PageRenderTime 9ms CodeModel.GetById 37ms app.highlight 138ms RepoModel.GetById 4ms app.codeStats 1ms

/net/socket.c

https://bitbucket.org/zarboz/ville-upstream-test-branch
C | 3385 lines | 2431 code | 488 blank | 466 comment | 378 complexity | e32ca468343846396f70ce384d0d5075 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 *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 *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	return 0;
 542}
 543EXPORT_SYMBOL(sock_tx_timestamp);
 544
 545static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 546				       struct msghdr *msg, size_t size)
 547{
 548	struct sock_iocb *si = kiocb_to_siocb(iocb);
 549
 550	sock_update_classid(sock->sk);
 551
 552	si->sock = sock;
 553	si->scm = NULL;
 554	si->msg = msg;
 555	si->size = size;
 556
 557	return sock->ops->sendmsg(iocb, sock, msg, size);
 558}
 559
 560static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 561				 struct msghdr *msg, size_t size)
 562{
 563	int err = security_socket_sendmsg(sock, msg, size);
 564
 565	return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 566}
 567
 568int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 569{
 570	struct kiocb iocb;
 571	struct sock_iocb siocb;
 572	int ret;
 573
 574	init_sync_kiocb(&iocb, NULL);
 575	iocb.private = &siocb;
 576	ret = __sock_sendmsg(&iocb, sock, msg, size);
 577	if (-EIOCBQUEUED == ret)
 578		ret = wait_on_sync_kiocb(&iocb);
 579	return ret;
 580}
 581EXPORT_SYMBOL(sock_sendmsg);
 582
 583static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 584{
 585	struct kiocb iocb;
 586	struct sock_iocb siocb;
 587	int ret;
 588
 589	init_sync_kiocb(&iocb, NULL);
 590	iocb.private = &siocb;
 591	ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 592	if (-EIOCBQUEUED == ret)
 593		ret = wait_on_sync_kiocb(&iocb);
 594	return ret;
 595}
 596
 597int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 598		   struct kvec *vec, size_t num, size_t size)
 599{
 600	mm_segment_t oldfs = get_fs();
 601	int result;
 602
 603	set_fs(KERNEL_DS);
 604	/*
 605	 * the following is safe, since for compiler definitions of kvec and
 606	 * iovec are identical, yielding the same in-core layout and alignment
 607	 */
 608	msg->msg_iov = (struct iovec *)vec;
 609	msg->msg_iovlen = num;
 610	result = sock_sendmsg(sock, msg, size);
 611	set_fs(oldfs);
 612	return result;
 613}
 614EXPORT_SYMBOL(kernel_sendmsg);
 615
 616static int ktime2ts(ktime_t kt, struct timespec *ts)
 617{
 618	if (kt.tv64) {
 619		*ts = ktime_to_timespec(kt);
 620		return 1;
 621	} else {
 622		return 0;
 623	}
 624}
 625
 626/*
 627 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 628 */
 629void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 630	struct sk_buff *skb)
 631{
 632	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 633	struct timespec ts[3];
 634	int empty = 1;
 635	struct skb_shared_hwtstamps *shhwtstamps =
 636		skb_hwtstamps(skb);
 637
 638	/* Race occurred between timestamp enabling and packet
 639	   receiving.  Fill in the current time for now. */
 640	if (need_software_tstamp && skb->tstamp.tv64 == 0)
 641		__net_timestamp(skb);
 642
 643	if (need_software_tstamp) {
 644		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 645			struct timeval tv;
 646			skb_get_timestamp(skb, &tv);
 647			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 648				 sizeof(tv), &tv);
 649		} else {
 650			skb_get_timestampns(skb, &ts[0]);
 651			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 652				 sizeof(ts[0]), &ts[0]);
 653		}
 654	}
 655
 656
 657	memset(ts, 0, sizeof(ts));
 658	if (skb->tstamp.tv64 &&
 659	    sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 660		skb_get_timestampns(skb, ts + 0);
 661		empty = 0;
 662	}
 663	if (shhwtstamps) {
 664		if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 665		    ktime2ts(shhwtstamps->syststamp, ts + 1))
 666			empty = 0;
 667		if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 668		    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 669			empty = 0;
 670	}
 671	if (!empty)
 672		put_cmsg(msg, SOL_SOCKET,
 673			 SCM_TIMESTAMPING, sizeof(ts), &ts);
 674}
 675EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 676
 677static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 678				   struct sk_buff *skb)
 679{
 680	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 681		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 682			sizeof(__u32), &skb->dropcount);
 683}
 684
 685void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 686	struct sk_buff *skb)
 687{
 688	sock_recv_timestamp(msg, sk, skb);
 689	sock_recv_drops(msg, sk, skb);
 690}
 691EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 692
 693static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 694				       struct msghdr *msg, size_t size, int flags)
 695{
 696	struct sock_iocb *si = kiocb_to_siocb(iocb);
 697
 698	sock_update_classid(sock->sk);
 699
 700	si->sock = sock;
 701	si->scm = NULL;
 702	si->msg = msg;
 703	si->size = size;
 704	si->flags = flags;
 705
 706	return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 707}
 708
 709static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 710				 struct msghdr *msg, size_t size, int flags)
 711{
 712	int err = security_socket_recvmsg(sock, msg, size, flags);
 713
 714	return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 715}
 716
 717int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 718		 size_t size, int flags)
 719{
 720	struct kiocb iocb;
 721	struct sock_iocb siocb;
 722	int ret;
 723
 724	init_sync_kiocb(&iocb, NULL);
 725	iocb.private = &siocb;
 726	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 727	if (-EIOCBQUEUED == ret)
 728		ret = wait_on_sync_kiocb(&iocb);
 729	return ret;
 730}
 731EXPORT_SYMBOL(sock_recvmsg);
 732
 733static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 734			      size_t size, int flags)
 735{
 736	struct kiocb iocb;
 737	struct sock_iocb siocb;
 738	int ret;
 739
 740	init_sync_kiocb(&iocb, NULL);
 741	iocb.private = &siocb;
 742	ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 743	if (-EIOCBQUEUED == ret)
 744		ret = wait_on_sync_kiocb(&iocb);
 745	return ret;
 746}
 747
 748/**
 749 * kernel_recvmsg - Receive a message from a socket (kernel space)
 750 * @sock:       The socket to receive the message from
 751 * @msg:        Received message
 752 * @vec:        Input s/g array for message data
 753 * @num:        Size of input s/g array
 754 * @size:       Number of bytes to read
 755 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 756 *
 757 * On return the msg structure contains the scatter/gather array passed in the
 758 * vec argument. The array is modified so that it consists of the unfilled
 759 * portion of the original array.
