; 373 374 path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name); 375 if (unlikely(!path.dentry)) { 376 put_unused_fd(fd); 377 return -ENOMEM; 378 } 379 path.mnt = mntget(sock_mnt); 380 381 path.dentry->d_op = &sockfs_dentry_operations; 382 d_instantiate(path.dentry, SOCK_INODE(sock)); 383 SOCK_INODE(sock)->i_fop = &socket_file_ops; 384 385 file = alloc_file(&path, FMODE_READ | FMODE_WRITE, 386 &socket_file_ops); 387 if (unlikely(!file)) { 388 /* drop dentry, keep inode */ 389 atomic_inc(&path.dentry->d_inode->i_count); 390 path_put(&path); 391 put_unused_fd(fd); 392 return -ENFILE; 393 } 394 395 sock->file = file; 396 file->f_flags = O_RDWR | (flags & O_NONBLOCK); 397 file->f_pos = 0; 398 file->private_data = sock; 399 400 *f = file; 401 return fd; 402 } 403 404 int sock_map_fd(struct socket *sock, int flags) 405 { 406 struct file *newfile; 407 int fd = sock_alloc_file(sock, &newfile, flags); 408 409 if (likely(fd >= 0)) 410 fd_install(fd, newfile); 411 412 return fd; 413 } 414 EXPORT_SYMBOL(sock_map_fd); 415 416 static struct socket *sock_from_file(struct file *file, int *err) 417 { 418 if (file->f_op == &socket_file_ops) 419 return file->private_data; /* set in sock_map_fd */ 420 421 *err = -ENOTSOCK; 422 return NULL; 423 } 424 425 /** 426 * sockfd_lookup - Go from a file number to its socket slot 427 * @fd: file handle 428 * @err: pointer to an error code return 429 * 430 * The file handle passed in is locked and the socket it is bound 431 * too is returned. If an error occurs the err pointer is overwritten 432 * with a negative errno code and NULL is returned. The function checks 433 * for both invalid handles and passing a handle which is not a socket. 434 * 435 * On a success the socket object pointer is returned. 436 */ 437 438 struct socket *sockfd_lookup(int fd, int *err) 439 { 440 struct file *file; 441 struct socket *sock; 442 443 file = fget(fd); 444 if (!file) { 445 *err = -EBADF; 446 return NULL; 447 } 448 449 sock = sock_from_file(file, err); 450 if (!sock) 451 fput(file); 452 return sock; 453 } 454 EXPORT_SYMBOL(sockfd_lookup); 455 456 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) 457 { 458 struct file *file; 459 struct socket *sock; 460 461 *err = -EBADF; 462 file = fget_light(fd, fput_needed); 463 if (file) { 464 sock = sock_from_file(file, err); 465 if (sock) 466 return sock; 467 fput_light(file, *fput_needed); 468 } 469 return NULL; 470 } 471 472 /** 473 * sock_alloc - allocate a socket 474 * 475 * Allocate a new inode and socket object. The two are bound together 476 * and initialised. The socket is then returned. If we are out of inodes 477 * NULL is returned. 478 */ 479 480 static struct socket *sock_alloc(void) 481 { 482 struct inode *inode; 483 struct socket *sock; 484 485 inode = new_inode(sock_mnt->mnt_sb); 486 if (!inode) 487 return NULL; 488 489 sock = SOCKET_I(inode); 490 491 kmemcheck_annotate_bitfield(sock, type); 492 inode->i_mode = S_IFSOCK | S_IRWXUGO; 493 inode->i_uid = current_fsuid(); 494 inode->i_gid = current_fsgid(); 495 496 percpu_add(sockets_in_use, 1); 497 return sock; 498 } 499 500 /* 501 * In theory you can't get an open on this inode, but /proc provides 502 * a back door. Remember to keep it shut otherwise you'll let the 503 * creepy crawlies in. 504 */ 505 506 static int sock_no_open(struct inode *irrelevant, struct file *dontcare) 507 { 508 return -ENXIO; 509 } 510 511 const struct file_operations bad_sock_fops = { 512 .owner = THIS_MODULE, 513 .open = sock_no_open, 514 }; 515 516 /** 517 * sock_release - close a socket 518 * @sock: socket to close 519 * 520 * The socket is released from the protocol stack if it has a release 521 * callback, and the inode is then released if the socket is bound to 522 * an inode not a file. 523 */ 524 525 void sock_release(struct socket *sock) 526 { 527 if (sock->ops) { 528 struct module *owner = sock->ops->owner; 529 530 sock->ops->release(sock); 531 sock->ops = NULL; 532 module_put(owner); 533 } 534 535 if (sock->wq->fasync_list) 536 printk(KERN_ERR "sock_release: fasync list not empty!\n"); 537 538 percpu_sub(sockets_in_use, 1); 539 if (!sock->file) { 540 iput(SOCK_INODE(sock)); 541 return; 542 } 543 sock->file = NULL; 544 } 545 EXPORT_SYMBOL(sock_release); 546 547 int sock_tx_timestamp(struct msghdr *msg, struct sock *sk, 548 union skb_shared_tx *shtx) 549 { 550 shtx->flags = 0; 551 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE)) 552 shtx->hardware = 1; 553 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE)) 554 shtx->software = 1; 555 return 0; 556 } 557 EXPORT_SYMBOL(sock_tx_timestamp); 558 559 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock, 560 struct msghdr *msg, size_t size) 561 { 562 struct sock_iocb *si = kiocb_to_siocb(iocb); 563 int err; 564 565 sock_update_classid(sock->sk); 566 567 si->sock = sock; 568 si->scm = NULL; 569 si->msg = msg; 570 si->size = size; 571 572 err = security_socket_sendmsg(sock, msg, size); 573 if (err) 574 return err; 575 576 return sock->ops->sendmsg(iocb, sock, msg, size); 577 } 578 579 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 580 { 581 struct kiocb iocb; 582 struct sock_iocb siocb; 583 int ret; 584 585 init_sync_kiocb(&iocb, NULL); 586 iocb.private = &siocb; 587 ret = __sock_sendmsg(&iocb, sock, msg, size); 588 if (-EIOCBQUEUED == ret) 589 ret = wait_on_sync_kiocb(&iocb); 590 return ret; 591 } 592 EXPORT_SYMBOL(sock_sendmsg); 593 594 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, 595 struct kvec *vec, size_t num, size_t size) 596 { 597 mm_segment_t oldfs = get_fs(); 598 int result; 599 600 set_fs(KERNEL_DS); 601 /* 602 * the following is safe, since for compiler definitions of kvec and 603 * iovec are identical, yielding the same in-core layout and alignment 604 */ 605 msg->msg_iov = (struct iovec *)vec; 606 msg->msg_iovlen = num; 607 result = sock_sendmsg(sock, msg, size); 608 set_fs(oldfs); 609 return result; 610 } 611 EXPORT_SYMBOL(kernel_sendmsg); 612 613 static int ktime2ts(ktime_t kt, struct timespec *ts) 614 { 615 if (kt.tv64) { 616 *ts = ktime_to_timespec(kt); 617 return 1; 618 } else { 619 return 0; 620 } 621 } 622 623 /* 624 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) 625 */ 626 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 627 struct sk_buff *skb) 628 { 629 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); 630 struct timespec ts[3]; 631 int empty = 1; 632 struct skb_shared_hwtstamps *shhwtstamps = 633 skb_hwtstamps(skb); 634 635 /* Race occurred between timestamp enabling and packet 636 receiving. Fill in the current time for now. */ 637 if (need_software_tstamp && skb->tstamp.tv64 == 0) 638 __net_timestamp(skb); 639 640 if (need_software_tstamp) { 641 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { 642 struct timeval tv; 643 skb_get_timestamp(skb, &tv); 644 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, 645 sizeof(tv), &tv); 646 } else { 647 skb_get_timestampns(skb, &ts[0]); 648 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, 649 sizeof(ts[0]), &ts[0]); 650 } 651 } 652 653 654 memset(ts, 0, sizeof(ts)); 655 if (skb->tstamp.tv64 && 656 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) { 657 skb_get_timestampns(skb, ts + 0); 658 empty = 0; 659 } 660 if (shhwtstamps) { 661 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) && 662 ktime2ts(shhwtstamps->syststamp, ts + 1)) 663 empty = 0; 664 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) && 665 ktime2ts(shhwtstamps->hwtstamp, ts + 2)) 666 empty = 0; 667 } 668 if (!empty) 669 put_cmsg(msg, SOL_SOCKET, 670 SCM_TIMESTAMPING, sizeof(ts), &ts); 671 } 672 EXPORT_SYMBOL_GPL(__sock_recv_timestamp); 673 674 inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) 675 { 676 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount) 677 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, 678 sizeof(__u32), &skb->dropcount); 679 } 680 681 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 682 struct sk_buff *skb) 683 { 684 sock_recv_timestamp(msg, sk, skb); 685 sock_recv_drops(msg, sk, skb); 686 } 687 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops); 688 689 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock, 690 struct msghdr *msg, size_t size, int flags) 691 { 692 struct sock_iocb *si = kiocb_to_siocb(iocb); 693 694 sock_update_classid(sock->sk); 695 696 si->sock = sock; 697 si->scm = NULL; 698 si->msg = msg; 699 si->size = size; 700 si->flags = flags; 701 702 return sock->ops->recvmsg(iocb, sock, msg, size, flags); 703 } 704 705 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock, 706 struct msghdr *msg, size_t size, int flags) 707 { 708 int err = security_socket_recvmsg(sock, msg, size, flags); 709 710 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags); 711 } 712 713 int sock_recvmsg(struct socket *sock, struct msghdr *msg, 714 size_t size, int flags) 715 { 716 struct kiocb iocb; 717 struct sock_iocb siocb; 718 int ret; 719 720 init_sync_kiocb(&iocb, NULL); 721 iocb.private = &siocb; 722 ret = __sock_recvmsg(&iocb, sock, msg, size, flags); 723 if (-EIOCBQUEUED == ret) 724 ret = wait_on_sync_kiocb(&iocb); 725 return ret; 726 } 727 EXPORT_SYMBOL(sock_recvmsg); 728 