PageRenderTime 124ms CodeModel.GetById 4ms app.highlight 105ms RepoModel.GetById 1ms app.codeStats 1ms

/net/socket.c

https://bitbucket.org/codefirex/kernel_lge_gee
C | 3384 lines | 2426 code | 494 blank | 464 comment | 381 complexity | 93dd3d898071436bd4987eeea5a7ff22 MD5 | raw file

Large files files are truncated, but you can click here to view the full 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	percpu_sub(sockets_in_use, 1);
 526	if (!sock->file) {
 527		iput(SOCK_INODE(sock));
 528		return;
 529	}
 530	sock->file = NULL;
 531}
 532EXPORT_SYMBOL(sock_release);
 533
 534int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 535{
 536	*tx_flags = 0;
 537	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 538		*tx_flags |= SKBTX_HW_TSTAMP;
 539	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 540		*tx_flags |= SKBTX_SW_TSTAMP;
 541	if (sock_flag(sk, SOCK_WIFI_STATUS))
 542		*tx_flags |= SKBTX_WIFI_STATUS;
 543	return 0;
 544}
 545EXPORT_SYMBOL(sock_tx_timestamp);
 546
 547static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 548				       struct msghdr *msg, size_t size)
 549{
 550	struct sock_iocb *si = kiocb_to_siocb(iocb);
 551
 552	sock_update_classid(sock->sk);
 553
 554	sock_update_netprioidx(sock->sk);
 555
 556	si->sock = sock;
 557	si->scm = NULL;
 558	si->msg = msg;
 559	si->size = size;
 560
 561	return sock->ops->sendmsg(iocb, sock, msg, size);
 562}
 563
 564static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 565				 struct msghdr *msg, size_t size)
 566{
 567	int err = security_socket_sendmsg(sock, msg, size);
 568
 569	return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 570}
 571
 572int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 573{
 574	struct kiocb iocb;
 575	struct sock_iocb siocb;
 576	int ret;
 577
 578	init_sync_kiocb(&iocb, NULL);
 579	iocb.private = &siocb;
 580	ret = __sock_sendmsg(&iocb, sock, msg, size);
 581	if (-EIOCBQUEUED == ret)
 582		ret = wait_on_sync_kiocb(&iocb);
 583	return ret;
 584}
 585EXPORT_SYMBOL(sock_sendmsg);
 586
 587static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 588{
 589	struct kiocb iocb;
 590	struct sock_iocb siocb;
 591	int ret;
 592
 593	init_sync_kiocb(&iocb, NULL);
 594	iocb.private = &siocb;
 595	ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 596	if (-EIOCBQUEUED == ret)
 597		ret = wait_on_sync_kiocb(&iocb);
 598	return ret;
 599}
 600
 601int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 602		   struct kvec *vec, size_t num, size_t size)
 603{
 604	mm_segment_t oldfs = get_fs();
 605	int result;
 606
 607	set_fs(KERNEL_DS);
 608	/*
 609	 * the following is safe, since for compiler definitions of kvec and
 610	 * iovec are identical, yielding the same in-core layout and alignment
 611	 */
 612	msg->msg_iov = (struct iovec *)vec;
 613	msg->msg_iovlen = num;
 614	result = sock_sendmsg(sock, msg, size);
 615	set_fs(oldfs);
 616	return result;
 617}
 618EXPORT_SYMBOL(kernel_sendmsg);
 619
 620static int ktime2ts(ktime_t kt, struct timespec *ts)
 621{
 622	if (kt.tv64) {
 623		*ts = ktime_to_timespec(kt);
 624		return 1;
 625	} else {
 626		return 0;
 627	}
 628}
 629
 630/*
 631 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 632 */
 633void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 634	struct sk_buff *skb)
 635{
 636	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 637	struct timespec ts[3];
 638	int empty = 1;
 639	struct skb_shared_hwtstamps *shhwtstamps =
 640		skb_hwtstamps(skb);
 641
 642	/* Race occurred between timestamp enabling and packet
 643	   receiving.  Fill in the current time for now. */
 644	if (need_software_tstamp && skb->tstamp.tv64 == 0)
 645		__net_timestamp(skb);
 646
 647	if (need_software_tstamp) {
 648		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 649			struct timeval tv;
 650			skb_get_timestamp(skb, &tv);
 651			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 652				 sizeof(tv), &tv);
 653		} else {
 654			skb_get_timestampns(skb, &ts[0]);
 655			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 656				 sizeof(ts[0]), &ts[0]);
 657		}
 658	}
 659
 660
 661	memset(ts, 0, sizeof(ts));
 662	if (skb->tstamp.tv64 &&
 663	    sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 664		skb_get_timestampns(skb, ts + 0);
 665		empty = 0;
 666	}
 667	if (shhwtstamps) {
 668		if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 669		    ktime2ts(shhwtstamps->syststamp, ts + 1))
 670			empty = 0;
 671		if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 672		    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 673			empty = 0;
 674	}
 675	if (!empty)
 676		put_cmsg(msg, SOL_SOCKET,
 677			 SCM_TIMESTAMPING, sizeof(ts), &ts);
 678}
 679EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 680
 681void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
 682	struct sk_buff *skb)
 683{
 684	int ack;
 685
 686	if (!sock_flag(sk, SOCK_WIFI_STATUS))
 687		return;
 688	if (!skb->wifi_acked_valid)
 689		return;
 690
 691	ack = skb->wifi_acked;
 692
 693	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
 694}
 695EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
 696
 697static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 698				   struct sk_buff *skb)
 699{
 700	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 701		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 702			sizeof(__u32), &skb->dropcount);
 703}
 704
 705void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 706	struct sk_buff *skb)
 707{
 708	sock_recv_timestamp(msg, sk, skb);
 709	sock_recv_drops(msg, sk, skb);
 710}
 711EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 712
 713static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 714				       struct msghdr *msg, size_t size, int flags)
 715{
 716	struct sock_iocb *si = kiocb_to_siocb(iocb);
 717
 718	sock_update_classid(sock->sk);
 719
 720	si->sock = sock;
 721	si->scm = NULL;