 760 *
 761 * The returned value is the total number of bytes received, or an error.
 762 */
 763int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 764		   struct kvec *vec, size_t num, size_t size, int flags)
 765{
 766	mm_segment_t oldfs = get_fs();
 767	int result;
 768
 769	set_fs(KERNEL_DS);
 770	/*
 771	 * the following is safe, since for compiler definitions of kvec and
 772	 * iovec are identical, yielding the same in-core layout and alignment
 773	 */
 774	msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 775	result = sock_recvmsg(sock, msg, size, flags);
 776	set_fs(oldfs);
 777	return result;
 778}
 779EXPORT_SYMBOL(kernel_recvmsg);
 780
 781static void sock_aio_dtor(struct kiocb *iocb)
 782{
 783	kfree(iocb->private);
 784}
 785
 786static ssize_t sock_sendpage(struct file *file, struct page *page,
 787			     int offset, size_t size, loff_t *ppos, int more)
 788{
 789	struct socket *sock;
 790	int flags;
 791
 792	sock = file->private_data;
 793
 794	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 795	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 796	flags |= more;
 797
 798	return kernel_sendpage(sock, page, offset, size, flags);
 799}
 800
 801static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 802				struct pipe_inode_info *pipe, size_t len,
 803				unsigned int flags)
 804{
 805	struct socket *sock = file->private_data;
 806
 807	if (unlikely(!sock->ops->splice_read))
 808		return -EINVAL;
 809
 810	sock_update_classid(sock->sk);
 811
 812	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 813}
 814
 815static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 816					 struct sock_iocb *siocb)
 817{
 818	if (!is_sync_kiocb(iocb)) {
 819		siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 820		if (!siocb)
 821			return NULL;
 822		iocb->ki_dtor = sock_aio_dtor;
 823	}
 824
 825	siocb->kiocb = iocb;
 826	iocb->private = siocb;
 827	return siocb;
 828}
 829
 830static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 831		struct file *file, const struct iovec *iov,
 832		unsigned long nr_segs)
 833{
 834	struct socket *sock = file->private_data;
 835	size_t size = 0;
 836	int i;
 837
 838	for (i = 0; i < nr_segs; i++)
 839		size += iov[i].iov_len;
 840
 841	msg->msg_name = NULL;
 842	msg->msg_namelen = 0;
 843	msg->msg_control = NULL;
 844	msg->msg_controllen = 0;
 845	msg->msg_iov = (struct iovec *)iov;
 846	msg->msg_iovlen = nr_segs;
 847	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 848
 849	return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 850}
 851
 852static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 853				unsigned long nr_segs, loff_t pos)
 854{
 855	struct sock_iocb siocb, *x;
 856
 857	if (pos != 0)
 858		return -ESPIPE;
 859
 860	if (iocb->ki_left == 0)	/* Match SYS5 behaviour */
 861		return 0;
 862
 863
 864	x = alloc_sock_iocb(iocb, &siocb);
 865	if (!x)
 866		return -ENOMEM;
 867	return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 868}
 869
 870static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 871			struct file *file, const struct iovec *iov,
 872			unsigned long nr_segs)
 873{
 874	struct socket *sock = file->private_data;
 875	size_t size = 0;
 876	int i;
 877
 878	for (i = 0; i < nr_segs; i++)
 879		size += iov[i].iov_len;
 880
 881	msg->msg_name = NULL;
 882	msg->msg_namelen = 0;
 883	msg->msg_control = NULL;
 884	msg->msg_controllen = 0;
 885	msg->msg_iov = (struct iovec *)iov;
 886	msg->msg_iovlen = nr_segs;
 887	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 888	if (sock->type == SOCK_SEQPACKET)
 889		msg->msg_flags |= MSG_EOR;
 890
 891	return __sock_sendmsg(iocb, sock, msg, size);
 892}
 893
 894static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 895			  unsigned long nr_segs, loff_t pos)
 896{
 897	struct sock_iocb siocb, *x;
 898
 899	if (pos != 0)
 900		return -ESPIPE;
 901
 902	x = alloc_sock_iocb(iocb, &siocb);
 903	if (!x)
 904		return -ENOMEM;
 905
 906	return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 907}
 908
 909/*
 910 * Atomic setting of ioctl hooks to avoid race
 911 * with module unload.