729 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 730 size_t size, int flags) 731 { 732 struct kiocb iocb; 733 struct sock_iocb siocb; 734 int ret; 735 736 init_sync_kiocb(&iocb, NULL); 737 iocb.private = &siocb; 738 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags); 739 if (-EIOCBQUEUED == ret) 740 ret = wait_on_sync_kiocb(&iocb); 741 return ret; 742 } 743 744 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 745 struct kvec *vec, size_t num, size_t size, int flags) 746 { 747 mm_segment_t oldfs = get_fs(); 748 int result; 749 750 set_fs(KERNEL_DS); 751 /* 752 * the following is safe, since for compiler definitions of kvec and 753 * iovec are identical, yielding the same in-core layout and alignment 754 */ 755 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num; 756 result = sock_recvmsg(sock, msg, size, flags); 757 set_fs(oldfs); 758 return result; 759 } 760 EXPORT_SYMBOL(kernel_recvmsg); 761 762 static void sock_aio_dtor(struct kiocb *iocb) 763 { 764 kfree(iocb->private); 765 } 766 767 static ssize_t sock_sendpage(struct file *file, struct page *page, 768 int offset, size_t size, loff_t *ppos, int more) 769 { 770 struct socket *sock; 771 int flags; 772 773 sock = file->private_data; 774 775 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT; 776 if (more) 777 flags |= MSG_MORE; 778 779 return kernel_sendpage(sock, page, offset, size, flags); 780 } 781 782 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 783 struct pipe_inode_info *pipe, size_t len, 784 unsigned int flags) 785 { 786 struct socket *sock = file->private_data; 787 788 if (unlikely(!sock->ops->splice_read)) 789 return -EINVAL; 790 791 sock_update_classid(sock->sk); 792 793 return sock->ops->splice_read(sock, ppos, pipe, len, flags); 794 } 795 796 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb, 797 struct sock_iocb *siocb) 798 { 799 if (!is_sync_kiocb(iocb)) { 800 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL); 801 if (!siocb) 802 return NULL; 803 iocb->ki_dtor = sock_aio_dtor; 804 } 805 806 siocb->kiocb = iocb; 807 iocb->private = siocb; 808 return siocb; 809 } 810 811 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb, 812 struct file *file, const struct iovec *iov, 813 unsigned long nr_segs) 814 { 815 struct socket *sock = file->private_data; 816 size_t size = 0; 817 int i; 818 819 for (i = 0; i < nr_segs; i++) 820 size += iov[i].iov_len; 821 822 msg->msg_name = NULL; 823 msg->msg_namelen = 0; 824 msg->msg_control = NULL; 825 msg->msg_controllen = 0; 826 msg->msg_iov = (struct iovec *)iov; 827 msg->msg_iovlen = nr_segs; 828 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 829 830 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags); 831 } 832 833 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, 834 unsigned long nr_segs, loff_t pos) 835 { 836 struct sock_iocb siocb, *x; 837 838 if (pos != 0) 839 return -ESPIPE; 840 841 if (iocb->ki_left == 0) /* Match SYS5 behaviour */ 842 return 0; 843 844 845 x = alloc_sock_iocb(iocb, &siocb); 846 if (!x) 847 return -ENOMEM; 848 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); 849 } 850 851 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb, 852 struct file *file, const struct iovec *iov, 853 unsigned long nr_segs) 854 { 855 struct socket *sock = file->private_data; 856 size_t size = 0; 857 int i; 858 859 for (i = 0; i < nr_segs; i++) 860 size += iov[i].iov_len; 861 862 msg->msg_name = NULL; 863 msg->msg_namelen = 0; 864 msg->msg_control = NULL; 865 msg->msg_controllen = 0; 866 msg->msg_iov = (struct iovec *)iov; 867 msg->msg_iovlen = nr_segs; 868 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 869 if (sock->type == SOCK_SEQPACKET) 870 msg->msg_flags |= MSG_EOR; 871 872 return __sock_sendmsg(iocb, sock, msg, size); 873 } 874 875 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, 876 unsigned long nr_segs, loff_t pos) 877 { 878 struct sock_iocb siocb, *x; 879 880 if (pos != 0) 881 return -ESPIPE; 882 883 x = alloc_sock_iocb(iocb, &siocb); 884 if (!x) 885 return -ENOMEM; 886 887 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); 888 } 889 890 /* 891 * Atomic setting of ioctl hooks to avoid race 892 * with module unload. 893 */ 894 895 static DEFINE_MUTEX(br_ioctl_mutex); 896 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg); 897 898 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) 899 { 900 mutex_lock(&br_ioctl_mutex); 901 br_ioctl_hook = hook; 902 mutex_unlock(&br_ioctl_mutex); 903 } 904 EXPORT_SYMBOL(brioctl_set); 905 906 static DEFINE_MUTEX(vlan_ioctl_mutex); 907 static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 908 909 void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 910 { 911 mutex_lock(&vlan_ioctl_mutex); 912 vlan_ioctl_hook = hook; 913 mutex_unlock(&vlan_ioctl_mutex); 914 } 915 EXPORT_SYMBOL(vlan_ioctl_set); 916 917 static DEFINE_MUTEX(dlci_ioctl_mutex); 918 static int (*dlci_ioctl_hook) (unsigned int, void __user *); 919 920 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) 921 { 922 mutex_lock(&dlci_ioctl_mutex); 923 dlci_ioctl_hook = hook; 924 mutex_unlock(&dlci_ioctl_mutex); 925 } 926 EXPORT_SYMBOL(dlci_ioctl_set); 927 928 static long sock_do_ioctl(struct net *net, struct socket *sock, 929 unsigned int cmd, unsigned long arg) 930 { 931 int err; 932 void __user *argp = (void __user *)arg; 933 934 err = sock->ops->ioctl(sock, cmd, arg); 935 936 /* 937 * If this ioctl is unknown try to hand it down 938 * to the NIC driver. 939 */ 940 if (err == -ENOIOCTLCMD) 941 err = dev_ioctl(net, cmd, argp); 942 943 return err; 944 } 945 946 /* 947 * With an ioctl, arg may well be a user mode pointer, but we don't know 948 * what to do with it - that's up to the protocol still. 949 */ 950 951 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 952 { 953 struct socket *sock; 954 struct sock *sk; 955 void __user *argp = (void __user *)arg; 956 int pid, err; 957 struct net *net; 958 959 sock = file->private_data; 960 sk = sock->sk; 961 net = sock_net(sk); 962 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) { 963 err = dev_ioctl(net, cmd, argp); 964 } else 965 #ifdef CONFIG_WEXT_CORE 966 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 967 err = dev_ioctl(net, cmd, argp); 968 } else 969 #endif 970 switch (cmd) { 971 case FIOSETOWN: 972 case SIOCSPGRP: 973 err = -EFAULT; 974 if (get_user(pid, (int __user *)argp)) 975 break; 976 err = f_setown(sock->file, pid, 1); 977 break; 978 case FIOGETOWN: 979 case SIOCGPGRP: 980 err = put_user(f_getown(sock->file), 981 (int __user *)argp); 982 break; 983 case SIOCGIFBR: 984 case SIOCSIFBR: 985 case SIOCBRADDBR: 986 case SIOCBRDELBR: 987 err = -ENOPKG; 988 if (!br_ioctl_hook) 989 request_module("bridge"); 990 991 mutex_lock(&br_ioctl_mutex); 992 if (br_ioctl_hook) 993 err = br_ioctl_hook(net, cmd, argp); 994 mutex_unlock(&br_ioctl_mutex); 995 break; 996 case SIOCGIFVLAN: 997 case SIOCSIFVLAN: 998 err = -ENOPKG; 999 if (!vlan_ioctl_hook) 1000 request_module("8021q"); 1001 1002 mutex_lock(&vlan_ioctl_mutex); 1003 if (vlan_ioctl_hook) 1004 err = vlan_ioctl_hook(net, argp); 1005 mutex_unlock(&vlan_ioctl_mutex); 1006 break; 1007 case SIOCADDDLCI: 1008 case SIOCDELDLCI: 1009 err = -ENOPKG; 1010 if (!dlci_ioctl_hook) 1011 request_module("dlci"); 1012 1013 mutex_lock(&dlci_ioctl_mutex); 1014 if (dlci_ioctl_hook) 1015 err = dlci_ioctl_hook(cmd, argp); 1016 mutex_unlock(&dlci_ioctl_mutex); 1017 break; 1018 default: 1019 err = sock_do_ioctl(net, sock, cmd, arg); 1020 break; 1021 } 1022 return err; 1023 } 1024 1025 int sock_create_lite(int family, int type, int protocol, struct socket **res) 1026 { 1027 int err; 1028 struct socket *sock = NULL; 1029 1030 err = security_socket_create(family, type, protocol, 1); 1031 if (err) 1032 goto out; 1033 1034 sock = sock_alloc(); 1035 if (!sock) { 1036 err = -ENOMEM; 1037 goto out; 1038 } 1039 1040 sock->type = type; 1041 err = security_socket_post_create(sock, family, type, protocol, 1); 1042 if (err) 1043 goto out_release; 1044 1045 out: 1046 *res = sock; 1047 return err; 1048 out_release: 1049 sock_release(sock); 1050 sock = NULL; 1051 goto out; 1052 } 1053 EXPORT_SYMBOL(sock_create_lite); 1054 1055 /* No kernel lock held - perfect */ 1056 static unsigned int sock_poll(struct file *file, poll_table *wait) 1057 { 1058 struct socket *sock; 1059 1060 /* 1061 * We can't return errors to poll, so it's either yes or no. 1062 */ 1063 sock = file->private_data; 1064 return sock->ops->poll(file, sock, wait); 1065 } 1066 1067 static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1068 { 1069 struct socket *sock = file->private_data; 1070 1071 return sock->ops->mmap(file, sock, vma); 1072 } 1073 1074 static int sock_close(struct inode *inode, struct file *filp) 1075 { 1076 /* 1077 * It was possible the inode is NULL we were 1078 * closing an unfinished socket. 1079 */ 1080 1081 if (!inode) { 1082 printk(KERN_DEBUG "sock_close: NULL inode\n"); 1083 return 0; 1084 } 1085 sock_release(SOCKET_I(inode)); 1086 return 0; 1087 } 1088 1089 /* 1090 * Update the socket async list 1091 * 1092 * Fasync_list locking strategy. 1093 * 1094 * 1. fasync_list is modified only under process context socket lock 1095 * i.e. under semaphore. 