/net/socket.c
C | 3505 lines | 2541 code | 501 blank | 463 comment | 401 complexity | 8d5656caf9ddf051858bf9f47c7d6911 MD5 | raw file
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, <obz@Kodak.COM> 7 * Ross Biro 8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 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#include <linux/xattr.h> 92 93#include <asm/uaccess.h> 94#include <asm/unistd.h> 95 96#include <net/compat.h> 97#include <net/wext.h> 98#include <net/cls_cgroup.h> 99 100#include <net/sock.h> 101#include <linux/netfilter.h> 102 103#include <linux/if_tun.h> 104#include <linux/ipv6_route.h> 105#include <linux/route.h> 106#include <linux/sockios.h> 107#include <linux/atalk.h> 108 109static int sock_no_open(struct inode *irrelevant, struct file *dontcare); 110static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, 111 unsigned long nr_segs, loff_t pos); 112static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, 113 unsigned long nr_segs, loff_t pos); 114static int sock_mmap(struct file *file, struct vm_area_struct *vma); 115 116static int sock_close(struct inode *inode, struct file *file); 117static unsigned int sock_poll(struct file *file, 118 struct poll_table_struct *wait); 119static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 120#ifdef CONFIG_COMPAT 121static long compat_sock_ioctl(struct file *file, 122 unsigned int cmd, unsigned long arg); 123#endif 124static int sock_fasync(int fd, struct file *filp, int on); 125static ssize_t sock_sendpage(struct file *file, struct page *page, 126 int offset, size_t size, loff_t *ppos, int more); 127static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 128 struct pipe_inode_info *pipe, size_t len, 129 unsigned int flags); 130 131/* 132 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear 133 * in the operation structures but are done directly via the socketcall() multiplexor. 134 */ 135 136static const struct file_operations socket_file_ops = { 137 .owner = THIS_MODULE, 138 .llseek = no_llseek, 139 .aio_read = sock_aio_read, 140 .aio_write = sock_aio_write, 141 .poll = sock_poll, 142 .unlocked_ioctl = sock_ioctl, 143#ifdef CONFIG_COMPAT 144 .compat_ioctl = compat_sock_ioctl, 145#endif 146 .mmap = sock_mmap, 147 .open = sock_no_open, /* special open code to disallow open via /proc */ 148 .release = sock_close, 149 .fasync = sock_fasync, 150 .sendpage = sock_sendpage, 151 .splice_write = generic_splice_sendpage, 152 .splice_read = sock_splice_read, 153}; 154 155/* 156 * The protocol list. Each protocol is registered in here. 157 */ 158 159static DEFINE_SPINLOCK(net_family_lock); 160static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; 161 162/* 163 * Statistics counters of the socket lists 164 */ 165 166static DEFINE_PER_CPU(int, sockets_in_use); 167 168/* 169 * Support routines. 170 * Move socket addresses back and forth across the kernel/user 171 * divide and look after the messy bits. 172 */ 173 174/** 175 * move_addr_to_kernel - copy a socket address into kernel space 176 * @uaddr: Address in user space 177 * @kaddr: Address in kernel space 178 * @ulen: Length in user space 179 * 180 * The address is copied into kernel space. If the provided address is 181 * too long an error code of -EINVAL is returned. If the copy gives 182 * invalid addresses -EFAULT is returned. On a success 0 is returned. 183 */ 184 185int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr) 186{ 187 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) 188 return -EINVAL; 189 if (ulen == 0) 190 return 0; 191 if (copy_from_user(kaddr, uaddr, ulen)) 192 return -EFAULT; 193 return audit_sockaddr(ulen, kaddr); 194} 195 196/** 197 * move_addr_to_user - copy an address to user space 198 * @kaddr: kernel space address 199 * @klen: length of address in kernel 200 * @uaddr: user space address 201 * @ulen: pointer to user length field 202 * 203 * The value pointed to by ulen on entry is the buffer length available. 204 * This is overwritten with the buffer space used. -EINVAL is returned 205 * if an overlong buffer is specified or a negative buffer size. -EFAULT 206 * is returned if either the buffer or the length field are not 207 * accessible. 208 * After copying the data up to the limit the user specifies, the true 209 * length of the data is written over the length limit the user 210 * specified. Zero is returned for a success. 211 */ 212 213static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen, 214 void __user *uaddr, int __user *ulen) 215{ 216 int err; 217 int len; 218 219 BUG_ON(klen > sizeof(struct sockaddr_storage)); 220 err = get_user(len, ulen); 221 if (err) 222 return err; 223 if (len > klen) 224 len = klen; 225 if (len < 0) 226 return -EINVAL; 227 if (len) { 228 if (audit_sockaddr(klen, kaddr)) 229 return -ENOMEM; 230 if (copy_to_user(uaddr, kaddr, len)) 231 return -EFAULT; 232 } 233 /* 234 * "fromlen shall refer to the value before truncation.." 235 * 1003.1g 236 */ 237 return __put_user(klen, ulen); 238} 239 240static struct kmem_cache *sock_inode_cachep __read_mostly; 241 242static struct inode *sock_alloc_inode(struct super_block *sb) 243{ 244 struct socket_alloc *ei; 245 struct socket_wq *wq; 246 247 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL); 248 if (!ei) 249 return NULL; 250 wq = kmalloc(sizeof(*wq), GFP_KERNEL); 251 if (!wq) { 252 kmem_cache_free(sock_inode_cachep, ei); 253 return NULL; 254 } 255 init_waitqueue_head(&wq->wait); 256 wq->fasync_list = NULL; 257 RCU_INIT_POINTER(ei->socket.wq, wq); 258 259 ei->socket.state = SS_UNCONNECTED; 260 ei->socket.flags = 0; 261 ei->socket.ops = NULL; 262 ei->socket.sk = NULL; 263 ei->socket.file = NULL; 264 265 return &ei->vfs_inode; 266} 267 268static void sock_destroy_inode(struct inode *inode) 269{ 270 struct socket_alloc *ei; 271 struct socket_wq *wq; 272 273 ei = container_of(inode, struct socket_alloc, vfs_inode); 274 wq = rcu_dereference_protected(ei->socket.wq, 1); 275 kfree_rcu(wq, rcu); 276 kmem_cache_free(sock_inode_cachep, ei); 277} 278 279static void init_once(void *foo) 280{ 281 struct socket_alloc *ei = (struct socket_alloc *)foo; 282 283 inode_init_once(&ei->vfs_inode); 284} 285 286static int init_inodecache(void) 287{ 288 sock_inode_cachep = kmem_cache_create("sock_inode_cache", 289 sizeof(struct socket_alloc), 290 0, 291 (SLAB_HWCACHE_ALIGN | 292 SLAB_RECLAIM_ACCOUNT | 293 SLAB_MEM_SPREAD), 294 init_once); 295 if (sock_inode_cachep == NULL) 296 return -ENOMEM; 297 return 0; 298} 299 300static const struct super_operations sockfs_ops = { 301 .alloc_inode = sock_alloc_inode, 302 .destroy_inode = sock_destroy_inode, 303 .statfs = simple_statfs, 304}; 305 306/* 307 * sockfs_dname() is called from d_path(). 308 */ 309static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) 310{ 311 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]", 312 dentry->d_inode->i_ino); 313} 314 315static const struct dentry_operations sockfs_dentry_operations = { 316 .d_dname = sockfs_dname, 317}; 318 319static struct dentry *sockfs_mount(struct file_system_type *fs_type, 320 int flags, const char *dev_name, void *data) 321{ 322 return mount_pseudo(fs_type, "socket:", &sockfs_ops, 323 &sockfs_dentry_operations, SOCKFS_MAGIC); 324} 325 326static struct vfsmount *sock_mnt __read_mostly; 327 328static struct file_system_type sock_fs_type = { 329 .name = "sockfs", 330 .mount = sockfs_mount, 331 .kill_sb = kill_anon_super, 332}; 333 334/* 335 * Obtains the first available file descriptor and sets it up for use. 336 * 337 * These functions create file structures and maps them to fd space 338 * of the current process. On success it returns file descriptor 339 * and file struct implicitly stored in sock->file. 340 * Note that another thread may close file descriptor before we return 341 * from this function. We use the fact that now we do not refer 342 * to socket after mapping. If one day we will need it, this 343 * function will increment ref. count on file by 1. 344 * 345 * In any case returned fd MAY BE not valid! 346 * This race condition is unavoidable 347 * with shared fd spaces, we cannot solve it inside kernel, 348 * but we take care of internal coherence yet. 349 */ 350 351struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname) 352{ 353 struct qstr name = { .name = "" }; 354 struct path path; 355 struct file *file; 356 357 if (dname) { 358 name.name = dname; 359 name.len = strlen(name.name); 360 } else if (sock->sk) { 361 name.name = sock->sk->sk_prot_creator->name; 362 name.len = strlen(name.name); 363 } 364 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name); 365 if (unlikely(!path.dentry)) 366 return ERR_PTR(-ENOMEM); 367 path.mnt = mntget(sock_mnt); 368 369 d_instantiate(path.dentry, SOCK_INODE(sock)); 370 SOCK_INODE(sock)->i_fop = &socket_file_ops; 371 372 file = alloc_file(&path, FMODE_READ | FMODE_WRITE, 373 &socket_file_ops); 374 if (unlikely(!file)) { 375 /* drop dentry, keep inode */ 376 ihold(path.dentry->d_inode); 377 path_put(&path); 378 return ERR_PTR(-ENFILE); 379 } 380 381 sock->file = file; 382 file->f_flags = O_RDWR | (flags & O_NONBLOCK); 383 file->f_pos = 0; 384 file->private_data = sock; 385 return file; 386} 387EXPORT_SYMBOL(sock_alloc_file); 388 389static int sock_map_fd(struct socket *sock, int flags) 390{ 391 struct file *newfile; 392 int fd = get_unused_fd_flags(flags); 393 if (unlikely(fd < 0)) 394 return fd; 395 396 newfile = sock_alloc_file(sock, flags, NULL); 397 if (likely(!IS_ERR(newfile))) { 398 fd_install(fd, newfile); 399 return fd; 400 } 401 402 put_unused_fd(fd); 403 return PTR_ERR(newfile); 404} 405 406struct socket *sock_from_file(struct file *file, int *err) 407{ 408 if (file->f_op == &socket_file_ops) 409 return file->private_data; /* set in sock_map_fd */ 410 411 *err = -ENOTSOCK; 412 return NULL; 413} 414EXPORT_SYMBOL(sock_from_file); 415 416/** 417 * sockfd_lookup - Go from a file number to its socket slot 418 * @fd: file handle 419 * @err: pointer to an error code return 420 * 421 * The file handle passed in is locked and the socket it is bound 422 * too is returned. If an error occurs the err pointer is overwritten 423 * with a negative errno code and NULL is returned. The function checks 424 * for both invalid handles and passing a handle which is not a socket. 425 * 426 * On a success the socket object pointer is returned. 427 */ 428 429struct socket *sockfd_lookup(int fd, int *err) 430{ 431 struct file *file; 432 struct socket *sock; 433 434 file = fget(fd); 435 if (!file) { 436 *err = -EBADF; 437 return NULL; 438 } 439 440 sock = sock_from_file(file, err); 441 if (!sock) 442 fput(file); 443 return sock; 444} 445EXPORT_SYMBOL(sockfd_lookup); 446 447static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) 448{ 449 struct file *file; 450 struct socket *sock; 451 452 *err = -EBADF; 453 file = fget_light(fd, fput_needed); 454 if (file) { 455 sock = sock_from_file(file, err); 456 if (sock) 457 return sock; 458 fput_light(file, *fput_needed); 459 } 460 return NULL; 461} 462 463#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname" 464#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX) 465#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1) 466static ssize_t sockfs_getxattr(struct dentry *dentry, 467 const char *name, void *value, size_t size) 468{ 469 const char *proto_name; 470 size_t proto_size; 471 int error; 472 473 error = -ENODATA; 474 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) { 475 proto_name = dentry->d_name.name; 476 proto_size = strlen(proto_name); 477 478 if (value) { 479 error = -ERANGE; 480 if (proto_size + 1 > size) 481 goto out; 482 483 strncpy(value, proto_name, proto_size + 1); 484 } 485 error = proto_size + 1; 486 } 487 488out: 489 return error; 490} 491 492static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer, 493 size_t size) 494{ 495 ssize_t len; 496 ssize_t used = 0; 497 498 len = security_inode_listsecurity(dentry->d_inode, buffer, size); 499 if (len < 0) 500 return len; 501 used += len; 502 if (buffer) { 503 if (size < used) 504 return -ERANGE; 505 buffer += len; 506 } 507 508 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1); 509 used += len; 510 if (buffer) { 511 if (size < used) 512 return -ERANGE; 513 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len); 514 buffer += len; 515 } 516 517 return used; 518} 519 520static const struct inode_operations sockfs_inode_ops = { 521 .getxattr = sockfs_getxattr, 522 .listxattr = sockfs_listxattr, 523}; 524 525/** 526 * sock_alloc - allocate a socket 527 * 528 * Allocate a new inode and socket object. The two are bound together 529 * and initialised. The socket is then returned. If we are out of inodes 530 * NULL is returned. 