 912 */
 913
 914static DEFINE_MUTEX(br_ioctl_mutex);
 915static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 916
 917void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 918{
 919	mutex_lock(&br_ioctl_mutex);
 920	br_ioctl_hook = hook;
 921	mutex_unlock(&br_ioctl_mutex);
 922}
 923EXPORT_SYMBOL(brioctl_set);
 924
 925static DEFINE_MUTEX(vlan_ioctl_mutex);
 926static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 927
 928void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 929{
 930	mutex_lock(&vlan_ioctl_mutex);
 931	vlan_ioctl_hook = hook;
 932	mutex_unlock(&vlan_ioctl_mutex);
 933}
 934EXPORT_SYMBOL(vlan_ioctl_set);
 935
 936static DEFINE_MUTEX(dlci_ioctl_mutex);
 937static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 938
 939void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 940{
 941	mutex_lock(&dlci_ioctl_mutex);
 942	dlci_ioctl_hook = hook;
 943	mutex_unlock(&dlci_ioctl_mutex);
 944}
 945EXPORT_SYMBOL(dlci_ioctl_set);
 946
 947static long sock_do_ioctl(struct net *net, struct socket *sock,
 948				 unsigned int cmd, unsigned long arg)
 949{
 950	int err;
 951	void __user *argp = (void __user *)arg;
 952
 953	err = sock->ops->ioctl(sock, cmd, arg);
 954
 955	/*
 956	 * If this ioctl is unknown try to hand it down
 957	 * to the NIC driver.
 958	 */
 959	if (err == -ENOIOCTLCMD)
 960		err = dev_ioctl(net, cmd, argp);
 961
 962	return err;
 963}
 964
 965/*
 966 *	With an ioctl, arg may well be a user mode pointer, but we don't know
 967 *	what to do with it - that's up to the protocol still.
 968 */
 969
 970static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 971{
 972	struct socket *sock;
 973	struct sock *sk;
 974	void __user *argp = (void __user *)arg;
 975	int pid, err;
 976	struct net *net;
 977
 978	sock = file->private_data;
 979	sk = sock->sk;
 980	net = sock_net(sk);
 981	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
 982		err = dev_ioctl(net, cmd, argp);
 983	} else
 984#ifdef CONFIG_WEXT_CORE
 985	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
 986		err = dev_ioctl(net, cmd, argp);
 987	} else
 988#endif
 989		switch (cmd) {
 990		case FIOSETOWN:
 991		case SIOCSPGRP:
 992			err = -EFAULT;
 993			if (get_user(pid, (int __user *)argp))
 994				break;
 995			err = f_setown(sock->file, pid, 1);
 996			break;
 997		case FIOGETOWN:
 998		case SIOCGPGRP:
 999			err = put_user(f_getown(sock->file),
1000				       (int __user *)argp);
1001			break;
1002		case SIOCGIFBR:
1003		case SIOCSIFBR:
1004		case SIOCBRADDBR:
1005		case SIOCBRDELBR:
1006			err = -ENOPKG;
1007			if (!br_ioctl_hook)
1008				request_module("bridge");
1009
1010			mutex_lock(&br_ioctl_mutex);
1011			if (br_ioctl_hook)
1012				err = br_ioctl_hook(net, cmd, argp);
1013			mutex_unlock(&br_ioctl_mutex);
1014			break;
1015		case SIOCGIFVLAN:
1016		case SIOCSIFVLAN:
1017			err = -ENOPKG;
1018			if (!vlan_ioctl_hook)
1019				request_module("8021q");
1020
1021			mutex_lock(&vlan_ioctl_mutex);
1022			if (vlan_ioctl_hook)
1023				err = vlan_ioctl_hook(net, argp);
1024			mutex_unlock(&vlan_ioctl_mutex);
1025			break;
1026		case SIOCADDDLCI:
1027		case SIOCDELDLCI:
1028			err = -ENOPKG;
1029			if (!dlci_ioctl_hook)
1030				request_module("dlci");
1031
1032			mutex_lock(&dlci_ioctl_mutex);
1033			if (dlci_ioctl_hook)
1034				err = dlci_ioctl_hook(cmd, argp);
1035			mutex_unlock(&dlci_ioctl_mutex);
1036			break;
1037		default:
1038			err = sock_do_ioctl(net, sock, cmd, arg);
1039			break;
1040		}
1041	return err;
1042}
1043
1044int sock_create_lite(int family, int type, int protocol, struct socket **res)
1045{
1046	int err;
1047	struct socket *sock = NULL;
1048
1049	err = security_socket_create(family, type, protocol, 1);
1050	if (err)
1051		goto out;
1052
1053	sock = sock_alloc();
1054	if (!sock) {
1055		err = -ENOMEM;
1056		goto out;
1057	}
1058
1059	sock->type = type;
1060	err = security_socket_post_create(sock, family, type, protocol, 1);
1061	if (err)
1062		goto out_release;
1063
1064out:
1065	*res = sock;
1066	return err;
1067out_release:
1068	sock_release(sock);
1069	sock = NULL;
1070	goto out;
1071}
1072EXPORT_SYMBOL(sock_create_lite);
1073
1074/* No kernel lock held - perfect */
1075static unsigned int sock_poll(struct file *file, poll_table *wait)
1076{
1077	struct socket *sock;
1078
1079	/*
1080	 *      We can't return errors to poll, so it's either yes or no.
