PageRenderTime 14ms CodeModel.GetById 3ms app.highlight 143ms RepoModel.GetById 3ms app.codeStats 0ms

/net/socket.c

https://bitbucket.org/codefirex/kernel_lge_kamakogeebang
C | 3384 lines | 2425 code | 495 blank | 464 comment | 381 complexity | 0e0e00142ecff79c83e1293e19535b2f 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
  92#include <asm/uaccess.h>
  93#include <asm/unistd.h>
  94
  95#include <net/compat.h>
  96#include <net/wext.h>
  97#include <net/cls_cgroup.h>
  98
  99#include <net/sock.h>
 100#include <linux/netfilter.h>
 101
 102#include <linux/if_tun.h>
 103#include <linux/ipv6_route.h>
 104#include <linux/route.h>
 105#include <linux/sockios.h>
 106#include <linux/atalk.h>
 107
 108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
 109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 110			 unsigned long nr_segs, loff_t pos);
 111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 112			  unsigned long nr_segs, loff_t pos);
 113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
 114
 115static int sock_close(struct inode *inode, struct file *file);
 116static unsigned int sock_poll(struct file *file,
 117			      struct poll_table_struct *wait);
 118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 119#ifdef CONFIG_COMPAT
 120static long compat_sock_ioctl(struct file *file,
 121			      unsigned int cmd, unsigned long arg);
 122#endif
 123static int sock_fasync(int fd, struct file *filp, int on);
 124static ssize_t sock_sendpage(struct file *file, struct page *page,
 125			     int offset, size_t size, loff_t *ppos, int more);
 126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 127				struct pipe_inode_info *pipe, size_t len,
 128				unsigned int flags);
 129
 130/*
 131 *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
 132 *	in the operation structures but are done directly via the socketcall() multiplexor.
 133 */
 134
 135static const struct file_operations socket_file_ops = {
 136	.owner =	THIS_MODULE,
 137	.llseek =	no_llseek,
 138	.aio_read =	sock_aio_read,
 139	.aio_write =	sock_aio_write,
 140	.poll =		sock_poll,
 141	.unlocked_ioctl = sock_ioctl,
 142#ifdef CONFIG_COMPAT
 143	.compat_ioctl = compat_sock_ioctl,
 144#endif
 145	.mmap =		sock_mmap,
 146	.open =		sock_no_open,	/* special open code to disallow open via /proc */
 147	.release =	sock_close,
 148	.fasync =	sock_fasync,
 149	.sendpage =	sock_sendpage,
 150	.splice_write = generic_splice_sendpage,
 151	.splice_read =	sock_splice_read,
 152};
 153
 154/*
 155 *	The protocol list. Each protocol is registered in here.
 156 */
 157
 158static DEFINE_SPINLOCK(net_family_lock);
 159static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
 160
 161/*
 162 *	Statistics counters of the socket lists
 163 */
 164
 165static DEFINE_PER_CPU(int, sockets_in_use);
 166
 167/*
 168 * Support routines.
 169 * Move socket addresses back and forth across the kernel/user
 170 * divide and look after the messy bits.
 171 */
 172
 173/**
 174 *	move_addr_to_kernel	-	copy a socket address into kernel space
 175 *	@uaddr: Address in user space
 176 *	@kaddr: Address in kernel space
 177 *	@ulen: Length in user space
 178 *
 179 *	The address is copied into kernel space. If the provided address is
 180 *	too long an error code of -EINVAL is returned. If the copy gives
 181 *	invalid addresses -EFAULT is returned. On a success 0 is returned.
 182 */
 183
 184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
 185{
 186	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
 187		return -EINVAL;
 188	if (ulen == 0)
 189		return 0;
 190	if (copy_from_user(kaddr, uaddr, ulen))
 191		return -EFAULT;
 192	return audit_sockaddr(ulen, kaddr);
 193}
 194
 195/**
 196 *	move_addr_to_user	-	copy an address to user space
 197 *	@kaddr: kernel space address
 198 *	@klen: length of address in kernel
 199 *	@uaddr: user space address
 200 *	@ulen: pointer to user length field
 201 *
 202 *	The value pointed to by ulen on entry is the buffer length available.
 203 *	This is overwritten with the buffer space used. -EINVAL is returned
 204 *	if an overlong buffer is specified or a negative buffer size. -EFAULT
 205 *	is returned if either the buffer or the length field are not
 206 *	accessible.
 207 *	After copying the data up to the limit the user specifies, the true
 208 *	length of the data is written over the length limit the user
 209 *	specified. Zero is returned for a success.
 210 */
 211
 212static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
 213			     void __user *uaddr, int __user *ulen)
 214{
 215	int err;
 216	int len;
 217
 218	err = get_user(len, ulen);
 219	if (err)
 220		return err;
 221	if (len > klen)
 222		len = klen;
 223	if (len < 0 || len > sizeof(struct sockaddr_storage))
 224		return -EINVAL;
 225	if (len) {
 226		if (audit_sockaddr(klen, kaddr))
 227			return -ENOMEM;
 228		if (copy_to_user(uaddr, kaddr, len))
 229			return -EFAULT;
 230	}
 231	/*
 232	 *      "fromlen shall refer to the value before truncation.."
 233	 *                      1003.1g
 234	 */
 235	return __put_user(klen, ulen);
 236}
 237
 238static struct kmem_cache *sock_inode_cachep __read_mostly;
 239
 240static struct inode *sock_alloc_inode(struct super_block *sb)
 241{
 242	struct socket_alloc *ei;
 243	struct socket_wq *wq;
 244
 245	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 246	if (!ei)
 247		return NULL;
 248	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
 249	if (!wq) {
 250		kmem_cache_free(sock_inode_cachep, ei);
 251		return NULL;
 252	}
 253	init_waitqueue_head(&wq->wait);
 254	wq->fasync_list = NULL;
 255	RCU_INIT_POINTER(ei->socket.wq, wq);
 256
 257	ei->socket.state = SS_UNCONNECTED;
 258	ei->socket.flags = 0;
 259	ei->socket.ops = NULL;
 260	ei->socket.sk = NULL;
 261	ei->socket.file = NULL;
 262
 263	return &ei->vfs_inode;
 264}
 265
 266static void sock_destroy_inode(struct inode *inode)
 267{
 268	struct socket_alloc *ei;
 269	struct socket_wq *wq;
 270
 271	ei = container_of(inode, struct socket_alloc, vfs_inode);
 272	wq = rcu_dereference_protected(ei->socket.wq, 1);
 273	kfree_rcu(wq, rcu);
 274	kmem_cache_free(sock_inode_cachep, ei);
 275}
 276
 277static void init_once(void *foo)
 278{
 279	struct socket_alloc *ei = (struct socket_alloc *)foo;
 280
 281	inode_init_once(&ei->vfs_inode);
 282}
 283
 284static int init_inodecache(void)
 285{
 286	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 287					      sizeof(struct socket_alloc),
 288					      0,
 289					      (SLAB_HWCACHE_ALIGN |
 290					       SLAB_RECLAIM_ACCOUNT |
 291					       SLAB_MEM_SPREAD),
 292					      init_once);
 293	if (sock_inode_cachep == NULL)
 294		return -ENOMEM;
 295	return 0;
 296}
 297
 298static const struct super_operations sockfs_ops = {
 299	.alloc_inode	= sock_alloc_inode,
 300	.destroy_inode	= sock_destroy_inode,
 301	.statfs		= simple_statfs,
 302};
 303
 304/*
 305 * sockfs_dname() is called from d_path().
 306 */
 307static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 308{
 309	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 310				dentry->d_inode->i_ino);
 311}
 312
 313static const struct dentry_operations sockfs_dentry_operations = {
 314	.d_dname  = sockfs_dname,
 315};
 316
 317static struct dentry *sockfs_mount(struct file_system_type *fs_type,
 318			 int flags, const char *dev_name, void *data)
 319{
 320	return mount_pseudo(fs_type, "socket:", &sockfs_ops,
 321		&sockfs_dentry_operations, SOCKFS_MAGIC);
 322}
 323
 324static struct vfsmount *sock_mnt __read_mostly;
 325
 326static struct file_system_type sock_fs_type = {
 327	.name =		"sockfs",
 328	.mount =	sockfs_mount,
 329	.kill_sb =	kill_anon_super,
 330};
 331
 332/*
 333 *	Obtains the first available file descriptor and sets it up for use.
 334 *
 335 *	These functions create file structures and maps them to fd space
 336 *	of the current process. On success it returns file descriptor
 337 *	and file struct implicitly stored in sock->file.
 338 *	Note that another thread may close file descriptor before we return
 339 *	from this function. We use the fact that now we do not refer
 340 *	to socket after mapping. If one day we will need it, this
 341 *	function will increment ref. count on file by 1.
 342 *
 343 *	In any case returned fd MAY BE not valid!
 344 *	This race condition is unavoidable
 345 *	with shared fd spaces, we cannot solve it inside kernel,
 346 *	but we take care of internal coherence yet.
 347 */
 348
 349static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
 350{
 351	struct qstr name = { .name = "" };
 352	struct path path;
 353	struct file *file;
 354	int fd;
 355
 356	fd = get_unused_fd_flags(flags);
 357	if (unlikely(fd < 0))
 358		return fd;
 359
 360	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
 361	if (unlikely(!path.dentry)) {
 362		put_unused_fd(fd);
 363		return -ENOMEM;
 364	}
 365	path.mnt = mntget(sock_mnt);
 366
 367	d_instantiate(path.dentry, SOCK_INODE(sock));
 368	SOCK_INODE(sock)->i_fop = &socket_file_ops;
 369
 370	file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 371		  &socket_file_ops);
 372	if (unlikely(!file)) {
 373		/* drop dentry, keep inode */
 374		ihold(path.dentry->d_inode);
 375		path_put(&path);
 376		put_unused_fd(fd);
 377		return -ENFILE;
 378	}
 379
 380	sock->file = file;
 381	file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 382	file->f_pos = 0;
 383	file->private_data = sock;
 384
 385	*f = file;
 386	return fd;
 387}
 388
 389int sock_map_fd(struct socket *sock, int flags)
 390{
 391	struct file *newfile;
 392	int fd = sock_alloc_file(sock, &newfile, flags);
 393
 394	if (likely(fd >= 0))
 395		fd_install(fd, newfile);
 396
 397	return fd;
 398}
 399EXPORT_SYMBOL(sock_map_fd);
 400
 401static struct socket *sock_from_file(struct file *file, int *err)
 402{
 403	if (file->f_op == &socket_file_ops)
 404		return file->private_data;	/* set in sock_map_fd */
 405
 406	*err = -ENOTSOCK;
 407	return NULL;
 408}
 409
 410/**
 411 *	sockfd_lookup - Go from a file number to its socket slot
 412 *	@fd: file handle
 413 *	@err: pointer to an error code return
 414 *
 415 *	The file handle passed in is locked and the socket it is bound
 416 *	too is returned. If an error occurs the err pointer is overwritten
 417 *	with a negative errno code and NULL is returned. The function checks
 418 *	for both invalid handles and passing a handle which is not a socket.
 419 *
 420 *	On a success the socket object pointer is returned.
 421 */
 422
 423struct socket *sockfd_lookup(int fd, int *err)
 424{
 425	struct file *file;
 426	struct socket *sock;
 427
 428	file = fget(fd);
 429	if (!file) {
 430		*err = -EBADF;
 431		return NULL;
 432	}
 433
 434	sock = sock_from_file(file, err);
 435	if (!sock)
 436		fput(file);
 437	return sock;
 438}
 439EXPORT_SYMBOL(sockfd_lookup);
 440
 441static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 442{
 443	struct file *file;
 444	struct socket *sock;
 445
 446	*err = -EBADF;
 447	file = fget_light(fd, fput_needed);
 448	if (file) {
 449		sock = sock_from_file(file, err);
 450		if (sock)
 451			return sock;
 452		fput_light(file, *fput_needed);
 453	}
 454	return NULL;
 455}
 456
 457/**
 458 *	sock_alloc	-	allocate a socket
 459 *
 460 *	Allocate a new inode and socket object. The two are bound together
 461 *	and initialised. The socket is then returned. If we are out of inodes
 462 *	NULL is returned.