 722	si->msg = msg;
 723	si->size = size;
 724	si->flags = flags;
 725
 726	return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 727}
 728
 729static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 730				 struct msghdr *msg, size_t size, int flags)
 731{
 732	int err = security_socket_recvmsg(sock, msg, size, flags);
 733
 734	return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 735}
 736
 737int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 738		 size_t size, int flags)
 739{
 740	struct kiocb iocb;
 741	struct sock_iocb siocb;
 742	int ret;
 743
 744	init_sync_kiocb(&iocb, NULL);
 745	iocb.private = &siocb;
 746	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 747	if (-EIOCBQUEUED == ret)
 748		ret = wait_on_sync_kiocb(&iocb);
 749	return ret;
 750}
 751EXPORT_SYMBOL(sock_recvmsg);
 752
 753static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 754			      size_t size, int flags)
 755{
 756	struct kiocb iocb;
 757	struct sock_iocb siocb;
 758	int ret;
 759
 760	init_sync_kiocb(&iocb, NULL);
 761	iocb.private = &siocb;
 762	ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 763	if (-EIOCBQUEUED == ret)
 764		ret = wait_on_sync_kiocb(&iocb);
 765	return ret;
 766}
 767
 768/**
 769 * kernel_recvmsg - Receive a message from a socket (kernel space)
 770 * @sock:       The socket to receive the message from
 771 * @msg:        Received message
 772 * @vec:        Input s/g array for message data
 773 * @num:        Size of input s/g array
 774 * @size:       Number of bytes to read
 775 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 776 *
 777 * On return the msg structure contains the scatter/gather array passed in the
 778 * vec argument. The array is modified so that it consists of the unfilled
 779 * portion of the original array.
 780 *
 781 * The returned value is the total number of bytes received, or an error.
 782 */
 783int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 784		   struct kvec *vec, size_t num, size_t size, int flags)
 785{
 786	mm_segment_t oldfs = get_fs();
 787	int result;
 788
 789	set_fs(KERNEL_DS);
 790	/*
 791	 * the following is safe, since for compiler definitions of kvec and
 792	 * iovec are identical, yielding the same in-core layout and alignment
 793	 */
 794	msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 795	result = sock_recvmsg(sock, msg, size, flags);
 796	set_fs(oldfs);
 797	return result;
 798}
 799EXPORT_SYMBOL(kernel_recvmsg);
 800
 801static void sock_aio_dtor(struct kiocb *iocb)
 802{
 803	kfree(iocb->private);
 804}
 805
 806static ssize_t sock_sendpage(struct file *file, struct page *page,
 807			     int offset, size_t size, loff_t *ppos, int more)
 808{
 809	struct socket *sock;
 810	int flags;
 811
 812	sock = file->private_data;
 813
 814	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 815	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 816	flags |= more;
 817
 818	return kernel_sendpage(sock, page, offset, size, flags);
 819}
 820
 821static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 822				struct pipe_inode_info *pipe, size_t len,
 823				unsigned int flags)
 824{
 825	struct socket *sock = file->private_data;
 826
 827	if (unlikely(!sock->ops->splice_read))
 828		return -EINVAL;
 829
 830	sock_update_classid(sock->sk);
 831
 832	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 833}
 834
 835static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 836					 struct sock_iocb *siocb)
 837{
 838	if (!is_sync_kiocb(iocb)) {
 839		siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 840		if (!siocb)
 841			return NULL;
 842		iocb->ki_dtor = sock_aio_dtor;
 843	}
 844
 845	siocb->kiocb = iocb;
 846	iocb->private = siocb;
 847	return siocb;
 848}
 849
 850static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 851		struct file *file, const struct iovec *iov,
 852		unsigned long nr_segs)
 853{
 854	struct socket *sock = file->private_data;
 855	size_t size = 0;
 856	int i;
 857
 858	for (i = 0; i < nr_segs; i++)
 859		size += iov[i].iov_len;
 860
 861	msg->msg_name = NULL;
 862	msg->msg_namelen = 0;
 863	msg->msg_control = NULL;
 864	msg->msg_controllen = 0;
 865	msg->msg_iov = (struct iovec *)iov;
 866	msg->msg_iovlen = nr_segs;
 867	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 868
 869	return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 870}
 871
 872static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 873				unsigned long nr_segs, loff_t pos)
 874{
 875	struct sock_iocb siocb, *x;
 876
 877	if (pos != 0)
 878		return -ESPIPE;
 879
 880	if (iocb->ki_left == 0)	/* Match SYS5 behaviour */
 881		return 0;
 882
 883
 884	x = alloc_sock_iocb(iocb, &siocb);
 885	if (!x)
 886		return -ENOMEM;
 887	return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 888}
 889
 890static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 891			struct file *file, const struct iovec *iov,
 892			unsigned long nr_segs)
 893{
 894	struct socket *sock = file->private_data;
 895	size_t size = 0;
 896	int i;
 897
 898	for (i = 0; i < nr_segs; i++)
 899		size += iov[i].iov_len;
 900
 901	msg->msg_name = NULL;
 902	msg->msg_namelen = 0;
 903	msg->msg_control = NULL;
 904	msg->msg_controllen = 0;
 905	msg->msg_iov = (struct iovec *)iov;
 906	msg->msg_iovlen = nr_segs;
 907	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 908	if (sock->type == SOCK_SEQPACKET)
 909		msg->msg_flags |= MSG_EOR;
 910
 911	return __sock_sendmsg(iocb, sock, msg, size);
 912}
 913
 914static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 915			  unsigned long nr_segs, loff_t pos)
 916{
 917	struct sock_iocb siocb, *x;
 918
 919	if (pos != 0)
 920		return -ESPIPE;
 921
 922	x = alloc_sock_iocb(iocb, &siocb);
 923	if (!x)
 924		return -ENOMEM;
 925
 926	return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 927}
 928
 929/*
 930 * Atomic setting of ioctl hooks to avoid race
 931 * with module unload.