1081	 */
1082	sock = file->private_data;
1083	return sock->ops->poll(file, sock, wait);
1084}
1085
1086static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1087{
1088	struct socket *sock = file->private_data;
1089
1090	return sock->ops->mmap(file, sock, vma);
1091}
1092
1093static int sock_close(struct inode *inode, struct file *filp)
1094{
1095	/*
1096	 *      It was possible the inode is NULL we were
1097	 *      closing an unfinished socket.
1098	 */
1099
1100	if (!inode) {
1101		printk(KERN_DEBUG "sock_close: NULL inode\n");
1102		return 0;
1103	}
1104	sock_release(SOCKET_I(inode));
1105	return 0;
1106}
1107
1108/*
1109 *	Update the socket async list
1110 *
1111 *	Fasync_list locking strategy.
1112 *
1113 *	1. fasync_list is modified only under process context socket lock
1114 *	   i.e. under semaphore.
1115 *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1116 *	   or under socket lock
1117 */
1118
1119static int sock_fasync(int fd, struct file *filp, int on)
1120{
1121	struct socket *sock = filp->private_data;
1122	struct sock *sk = sock->sk;
1123	struct socket_wq *wq;
1124
1125	if (sk == NULL)
1126		return -EINVAL;
1127
1128	lock_sock(sk);
1129	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1130	fasync_helper(fd, filp, on, &wq->fasync_list);
1131
1132	if (!wq->fasync_list)
1133		sock_reset_flag(sk, SOCK_FASYNC);
1134	else
1135		sock_set_flag(sk, SOCK_FASYNC);
1136
1137	release_sock(sk);
1138	return 0;
1139}
1140
1141/* This function may be called only under socket lock or callback_lock or rcu_lock */
1142
1143int sock_wake_async(struct socket *sock, int how, int band)
1144{
1145	struct socket_wq *wq;
1146
1147	if (!sock)
1148		return -1;
1149	rcu_read_lock();
1150	wq = rcu_dereference(sock->wq);
1151	if (!wq || !wq->fasync_list) {
1152		rcu_read_unlock();
1153		return -1;
1154	}
1155	switch (how) {
1156	case SOCK_WAKE_WAITD:
1157		if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1158			break;
1159		goto call_kill;
1160	case SOCK_WAKE_SPACE:
1161		if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1162			break;
1163		/* fall through */
1164	case SOCK_WAKE_IO:
1165call_kill:
1166		kill_fasync(&wq->fasync_list, SIGIO, band);
1167		break;
1168	case SOCK_WAKE_URG:
1169		kill_fasync(&wq->fasync_list, SIGURG, band);
1170	}
1171	rcu_read_unlock();
1172	return 0;
1173}
1174EXPORT_SYMBOL(sock_wake_async);
1175
1176int __sock_create(struct net *net, int family, int type, int protocol,
1177			 struct socket **res, int kern)
1178{
1179	int err;
1180	struct socket *sock;
1181	const struct net_proto_family *pf;
1182
1183	/*
1184	 *      Check protocol is in range
1185	 */
1186	if (family < 0 || family >= NPROTO)
1187		return -EAFNOSUPPORT;
1188	if (type < 0 || type >= SOCK_MAX)
1189		return -EINVAL;
1190
1191	/* Compatibility.
1192
1193	   This uglymoron is moved from INET layer to here to avoid
1194	   deadlock in module load.
1195	 */
1196	if (family == PF_INET && type == SOCK_PACKET) {
1197		static int warned;
1198		if (!warned) {
1199			warned = 1;
1200			printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1201			       current->comm);
1202		}
1203		family = PF_PACKET;
1204	}
1205
1206	err = security_socket_create(family, type, protocol, kern);
1207	if (err)
1208		return err;
1209
1210	/*
1211	 *	Allocate the socket and allow the family to set things up. if
1212	 *	the protocol is 0, the family is instructed to select an appropriate
1213	 *	default.
1214	 */
1215	sock = sock_alloc();
1216	if (!sock) {
1217		if (net_ratelimit())
1218			printk(KERN_WARNING "socket: no more sockets\n");
1219		return -ENFILE;	/* Not exactly a match, but its the
1220				   closest posix thing */
1221	}
1222
1223	sock->type = type;
1224
1225#ifdef CONFIG_MODULES
1226	/* Attempt to load a protocol module if the find failed.
1227	 *
1228	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1229	 * requested real, full-featured networking support upon configuration.
1230	 * Otherwise module support will break!
1231	 */
1232	if (rcu_access_pointer(net_families[family]) == NULL)
1233		request_module("net-pf-%d", family);
1234#endif
1235
1236	rcu_read_lock();
1237	pf = rcu_dereference(net_families[family]);
1238	err = -EAFNOSUPPORT;
1239	if (!pf)
1240		goto out_release;
1241
1242	/*
1243	 * We will call the ->create function, that possibly is in a loadable
1244	 * module, so we have to bump that loadable module refcnt first.