 463 */
 464
 465static struct socket *sock_alloc(void)
 466{
 467	struct inode *inode;
 468	struct socket *sock;
 469
 470	inode = new_inode_pseudo(sock_mnt->mnt_sb);
 471	if (!inode)
 472		return NULL;
 473
 474	sock = SOCKET_I(inode);
 475
 476	kmemcheck_annotate_bitfield(sock, type);
 477	inode->i_ino = get_next_ino();
 478	inode->i_mode = S_IFSOCK | S_IRWXUGO;
 479	inode->i_uid = current_fsuid();
 480	inode->i_gid = current_fsgid();
 481
 482	percpu_add(sockets_in_use, 1);
 483	return sock;
 484}
 485
 486/*
 487 *	In theory you can't get an open on this inode, but /proc provides
 488 *	a back door. Remember to keep it shut otherwise you'll let the
 489 *	creepy crawlies in.
 490 */
 491
 492static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 493{
 494	return -ENXIO;
 495}
 496
 497const struct file_operations bad_sock_fops = {
 498	.owner = THIS_MODULE,
 499	.open = sock_no_open,
 500	.llseek = noop_llseek,
 501};
 502
 503/**
 504 *	sock_release	-	close a socket
 505 *	@sock: socket to close
 506 *
 507 *	The socket is released from the protocol stack if it has a release
 508 *	callback, and the inode is then released if the socket is bound to
 509 *	an inode not a file.
 510 */
 511
 512void sock_release(struct socket *sock)
 513{
 514	if (sock->ops) {
 515		struct module *owner = sock->ops->owner;
 516
 517		sock->ops->release(sock);
 518		sock->ops = NULL;
 519		module_put(owner);
 520	}
 521
 522	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
 523		printk(KERN_ERR "sock_release: fasync list not empty!\n");
 524
 525	if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
 526		return;
 527
 528	percpu_sub(sockets_in_use, 1);
 529	if (!sock->file) {
 530		iput(SOCK_INODE(sock));
 531		return;
 532	}
 533	sock->file = NULL;
 534}
 535EXPORT_SYMBOL(sock_release);
 536
 537int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 538{
 539	*tx_flags = 0;
 540	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 541		*tx_flags |= SKBTX_HW_TSTAMP;
 542	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 543		*tx_flags |= SKBTX_SW_TSTAMP;
 544	if (sock_flag(sk, SOCK_WIFI_STATUS))
 545		*tx_flags |= SKBTX_WIFI_STATUS;
 546	return 0;
 547}
 548EXPORT_SYMBOL(sock_tx_timestamp);
 549
 550static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 551				       struct msghdr *msg, size_t size)
 552{
 553	struct sock_iocb *si = kiocb_to_siocb(iocb);
 554
 555	sock_update_classid(sock->sk);
 556
 557	sock_update_netprioidx(sock->sk);
 558
 559	si->sock = sock;
 560	si->scm = NULL;
 561	si->msg = msg;
 562	si->size = size;
 563
 564	return sock->ops->sendmsg(iocb, sock, msg, size);
 565}
 566
 567static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 568				 struct msghdr *msg, size_t size)
 569{
 570	int err = security_socket_sendmsg(sock, msg, size);
 571
 572	return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 573}
 574
 575int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 576{
 577	struct kiocb iocb;
 578	struct sock_iocb siocb;
 579	int ret;
 580
 581	init_sync_kiocb(&iocb, NULL);
 582	iocb.private = &siocb;
 583	ret = __sock_sendmsg(&iocb, sock, msg, size);
 584	if (-EIOCBQUEUED == ret)
 585		ret = wait_on_sync_kiocb(&iocb);
 586	return ret;
 587}
 588EXPORT_SYMBOL(sock_sendmsg);
 589
 590static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 591{
 592	struct kiocb iocb;
 593	struct sock_iocb siocb;
 594	int ret;
 595
 596	init_sync_kiocb(&iocb, NULL);
 597	iocb.private = &siocb;
 598	ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 599	if (-EIOCBQUEUED == ret)
 600		ret = wait_on_sync_kiocb(&iocb);
 601	return ret;
 602}
 603
 604int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 605		   struct kvec *vec, size_t num, size_t size)
 606{
 607	mm_segment_t oldfs = get_fs();
 608	int result;
 609
 610	set_fs(KERNEL_DS);
 611	/*
 612	 * the following is safe, since for compiler definitions of kvec and
 613	 * iovec are identical, yielding the same in-core layout and alignment
 614	 */
 615	msg->msg_iov = (struct iovec *)vec;
 616	msg->msg_iovlen = num;
 617	result = sock_sendmsg(sock, msg, size);
 618	set_fs(oldfs);
 619	return result;
 620}
 621EXPORT_SYMBOL(kernel_sendmsg);
 622
 623static int ktime2ts(ktime_t kt, struct timespec *ts)
 624{
 625	if (kt.tv64) {
 626		*ts = ktime_to_timespec(kt);
 627		return 1;
 628	} else {
 629		return 0;
 630	}
 631}
 632
 633/*
 634 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 635 */
 636void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 637	struct sk_buff *skb)
 638{
 639	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 640	struct timespec ts[3];
 641	int empty = 1;
 642	struct skb_shared_hwtstamps *shhwtstamps =
 643		skb_hwtstamps(skb);
 644
 645	/* Race occurred between timestamp enabling and packet
 646	   receiving.  Fill in the current time for now. */
 647	if (need_software_tstamp && skb->tstamp.tv64 == 0)
 648		__net_timestamp(skb);
 649
 650	if (need_software_tstamp) {
 651		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 652			struct timeval tv;
 653			skb_get_timestamp(skb, &tv);
 654			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 655				 sizeof(tv), &tv);
 656		} else {
 657			skb_get_timestampns(skb, &ts[0]);
 658			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 659				 sizeof(ts[0]), &ts[0]);
 660		}
 661	}
 662
 663
 664	memset(ts, 0, sizeof(ts));
 665	if (skb->tstamp.tv64 &&
 666	    sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 667		skb_get_timestampns(skb, ts + 0);
 668		empty = 0;
 669	}
 670	if (shhwtstamps) {
 671		if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 672		    ktime2ts(shhwtstamps->syststamp, ts + 1))
 673			empty = 0;
 674		if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 675		    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 676			empty = 0;
 677	}
 678	if (!empty)
 679		put_cmsg(msg, SOL_SOCKET,
 680			 SCM_TIMESTAMPING, sizeof(ts), &ts);
 681}
 682EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 683
 684void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
 685	struct sk_buff *skb)
 686{
 687	int ack;
 688
 689	if (!sock_flag(sk, SOCK_WIFI_STATUS))
 690		return;
 691	if (!skb->wifi_acked_valid)
 692		return;
 693
 694	ack = skb->wifi_acked;
 695
 696	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
 697}
 698EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
 699
 700static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 701				   struct sk_buff *skb)
 702{
 703	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 704		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 705			sizeof(__u32), &skb->dropcount);
 706}
 707
 708void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 709	struct sk_buff *skb)
 710{
 711	sock_recv_timestamp(msg, sk, skb);
 712	sock_recv_drops(msg, sk, skb);
 713}
 714EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 715
 716static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 717				       struct msghdr *msg, size_t size, int flags)
 718{
 719	struct sock_iocb *si = kiocb_to_siocb(iocb);
 720
 721	sock_update_classid(sock->sk);
 722
 723	si->sock = sock;
 724	si->scm = NULL;
 725	si->msg = msg;
 726	si->size = size;
 727	si->flags = flags;
 728
 729	return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 730}
 731
 732static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 733				 struct msghdr *msg, size_t size, int flags)
 734{
 735	int err = security_socket_recvmsg(sock, msg, size, flags);
 736
 737	return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 738}
 739
 740int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 741		 size_t size, int flags)
 742{
 743	struct kiocb iocb;
 744	struct sock_iocb siocb;
 745	int ret;
 746
 747	init_sync_kiocb(&iocb, NULL);
 748	iocb.private = &siocb;
 749	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 750	if (-EIOCBQUEUED == ret)
 751		ret = wait_on_sync_kiocb(&iocb);
 752	return ret;
 753}
 754EXPORT_SYMBOL(sock_recvmsg);
 755
 756static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 757			      size_t size, int flags)
 758{
 759	struct kiocb iocb;
 760	struct sock_iocb siocb;
 761	int ret;
 762
 763	init_sync_kiocb(&iocb, NULL);
 764	iocb.private = &siocb;
 765	ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 766	if (-EIOCBQUEUED == ret)
 767		ret = wait_on_sync_kiocb(&iocb);
 768	return ret;
 769}
 770
 771/**
 772 * kernel_recvmsg - Receive a message from a socket (kernel space)
 773 * @sock:       The socket to receive the message from
 774 * @msg:        Received message
 775 * @vec:        Input s/g array for message data
 776 * @num:        Size of input s/g array
 777 * @size:       Number of bytes to read
 778 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 779 *
 780 * On return the msg structure contains the scatter/gather array passed in the
 781 * vec argument. The array is modified so that it consists of the unfilled
 782 * portion of the original array.
 783 *
 784 * The returned value is the total number of bytes received, or an error.
 785 */
 786int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 787		   struct kvec *vec, size_t num, size_t size, int flags)
 788{
 789	mm_segment_t oldfs = get_fs();
 790	int result;
 791
 792	set_fs(KERNEL_DS);
 793	/*
 794	 * the following is safe, since for compiler definitions of kvec and
 795	 * iovec are identical, yielding the same in-core layout and alignment
 796	 */
 797	msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 798	result = sock_recvmsg(sock, msg, size, flags);
 799	set_fs(oldfs);
 800	return result;
 801}
 802EXPORT_SYMBOL(kernel_recvmsg);
 803
 804static void sock_aio_dtor(struct kiocb *iocb)
 805{
 806	kfree(iocb->private);
 807}
 808
 809static ssize_t sock_sendpage(struct file *file, struct page *page,
 810			     int offset, size_t size, loff_t *ppos, int more)
 811{
 812	struct socket *sock;
 813	int flags;
 814
 815	sock = file->private_data;
 816
 817	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 818	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 819	flags |= more;
 820
 821	return kernel_sendpage(sock, page, offset, size, flags);
 822}
 823
 824static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 825				struct pipe_inode_info *pipe, size_t len,
 826				unsigned int flags)
 827{
 828	struct socket *sock = file->private_data;
 829
 830	if (unlikely(!sock->ops->splice_read))
 831		return -EINVAL;
 832
 833	sock_update_classid(sock->sk);
 834
 835	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 836}
 837
 838static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 839					 struct sock_iocb *siocb)
 840{
 841	if (!is_sync_kiocb(iocb)) {
 842		siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 843		if (!siocb)
 844			return NULL;
 845		iocb->ki_dtor = sock_aio_dtor;
 846	}
 847
 848	siocb->kiocb = iocb;
 849	iocb->private = siocb;
 850	return siocb;
 851}
 852
 853static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 854		struct file *file, const struct iovec *iov,
 855		unsigned long nr_segs)
 856{
 857	struct socket *sock = file->private_data;
 858	size_t size = 0;
 859	int i;
 860
 861	for (i = 0; i < nr_segs; i++)
 862		size += iov[i].iov_len;
 863
 864	msg->msg_name = NULL;
 865	msg->msg_namelen = 0;
 866	msg->msg_control = NULL;
 867	msg->msg_controllen = 0;
 868	msg->msg_iov = (struct iovec *)iov;
 869	msg->msg_iovlen = nr_segs;
 870	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 871
 872	return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 873}
 874
 875static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 876				unsigned long nr_segs, loff_t pos)
 877{
 878	struct sock_iocb siocb, *x;
 879
 880	if (pos != 0)
 881		return -ESPIPE;
 882
 883	if (iocb->ki_left == 0)	/* Match SYS5 behaviour */
 884		return 0;
 885
 886
 887	x = alloc_sock_iocb(iocb, &siocb);
 888	if (!x)
 889		return -ENOMEM;
 890	return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 891}
 892
 893static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 894			struct file *file, const struct iovec *iov,
 895			unsigned long nr_segs)
 896{
 897	struct socket *sock = file->private_data;
 898	size_t size = 0;
 899	int i;
 900
 901	for (i = 0; i < nr_segs; i++)
 902		size += iov[i].iov_len;
 903
 904	msg->msg_name = NULL;
 905	msg->msg_namelen = 0;
 906	msg->msg_control = NULL;
 907	msg->msg_controllen = 0;
 908	msg->msg_iov = (struct iovec *)iov;
 909	msg->msg_iovlen = nr_segs;
 910	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 911	if (sock->type == SOCK_SEQPACKET)
 912		msg->msg_flags |= MSG_EOR;
 913
 914	return __sock_sendmsg(iocb, sock, msg, size);
 915}
 916
 917static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 918			  unsigned long nr_segs, loff_t pos)
 919{
 920	struct sock_iocb siocb, *x;
 921
 922	if (pos != 0)
 923		return -ESPIPE;
 924
 925	x = alloc_sock_iocb(iocb, &siocb);
 926	if (!x)
 927		return -ENOMEM;
 928
 929	return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 930}
 931
 932/*
 933 * Atomic setting of ioctl hooks to avoid race
 934 * with module unload.