 932 */
 933
 934static DEFINE_MUTEX(br_ioctl_mutex);
 935static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 936
 937void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 938{
 939	mutex_lock(&br_ioctl_mutex);
 940	br_ioctl_hook = hook;
 941	mutex_unlock(&br_ioctl_mutex);
 942}
 943EXPORT_SYMBOL(brioctl_set);
 944
 945static DEFINE_MUTEX(vlan_ioctl_mutex);
 946static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 947
 948void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 949{
 950	mutex_lock(&vlan_ioctl_mutex);
 951	vlan_ioctl_hook = hook;
 952	mutex_unlock(&vlan_ioctl_mutex);
 953}
 954EXPORT_SYMBOL(vlan_ioctl_set);
 955
 956static DEFINE_MUTEX(dlci_ioctl_mutex);
 957static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 958
 959void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 960{
 961	mutex_lock(&dlci_ioctl_mutex);
 962	dlci_ioctl_hook = hook;
 963	mutex_unlock(&dlci_ioctl_mutex);
 964}
 965EXPORT_SYMBOL(dlci_ioctl_set);
 966
 967static long sock_do_ioctl(struct net *net, struct socket *sock,
 968				 unsigned int cmd, unsigned long arg)
 969{
 970	int err;
 971	void __user *argp = (void __user *)arg;
 972
 973	err = sock->ops->ioctl(sock, cmd, arg);
 974
 975	/*
 976	 * If this ioctl is unknown try to hand it down
 977	 * to the NIC driver.
 978	 */
 979	if (err == -ENOIOCTLCMD)
 980		err = dev_ioctl(net, cmd, argp);
 981
 982	return err;
 983}
 984
 985/*
 986 *	With an ioctl, arg may well be a user mode pointer, but we don't know
 987 *	what to do with it - that's up to the protocol still.
 988 */
 989
 990static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 991{
 992	struct socket *sock;
 993	struct sock *sk;
 994	void __user *argp = (void __user *)arg;
 995	int pid, err;
 996	struct net *net;
 997
 998	sock = file->private_data;
 999	sk = sock->sk;
1000	net = sock_net(sk);
1001	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1002		err = dev_ioctl(net, cmd, argp);
1003	} else
1004#ifdef CONFIG_WEXT_CORE
1005	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1006		err = dev_ioctl(net, cmd, argp);
1007	} else
1008#endif
1009		switch (cmd) {
1010		case FIOSETOWN:
1011		case SIOCSPGRP:
1012			err = -EFAULT;
1013			if (get_user(pid, (int __user *)argp))
1014				break;
1015			err = f_setown(sock->file, pid, 1);
1016			break;
1017		case FIOGETOWN:
1018		case SIOCGPGRP:
1019			err = put_user(f_getown(sock->file),
1020				       (int __user *)argp);
1021			break;
1022		case SIOCGIFBR:
1023		case SIOCSIFBR:
1024		case SIOCBRADDBR:
1025		case SIOCBRDELBR:
1026			err = -ENOPKG;
1027			if (!br_ioctl_hook)
1028				request_module("bridge");
1029
1030			mutex_lock(&br_ioctl_mutex);
1031			if (br_ioctl_hook)
1032				err = br_ioctl_hook(net, cmd, argp);
1033			mutex_unlock(&br_ioctl_mutex);
1034			break;
1035		case SIOCGIFVLAN:
1036		case SIOCSIFVLAN:
1037			err = -ENOPKG;
1038			if (!vlan_ioctl_hook)
1039				request_module("8021q");
1040
1041			mutex_lock(&vlan_ioctl_mutex);
1042			if (vlan_ioctl_hook)
1043				err = vlan_ioctl_hook(net, argp);
1044			mutex_unlock(&vlan_ioctl_mutex);
1045			break;
1046		case SIOCADDDLCI:
1047		case SIOCDELDLCI:
1048			err = -ENOPKG;
1049			if (!dlci_ioctl_hook)
1050				request_module("dlci");
1051
1052			mutex_lock(&dlci_ioctl_mutex);
1053			if (dlci_ioctl_hook)
1054				err = dlci_ioctl_hook(cmd, argp);
1055			mutex_unlock(&dlci_ioctl_mutex);
1056			break;
1057		default:
1058			err = sock_do_ioctl(net, sock, cmd, arg);
1059			break;
1060		}
1061	return err;
1062}
1063
1064int sock_create_lite(int family, int type, int protocol, struct socket **res)
1065{
1066	int err;
1067	struct socket *sock = NULL;
1068
1069	err = security_socket_create(family, type, protocol, 1);
1070	if (err)
1071		goto out;
1072
1073	sock = sock_alloc();
1074	if (!sock) {
1075		err = -ENOMEM;
1076		goto out;
1077	}
1078
1079	sock->type = type;
1080	err = security_socket_post_create(sock, family, type, protocol, 1);
1081	if (err)
1082		goto out_release;
1083
1084out:
1085	*res = sock;
1086	return err;
1087out_release:
1088	sock_release(sock);
1089	sock = NULL;
1090	goto out;
1091}
1092EXPORT_SYMBOL(sock_create_lite);
1093
1094/* No kernel lock held - perfect */
1095static unsigned int sock_poll(struct file *file, poll_table *wait)
1096{
1097	struct socket *sock;
1098
1099	/*
1100	 *      We can't return errors to poll, so it's either yes or no.