1245	 */
1246	if (!try_module_get(pf->owner))
1247		goto out_release;
1248
1249	/* Now protected by module ref count */
1250	rcu_read_unlock();
1251
1252	err = pf->create(net, sock, protocol, kern);
1253	if (err < 0)
1254		goto out_module_put;
1255
1256	/*
1257	 * Now to bump the refcnt of the [loadable] module that owns this
1258	 * socket at sock_release time we decrement its refcnt.
1259	 */
1260	if (!try_module_get(sock->ops->owner))
1261		goto out_module_busy;
1262
1263	/*
1264	 * Now that we're done with the ->create function, the [loadable]
1265	 * module can have its refcnt decremented
1266	 */
1267	module_put(pf->owner);
1268	err = security_socket_post_create(sock, family, type, protocol, kern);
1269	if (err)
1270		goto out_sock_release;
1271	*res = sock;
1272
1273	return 0;
1274
1275out_module_busy:
1276	err = -EAFNOSUPPORT;
1277out_module_put:
1278	sock->ops = NULL;
1279	module_put(pf->owner);
1280out_sock_release:
1281	sock_release(sock);
1282	return err;
1283
1284out_release:
1285	rcu_read_unlock();
1286	goto out_sock_release;
1287}
1288EXPORT_SYMBOL(__sock_create);
1289
1290int sock_create(int family, int type, int protocol, struct socket **res)
1291{
1292	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1293}
1294EXPORT_SYMBOL(sock_create);
1295
1296int sock_create_kern(int family, int type, int protocol, struct socket **res)
1297{
1298	return __sock_create(&init_net, family, type, protocol, res, 1);
1299}
1300EXPORT_SYMBOL(sock_create_kern);
1301
1302SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1303{
1304	int retval;
1305	struct socket *sock;
1306	int flags;
1307
1308	/* Check the SOCK_* constants for consistency.  */
1309	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1310	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1311	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1312	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1313
1314	flags = type & ~SOCK_TYPE_MASK;
1315	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1316		return -EINVAL;
1317	type &= SOCK_TYPE_MASK;
1318
1319	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1320		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1321
1322	retval = sock_create(family, type, protocol, &sock);
1323	if (retval < 0)
1324		goto out;
1325
1326	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1327	if (retval < 0)
1328		goto out_release;
1329
1330out:
1331	/* It may be already another descriptor 8) Not kernel problem. */
1332	return retval;
1333
1334out_release:
1335	sock_release(sock);
1336	return retval;
1337}
1338
1339/*
1340 *	Create a pair of connected sockets.
1341 */
1342
1343SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1344		int __user *, usockvec)
1345{
1346	struct socket *sock1, *sock2;
1347	int fd1, fd2, err;
1348	struct file *newfile1, *newfile2;
1349	int flags;
1350
1351	flags = type & ~SOCK_TYPE_MASK;
1352	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1353		return -EINVAL;
1354	type &= SOCK_TYPE_MASK;
1355
1356	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1357		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1358
1359	/*
1360	 * Obtain the first socket and check if the underlying protocol
1361	 * supports the socketpair call.
1362	 */
1363
1364	err = sock_create(family, type, protocol, &sock1);
1365	if (err < 0)
1366		goto out;
1367
1368	err = sock_create(family, type, protocol, &sock2);
1369	if (err < 0)
1370		goto out_release_1;
1371
1372	err = sock1->ops->socketpair(sock1, sock2);
1373	if (err < 0)
1374		goto out_release_both;
1375
1376	fd1 = sock_alloc_file(sock1, &newfile1, flags);
1377	if (unlikely(fd1 < 0)) {
1378		err = fd1;
1379		goto out_release_both;
1380	}
1381
1382	fd2 = sock_alloc_file(sock2, &newfile2, flags);
1383	if (unlikely(fd2 < 0)) {
1384		err = fd2;
1385		fput(newfile1);
1386		put_unused_fd(fd1);
1387		sock_release(sock2);
1388		goto out;
1389	}
1390
1391	audit_fd_pair(fd1, fd2);
1392	fd_install(fd1, newfile1);
1393	fd_install(fd2, newfile2);
1394	/* fd1 and fd2 may be already another descriptors.
1395	 * Not kernel problem.
1396	 */
1397
1398	err = put_user(fd1, &usockvec[0]);
1399	if (!err)
1400		err = put_user(fd2, &usockvec[1]);
1401	if (!err)
1402		return 0;
1403
1404	sys_close(fd2);
1405	sys_close(fd1);
1406	return err;
1407
1408out_release_both:
1409	sock_release(sock2);
1410out_release_1:
1411	sock_release(sock1);
1412out:
1413	return err;
1414}
1415
1416/*
1417 *	Bind a name to a socket. Nothing much to do here since it's
1418 *	the protocol's responsibility to handle the local address.
1419 *
1420 *	We move the socket address to kernel space before we call
1421 *	the protocol layer (having also checked the address is ok).