531 */ 532 533static struct socket *sock_alloc(void) 534{ 535 struct inode *inode; 536 struct socket *sock; 537 538 inode = new_inode_pseudo(sock_mnt->mnt_sb); 539 if (!inode) 540 return NULL; 541 542 sock = SOCKET_I(inode); 543 544 kmemcheck_annotate_bitfield(sock, type); 545 inode->i_ino = get_next_ino(); 546 inode->i_mode = S_IFSOCK | S_IRWXUGO; 547 inode->i_uid = current_fsuid(); 548 inode->i_gid = current_fsgid(); 549 inode->i_op = &sockfs_inode_ops; 550 551 this_cpu_add(sockets_in_use, 1); 552 return sock; 553} 554 555/* 556 * In theory you can't get an open on this inode, but /proc provides 557 * a back door. Remember to keep it shut otherwise you'll let the 558 * creepy crawlies in. 559 */ 560 561static int sock_no_open(struct inode *irrelevant, struct file *dontcare) 562{ 563 return -ENXIO; 564} 565 566const struct file_operations bad_sock_fops = { 567 .owner = THIS_MODULE, 568 .open = sock_no_open, 569 .llseek = noop_llseek, 570}; 571 572/** 573 * sock_release - close a socket 574 * @sock: socket to close 575 * 576 * The socket is released from the protocol stack if it has a release 577 * callback, and the inode is then released if the socket is bound to 578 * an inode not a file. 579 */ 580 581void sock_release(struct socket *sock) 582{ 583 if (sock->ops) { 584 struct module *owner = sock->ops->owner; 585 586 sock->ops->release(sock); 587 sock->ops = NULL; 588 module_put(owner); 589 } 590 591 if (rcu_dereference_protected(sock->wq, 1)->fasync_list) 592 printk(KERN_ERR "sock_release: fasync list not empty!\n"); 593 594 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags)) 595 return; 596 597 this_cpu_sub(sockets_in_use, 1); 598 if (!sock->file) { 599 iput(SOCK_INODE(sock)); 600 return; 601 } 602 sock->file = NULL; 603} 604EXPORT_SYMBOL(sock_release); 605 606int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags) 607{ 608 *tx_flags = 0; 609 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE)) 610 *tx_flags |= SKBTX_HW_TSTAMP; 611 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE)) 612 *tx_flags |= SKBTX_SW_TSTAMP; 613 if (sock_flag(sk, SOCK_WIFI_STATUS)) 614 *tx_flags |= SKBTX_WIFI_STATUS; 615 return 0; 616} 617EXPORT_SYMBOL(sock_tx_timestamp); 618 619static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock, 620 struct msghdr *msg, size_t size) 621{ 622 struct sock_iocb *si = kiocb_to_siocb(iocb); 623 624 si->sock = sock; 625 si->scm = NULL; 626 si->msg = msg; 627 si->size = size; 628 629 return sock->ops->sendmsg(iocb, sock, msg, size); 630} 631 632static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock, 633 struct msghdr *msg, size_t size) 634{ 635 int err = security_socket_sendmsg(sock, msg, size); 636 637 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size); 638} 639 640int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 641{ 642 struct kiocb iocb; 643 struct sock_iocb siocb; 644 int ret; 645 646 init_sync_kiocb(&iocb, NULL); 647 iocb.private = &siocb; 648 ret = __sock_sendmsg(&iocb, sock, msg, size); 649 if (-EIOCBQUEUED == ret) 650 ret = wait_on_sync_kiocb(&iocb); 651 return ret; 652} 653EXPORT_SYMBOL(sock_sendmsg); 654 655static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size) 656{ 657 struct kiocb iocb; 658 struct sock_iocb siocb; 659 int ret; 660 661 init_sync_kiocb(&iocb, NULL); 662 iocb.private = &siocb; 663 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size); 664 if (-EIOCBQUEUED == ret) 665 ret = wait_on_sync_kiocb(&iocb); 666 return ret; 667} 668 669int kernel_sendmsg(struct socket *sock, struct msghdr *msg, 670 struct kvec *vec, size_t num, size_t size) 671{ 672 mm_segment_t oldfs = get_fs(); 673 int result; 674 675 set_fs(KERNEL_DS); 676 /* 677 * the following is safe, since for compiler definitions of kvec and 678 * iovec are identical, yielding the same in-core layout and alignment 679 */ 680 msg->msg_iov = (struct iovec *)vec; 681 msg->msg_iovlen = num; 682 result = sock_sendmsg(sock, msg, size); 683 set_fs(oldfs); 684 return result; 685} 686EXPORT_SYMBOL(kernel_sendmsg); 687 688static int ktime2ts(ktime_t kt, struct timespec *ts) 689{ 690 if (kt.tv64) { 691 *ts = ktime_to_timespec(kt); 692 return 1; 693 } else { 694 return 0; 695 } 696} 697 698/* 699 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) 700 */ 701void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 702 struct sk_buff *skb) 703{ 704 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); 705 struct timespec ts[3]; 706 int empty = 1; 707 struct skb_shared_hwtstamps *shhwtstamps = 708 skb_hwtstamps(skb); 709 710 /* Race occurred between timestamp enabling and packet 711 receiving. Fill in the current time for now. */ 712 if (need_software_tstamp && skb->tstamp.tv64 == 0) 713 __net_timestamp(skb); 714 715 if (need_software_tstamp) { 716 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { 717 struct timeval tv; 718 skb_get_timestamp(skb, &tv); 719 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, 720 sizeof(tv), &tv); 721 } else { 722 skb_get_timestampns(skb, &ts[0]); 723 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, 724 sizeof(ts[0]), &ts[0]); 725 } 726 } 727 728 729 memset(ts, 0, sizeof(ts)); 730 if (skb->tstamp.tv64 && 731 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) { 732 skb_get_timestampns(skb, ts + 0); 733 empty = 0; 734 } 735 if (shhwtstamps) { 736 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) && 737 ktime2ts(shhwtstamps->syststamp, ts + 1)) 738 empty = 0; 739 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) && 740 ktime2ts(shhwtstamps->hwtstamp, ts + 2)) 741 empty = 0; 742 } 743 if (!empty) 744 put_cmsg(msg, SOL_SOCKET, 745 SCM_TIMESTAMPING, sizeof(ts), &ts); 746} 747EXPORT_SYMBOL_GPL(__sock_recv_timestamp); 748 749void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, 750 struct sk_buff *skb) 751{ 752 int ack; 753 754 if (!sock_flag(sk, SOCK_WIFI_STATUS)) 755 return; 756 if (!skb->wifi_acked_valid) 757 return; 758 759 ack = skb->wifi_acked; 760 761 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); 762} 763EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); 764 765static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, 766 struct sk_buff *skb) 767{ 768 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount) 769 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, 770 sizeof(__u32), &skb->dropcount); 771} 772 773void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, 774 struct sk_buff *skb) 775{ 776 sock_recv_timestamp(msg, sk, skb); 777 sock_recv_drops(msg, sk, skb); 778} 779EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops); 780 781static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock, 782 struct msghdr *msg, size_t size, int flags) 783{ 784 struct sock_iocb *si = kiocb_to_siocb(iocb); 785 786 si->sock = sock; 787 si->scm = NULL; 788 si->msg = msg; 789 si->size = size; 790 si->flags = flags; 791 792 return sock->ops->recvmsg(iocb, sock, msg, size, flags); 793} 794 795static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock, 796 struct msghdr *msg, size_t size, int flags) 797{ 798 int err = security_socket_recvmsg(sock, msg, size, flags); 799 800 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags); 801} 802 803int sock_recvmsg(struct socket *sock, struct msghdr *msg, 804 size_t size, int flags) 805{ 806 struct kiocb iocb; 807 struct sock_iocb siocb; 808 int ret; 809 810 init_sync_kiocb(&iocb, NULL); 811 iocb.private = &siocb; 812 ret = __sock_recvmsg(&iocb, sock, msg, size, flags); 813 if (-EIOCBQUEUED == ret) 814 ret = wait_on_sync_kiocb(&iocb); 815 return ret; 816} 817EXPORT_SYMBOL(sock_recvmsg); 818 819static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 820 size_t size, int flags) 821{ 822 struct kiocb iocb; 823 struct sock_iocb siocb; 824 int ret; 825 826 init_sync_kiocb(&iocb, NULL); 827 iocb.private = &siocb; 828 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags); 829 if (-EIOCBQUEUED == ret) 830 ret = wait_on_sync_kiocb(&iocb); 831 return ret; 832} 833 834/** 835 * kernel_recvmsg - Receive a message from a socket (kernel space) 836 * @sock: The socket to receive the message from 837 * @msg: Received message 838 * @vec: Input s/g array for message data 839 * @num: Size of input s/g array 840 * @size: Number of bytes to read 841 * @flags: Message flags (MSG_DONTWAIT, etc...) 842 * 843 * On return the msg structure contains the scatter/gather array passed in the 844 * vec argument. The array is modified so that it consists of the unfilled 845 * portion of the original array. 846 * 847 * The returned value is the total number of bytes received, or an error. 848 */ 849int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 850 struct kvec *vec, size_t num, size_t size, int flags) 851{ 852 mm_segment_t oldfs = get_fs(); 853 int result; 854 855 set_fs(KERNEL_DS); 856 /* 857 * the following is safe, since for compiler definitions of kvec and 858 * iovec are identical, yielding the same in-core layout and alignment 859 */ 860 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num; 861 result = sock_recvmsg(sock, msg, size, flags); 862 set_fs(oldfs); 863 return result; 864} 865EXPORT_SYMBOL(kernel_recvmsg); 866 867static void sock_aio_dtor(struct kiocb *iocb) 868{ 869 kfree(iocb->private); 870} 871 872static ssize_t sock_sendpage(struct file *file, struct page *page, 873 int offset, size_t size, loff_t *ppos, int more) 874{ 875 struct socket *sock; 876 int flags; 877 878 sock = file->private_data; 879 880 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 881 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */ 882 flags |= more; 883 884 return kernel_sendpage(sock, page, offset, size, flags); 885} 886 887static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 888 struct pipe_inode_info *pipe, size_t len, 889 unsigned int flags) 890{ 891 struct socket *sock = file->private_data; 892 893 if (unlikely(!sock->ops->splice_read)) 894 return -EINVAL; 895 896 return sock->ops->splice_read(sock, ppos, pipe, len, flags); 897} 898 899static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb, 900 struct sock_iocb *siocb) 901{ 902 if (!is_sync_kiocb(iocb)) { 903 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL); 904 if (!siocb) 905 return NULL; 906 iocb->ki_dtor = sock_aio_dtor; 907 } 908 909 siocb->kiocb = iocb; 910 iocb->private = siocb; 911 return siocb; 912} 913 914static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb, 915 struct file *file, const struct iovec *iov, 916 unsigned long nr_segs) 917{ 918 struct socket *sock = file->private_data; 919 size_t size = 0; 920 int i; 921 922 for (i = 0; i < nr_segs; i++) 923 size += iov[i].iov_len; 924 925 msg->msg_name = NULL; 926 msg->msg_namelen = 0; 927 msg->msg_control = NULL; 928 msg->msg_controllen = 0; 929 msg->msg_iov = (struct iovec *)iov; 930 msg->msg_iovlen = nr_segs; 931 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 932 933 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags); 934} 935 936static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov, 937 unsigned long nr_segs, loff_t pos) 938{ 939 struct sock_iocb siocb, *x; 940 941 if (pos != 0) 942 return -ESPIPE; 943 944 if (iocb->ki_left == 0) /* Match SYS5 behaviour */ 945 return 0; 946 947 948 x = alloc_sock_iocb(iocb, &siocb); 949 if (!x) 950 return -ENOMEM; 951 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); 952} 953 954static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb, 955 struct file *file, const struct iovec *iov, 956 unsigned long nr_segs) 957{ 958 struct socket *sock = file->private_data; 959 size_t size = 0; 960 int i; 961 962 for (i = 0; i < nr_segs; i++) 963 size += iov[i].iov_len; 964 965 msg->msg_name = NULL; 966 msg->msg_namelen = 0; 967 msg->msg_control = NULL; 968 msg->msg_controllen = 0; 969 msg->msg_iov = (struct iovec *)iov; 970 msg->msg_iovlen = nr_segs; 971 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0; 972 if (sock->type == SOCK_SEQPACKET) 973 msg->msg_flags |= MSG_EOR; 974 975 return __sock_sendmsg(iocb, sock, msg, size); 976} 977 978static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov, 979 unsigned long nr_segs, loff_t pos) 980{ 981 struct sock_iocb siocb, *x; 982 983 if (pos != 0) 984 return -ESPIPE; 985 986 x = alloc_sock_iocb(iocb, &siocb); 987 if (!x) 988 return -ENOMEM; 989 990 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs); 991} 992 993/* 994 * Atomic setting of ioctl hooks to avoid race 995 * with module unload. 996 */ 997 998static DEFINE_MUTEX(br_ioctl_mutex); 999static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg); 1000 1001void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *)) 1002{ 1003 mutex_lock(&br_ioctl_mutex); 1004 br_ioctl_hook = hook; 1005 mutex_unlock(&br_ioctl_mutex); 1006} 1007EXPORT_SYMBOL(brioctl_set); 1008 1009static DEFINE_MUTEX(vlan_ioctl_mutex); 1010static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 1011 1012void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 1013{ 1014 mutex_lock(&vlan_ioctl_mutex); 1015 vlan_ioctl_hook = hook; 1016 mutex_unlock(&vlan_ioctl_mutex); 1017} 1018EXPORT_SYMBOL(vlan_ioctl_set); 1019 1020static DEFINE_MUTEX(dlci_ioctl_mutex); 1021static int (*dlci_ioctl_hook) (unsigned int, void __user *); 1022 1023void dlci_ioctl_set(int (*hook) (unsigned int, void __user *)) 1024{ 1025 mutex_lock(&dlci_ioctl_mutex); 1026 dlci_ioctl_hook = hook; 1027 mutex_unlock(&dlci_ioctl_mutex); 1028} 1029EXPORT_SYMBOL(dlci_ioctl_set); 1030 1031static long sock_do_ioctl(struct net *net, struct socket *sock, 1032 unsigned int cmd, unsigned long arg) 1033{ 1034 int err; 1035 void __user *argp = (void __user *)arg; 1036 1037 err = sock->ops->ioctl(sock, cmd, arg); 1038 1039 /* 1040 * If this ioctl is unknown try to hand it down 1041 * to the NIC driver. 1042 */ 1043 if (err == -ENOIOCTLCMD) 1044 err = dev_ioctl(net, cmd, argp); 1045 1046 return err; 1047} 1048 1049/* 1050 * With an ioctl, arg may well be a user mode pointer, but we don't know 1051 * what to do with it - that's up to the protocol still. 1052 */ 1053 1054static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 1055{ 1056 struct socket *sock; 1057 struct sock *sk; 1058 void __user *argp = (void __user *)arg; 1059 int pid, err; 1060 struct net *net; 1061 1062 sock = file->private_data; 1063 sk = sock->sk; 1064 net = sock_net(sk); 1065 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) { 1066 err = dev_ioctl(net, cmd, argp); 1067 } else 1068#ifdef CONFIG_WEXT_CORE 1069 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 1070 err = dev_ioctl(net, cmd, argp); 1071 } else 1072#endif 1073 switch (cmd) { 1074 case FIOSETOWN: 1075 case SIOCSPGRP: 1076 err = -EFAULT; 1077 if (get_user(pid, (int __user *)argp)) 1078 break; 1079 err = f_setown(sock->file, pid, 1); 1080 break; 1081 case FIOGETOWN: 1082 case SIOCGPGRP: 1083 err = put_user(f_getown(sock->file), 1084 (int __user *)argp); 1085 break; 1086 case SIOCGIFBR: 1087 case SIOCSIFBR: 1088 case SIOCBRADDBR: 1089 case SIOCBRDELBR: 1090 err = -ENOPKG; 1091 if (!br_ioctl_hook) 1092 request_module("bridge"); 1093 1094 mutex_lock(&br_ioctl_mutex); 1095 if (br_ioctl_hook) 1096 err = br_ioctl_hook(net, cmd, argp); 1097 mutex_unlock(&br_ioctl_mutex); 1098 break; 1099 case SIOCGIFVLAN: 1100 case SIOCSIFVLAN: 1101 err = -ENOPKG; 1102 if (!vlan_ioctl_hook) 1103 request_module("8021q"); 1104 1105 mutex_lock(&vlan_ioctl_mutex); 1106 if (vlan_ioctl_hook) 1107 err = vlan_ioctl_hook(net, argp); 1108 mutex_unlock(&vlan_ioctl_mutex); 1109 break; 1110 case SIOCADDDLCI: 1111 case SIOCDELDLCI: 1112 err = -ENOPKG; 1113 if (!dlci_ioctl_hook) 1114 request_module("dlci"); 1115 1116 mutex_lock(&dlci_ioctl_mutex); 1117 if (dlci_ioctl_hook) 1118 err = dlci_ioctl_hook(cmd, argp); 1119 mutex_unlock(&dlci_ioctl_mutex); 1120 break; 1121 default: 1122 err = sock_do_ioctl(net, sock, cmd, arg); 1123 break; 1124 } 1125 return err; 1126} 1127 1128int sock_create_lite(int family, int type, int protocol, struct socket **res) 1129{ 1130 int err; 1131 struct socket *sock = NULL; 1132 1133 err = security_socket_create(family, type, protocol, 1); 1134 if (err) 1135 goto out; 1136 1137 sock = sock_alloc(); 1138 if (!sock) { 1139 err = -ENOMEM; 1140 goto out; 1141 } 1142 1143 sock->type = type; 1144 err = security_socket_post_create(sock, family, type, protocol, 1); 1145 if (err) 1146 goto out_release; 1147 1148out: 1149 *res = sock; 1150 return err; 1151out_release: 1152 sock_release(sock); 1153 sock = NULL; 1154 goto out; 1155} 1156EXPORT_SYMBOL(sock_create_lite); 1157 1158/* No kernel lock held - perfect */ 1159static unsigned int sock_poll(struct file *file, poll_table *wait) 1160{ 1161 struct socket *sock; 1162 1163 /* 1164 * We can't return errors to poll, so it's either yes or no. 1165 */ 1166 sock = file->private_data; 1167 return sock->ops->poll(file, sock, wait); 1168} 1169 1170static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1171{ 1172 struct socket *sock = file->private_data; 1173 1174 return sock->ops->mmap(file, sock, vma); 1175} 1176 1177static int sock_close(struct inode *inode, struct file *filp) 1178{ 1179 /* 1180 * It was possible the inode is NULL we were 1181 * closing an unfinished socket. 1182 */ 1183 1184 if (!inode) { 1185 printk(KERN_DEBUG "sock_close: NULL inode\n"); 1186 return 0; 1187 } 1188 sock_release(SOCKET_I(inode)); 1189 return 0; 1190} 1191 1192/* 1193 * Update the socket async list 1194 * 1195 * Fasync_list locking strategy. 1196 * 1197 * 1. fasync_list is modified only under process context socket lock 1198 * i.e. under semaphore. 1199 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1200 * or under socket lock 1201 */ 1202 1203static int sock_fasync(int fd, struct file *filp, int on) 1204{ 1205 struct socket *sock = filp->private_data; 1206 struct sock *sk = sock->sk; 1207 struct socket_wq *wq; 1208 1209 if (sk == NULL) 1210 return -EINVAL; 1211 1212 lock_sock(sk); 1213 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk)); 1214 fasync_helper(fd, filp, on, &wq->fasync_list); 1215 1216 if (!wq->fasync_list) 1217 sock_reset_flag(sk, SOCK_FASYNC); 1218 else 1219 sock_set_flag(sk, SOCK_FASYNC); 1220 1221 release_sock(sk); 1222 return 0; 1223} 1224 1225/* This function may be called only under socket lock or callback_lock or rcu_lock */ 1226 1227int sock_wake_async(struct socket *sock, int how, int band) 1228{ 1229 struct socket_wq *wq; 1230 1231 if (!sock) 1232 return -1; 1233 rcu_read_lock(); 1234 wq = rcu_dereference(sock->wq); 1235 if (!wq || !wq->fasync_list) { 1236 rcu_read_unlock(); 1237 return -1; 1238 } 1239 switch (how) { 1240 case SOCK_WAKE_WAITD: 1241 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags)) 1242 break; 1243 goto call_kill; 1244 case SOCK_WAKE_SPACE: 1245 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags)) 1246 break; 1247 /* fall through */ 1248 case SOCK_WAKE_IO: 1249call_kill: 1250 kill_fasync(&wq->fasync_list, SIGIO, band); 1251 break; 1252 case SOCK_WAKE_URG: 1253 kill_fasync(&wq->fasync_list, SIGURG, band); 1254 } 1255 rcu_read_unlock(); 1256 return 0; 1257} 1258EXPORT_SYMBOL(sock_wake_async); 1259 1260int __sock_create(struct net *net, int family, int type, int protocol, 1261 struct socket **res, int kern) 1262{ 1263 int err; 1264 struct socket *sock; 1265 const struct net_proto_family *pf; 1266 1267 /* 1268 * Check protocol is in range 1269 */ 1270 if (family < 0 || family >= NPROTO) 1271 return -EAFNOSUPPORT; 1272 if (type < 0 || type >= SOCK_MAX) 1273 return -EINVAL; 1274 1275 /* Compatibility. 1276 1277 This uglymoron is moved from INET layer to here to avoid 1278 deadlock in module load. 1279 */ 1280 if (family == PF_INET && type == SOCK_PACKET) { 1281 static int warned; 1282 if (!warned) { 1283 warned = 1; 1284 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1285 current->comm); 1286 } 1287 family = PF_PACKET; 1288 } 1289 1290 err = security_socket_create(family, type, protocol, kern); 1291 if (err) 1292 return err; 1293 1294 /* 1295 * Allocate the socket and allow the family to set things up. if 1296 * the protocol is 0, the family is instructed to select an appropriate 1297 * default. 1298 */ 1299 sock = sock_alloc(); 1300 if (!sock) { 1301 net_warn_ratelimited("socket: no more sockets\n"); 1302 return -ENFILE; /* Not exactly a match, but its the 1303 closest posix thing */ 1304 } 1305 1306 sock->type = type; 1307 1308#ifdef CONFIG_MODULES 1309 /* Attempt to load a protocol module if the find failed. 1310 * 1311 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1312 * requested real, full-featured networking support upon configuration. 1313 * Otherwise module support will break! 1314 */ 1315 if (rcu_access_pointer(net_families[family]) == NULL) 1316 request_module("net-pf-%d", family); 1317#endif 1318 1319 rcu_read_lock(); 1320 pf = rcu_dereference(net_families[family]); 1321 err = -EAFNOSUPPORT; 1322 if (!pf) 1323 goto out_release; 1324 1325 /* 1326 * We will call the ->create function, that possibly is in a loadable 1327 * module, so we have to bump that loadable module refcnt first. 1328 */ 1329 if (!try_module_get(pf->owner)) 1330 goto out_release; 1331 1332 /* Now protected by module ref count */ 1333 rcu_read_unlock(); 1334 1335 err = pf->create(net, sock, protocol, kern); 1336 if (err < 0) 1337 goto out_module_put; 1338 1339 /* 1340 * Now to bump the refcnt of the [loadable] module that owns this 1341 * socket at sock_release time we decrement its refcnt. 