 935 */
 936
 937static DEFINE_MUTEX(br_ioctl_mutex);
 938static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 939
 940void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 941{
 942	mutex_lock(&br_ioctl_mutex);
 943	br_ioctl_hook = hook;
 944	mutex_unlock(&br_ioctl_mutex);
 945}
 946EXPORT_SYMBOL(brioctl_set);
 947
 948static DEFINE_MUTEX(vlan_ioctl_mutex);
 949static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 950
 951void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 952{
 953	mutex_lock(&vlan_ioctl_mutex);
 954	vlan_ioctl_hook = hook;
 955	mutex_unlock(&vlan_ioctl_mutex);
 956}
 957EXPORT_SYMBOL(vlan_ioctl_set);
 958
 959static DEFINE_MUTEX(dlci_ioctl_mutex);
 960static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 961
 962void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 963{
 964	mutex_lock(&dlci_ioctl_mutex);
 965	dlci_ioctl_hook = hook;
 966	mutex_unlock(&dlci_ioctl_mutex);
 967}
 968EXPORT_SYMBOL(dlci_ioctl_set);
 969
 970static long sock_do_ioctl(struct net *net, struct socket *sock,
 971				 unsigned int cmd, unsigned long arg)
 972{
 973	int err;
 974	void __user *argp = (void __user *)arg;
 975
 976	err = sock->ops->ioctl(sock, cmd, arg);
 977
 978	/*
 979	 * If this ioctl is unknown try to hand it down
 980	 * to the NIC driver.
 981	 */
 982	if (err == -ENOIOCTLCMD)
 983		err = dev_ioctl(net, cmd, argp);
 984
 985	return err;
 986}
 987
 988/*
 989 *	With an ioctl, arg may well be a user mode pointer, but we don't know
 990 *	what to do with it - that's up to the protocol still.
 991 */
 992
 993static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 994{
 995	struct socket *sock;
 996	struct sock *sk;
 997	void __user *argp = (void __user *)arg;
 998	int pid, err;
 999	struct net *net;
1000
1001	sock = file->private_data;
1002	sk = sock->sk;
1003	net = sock_net(sk);
1004	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1005		err = dev_ioctl(net, cmd, argp);
1006	} else
1007#ifdef CONFIG_WEXT_CORE
1008	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1009		err = dev_ioctl(net, cmd, argp);
1010	} else
1011#endif
1012		switch (cmd) {
1013		case FIOSETOWN:
1014		case SIOCSPGRP:
1015			err = -EFAULT;
1016			if (get_user(pid, (int __user *)argp))
1017				break;
1018			err = f_setown(sock->file, pid, 1);
1019			break;
1020		case FIOGETOWN:
1021		case SIOCGPGRP:
1022			err = put_user(f_getown(sock->file),
1023				       (int __user *)argp);
1024			break;
1025		case SIOCGIFBR:
1026		case SIOCSIFBR:
1027		case SIOCBRADDBR:
1028		case SIOCBRDELBR:
1029			err = -ENOPKG;
1030			if (!br_ioctl_hook)
1031				request_module("bridge");
1032
1033			mutex_lock(&br_ioctl_mutex);
1034			if (br_ioctl_hook)
1035				err = br_ioctl_hook(net, cmd, argp);
1036			mutex_unlock(&br_ioctl_mutex);
1037			break;
1038		case SIOCGIFVLAN:
1039		case SIOCSIFVLAN:
1040			err = -ENOPKG;
1041			if (!vlan_ioctl_hook)
1042				request_module("8021q");
1043
1044			mutex_lock(&vlan_ioctl_mutex);
1045			if (vlan_ioctl_hook)
1046				err = vlan_ioctl_hook(net, argp);
1047			mutex_unlock(&vlan_ioctl_mutex);
1048			break;
1049		case SIOCADDDLCI:
1050		case SIOCDELDLCI:
1051			err = -ENOPKG;
1052			if (!dlci_ioctl_hook)
1053				request_module("dlci");
1054
1055			mutex_lock(&dlci_ioctl_mutex);
1056			if (dlci_ioctl_hook)
1057				err = dlci_ioctl_hook(cmd, argp);
1058			mutex_unlock(&dlci_ioctl_mutex);
1059			break;
1060		default:
1061			err = sock_do_ioctl(net, sock, cmd, arg);
1062			break;
1063		}
1064	return err;
1065}
1066
1067int sock_create_lite(int family, int type, int protocol, struct socket **res)
1068{
1069	int err;
1070	struct socket *sock = NULL;
1071
1072	err = security_socket_create(family, type, protocol, 1);
1073	if (err)
1074		goto out;
1075
1076	sock = sock_alloc();
1077	if (!sock) {
1078		err = -ENOMEM;
1079		goto out;
1080	}
1081
1082	sock->type = type;
1083	err = security_socket_post_create(sock, family, type, protocol, 1);
1084	if (err)
1085		goto out_release;
1086
1087out:
1088	*res = sock;
1089	return err;
1090out_release:
1091	sock_release(sock);
1092	sock = NULL;
1093	goto out;
1094}
1095EXPORT_SYMBOL(sock_create_lite);
1096
1097/* No kernel lock held - perfect */
1098static unsigned int sock_poll(struct file *file, poll_table *wait)
1099{
1100	struct socket *sock;
1101
1102	/*
1103	 *      We can't return errors to poll, so it's either yes or no.
1104	 */
1105	sock = file->private_data;
1106	return sock->ops->poll(file, sock, wait);
1107}
1108
1109static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1110{
1111	struct socket *sock = file->private_data;
1112
1113	return sock->ops->mmap(file, sock, vma);
1114}
1115
1116static int sock_close(struct inode *inode, struct file *filp)
1117{
1118	/*
1119	 *      It was possible the inode is NULL we were
1120	 *      closing an unfinished socket.
1121	 */
1122
1123	if (!inode) {
1124		printk(KERN_DEBUG "sock_close: NULL inode\n");
1125		return 0;
1126	}
1127	sock_release(SOCKET_I(inode));
1128	return 0;
1129}
1130
1131/*
1132 *	Update the socket async list
1133 *
1134 *	Fasync_list locking strategy.
1135 *
1136 *	1. fasync_list is modified only under process context socket lock
1137 *	   i.e. under semaphore.
1138 *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1139 *	   or under socket lock
1140 */
1141
1142static int sock_fasync(int fd, struct file *filp, int on)
1143{
1144	struct socket *sock = filp->private_data;
1145	struct sock *sk = sock->sk;
1146	struct socket_wq *wq;
1147
1148	if (sk == NULL)
1149		return -EINVAL;
1150
1151	lock_sock(sk);
1152	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1153	fasync_helper(fd, filp, on, &wq->fasync_list);
1154
1155	if (!wq->fasync_list)
1156		sock_reset_flag(sk, SOCK_FASYNC);
1157	else
1158		sock_set_flag(sk, SOCK_FASYNC);
1159
1160	release_sock(sk);
1161	return 0;
1162}
1163
1164/* This function may be called only under socket lock or callback_lock or rcu_lock */
1165
1166int sock_wake_async(struct socket *sock, int how, int band)
1167{
1168	struct socket_wq *wq;
1169
1170	if (!sock)
1171		return -1;
1172	rcu_read_lock();
1173	wq = rcu_dereference(sock->wq);
1174	if (!wq || !wq->fasync_list) {
1175		rcu_read_unlock();
1176		return -1;
1177	}
1178	switch (how) {
1179	case SOCK_WAKE_WAITD:
1180		if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1181			break;
1182		goto call_kill;
1183	case SOCK_WAKE_SPACE:
1184		if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1185			break;
1186		/* fall through */
1187	case SOCK_WAKE_IO:
1188call_kill:
1189		kill_fasync(&wq->fasync_list, SIGIO, band);
1190		break;
1191	case SOCK_WAKE_URG:
1192		kill_fasync(&wq->fasync_list, SIGURG, band);
1193	}
1194	rcu_read_unlock();
1195	return 0;
1196}
1197EXPORT_SYMBOL(sock_wake_async);
1198
1199int __sock_create(struct net *net, int family, int type, int protocol,
1200			 struct socket **res, int kern)
1201{
1202	int err;
1203	struct socket *sock;
1204	const struct net_proto_family *pf;
1205
1206	/*
1207	 *      Check protocol is in range
1208	 */
1209	if (family < 0 || family >= NPROTO)
1210		return -EAFNOSUPPORT;
1211	if (type < 0 || type >= SOCK_MAX)
1212		return -EINVAL;
1213
1214	/* Compatibility.
1215
1216	   This uglymoron is moved from INET layer to here to avoid
1217	   deadlock in module load.
1218	 */
1219	if (family == PF_INET && type == SOCK_PACKET) {
1220		static int warned;
1221		if (!warned) {
1222			warned = 1;
1223			printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1224			       current->comm);
1225		}
1226		family = PF_PACKET;
1227	}
1228
1229	err = security_socket_create(family, type, protocol, kern);
1230	if (err)
1231		return err;
1232
1233	/*
1234	 *	Allocate the socket and allow the family to set things up. if
1235	 *	the protocol is 0, the family is instructed to select an appropriate
1236	 *	default.
1237	 */
1238	sock = sock_alloc();
1239	if (!sock) {
1240		if (net_ratelimit())
1241			printk(KERN_WARNING "socket: no more sockets\n");
1242		return -ENFILE;	/* Not exactly a match, but its the
1243				   closest posix thing */
1244	}
1245
1246	sock->type = type;
1247
1248#ifdef CONFIG_MODULES
1249	/* Attempt to load a protocol module if the find failed.
1250	 *
1251	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1252	 * requested real, full-featured networking support upon configuration.
1253	 * Otherwise module support will break!
1254	 */
1255	if (rcu_access_pointer(net_families[family]) == NULL)
1256		request_module("net-pf-%d", family);
1257#endif
1258
1259	rcu_read_lock();
1260	pf = rcu_dereference(net_families[family]);
1261	err = -EAFNOSUPPORT;
1262	if (!pf)
1263		goto out_release;
1264
1265	/*
1266	 * We will call the ->create function, that possibly is in a loadable
1267	 * module, so we have to bump that loadable module refcnt first.
1268	 */
1269	if (!try_module_get(pf->owner))
1270		goto out_release;
1271
1272	/* Now protected by module ref count */
1273	rcu_read_unlock();
1274
1275	err = pf->create(net, sock, protocol, kern);
1276	if (err < 0)
1277		goto out_module_put;
1278
1279	/*
1280	 * Now to bump the refcnt of the [loadable] module that owns this
1281	 * socket at sock_release time we decrement its refcnt.