1422 */
1423
1424SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1425{
1426	struct socket *sock;
1427	struct sockaddr_storage address;
1428	int err, fput_needed;
1429
1430	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1431	if (sock) {
1432		err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1433		if (err >= 0) {
1434			err = security_socket_bind(sock,
1435						   (struct sockaddr *)&address,
1436						   addrlen);
1437			if (!err)
1438				err = sock->ops->bind(sock,
1439						      (struct sockaddr *)
1440						      &address, addrlen);
1441		}
1442		fput_light(sock->file, fput_needed);
1443	}
1444	return err;
1445}
1446
1447/*
1448 *	Perform a listen. Basically, we allow the protocol to do anything
1449 *	necessary for a listen, and if that works, we mark the socket as
1450 *	ready for listening.
1451 */
1452
1453SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1454{
1455	struct socket *sock;
1456	int err, fput_needed;
1457	int somaxconn;
1458
1459	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1460	if (sock) {
1461		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1462		if ((unsigned)backlog > somaxconn)
1463			backlog = somaxconn;
1464
1465		err = security_socket_listen(sock, backlog);
1466		if (!err)
1467			err = sock->ops->listen(sock, backlog);
1468
1469		fput_light(sock->file, fput_needed);
1470	}
1471	return err;
1472}
1473
1474/*
1475 *	For accept, we attempt to create a new socket, set up the link
1476 *	with the client, wake up the client, then return the new
1477 *	connected fd. We collect the address of the connector in kernel
1478 *	space and move it to user at the very end. This is unclean because
1479 *	we open the socket then return an error.
1480 *
1481 *	1003.1g adds the ability to recvmsg() to query connection pending
1482 *	status to recvmsg. We need to add that support in a way thats
1483 *	clean when we restucture accept also.
1484 */
1485
1486SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1487		int __user *, upeer_addrlen, int, flags)
1488{
1489	struct socket *sock, *newsock;
1490	struct file *newfile;
1491	int err, len, newfd, fput_needed;
1492	struct sockaddr_storage address;
1493
1494	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1495		return -EINVAL;
1496
1497	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1498		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1499
1500	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1501	if (!sock)
1502		goto out;
1503
1504	err = -ENFILE;
1505	newsock = sock_alloc();
1506	if (!newsock)
1507		goto out_put;
1508
1509	newsock->type = sock->type;
1510	newsock->ops = sock->ops;
1511
1512	/*
1513	 * We don't need try_module_get here, as the listening socket (sock)
1514	 * has the protocol module (sock->ops->owner) held.
1515	 */
1516	__module_get(newsock->ops->owner);
1517
1518	newfd = sock_alloc_file(newsock, &newfile, flags);
1519	if (unlikely(newfd < 0)) {
1520		err = newfd;
1521		sock_release(newsock);
1522		goto out_put;
1523	}
1524
1525	err = security_socket_accept(sock, newsock);
1526	if (err)
1527		goto out_fd;
1528
1529	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1530	if (err < 0)
1531		goto out_fd;
1532
1533	if (upeer_sockaddr) {
1534		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1535					  &len, 2) < 0) {
1536			err = -ECONNABORTED;
1537			goto out_fd;
1538		}
1539		err = move_addr_to_user((struct sockaddr *)&address,
1540					len, upeer_sockaddr, upeer_addrlen);
1541		if (err < 0)
1542			goto out_fd;
1543	}
1544
1545	/* File flags are not inherited via accept() unlike another OSes. */
1546
1547	fd_install(newfd, newfile);
1548	err = newfd;
1549
1550out_put:
1551	fput_light(sock->file, fput_needed);
1552out:
1553	return err;
1554out_fd:
1555	fput(newfile);
1556	put_unused_fd(newfd);
1557	goto out_put;
1558}
1559
1560SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1561		int __user *, upeer_addrlen)
1562{
1563	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1564}
1565
1566/*
1567 *	Attempt to connect to a socket with the server address.  The address
1568 *	is in user space so we verify it is OK and move it to kernel space.
1569 *
1570 *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1571 *	break bindings
1572 *
1573 *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1574 *	other SEQPACKET protocols that take time to connect() as it doesn't
1575 *	include the -EINPROGRESS status for such sockets.
1576 */
1577
1578SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1579		int, addrlen)
1580{
1581	struct socket *sock;
1582	struct sockaddr_storage address;
1583	int err, fput_needed;
1584
1585	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1586	if (!sock)
1587		goto out;
1588	err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1589	if (err < 0)
1590		goto out_put;
1591
1592	err =
1593	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1594	if (err)
1595		goto out_put;
1596
1597	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1598				 sock->file->f_flags);
1599out_put:
1600	fput_light(sock->file, fput_needed);
1601out:
1602	return err;
1603}
1604
1605/*
1606 *	Get the local address ('name') of a socket object. Move the obtained
1607 *	name to user space.
1608 */
1609
1610SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1611		int __user *, usockaddr_len)
1612{
1613	struct socket *sock;
1614	struct sockaddr_storage address;
1615	int len, err, fput_needed;
1616
1617	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1618	if (!sock)
1619		goto out;
1620
1621	err = security_socket_getsockname(sock);
1622	if (err)
1623		goto out_put;
1624
1625	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1626	if (err)
1627		goto out_put;
1628	err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1629
1630out_put:
1631	fput_light(sock->file, fput_needed);
1632out:
1633	return err;
1634}
1635
1636/*
1637 *	Get the remote address ('name') of a socket object. Move the obtained
1638 *	name to user space.