1342 */ 1343 if (!try_module_get(sock->ops->owner)) 1344 goto out_module_busy; 1345 1346 /* 1347 * Now that we're done with the ->create function, the [loadable] 1348 * module can have its refcnt decremented 1349 */ 1350 module_put(pf->owner); 1351 err = security_socket_post_create(sock, family, type, protocol, kern); 1352 if (err) 1353 goto out_sock_release; 1354 *res = sock; 1355 1356 return 0; 1357 1358out_module_busy: 1359 err = -EAFNOSUPPORT; 1360out_module_put: 1361 sock->ops = NULL; 1362 module_put(pf->owner); 1363out_sock_release: 1364 sock_release(sock); 1365 return err; 1366 1367out_release: 1368 rcu_read_unlock(); 1369 goto out_sock_release; 1370} 1371EXPORT_SYMBOL(__sock_create); 1372 1373int sock_create(int family, int type, int protocol, struct socket **res) 1374{ 1375 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1376} 1377EXPORT_SYMBOL(sock_create); 1378 1379int sock_create_kern(int family, int type, int protocol, struct socket **res) 1380{ 1381 return __sock_create(&init_net, family, type, protocol, res, 1); 1382} 1383EXPORT_SYMBOL(sock_create_kern); 1384 1385SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1386{ 1387 int retval; 1388 struct socket *sock; 1389 int flags; 1390 1391 /* Check the SOCK_* constants for consistency. */ 1392 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1393 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1394 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1395 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1396 1397 flags = type & ~SOCK_TYPE_MASK; 1398 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1399 return -EINVAL; 1400 type &= SOCK_TYPE_MASK; 1401 1402 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1403 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1404 1405 retval = sock_create(family, type, protocol, &sock); 1406 if (retval < 0) 1407 goto out; 1408 1409 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1410 if (retval < 0) 1411 goto out_release; 1412 1413out: 1414 /* It may be already another descriptor 8) Not kernel problem. */ 1415 return retval; 1416 1417out_release: 1418 sock_release(sock); 1419 return retval; 1420} 1421 1422/* 1423 * Create a pair of connected sockets. 1424 */ 1425 1426SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1427 int __user *, usockvec) 1428{ 1429 struct socket *sock1, *sock2; 1430 int fd1, fd2, err; 1431 struct file *newfile1, *newfile2; 1432 int flags; 1433 1434 flags = type & ~SOCK_TYPE_MASK; 1435 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1436 return -EINVAL; 1437 type &= SOCK_TYPE_MASK; 1438 1439 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1440 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1441 1442 /* 1443 * Obtain the first socket and check if the underlying protocol 1444 * supports the socketpair call. 1445 */ 1446 1447 err = sock_create(family, type, protocol, &sock1); 1448 if (err < 0) 1449 goto out; 1450 1451 err = sock_create(family, type, protocol, &sock2); 1452 if (err < 0) 1453 goto out_release_1; 1454 1455 err = sock1->ops->socketpair(sock1, sock2); 1456 if (err < 0) 1457 goto out_release_both; 1458 1459 fd1 = get_unused_fd_flags(flags); 1460 if (unlikely(fd1 < 0)) { 1461 err = fd1; 1462 goto out_release_both; 1463 } 1464 fd2 = get_unused_fd_flags(flags); 1465 if (unlikely(fd2 < 0)) { 1466 err = fd2; 1467 put_unused_fd(fd1); 1468 goto out_release_both; 1469 } 1470 1471 newfile1 = sock_alloc_file(sock1, flags, NULL); 1472 if (unlikely(IS_ERR(newfile1))) { 1473 err = PTR_ERR(newfile1); 1474 put_unused_fd(fd1); 1475 put_unused_fd(fd2); 1476 goto out_release_both; 1477 } 1478 1479 newfile2 = sock_alloc_file(sock2, flags, NULL); 1480 if (IS_ERR(newfile2)) { 1481 err = PTR_ERR(newfile2); 1482 fput(newfile1); 1483 put_unused_fd(fd1); 1484 put_unused_fd(fd2); 1485 sock_release(sock2); 1486 goto out; 1487 } 1488 1489 audit_fd_pair(fd1, fd2); 1490 fd_install(fd1, newfile1); 1491 fd_install(fd2, newfile2); 1492 /* fd1 and fd2 may be already another descriptors. 1493 * Not kernel problem. 1494 */ 1495 1496 err = put_user(fd1, &usockvec[0]); 1497 if (!err) 1498 err = put_user(fd2, &usockvec[1]); 1499 if (!err) 1500 return 0; 1501 1502 sys_close(fd2); 1503 sys_close(fd1); 1504 return err; 1505 1506out_release_both: 1507 sock_release(sock2); 1508out_release_1: 1509 sock_release(sock1); 1510out: 1511 return err; 1512} 1513 1514/* 1515 * Bind a name to a socket. Nothing much to do here since it's 1516 * the protocol's responsibility to handle the local address. 1517 * 1518 * We move the socket address to kernel space before we call 1519 * the protocol layer (having also checked the address is ok). 1520 */ 1521 1522SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1523{ 1524 struct socket *sock; 1525 struct sockaddr_storage address; 1526 int err, fput_needed; 1527 1528 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1529 if (sock) { 1530 err = move_addr_to_kernel(umyaddr, addrlen, &address); 1531 if (err >= 0) { 1532 err = security_socket_bind(sock, 1533 (struct sockaddr *)&address, 1534 addrlen); 1535 if (!err) 1536 err = sock->ops->bind(sock, 1537 (struct sockaddr *) 1538 &address, addrlen); 1539 } 1540 fput_light(sock->file, fput_needed); 1541 } 1542 return err; 1543} 1544 1545/* 1546 * Perform a listen. Basically, we allow the protocol to do anything 1547 * necessary for a listen, and if that works, we mark the socket as 1548 * ready for listening. 1549 */ 1550 1551SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1552{ 1553 struct socket *sock; 1554 int err, fput_needed; 1555 int somaxconn; 1556 1557 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1558 if (sock) { 1559 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn; 1560 if ((unsigned int)backlog > somaxconn) 1561 backlog = somaxconn; 1562 1563 err = security_socket_listen(sock, backlog); 1564 if (!err) 1565 err = sock->ops->listen(sock, backlog); 1566 1567 fput_light(sock->file, fput_needed); 1568 } 1569 return err; 1570} 1571 1572/* 1573 * For accept, we attempt to create a new socket, set up the link 1574 * with the client, wake up the client, then return the new 1575 * connected fd. We collect the address of the connector in kernel 1576 * space and move it to user at the very end. This is unclean because 1577 * we open the socket then return an error. 1578 * 1579 * 1003.1g adds the ability to recvmsg() to query connection pending 1580 * status to recvmsg. We need to add that support in a way thats 1581 * clean when we restucture accept also. 1582 */ 1583 1584SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 1585 int __user *, upeer_addrlen, int, flags) 1586{ 1587 struct socket *sock, *newsock; 1588 struct file *newfile; 1589 int err, len, newfd, fput_needed; 1590 struct sockaddr_storage address; 1591 1592 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1593 return -EINVAL; 1594 1595 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1596 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1597 1598 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1599 if (!sock) 1600 goto out; 1601 1602 err = -ENFILE; 1603 newsock = sock_alloc(); 1604 if (!newsock) 1605 goto out_put; 1606 1607 newsock->type = sock->type; 1608 newsock->ops = sock->ops; 1609 1610 /* 1611 * We don't need try_module_get here, as the listening socket (sock) 1612 * has the protocol module (sock->ops->owner) held. 1613 */ 1614 __module_get(newsock->ops->owner); 1615 1616 newfd = get_unused_fd_flags(flags); 1617 if (unlikely(newfd < 0)) { 1618 err = newfd; 1619 sock_release(newsock); 1620 goto out_put; 1621 } 1622 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); 1623 if (unlikely(IS_ERR(newfile))) { 1624 err = PTR_ERR(newfile); 1625 put_unused_fd(newfd); 1626 sock_release(newsock); 1627 goto out_put; 1628 } 1629 1630 err = security_socket_accept(sock, newsock); 1631 if (err) 1632 goto out_fd; 1633 1634 err = sock->ops->accept(sock, newsock, sock->file->f_flags); 1635 if (err < 0) 1636 goto out_fd; 1637 1638 if (upeer_sockaddr) { 1639 if (newsock->ops->getname(newsock, (struct sockaddr *)&address, 1640 &len, 2) < 0) { 1641 err = -ECONNABORTED; 1642 goto out_fd; 1643 } 1644 err = move_addr_to_user(&address, 1645 len, upeer_sockaddr, upeer_addrlen); 1646 if (err < 0) 1647 goto out_fd; 1648 } 1649 1650 /* File flags are not inherited via accept() unlike another OSes. */ 1651 1652 fd_install(newfd, newfile); 1653 err = newfd; 1654 1655out_put: 1656 fput_light(sock->file, fput_needed); 1657out: 1658 return err; 1659out_fd: 1660 fput(newfile); 1661 put_unused_fd(newfd); 1662 goto out_put; 1663} 1664 1665SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 1666 int __user *, upeer_addrlen) 1667{ 1668 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 1669} 1670 1671/* 1672 * Attempt to connect to a socket with the server address. The address 1673 * is in user space so we verify it is OK and move it to kernel space. 1674 * 1675 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 1676 * break bindings 1677 * 1678 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 1679 * other SEQPACKET protocols that take time to connect() as it doesn't 1680 * include the -EINPROGRESS status for such sockets. 1681 */ 1682 1683SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 1684 int, addrlen) 1685{ 1686 struct socket *sock; 1687 struct sockaddr_storage address; 1688 int err, fput_needed; 1689 1690 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1691 if (!sock) 1692 goto out; 1693 err = move_addr_to_kernel(uservaddr, addrlen, &address); 1694 if (err < 0) 1695 goto out_put; 1696 1697 err = 1698 security_socket_connect(sock, (struct sockaddr *)&address, addrlen); 1699 if (err) 1700 goto out_put; 1701 1702 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, 1703 sock->file->f_flags); 1704out_put: 1705 fput_light(sock->file, fput_needed); 1706out: 1707 return err; 1708} 1709 1710/* 1711 * Get the local address ('name') of a socket object. Move the obtained 1712 * name to user space. 1713 */ 1714 1715SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 1716 int __user *, usockaddr_len) 1717{ 1718 struct socket *sock; 1719 struct sockaddr_storage address; 1720 int len, err, fput_needed; 1721 1722 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1723 if (!sock) 1724 goto out; 1725 1726 err = security_socket_getsockname(sock); 1727 if (err) 1728 goto out_put; 1729 1730 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0); 1731 if (err) 1732 goto out_put; 1733 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len); 1734 1735out_put: 1736 fput_light(sock->file, fput_needed); 1737out: 1738 return err; 1739} 1740 1741/* 1742 * Get the remote address ('name') of a socket object. Move the obtained 1743 * name to user space. 1744 */ 1745 1746SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 1747 int __user *, usockaddr_len) 1748{ 1749 struct socket *sock; 1750 struct sockaddr_storage address; 1751 int len, err, fput_needed; 1752 1753 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1754 if (sock != NULL) { 1755 err = security_socket_getpeername(sock); 1756 if (err) { 1757 fput_light(sock->file, fput_needed); 1758 return err; 1759 } 1760 1761 err = 1762 sock->ops->getname(sock, (struct sockaddr *)&address, &len, 1763 1); 1764 if (!err) 1765 err = move_addr_to_user(&address, len, usockaddr, 1766 usockaddr_len); 1767 fput_light(sock->file, fput_needed); 1768 } 1769 return err; 1770} 1771 1772/* 1773 * Send a datagram to a given address. We move the address into kernel 1774 * space and check the user space data area is readable before invoking 1775 * the protocol. 1776 */ 1777 1778SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 1779 unsigned int, flags, struct sockaddr __user *, addr, 1780 int, addr_len) 1781{ 1782 struct socket *sock; 1783 struct sockaddr_storage address; 1784 int err; 1785 struct msghdr msg; 1786 struct iovec iov; 1787 int fput_needed; 1788 1789 if (len > INT_MAX) 1790 len = INT_MAX; 1791 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1792 if (!sock) 1793 goto out; 1794 1795 iov.iov_base = buff; 1796 iov.iov_len = len; 1797 msg.msg_name = NULL; 1798 msg.msg_iov = &iov; 1799 msg.msg_iovlen = 1; 1800 msg.msg_control = NULL; 1801 msg.msg_controllen = 0; 1802 msg.msg_namelen = 0; 1803 if (addr) { 1804 err = move_addr_to_kernel(addr, addr_len, &address); 1805 if (err < 0) 1806 goto out_put; 1807 msg.msg_name = (struct sockaddr *)&address; 1808 msg.msg_namelen = addr_len; 1809 } 1810 if (sock->file->f_flags & O_NONBLOCK) 1811 flags |= MSG_DONTWAIT; 1812 msg.msg_flags = flags; 1813 err = sock_sendmsg(sock, &msg, len); 1814 1815out_put: 1816 fput_light(sock->file, fput_needed); 1817out: 1818 return err; 1819} 1820 1821/* 1822 * Send a datagram down a socket. 