1282	 */
1283	if (!try_module_get(sock->ops->owner))
1284		goto out_module_busy;
1285
1286	/*
1287	 * Now that we're done with the ->create function, the [loadable]
1288	 * module can have its refcnt decremented
1289	 */
1290	module_put(pf->owner);
1291	err = security_socket_post_create(sock, family, type, protocol, kern);
1292	if (err)
1293		goto out_sock_release;
1294	*res = sock;
1295
1296	return 0;
1297
1298out_module_busy:
1299	err = -EAFNOSUPPORT;
1300out_module_put:
1301	sock->ops = NULL;
1302	module_put(pf->owner);
1303out_sock_release:
1304	sock_release(sock);
1305	return err;
1306
1307out_release:
1308	rcu_read_unlock();
1309	goto out_sock_release;
1310}
1311EXPORT_SYMBOL(__sock_create);
1312
1313int sock_create(int family, int type, int protocol, struct socket **res)
1314{
1315	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1316}
1317EXPORT_SYMBOL(sock_create);
1318
1319int sock_create_kern(int family, int type, int protocol, struct socket **res)
1320{
1321	return __sock_create(&init_net, family, type, protocol, res, 1);
1322}
1323EXPORT_SYMBOL(sock_create_kern);
1324
1325SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1326{
1327	int retval;
1328	struct socket *sock;
1329	int flags;
1330
1331	/* Check the SOCK_* constants for consistency.  */
1332	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1333	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1334	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1335	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1336
1337	flags = type & ~SOCK_TYPE_MASK;
1338	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1339		return -EINVAL;
1340	type &= SOCK_TYPE_MASK;
1341
1342	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1343		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1344
1345	retval = sock_create(family, type, protocol, &sock);
1346	if (retval < 0)
1347		goto out;
1348
1349	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1350	if (retval < 0)
1351		goto out_release;
1352
1353out:
1354	/* It may be already another descriptor 8) Not kernel problem. */
1355	return retval;
1356
1357out_release:
1358	sock_release(sock);
1359	return retval;
1360}
1361
1362/*
1363 *	Create a pair of connected sockets.
1364 */
1365
1366SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1367		int __user *, usockvec)
1368{
1369	struct socket *sock1, *sock2;
1370	int fd1, fd2, err;
1371	struct file *newfile1, *newfile2;
1372	int flags;
1373
1374	flags = type & ~SOCK_TYPE_MASK;
1375	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1376		return -EINVAL;
1377	type &= SOCK_TYPE_MASK;
1378
1379	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1380		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1381
1382	/*
1383	 * Obtain the first socket and check if the underlying protocol
1384	 * supports the socketpair call.
1385	 */
1386
1387	err = sock_create(family, type, protocol, &sock1);
1388	if (err < 0)
1389		goto out;
1390
1391	err = sock_create(family, type, protocol, &sock2);
1392	if (err < 0)
1393		goto out_release_1;
1394
1395	err = sock1->ops->socketpair(sock1, sock2);
1396	if (err < 0)
1397		goto out_release_both;
1398
1399	fd1 = sock_alloc_file(sock1, &newfile1, flags);
1400	if (unlikely(fd1 < 0)) {
1401		err = fd1;
1402		goto out_release_both;
1403	}
1404
1405	fd2 = sock_alloc_file(sock2, &newfile2, flags);
1406	if (unlikely(fd2 < 0)) {
1407		err = fd2;
1408		fput(newfile1);
1409		put_unused_fd(fd1);
1410		sock_release(sock2);
1411		goto out;
1412	}
1413
1414	audit_fd_pair(fd1, fd2);
1415	fd_install(fd1, newfile1);
1416	fd_install(fd2, newfile2);
1417	/* fd1 and fd2 may be already another descriptors.
1418	 * Not kernel problem.
1419	 */
1420
1421	err = put_user(fd1, &usockvec[0]);
1422	if (!err)
1423		err = put_user(fd2, &usockvec[1]);
1424	if (!err)
1425		return 0;
1426
1427	sys_close(fd2);
1428	sys_close(fd1);
1429	return err;
1430
1431out_release_both:
1432	sock_release(sock2);
1433out_release_1:
1434	sock_release(sock1);
1435out:
1436	return err;
1437}
1438
1439/*
1440 *	Bind a name to a socket. Nothing much to do here since it's
1441 *	the protocol's responsibility to handle the local address.
1442 *
1443 *	We move the socket address to kernel space before we call
1444 *	the protocol layer (having also checked the address is ok).
1445 */
1446
1447SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1448{
1449	struct socket *sock;
1450	struct sockaddr_storage address;
1451	int err, fput_needed;
1452
1453	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1454	if (sock) {
1455		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1456		if (err >= 0) {
1457			err = security_socket_bind(sock,
1458						   (struct sockaddr *)&address,
1459						   addrlen);
1460			if (!err)
1461				err = sock->ops->bind(sock,
1462						      (struct sockaddr *)
1463						      &address, addrlen);
1464		}
1465		fput_light(sock->file, fput_needed);
1466	}
1467	return err;
1468}
1469
1470/*
1471 *	Perform a listen. Basically, we allow the protocol to do anything
1472 *	necessary for a listen, and if that works, we mark the socket as
1473 *	ready for listening.
1474 */
1475
1476SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1477{
1478	struct socket *sock;
1479	int err, fput_needed;
1480	int somaxconn;
1481
1482	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1483	if (sock) {
1484		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1485		if ((unsigned)backlog > somaxconn)
1486			backlog = somaxconn;
1487
1488		err = security_socket_listen(sock, backlog);
1489		if (!err)
1490			err = sock->ops->listen(sock, backlog);
1491
1492		fput_light(sock->file, fput_needed);
1493	}
1494	return err;
1495}
1496
1497/*
1498 *	For accept, we attempt to create a new socket, set up the link
1499 *	with the client, wake up the client, then return the new
1500 *	connected fd. We collect the address of the connector in kernel
1501 *	space and move it to user at the very end. This is unclean because
1502 *	we open the socket then return an error.
1503 *
1504 *	1003.1g adds the ability to recvmsg() to query connection pending
1505 *	status to recvmsg. We need to add that support in a way thats
1506 *	clean when we restucture accept also.
1507 */
1508
1509SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1510		int __user *, upeer_addrlen, int, flags)
1511{
1512	struct socket *sock, *newsock;
1513	struct file *newfile;
1514	int err, len, newfd, fput_needed;
1515	struct sockaddr_storage address;
1516
1517	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1518		return -EINVAL;
1519
1520	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1521		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1522
1523	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1524	if (!sock)
1525		goto out;
1526
1527	err = -ENFILE;
1528	newsock = sock_alloc();
1529	if (!newsock)
1530		goto out_put;
1531
1532	newsock->type = sock->type;
1533	newsock->ops = sock->ops;
1534
1535	/*
1536	 * We don't need try_module_get here, as the listening socket (sock)
1537	 * has the protocol module (sock->ops->owner) held.
1538	 */
1539	__module_get(newsock->ops->owner);
1540
1541	newfd = sock_alloc_file(newsock, &newfile, flags);
1542	if (unlikely(newfd < 0)) {
1543		err = newfd;
1544		sock_release(newsock);
1545		goto out_put;
1546	}
1547
1548	err = security_socket_accept(sock, newsock);
1549	if (err)
1550		goto out_fd;
1551
1552	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1553	if (err < 0)
1554		goto out_fd;
1555
1556	if (upeer_sockaddr) {
1557		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1558					  &len, 2) < 0) {
1559			err = -ECONNABORTED;
1560			goto out_fd;
1561		}
1562		err = move_addr_to_user(&address,
1563					len, upeer_sockaddr, upeer_addrlen);
1564		if (err < 0)
1565			goto out_fd;
1566	}
1567
1568	/* File flags are not inherited via accept() unlike another OSes. */
1569
1570	fd_install(newfd, newfile);
1571	err = newfd;
1572
1573out_put:
1574	fput_light(sock->file, fput_needed);
1575out:
1576	return err;
1577out_fd:
1578	fput(newfile);
1579	put_unused_fd(newfd);
1580	goto out_put;
1581}
1582
1583SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1584		int __user *, upeer_addrlen)
1585{
1586	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1587}
1588
1589/*
1590 *	Attempt to connect to a socket with the server address.  The address
1591 *	is in user space so we verify it is OK and move it to kernel space.
1592 *
1593 *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1594 *	break bindings
1595 *
1596 *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1597 *	other SEQPACKET protocols that take time to connect() as it doesn't
1598 *	include the -EINPROGRESS status for such sockets.
1599 */
1600
1601SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1602		int, addrlen)
1603{
1604	struct socket *sock;
1605	struct sockaddr_storage address;
1606	int err, fput_needed;
1607
1608	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1609	if (!sock)
1610		goto out;
1611	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1612	if (err < 0)
1613		goto out_put;
1614
1615	err =
1616	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1617	if (err)
1618		goto out_put;
1619
1620	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1621				 sock->file->f_flags);
1622out_put:
1623	fput_light(sock->file, fput_needed);
1624out:
1625	return err;
1626}
1627
1628/*
1629 *	Get the local address ('name') of a socket object. Move the obtained
1630 *	name to user space.
1631 */
1632
1633SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1634		int __user *, usockaddr_len)
1635{
1636	struct socket *sock;
1637	struct sockaddr_storage address;
1638	int len, err, fput_needed;
1639
1640	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1641	if (!sock)
1642		goto out;
1643
1644	err = security_socket_getsockname(sock);
1645	if (err)
1646		goto out_put;
1647
1648	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1649	if (err)
1650		goto out_put;
1651	err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1652
1653out_put:
1654	fput_light(sock->file, fput_needed);
1655out:
1656	return err;
1657}
1658
1659/*
1660 *	Get the remote address ('name') of a socket object. Move the obtained
1661 *	name to user space.
1662 */
1663
1664SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1665		int __user *, usockaddr_len)
1666{
1667	struct socket *sock;
1668	struct sockaddr_storage address;
1669	int len, err, fput_needed;
1670
1671	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1672	if (sock != NULL) {
1673		err = security_socket_getpeername(sock);
1674		if (err) {
1675			fput_light(sock->file, fput_needed);
1676			return err;
1677		}
1678
1679		err =
1680		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1681				       1);
1682		if (!err)
1683			err = move_addr_to_user(&address, len, usockaddr,
1684						usockaddr_len);
1685		fput_light(sock->file, fput_needed);
1686	}
1687	return err;
1688}
1689
1690/*
1691 *	Send a datagram to a given address. We move the address into kernel
1692 *	space and check the user space data area is readable before invoking
1693 *	the protocol.
1694 */
1695
1696SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1697		unsigned, flags, struct sockaddr __user *, addr,
1698		int, addr_len)
1699{
1700	struct socket *sock;
1701	struct sockaddr_storage address;
1702	int err;
1703	struct msghdr msg;
1704	struct iovec iov;
1705	int fput_needed;
1706
1707	if (len > INT_MAX)
1708		len = INT_MAX;
1709	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1710	if (!sock)
1711		goto out;
1712
1713	iov.iov_base = buff;
1714	iov.iov_len = len;
1715	msg.msg_name = NULL;
1716	msg.msg_iov = &iov;
1717	msg.msg_iovlen = 1;
1718	msg.msg_control = NULL;
1719	msg.msg_controllen = 0;
1720	msg.msg_namelen = 0;
1721	if (addr) {
1722		err = move_addr_to_kernel(addr, addr_len, &address);
1723		if (err < 0)
1724			goto out_put;
1725		msg.msg_name = (struct sockaddr *)&address;
1726		msg.msg_namelen = addr_len;
1727	}
1728	if (sock->file->f_flags & O_NONBLOCK)
1729		flags |= MSG_DONTWAIT;
1730	msg.msg_flags = flags;
1731	err = sock_sendmsg(sock, &msg, len);
1732
1733out_put:
1734	fput_light(sock->file, fput_needed);
1735out:
1736	return err;
1737}
1738
1739/*
1740 *	Send a datagram down a socket.