1096 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1097 * or under socket lock 1098 */ 1099 1100 static int sock_fasync(int fd, struct file *filp, int on) 1101 { 1102 struct socket *sock = filp->private_data; 1103 struct sock *sk = sock->sk; 1104 1105 if (sk == NULL) 1106 return -EINVAL; 1107 1108 lock_sock(sk); 1109 1110 fasync_helper(fd, filp, on, &sock->wq->fasync_list); 1111 1112 if (!sock->wq->fasync_list) 1113 sock_reset_flag(sk, SOCK_FASYNC); 1114 else 1115 sock_set_flag(sk, SOCK_FASYNC); 1116 1117 release_sock(sk); 1118 return 0; 1119 } 1120 1121 /* This function may be called only under socket lock or callback_lock or rcu_lock */ 1122 1123 int sock_wake_async(struct socket *sock, int how, int band) 1124 { 1125 struct socket_wq *wq; 1126 1127 if (!sock) 1128 return -1; 1129 rcu_read_lock(); 1130 wq = rcu_dereference(sock->wq); 1131 if (!wq || !wq->fasync_list) { 1132 rcu_read_unlock(); 1133 return -1; 1134 } 1135 switch (how) { 1136 case SOCK_WAKE_WAITD: 1137 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags)) 1138 break; 1139 goto call_kill; 1140 case SOCK_WAKE_SPACE: 1141 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags)) 1142 break; 1143 /* fall through */ 1144 case SOCK_WAKE_IO: 1145 call_kill: 1146 kill_fasync(&wq->fasync_list, SIGIO, band); 1147 break; 1148 case SOCK_WAKE_URG: 1149 kill_fasync(&wq->fasync_list, SIGURG, band); 1150 } 1151 rcu_read_unlock(); 1152 return 0; 1153 } 1154 EXPORT_SYMBOL(sock_wake_async); 1155 1156 static int __sock_create(struct net *net, int family, int type, int protocol, 1157 struct socket **res, int kern) 1158 { 1159 int err; 1160 struct socket *sock; 1161 const struct net_proto_family *pf; 1162 1163 /* 1164 * Check protocol is in range 1165 */ 1166 if (family < 0 || family >= NPROTO) 1167 return -EAFNOSUPPORT; 1168 if (type < 0 || type >= SOCK_MAX) 1169 return -EINVAL; 1170 1171 /* Compatibility. 1172 1173 This uglymoron is moved from INET layer to here to avoid 1174 deadlock in module load. 1175 */ 1176 if (family == PF_INET && type == SOCK_PACKET) { 1177 static int warned; 1178 if (!warned) { 1179 warned = 1; 1180 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1181 current->comm); 1182 } 1183 family = PF_PACKET; 1184 } 1185 1186 err = security_socket_create(family, type, protocol, kern); 1187 if (err) 1188 return err; 1189 1190 /* 1191 * Allocate the socket and allow the family to set things up. if 1192 * the protocol is 0, the family is instructed to select an appropriate 1193 * default. 1194 */ 1195 sock = sock_alloc(); 1196 if (!sock) { 1197 if (net_ratelimit()) 1198 printk(KERN_WARNING "socket: no more sockets\n"); 1199 return -ENFILE; /* Not exactly a match, but its the 1200 closest posix thing */ 1201 } 1202 1203 sock->type = type; 1204 1205 #ifdef CONFIG_MODULES 1206 /* Attempt to load a protocol module if the find failed. 1207 * 1208 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1209 * requested real, full-featured networking support upon configuration. 1210 * Otherwise module support will break! 1211 */ 1212 if (net_families[family] == NULL) 1213 request_module("net-pf-%d", family); 1214 #endif 1215 1216 rcu_read_lock(); 1217 pf = rcu_dereference(net_families[family]); 1218 err = -EAFNOSUPPORT; 1219 if (!pf) 1220 goto out_release; 1221 1222 /* 1223 * We will call the ->create function, that possibly is in a loadable 1224 * module, so we have to bump that loadable module refcnt first. 1225 */ 1226 if (!try_module_get(pf->owner)) 1227 goto out_release; 1228 1229 /* Now protected by module ref count */ 1230 rcu_read_unlock(); 1231 1232 err = pf->create(net, sock, protocol, kern); 1233 if (err < 0) 1234 goto out_module_put; 1235 1236 /* 1237 * Now to bump the refcnt of the [loadable] module that owns this 1238 * socket at sock_release time we decrement its refcnt. 1239 */ 1240 if (!try_module_get(sock->ops->owner)) 1241 goto out_module_busy; 1242 1243 /* 1244 * Now that we're done with the ->create function, the [loadable] 1245 * module can have its refcnt decremented 1246 */ 1247 module_put(pf->owner); 1248 err = security_socket_post_create(sock, family, type, protocol, kern); 1249 if (err) 1250 goto out_sock_release; 1251 *res = sock; 1252 1253 return 0; 1254 1255 out_module_busy: 1256 err = -EAFNOSUPPORT; 1257 out_module_put: 1258 sock->ops = NULL; 1259 module_put(pf->owner); 1260 out_sock_release: 1261 sock_release(sock); 1262 return err; 1263 1264 out_release: 1265 rcu_read_unlock(); 1266 goto out_sock_release; 1267 } 1268 1269 int sock_create(int family, int type, int protocol, struct socket **res) 1270 { 1271 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1272 } 1273 EXPORT_SYMBOL(sock_create); 1274 1275 int sock_create_kern(int family, int type, int protocol, struct socket **res) 1276 { 1277 return __sock_create(&init_net, family, type, protocol, res, 1); 1278 } 1279 EXPORT_SYMBOL(sock_create_kern); 1280 1281 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1282 { 1283 int retval; 1284 struct socket *sock; 1285 int flags; 1286 1287 /* Check the SOCK_* constants for consistency. */ 1288 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1289 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1290 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1291 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1292 1293 flags = type & ~SOCK_TYPE_MASK; 1294 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1295 return -EINVAL; 1296 type &= SOCK_TYPE_MASK; 1297 1298 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1299 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1300 1301 retval = sock_create(family, type, protocol, &sock); 1302 if (retval < 0) 1303 goto out; 1304 1305 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1306 if (retval < 0) 1307 goto out_release; 1308 1309 out: 1310 /* It may be already another descriptor 8) Not kernel problem. */ 1311 return retval; 1312 1313 out_release: 1314 sock_release(sock); 1315 return retval; 1316 } 1317 1318 /* 1319 * Create a pair of connected sockets. 1320 */ 1321 1322 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1323 int __user *, usockvec) 1324 { 1325 struct socket *sock1, *sock2; 1326 int fd1, fd2, err; 1327 struct file *newfile1, *newfile2; 1328 int flags; 1329 1330 flags = type & ~SOCK_TYPE_MASK; 1331 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1332 return -EINVAL; 1333 type &= SOCK_TYPE_MASK; 1334 1335 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1336 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1337 1338 /* 1339 * Obtain the first socket and check if the underlying protocol 1340 * supports the socketpair call. 1341 */ 1342 1343 err = sock_create(family, type, protocol, &sock1); 1344 if (err < 0) 1345 goto out; 1346 1347 err = sock_create(family, type, protocol, &sock2); 1348 if (err < 0) 1349 goto out_release_1; 1350 1351 err = sock1->ops->socketpair(sock1, sock2); 1352 if (err < 0) 1353 goto out_release_both; 1354 1355 fd1 = sock_alloc_file(sock1, &newfile1, flags); 1356 if (unlikely(fd1 < 0)) { 1357 err = fd1; 1358 goto out_release_both; 1359 } 1360 1361 fd2 = sock_alloc_file(sock2, &newfile2, flags); 1362 if (unlikely(fd2 < 0)) { 1363 err = fd2; 1364 fput(newfile1); 1365 put_unused_fd(fd1); 1366 sock_release(sock2); 1367 goto out; 1368 } 1369 1370 audit_fd_pair(fd1, fd2); 1371 fd_install(fd1, newfile1); 1372 fd_install(fd2, newfile2); 1373 /* fd1 and fd2 may be already another descriptors. 1374 * Not kernel problem. 1375 */ 1376 1377 err = put_user(fd1, &usockvec[0]); 1378 if (!err) 1379 err = put_user(fd2, &usockvec[1]); 1380 if (!err) 1381 return 0; 1382 1383 sys_close(fd2); 1384 sys_close(fd1); 1385 return err; 1386 1387 out_release_both: 1388 sock_release(sock2); 1389 out_release_1: 1390 sock_release(sock1); 1391 out: 1392 return err; 1393 } 1394 1395 /* 1396 * Bind a name to a socket. Nothing much to do here since it's 1397 * the protocol's responsibility to handle the local address. 1398 * 1399 * We move the socket address to kernel space before we call 1400 * the protocol layer (having also checked the address is ok). 1401 */ 1402 1403 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1404 { 1405 struct socket *sock; 1406 struct sockaddr_storage address; 1407 int err, fput_needed; 1408 1409 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1410 if (sock) { 1411 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address); 1412 if (err >= 0) { 1413 err = security_socket_bind(sock, 1414 (struct sockaddr *)&address, 1415 addrlen); 1416 if (!err) 1417 err = sock->ops->bind(sock, 1418 (struct sockaddr *) 1419 &address, addrlen); 1420 } 1421 fput_light(sock->file, fput_needed); 1422 } 1423 return err; 1424 } 1425 1426 /* 1427 * Perform a listen. Basically, we allow the protocol to do anything 1428 * necessary for a listen, and if that works, we mark the socket as 1429 * ready for listening. 