1823 */ 1824 1825SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 1826 unsigned int, flags) 1827{ 1828 return sys_sendto(fd, buff, len, flags, NULL, 0); 1829} 1830 1831/* 1832 * Receive a frame from the socket and optionally record the address of the 1833 * sender. We verify the buffers are writable and if needed move the 1834 * sender address from kernel to user space. 1835 */ 1836 1837SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 1838 unsigned int, flags, struct sockaddr __user *, addr, 1839 int __user *, addr_len) 1840{ 1841 struct socket *sock; 1842 struct iovec iov; 1843 struct msghdr msg; 1844 struct sockaddr_storage address; 1845 int err, err2; 1846 int fput_needed; 1847 1848 if (size > INT_MAX) 1849 size = INT_MAX; 1850 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1851 if (!sock) 1852 goto out; 1853 1854 msg.msg_control = NULL; 1855 msg.msg_controllen = 0; 1856 msg.msg_iovlen = 1; 1857 msg.msg_iov = &iov; 1858 iov.iov_len = size; 1859 iov.iov_base = ubuf; 1860 /* Save some cycles and don't copy the address if not needed */ 1861 msg.msg_name = addr ? (struct sockaddr *)&address : NULL; 1862 /* We assume all kernel code knows the size of sockaddr_storage */ 1863 msg.msg_namelen = 0; 1864 if (sock->file->f_flags & O_NONBLOCK) 1865 flags |= MSG_DONTWAIT; 1866 err = sock_recvmsg(sock, &msg, size, flags); 1867 1868 if (err >= 0 && addr != NULL) { 1869 err2 = move_addr_to_user(&address, 1870 msg.msg_namelen, addr, addr_len); 1871 if (err2 < 0) 1872 err = err2; 1873 } 1874 1875 fput_light(sock->file, fput_needed); 1876out: 1877 return err; 1878} 1879 1880/* 1881 * Receive a datagram from a socket. 1882 */ 1883 1884asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size, 1885 unsigned int flags) 1886{ 1887 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 1888} 1889 1890/* 1891 * Set a socket option. Because we don't know the option lengths we have 1892 * to pass the user mode parameter for the protocols to sort out. 1893 */ 1894 1895SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 1896 char __user *, optval, int, optlen) 1897{ 1898 int err, fput_needed; 1899 struct socket *sock; 1900 1901 if (optlen < 0) 1902 return -EINVAL; 1903 1904 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1905 if (sock != NULL) { 1906 err = security_socket_setsockopt(sock, level, optname); 1907 if (err) 1908 goto out_put; 1909 1910 if (level == SOL_SOCKET) 1911 err = 1912 sock_setsockopt(sock, level, optname, optval, 1913 optlen); 1914 else 1915 err = 1916 sock->ops->setsockopt(sock, level, optname, optval, 1917 optlen); 1918out_put: 1919 fput_light(sock->file, fput_needed); 1920 } 1921 return err; 1922} 1923 1924/* 1925 * Get a socket option. Because we don't know the option lengths we have 1926 * to pass a user mode parameter for the protocols to sort out. 1927 */ 1928 1929SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 1930 char __user *, optval, int __user *, optlen) 1931{ 1932 int err, fput_needed; 1933 struct socket *sock; 1934 1935 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1936 if (sock != NULL) { 1937 err = security_socket_getsockopt(sock, level, optname); 1938 if (err) 1939 goto out_put; 1940 1941 if (level == SOL_SOCKET) 1942 err = 1943 sock_getsockopt(sock, level, optname, optval, 1944 optlen); 1945 else 1946 err = 1947 sock->ops->getsockopt(sock, level, optname, optval, 1948 optlen); 1949out_put: 1950 fput_light(sock->file, fput_needed); 1951 } 1952 return err; 1953} 1954 1955/* 1956 * Shutdown a socket. 1957 */ 1958 1959SYSCALL_DEFINE2(shutdown, int, fd, int, how) 1960{ 1961 int err, fput_needed; 1962 struct socket *sock; 1963 1964 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1965 if (sock != NULL) { 1966 err = security_socket_shutdown(sock, how); 1967 if (!err) 1968 err = sock->ops->shutdown(sock, how); 1969 fput_light(sock->file, fput_needed); 1970 } 1971 return err; 1972} 1973 1974/* A couple of helpful macros for getting the address of the 32/64 bit 1975 * fields which are the same type (int / unsigned) on our platforms. 1976 */ 1977#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 1978#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 1979#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 1980 1981struct used_address { 1982 struct sockaddr_storage name; 1983 unsigned int name_len; 1984}; 1985 1986static int copy_msghdr_from_user(struct msghdr *kmsg, 1987 struct msghdr __user *umsg) 1988{ 1989 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr))) 1990 return -EFAULT; 1991 1992 if (kmsg->msg_namelen < 0) 1993 return -EINVAL; 1994 1995 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) 1996 kmsg->msg_namelen = sizeof(struct sockaddr_storage); 1997 return 0; 1998} 1999 2000static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg, 2001 struct msghdr *msg_sys, unsigned int flags, 2002 struct used_address *used_address) 2003{ 2004 struct compat_msghdr __user *msg_compat = 2005 (struct compat_msghdr __user *)msg; 2006 struct sockaddr_storage address; 2007 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2008 unsigned char ctl[sizeof(struct cmsghdr) + 20] 2009 __attribute__ ((aligned(sizeof(__kernel_size_t)))); 2010 /* 20 is size of ipv6_pktinfo */ 2011 unsigned char *ctl_buf = ctl; 2012 int err, ctl_len, total_len; 2013 2014 err = -EFAULT; 2015 if (MSG_CMSG_COMPAT & flags) { 2016 if (get_compat_msghdr(msg_sys, msg_compat)) 2017 return -EFAULT; 2018 } else { 2019 err = copy_msghdr_from_user(msg_sys, msg); 2020 if (err) 2021 return err; 2022 } 2023 2024 if (msg_sys->msg_iovlen > UIO_FASTIOV) { 2025 err = -EMSGSIZE; 2026 if (msg_sys->msg_iovlen > UIO_MAXIOV) 2027 goto out; 2028 err = -ENOMEM; 2029 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec), 2030 GFP_KERNEL); 2031 if (!iov) 2032 goto out; 2033 } 2034 2035 /* This will also move the address data into kernel space */ 2036 if (MSG_CMSG_COMPAT & flags) { 2037 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ); 2038 } else 2039 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ); 2040 if (err < 0) 2041 goto out_freeiov; 2042 total_len = err; 2043 2044 err = -ENOBUFS; 2045 2046 if (msg_sys->msg_controllen > INT_MAX) 2047 goto out_freeiov; 2048 ctl_len = msg_sys->msg_controllen; 2049 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 2050 err = 2051 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, 2052 sizeof(ctl)); 2053 if (err) 2054 goto out_freeiov; 2055 ctl_buf = msg_sys->msg_control; 2056 ctl_len = msg_sys->msg_controllen; 2057 } else if (ctl_len) { 2058 if (ctl_len > sizeof(ctl)) { 2059 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 2060 if (ctl_buf == NULL) 2061 goto out_freeiov; 2062 } 2063 err = -EFAULT; 2064 /* 2065 * Careful! Before this, msg_sys->msg_control contains a user pointer. 2066 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted 2067 * checking falls down on this. 2068 */ 2069 if (copy_from_user(ctl_buf, 2070 (void __user __force *)msg_sys->msg_control, 2071 ctl_len)) 2072 goto out_freectl; 2073 msg_sys->msg_control = ctl_buf; 2074 } 2075 msg_sys->msg_flags = flags; 2076 2077 if (sock->file->f_flags & O_NONBLOCK) 2078 msg_sys->msg_flags |= MSG_DONTWAIT; 2079 /* 2080 * If this is sendmmsg() and current destination address is same as 2081 * previously succeeded address, omit asking LSM's decision. 2082 * used_address->name_len is initialized to UINT_MAX so that the first 2083 * destination address never matches. 2084 */ 2085 if (used_address && msg_sys->msg_name && 2086 used_address->name_len == msg_sys->msg_namelen && 2087 !memcmp(&used_address->name, msg_sys->msg_name, 2088 used_address->name_len)) { 2089 err = sock_sendmsg_nosec(sock, msg_sys, total_len); 2090 goto out_freectl; 2091 } 2092 err = sock_sendmsg(sock, msg_sys, total_len); 2093 /* 2094 * If this is sendmmsg() and sending to current destination address was 2095 * successful, remember it. 2096 */ 2097 if (used_address && err >= 0) { 2098 used_address->name_len = msg_sys->msg_namelen; 2099 if (msg_sys->msg_name) 2100 memcpy(&used_address->name, msg_sys->msg_name, 2101 used_address->name_len); 2102 } 2103 2104out_freectl: 2105 if (ctl_buf != ctl) 2106 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 2107out_freeiov: 2108 if (iov != iovstack) 2109 kfree(iov); 2110out: 2111 return err; 2112} 2113 2114/* 2115 * BSD sendmsg interface 2116 */ 2117 2118long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags) 2119{ 2120 int fput_needed, err; 2121 struct msghdr msg_sys; 2122 struct socket *sock; 2123 2124 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2125 if (!sock) 2126 goto out; 2127 2128 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL); 2129 2130 fput_light(sock->file, fput_needed); 2131out: 2132 return err; 2133} 2134 2135SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags) 2136{ 2137 if (flags & MSG_CMSG_COMPAT) 2138 return -EINVAL; 2139 return __sys_sendmsg(fd, msg, flags); 2140} 2141 2142/* 2143 * Linux sendmmsg interface 2144 */ 2145 2146int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2147 unsigned int flags) 2148{ 2149 int fput_needed, err, datagrams; 2150 struct socket *sock; 2151 struct mmsghdr __user *entry; 2152 struct compat_mmsghdr __user *compat_entry; 2153 struct msghdr msg_sys; 2154 struct used_address used_address; 2155 2156 if (vlen > UIO_MAXIOV) 2157 vlen = UIO_MAXIOV; 2158 2159 datagrams = 0; 2160 2161 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2162 if (!sock) 2163 return err; 2164 2165 used_address.name_len = UINT_MAX; 2166 entry = mmsg; 2167 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2168 err = 0; 2169 2170 while (datagrams < vlen) { 2171 if (MSG_CMSG_COMPAT & flags) { 2172 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry, 2173 &msg_sys, flags, &used_address); 2174 if (err < 0) 2175 break; 2176 err = __put_user(err, &compat_entry->msg_len); 2177 ++compat_entry; 2178 } else { 2179 err = ___sys_sendmsg(sock, 2180 (struct msghdr __user *)entry, 2181 &msg_sys, flags, &used_address); 2182 if (err < 0) 2183 break; 2184 err = put_user(err, &entry->msg_len); 2185 ++entry; 2186 } 2187 2188 if (err) 2189 break; 2190 ++datagrams; 2191 } 2192 2193 fput_light(sock->file, fput_needed); 2194 2195 /* We only return an error if no datagrams were able to be sent */ 2196 if (datagrams != 0) 2197 return datagrams; 2198 2199 return err; 2200} 2201 2202SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, 2203 unsigned int, vlen, unsigned int, flags) 2204{ 2205 if (flags & MSG_CMSG_COMPAT) 2206 return -EINVAL; 2207 return __sys_sendmmsg(fd, mmsg, vlen, flags); 2208} 2209 2210static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg, 2211 struct msghdr *msg_sys, unsigned int flags, int nosec) 2212{ 2213 struct compat_msghdr __user *msg_compat = 2214 (struct compat_msghdr __user *)msg; 2215 struct iovec iovstack[UIO_FASTIOV]; 2216 struct iovec *iov = iovstack; 2217 unsigned long cmsg_ptr; 2218 int err, total_len, len; 2219 2220 /* kernel mode address */ 2221 struct sockaddr_storage addr; 2222 2223 /* user mode address pointers */ 2224 struct sockaddr __user *uaddr; 2225 int __user *uaddr_len; 2226 2227 if (MSG_CMSG_COMPAT & flags) { 2228 if (get_compat_msghdr(msg_sys, msg_compat)) 2229 return -EFAULT; 2230 } else { 2231 err = copy_msghdr_from_user(msg_sys, msg); 2232 if (err) 2233 return err; 2234 } 2235 2236 if (msg_sys->msg_iovlen > UIO_FASTIOV) { 2237 err = -EMSGSIZE; 2238 if (msg_sys->msg_iovlen > UIO_MAXIOV) 2239 goto out; 2240 err = -ENOMEM; 2241 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec), 2242 GFP_KERNEL); 2243 if (!iov) 2244 goto out; 2245 } 2246 2247 /* Save the user-mode address (verify_iovec will change the 2248 * kernel msghdr to use the kernel address space) 2249 */ 2250 uaddr = (__force void __user *)msg_sys->msg_name; 2251 uaddr_len = COMPAT_NAMELEN(msg); 2252 if (MSG_CMSG_COMPAT & flags) 2253 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE); 2254 else 2255 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE); 2256 if (err < 0) 2257 goto out_freeiov; 2258 total_len = err; 2259 2260 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2261 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2262 2263 /* We assume all kernel code knows the size of sockaddr_storage */ 2264 msg_sys->msg_namelen = 0; 2265 2266 if (sock->file->f_flags & O_NONBLOCK) 2267 flags |= MSG_DONTWAIT; 2268 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, 2269 total_len, flags); 2270 if (err < 0) 2271 goto out_freeiov; 2272 len = err; 2273 2274 if (uaddr != NULL) { 2275 err = move_addr_to_user(&addr, 2276 msg_sys->msg_namelen, uaddr, 2277 uaddr_len); 2278 if (err < 0) 2279 goto out_freeiov; 2280 } 2281 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2282 COMPAT_FLAGS(msg)); 2283 if (err) 2284 goto out_freeiov; 2285 if (MSG_CMSG_COMPAT & flags) 2286 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2287 &msg_compat->msg_controllen); 2288 else 2289 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2290 &msg->msg_controllen); 2291 if (err) 2292 goto out_freeiov; 2293 err = len; 2294 2295out_freeiov: 2296 if (iov != iovstack) 2297 kfree(iov); 2298out: 2299 return err; 2300} 2301 2302/* 2303 * BSD recvmsg interface 2304 */ 2305 2306long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags) 2307{ 2308 int fput_needed, err; 2309 struct msghdr msg_sys; 2310 struct socket *sock; 2311 2312 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2313 if (!sock) 2314 goto out; 2315 2316 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2317 2318 fput_light(sock->file, fput_needed); 2319out: 2320 return err; 2321} 2322 2323SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg, 2324 unsigned int, flags) 2325{ 2326 if (flags & MSG_CMSG_COMPAT) 2327 return -EINVAL; 2328 return __sys_recvmsg(fd, msg, flags); 2329} 2330 2331/* 2332 * Linux recvmmsg interface 2333 */ 2334 2335int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2336 unsigned int flags, struct timespec *timeout) 2337{ 2338 int fput_needed, err, datagrams; 2339 struct socket *sock; 2340 struct mmsghdr __user *entry; 2341 struct compat_mmsghdr __user *compat_entry; 2342 struct msghdr msg_sys; 2343 struct timespec end_time; 2344 2345 if (timeout && 2346 poll_select_set_timeout(&end_time, timeout->tv_sec, 2347 timeout->tv_nsec)) 2348 return -EINVAL; 2349 2350 datagrams = 0; 2351 2352 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2353 if (!sock) 2354 return err; 2355 2356 err = sock_error(sock->sk); 2357 if (err) 2358 goto out_put; 2359 2360 entry = mmsg; 2361 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2362 2363 while (datagrams < vlen) { 2364 /* 2365 * No need to ask LSM for more than the first datagram. 2366 */ 2367 if (MSG_CMSG_COMPAT & flags) { 2368 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry, 2369 &msg_sys, flags & ~MSG_WAITFORONE, 2370 datagrams); 2371 if (err < 0) 2372 break; 2373 err = __put_user(err, &compat_entry->msg_len); 2374 ++compat_entry; 2375 } else { 2376 err = ___sys_recvmsg(sock, 2377 (struct msghdr __user *)entry, 2378 &msg_sys, flags & ~MSG_WAITFORONE, 2379 datagrams); 2380 if (err < 0) 2381 break; 2382 err = put_user(err, &entry->msg_len); 2383 ++entry; 2384 } 2385 2386 if (err) 2387 break; 2388 ++datagrams; 2389 2390 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2391 if (flags & MSG_WAITFORONE) 2392 flags |= MSG_DONTWAIT; 2393 2394 if (timeout) { 2395 ktime_get_ts(timeout); 2396 *timeout = timespec_sub(end_time, *timeout); 2397 if (timeout->tv_sec < 0) { 2398 timeout->tv_sec = timeout->tv_nsec = 0; 2399 break; 2400 } 2401 2402 /* Timeout, return less than vlen datagrams */ 2403 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2404 break; 2405 } 2406 2407 /* Out of band data, return right away */ 2408 if (msg_sys.msg_flags & MSG_OOB) 2409 break; 2410 } 2411 2412out_put: 2413 fput_light(sock->file, fput_needed); 2414 2415 if (err == 0) 2416 return datagrams; 2417 2418 if (datagrams != 0) { 2419 /* 2420 * We may return less entries than requested (vlen) if the 2421 * sock is non block and there aren't enough datagrams... 2422 */ 2423 if (err != -EAGAIN) { 2424 /* 2425 * ... or if recvmsg returns an error after we 2426 * received some datagrams, where we record the 2427 * error to return on the next call or if the 2428 * app asks about it using getsockopt(SO_ERROR). 2429 */ 2430 sock->sk->sk_err = -err; 2431 } 2432 2433 return datagrams; 2434 } 2435 2436 return err; 2437} 2438 2439SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 2440 unsigned int, vlen, unsigned int, flags, 2441 struct timespec __user *, timeout) 2442{ 2443 int datagrams; 2444 struct timespec timeout_sys; 2445 2446 if (flags & MSG_CMSG_COMPAT) 2447 return -EINVAL; 2448 2449 if (!timeout) 2450 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL); 2451 2452 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys))) 2453 return -EFAULT; 2454 2455 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 2456 2457 if (datagrams > 0 && 2458 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys))) 2459 datagrams = -EFAULT; 2460 2461 return datagrams; 2462} 2463 2464#ifdef __ARCH_WANT_SYS_SOCKETCALL 2465/* Argument list sizes for sys_socketcall */ 2466#define AL(x) ((x) * sizeof(unsigned long)) 2467static const unsigned char nargs[21] = { 2468 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 2469 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 2470 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 2471 AL(4), AL(5), AL(4) 2472}; 2473 2474#undef AL 2475 2476/* 2477 * System call vectors. 2478 * 2479 * Argument checking cleaned up. Saved 20% in size. 2480 * This function doesn't need to set the kernel lock because 2481 * it is set by the callees. 2482 */ 2483 2484SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 2485{ 2486 unsigned long a[6]; 2487 unsigned long a0, a1; 2488 int err; 2489 unsigned int len; 2490 2491 if (call < 1 || call > SYS_SENDMMSG) 2492 return -EINVAL; 2493 2494 len = nargs[call]; 2495 if (len > sizeof(a)) 2496 return -EINVAL; 2497 2498 /* copy_from_user should be SMP safe. */ 2499 if (copy_from_user(a, args, len)) 2500 return -EFAULT; 2501 2502 audit_socketcall(nargs[call] / sizeof(unsigned long), a); 2503 2504 a0 = a[0]; 2505 a1 = a[1]; 2506 2507 switch (call) { 2508 case SYS_SOCKET: 2509 err = sys_socket(a0, a1, a[2]); 2510 break; 2511 case SYS_BIND: 2512 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 2513 break; 2514 case SYS_CONNECT: 2515 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 2516 break; 2517 case SYS_LISTEN: 2518 err = sys_listen(a0, a1); 2519 break; 2520 case SYS_ACCEPT: 2521 err = sys_accept4(a0, (struct sockaddr __user *)a1, 2522 (int __user *)a[2], 0); 2523 break; 2524 case SYS_GETSOCKNAME: 2525 err = 2526 sys_getsockname(a0, (struct sockaddr __user *)a1, 2527 (int __user *)a[2]); 2528 break; 2529 case SYS_GETPEERNAME: 2530 err = 2531 sys_getpeername(a0, (struct sockaddr __user *)a1, 2532 (int __user *)a[2]); 2533 break; 2534 case SYS_SOCKETPAIR: 2535 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 2536 break; 2537 case SYS_SEND: 2538 err = sys_send(a0, (void __user *)a1, a[2], a[3]); 2539 break; 2540 case SYS_SENDTO: 2541 err = sys_sendto(a0, (void __user *)a1, a[2], a[3], 2542 (struct sockaddr __user *)a[4], a[5]); 2543 break; 2544 case SYS_RECV: 2545 err = sys_recv(a0, (void __user *)a1, a[2], a[3]); 2546 break; 2547 case SYS_RECVFROM: 2548 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 2549 (struct sockaddr __user *)a[4], 2550 (int __user *)a[5]); 2551 break; 2552 case SYS_SHUTDOWN: 2553 err = sys_shutdown(a0, a1); 2554 break; 2555 case SYS_SETSOCKOPT: 2556 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]); 2557 break; 2558 case SYS_GETSOCKOPT: 2559 err = 2560 sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 2561 (int __user *)a[4]); 2562 break; 2563 case SYS_SENDMSG: 2564 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]); 2565 break; 2566 case SYS_SENDMMSG: 2567 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]); 2568 break; 2569 case SYS_RECVMSG: 2570 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]); 2571 break; 2572 case SYS_RECVMMSG: 2573 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3], 2574 (struct timespec __user *)a[4]); 2575 break; 2576 case SYS_ACCEPT4: 2577 err = sys_accept4(a0, (struct sockaddr __user *)a1, 2578 (int __user *)a[2], a[3]); 2579 break; 2580 default: 2581 err = -EINVAL; 2582 break; 2583 } 2584 return err; 2585} 2586 2587#endif /* __ARCH_WANT_SYS_SOCKETCALL */ 2588 2589/** 2590 * sock_register - add a socket protocol handler 2591 * @ops: description of protocol 2592 * 2593 * This function is called by a protocol handler that wants to 2594 * advertise its address family, and have it linked into the 2595 * socket interface. The value ops->family coresponds to the 2596 * socket system call protocol family. 2597 */ 2598int sock_register(const struct net_proto_family *ops) 2599{ 2600 int err; 2601 2602 if (ops->family >= NPROTO) { 2603 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, 2604 NPROTO); 2605 return -ENOBUFS; 2606 } 2607 2608 spin_lock(&net_family_lock); 2609 if (rcu_dereference_protected(net_families[ops->family], 2610 lockdep_is_held(&net_family_lock))) 2611 err = -EEXIST; 2612 else { 2613 rcu_assign_pointer(net_families[ops->family], ops); 2614 err = 0; 2615 } 2616 spin_unlock(&net_family_lock); 2617 2618 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family); 2619 return err; 2620} 2621EXPORT_SYMBOL(sock_register); 2622 2623/** 2624 * sock_unregister - remove a protocol handler 2625 * @family: protocol family to remove 2626 * 2627 * This function is called by a protocol handler that wants to 2628 * remove its address family, and have it unlinked from the 2629 * new socket creation. 2630 * 2631 * If protocol handler is a module, then it can use module reference 2632 * counts to protect against new references. If protocol handler is not 2633 * a module then it needs to provide its own protection in 2634 * the ops->create routine. 2635 */ 2636void sock_unregister(int family) 2637{ 2638 BUG_ON(family < 0 || family >= NPROTO); 2639 2640 spin_lock(&net_family_lock); 2641 RCU_INIT_POINTER(net_families[family], NULL); 2642 spin_unlock(&net_family_lock); 2643 2644 synchronize_rcu(); 2645 2646 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family); 2647} 2648EXPORT_SYMBOL(sock_unregister); 2649 2650static int __init sock_init(void) 2651{ 2652 int err; 2653 /* 2654 * Initialize the network sysctl infrastructure. 2655 */ 2656 err = net_sysctl_init(); 2657 if (err) 2658 goto out; 2659 2660 /* 2661 * Initialize skbuff SLAB cache 2662 */ 2663 skb_init(); 2664 2665 /* 2666 * Initialize the protocols module. 2667 */ 2668 2669 init_inodecache(); 2670 2671 err = register_filesystem(&sock_fs_type); 2672 if (err) 2673 goto out_fs; 2674 sock_mnt = kern_mount(&sock_fs_type); 2675 if (IS_ERR(sock_mnt)) { 2676 err = PTR_ERR(sock_mnt); 2677 goto out_mount; 2678 } 2679 2680 /* The real protocol initialization is performed in later initcalls. 