1741 */
1742
1743SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1744		unsigned, flags)
1745{
1746	return sys_sendto(fd, buff, len, flags, NULL, 0);
1747}
1748
1749/*
1750 *	Receive a frame from the socket and optionally record the address of the
1751 *	sender. We verify the buffers are writable and if needed move the
1752 *	sender address from kernel to user space.
1753 */
1754
1755SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1756		unsigned, flags, struct sockaddr __user *, addr,
1757		int __user *, addr_len)
1758{
1759	struct socket *sock;
1760	struct iovec iov;
1761	struct msghdr msg;
1762	struct sockaddr_storage address;
1763	int err, err2;
1764	int fput_needed;
1765
1766	if (size > INT_MAX)
1767		size = INT_MAX;
1768	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769	if (!sock)
1770		goto out;
1771
1772	msg.msg_control = NULL;
1773	msg.msg_controllen = 0;
1774	msg.msg_iovlen = 1;
1775	msg.msg_iov = &iov;
1776	iov.iov_len = size;
1777	iov.iov_base = ubuf;
1778	msg.msg_name = (struct sockaddr *)&address;
1779	msg.msg_namelen = sizeof(address);
1780	if (sock->file->f_flags & O_NONBLOCK)
1781		flags |= MSG_DONTWAIT;
1782	err = sock_recvmsg(sock, &msg, size, flags);
1783
1784	if (err >= 0 && addr != NULL) {
1785		err2 = move_addr_to_user(&address,
1786					 msg.msg_namelen, addr, addr_len);
1787		if (err2 < 0)
1788			err = err2;
1789	}
1790
1791	fput_light(sock->file, fput_needed);
1792out:
1793	return err;
1794}
1795
1796/*
1797 *	Receive a datagram from a socket.
1798 */
1799
1800asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1801			 unsigned flags)
1802{
1803	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1804}
1805
1806/*
1807 *	Set a socket option. Because we don't know the option lengths we have
1808 *	to pass the user mode parameter for the protocols to sort out.
1809 */
1810
1811SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1812		char __user *, optval, int, optlen)
1813{
1814	int err, fput_needed;
1815	struct socket *sock;
1816
1817	if (optlen < 0)
1818		return -EINVAL;
1819
1820	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1821	if (sock != NULL) {
1822		err = security_socket_setsockopt(sock, level, optname);
1823		if (err)
1824			goto out_put;
1825
1826		if (level == SOL_SOCKET)
1827			err =
1828			    sock_setsockopt(sock, level, optname, optval,
1829					    optlen);
1830		else
1831			err =
1832			    sock->ops->setsockopt(sock, level, optname, optval,
1833						  optlen);
1834out_put:
1835		fput_light(sock->file, fput_needed);
1836	}
1837	return err;
1838}
1839
1840/*
1841 *	Get a socket option. Because we don't know the option lengths we have
1842 *	to pass a user mode parameter for the protocols to sort out.
1843 */
1844
1845SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1846		char __user *, optval, int __user *, optlen)
1847{
1848	int err, fput_needed;
1849	struct socket *sock;
1850
1851	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1852	if (sock != NULL) {
1853		err = security_socket_getsockopt(sock, level, optname);
1854		if (err)
1855			goto out_put;
1856
1857		if (level == SOL_SOCKET)
1858			err =
1859			    sock_getsockopt(sock, level, optname, optval,
1860					    optlen);
1861		else
1862			err =
1863			    sock->ops->getsockopt(sock, level, optname, optval,
1864						  optlen);
1865out_put:
1866		fput_light(sock->file, fput_needed);
1867	}
1868	return err;
1869}
1870
1871/*
1872 *	Shutdown a socket.
1873 */
1874
1875SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1876{
1877	int err, fput_needed;
1878	struct socket *sock;
1879
1880	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1881	if (sock != NULL) {
1882		err = security_socket_shutdown(sock, how);
1883		if (!err)
1884			err = sock->ops->shutdown(sock, how);
1885		fput_light(sock->file, fput_needed);
1886	}
1887	return err;
1888}
1889
1890/* A couple of helpful macros for getting the address of the 32/64 bit
1891 * fields which are the same type (int / unsigned) on our platforms.
1892 */
1893#define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1894#define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1895#define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1896
1897struct used_address {
1898	struct sockaddr_storage name;
1899	unsigned int name_len;
1900};
1901
1902static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1903			 struct msghdr *msg_sys, unsigned flags,
1904			 struct used_address *used_address)
1905{
1906	struct compat_msghdr __user *msg_compat =
1907	    (struct compat_msghdr __user *)msg;
1908	struct sockaddr_storage address;
1909	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1910	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1911	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1912	/* 20 is size of ipv6_pktinfo */
1913	unsigned char *ctl_buf = ctl;
1914	int err, ctl_len, iov_size, total_len;
1915
1916	err = -EFAULT;
1917	if (MSG_CMSG_COMPAT & flags) {
1918		if (get_compat_msghdr(msg_sys, msg_compat))
1919			return -EFAULT;
1920	} else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1921		return -EFAULT;
1922
1923	/* do not move before msg_sys is valid */
1924	err = -EMSGSIZE;
1925	if (msg_sys->msg_iovlen > UIO_MAXIOV)
1926		goto out;
1927
1928	/* Check whether to allocate the iovec area */
1929	err = -ENOMEM;
1930	iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1931	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1932		iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1933		if (!iov)
1934			goto out;
1935	}
1936
1937	/* This will also move the address data into kernel space */
1938	if (MSG_CMSG_COMPAT & flags) {
1939		err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
1940	} else
1941		err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
1942	if (err < 0)
1943		goto out_freeiov;
1944	total_len = err;
1945
1946	err = -ENOBUFS;
1947
1948	if (msg_sys->msg_controllen > INT_MAX)
1949		goto out_freeiov;
1950	ctl_len = msg_sys->msg_controllen;
1951	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1952		err =
1953		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1954						     sizeof(ctl));
1955		if (err)
1956			goto out_freeiov;
1957		ctl_buf = msg_sys->msg_control;
1958		ctl_len = msg_sys->msg_controllen;
1959	} else if (ctl_len) {
1960		if (ctl_len > sizeof(ctl)) {
1961			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1962			if (ctl_buf == NULL)
1963				goto out_freeiov;
1964		}
1965		err = -EFAULT;
1966		/*
1967		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1968		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1969		 * checking falls down on this.
1970		 */
1971		if (copy_from_user(ctl_buf,
1972				   (void __user __force *)msg_sys->msg_control,
1973				   ctl_len))
1974			goto out_freectl;
1975		msg_sys->msg_control = ctl_buf;
1976	}
1977	msg_sys->msg_flags = flags;
1978
1979	if (sock->file->f_flags & O_NONBLOCK)
1980		msg_sys->msg_flags |= MSG_DONTWAIT;
1981	/*
1982	 * If this is sendmmsg() and current destination address is same as
1983	 * previously succeeded address, omit asking LSM's decision.
1984	 * used_address->name_len is initialized to UINT_MAX so that the first
1985	 * destination address never matches.
1986	 */
1987	if (used_address && msg_sys->msg_name &&
1988	    used_address->name_len == msg_sys->msg_namelen &&
1989	    !memcmp(&used_address->name, msg_sys->msg_name,
1990		    used_address->name_len)) {
1991		err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1992		goto out_freectl;
1993	}
1994	err = sock_sendmsg(sock, msg_sys, total_len);
1995	/*
1996	 * If this is sendmmsg() and sending to current destination address was
1997	 * successful, remember it.
1998	 */
1999	if (used_address && err >= 0) {
2000		used_address->name_len = msg_sys->msg_namelen;
2001		if (msg_sys->msg_name)
2002			memcpy(&used_address->name, msg_sys->msg_name,
2003			       used_address->name_len);
2004	}
2005
2006out_freectl:
2007	if (ctl_buf != ctl)
2008		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2009out_freeiov:
2010	if (iov != iovstack)
2011		sock_kfree_s(sock->sk, iov, iov_size);
2012out:
2013	return err;
2014}
2015
2016/*
2017 *	BSD sendmsg interface
2018 */
2019
2020SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
2021{
2022	int fput_needed, err;
2023	struct msghdr msg_sys;
2024	struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2025
2026	if (!sock)
2027		goto out;
2028
2029	err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2030
2031	fput_light(sock->file, fput_needed);
2032out:
2033	return err;
2034}
2035
2036/*
2037 *	Linux sendmmsg interface
2038 */
2039
2040int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2041		   unsigned int flags)
2042{
2043	int fput_needed, err, datagrams;
2044	struct socket *sock;
2045	struct mmsghdr __user *entry;
2046	struct compat_mmsghdr __user *compat_entry;
2047	struct msghdr msg_sys;
2048	struct used_address used_address;
2049
2050	if (vlen > UIO_MAXIOV)
2051		vlen = UIO_MAXIOV;
2052
2053	datagrams = 0;
2054
2055	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2056	if (!sock)
2057		return err;
2058
2059	used_address.name_len = UINT_MAX;
2060	entry = mmsg;
2061	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2062	err = 0;
2063
2064	while (datagrams < vlen) {
2065		if (MSG_CMSG_COMPAT & flags) {
2066			err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2067					    &msg_sys, flags, &used_address);
2068			if (err < 0)
2069				break;
2070			err = __put_user(err, &compat_entry->msg_len);
2071			++compat_entry;
2072		} else {
2073			err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
2074					    &msg_sys, flags, &used_address);
2075			if (err < 0)
2076				break;
2077			err = put_user(err, &entry->msg_len);
2078			++entry;
2079		}
2080
2081		if (err)
2082			break;
2083		++datagrams;
2084	}
2085
2086	fput_light(sock->file, fput_needed);
2087
2088	/* We only return an error if no datagrams were able to be sent */
2089	if (datagrams != 0)
2090		return datagrams;
2091
2092	return err;
2093}
2094
2095SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2096		unsigned int, vlen, unsigned int, flags)
2097{
2098	return __sys_sendmmsg(fd, mmsg, vlen, flags);
2099}
2100
2101static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2102			 struct msghdr *msg_sys, unsigned flags, int nosec)
2103{
2104	struct compat_msghdr __user *msg_compat =
2105	    (struct compat_msghdr __user *)msg;
2106	struct iovec iovstack[UIO_FASTIOV];
2107	struct iovec *iov = iovstack;
2108	unsigned long cmsg_ptr;
2109	int err, iov_size, total_len, len;
2110
2111	/* kernel mode address */
2112	struct sockaddr_storage addr;
2113
2114	/* user mode address pointers */
2115	struct sockaddr __user *uaddr;
2116	int __user *uaddr_len;
2117
2118	if (MSG_CMSG_COMPAT & flags) {
2119		if (get_compat_msghdr(msg_sys, msg_compat))
2120			return -EFAULT;
2121	} else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
2122		return -EFAULT;
2123
2124	err = -EMSGSIZE;
2125	if (msg_sys->msg_iovlen > UIO_MAXIOV)
2126		goto out;
2127
2128	/* Check whether to allocate the iovec area */
2129	err = -ENOMEM;
2130	iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2131	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2132		iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2133		if (!iov)
2134			goto out;
2135	}
2136
2137	/*
2138	 *      Save the user-mode address (verify_iovec will change the
2139	 *      kernel msghdr to use the kernel address space)
2140	 */
2141
2142	uaddr = (__force void __user *)msg_sys->msg_name;
2143	uaddr_len = COMPAT_NAMELEN(msg);
2144	if (MSG_CMSG_COMPAT & flags) {
2145		err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2146	} else
2147		err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2148	if (err < 0)
2149		goto out_freeiov;
2150	total_len = err;
2151
2152	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2153	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2154
2155	if (sock->file->f_flags & O_NONBLOCK)
2156		flags |= MSG_DONTWAIT;
2157	err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2158							  total_len, flags);
2159	if (err < 0)
2160		goto out_freeiov;
2161	len = err;
2162
2163	if (uaddr != NULL) {
2164		err = move_addr_to_user(&addr,
2165					msg_sys->msg_namelen, uaddr,
2166					uaddr_len);
2167		if (err < 0)
2168			goto out_freeiov;
2169	}
2170	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2171			 COMPAT_FLAGS(msg));
2172	if (err)
2173		goto out_freeiov;
2174	if (MSG_CMSG_COMPAT & flags)
2175		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2176				 &msg_compat->msg_controllen);
2177	else
2178		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2179				 &msg->msg_controllen);
2180	if (err)
2181		goto out_freeiov;
2182	err = len;
2183
2184out_freeiov:
2185	if (iov != iovstack)
2186		sock_kfree_s(sock->sk, iov, iov_size);
2187out:
2188	return err;
2189}
2190
2191/*
2192 *	BSD recvmsg interface
2193 */
2194
2195SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2196		unsigned int, flags)
2197{
2198	int fput_needed, err;
2199	struct msghdr msg_sys;
2200	struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2201
2202	if (!sock)
2203		goto out;
2204
2205	err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2206
2207	fput_light(sock->file, fput_needed);
2208out:
2209	return err;
2210}
2211
2212/*
2213 *     Linux recvmmsg interface
2214 */
2215
2216int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2217		   unsigned int flags, struct timespec *timeout)
2218{
2219	int fput_needed, err, datagrams;
2220	struct socket *sock;
2221	struct mmsghdr __user *entry;
2222	struct compat_mmsghdr __user *compat_entry;
2223	struct msghdr msg_sys;
2224	struct timespec end_time;
2225
2226	if (timeout &&
2227	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2228				    timeout->tv_nsec))
2229		return -EINVAL;
2230
2231	datagrams = 0;
2232
2233	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2234	if (!sock)
2235		return err;
2236
2237	err = sock_error(sock->sk);
2238	if (err)
2239		goto out_put;
2240
2241	entry = mmsg;
2242	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2243
2244	while (datagrams < vlen) {
2245		/*
2246		 * No need to ask LSM for more than the first datagram.