1639 */
1640
1641SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1642		int __user *, usockaddr_len)
1643{
1644	struct socket *sock;
1645	struct sockaddr_storage address;
1646	int len, err, fput_needed;
1647
1648	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1649	if (sock != NULL) {
1650		err = security_socket_getpeername(sock);
1651		if (err) {
1652			fput_light(sock->file, fput_needed);
1653			return err;
1654		}
1655
1656		err =
1657		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1658				       1);
1659		if (!err)
1660			err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1661						usockaddr_len);
1662		fput_light(sock->file, fput_needed);
1663	}
1664	return err;
1665}
1666
1667/*
1668 *	Send a datagram to a given address. We move the address into kernel
1669 *	space and check the user space data area is readable before invoking
1670 *	the protocol.
1671 */
1672
1673SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1674		unsigned, flags, struct sockaddr __user *, addr,
1675		int, addr_len)
1676{
1677	struct socket *sock;
1678	struct sockaddr_storage address;
1679	int err;
1680	struct msghdr msg;
1681	struct iovec iov;
1682	int fput_needed;
1683
1684	if (len > INT_MAX)
1685		len = INT_MAX;
1686	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1687	if (!sock)
1688		goto out;
1689
1690	iov.iov_base = buff;
1691	iov.iov_len = len;
1692	msg.msg_name = NULL;
1693	msg.msg_iov = &iov;
1694	msg.msg_iovlen = 1;
1695	msg.msg_control = NULL;
1696	msg.msg_controllen = 0;
1697	msg.msg_namelen = 0;
1698	if (addr) {
1699		err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1700		if (err < 0)
1701			goto out_put;
1702		msg.msg_name = (struct sockaddr *)&address;
1703		msg.msg_namelen = addr_len;
1704	}
1705	if (sock->file->f_flags & O_NONBLOCK)
1706		flags |= MSG_DONTWAIT;
1707	msg.msg_flags = flags;
1708	err = sock_sendmsg(sock, &msg, len);
1709
1710out_put:
1711	fput_light(sock->file, fput_needed);
1712out:
1713	return err;
1714}
1715
1716/*
1717 *	Send a datagram down a socket.
1718 */
1719
1720SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1721		unsigned, flags)
1722{
1723	return sys_sendto(fd, buff, len, flags, NULL, 0);
1724}
1725
1726/*
1727 *	Receive a frame from the socket and optionally record the address of the
1728 *	sender. We verify the buffers are writable and if needed move the
1729 *	sender address from kernel to user space.
1730 */
1731
1732SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1733		unsigned, flags, struct sockaddr __user *, addr,
1734		int __user *, addr_len)
1735{
1736	struct socket *sock;
1737	struct iovec iov;
1738	struct msghdr msg;
1739	struct sockaddr_storage address;
1740	int err, err2;
1741	int fput_needed;
1742
1743	if (size > INT_MAX)
1744		size = INT_MAX;
1745	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1746	if (!sock)
1747		goto out;
1748
1749	msg.msg_control = NULL;
1750	msg.msg_controllen = 0;
1751	msg.msg_iovlen = 1;
1752	msg.msg_iov = &iov;
1753	iov.iov_len = size;
1754	iov.iov_base = ubuf;
1755	msg.msg_name = (struct sockaddr *)&address;
1756	msg.msg_namelen = sizeof(address);
1757	if (sock->file->f_flags & O_NONBLOCK)
1758		flags |= MSG_DONTWAIT;
1759	err = sock_recvmsg(sock, &msg, size, flags);
1760
1761	if (err >= 0 && addr != NULL) {
1762		err2 = move_addr_to_user((struct sockaddr *)&address,
1763					 msg.msg_namelen, addr, addr_len);
1764		if (err2 < 0)
1765			err = err2;
1766	}
1767
1768	fput_light(sock->file, fput_needed);
1769out:
1770	return err;
1771}
1772
1773/*
1774 *	Receive a datagram from a socket.
1775 */
1776
1777asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1778			 unsigned flags)
1779{
1780	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1781}
1782
1783/*
1784 *	Set a socket option. Because we don't know the option lengths we have
1785 *	to pass the user mode parameter for the protocols to sort out.
1786 */
1787
1788SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1789		char __user *, optval, int, optlen)
1790{
1791	int err, fput_needed;
1792	struct socket *sock;
1793
1794	if (optlen < 0)
1795		return -EINVAL;
1796
1797	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1798	if (sock != NULL) {
1799		err = security_socket_setsockopt(sock, level, optname);
1800		if (err)
1801			goto out_put;
1802
1803		if (level == SOL_SOCKET)
1804			err =
1805			    sock_setsockopt(sock, level, optname, optval,
1806					    optlen);
1807		else
1808			err =
1809			    sock->ops->setsockopt(sock, level, optname, optval,
1810						  optlen);
1811out_put:
1812		fput_light(sock->file, fput_needed);
1813	}
1814	return err;
1815}
1816
1817/*
1818 *	Get a socket option. Because we don't know the option lengths we have
1819 *	to pass a user mode parameter for the protocols to sort out.
1820 */
1821
1822SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1823		char __user *, optval, int __user *, optlen)
1824{
1825	int err, fput_needed;
1826	struct socket *sock;
1827
1828	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1829	if (sock != NULL) {
1830		err = security_socket_getsockopt(sock, level, optname);
1831		if (err)
1832			goto out_put;
1833
1834		if (level == SOL_SOCKET)
1835			err =
1836			    sock_getsockopt(sock, level, optname, optval,
1837					    optlen);
1838		else
1839			err =
1840			    sock->ops->getsockopt(sock, level, optname, optval,
1841						  optlen);
1842out_put:
1843		fput_light(sock->file, fput_needed);
1844	}
1845	return err;
1846}
1847
1848/*
1849 *	Shutdown a socket.