1101	 */
1102	sock = file->private_data;
1103	return sock->ops->poll(file, sock, wait);
1104}
1105
1106static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1107{
1108	struct socket *sock = file->private_data;
1109
1110	return sock->ops->mmap(file, sock, vma);
1111}
1112
1113static int sock_close(struct inode *inode, struct file *filp)
1114{
1115	/*
1116	 *      It was possible the inode is NULL we were
1117	 *      closing an unfinished socket.
1118	 */
1119
1120	if (!inode) {
1121		printk(KERN_DEBUG "sock_close: NULL inode\n");
1122		return 0;
1123	}
1124	sock_release(SOCKET_I(inode));
1125	return 0;
1126}
1127
1128/*
1129 *	Update the socket async list
1130 *
1131 *	Fasync_list locking strategy.
1132 *
1133 *	1. fasync_list is modified only under process context socket lock
1134 *	   i.e. under semaphore.
1135 *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1136 *	   or under socket lock
1137 */
1138
1139static int sock_fasync(int fd, struct file *filp, int on)
1140{
1141	struct socket *sock = filp->private_data;
1142	struct sock *sk = sock->sk;
1143	struct socket_wq *wq;
1144
1145	if (sk == NULL)
1146		return -EINVAL;
1147
1148	lock_sock(sk);
1149	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1150	fasync_helper(fd, filp, on, &wq->fasync_list);
1151
1152	if (!wq->fasync_list)
1153		sock_reset_flag(sk, SOCK_FASYNC);
1154	else
1155		sock_set_flag(sk, SOCK_FASYNC);
1156
1157	release_sock(sk);
1158	return 0;
1159}
1160
1161/* This function may be called only under socket lock or callback_lock or rcu_lock */
1162
1163int sock_wake_async(struct socket *sock, int how, int band)
1164{
1165	struct socket_wq *wq;
1166
1167	if (!sock)
1168		return -1;
1169	rcu_read_lock();
1170	wq = rcu_dereference(sock->wq);
1171	if (!wq || !wq->fasync_list) {
1172		rcu_read_unlock();
1173		return -1;
1174	}
1175	switch (how) {
1176	case SOCK_WAKE_WAITD:
1177		if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1178			break;
1179		goto call_kill;
1180	case SOCK_WAKE_SPACE:
1181		if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1182			break;
1183		/* fall through */
1184	case SOCK_WAKE_IO:
1185call_kill:
1186		kill_fasync(&wq->fasync_list, SIGIO, band);
1187		break;
1188	case SOCK_WAKE_URG:
1189		kill_fasync(&wq->fasync_list, SIGURG, band);
1190	}
1191	rcu_read_unlock();
1192	return 0;
1193}
1194EXPORT_SYMBOL(sock_wake_async);
1195
1196int __sock_create(struct net *net, int family, int type, int protocol,
1197			 struct socket **res, int kern)
1198{
1199	int err;
1200	struct socket *sock;
1201	const struct net_proto_family *pf;
1202
1203	/*
1204	 *      Check protocol is in range
1205	 */
1206	if (family < 0 || family >= NPROTO)
1207		return -EAFNOSUPPORT;
1208	if (type < 0 || type >= SOCK_MAX)
1209		return -EINVAL;
1210
1211	/* Compatibility.
1212
1213	   This uglymoron is moved from INET layer to here to avoid
1214	   deadlock in module load.
1215	 */
1216	if (family == PF_INET && type == SOCK_PACKET) {
1217		static int warned;
1218		if (!warned) {
1219			warned = 1;
1220			printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1221			       current->comm);
1222		}
1223		family = PF_PACKET;
1224	}
1225
1226	err = security_socket_create(family, type, protocol, kern);
1227	if (err)
1228		return err;
1229
1230	/*
1231	 *	Allocate the socket and allow the family to set things up. if
1232	 *	the protocol is 0, the family is instructed to select an appropriate
1233	 *	default.
1234	 */
1235	sock = sock_alloc();
1236	if (!sock) {
1237		if (net_ratelimit())
1238			printk(KERN_WARNING "socket: no more sockets\n");
1239		return -ENFILE;	/* Not exactly a match, but its the
1240				   closest posix thing */
1241	}
1242
1243	sock->type = type;
1244
1245#ifdef CONFIG_MODULES
1246	/* Attempt to load a protocol module if the find failed.