1430 */ 1431 1432 SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1433 { 1434 struct socket *sock; 1435 int err, fput_needed; 1436 int somaxconn; 1437 1438 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1439 if (sock) { 1440 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; 1441 if ((unsigned)backlog > somaxconn) 1442 backlog = somaxconn; 1443 1444 err = security_socket_listen(sock, backlog); 1445 if (!err) 1446 err = sock->ops->listen(sock, backlog); 1447 1448 fput_light(sock->file, fput_needed); 1449 } 1450 return err; 1451 } 1452 1453 /* 1454 * For accept, we attempt to create a new socket, set up the link 1455 * with the client, wake up the client, then return the new 1456 * connected fd. We collect the address of the connector in kernel 1457 * space and move it to user at the very end. This is unclean because 1458 * we open the socket then return an error. 1459 * 1460 * 1003.1g adds the ability to recvmsg() to query connection pending 1461 * status to recvmsg. We need to add that support in a way thats 1462 * clean when we restucture accept also. 1463 */ 1464 1465 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 1466 int __user *, upeer_addrlen, int, flags) 1467 { 1468 struct socket *sock, *newsock; 1469 struct file *newfile; 1470 int err, len, newfd, fput_needed; 1471 struct sockaddr_storage address; 1472 1473 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1474 return -EINVAL; 1475 1476 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1477 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1478 1479 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1480 if (!sock) 1481 goto out; 1482 1483 err = -ENFILE; 1484 newsock = sock_alloc(); 1485 if (!newsock) 1486 goto out_put; 1487 1488 newsock->type = sock->type; 1489 newsock->ops = sock->ops; 1490 1491 /* 1492 * We don't need try_module_get here, as the listening socket (sock) 1493 * has the protocol module (sock->ops->owner) held. 1494 */ 1495 __module_get(newsock->ops->owner); 1496 1497 newfd = sock_alloc_file(newsock, &newfile, flags); 1498 if (unlikely(newfd < 0)) { 1499 err = newfd; 1500 sock_release(newsock); 1501 goto out_put; 1502 } 1503 1504 err = security_socket_accept(sock, newsock); 1505 if (err) 1506 goto out_fd; 1507 1508 err = sock->ops->accept(sock, newsock, sock->file->f_flags); 1509 if (err < 0) 1510 goto out_fd; 1511 1512 if (upeer_sockaddr) { 1513 if (newsock->ops->getname(newsock, (struct sockaddr *)&address, 1514 &len, 2) < 0) { 1515 err = -ECONNABORTED; 1516 goto out_fd; 1517 } 1518 err = move_addr_to_user((struct sockaddr *)&address, 1519 len, upeer_sockaddr, upeer_addrlen); 1520 if (err < 0) 1521 goto out_fd; 1522 } 1523 1524 /* File flags are not inherited via accept() unlike another OSes. */ 1525 1526 fd_install(newfd, newfile); 1527 err = newfd; 1528 1529 out_put: 1530 fput_light(sock->file, fput_needed); 1531 out: 1532 return err; 1533 out_fd: 1534 fput(newfile); 1535 put_unused_fd(newfd); 1536 goto out_put; 1537 } 1538 1539 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 1540 int __user *, upeer_addrlen) 1541 { 1542 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 1543 } 1544 1545 /* 1546 * Attempt to connect to a socket with the server address. The address 1547 * is in user space so we verify it is OK and move it to kernel space. 1548 * 1549 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 1550 * break bindings 1551 * 1552 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 1553 * other SEQPACKET protocols that take time to connect() as it doesn't 1554 * include the -EINPROGRESS status for such sockets. 1555 */ 1556 1557 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 1558 int, addrlen) 1559 { 1560 struct socket *sock; 1561 struct sockaddr_storage address; 1562 int err, fput_needed; 1563 1564 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1565 if (!sock) 1566 goto out; 1567 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address); 1568 if (err < 0) 1569 goto out_put; 1570 1571 err = 1572 security_socket_connect(sock, (struct sockaddr *)&address, addrlen); 1573 if (err) 1574 goto out_put; 1575 1576 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, 1577 sock->file->f_flags); 1578 out_put: 1579 fput_light(sock->file, fput_needed); 1580 out: 1581 return err; 1582 } 1583 1584 /* 1585 * Get the local address ('name') of a socket object. Move the obtained 1586 * name to user space. 1587 */ 1588 1589 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 1590 int __user *, usockaddr_len) 1591 { 1592 struct socket *sock; 1593 struct sockaddr_storage address; 1594 int len, err, fput_needed; 1595 1596 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1597 if (!sock) 1598 goto out; 1599 1600 err = security_socket_getsockname(sock); 1601 if (err) 1602 goto out_put; 1603 1604 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0); 1605 if (err) 1606 goto out_put; 1607 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len); 1608 1609 out_put: 1610 fput_light(sock->file, fput_needed); 1611 out: 1612 return err; 1613 } 1614 1615 /* 1616 * Get the remote address ('name') of a socket object. Move the obtained 1617 * name to user space. 1618 */ 1619 1620 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 1621 int __user *, usockaddr_len) 1622 { 1623 struct socket *sock; 1624 struct sockaddr_storage address; 1625 int len, err, fput_needed; 1626 1627 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1628 if (sock != NULL) { 1629 err = security_socket_getpeername(sock); 1630 if (err) { 1631 fput_light(sock->file, fput_needed); 1632 return err; 1633 } 1634 1635 err = 1636 sock->ops->getname(sock, (struct sockaddr *)&address, &len, 1637 1); 1638 if (!err) 1639 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, 1640 usockaddr_len); 1641 fput_light(sock->file, fput_needed); 1642 } 1643 return err; 1644 } 1645 1646 /* 1647 * Send a datagram to a given address. We move the address into kernel 1648 * space and check the user space data area is readable before invoking 1649 * the protocol. 1650 */ 1651 1652 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 1653 unsigned, flags, struct sockaddr __user *, addr, 1654 int, addr_len) 1655 { 1656 struct socket *sock; 1657 struct sockaddr_storage address; 1658 int err; 1659 struct msghdr msg; 1660 struct iovec iov; 1661 int fput_needed; 1662 1663 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1664 if (!sock) 1665 goto out; 1666 1667 iov.iov_base = buff; 1668 iov.iov_len = len; 1669 msg.msg_name = NULL; 1670 msg.msg_iov = &iov; 1671 msg.msg_iovlen = 1; 1672 msg.msg_control = NULL; 1673 msg.msg_controllen = 0; 1674 msg.msg_namelen = 0; 1675 if (addr) { 1676 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address); 1677 if (err < 0) 1678 goto out_put; 1679 msg.msg_name = (struct sockaddr *)&address; 1680 msg.msg_namelen = addr_len; 1681 } 1682 if (sock->file->f_flags & O_NONBLOCK) 1683 flags |= MSG_DONTWAIT; 1684 msg.msg_flags = flags; 1685 err = sock_sendmsg(sock, &msg, len); 1686 1687 out_put: 1688 fput_light(sock->file, fput_needed); 1689 out: 1690 return err; 1691 } 1692 1693 /* 1694 * Send a datagram down a socket. 1695 */ 1696 1697 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 1698 unsigned, flags) 1699 { 1700 return sys_sendto(fd, buff, len, flags, NULL, 0); 1701 } 1702 1703 /* 1704 * Receive a frame from the socket and optionally record the address of the 1705 * sender. We verify the buffers are writable and if needed move the 1706 * sender address from kernel to user space. 1707 */ 1708 1709 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 1710 unsigned, flags, struct sockaddr __user *, addr, 1711 int __user *, addr_len) 1712 { 1713 struct socket *sock; 1714 struct iovec iov; 1715 struct msghdr msg; 1716 struct sockaddr_storage address; 1717 int err, err2; 1718 int fput_needed; 1719 1720 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1721 if (!sock) 1722 goto out; 1723 1724 msg.msg_control = NULL; 1725 msg.msg_controllen = 0; 1726 msg.msg_iovlen = 1; 1727 msg.msg_iov = &iov; 1728 iov.iov_len = size; 1729 iov.iov_base = ubuf; 1730 msg.msg_name = (struct sockaddr *)&address; 1731 msg.msg_namelen = sizeof(address); 1732 if (sock->file->f_flags & O_NONBLOCK) 1733 flags |= MSG_DONTWAIT; 1734 err = sock_recvmsg(sock, &msg, size, flags); 1735 1736 if (err >= 0 && addr != NULL) { 1737 err2 = move_addr_to_user((struct sockaddr *)&address, 1738 msg.msg_namelen, addr, addr_len); 1739 if (err2 < 0) 1740 err = err2; 1741 } 1742 1743 fput_light(sock->file, fput_needed); 1744 out: 1745 return err; 1746 } 1747 1748 /* 1749 * Receive a datagram from a socket. 