2247		 */
2248		if (MSG_CMSG_COMPAT & flags) {
2249			err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2250					    &msg_sys, flags & ~MSG_WAITFORONE,
2251					    datagrams);
2252			if (err < 0)
2253				break;
2254			err = __put_user(err, &compat_entry->msg_len);
2255			++compat_entry;
2256		} else {
2257			err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2258					    &msg_sys, flags & ~MSG_WAITFORONE,
2259					    datagrams);
2260			if (err < 0)
2261				break;
2262			err = put_user(err, &entry->msg_len);
2263			++entry;
2264		}
2265
2266		if (err)
2267			break;
2268		++datagrams;
2269
2270		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2271		if (flags & MSG_WAITFORONE)
2272			flags |= MSG_DONTWAIT;
2273
2274		if (timeout) {
2275			ktime_get_ts(timeout);
2276			*timeout = timespec_sub(end_time, *timeout);
2277			if (timeout->tv_sec < 0) {
2278				timeout->tv_sec = timeout->tv_nsec = 0;
2279				break;
2280			}
2281
2282			/* Timeout, return less than vlen datagrams */
2283			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2284				break;
2285		}
2286
2287		/* Out of band data, return right away */
2288		if (msg_sys.msg_flags & MSG_OOB)
2289			break;
2290	}
2291
2292out_put:
2293	fput_light(sock->file, fput_needed);
2294
2295	if (err == 0)
2296		return datagrams;
2297
2298	if (datagrams != 0) {
2299		/*
2300		 * We may return less entries than requested (vlen) if the
2301		 * sock is non block and there aren't enough datagrams...
2302		 */
2303		if (err != -EAGAIN) {
2304			/*
2305			 * ... or  if recvmsg returns an error after we
2306			 * received some datagrams, where we record the
2307			 * error to return on the next call or if the
2308			 * app asks about it using getsockopt(SO_ERROR).
2309			 */
2310			sock->sk->sk_err = -err;
2311		}
2312
2313		return datagrams;
2314	}
2315
2316	return err;
2317}
2318
2319SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2320		unsigned int, vlen, unsigned int, flags,
2321		struct timespec __user *, timeout)
2322{
2323	int datagrams;
2324	struct timespec timeout_sys;
2325
2326	if (!timeout)
2327		return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2328
2329	if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2330		return -EFAULT;
2331
2332	datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2333
2334	if (datagrams > 0 &&
2335	    copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2336		datagrams = -EFAULT;
2337
2338	return datagrams;
2339}
2340
2341#ifdef __ARCH_WANT_SYS_SOCKETCALL
2342/* Argument list sizes for sys_socketcall */
2343#define AL(x) ((x) * sizeof(unsigned long))
2344static const unsigned char nargs[21] = {
2345	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2346	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2347	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2348	AL(4), AL(5), AL(4)
2349};
2350
2351#undef AL
2352
2353/*
2354 *	System call vectors.
2355 *
2356 *	Argument checking cleaned up. Saved 20% in size.
2357 *  This function doesn't need to set the kernel lock because
2358 *  it is set by the callees.
2359 */
2360
2361SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2362{
2363	unsigned long a[6];
2364	unsigned long a0, a1;
2365	int err;
2366	unsigned int len;
2367
2368	if (call < 1 || call > SYS_SENDMMSG)
2369		return -EINVAL;
2370
2371	len = nargs[call];
2372	if (len > sizeof(a))
2373		return -EINVAL;
2374
2375	/* copy_from_user should be SMP safe. */
2376	if (copy_from_user(a, args, len))
2377		return -EFAULT;
2378
2379	audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2380
2381	a0 = a[0];
2382	a1 = a[1];
2383
2384	switch (call) {
2385	case SYS_SOCKET:
2386		err = sys_socket(a0, a1, a[2]);
2387		break;
2388	case SYS_BIND:
2389		err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2390		break;
2391	case SYS_CONNECT:
2392		err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2393		break;
2394	case SYS_LISTEN:
2395		err = sys_listen(a0, a1);
2396		break;
2397	case SYS_ACCEPT:
2398		err = sys_accept4(a0, (struct sockaddr __user *)a1,
2399				  (int __user *)a[2], 0);
2400		break;
2401	case SYS_GETSOCKNAME:
2402		err =
2403		    sys_getsockname(a0, (struct sockaddr __user *)a1,
2404				    (int __user *)a[2]);
2405		break;
2406	case SYS_GETPEERNAME:
2407		err =
2408		    sys_getpeername(a0, (struct sockaddr __user *)a1,
2409				    (int __user *)a[2]);
2410		break;
2411	case SYS_SOCKETPAIR:
2412		err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2413		break;
2414	case SYS_SEND:
2415		err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2416		break;
2417	case SYS_SENDTO:
2418		err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2419				 (struct sockaddr __user *)a[4], a[5]);
2420		break;
2421	case SYS_RECV:
2422		err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2423		break;
2424	case SYS_RECVFROM:
2425		err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2426				   (struct sockaddr __user *)a[4],
2427				   (int __user *)a[5]);
2428		break;
2429	case SYS_SHUTDOWN:
2430		err = sys_shutdown(a0, a1);
2431		break;
2432	case SYS_SETSOCKOPT:
2433		err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2434		break;
2435	case SYS_GETSOCKOPT:
2436		err =
2437		    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2438				   (int __user *)a[4]);
2439		break;
2440	case SYS_SENDMSG:
2441		err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2442		break;
2443	case SYS_SENDMMSG:
2444		err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2445		break;
2446	case SYS_RECVMSG:
2447		err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2448		break;
2449	case SYS_RECVMMSG:
2450		err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2451				   (struct timespec __user *)a[4]);
2452		break;
2453	case SYS_ACCEPT4:
2454		err = sys_accept4(a0, (struct sockaddr __user *)a1,
2455				  (int __user *)a[2], a[3]);
2456		break;
2457	default:
2458		err = -EINVAL;
2459		break;
2460	}
2461	return err;
2462}
2463
2464#endif				/* __ARCH_WANT_SYS_SOCKETCALL */
2465
2466/**
2467 *	sock_register - add a socket protocol handler
2468 *	@ops: description of protocol
2469 *
2470 *	This function is called by a protocol handler that wants to
2471 *	advertise its address family, and have it linked into the
2472 *	socket interface. The value ops->family coresponds to the
2473 *	socket system call protocol family.
2474 */
2475int sock_register(const struct net_proto_family *ops)
2476{
2477	int err;
2478
2479	if (ops->family >= NPROTO) {
2480		printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2481		       NPROTO);
2482		return -ENOBUFS;
2483	}
2484
2485	spin_lock(&net_family_lock);
2486	if (rcu_dereference_protected(net_families[ops->family],
2487				      lockdep_is_held(&net_family_lock)))
2488		err = -EEXIST;
2489	else {
2490		rcu_assign_pointer(net_families[ops->family], ops);
2491		err = 0;
2492	}
2493	spin_unlock(&net_family_lock);
2494
2495	printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2496	return err;
2497}
2498EXPORT_SYMBOL(sock_register);
2499
2500/**
2501 *	sock_unregister - remove a protocol handler
2502 *	@family: protocol family to remove
2503 *
2504 *	This function is called by a protocol handler that wants to
2505 *	remove its address family, and have it unlinked from the
2506 *	new socket creation.
2507 *
2508 *	If protocol handler is a module, then it can use module reference
2509 *	counts to protect against new references. If protocol handler is not
2510 *	a module then it needs to provide its own protection in
2511 *	the ops->create routine.
2512 */
2513void sock_unregister(int family)
2514{
2515	BUG_ON(family < 0 || family >= NPROTO);
2516
2517	spin_lock(&net_family_lock);
2518	RCU_INIT_POINTER(net_families[family], NULL);
2519	spin_unlock(&net_family_lock);
2520
2521	synchronize_rcu();
2522
2523	printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2524}
2525EXPORT_SYMBOL(sock_unregister);
2526
2527static int __init sock_init(void)
2528{
2529	int err;
2530
2531	/*
2532	 *      Initialize sock SLAB cache.
2533	 */
2534
2535	sk_init();
2536
2537	/*
2538	 *      Initialize skbuff SLAB cache
2539	 */
2540	skb_init();
2541
2542	/*
2543	 *      Initialize the protocols module.
2544	 */
2545
2546	init_inodecache();
2547
2548	err = register_filesystem(&sock_fs_type);
2549	if (err)
2550		goto out_fs;
2551	sock_mnt = kern_mount(&sock_fs_type);
2552	if (IS_ERR(sock_mnt)) {
2553		err = PTR_ERR(sock_mnt);
2554		goto out_mount;
2555	}
2556
2557	/* The real protocol initialization is performed in later initcalls.