1247	 *
1248	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1249	 * requested real, full-featured networking support upon configuration.
1250	 * Otherwise module support will break!
1251	 */
1252	if (rcu_access_pointer(net_families[family]) == NULL)
1253		request_module("net-pf-%d", family);
1254#endif
1255
1256	rcu_read_lock();
1257	pf = rcu_dereference(net_families[family]);
1258	err = -EAFNOSUPPORT;
1259	if (!pf)
1260		goto out_release;
1261
1262	/*
1263	 * We will call the ->create function, that possibly is in a loadable
1264	 * module, so we have to bump that loadable module refcnt first.
1265	 */
1266	if (!try_module_get(pf->owner))
1267		goto out_release;
1268
1269	/* Now protected by module ref count */
1270	rcu_read_unlock();
1271
1272	err = pf->create(net, sock, protocol, kern);
1273	if (err < 0)
1274		goto out_module_put;
1275
1276	/*
1277	 * Now to bump the refcnt of the [loadable] module that owns this
1278	 * socket at sock_release time we decrement its refcnt.
1279	 */
1280	if (!try_module_get(sock->ops->owner))
1281		goto out_module_busy;
1282
1283	/*
1284	 * Now that we're done with the ->create function, the [loadable]
1285	 * module can have its refcnt decremented
1286	 */
1287	module_put(pf->owner);
1288	err = security_socket_post_create(sock, family, type, protocol, kern);
1289	if (err)
1290		goto out_sock_release;
1291	*res = sock;
1292
1293	return 0;
1294
1295out_module_busy:
1296	err = -EAFNOSUPPORT;
1297out_module_put:
1298	sock->ops = NULL;
1299	module_put(pf->owner);
1300out_sock_release:
1301	sock_release(sock);
1302	return err;
1303
1304out_release:
1305	rcu_read_unlock();
1306	goto out_sock_release;
1307}
1308EXPORT_SYMBOL(__sock_create);
1309
1310int sock_create(int family, int type, int protocol, struct socket **res)
1311{
1312	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1313}
1314EXPORT_SYMBOL(sock_create);
1315
1316int sock_create_kern(int family, int type, int protocol, struct socket **res)
1317{
1318	return __sock_create(&init_net, family, type, protocol, res, 1);
1319}
1320EXPORT_SYMBOL(sock_create_kern);
1321
1322SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1323{
1324	int retval;
1325	struct socket *sock;
1326	int flags;
1327
1328	/* Check the SOCK_* constants for consistency.  */
1329	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1330	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1331	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1332	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1333
1334	flags = type & ~SOCK_TYPE_MASK;
1335	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1336		return -EINVAL;
1337	type &= SOCK_TYPE_MASK;
1338
1339	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1340		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1341
1342	retval = sock_create(family, type, protocol, &sock);
1343	if (retval < 0)
1344		goto out;
1345
1346	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1347	if (retval < 0)
1348		goto out_release;
1349
1350out:
1351	/* It may be already another descriptor 8) Not kernel problem. */
1352	return retval;
1353
1354out_release:
1355	sock_release(sock);
1356	return retval;
1357}
1358
1359/*
1360 *	Create a pair of connected sockets.
1361 */
1362
1363SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1364		int __user *, usockvec)
1365{
1366	struct socket *sock1, *sock2;
1367	int fd1, fd2, err;
1368	struct file *newfile1, *newfile2;
1369	int flags;
1370
1371	flags = type & ~SOCK_TYPE_MASK;
1372	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1373		return -EINVAL;
1374	type &= SOCK_TYPE_MASK;
1375
1376	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1377		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1378
1379	/*
1380	 * Obtain the first socket and check if the underlying protocol
1381	 * supports the socketpair call.
1382	 */
1383
1384	err = sock_create(family, type, protocol, &sock1);
1385	if (err < 0)
1386		goto out;
1387
1388	err = sock_create(family, type, protocol, &sock2);
1389	if (err < 0)
1390		goto out_release_1;
1391
1392	err = sock1->ops->socketpair(sock1, sock2);
1393	if (err < 0)
1394		goto out_release_both;
1395
1396	fd1 = sock_alloc_file(sock1, &newfile1, flags);
1397	if (unlikely(fd1 < 0)) {
1398		err = fd1;
1399		goto out_release_both;
1400	}
1401
1402	fd2 = sock_alloc_file(sock2, &newfile2, flags);
1403	if (unlikely(fd2 < 0)) {
1404		err = fd2;
1405		fput(newfile1);
1406		put_unused_fd(fd1);
1407		sock_release(sock2);
1408		goto out;
1409	}
1410
1411	audit_fd_pair(fd1, fd2);
1412	fd_install(fd1, newfile1);
1413	fd_install(fd2, newfile2);
1414	/* fd1 and fd2 may be already another descriptors.
1415	 * Not kernel problem.
1416	 */
1417
1418	err = put_user(fd1, &usockvec[0]);
1419	if (!err)
1420		err = put_user(fd2, &usockvec[1]);
1421	if (!err)
1422		return 0;
1423
1424	sys_close(fd2);
1425	sys_close(fd1);
1426	return err;
1427
1428out_release_both:
1429	sock_release(sock2);
1430out_release_1:
1431	sock_release(sock1);
1432out:
1433	return err;
1434}
1435
1436/*
1437 *	Bind a name to a socket. Nothing much to do here since it's
1438 *	the protocol's responsibility to handle the local address.