1750 */ 1751 1752 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size, 1753 unsigned flags) 1754 { 1755 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 1756 } 1757 1758 /* 1759 * Set a socket option. Because we don't know the option lengths we have 1760 * to pass the user mode parameter for the protocols to sort out. 1761 */ 1762 1763 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 1764 char __user *, optval, int, optlen) 1765 { 1766 int err, fput_needed; 1767 struct socket *sock; 1768 1769 if (optlen < 0) 1770 return -EINVAL; 1771 1772 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1773 if (sock != NULL) { 1774 err = security_socket_setsockopt(sock, level, optname); 1775 if (err) 1776 goto out_put; 1777 1778 if (level == SOL_SOCKET) 1779 err = 1780 sock_setsockopt(sock, level, optname, optval, 1781 optlen); 1782 else 1783 err = 1784 sock->ops->setsockopt(sock, level, optname, optval, 1785 optlen); 1786 out_put: 1787 fput_light(sock->file, fput_needed); 1788 } 1789 return err; 1790 } 1791 1792 /* 1793 * Get a socket option. Because we don't know the option lengths we have 1794 * to pass a user mode parameter for the protocols to sort out. 1795 */ 1796 1797 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 1798 char __user *, optval, int __user *, optlen) 1799 { 1800 int err, fput_needed; 1801 struct socket *sock; 1802 1803 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1804 if (sock != NULL) { 1805 err = security_socket_getsockopt(sock, level, optname); 1806 if (err) 1807 goto out_put; 1808 1809 if (level == SOL_SOCKET) 1810 err = 1811 sock_getsockopt(sock, level, optname, optval, 1812 optlen); 1813 else 1814 err = 1815 sock->ops->getsockopt(sock, level, optname, optval, 1816 optlen); 1817 out_put: 1818 fput_light(sock->file, fput_needed); 1819 } 1820 return err; 1821 } 1822 1823 /* 1824 * Shutdown a socket. 1825 */ 1826 1827 SYSCALL_DEFINE2(shutdown, int, fd, int, how) 1828 { 1829 int err, fput_needed; 1830 struct socket *sock; 1831 1832 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1833 if (sock != NULL) { 1834 err = security_socket_shutdown(sock, how); 1835 if (!err) 1836 err = sock->ops->shutdown(sock, how); 1837 fput_light(sock->file, fput_needed); 1838 } 1839 return err; 1840 } 1841 1842 /* A couple of helpful macros for getting the address of the 32/64 bit 1843 * fields which are the same type (int / unsigned) on our platforms. 1844 */ 1845 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 1846 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 1847 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 1848 1849 /* 1850 * BSD sendmsg interface 1851 */ 1852 1853 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags) 1854 { 1855 struct compat_msghdr __user *msg_compat = 1856 (struct compat_msghdr __user *)msg; 1857 struct socket *sock; 1858 struct sockaddr_storage address; 1859 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 1860 unsigned char ctl[sizeof(struct cmsghdr) + 20] 1861 __attribute__ ((aligned(sizeof(__kernel_size_t)))); 1862 /* 20 is size of ipv6_pktinfo */ 1863 unsigned char *ctl_buf = ctl; 1864 struct msghdr msg_sys; 1865 int err, ctl_len, iov_size, total_len; 1866 int fput_needed; 1867 1868 err = -EFAULT; 1869 if (MSG_CMSG_COMPAT & flags) { 1870 if (get_compat_msghdr(&msg_sys, msg_compat)) 1871 return -EFAULT; 1872 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr))) 1873 return -EFAULT; 1874 1875 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1876 if (!sock) 1877 goto out; 1878 1879 /* do not move before msg_sys is valid */ 1880 err = -EMSGSIZE; 1881 if (msg_sys.msg_iovlen > UIO_MAXIOV) 1882 goto out_put; 1883 1884 /* Check whether to allocate the iovec area */ 1885 err = -ENOMEM; 1886 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec); 1887 if (msg_sys.msg_iovlen > UIO_FASTIOV) { 1888 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); 1889 if (!iov) 1890 goto out_put; 1891 } 1892 1893 /* This will also move the address data into kernel space */ 1894 if (MSG_CMSG_COMPAT & flags) { 1895 err = verify_compat_iovec(&msg_sys, iov, 1896 (struct sockaddr *)&address, 1897 VERIFY_READ); 1898 } else 1899 err = verify_iovec(&msg_sys, iov, 1900 (struct sockaddr *)&address, 1901 VERIFY_READ); 1902 if (err < 0) 1903 goto out_freeiov; 1904 total_len = err; 1905 1906 err = -ENOBUFS; 1907 1908 if (msg_sys.msg_controllen > INT_MAX) 1909 goto out_freeiov; 1910 ctl_len = msg_sys.msg_controllen; 1911 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 1912 err = 1913 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, 1914 sizeof(ctl)); 1915 if (err) 1916 goto out_freeiov; 1917 ctl_buf = msg_sys.msg_control; 1918 ctl_len = msg_sys.msg_controllen; 1919 } else if (ctl_len) { 1920 if (ctl_len > sizeof(ctl)) { 1921 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 1922 if (ctl_buf == NULL) 1923 goto out_freeiov; 1924 } 1925 err = -EFAULT; 1926 /* 1927 * Careful! Before this, msg_sys.msg_control contains a user pointer. 1928 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted 1929 * checking falls down on this. 1930 */ 1931 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control, 1932 ctl_len)) 1933 goto out_freectl; 1934 msg_sys.msg_control = ctl_buf; 1935 } 1936 msg_sys.msg_flags = flags; 1937 1938 if (sock->file->f_flags & O_NONBLOCK) 1939 msg_sys.msg_flags |= MSG_DONTWAIT; 1940 err = sock_sendmsg(sock, &msg_sys, total_len); 1941 1942 out_freectl: 1943 if (ctl_buf != ctl) 1944 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 1945 out_freeiov: 1946 if (iov != iovstack) 1947 sock_kfree_s(sock->sk, iov, iov_size); 1948 out_put: 1949 fput_light(sock->file, fput_needed); 1950 out: 1951 return err; 1952 } 1953 1954 static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg, 1955 struct msghdr *msg_sys, unsigned flags, int nosec) 1956 { 1957 struct compat_msghdr __user *msg_compat = 1958 (struct compat_msghdr __user *)msg; 1959 struct iovec iovstack[UIO_FASTIOV]; 1960 struct iovec *iov = iovstack; 1961 unsigned long cmsg_ptr; 1962 int err, iov_size, total_len, len; 1963 1964 /* kernel mode address */ 1965 struct sockaddr_storage addr; 1966 1967 /* user mode address pointers */ 1968 struct sockaddr __user *uaddr; 1969 int __user *uaddr_len; 1970 1971 if (MSG_CMSG_COMPAT & flags) { 1972 if (get_compat_msghdr(msg_sys, msg_compat)) 1973 return -EFAULT; 1974 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr))) 1975 return -EFAULT; 1976 1977 err = -EMSGSIZE; 1978 if (msg_sys->msg_iovlen > UIO_MAXIOV) 1979 goto out; 1980 1981 /* Check whether to allocate the iovec area */ 1982 err = -ENOMEM; 1983 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec); 1984 if (msg_sys->msg_iovlen > UIO_FASTIOV) { 1985 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL); 1986 if (!iov) 1987 goto out; 1988 } 1989 1990 /* 1991 * Save the user-mode address (verify_iovec will change the 1992 * kernel msghdr to use the kernel address space) 1993 */ 1994 1995 uaddr = (__force void __user *)msg_sys->msg_name; 1996 uaddr_len = COMPAT_NAMELEN(msg); 1997 if (MSG_CMSG_COMPAT & flags) { 1998 err = verify_compat_iovec(msg_sys, iov, 1999 (struct sockaddr *)&addr, 2000 VERIFY_WRITE); 2001 } else 2002 err = verify_iovec(msg_sys, iov, 2003 (struct sockaddr *)&addr, 2004 VERIFY_WRITE); 2005 if (err < 0) 2006 goto out_freeiov; 2007 total_len = err; 2008 2009 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2010 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2011 2012 if (sock->file->f_flags & O_NONBLOCK) 2013 flags |= MSG_DONTWAIT; 2014 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, 2015 total_len, flags); 2016 if (err < 0) 2017 goto out_freeiov; 2018 len = err; 2019 2020 if (uaddr != NULL) { 2021 err = move_addr_to_user((struct sockaddr *)&addr, 2022 msg_sys->msg_namelen, uaddr, 2023 uaddr_len); 2024 if (err < 0) 2025 goto out_freeiov; 2026 } 2027 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2028 COMPAT_FLAGS(msg)); 2029 if (err) 2030 goto out_freeiov; 2031 if (MSG_CMSG_COMPAT & flags) 2032 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2033 &msg_compat->msg_controllen); 2034 else 2035 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2036 &msg->msg_controllen); 2037 if (err) 2038 goto out_freeiov; 2039 err = len; 2040 2041 out_freeiov: 2042 if (iov != iovstack) 2043 sock_kfree_s(sock->sk, iov, iov_size); 2044 out: 2045 return err; 2046 } 2047 2048 /* 2049 * BSD recvmsg interface 2050 */ 2051 2052 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg, 2053 unsigned int, flags) 2054 { 2055 int fput_needed, err; 2056 struct msghdr msg_sys; 2057 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed); 2058 2059 if (!sock) 2060 goto out; 2061 2062 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2063 2064 fput_light(sock->file, fput_needed); 2065 out: 2066 return err; 2067 } 2068 2069 /* 2070 * Linux recvmmsg interface 2071 */ 2072 2073 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2074 unsigned int flags, struct timespec *timeout) 2075 { 2076 int fput_needed, err, datagrams; 2077 struct socket *sock; 2078 struct mmsghdr __user *entry; 2079 struct compat_mmsghdr __user *compat_entry; 2080 struct msghdr msg_sys; 2081 struct timespec end_time; 2082 2083 if (timeout && 2084 poll_select_set_timeout(&end_time, timeout->tv_sec, 2085 timeout->tv_nsec)) 2086 return -EINVAL; 2087 2088 datagrams = 0; 2089 2090 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2091 if (!sock) 2092 return err; 2093 2094 err = sock_error(sock->sk); 2095 if (err) 2096 goto out_put; 2097 2098 entry = mmsg; 2099 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2100 2101 while (datagrams < vlen) { 2102 /* 2103 * No need to ask LSM for more than the first datagram. 