2681 */ 2682 2683#ifdef CONFIG_NETFILTER 2684 netfilter_init(); 2685#endif 2686 2687#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING 2688 skb_timestamping_init(); 2689#endif 2690 2691out: 2692 return err; 2693 2694out_mount: 2695 unregister_filesystem(&sock_fs_type); 2696out_fs: 2697 goto out; 2698} 2699 2700core_initcall(sock_init); /* early initcall */ 2701 2702#ifdef CONFIG_PROC_FS 2703void socket_seq_show(struct seq_file *seq) 2704{ 2705 int cpu; 2706 int counter = 0; 2707 2708 for_each_possible_cpu(cpu) 2709 counter += per_cpu(sockets_in_use, cpu); 2710 2711 /* It can be negative, by the way. 8) */ 2712 if (counter < 0) 2713 counter = 0; 2714 2715 seq_printf(seq, "sockets: used %d\n", counter); 2716} 2717#endif /* CONFIG_PROC_FS */ 2718 2719#ifdef CONFIG_COMPAT 2720static int do_siocgstamp(struct net *net, struct socket *sock, 2721 unsigned int cmd, void __user *up) 2722{ 2723 mm_segment_t old_fs = get_fs(); 2724 struct timeval ktv; 2725 int err; 2726 2727 set_fs(KERNEL_DS); 2728 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv); 2729 set_fs(old_fs); 2730 if (!err) 2731 err = compat_put_timeval(&ktv, up); 2732 2733 return err; 2734} 2735 2736static int do_siocgstampns(struct net *net, struct socket *sock, 2737 unsigned int cmd, void __user *up) 2738{ 2739 mm_segment_t old_fs = get_fs(); 2740 struct timespec kts; 2741 int err; 2742 2743 set_fs(KERNEL_DS); 2744 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts); 2745 set_fs(old_fs); 2746 if (!err) 2747 err = compat_put_timespec(&kts, up); 2748 2749 return err; 2750} 2751 2752static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32) 2753{ 2754 struct ifreq __user *uifr; 2755 int err; 2756 2757 uifr = compat_alloc_user_space(sizeof(struct ifreq)); 2758 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) 2759 return -EFAULT; 2760 2761 err = dev_ioctl(net, SIOCGIFNAME, uifr); 2762 if (err) 2763 return err; 2764 2765 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq))) 2766 return -EFAULT; 2767 2768 return 0; 2769} 2770 2771static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32) 2772{ 2773 struct compat_ifconf ifc32; 2774 struct ifconf ifc; 2775 struct ifconf __user *uifc; 2776 struct compat_ifreq __user *ifr32; 2777 struct ifreq __user *ifr; 2778 unsigned int i, j; 2779 int err; 2780 2781 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf))) 2782 return -EFAULT; 2783 2784 memset(&ifc, 0, sizeof(ifc)); 2785 if (ifc32.ifcbuf == 0) { 2786 ifc32.ifc_len = 0; 2787 ifc.ifc_len = 0; 2788 ifc.ifc_req = NULL; 2789 uifc = compat_alloc_user_space(sizeof(struct ifconf)); 2790 } else { 2791 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) * 2792 sizeof(struct ifreq); 2793 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len); 2794 ifc.ifc_len = len; 2795 ifr = ifc.ifc_req = (void __user *)(uifc + 1); 2796 ifr32 = compat_ptr(ifc32.ifcbuf); 2797 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) { 2798 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq))) 2799 return -EFAULT; 2800 ifr++; 2801 ifr32++; 2802 } 2803 } 2804 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf))) 2805 return -EFAULT; 2806 2807 err = dev_ioctl(net, SIOCGIFCONF, uifc); 2808 if (err) 2809 return err; 2810 2811 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf))) 2812 return -EFAULT; 2813 2814 ifr = ifc.ifc_req; 2815 ifr32 = compat_ptr(ifc32.ifcbuf); 2816 for (i = 0, j = 0; 2817 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len; 2818 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) { 2819 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq))) 2820 return -EFAULT; 2821 ifr32++; 2822 ifr++; 2823 } 2824 2825 if (ifc32.ifcbuf == 0) { 2826 /* Translate from 64-bit structure multiple to 2827 * a 32-bit one. 2828 */ 2829 i = ifc.ifc_len; 2830 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq)); 2831 ifc32.ifc_len = i; 2832 } else { 2833 ifc32.ifc_len = i; 2834 } 2835 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf))) 2836 return -EFAULT; 2837 2838 return 0; 2839} 2840 2841static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32) 2842{ 2843 struct compat_ethtool_rxnfc __user *compat_rxnfc; 2844 bool convert_in = false, convert_out = false; 2845 size_t buf_size = ALIGN(sizeof(struct ifreq), 8); 2846 struct ethtool_rxnfc __user *rxnfc; 2847 struct ifreq __user *ifr; 2848 u32 rule_cnt = 0, actual_rule_cnt; 2849 u32 ethcmd; 2850 u32 data; 2851 int ret; 2852 2853 if (get_user(data, &ifr32->ifr_ifru.ifru_data)) 2854 return -EFAULT; 2855 2856 compat_rxnfc = compat_ptr(data); 2857 2858 if (get_user(ethcmd, &compat_rxnfc->cmd)) 2859 return -EFAULT; 2860 2861 /* Most ethtool structures are defined without padding. 2862 * Unfortunately struct ethtool_rxnfc is an exception. 2863 */ 2864 switch (ethcmd) { 2865 default: 2866 break; 2867 case ETHTOOL_GRXCLSRLALL: 2868 /* Buffer size is variable */ 2869 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt)) 2870 return -EFAULT; 2871 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32)) 2872 return -ENOMEM; 2873 buf_size += rule_cnt * sizeof(u32); 2874 /* fall through */ 2875 case ETHTOOL_GRXRINGS: 2876 case ETHTOOL_GRXCLSRLCNT: 2877 case ETHTOOL_GRXCLSRULE: 2878 case ETHTOOL_SRXCLSRLINS: 2879 convert_out = true; 2880 /* fall through */ 2881 case ETHTOOL_SRXCLSRLDEL: 2882 buf_size += sizeof(struct ethtool_rxnfc); 2883 convert_in = true; 2884 break; 2885 } 2886 2887 ifr = compat_alloc_user_space(buf_size); 2888 rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8); 2889 2890 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ)) 2891 return -EFAULT; 2892 2893 if (put_user(convert_in ? rxnfc : compat_ptr(data), 2894 &ifr->ifr_ifru.ifru_data)) 2895 return -EFAULT; 2896 2897 if (convert_in) { 2898 /* We expect there to be holes between fs.m_ext and 2899 * fs.ring_cookie and at the end of fs, but nowhere else. 2900 */ 2901 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) + 2902 sizeof(compat_rxnfc->fs.m_ext) != 2903 offsetof(struct ethtool_rxnfc, fs.m_ext) + 2904 sizeof(rxnfc->fs.m_ext)); 2905 BUILD_BUG_ON( 2906 offsetof(struct compat_ethtool_rxnfc, fs.location) - 2907 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) != 2908 offsetof(struct ethtool_rxnfc, fs.location) - 2909 offsetof(struct ethtool_rxnfc, fs.ring_cookie)); 2910 2911 if (copy_in_user(rxnfc, compat_rxnfc, 2912 (void *)(&rxnfc->fs.m_ext + 1) - 2913 (void *)rxnfc) || 2914 copy_in_user(&rxnfc->fs.ring_cookie, 2915 &compat_rxnfc->fs.ring_cookie, 2916 (void *)(&rxnfc->fs.location + 1) - 2917 (void *)&rxnfc->fs.ring_cookie) || 2918 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt, 2919 sizeof(rxnfc->rule_cnt))) 2920 return -EFAULT; 2921 } 2922 2923 ret = dev_ioctl(net, SIOCETHTOOL, ifr); 2924 if (ret) 2925 return ret; 2926 2927 if (convert_out) { 2928 if (copy_in_user(compat_rxnfc, rxnfc, 2929 (const void *)(&rxnfc->fs.m_ext + 1) - 2930 (const void *)rxnfc) || 2931 copy_in_user(&compat_rxnfc->fs.ring_cookie, 2932 &rxnfc->fs.ring_cookie, 2933 (const void *)(&rxnfc->fs.location + 1) - 2934 (const void *)&rxnfc->fs.ring_cookie) || 2935 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt, 2936 sizeof(rxnfc->rule_cnt))) 2937 return -EFAULT; 2938 2939 if (ethcmd == ETHTOOL_GRXCLSRLALL) { 2940 /* As an optimisation, we only copy the actual 2941 * number of rules that the underlying 2942 * function returned. Since Mallory might 2943 * change the rule count in user memory, we 2944 * check that it is less than the rule count 2945 * originally given (as the user buffer size), 2946 * which has been range-checked. 2947 */ 2948 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt)) 2949 return -EFAULT; 2950 if (actual_rule_cnt < rule_cnt) 2951 rule_cnt = actual_rule_cnt; 2952 if (copy_in_user(&compat_rxnfc->rule_locs[0], 2953 &rxnfc->rule_locs[0], 2954 rule_cnt * sizeof(u32))) 2955 return -EFAULT; 2956 } 2957 } 2958 2959 return 0; 2960} 2961 2962static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 2963{ 2964 void __user *uptr; 2965 compat_uptr_t uptr32; 2966 struct ifreq __user *uifr; 2967 2968 uifr = compat_alloc_user_space(sizeof(*uifr)); 2969 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) 2970 return -EFAULT; 2971 2972 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 2973 return -EFAULT; 2974 2975 uptr = compat_ptr(uptr32); 2976 2977 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc)) 2978 return -EFAULT; 2979 2980 return dev_ioctl(net, SIOCWANDEV, uifr); 2981} 2982 2983static int bond_ioctl(struct net *net, unsigned int cmd, 2984 struct compat_ifreq __user *ifr32) 2985{ 2986 struct ifreq kifr; 2987 struct ifreq __user *uifr; 2988 mm_segment_t old_fs; 2989 int err; 2990 u32 data; 2991 void __user *datap; 2992 2993 switch (cmd) { 2994 case SIOCBONDENSLAVE: 2995 case SIOCBONDRELEASE: 2996 case SIOCBONDSETHWADDR: 2997 case SIOCBONDCHANGEACTIVE: 2998 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq))) 2999 return -EFAULT; 3000 3001 old_fs = get_fs(); 3002 set_fs(KERNEL_DS); 3003 err = dev_ioctl(net, cmd, 3004 (struct ifreq __user __force *) &kifr); 3005 set_fs(old_fs); 3006 3007 return err; 3008 case SIOCBONDSLAVEINFOQUERY: 3009 case SIOCBONDINFOQUERY: 3010 uifr = compat_alloc_user_space(sizeof(*uifr)); 3011 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ)) 3012 return -EFAULT; 3013 3014 if (get_user(data, &ifr32->ifr_ifru.ifru_data)) 3015 return -EFAULT; 3016 3017 datap = compat_ptr(data); 3018 if (put_user(datap, &uifr->ifr_ifru.ifru_data)) 3019 return -EFAULT; 3020 3021 return dev_ioctl(net, cmd, uifr); 3022 default: 3023 return -ENOIOCTLCMD; 3024 } 3025} 3026 3027static int siocdevprivate_ioctl(struct net *net, unsigned int cmd, 3028 struct compat_ifreq __user *u_ifreq32) 3029{ 3030 struct ifreq __user *u_ifreq64; 3031 char tmp_buf[IFNAMSIZ]; 3032 void __user *data64; 3033 u32 data32; 3034 3035 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]), 3036 IFNAMSIZ)) 3037 return -EFAULT; 3038 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data)) 3039 return -EFAULT; 3040 data64 = compat_ptr(data32); 3041 3042 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64)); 3043 3044 /* Don't check these user accesses, just let that get trapped 3045 * in the ioctl handler instead. 