1439 *
1440 *	We move the socket address to kernel space before we call
1441 *	the protocol layer (having also checked the address is ok).
1442 */
1443
1444SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1445{
1446	struct socket *sock;
1447	struct sockaddr_storage address;
1448	int err, fput_needed;
1449
1450	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1451	if (sock) {
1452		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1453		if (err >= 0) {
1454			err = security_socket_bind(sock,
1455						   (struct sockaddr *)&address,
1456						   addrlen);
1457			if (!err)
1458				err = sock->ops->bind(sock,
1459						      (struct sockaddr *)
1460						      &address, addrlen);
1461		}
1462		fput_light(sock->file, fput_needed);
1463	}
1464	return err;
1465}
1466
1467/*
1468 *	Perform a listen. Basically, we allow the protocol to do anything
1469 *	necessary for a listen, and if that works, we mark the socket as
1470 *	ready for listening.
1471 */
1472
1473SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1474{
1475	struct socket *sock;
1476	int err, fput_needed;
1477	int somaxconn;
1478
1479	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1480	if (sock) {
1481		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1482		if ((unsigned)backlog > somaxconn)
1483			backlog = somaxconn;
1484
1485		err = security_socket_listen(sock, backlog);
1486		if (!err)
1487			err = sock->ops->listen(sock, backlog);
1488
1489		fput_light(sock->file, fput_needed);
1490	}
1491	return err;
1492}
1493
1494/*
1495 *	For accept, we attempt to create a new socket, set up the link
1496 *	with the client, wake up the client, then return the new
1497 *	connected fd. We collect the address of the connector in kernel
1498 *	space and move it to user at the very end. This is unclean because
1499 *	we open the socket then return an error.
1500 *
1501 *	1003.1g adds the ability to recvmsg() to query connection pending
1502 *	status to recvmsg. We need to add that support in a way thats
1503 *	clean when we restucture accept also.
1504 */
1505
1506SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1507		int __user *, upeer_addrlen, int, flags)
1508{
1509	struct socket *sock, *newsock;
1510	struct file *newfile;
1511	int err, len, newfd, fput_needed;
1512	struct sockaddr_storage address;
1513
1514	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1515		return -EINVAL;
1516
1517	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1518		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1519
1520	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1521	if (!sock)
1522		goto out;
1523
1524	err = -ENFILE;
1525	newsock = sock_alloc();
1526	if (!newsock)
1527		goto out_put;
1528
1529	newsock->type = sock->type;
1530	newsock->ops = sock->ops;
1531
1532	/*
1533	 * We don't need try_module_get here, as the listening socket (sock)
1534	 * has the protocol module (sock->ops->owner) held.
1535	 */
1536	__module_get(newsock->ops->owner);
1537
1538	newfd = sock_alloc_file(newsock, &newfile, flags);
1539	if (unlikely(newfd < 0)) {
1540		err = newfd;
1541		sock_release(newsock);
1542		goto out_put;
1543	}
1544
1545	err = security_socket_accept(sock, newsock);
1546	if (err)
1547		goto out_fd;
1548
1549	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1550	if (err < 0)
1551		goto out_fd;
1552
1553	if (ccs_socket_post_accept_permission(sock, newsock)) {
1554		err = -EAGAIN; /* Hope less harmful than -EPERM. */
1555		goto out_fd;
1556	}
1557	if (upeer_sockaddr) {
1558		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1559					  &len, 2) < 0) {
1560			err = -ECONNABORTED;
1561			goto out_fd;
1562		}
1563		err = move_addr_to_user(&address,
1564					len, upeer_sockaddr, upeer_addrlen);
1565		if (err < 0)
1566			goto out_fd;
1567	}
1568
1569	/* File flags are not inherited via accept() unlike another OSes. */
1570
1571	fd_install(newfd, newfile);
1572	err = newfd;
1573
1574out_put:
1575	fput_light(sock->file, fput_needed);
1576out:
1577	return err;
1578out_fd:
1579	fput(newfile);
1580	put_unused_fd(newfd);
1581	goto out_put;
1582}
1583
1584SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1585		int __user *, upeer_addrlen)
1586{
1587	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1588}
1589
1590/*
1591 *	Attempt to connect to a socket with the server address.  The address
1592 *	is in user space so we verify it is OK and move it to kernel space.
1593 *
1594 *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1595 *	break bindings
1596 *
1597 *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1598 *	other SEQPACKET protocols that take time to connect() as it doesn't
1599 *	include the -EINPROGRESS status for such sockets.
1600 */
1601
1602SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1603		int, addrlen)
1604{
1605	struct socket *sock;
1606	struct sockaddr_storage address;
1607	int err, fput_needed;
1608
1609	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1610	if (!sock)
1611		goto out;
1612	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1613	if (err < 0)
1614		goto out_put;
1615
1616	err =
1617	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1618	if (err)
1619		goto out_put;
1620
1621	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1622				 sock->file->f_flags);
1623out_put:
1624	fput_light(sock->file, fput_needed);
1625out:
1626	return err;
1627}
1628
1629/*
1630 *	Get the local address ('name') of a socket object. Move the obtained
1631 *	name to user space.