2104 */ 2105 if (MSG_CMSG_COMPAT & flags) { 2106 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry, 2107 &msg_sys, flags, datagrams); 2108 if (err < 0) 2109 break; 2110 err = __put_user(err, &compat_entry->msg_len); 2111 ++compat_entry; 2112 } else { 2113 err = __sys_recvmsg(sock, (struct msghdr __user *)entry, 2114 &msg_sys, flags, datagrams); 2115 if (err < 0) 2116 break; 2117 err = put_user(err, &entry->msg_len); 2118 ++entry; 2119 } 2120 2121 if (err) 2122 break; 2123 ++datagrams; 2124 2125 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2126 if (flags & MSG_WAITFORONE) 2127 flags |= MSG_DONTWAIT; 2128 2129 if (timeout) { 2130 ktime_get_ts(timeout); 2131 *timeout = timespec_sub(end_time, *timeout); 2132 if (timeout->tv_sec < 0) { 2133 timeout->tv_sec = timeout->tv_nsec = 0; 2134 break; 2135 } 2136 2137 /* Timeout, return less than vlen datagrams */ 2138 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2139 break; 2140 } 2141 2142 /* Out of band data, return right away */ 2143 if (msg_sys.msg_flags & MSG_OOB) 2144 break; 2145 } 2146 2147 out_put: 2148 fput_light(sock->file, fput_needed); 2149 2150 if (err == 0) 2151 return datagrams; 2152 2153 if (datagrams != 0) { 2154 /* 2155 * We may return less entries than requested (vlen) if the 2156 * sock is non block and there aren't enough datagrams... 2157 */ 2158 if (err != -EAGAIN) { 2159 /* 2160 * ... or if recvmsg returns an error after we 2161 * received some datagrams, where we record the 2162 * error to return on the next call or if the 2163 * app asks about it using getsockopt(SO_ERROR). 2164 */ 2165 sock->sk->sk_err = -err; 2166 } 2167 2168 return datagrams; 2169 } 2170 2171 return err; 2172 } 2173 2174 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 2175 unsigned int, vlen, unsigned int, flags, 2176 struct timespec __user *, timeout) 2177 { 2178 int datagrams; 2179 struct timespec timeout_sys; 2180 2181 if (!timeout) 2182 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL); 2183 2184 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys))) 2185 return -EFAULT; 2186 2187 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 2188 2189 if (datagrams > 0 && 2190 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys))) 2191 datagrams = -EFAULT; 2192 2193 return datagrams; 2194 } 2195 2196 #ifdef __ARCH_WANT_SYS_SOCKETCALL 2197 /* Argument list sizes for sys_socketcall */ 2198 #define AL(x) ((x) * sizeof(unsigned long)) 2199 static const unsigned char nargs[20] = { 2200 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 2201 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 2202 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 2203 AL(4), AL(5) 2204 }; 2205 2206 #undef AL 2207 2208 /* 2209 * System call vectors. 2210 * 2211 * Argument checking cleaned up. Saved 20% in size. 2212 * This function doesn't need to set the kernel lock because 2213 * it is set by the callees. 2214 */ 2215 2216 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 2217 { 2218 unsigned long a[6]; 2219 unsigned long a0, a1; 2220 int err; 2221 unsigned int len; 2222 2223 if (call < 1 || call > SYS_RECVMMSG) 2224 return -EINVAL; 2225 2226 len = nargs[call]; 2227 if (len > sizeof(a)) 2228 return -EINVAL; 2229 2230 /* copy_from_user should be SMP safe. */ 2231 if (copy_from_user(a, args, len)) 2232 return -EFAULT; 2233 2234 audit_socketcall(nargs[call] / sizeof(unsigned long), a); 2235 2236 a0 = a[0]; 2237 a1 = a[1]; 2238 2239 switch (call) { 2240 case SYS_SOCKET: 2241 err = sys_socket(a0, a1, a[2]); 2242 break; 2243 case SYS_BIND: 2244 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 2245 break; 2246 case SYS_CONNECT: 2247 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 2248 break; 2249 case SYS_LISTEN: 2250 err = sys_listen(a0, a1); 2251 break; 2252 case SYS_ACCEPT: 2253 err = sys_accept4(a0, (struct sockaddr __user *)a1, 2254 (int __user *)a[2], 0); 2255 break; 2256 case SYS_GETSOCKNAME: 2257 err = 2258 sys_getsockname(a0, (struct sockaddr __user *)a1, 2259 (int __user *)a[2]); 2260 break; 2261 case SYS_GETPEERNAME: 2262 err = 2263 sys_getpeername(a0, (struct sockaddr __user *)a1, 2264 (int __user *)a[2]); 2265 break; 2266 case SYS_SOCKETPAIR: 2267 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 2268 break; 2269 case SYS_SEND: 2270 err = sys_send(a0, (void __user *)a1, a[2], a[3]); 2271 break; 2272 case SYS_SENDTO: 2273 err = sys_sendto(a0, (void __user *)a1, a[2], a[3], 2274 (struct sockaddr __user *)a[4], a[5]); 2275 break; 2276 case SYS_RECV: 2277 err = sys_recv(a0, (void __user *)a1, a[2], a[3]); 2278 break; 2279 case SYS_RECVFROM: 2280 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 2281 (struct sockaddr __user *)a[4], 2282 (int __user *)a[5]); 2283 break; 2284 case SYS_SHUTDOWN: 2285 err = sys_shutdown(a0, a1); 2286 break; 2287 case SYS_SETSOCKOPT: 2288 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]); 2289 break; 2290 case SYS_GETSOCKOPT: 2291 err = 2292 sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 2293 (int __user *)a[4]); 2294 break; 2295 case SYS_SENDMSG: 2296 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]); 2297 break; 2298 case SYS_RECVMSG: 2299 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]); 2300 break; 2301 case SYS_RECVMMSG: 2302 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3], 2303 (struct timespec __user *)a[4]); 2304 break; 2305 case SYS_ACCEPT4: 2306 err = sys_accept4(a0, (struct sockaddr __user *)a1, 2307 (int __user *)a[2], a[3]); 2308 break; 2309 default: 2310 err = -EINVAL; 2311 break; 2312 } 2313 return err; 2314 } 2315 2316 #endif /* __ARCH_WANT_SYS_SOCKETCALL */ 2317 2318 /** 2319 * sock_register - add a socket protocol handler 2320 * @ops: description of protocol 2321 * 2322 * This function is called by a protocol handler that wants to 2323 * advertise its address family, and have it linked into the 2324 * socket interface. The value ops->family coresponds to the 2325 * socket system call protocol family. 2326 */ 2327 int sock_register(const struct net_proto_family *ops) 2328 { 2329 int err; 2330 2331 if (ops->family >= NPROTO) { 2332 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, 2333 NPROTO); 2334 return -ENOBUFS; 2335 } 2336 2337 spin_lock(&net_family_lock); 2338 if (net_families[ops->family]) 2339 err = -EEXIST; 2340 else { 2341 net_families[ops->family] = ops; 2342 err = 0; 2343 } 2344 spin_unlock(&net_family_lock); 2345 2346 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family); 2347 return err; 2348 } 2349 EXPORT_SYMBOL(sock_register); 2350 2351 /** 2352 * sock_unregister - remove a protocol handler 2353 * @family: protocol family to remove 2354 * 2355 * This function is called by a protocol handler that wants to 2356 * remove its address family, and have it unlinked from the 2357 * new socket creation. 2358 * 2359 * If protocol handler is a module, then it can use module reference 2360 * counts to protect against new references. If protocol handler is not 2361 * a module then it needs to provide its own protection in 2362 * the ops->create routine. 2363 */ 2364 void sock_unregister(int family) 2365 { 2366 BUG_ON(family < 0 || family >= NPROTO); 2367 2368 spin_lock(&net_family_lock); 2369 net_families[family] = NULL; 2370 spin_unlock(&net_family_lock); 2371 2372 synchronize_rcu(); 2373 2374 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family); 2375 } 2376 EXPORT_SYMBOL(sock_unregister); 2377 2378 static int __init sock_init(void) 2379 { 2380 /* 2381 * Initialize sock SLAB cache. 2382 */ 2383 2384 sk_init(); 2385 2386 /* 2387 * Initialize skbuff SLAB cache 2388 */ 2389 skb_init(); 2390 2391 /* 2392 * Initialize the protocols module. 2393 */ 2394 2395 init_inodecache(); 2396 register_filesystem(&sock_fs_type); 2397 sock_mnt = kern_mount(&sock_fs_type); 2398 2399 /* The real protocol initialization is performed in later initcalls. 2400 */ 2401 2402 #ifdef CONFIG_NETFILTER 2403 netfilter_init(); 2404 #endif 2405 2406 return 0; 2407 } 2408 2409 core_initcall(sock_init); /* early initcall */ 2410 2411 #ifdef CONFIG_PROC_FS 2412 void socket_seq_show(struct seq_file *seq) 2413 { 2414 int cpu; 2415 int counter = 0; 2416 2417 for_each_possible_cpu(cpu) 2418 counter += per_cpu(sockets_in_use, cpu); 2419 2420 /* It can be negative, by the way. 8) */ 2421 if (counter < 0) 2422 counter = 0; 2423 2424 seq_printf(seq, "sockets: used %d\n", counter); 2425 } 2426 #endif /* CONFIG_PROC_FS */ 2427 2428 #ifdef CONFIG_COMPAT 2429 static int do_siocgstamp(struct net *net, struct socket *sock, 2430 unsigned int cmd, struct compat_timeval __user *up) 2431 { 2432 mm_segment_t old_fs = get_fs(); 2433 struct timeval ktv; 2434 int err; 2435 2436 set_fs(KERNEL_DS); 2437 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv); 2438 set_fs(old_fs); 2439 if (!err) { 2440 err = put_user(ktv.tv_sec, &up->tv_sec); 2441 err |= __put_user(ktv.tv_usec, &up->tv_usec); 2442 } 2443 return err; 2444 } 2445 2446 static int do_siocgstampns(struct net *net, struct socket *sock, 2447 unsigned int cmd, struct compat_timespec __user *up) 2448 { 2449 mm_segment_t old_fs = get_fs(); 2450 struct timespec kts; 2451 int err; 2452 2453 set_fs(KERNEL_DS); 2454 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts); 2455 set_fs(old_fs); 2456 if (!err) { 2457 err = put_user(kts.tv_sec, &up->tv_sec); 2458 err |= __put_user(kts.tv_nsec, &up->tv_nsec); 2459 } 2460 return err; 2461 } 2462 2463 