1850 */
1851
1852SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1853{
1854	int err, fput_needed;
1855	struct socket *sock;
1856
1857	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1858	if (sock != NULL) {
1859		err = security_socket_shutdown(sock, how);
1860		if (!err)
1861			err = sock->ops->shutdown(sock, how);
1862		fput_light(sock->file, fput_needed);
1863	}
1864	return err;
1865}
1866
1867/* A couple of helpful macros for getting the address of the 32/64 bit
1868 * fields which are the same type (int / unsigned) on our platforms.
1869 */
1870#define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1871#define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1872#define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1873
1874struct used_address {
1875	struct sockaddr_storage name;
1876	unsigned int name_len;
1877};
1878
1879static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1880			 struct msghdr *msg_sys, unsigned flags,
1881			 struct used_address *used_address)
1882{
1883	struct compat_msghdr __user *msg_compat =
1884	    (struct compat_msghdr __user *)msg;
1885	struct sockaddr_storage address;
1886	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1887	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1888	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1889	/* 20 is size of ipv6_pktinfo */
1890	unsigned char *ctl_buf = ctl;
1891	int err, ctl_len, iov_size, total_len;
1892
1893	err = -EFAULT;
1894	if (MSG_CMSG_COMPAT & flags) {
1895		if (get_compat_msghdr(msg_sys, msg_compat))
1896			return -EFAULT;
1897	} else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1898		return -EFAULT;
1899
1900	/* do not move before msg_sys is valid */
1901	err = -EMSGSIZE;
1902	if (msg_sys->msg_iovlen > UIO_MAXIOV)
1903		goto out;
1904
1905	/* Check whether to allocate the iovec area */
1906	err = -ENOMEM;
1907	iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1908	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1909		iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1910		if (!iov)
1911			goto out;
1912	}
1913
1914	/* This will also move the address data into kernel space */
1915	if (MSG_CMSG_COMPAT & flags) {
1916		err = verify_compat_iovec(msg_sys, iov,
1917					  (struct sockaddr *)&address,
1918					  VERIFY_READ);
1919	} else
1920		err = verify_iovec(msg_sys, iov,
1921				   (struct sockaddr *)&address,
1922				   VERIFY_READ);
1923	if (err < 0)
1924		goto out_freeiov;
1925	total_len = err;
1926
1927	err = -ENOBUFS;
1928
1929	if (msg_sys->msg_controllen > INT_MAX)
1930		goto out_freeiov;
1931	ctl_len = msg_sys->msg_controllen;
1932	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1933		err =
1934		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1935						     sizeof(ctl));
1936		if (err)
1937			goto out_freeiov;
1938		ctl_buf = msg_sys->msg_control;
1939		ctl_len = msg_sys->msg_controllen;
1940	} else if (ctl_len) {
1941		if (ctl_len > sizeof(ctl)) {
1942			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1943			if (ctl_buf == NULL)
1944				goto out_freeiov;
1945		}
1946		err = -EFAULT;
1947		/*
1948		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1949		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1950		 * checking falls down on this.
1951		 */
1952		if (copy_from_user(ctl_buf,
1953				   (void __user __force *)msg_sys->msg_control,
1954				   ctl_len))
1955			goto out_freectl;
1956		msg_sys->msg_control = ctl_buf;
1957	}
1958	msg_sys->msg_flags = flags;
1959
1960	if (sock->file->f_flags & O_NONBLOCK)
1961		msg_sys->msg_flags |= MSG_DONTWAIT;
1962	/*
1963	 * If this is sendmmsg() and current destination address is same as
1964	 * previously succeeded address, omit asking LSM's decision.
1965	 * used_address->name_len is initialized to UINT_MAX so that the first
1966	 * destination address never matches.
1967	 */
1968	if (used_address && msg_sys->msg_name &&
1969	    used_address->name_len == msg_sys->msg_namelen &&
1970	    !memcmp(&used_address->name, msg_sys->msg_name,
1971		    used_address->name_len)) {
1972		err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1973		goto out_freectl;
1974	}
1975	err = sock_sendmsg(sock, msg_sys, total_len);
1976	/*
1977	 * If this is sendmmsg() and sending to current destination address was
1978	 * successful, remember it.
1979	 */
1980	if (used_address && err >= 0) {
1981		used_address->name_len = msg_sys->msg_namelen;
1982		if (msg_sys->msg_name)
1983			memcpy(&used_address->name, msg_sys->msg_name,
1984			       used_address->name_len);
1985	}
1986
1987out_freectl:
1988	if (ctl_buf != ctl)
1989		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1990out_freeiov:
1991	if (iov != iovstack)
1992		sock_kfree_s(sock->sk, iov, iov_size);
1993out:
1994	return err;
1995}
1996
1997/*
1998 *	BSD sendmsg interface
1999 */
2000
2001SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
2002{
2003	int fput_needed, err;
2004	struct msghdr msg_sys;
2005	struct socket *sock = sockfd_lookup_light

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