1632 */
1633
1634SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1635		int __user *, usockaddr_len)
1636{
1637	struct socket *sock;
1638	struct sockaddr_storage address;
1639	int len, err, fput_needed;
1640
1641	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1642	if (!sock)
1643		goto out;
1644
1645	err = security_socket_getsockname(sock);
1646	if (err)
1647		goto out_put;
1648
1649	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1650	if (err)
1651		goto out_put;
1652	err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1653
1654out_put:
1655	fput_light(sock->file, fput_needed);
1656out:
1657	return err;
1658}
1659
1660/*
1661 *	Get the remote address ('name') of a socket object. Move the obtained
1662 *	name to user space.
1663 */
1664
1665SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1666		int __user *, usockaddr_len)
1667{
1668	struct socket *sock;
1669	struct sockaddr_storage address;
1670	int len, err, fput_needed;
1671
1672	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1673	if (sock != NULL) {
1674		err = security_socket_getpeername(sock);
1675		if (err) {
1676			fput_light(sock->file, fput_needed);
1677			return err;
1678		}
1679
1680		err =
1681		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1682				       1);
1683		if (!err)
1684			err = move_addr_to_user(&address, len, usockaddr,
1685						usockaddr_len);
1686		fput_light(sock->file, fput_needed);
1687	}
1688	return err;
1689}
1690
1691/*
1692 *	Send a datagram to a given address. We move the address into kernel
1693 *	space and check the user space data area is readable before invoking
1694 *	the protocol.
1695 */
1696
1697SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1698		unsigned, flags, struct sockaddr __user *, addr,
1699		int, addr_len)
1700{
1701	struct socket *sock;
1702	struct sockaddr_storage address;
1703	int err;
1704	struct msghdr msg;
1705	struct iovec iov;
1706	int fput_needed;
1707
1708	if (len > INT_MAX)
1709		len = INT_MAX;
1710	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1711	if (!sock)
1712		goto out;
1713
1714	iov.iov_base = buff;
1715	iov.iov_len = len;
1716	msg.msg_name = NULL;
1717	msg.msg_iov = &iov;
1718	msg.msg_iovlen = 1;
1719	msg.msg_control = NULL;
1720	msg.msg_controllen = 0;
1721	msg.msg_namelen = 0;
1722	if (addr) {
1723		err = move_addr_to_kernel(addr, addr_len, &address);
1724		if (err < 0)
1725			goto out_put;
1726		msg.msg_name = (struct sockaddr *)&address;
1727		msg.msg_namelen = addr_len;
1728	}
1729	if (sock->file->f_flags & O_NONBLOCK)
1730		flags |= MSG_DONTWAIT;
1731	msg.msg_flags = flags;
1732	err = sock_sendmsg(sock, &msg, len);
1733
1734out_put:
1735	fput_light(sock->file, fput_needed);
1736out:
1737	return err;
1738}
1739
1740/*
1741 *	Send a datagram down a socket.
1742 */
1743
1744SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1745		unsigned, flags)
1746{
1747	return sys_sendto(fd, buff, len, flags, NULL, 0);
1748}
1749
1750/*
1751 *	Receive a frame from the socket and optionally record the address of the
1752 *	sender. We verify the buffers are writable and if needed move the
1753 *	sender address from kernel to user space.
1754 */
1755
1756SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1757		unsigned, flags, struct sockaddr __user *, addr,
1758		int __user *, addr_len)
1759{
1760	struct socket *sock;
1761	struct iovec iov;
1762	struct msghdr msg;
1763	struct sockaddr_storage address;
1764	int err, err2;
1765	int fput_needed;
1766
1767	if (size > INT_MAX)
1768		size = INT_MAX;
1769	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1770	if (!sock)
1771		goto out;
1772
1773	msg.msg_control = NULL;
1774	msg.msg_controllen = 0;
1775	msg.msg_iovlen = 1;
1776	msg.msg_iov = &iov;
1777	iov.iov_len = size;
1778	iov.iov_base = ubuf;
1779	msg.msg_name = (struct sockaddr *)&address;
1780	msg.msg_namelen = sizeof(address);
1781	if (sock->file->f_flags & O_NONBLOCK)
1782		flags |= MSG_DONTWAIT;
1783	err = sock_recvmsg(sock, &msg, size, flags);
1784
1785	if (err >= 0 && addr != NULL) {
1786		err2 = move_addr_to_user(&address,
1787					 msg.msg_namelen, addr, addr_len);
1788		if (err2 < 0)
1789			err = err2;
1790	}
1791
1792	fput_light(sock->file, fput_needed);
1793out:
1794	return err;
1795}
1796
1797/*
1798 *	Receive a datagram from a socket.
1799 */
1800
1801asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1802			 unsigned flags)
1803{
1804	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1805}
1806
1807/*
1808 *	Set a socket option. Because we don't know the option lengths we have
1809 *	to pass the user mode parameter for the protocols to sort out.
1810 */
1811
1812SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1813		char __user *, optval, int, optlen)
1814{
1815	int err, fput_needed;
1816	struct socket *sock;
1817
1818	if (optlen < 0)
1819		return -EINVAL;
1820
1821	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1822	if (sock != NULL) {
1823		err = security_socket_setsockopt(sock, level, optname);
1824		if (err)
1825			goto out_put;
1826
1827		if (level == SOL_SOCKET)
1828			err =
1829			    sock_setsockopt(sock, level, optname, optval,
1830					    optlen);
1831		else
1832			err =
1833			    sock->ops->setsockopt(sock, level, optname, optval,
1834						  optlen);
1835out_put:
1836		fput_light(sock->file, fput_needed);
1837	}
1838	return err;
1839}
1840
1841/*
1842 *	Get a socket option. Because we don't know the option lengths we have
1843 *	to pass a user mode parameter for the protocols to sort out.