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32) 2464 { 2465 struct ifreq __user *uifr; 2466 int err; 2467 2468 uifr = compat_alloc_user_space(sizeof(struct ifreq)); 2469 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) 2470 return -EFAULT; 2471 2472 err = dev_ioctl(net, SIOCGIFNAME, uifr); 2473 if (err) 2474 return err; 2475 2476 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq))) 2477 return -EFAULT; 2478 2479 return 0; 2480 } 2481 2482 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32) 2483 { 2484 struct compat_ifconf ifc32; 2485 struct ifconf ifc; 2486 struct ifconf __user *uifc; 2487 struct compat_ifreq __user *ifr32; 2488 struct ifreq __user *ifr; 2489 unsigned int i, j; 2490 int err; 2491 2492 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf))) 2493 return -EFAULT; 2494 2495 if (ifc32.ifcbuf == 0) { 2496 ifc32.ifc_len = 0; 2497 ifc.ifc_len = 0; 2498 ifc.ifc_req = NULL; 2499 uifc = compat_alloc_user_space(sizeof(struct ifconf)); 2500 } else { 2501 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) * 2502 sizeof(struct ifreq); 2503 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len); 2504 ifc.ifc_len = len; 2505 ifr = ifc.ifc_req = (void __user *)(uifc + 1); 2506 ifr32 = compat_ptr(ifc32.ifcbuf); 2507 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) { 2508 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq))) 2509 return -EFAULT; 2510 ifr++; 2511 ifr32++; 2512 } 2513 } 2514 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf))) 2515 return -EFAULT; 2516 2517 err = dev_ioctl(net, SIOCGIFCONF, uifc); 2518 if (err) 2519 return err; 2520 2521 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf))) 2522 return -EFAULT; 2523 2524 ifr = ifc.ifc_req; 2525 ifr32 = compat_ptr(ifc32.ifcbuf); 2526 for (i = 0, j = 0; 2527 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len; 2528 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) { 2529 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq))) 2530 return -EFAULT; 2531 ifr32++; 2532 ifr++; 2533 } 2534 2535 if (ifc32.ifcbuf == 0) { 2536 /* Translate from 64-bit structure multiple to 2537 * a 32-bit one. 2538 */ 2539 i = ifc.ifc_len; 2540 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq)); 2541 ifc32.ifc_len = i; 2542 } else { 2543 ifc32.ifc_len = i; 2544 } 2545 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf))) 2546 return -EFAULT; 2547 2548 return 0; 2549 } 2550 2551 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32) 2552 { 2553 struct ifreq __user *ifr; 2554 u32 data; 2555 void __user *datap; 2556 2557 ifr = compat_alloc_user_space(sizeof(*ifr)); 2558 2559 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ)) 2560 return -EFAULT; 2561 2562 if (get_user(data, &ifr32->ifr_ifru.ifru_data)) 2563 return -EFAULT; 2564 2565 datap = compat_ptr(data); 2566 if (put_user(datap, &ifr->ifr_ifru.ifru_data)) 2567 return -EFAULT; 2568 2569 return dev_ioctl(net, SIOCETHTOOL, ifr); 2570 } 2571 2572 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 2573 { 2574 void __user *uptr; 2575 compat_uptr_t uptr32; 2576 struct ifreq __user *uifr; 2577 2578 uifr = compat_alloc_user_space(sizeof(*uifr)); 2579 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) 2580 return -EFAULT; 2581 2582 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 2583 return -EFAULT; 2584 2585 uptr = compat_ptr(uptr32); 2586 2587 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc)) 2588 return -EFAULT; 2589 2590 return dev_ioctl(net, SIOCWANDEV, uifr); 2591 } 2592 2593 static int bond_ioctl(struct net *net, unsigned int cmd, 2594 struct compat_ifreq __user *ifr32) 2595 { 2596 struct ifreq kifr; 2597 struct ifreq __user *uifr; 2598 mm_segment_t old_fs; 2599 int err; 2600 u32 data; 2601 void __user *datap; 2602 2603 switch (cmd) { 2604 case SIOCBONDENSLAVE: 2605 case SIOCBONDRELEASE: 2606 case SIOCBONDSETHWADDR: 2607 case SIOCBONDCHANGEACTIVE: 2608 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq))) 2609 return -EFAULT; 2610 2611 old_fs = get_fs(); 2612 set_fs(KERNEL_DS); 2613 err = dev_ioctl(net, cmd, &kifr); 2614 set_fs(old_fs); 2615 2616 return err; 2617 case SIOCBONDSLAVEINFOQUERY: 2618 case SIOCBONDINFOQUERY: 2619 uifr = compat_alloc_user_space(sizeof(*uifr)); 2620 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ)) 2621 return -EFAULT; 2622 2623 if (get_user(data, &ifr32->ifr_ifru.ifru_data)) 2624 return -EFAULT; 2625 2626 datap = compat_ptr(data); 2627 if (put_user(datap, &uifr->ifr_ifru.ifru_data)) 2628 return -EFAULT; 2629 2630 return dev_ioctl(net, cmd, uifr); 2631 default: 2632 return -EINVAL; 2633 } 2634 } 2635 2636 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd, 2637 struct compat_ifreq __user *u_ifreq32) 2638 { 2639 struct ifreq __user *u_ifreq64; 2640 char tmp_buf[IFNAMSIZ]; 2641 void __user *data64; 2642 u32 data32; 2643 2644 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]), 2645 IFNAMSIZ)) 2646 return -EFAULT; 2647 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data)) 2648 return -EFAULT; 2649 data64 = compat_ptr(data32); 2650 2651 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64)); 2652 2653 /* Don't check these user accesses, just let that get trapped 2654 * in the ioctl handler instead. 2655 */ 2656 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0], 2657 IFNAMSIZ)) 2658 return -EFAULT; 2659 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data)) 2660 return -EFAULT; 2661 2662 return dev_ioctl(net, cmd, u_ifreq64); 2663 } 2664 2665 static int dev_ifsioc(struct net *net, struct socket *sock, 2666 unsigned int cmd, struct compat_ifreq __user *uifr32) 2667 { 2668 struct ifreq __user *uifr; 2669 int err; 2670 2671 uifr = compat_alloc_user_space(sizeof(*uifr)); 2672 if (copy_in_user(uifr, uifr32, sizeof(*uifr32))) 2673 return -EFAULT; 2674 2675 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr); 2676 2677 if (!err) { 2678 switch (cmd) { 2679 case SIOCGIFFLAGS: 2680 case SIOCGIFMETRIC: 2681 case SIOCGIFMTU: 2682 case SIOCGIFMEM: 2683 case SIOCGIFHWADDR: 2684 case SIOCGIFINDEX: 2685 case SIOCGIFADDR: 2686 case SIOCGIFBRDADDR: 2687 case SIOCGIFDSTADDR: 2688 case SIOCGIFNETMASK: 2689 case SIOCGIFPFLAGS: 2690 case SIOCGIFTXQLEN: 2691 case SIOCGMIIPHY: 2692 case SIOCGMIIREG: 2693 if (copy_in_user(uifr32, uifr, sizeof(*uifr32))) 2694 err = -EFAULT; 2695 break; 2696 } 2697 } 2698 return err; 2699 } 2700 2701 static int compat_sioc_ifmap(struct net *net, unsigned int cmd, 2702 struct compat_ifreq __user *uifr32) 2703 { 2704 struct ifreq ifr; 2705 struct compat_ifmap __user *uifmap32; 2706 mm_segment_t old_fs; 2707 int err; 2708 2709 uifmap32 = &uifr32->ifr_ifru.ifru_map; 2710 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name)); 2711 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 2712 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 2713 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 2714 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq); 2715 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma); 2716 err |= __get_user(ifr.ifr_map.port, &uifmap32->port); 2717 if (err) 2718 return -EFAULT; 2719 2720 old_fs = get_fs(); 2721 set_fs(KERNEL_DS); 2722 err = dev_ioctl(net, cmd, (void __user *)&ifr); 2723 set_fs(old_fs); 2724 2725 if (cmd == SIOCGIFMAP && !err) { 2726 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name)); 2727 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 2728 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 2729 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 2730 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq); 2731 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma); 2732 err |= __put_user(ifr.ifr_map.port, &uifmap32->port); 2733 if (err) 2734 err = -EFAULT; 2735 } 2736 return err; 2737 } 2738 2739 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32) 2740 { 2741 void __user *uptr; 2742 compat_uptr_t uptr32; 2743 struct ifreq __user *uifr; 2744 2745 uifr = compat_alloc_user_space(sizeof(*uifr)); 2746 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) 2747 return -EFAULT; 2748 2749 if (get_user(uptr32, &uifr32->ifr_data)) 2750 return -EFAULT; 2751 2752 uptr = compat_ptr(uptr32); 2753 2754 if (put_user(uptr, &uifr->ifr_data)) 2755 return -EFAULT; 2756 2757 return dev_ioctl(net, SIOCSHWTSTAMP, uifr); 2758 } 2759 2760 struct rtentry32 { 2761 u32 rt_pad1; 2762 struct sockaddr rt_dst; /* target address */ 2763 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */ 2764 struct sockaddr rt_genmask; /* target network mask (IP) */ 2765 unsigned short rt_flags; 2766 short rt_pad2; 2767 u32 rt_pad3; 2768 unsigned char rt_tos; 2769 unsigned char rt_class; 2770 short rt_pad4; 2771 short rt_metric; /* +1 for binary compatibility! */ 2772 /* char * */ u32 rt_dev; /* forcing the device at add */ 2773 u32 rt_mtu; /* per route MTU/Window */ 2774 u32 rt_window; /* Window clamping */ 2775 unsigned short rt_irtt; /* Initial RTT */ 2776 }; 2777 2778 struct in6_rtmsg32 { 2779 struct in6_addr rtmsg_dst; 2780 struct in6_addr rtmsg_src; 2781 struct in6_addr rtmsg_gateway; 2782 u32 rtmsg_type; 2783 u16 rtmsg_dst_len; 2784 u16 rtmsg_src_len; 2785 u32 rtmsg_metric; 2786 u32 rtmsg_info; 2787 u32 rtmsg_flags; 