2558	 */
2559
2560#ifdef CONFIG_NETFILTER
2561	netfilter_init();
2562#endif
2563
2564#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2565	skb_timestamping_init();
2566#endif
2567
2568out:
2569	return err;
2570
2571out_mount:
2572	unregister_filesystem(&sock_fs_type);
2573out_fs:
2574	goto out;
2575}
2576
2577core_initcall(sock_init);	/* early initcall */
2578
2579#ifdef CONFIG_PROC_FS
2580void socket_seq_show(struct seq_file *seq)
2581{
2582	int cpu;
2583	int counter = 0;
2584
2585	for_each_possible_cpu(cpu)
2586	    counter += per_cpu(sockets_in_use, cpu);
2587
2588	/* It can be negative, by the way. 8) */
2589	if (counter < 0)
2590		counter = 0;
2591
2592	seq_printf(seq, "sockets: used %d\n", counter);
2593}
2594#endif				/* CONFIG_PROC_FS */
2595
2596#ifdef CONFIG_COMPAT
2597static int do_siocgstamp(struct net *net, struct socket *sock,
2598			 unsigned int cmd, void __user *up)
2599{
2600	mm_segment_t old_fs = get_fs();
2601	struct timeval ktv;
2602	int err;
2603
2604	set_fs(KERNEL_DS);
2605	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2606	set_fs(old_fs);
2607	if (!err)
2608		err = compat_put_timeval(&ktv, up);
2609
2610	return err;
2611}
2612
2613static int do_siocgstampns(struct net *net, struct socket *sock,
2614			   unsigned int cmd, void __user *up)
2615{
2616	mm_segment_t old_fs = get_fs();
2617	struct timespec kts;
2618	int err;
2619
2620	set_fs(KERNEL_DS);
2621	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2622	set_fs(old_fs);
2623	if (!err)
2624		err = compat_put_timespec(&kts, up);
2625
2626	return err;
2627}
2628
2629static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2630{
2631	struct ifreq __user *uifr;
2632	int err;
2633
2634	uifr = compat_alloc_user_space(sizeof(struct ifreq));
2635	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2636		return -EFAULT;
2637
2638	err = dev_ioctl(net, SIOCGIFNAME, uifr);
2639	if (err)
2640		return err;
2641
2642	if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2643		return -EFAULT;
2644
2645	return 0;
2646}
2647
2648static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2649{
2650	struct compat_ifconf ifc32;
2651	struct ifconf ifc;
2652	struct ifconf __user *uifc;
2653	struct compat_ifreq __user *ifr32;
2654	struct ifreq __user *ifr;
2655	unsigned int i, j;
2656	int err;
2657
2658	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2659		return -EFAULT;
2660
2661	memset(&ifc, 0, sizeof(ifc));
2662	if (ifc32.ifcbuf == 0) {
2663		ifc32.ifc_len = 0;
2664		ifc.ifc_len = 0;
2665		ifc.ifc_req = NULL;
2666		uifc = compat_alloc_user_space(sizeof(struct ifconf));
2667	} else {
2668		size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2669			sizeof(struct ifreq);
2670		uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2671		ifc.ifc_len = len;
2672		ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2673		ifr32 = compat_ptr(ifc32.ifcbuf);
2674		for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2675			if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2676				return -EFAULT;
2677			ifr++;
2678			ifr32++;
2679		}
2680	}
2681	if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2682		return -EFAULT;
2683
2684	err = dev_ioctl(net, SIOCGIFCONF, uifc);
2685	if (err)
2686		return err;
2687
2688	if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2689		return -EFAULT;
2690
2691	ifr = ifc.ifc_req;
2692	ifr32 = compat_ptr(ifc32.ifcbuf);
2693	for (i = 0, j = 0;
2694	     i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2695	     i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2696		if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2697			return -EFAULT;
2698		ifr32++;
2699		ifr++;
2700	}
2701
2702	if (ifc32.ifcbuf == 0) {
2703		/* Translate from 64-bit structure multiple to
2704		 * a 32-bit one.
2705		 */
2706		i = ifc.ifc_len;
2707		i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2708		ifc32.ifc_len = i;
2709	} else {
2710		ifc32.ifc_len = i;
2711	}
2712	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2713		return -EFAULT;
2714
2715	return 0;
2716}
2717
2718static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2719{
2720	struct compat_ethtool_rxnfc __user *compat_rxnfc;
2721	bool convert_in = false, convert_out = false;
2722	size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2723	struct ethtool_rxnfc __user *rxnfc;
2724	struct ifreq __user *ifr;
2725	u32 rule_cnt = 0, actual_rule_cnt;
2726	u32 ethcmd;
2727	u32 data;
2728	int ret;
2729
2730	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2731		return -EFAULT;
2732
2733	compat_rxnfc = compat_ptr(data);
2734
2735	if (get_user(ethcmd, &compat_rxnfc->cmd))
2736		return -EFAULT;
2737
2738	/* Most ethtool structures are defined without padding.
2739	 * Unfortunately struct ethtool_rxnfc is an exception.
2740	 */
2741	switch (ethcmd) {
2742	default:
2743		break;
2744	case ETHTOOL_GRXCLSRLALL:
2745		/* Buffer size is variable */
2746		if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2747			return -EFAULT;
2748		if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2749			return -ENOMEM;
2750		buf_size += rule_cnt * sizeof(u32);
2751		/* fall through */
2752	case ETHTOOL_GRXRINGS:
2753	case ETHTOOL_GRXCLSRLCNT:
2754	case ETHTOOL_GRXCLSRULE:
2755	case ETHTOOL_SRXCLSRLINS:
2756		convert_out = true;
2757		/* fall through */
2758	case ETHTOOL_SRXCLSRLDEL:
2759		buf_size += sizeof(struct ethtool_rxnfc);
2760		convert_in = true;
2761		break;
2762	}
2763
2764	ifr = compat_alloc_user_space(buf_size);
2765	rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2766
2767	if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2768		return -EFAULT;
2769
2770	if (put_user(convert_in ? rxnfc : compat_ptr(data),
2771		     &ifr->ifr_ifru.ifru_data))
2772		return -EFAULT;
2773
2774	if (convert_in) {
2775		/* We expect there to be holes between fs.m_ext and
2776		 * fs.ring_cookie and at the end of fs, but nowhere else.
2777		 */
2778		BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2779			     sizeof(compat_rxnfc->fs.m_ext) !=
2780			     offsetof(struct ethtool_rxnfc, fs.m_ext) +
2781			     sizeof(rxnfc->fs.m_ext));
2782		BUILD_BUG_ON(
2783			offsetof(struct compat_ethtool_rxnfc, fs.location) -
2784			offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2785			offsetof(struct ethtool_rxnfc, fs.location) -
2786			offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2787
2788		if (copy_in_user(rxnfc, compat_rxnfc,
2789				 (void *)(&rxnfc->fs.m_ext + 1) -
2790				 (void *)rxnfc) ||
2791		    copy_in_user(&rxnfc->fs.ring_cookie,
2792				 &compat_rxnfc->fs.ring_cookie,
2793				 (void *)(&rxnfc->fs.location + 1) -
2794				 (void *)&rxnfc->fs.ring_cookie) ||
2795		    copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2796				 sizeof(rxnfc->rule_cnt)))
2797			return -EFAULT;
2798	}
2799
2800	ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2801	if (ret)
2802		return ret;
2803
2804	if (convert_out) {
2805		if (copy_in_user(compat_rxnfc, rxnfc,
2806				 (const void *)(&rxnfc->fs.m_ext + 1) -
2807				 (const void *)rxnfc) ||
2808		    copy_in_user(&compat_rxnfc->fs.ring_cookie,
2809				 &rxnfc->fs.ring_cookie,
2810				 (const void *)(&rxnfc->fs.location + 1) -
2811				 (const void *)&rxnfc->fs.ring_cookie) ||
2812		    copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2813				 sizeof(rxnfc->rule_cnt)))
2814			return -EFAULT;
2815
2816		if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2817			/* As an optimisation, we only copy the actual
2818			 * number of rules that the underlying
2819			 * function returned.  Since Mallory might
2820			 * change the rule count in user memory, we
2821			 * check that it is less than the rule count
2822			 * originally given (as the user buffer size),
2823			 * which has been range-checked.
2824			 */
2825			if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2826				return -EFAULT;
2827			if (actual_rule_cnt < rule_cnt)
2828				rule_cnt = actual_rule_cnt;
2829			if (copy_in_user(&compat_rxnfc->rule_locs[0],
2830					 &rxnfc->rule_locs[0],
2831					 rule_cnt * sizeof(u32)))
2832				return -EFAULT;
2833		}
2834	}
2835
2836	return 0;
2837}
2838
2839static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2840{
2841	void __user *uptr;
2842	compat_uptr_t uptr32;
2843	struct ifreq __user *uifr;
2844
2845	uifr = compat_alloc_user_space(sizeof(*uifr));
2846	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2847		return -EFAULT;
2848
2849	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2850		return -EFAULT;
2851
2852	uptr = compat_ptr(uptr32);
2853
2854	if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2855		return -EFAULT;
2856
2857	return dev_ioctl(net, SIOCWANDEV, uifr);
2858}
2859
2860static int bond_ioctl(struct net *net, unsigned int cmd,
2861			 struct compat_ifreq __user *ifr32)
2862{
2863	struct ifreq kifr;
2864	struct ifreq __user *uifr;
2865	mm_segment_t old_fs;
2866	int err;
2867	u32 data;
2868	void __user *datap;
2869
2870	switch (cmd) {
2871	case SIOCBONDENSLAVE:
2872	case SIOCBONDRELEASE:
2873	case SIOCBONDSETHWADDR:
2874	case SIOCBONDCHANGEACTIVE:
2875		if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2876			return -EFAULT;
2877
2878		old_fs = get_fs();
2879		set_fs(KERNEL_DS);
2880		err = dev_ioctl(net, cmd,
2881				(struct ifreq __user __force *) &kifr);
2882		set_fs(old_fs);
2883
2884		return err;
2885	case SIOCBONDSLAVEINFOQUERY:
2886	case SIOCBONDINFOQUERY:
2887		uifr = compat_alloc_user_space(sizeof(*uifr));
2888		if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2889			return -EFAULT;
2890
2891		if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2892			return -EFAULT;
2893
2894		datap = compat_ptr(data);
2895		if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2896			return -EFAULT;
2897
2898		return dev_ioctl(net, cmd, uifr);
2899	default:
2900		return -ENOIOCTLCMD;
2901	}
2902}
2903
2904static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2905				 struct compat_ifreq __user *u_ifreq32)
2906{
2907	struct ifreq __user *u_ifreq64;
2908	char tmp_buf[IFNAMSIZ];
2909	void __user *data64;
2910	u32 data32;
2911
2912	if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2913			   IFNAMSIZ))
2914		return -EFAULT;
2915	if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2916		return -EFAULT;
2917	data64 = compat_ptr(data32);
2918
2919	u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2920
2921	/* Don't check these user accesses, just let that get trapped
2922	 * in the ioctl handler instead.