1844 */
1845
1846SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1847		char __user *, optval, int __user *, optlen)
1848{
1849	int err, fput_needed;
1850	struct socket *sock;
1851
1852	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1853	if (sock != NULL) {
1854		err = security_socket_getsockopt(sock, level, optname);
1855		if (err)
1856			goto out_put;
1857
1858		if (level == SOL_SOCKET)
1859			err =
1860			    sock_getsockopt(sock, level, optname, optval,
1861					    optlen);
1862		else
1863			err =
1864			    sock->ops->getsockopt(sock, level, optname, optval,
1865						  optlen);
1866out_put:
1867		fput_light(sock->file, fput_needed);
1868	}
1869	return err;
1870}
1871
1872/*
1873 *	Shutdown a socket.
1874 */
1875
1876SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1877{
1878	int err, fput_needed;
1879	struct socket *sock;
1880
1881	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1882	if (sock != NULL) {
1883		err = security_socket_shutdown(sock, how);
1884		if (!err)
1885			err = sock->ops->shutdown(sock, how);
1886		fput_light(sock->file, fput_needed);
1887	}
1888	return err;
1889}
1890
1891/* A couple of helpful macros for getting the address of the 32/64 bit
1892 * fields which are the same type (int / unsigned) on our platforms.
1893 */
1894#define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1895#define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1896#define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1897
1898struct used_address {
1899	struct sockaddr_storage name;
1900	unsigned int name_len;
1901};
1902
1903static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1904			 struct msghdr *msg_sys, unsigned flags,
1905			 struct used_address *used_address)
1906{
1907	struct compat_msghdr __user *msg_compat =
1908	    (struct compat_msghdr __user *)msg;
1909	struct sockaddr_storage address;
1910	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1911	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1912	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1913	/* 20 is size of ipv6_pktinfo */
1914	unsigned char *ctl_buf = ctl;
1915	int err, ctl_len, iov_size, total_len;
1916
1917	err = -EFAULT;
1918	if (MSG_CMSG_COMPAT & flags) {
1919		if (get_compat_msghdr(msg_sys, msg_compat))
1920			return -EFAULT;
1921	} else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1922		return -EFAULT;
1923
1924	/* do not move before msg_sys is valid */
1925	err = -EMSGSIZE;
1926	if (msg_sys->msg_iovlen > UIO_MAXIOV)
1927		goto out;
1928
1929	/* Check whether to allocate the iovec area */
1930	err = -ENOMEM;
1931	iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1932	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1933		iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1934		if (!iov)
1935			goto out;
1936	}
1937
1938	/* This will also move the address data into kernel space */
1939	if (MSG_CMSG_COMPAT & flags) {
1940		err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
1941	} else
1942		err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
1943	if (err < 0)
1944		goto out_freeiov;
1945	total_len = err;
1946
1947	err = -ENOBUFS;
1948
1949	if (msg_sys->msg_controllen > INT_MAX)
1950		goto out_freeiov;
1951	ctl_len = msg_sys->msg_controllen;
1952	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1953		err =
1954		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1955						     sizeof(ctl));
1956		if (err)
1957			goto out_freeiov;
1958		ctl_buf = msg_sys->msg_control;
1959		ctl_len = msg_sys->msg_controllen;
1960	} else if (ctl_len) {
1961		if (ctl_len > sizeof(ctl)) {
1962			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1963			if (ctl_buf == NULL)
1964				goto out_freeiov;
1965		}
1966		err = -EFAULT;
1967		/*
1968		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1969		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1970		 * checking falls down on this.
1971		 */
1972		if (copy_from_user(ctl_buf,
1973				   (void __user __force *)msg_sys->msg_control,
1974				   ctl_len))
1975			goto out_freectl;
1976		msg_sys->msg_control = ctl_buf;
1977	}
1978	msg_sys->msg_flags = flags;
1979
1980	if (sock->file->f_flags & O_NONBLOCK)
1981		msg_sys->msg_flags |= MSG_DONTWAIT;
1982	/*
1983	 * If this is sendmmsg() and current destination address is same as
1984	 * previously succeeded address, omit asking LSM's decision.
1985	 * used_address->name_len is initialized to UINT_MAX so that the first
1986	 * destination address never matches.
1987	 */
1988	if (used_address && msg_sys->msg_name &&
1989	    used_address->name_len == msg_sys->msg_namelen &&
1990	    !memcmp(&used_address->name, msg_sys->msg_name,
1991		    used_address->name_len)) {
1992		err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1993		goto out_freectl;
1994	}
1995	err = sock_sendmsg(sock, msg_sys, total_len);
1996	/*
1997	 * If this is sendmmsg() and sending to current destination address was
1998	 * successful, remember it.
1999	 */
2000	if (used_address && err >= 0) {
2001		used_address->name_len = msg_sys->msg_namelen;
2002		if (msg_sys->msg_name)
2003			memcpy(&used_address->name, msg_sys->msg_name,
2004			       used_address->name_len);
2005	}
2006

Large files files are truncated, but you can click here to view the full file