2788 s32 rtmsg_ifindex; 2789 }; 2790 2791 static int routing_ioctl(struct net *net, struct socket *sock, 2792 unsigned int cmd, void __user *argp) 2793 { 2794 int ret; 2795 void *r = NULL; 2796 struct in6_rtmsg r6; 2797 struct rtentry r4; 2798 char devname[16]; 2799 u32 rtdev; 2800 mm_segment_t old_fs = get_fs(); 2801 2802 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */ 2803 struct in6_rtmsg32 __user *ur6 = argp; 2804 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst), 2805 3 * sizeof(struct in6_addr)); 2806 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type)); 2807 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len)); 2808 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len)); 2809 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric)); 2810 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info)); 2811 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags)); 2812 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex)); 2813 2814 r = (void *) &r6; 2815 } else { /* ipv4 */ 2816 struct rtentry32 __user *ur4 = argp; 2817 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst), 2818 3 * sizeof(struct sockaddr)); 2819 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags)); 2820 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric)); 2821 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu)); 2822 ret |= __get_user(r4.rt_window, &(ur4->rt_window)); 2823 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt)); 2824 ret |= __get_user(rtdev, &(ur4->rt_dev)); 2825 if (rtdev) { 2826 ret |= copy_from_user(devname, compat_ptr(rtdev), 15); 2827 r4.rt_dev = devname; devname[15] = 0; 2828 } else 2829 r4.rt_dev = NULL; 2830 2831 r = (void *) &r4; 2832 } 2833 2834 if (ret) { 2835 ret = -EFAULT; 2836 goto out; 2837 } 2838 2839 set_fs(KERNEL_DS); 2840 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r); 2841 set_fs(old_fs); 2842 2843 out: 2844 return ret; 2845 } 2846 2847 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE 2848 * for some operations; this forces use of the newer bridge-utils that 2849 * use compatiable ioctls 2850 */ 2851 static int old_bridge_ioctl(compat_ulong_t __user *argp) 2852 { 2853 compat_ulong_t tmp; 2854 2855 if (get_user(tmp, argp)) 2856 return -EFAULT; 2857 if (tmp == BRCTL_GET_VERSION) 2858 return BRCTL_VERSION + 1; 2859 return -EINVAL; 2860 } 2861 2862 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 2863 unsigned int cmd, unsigned long arg) 2864 { 2865 void __user *argp = compat_ptr(arg); 2866 struct sock *sk = sock->sk; 2867 struct net *net = sock_net(sk); 2868 2869 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 2870 return siocdevprivate_ioctl(net, cmd, argp); 2871 2872 switch (cmd) { 2873 case SIOCSIFBR: 2874 case SIOCGIFBR: 2875 return old_bridge_ioctl(argp); 2876 case SIOCGIFNAME: 2877 return dev_ifname32(net, argp); 2878 case SIOCGIFCONF: 2879 return dev_ifconf(net, argp); 2880 case SIOCETHTOOL: 2881 return ethtool_ioctl(net, argp); 2882 case SIOCWANDEV: 2883 return compat_siocwandev(net, argp); 2884 case SIOCGIFMAP: 2885 case SIOCSIFMAP: 2886 return compat_sioc_ifmap(net, cmd, argp); 2887 case SIOCBONDENSLAVE: 2888 case SIOCBONDRELEASE: 2889 case SIOCBONDSETHWADDR: 2890 case SIOCBONDSLAVEINFOQUERY: 2891 case SIOCBONDINFOQUERY: 2892 case SIOCBONDCHANGEACTIVE: 2893 return bond_ioctl(net, cmd, argp); 2894 case SIOCADDRT: 2895 case SIOCDELRT: 2896 return routing_ioctl(net, sock, cmd, argp); 2897 case SIOCGSTAMP: 2898 return do_siocgstamp(net, sock, cmd, argp); 2899 case SIOCGSTAMPNS: 2900 return do_siocgstampns(net, sock, cmd, argp); 2901 case SIOCSHWTSTAMP: 2902 return compat_siocshwtstamp(net, argp); 2903 2904 case FIOSETOWN: 2905 case SIOCSPGRP: 2906 case FIOGETOWN: 2907 case SIOCGPGRP: 2908 case SIOCBRADDBR: 2909 case SIOCBRDELBR: 2910 case SIOCGIFVLAN: 2911 case SIOCSIFVLAN: 2912 case SIOCADDDLCI: 2913 case SIOCDELDLCI: 2914 return sock_ioctl(file, cmd, arg); 2915 2916 case SIOCGIFFLAGS: 2917 case SIOCSIFFLAGS: 2918 case SIOCGIFMETRIC: 2919 case SIOCSIFMETRIC: 2920 case SIOCGIFMTU: 2921 case SIOCSIFMTU: 2922 case SIOCGIFMEM: 2923 case SIOCSIFMEM: 2924 case SIOCGIFHWADDR: 2925 case SIOCSIFHWADDR: 2926 case SIOCADDMULTI: 2927 case SIOCDELMULTI: 2928 case SIOCGIFINDEX: 2929 case SIOCGIFADDR: 2930 case SIOCSIFADDR: 2931 case SIOCSIFHWBROADCAST: 2932 case SIOCDIFADDR: 2933 case SIOCGIFBRDADDR: 2934 case SIOCSIFBRDADDR: 2935 case SIOCGIFDSTADDR: 2936 case SIOCSIFDSTADDR: 2937 case SIOCGIFNETMASK: 2938 case SIOCSIFNETMASK: 2939 case SIOCSIFPFLAGS: 2940 case SIOCGIFPFLAGS: 2941 case SIOCGIFTXQLEN: 2942 case SIOCSIFTXQLEN: 2943 case SIOCBRADDIF: 2944 case SIOCBRDELIF: 2945 case SIOCSIFNAME: 2946 case SIOCGMIIPHY: 2947 case SIOCGMIIREG: 2948 case SIOCSMIIREG: 2949 return dev_ifsioc(net, sock, cmd, argp); 2950 2951 case SIOCSARP: 2952 case SIOCGARP: 2953 case SIOCDARP: 2954 case SIOCATMARK: 2955 return sock_do_ioctl(net, sock, cmd, arg); 2956 } 2957 2958 /* Prevent warning from compat_sys_ioctl, these always 2959 * result in -EINVAL in the native case anyway. */ 2960 switch (cmd) { 2961 case SIOCRTMSG: 2962 case SIOCGIFCOUNT: 2963 case SIOCSRARP: 2964 case SIOCGRARP: 2965 case SIOCDRARP: 2966 case SIOCSIFLINK: 2967 case SIOCGIFSLAVE: 2968 case SIOCSIFSLAVE: 2969 return -EINVAL; 2970 } 2971 2972 return -ENOIOCTLCMD; 2973 } 2974 2975 static long compat_sock_ioctl(struct file *file, unsigned cmd, 2976 unsigned long arg) 2977 { 2978 struct socket *sock = file->private_data; 2979 int ret = -ENOIOCTLCMD; 2980 struct sock *sk; 2981 struct net *net; 2982 2983 sk = sock->sk; 2984 net = sock_net(sk); 2985 2986 if (sock->ops->compat_ioctl) 2987 ret = sock->ops->compat_ioctl(sock, cmd, arg); 2988 2989 if (ret == -ENOIOCTLCMD && 2990 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 2991 ret = compat_wext_handle_ioctl(net, cmd, arg); 2992 2993 if (ret == -ENOIOCTLCMD) 2994 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 2995 2996 return ret; 2997 } 2998 #endif 2999 3000 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3001 { 3002 return sock->ops->bind(sock, addr, addrlen); 3003 } 3004 EXPORT_SYMBOL(kernel_bind); 3005 3006 int kernel_listen(struct socket *sock, int backlog) 3007 { 3008 return sock->ops->listen(sock, backlog); 3009 } 3010 EXPORT_SYMBOL(kernel_listen); 3011 3012 int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3013 { 3014 struct sock *sk = sock->sk; 3015 int err; 3016 3017 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3018 newsock); 3019 if (err < 0) 3020 goto done; 3021 3022 err = sock->ops->accept(sock, *newsock, flags); 3023 if (err < 0) { 3024 sock_release(*newsock); 3025 *newsock = NULL; 3026 goto done; 3027 } 3028 3029 (*newsock)->ops = sock->ops; 3030 __module_get((*newsock)->ops->owner); 3031 3032 done: 3033 return err; 3034 } 3035 EXPORT_SYMBOL(kernel_accept); 3036 3037 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3038 int flags) 3039 { 3040 return sock->ops->connect(sock, addr, addrlen, flags); 3041 } 3042 EXPORT_SYMBOL(kernel_connect); 3043 3044 int kernel_getsockname(struct socket *sock, struct sockaddr *addr, 3045 int *addrlen) 3046 { 3047 return sock->ops->getname(sock, addr, addrlen, 0); 3048 } 3049 EXPORT_SYMBOL(kernel_getsockname); 3050 3051 int kernel_getpeername(struct socket *sock, struct sockaddr *addr, 3052 int *addrlen) 3053 { 3054 return sock->ops->getname(sock, addr, addrlen, 1); 3055 } 3056 EXPORT_SYMBOL(kernel_getpeername); 3057 3058 int kernel_getsockopt(struct socket *sock, int level, int optname, 3059 char *optval, int *optlen) 3060 { 3061 mm_segment_t oldfs = get_fs(); 3062 int err; 3063 3064 set_fs(KERNEL_DS); 3065 if (level == SOL_SOCKET) 3066 err = sock_getsockopt(sock, level, optname, optval, optlen); 3067 else 3068 err = sock->ops->getsockopt(sock, level, optname, optval, 3069 optlen); 3070 set_fs(oldfs); 3071 return err; 3072 } 3073 EXPORT_SYMBOL(kernel_getsockopt); 3074 3075 int kernel_setsockopt(struct socket *sock, int level, int optname, 3076 char *optval, unsigned int optlen) 3077 { 3078 mm_segment_t oldfs = get_fs(); 3079 int err; 3080 3081 set_fs(KERNEL_DS); 3082 if (level == SOL_SOCKET) 3083 err = sock_setsockopt(sock, level, optname, optval, optlen); 3084 else 3085 err = sock->ops->setsockopt(sock, level, optname, optval, 3086 optlen); 3087 set_fs(oldfs); 3088 return err; 3089 } 3090 EXPORT_SYMBOL(kernel_setsockopt); 3091 3092 int kernel_sendpage(struct socket *sock, struct page *page, int offset, 3093 size_t size, int flags) 3094 { 3095 sock_update_classid(sock->sk); 3096 3097 if (sock->ops->sendpage) 3098 return sock->ops->sendpage(sock, page, offset, size, flags); 3099 3100 return sock_no_sendpage(sock, page, offset, size, flags); 3101 } 3102 EXPORT_SYMBOL(kernel_sendpage); 3103 3104 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg) 3105 { 3106 mm_segment_t oldfs = get_fs(); 3107 int err; 3108 3109 set_fs(KERNEL_DS); 3110 err = sock->ops->ioctl(sock, cmd, arg); 3111 set_fs(oldfs); 3112 3113 return err; 3114 } 3115 EXPORT_SYMBOL(kernel_sock_ioctl); 3116 3117 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3118 { 3119 return sock->ops->shutdown(sock, how); 3120 } 3121 EXPORT_SYMBOL(kernel_sock_shutdown); 3122