3046 */ 3047 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0], 3048 IFNAMSIZ)) 3049 return -EFAULT; 3050 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data)) 3051 return -EFAULT; 3052 3053 return dev_ioctl(net, cmd, u_ifreq64); 3054} 3055 3056static int dev_ifsioc(struct net *net, struct socket *sock, 3057 unsigned int cmd, struct compat_ifreq __user *uifr32) 3058{ 3059 struct ifreq __user *uifr; 3060 int err; 3061 3062 uifr = compat_alloc_user_space(sizeof(*uifr)); 3063 if (copy_in_user(uifr, uifr32, sizeof(*uifr32))) 3064 return -EFAULT; 3065 3066 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr); 3067 3068 if (!err) { 3069 switch (cmd) { 3070 case SIOCGIFFLAGS: 3071 case SIOCGIFMETRIC: 3072 case SIOCGIFMTU: 3073 case SIOCGIFMEM: 3074 case SIOCGIFHWADDR: 3075 case SIOCGIFINDEX: 3076 case SIOCGIFADDR: 3077 case SIOCGIFBRDADDR: 3078 case SIOCGIFDSTADDR: 3079 case SIOCGIFNETMASK: 3080 case SIOCGIFPFLAGS: 3081 case SIOCGIFTXQLEN: 3082 case SIOCGMIIPHY: 3083 case SIOCGMIIREG: 3084 if (copy_in_user(uifr32, uifr, sizeof(*uifr32))) 3085 err = -EFAULT; 3086 break; 3087 } 3088 } 3089 return err; 3090} 3091 3092static int compat_sioc_ifmap(struct net *net, unsigned int cmd, 3093 struct compat_ifreq __user *uifr32) 3094{ 3095 struct ifreq ifr; 3096 struct compat_ifmap __user *uifmap32; 3097 mm_segment_t old_fs; 3098 int err; 3099 3100 uifmap32 = &uifr32->ifr_ifru.ifru_map; 3101 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name)); 3102 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 3103 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 3104 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 3105 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq); 3106 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma); 3107 err |= __get_user(ifr.ifr_map.port, &uifmap32->port); 3108 if (err) 3109 return -EFAULT; 3110 3111 old_fs = get_fs(); 3112 set_fs(KERNEL_DS); 3113 err = dev_ioctl(net, cmd, (void __user __force *)&ifr); 3114 set_fs(old_fs); 3115 3116 if (cmd == SIOCGIFMAP && !err) { 3117 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name)); 3118 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start); 3119 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end); 3120 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr); 3121 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq); 3122 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma); 3123 err |= __put_user(ifr.ifr_map.port, &uifmap32->port); 3124 if (err) 3125 err = -EFAULT; 3126 } 3127 return err; 3128} 3129 3130static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32) 3131{ 3132 void __user *uptr; 3133 compat_uptr_t uptr32; 3134 struct ifreq __user *uifr; 3135 3136 uifr = compat_alloc_user_space(sizeof(*uifr)); 3137 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq))) 3138 return -EFAULT; 3139 3140 if (get_user(uptr32, &uifr32->ifr_data)) 3141 return -EFAULT; 3142 3143 uptr = compat_ptr(uptr32); 3144 3145 if (put_user(uptr, &uifr->ifr_data)) 3146 return -EFAULT; 3147 3148 return dev_ioctl(net, SIOCSHWTSTAMP, uifr); 3149} 3150 3151struct rtentry32 { 3152 u32 rt_pad1; 3153 struct sockaddr rt_dst; /* target address */ 3154 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */ 3155 struct sockaddr rt_genmask; /* target network mask (IP) */ 3156 unsigned short rt_flags; 3157 short rt_pad2; 3158 u32 rt_pad3; 3159 unsigned char rt_tos; 3160 unsigned char rt_class; 3161 short rt_pad4; 3162 short rt_metric; /* +1 for binary compatibility! */ 3163 /* char * */ u32 rt_dev; /* forcing the device at add */ 3164 u32 rt_mtu; /* per route MTU/Window */ 3165 u32 rt_window; /* Window clamping */ 3166 unsigned short rt_irtt; /* Initial RTT */ 3167}; 3168 3169struct in6_rtmsg32 { 3170 struct in6_addr rtmsg_dst; 3171 struct in6_addr rtmsg_src; 3172 struct in6_addr rtmsg_gateway; 3173 u32 rtmsg_type; 3174 u16 rtmsg_dst_len; 3175 u16 rtmsg_src_len; 3176 u32 rtmsg_metric; 3177 u32 rtmsg_info; 3178 u32 rtmsg_flags; 3179 s32 rtmsg_ifindex; 3180}; 3181 3182static int routing_ioctl(struct net *net, struct socket *sock, 3183 unsigned int cmd, void __user *argp) 3184{ 3185 int ret; 3186 void *r = NULL; 3187 struct in6_rtmsg r6; 3188 struct rtentry r4; 3189 char devname[16]; 3190 u32 rtdev; 3191 mm_segment_t old_fs = get_fs(); 3192 3193 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */ 3194 struct in6_rtmsg32 __user *ur6 = argp; 3195 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst), 3196 3 * sizeof(struct in6_addr)); 3197 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type)); 3198 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len)); 3199 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len)); 3200 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric)); 3201 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info)); 3202 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags)); 3203 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex)); 3204 3205 r = (void *) &r6; 3206 } else { /* ipv4 */ 3207 struct rtentry32 __user *ur4 = argp; 3208 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst), 3209 3 * sizeof(struct sockaddr)); 3210 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags)); 3211 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric)); 3212 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu)); 3213 ret |= __get_user(r4.rt_window, &(ur4->rt_window)); 3214 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt)); 3215 ret |= __get_user(rtdev, &(ur4->rt_dev)); 3216 if (rtdev) { 3217 ret |= copy_from_user(devname, compat_ptr(rtdev), 15); 3218 r4.rt_dev = (char __user __force *)devname; 3219 devname[15] = 0; 3220 } else 3221 r4.rt_dev = NULL; 3222 3223 r = (void *) &r4; 3224 } 3225 3226 if (ret) { 3227 ret = -EFAULT; 3228 goto out; 3229 } 3230 3231 set_fs(KERNEL_DS); 3232 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r); 3233 set_fs(old_fs); 3234 3235out: 3236 return ret; 3237} 3238 3239/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE 3240 * for some operations; this forces use of the newer bridge-utils that 3241 * use compatible ioctls 3242 */ 3243static int old_bridge_ioctl(compat_ulong_t __user *argp) 3244{ 3245 compat_ulong_t tmp; 3246 3247 if (get_user(tmp, argp)) 3248 return -EFAULT; 3249 if (tmp == BRCTL_GET_VERSION) 3250 return BRCTL_VERSION + 1; 3251 return -EINVAL; 3252} 3253 3254static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 3255 unsigned int cmd, unsigned long arg) 3256{ 3257 void __user *argp = compat_ptr(arg); 3258 struct sock *sk = sock->sk; 3259 struct net *net = sock_net(sk); 3260 3261 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 3262 return siocdevprivate_ioctl(net, cmd, argp); 3263 3264 switch (cmd) { 3265 case SIOCSIFBR: 3266 case SIOCGIFBR: 3267 return old_bridge_ioctl(argp); 3268 case SIOCGIFNAME: 3269 return dev_ifname32(net, argp); 3270 case SIOCGIFCONF: 3271 return dev_ifconf(net, argp); 3272 case SIOCETHTOOL: 3273 return ethtool_ioctl(net, argp); 3274 case SIOCWANDEV: 3275 return compat_siocwandev(net, argp); 3276 case SIOCGIFMAP: 3277 case SIOCSIFMAP: 3278 return compat_sioc_ifmap(net, cmd, argp); 3279 case SIOCBONDENSLAVE: 3280 case SIOCBONDRELEASE: 3281 case SIOCBONDSETHWADDR: 3282 case SIOCBONDSLAVEINFOQUERY: 3283 case SIOCBONDINFOQUERY: 3284 case SIOCBONDCHANGEACTIVE: 3285 return bond_ioctl(net, cmd, argp); 3286 case SIOCADDRT: 3287 case SIOCDELRT: 3288 return routing_ioctl(net, sock, cmd, argp); 3289 case SIOCGSTAMP: 3290 return do_siocgstamp(net, sock, cmd, argp); 3291 case SIOCGSTAMPNS: 3292 return do_siocgstampns(net, sock, cmd, argp); 3293 case SIOCSHWTSTAMP: 3294 return compat_siocshwtstamp(net, argp); 3295 3296 case FIOSETOWN: 3297 case SIOCSPGRP: 3298 case FIOGETOWN: 3299 case SIOCGPGRP: 3300 case SIOCBRADDBR: 3301 case SIOCBRDELBR: 3302 case SIOCGIFVLAN: 3303 case SIOCSIFVLAN: 3304 case SIOCADDDLCI: 3305 case SIOCDELDLCI: 3306 return sock_ioctl(file, cmd, arg); 3307 3308 case SIOCGIFFLAGS: 3309 case SIOCSIFFLAGS: 3310 case SIOCGIFMETRIC: 3311 case SIOCSIFMETRIC: 3312 case SIOCGIFMTU: 3313 case SIOCSIFMTU: 3314 case SIOCGIFMEM: 3315 case SIOCSIFMEM: 3316 case SIOCGIFHWADDR: 3317 case SIOCSIFHWADDR: 3318 case SIOCADDMULTI: 3319 case SIOCDELMULTI: 3320 case SIOCGIFINDEX: 3321 case SIOCGIFADDR: 3322 case SIOCSIFADDR: 3323 case SIOCSIFHWBROADCAST: 3324 case SIOCDIFADDR: 3325 case SIOCGIFBRDADDR: 3326 case SIOCSIFBRDADDR: 3327 case SIOCGIFDSTADDR: 3328 case SIOCSIFDSTADDR: 3329 case SIOCGIFNETMASK: 3330 case SIOCSIFNETMASK: 3331 case SIOCSIFPFLAGS: 3332 case SIOCGIFPFLAGS: 3333 case SIOCGIFTXQLEN: 3334 case SIOCSIFTXQLEN: 3335 case SIOCBRADDIF: 3336 case SIOCBRDELIF: 3337 case SIOCSIFNAME: 3338 case SIOCGMIIPHY: 3339 case SIOCGMIIREG: 3340 case SIOCSMIIREG: 3341 return dev_ifsioc(net, sock, cmd, argp); 3342 3343 case SIOCSARP: 3344 case SIOCGARP: 3345 case SIOCDARP: 3346 case SIOCATMARK: 3347 return sock_do_ioctl(net, sock, cmd, arg); 3348 } 3349 3350 return -ENOIOCTLCMD; 3351} 3352 3353static long compat_sock_ioctl(struct file *file, unsigned int cmd, 3354 unsigned long arg) 3355{ 3356 struct socket *sock = file->private_data; 3357 int ret = -ENOIOCTLCMD; 3358 struct sock *sk; 3359 struct net *net; 3360 3361 sk = sock->sk; 3362 net = sock_net(sk); 3363 3364 if (sock->ops->compat_ioctl) 3365 ret = sock->ops->compat_ioctl(sock, cmd, arg); 3366 3367 if (ret == -ENOIOCTLCMD && 3368 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 3369 ret = compat_wext_handle_ioctl(net, cmd, arg); 3370 3371 if (ret == -ENOIOCTLCMD) 3372 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 3373 3374 return ret; 3375} 3376#endif 3377 3378int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3379{ 3380 return sock->ops->bind(sock, addr, addrlen); 3381} 3382EXPORT_SYMBOL(kernel_bind); 3383 3384int kernel_listen(struct socket *sock, int backlog) 3385{ 3386 return sock->ops->listen(sock, backlog); 3387} 3388EXPORT_SYMBOL(kernel_listen); 3389 3390int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3391{ 3392 struct sock *sk = sock->sk; 3393 int err; 3394 3395 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3396 newsock); 3397 if (err < 0) 3398 goto done; 3399 3400 err = sock->ops->accept(sock, *newsock, flags); 3401 if (err < 0) { 3402 sock_release(*newsock); 3403 *newsock = NULL; 3404 goto done; 3405 } 3406 3407 (*newsock)->ops = sock->ops; 3408 __module_get((*newsock)->ops->owner); 3409 3410done: 3411 return err; 3412} 3413EXPORT_SYMBOL(kernel_accept); 3414 3415int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3416 int flags) 3417{ 3418 return sock->ops->connect(sock, addr, addrlen, flags); 3419} 3420EXPORT_SYMBOL(kernel_connect); 3421 3422int kernel_getsockname(struct socket *sock, struct sockaddr *addr, 3423 int *addrlen) 3424{ 3425 return sock->ops->getname(sock, addr, addrlen, 0); 3426} 3427EXPORT_SYMBOL(kernel_getsockname); 3428 3429int kernel_getpeername(struct socket *sock, struct sockaddr *addr, 3430 int *addrlen) 3431{ 3432 return sock->ops->getname(sock, addr, addrlen, 1); 3433} 3434EXPORT_SYMBOL(kernel_getpeername); 3435 3436int kernel_getsockopt(struct socket *sock, int level, int optname, 3437 char *optval, int *optlen) 3438{ 3439 mm_segment_t oldfs = get_fs(); 3440 char __user *uoptval; 3441 int __user *uoptlen; 3442 int err; 3443 3444 uoptval = (char __user __force *) optval; 3445 uoptlen = (int __user __force *) optlen; 3446 3447 set_fs(KERNEL_DS); 3448 if (level == SOL_SOCKET) 3449 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen); 3450 else 3451 err = sock->ops->getsockopt(sock, level, optname, uoptval, 3452 uoptlen); 3453 set_fs(oldfs); 3454 return err; 3455} 3456EXPORT_SYMBOL(kernel_getsockopt); 3457 3458int kernel_setsockopt(struct socket *sock, int level, int optname, 3459 char *optval, unsigned int optlen) 3460{ 3461 mm_segment_t oldfs = get_fs(); 3462 char __user *uoptval; 3463 int err; 3464 3465 uoptval = (char __user __force *) optval; 3466 3467 set_fs(KERNEL_DS); 3468 if (level == SOL_SOCKET) 3469 err = sock_setsockopt(sock, level, optname, uoptval, optlen); 3470 else 3471 err = sock->ops->setsockopt(sock, level, optname, uoptval, 3472 optlen); 3473 set_fs(oldfs); 3474 return err; 3475} 3476EXPORT_SYMBOL(kernel_setsockopt); 3477 3478int kernel_sendpage(struct socket *sock, struct page *page, int offset, 3479 size_t size, int flags) 3480{ 3481 if (sock->ops->sendpage) 3482 return sock->ops->sendpage(sock, page, offset, size, flags); 3483 3484 return sock_no_sendpage(sock, page, offset, size, flags); 3485} 3486EXPORT_SYMBOL(kernel_sendpage); 3487 3488int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg) 3489{ 3490 mm_segment_t oldfs = get_fs(); 3491 int err; 3492 3493 set_fs(KERNEL_DS); 3494 err = sock->ops->ioctl(sock, cmd, arg); 3495 set_fs(oldfs); 3496 3497 return err; 3498} 3499EXPORT_SYMBOL(kernel_sock_ioctl); 3500 3501int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3502{ 3503 return sock->ops->shutdown(sock, how); 3504} 3505EXPORT_SYMBOL(kernel_sock_shutdown);