2923	 */
2924	if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2925			 IFNAMSIZ))
2926		return -EFAULT;
2927	if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2928		return -EFAULT;
2929
2930	return dev_ioctl(net, cmd, u_ifreq64);
2931}
2932
2933static int dev_ifsioc(struct net *net, struct socket *sock,
2934			 unsigned int cmd, struct compat_ifreq __user *uifr32)
2935{
2936	struct ifreq __user *uifr;
2937	int err;
2938
2939	uifr = compat_alloc_user_space(sizeof(*uifr));
2940	if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2941		return -EFAULT;
2942
2943	err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2944
2945	if (!err) {
2946		switch (cmd) {
2947		case SIOCGIFFLAGS:
2948		case SIOCGIFMETRIC:
2949		case SIOCGIFMTU:
2950		case SIOCGIFMEM:
2951		case SIOCGIFHWADDR:
2952		case SIOCGIFINDEX:
2953		case SIOCGIFADDR:
2954		case SIOCGIFBRDADDR:
2955		case SIOCGIFDSTADDR:
2956		case SIOCGIFNETMASK:
2957		case SIOCGIFPFLAGS:
2958		case SIOCGIFTXQLEN:
2959		case SIOCGMIIPHY:
2960		case SIOCGMIIREG:
2961			if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2962				err = -EFAULT;
2963			break;
2964		}
2965	}
2966	return err;
2967}
2968
2969static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2970			struct compat_ifreq __user *uifr32)
2971{
2972	struct ifreq ifr;
2973	struct compat_ifmap __user *uifmap32;
2974	mm_segment_t old_fs;
2975	int err;
2976
2977	uifmap32 = &uifr32->ifr_ifru.ifru_map;
2978	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2979	err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2980	err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2981	err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2982	err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2983	err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2984	err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2985	if (err)
2986		return -EFAULT;
2987
2988	old_fs = get_fs();
2989	set_fs(KERNEL_DS);
2990	err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
2991	set_fs(old_fs);
2992
2993	if (cmd == SIOCGIFMAP && !err) {
2994		err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2995		err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2996		err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2997		err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2998		err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2999		err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3000		err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3001		if (err)
3002			err = -EFAULT;
3003	}
3004	return err;
3005}
3006
3007static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3008{
3009	void __user *uptr;
3010	compat_uptr_t uptr32;
3011	struct ifreq __user *uifr;
3012
3013	uifr = compat_alloc_user_space(sizeof(*uifr));
3014	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3015		return -EFAULT;
3016
3017	if (get_user(uptr32, &uifr32->ifr_data))
3018		return -EFAULT;
3019
3020	uptr = compat_ptr(uptr32);
3021
3022	if (put_user(uptr, &uifr->ifr_data))
3023		return -EFAULT;
3024
3025	return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3026}
3027
3028struct rtentry32 {
3029	u32		rt_pad1;
3030	struct sockaddr rt_dst;         /* target address               */
3031	struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3032	struct sockaddr rt_genmask;     /* target network mask (IP)     */
3033	unsigned short	rt_flags;
3034	short		rt_pad2;
3035	u32		rt_pad3;
3036	unsigned char	rt_tos;
3037	unsigned char	rt_class;
3038	short		rt_pad4;
3039	short		rt_metric;      /* +1 for binary compatibility! */
3040	/* char * */ u32 rt_dev;        /* forcing the device at add    */
3041	u32		rt_mtu;         /* per route MTU/Window         */
3042	u32		rt_window;      /* Window clamping              */
3043	unsigned short  rt_irtt;        /* Initial RTT                  */
3044};
3045
3046struct in6_rtmsg32 {
3047	struct in6_addr		rtmsg_dst;
3048	struct in6_addr		rtmsg_src;
3049	struct in6_addr		rtmsg_gateway;
3050	u32			rtmsg_type;
3051	u16			rtmsg_dst_len;
3052	u16			rtmsg_src_len;
3053	u32			rtmsg_metric;
3054	u32			rtmsg_info;
3055	u32			rtmsg_flags;
3056	s32			rtmsg_ifindex;
3057};
3058
3059static int routing_ioctl(struct net *net, struct socket *sock,
3060			 unsigned int cmd, void __user *argp)
3061{
3062	int ret;
3063	void *r = NULL;
3064	struct in6_rtmsg r6;
3065	struct rtentry r4;
3066	char devname[16];
3067	u32 rtdev;
3068	mm_segment_t old_fs = get_fs();
3069
3070	if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3071		struct in6_rtmsg32 __user *ur6 = argp;
3072		ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3073			3 * sizeof(struct in6_addr));
3074		ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3075		ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3076		ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3077		ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3078		ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3079		ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3080		ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3081
3082		r = (void *) &r6;
3083	} else { /* ipv4 */
3084		struct rtentry32 __user *ur4 = argp;
3085		ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3086					3 * sizeof(struct sockaddr));
3087		ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3088		ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3089		ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3090		ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3091		ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3092		ret |= __get_user(rtdev, &(ur4->rt_dev));
3093		if (rtdev) {
3094			ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3095			r4.rt_dev = (char __user __force *)devname;
3096			devname[15] = 0;
3097		} else
3098			r4.rt_dev = NULL;
3099
3100		r = (void *) &r4;
3101	}
3102
3103	if (ret) {
3104		ret = -EFAULT;
3105		goto out;
3106	}
3107
3108	set_fs(KERNEL_DS);
3109	ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3110	set_fs(old_fs);
3111
3112out:
3113	return ret;
3114}
3115
3116/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3117 * for some operations; this forces use of the newer bridge-utils that
3118 * use compatible ioctls
3119 */
3120static int old_bridge_ioctl(compat_ulong_t __user *argp)
3121{
3122	compat_ulong_t tmp;
3123
3124	if (get_user(tmp, argp))
3125		return -EFAULT;
3126	if (tmp == BRCTL_GET_VERSION)
3127		return BRCTL_VERSION + 1;
3128	return -EINVAL;
3129}
3130
3131static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3132			 unsigned int cmd, unsigned long arg)
3133{
3134	void __user *argp = compat_ptr(arg);
3135	struct sock *sk = sock->sk;
3136	struct net *net = sock_net(sk);
3137
3138	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3139		return siocdevprivate_ioctl(net, cmd, argp);
3140
3141	switch (cmd) {
3142	case SIOCSIFBR:
3143	case SIOCGIFBR:
3144		return old_bridge_ioctl(argp);
3145	case SIOCGIFNAME:
3146		return dev_ifname32(net, argp);
3147	case SIOCGIFCONF:
3148		return dev_ifconf(net, argp);
3149	case SIOCETHTOOL:
3150		return ethtool_ioctl(net, argp);
3151	case SIOCWANDEV:
3152		return compat_siocwandev(net, argp);
3153	case SIOCGIFMAP:
3154	case SIOCSIFMAP:
3155		return compat_sioc_ifmap(net, cmd, argp);
3156	case SIOCBONDENSLAVE:
3157	case SIOCBONDRELEASE:
3158	case SIOCBONDSETHWADDR:
3159	case SIOCBONDSLAVEINFOQUERY:
3160	case SIOCBONDINFOQUERY:
3161	case SIOCBONDCHANGEACTIVE:
3162		return bond_ioctl(net, cmd, argp);
3163	case SIOCADDRT:
3164	case SIOCDELRT:
3165		return routing_ioctl(net, sock, cmd, argp);
3166	case SIOCGSTAMP:
3167		return do_siocgstamp(net, sock, cmd, argp);
3168	case SIOCGSTAMPNS:
3169		return do_siocgstampns(net, sock, cmd, argp);
3170	case SIOCSHWTSTAMP:
3171		return compat_siocshwtstamp(net, argp);
3172
3173	case FIOSETOWN:
3174	case SIOCSPGRP:
3175	case FIOGETOWN:
3176	case SIOCGPGRP:
3177	case SIOCBRADDBR:
3178	case SIOCBRDELBR:
3179	case SIOCGIFVLAN:
3180	case SIOCSIFVLAN:
3181	case SIOCADDDLCI:
3182	case SIOCDELDLCI:
3183		return sock_ioctl(file, cmd, arg);
3184
3185	case SIOCGIFFLAGS:
3186	case SIOCSIFFLAGS:
3187	case SIOCGIFMETRIC:
3188	case SIOCSIFMETRIC:
3189	case SIOCGIFMTU:
3190	case SIOCSIFMTU:
3191	case SIOCGIFMEM:
3192	case SIOCSIFMEM:
3193	case SIOCGIFHWADDR:
3194	case SIOCSIFHWADDR:
3195	case SIOCADDMULTI:
3196	case SIOCDELMULTI:
3197	case SIOCGIFINDEX:
3198	case SIOCGIFADDR:
3199	case SIOCSIFADDR:
3200	case SIOCSIFHWBROADCAST:
3201	case SIOCDIFADDR:
3202	case SIOCGIFBRDADDR:
3203	case SIOCSIFBRDADDR:
3204	case SIOCGIFDSTADDR:
3205	case SIOCSIFDSTADDR:
3206	case SIOCGIFNETMASK:
3207	case SIOCSIFNETMASK:
3208	case SIOCSIFPFLAGS:
3209	case SIOCGIFPFLAGS:
3210	case SIOCGIFTXQLEN:
3211	case SIOCSIFTXQLEN:
3212	case SIOCBRADDIF:
3213	case SIOCBRDELIF:
3214	case SIOCSIFNAME:
3215	case SIOCGMIIPHY:
3216	case SIOCGMIIREG:
3217	case SIOCSMIIREG:
3218		return dev_ifsioc(net, sock, cmd, argp);
3219
3220	case SIOCSARP:
3221	case SIOCGARP:
3222	case SIOCDARP:
3223	case SIOCATMARK:
3224		return sock_do_ioctl(net, sock, cmd, arg);
3225	}
3226
3227	return -ENOIOCTLCMD;
3228}
3229
3230static long compat_sock_ioctl(struct file *file, unsigned cmd,
3231			      unsigned long arg)
3232{
3233	struct socket *sock = file->private_data;
3234	int ret = -ENOIOCTLCMD;
3235	struct sock *sk;
3236	struct net *net;
3237
3238	sk = sock->sk;
3239	net = sock_net(sk);
3240
3241	if (sock->ops->compat_ioctl)
3242		ret = sock->ops->compat_ioctl(sock, cmd, arg);
3243
3244	if (ret == -ENOIOCTLCMD &&
3245	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3246		ret = compat_wext_handle_ioctl(net, cmd, arg);
3247
3248	if (ret == -ENOIOCTLCMD)
3249		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3250
3251	return ret;
3252}
3253#endif
3254
3255int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3256{
3257	return sock->ops->bind(sock, addr, addrlen);
3258}
3259EXPORT_SYMBOL(kernel_bind);
3260
3261int kernel_listen(struct socket *sock, int backlog)
3262{
3263	return sock->ops->listen(sock, backlog);
3264}
3265EXPORT_SYMBOL(kernel_listen);
3266
3267int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3268{
3269	struct sock *sk = sock->sk;
3270	int err;
3271
3272	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3273			       newsock);
3274	if (err < 0)
3275		goto done;
3276
3277	err = sock->ops->accept(sock, *newsock, flags);
3278	if (err < 0) {
3279		sock_release(*newsock);
3280		*newsock = NULL;
3281		goto done;
3282	}
3283
3284	(*newsock)->ops = sock->ops;
3285	__module_get((*newsock)->ops->owner);
3286
3287done:
3288	return err;
3289}
3290EXPORT_SYMBOL(kernel_accept);
3291
3292int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3293		   int flags)
3294{
3295	return sock->ops->connect(sock, addr, addrlen, flags);
3296}
3297EXPORT_SYMBOL(kernel_connect);
3298
3299int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3300			 int *addrlen)
3301{
3302	return sock->ops->getname(sock, addr, addrlen, 0);
3303}
3304EXPORT_SYMBOL(kernel_getsockname);
3305
3306int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3307			 int *addrlen)
3308{
3309	return sock->ops->getname(sock, addr, addrlen, 1);
3310}
3311EXPORT_SYMBOL(kernel_getpeername);
3312
3313int kernel_getsockopt(struct socket *sock, int level, int optname,
3314			char *optval, int *optlen)
3315{
3316	mm_segment_t oldfs = get_fs();
3317	char __user *uoptval;
3318	int __user *uoptlen;
3319	int err;
3320
3321	uoptval = (char __user __force *) optval;
3322	uoptlen = (int __user __force *) optlen;
3323
3324	set_fs(KERNEL_DS);
3325	if (level == SOL_SOCKET)
3326		err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3327	else
3328		err = sock->ops->getsockopt(sock, level, optname, uoptval,
3329					    uoptlen);
3330	set_fs(oldfs);
3331	return err;
3332}
3333EXPORT_SYMBOL(kernel_getsockopt);
3334
3335int kernel_setsockopt(struct socket *sock, int level, int optname,
3336			char *optval, unsigned int optlen)
3337{
3338	mm_segment_t oldfs = get_fs();
3339	char __user *uoptval;
3340	int err;
3341
3342	uoptval = (char __user __force *) optval;
3343
3344	set_fs(KERNEL_DS);
3345	if (level == SOL_SOCKET)
3346		err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3347	else
3348		err = sock->ops->setsockopt(sock, level, optname, uoptval,
3349					    optlen);
3350	set_fs(oldfs);
3351	return err;
3352}
3353EXPORT_SYMBOL(kernel_setsockopt);
3354
3355int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3356		    size_t size, int flags)
3357{
3358	sock_update_classid(sock->sk);
3359
3360	if (sock->ops->sendpage)
3361		return sock->ops->sendpage(sock, page, offset, size, flags);
3362
3363	return sock_no_sendpage(sock, page, offset, size, flags);
3364}
3365EXPORT_SYMBOL(kernel_sendpage);
3366
3367int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3368{
3369	mm_segment_t oldfs = get_fs();
3370	int err;
3371
3372	set_fs(KERNEL_DS);
3373	err = sock->ops->ioctl(sock, cmd, arg);
3374	set_fs(oldfs);
3375
3376	return err;
3377}
3378EXPORT_SYMBOL(kernel_sock_ioctl);
3379
3380int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3381{
3382	return sock->ops->shutdown(sock, how);
3383}
3384EXPORT_SYMBOL(kernel_sock_shutdown);